Digital Radio and Its Application in the HF (2-30 MHz) Band

Digital Radio and Its Application

in the HF (2-30 MHz) Band

Nigel Clement Davies B. Eng(Hons) CEng WEE

Submitted in accordance with the Requirementsfor the Degree of Doctor of Philosophy

The University of Leeds School of Electronic and Electrical Engineering May 2004

The candidate confirms that the work submitted is his own and that appropriate credit has been given where reference has been made to the work of others. This copy has been supplied on the understanding that it is copyright material and that from be the thesis no quotation may published without proper acknowledgement

For Jo, my wonderful wife

, ýý,,..r

: r:.

..

Acknowledgements The research presented in this thesis was carried out for my employer, QinetiQ Ltd (formally DERA), and principally sponsored by the UK Ministry of Defence under its Corporate Research programme.

The research was undertaken as part of a working

collaboration with scientists and engineers at the Communications Research Centre (CRC) in Ottawa, Ontario, Canada under a UK-Canadian government agreement on collaboration in defence science and technology.

Joe Schlesak, CRC's Terrestrial

Wireless Systems group leader, and Dr Trisha Willink strongly supported the work and Canadian involvement in it. The research effort has included important contributions from a number of individuals. CRC researchers Dr Mark Jorgenson, Dr Bob Johnson and Bill

Moreland were

responsible for the realisation of the high throughput 16 kbps modem from my initial concept. CRC's Michael Bova made significant contributions to the HF software radio programmable logic design and the software architecture that we conceived, refined and implemented together. A number of Canadian under-graduate `co-op' work placement students also helped with hardware and software development tasks under my supervision and direction; specifically Chris Taylor, Chris Squires, Mike Osmond and Jason Chau. The complex multi-layer printed circuit boards for the transceiver were laid-out by Minh Huynh of CRC under my direction.

The RF shielding enclosures

utilised in the transceiver were designed by Andre Giroux and built by CRC workshops. Thanks are due to my colleagues Prof Paul Cannon, Dr Mathew Angling and Mel Maundrell who have always been sources of inspiration and good advice. A heartfelt thank-you goes to Prof Mike Darnell, for being my supervisor and keeping faith in me over so many years. Thanks also to George Vongas who started me down this path. Finally, I owe a true debt of gratitude to my parents for (as it has been said) doing so much for me, and to my wife, Jo, for great patience and understanding through the many months it has taken to prepare this thesis.

NC Davies, May 2004

Abstract

Digital Radio and Its Application

in the HF (2-30 MHz) Band

Nigel Davies

The propagation environment at high frequencies (HF, 2-30 MHz) has a significant impact on the performance of radio systems (especially data communications). However, the ability

to communicate information

over very long ranges using

ionospheric propagation paths without any intermediate infrastructure makes the use of HF attractive for many applications.

In order to increase the utility

of HF

is there communications a strong desire to increase HF data rates. Currently data rates of up to -2400 bps can be reliably achieved in standard 3 kHz HF channel allocations. Whilst further increases in data rate within the confines of these narrowband frequency allocations is likely, the use of larger bandwidths (contiguous or otherwise) appears to offer potential for much greater throughputs. This requires a greater understanding of the characteristics of wideband channels and also requires transmitting and receiving equipment capable of wideband/multi-channel operation.

New waveforms have been proposed for the transmission of higher data rates in extended channel bandwidths (6 kHz).

The results of laboratory measurements and

analysis of data collected during on-air trials of a number of 16 kbps waveforms are presented. Analysis indicates that operation over surface wave and benign skywave channels is possible, demonstrating

the benefit

of

exploiting

greater channel

bandwidths.

Suitable architectures for the implementation of wideband and multi-channel digital HF radios (software radios) have been investigated. The work presented indicates that it is for the first time, to construct high performance, direct sampling now possible, digital HF wideband receivers. In such a receiver the entire HF band is digitised and then all subsequent processing is undertaken digitally.

Conceptually this would allow

an arbitrary number of channels to be simultaneously received using a single RF frontdigitiser. end and

With careful design performance comparable with that of the high iv

V

performance conventional super-heterodyne single channel receivers can be obtained. A prototype wideband multi-channel digital HF transceiver with this architecture has been implemented and its performance shown to agree with that predicted.

A particular challenge in complex systems such as software radios is the deployment of software across a number of heterogeneous processors. A new asynchronous, eventbased, processing architecture which employs messaging to allow processing tasks to be effectively distributed across a multiple processors and buses is proposed. It has been implemented on the digital

transceiver platform

and its effectiveness has been

demonstrated.

A new low-power pulse-compression oblique HF ionospheric sounder, known as WHISPER, has been developed. This sounder has been implemented as a software application on the wideband HF digital transceiver.

Waveforms suitable for making

kHz) (-80 measurements of the channel time varying complex impulse wideband response have been designed.

These have been used to make measurements on a

170 km path in the UK during Spring 2001. The results of these measurements have been analysed and confirm the ability of the sounding instrument to measure the channel scattering function and the amplitude and phase within individual modes. A design directions for further to the analysis, pertinent of wideband number of possible HF modems, have been proposed.

V

vi

Contents

Acknowledgements

Abstract

................

'

iv ...........................................................................................................................

Contents

........................................................................................

List of Figures ix ................................................................................................................. List of Abbreviations

Chapter 1. 1.1 1.2 Chapter

2.

2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 Chapter 3. 3.1 3.2 3.3 3.4 3.5 Chapter 4. 4.1 4.2 4.3 4.4 4.5

xv ....................................................................................................

Introduction

1

The HF Propagation Environment and Its Impact on Communications ......................................................................................

7

............................................................................................. Structure of Thesis 2 .................................................................................... Original Work 4 ...........................................................................................

Surface Wave Propagation 7 ........................................................................ Sky Wave Propagation 7 .............................................................................. 16 NVIS Propagation ................................................................................... 16 Impact of Propagation on Radio Waves .................................................. 23 Propagation Diversity .............................................................................. 24 Propagation at Different Latitudes .......................................................... 25 Propagation of Wideband Signals ........................................................... 27 Noise and Interference ............................................................................ 30 HF Channel Models and their Application ............................................. 35 HF Propagation - Summary of Principal Characteristics ....................... HF Data Communications

37

.................................................................... Waveforms for Data Communication Over Fading, Multipath 37 Channels .................................................................................................. 38 Modulation Schemes for Data Communications .................................... Communications HF Data Serial Tone MIL-STD-188-IIOA -a 54 Waveform ................................................................................................ 56 High Data Rate HF Communications ..................................................... 58 Summary ................................................................................................. 59 A High Data Rate Modem for Extended Bandwidth Channels ........ 60 Waveform and Modem Processing Description ..................................... 62 16 kbps Modem Performance ................................................................. 69 On-air Trials ............................................................................................ 72 Potential Improvements to the Experimental 16 kbps Modem ............... 73 Comparison of Extended Bandwidth Modem Performance ................... VI

4.6 4.7 4.8 Chapter 5. 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 Chapter 6. 6.1 6.2 6.3 6.4 6.5 Chapter 7. 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 7.10 7.11 7.12 7.13 7.14 7.15 7.16 Chapter 8. 8.1

Application of Extended Bandwidth HDR Modems 75 .............................. Standardisation of Extended Bandwidth HDR HF Modems 76 .................. Chapter Summary 77 .................................................................................... On the Specification and Design of Digital HF Radios 78 ...................... Applicable Technology Developments 79 ................................................... Wideband Digital Radio Architectures 80 ................................................... HF Receiver Performance Requirements 87 ................................................ HF Transmitter Performance Requirements 96 ........................................... A Direct Sampling Digital HF Receiver 97 ................................................. Front End Filter 97 ....................................................................................... Digitally Controlled RF Amplifier 99 .......................................................... Analogue-to-Digital Converter (ADC) Performance 99 .............................. Digital Down-Converter (DDC) 110 ........................................................... 115 A High Performance Direct Sampling Digital HF Receiver ................. Single Conversion Wideband Receiver An Alternative 117 Architecture ........................................................................................... 119 Chapter Summary .................................................................................. Performance of a Prototype Direct Sampling Digital HF 120 Receiver ................................................................................................ 120 Description of Prototype Receiver ........................................................ 134 Predicted Performance of Prototype Digital HF Receiver .................... 136 Prototype Receiver Performance Measurements .................................. 143 Improving the Performance of the Prototype Receiver ......................... 145 Chapter Summary .................................................................................. A Wideband, Multi-Channel,

HF Software Radio

...........................

146

148 Wideband Digital Transceiver Architecture ......................................... 150 Digital Transceiver PCI Bus Interface .................................................. 150 Digital Transceiver Bus Arbitration CPLD ........................................... 151 Digital Signal Processing (DSP) Sub-System ....................................... 154 Dual SHARC DSP Processor Module .................................................. Digital Transceiver Configuration and Software Download ................ 156 158 Transceiver Digital Interfaces ............................................................... 160 Frequency Standard Sub-System .......................................................... 166 Digital Receiver Sub-System ................................................................ 171 Transmitter Exciter Sub-System ........................................................... 175 Power Supplies for Analogue Sub-Systems .......................................... 175 Front End Protection and Filtering Module .......................................... 176 Construction Techniques ....................................................................... 178 Digital Transceiver Control Software Architecture .............................. 179 Built-in Self Test and Diagnostics ........................................................ 181 Chapter Summary .................................................................................. A High Performance Event Driven Processing Architecture Asynchronous, Event Based, Processing Architecture

.........

.........................

182 183

vii

'iii 8.2 8.3 8.4 8.5 8.6 8.7 8.8

Digital Transceiver Platform 184 ................................................................. Intelligent Input/Output (120) Messaging 186 ............................................. Asynchronous Messaging - Software Implementation 190 ......................... Implementation Issues 197 ........................................................................... Practical Demonstration of Event Based Processing Architecture 198 ....... Possible Extensions to the Event Based Processing Architecture 199 ........ Chapter Summary 200 ..................................................................................

Chapter 9. 9.1 9.2

of Digital Radio to HF Channel Characterisation

......

201

Introduction to Ionospheric Channel Sounding 201 .................................... Measuring the Time Varying Complex Channel Impulse Response (C IR) 205 ..................................................................................................... Implementation of WHISPER -A Wideband HF Channel Sounder 207 ... Pulse Compression Sounding Waveform Design 210 ................................. Receive Digital Signal Processing 218 ........................................................

9.3 9.4 9.5 9.6

Laboratory Measurements to Verify Sounder Performance

9.7 9.8

.................

219

Suggestions for Improvements to the WHISPER Sounder 222 ................... Chapter Summary 223 ..................................................................................

Chapter 10. 10.1 10.2 10.3 10.4 Chapter

Application

11.

11.1 11.2 11.3 11.4 11.5 11.6

Measurement of the Wideband HF Channel using WHISPER...... 224 Experiment Deployment 224 ....................................................................... Results 225 ................................................................................................... Suggested Routes for Further Data Analysis 237 ........................................ Chapter Summary 237 .................................................................................. Applications

of Digital

Radio in the HF Band

..................................

239

Digital Broadcast Receivers 239 .................................................................. Chirp Sounder 240 ....................................................................................... RF Channel Simulator 240 ........................................................................... 241 High Performance ALE Systems .......................................................... 242 Applications in Other Frequency Bands ............................................... 242 Applications to Weather Radar .............................................................

244 Conclusions and Recommendations for Further Work .................. 245 12.1 Recommendations for Further Investigation ......................................... 250 References .................................................................................................................... Chapter 12.

viii

List of Figures Chapter 2 Figure Figure Figure Figure Figure

2-1 2-2 2-3 2-4 2-5

Figure 2-6

Figure 2-7 Figure 2-8 Figure 2-9 Figure 2-10 Figure 2-11 Figure 2-12

The Earth's Ionosphere and its Principal Regions [Maslin, 5] 8 ................. Typical Electron Concentration within the Ionosphere [Davies. 3].......... 9 Daily Smoothed Sunspot Numbers Illustrating I1 Year Solar Cycle 11 ..... Terrestrial Effects of a Solar Disturbance [5,3] 12 ..................................... Ray Paths as a Function of Elevation at a Single Frequency [Maslin, 5] ............................................................................................... Sky Predicted Propagation Loss and MUF for Frankfurt-to-London wave Circuit ............................................................................................

14 15

Position of Day-Night Terminator at 1200Z in December 16 ..................... Summary of the Causes of Multipath and Dispersion [after Maslin, 19 5] ............................................................................................................. Typical Oblique HF lonogram (Malvern to Farnborough 17:31 2020 03-200) .................................................................................................... Effective Noise Figure of a Loss-less Isotropic Antenna Due to 28 External Noise in the HF Band [ITU, 29] ............................................... 29 Predicted Congestion from Gott-Laycock Occupancy Model ................ 32 Block diagram of Watterson model ........................................................

Chapter 3 Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure

3-1 3-2 3-3 3-4 3-5 3-6 3-7 3-8 3-9 3-10

Figure Figure Figure Figure Figure Figure

3-11 3-12 3-13 3-14 3-15 3-16

Figure 3-17

38 Components of a Generic Data Communications System ...................... 40 On-Off Keying (OOK) ............................................................................ 41 Non-coherent Matched Filter Structure .................................................. 42 Frequency Shift Keying (FSK) ............................................................... 43 M=8, M-ary Multi-Frequency Shift Keying (MFSK) ............................. 44 Binary Phase Shift Keying (BPSK) ........................................................ 45 8-PSK Constellation ................................................................................ 45 QPSK/MPSK/QAM Modulator .............................................................. 46 16-QAM Constellation ............................................................................ BER Performance of Various Modulation Schemes in an AWGN 48 Channel ................................................................................................... 49 Linear Transverse Equaliser .................................................................... 50 Decision Feedback Equaliser (DFE) ....................................................... Performance of Various Forward Error Correction Codes ..................... 53 54 Operation of an (mxn) Block Interleaver ................................................ 55 Structure of MIL-STD-188-11 OA waveform .......................................... Measured Performance of 1200 bps MIL-STD-188-11OA Modem 56 (BER=10-3) .............................................................................................. Current Progress in Development of HF Data Communications (to 57 2001) .......................................................................................................

Chapter 4 Figure 4-1

16 kbps Modem Waveform Structure

.....................................................

61

ix

Figure Figure Figure Figure

4-2 4-3 4-4 4-5

Figure 4-6 Figure 4-7 Figure 4-8 Figure 4-9 Figure 4-10 Figure 4-11 Figure 4-12 Figure 4-13 Figure 4-14

Figure 4-15 Figure 4-16

16 QAM Constellation Employed in the 16 kbps Modem 62 ...................... 16 kbps Modem Laboratory Characterisation 63 ......................................... 16 kbps ISB Modem Performance in a Gaussian Noise Channel........... 64 16 kbps ISB Modem Performance in a Flat Fading Gaussian Noise Channel 64 ................................................................................................... 16 kbps ISB Modem Performance in CCIR Good Channel 65 ................... 16 kbps ISB Modem Performance in CCIR Poor Channel 65 ..................... Constant BER Surface for 16 kbps ISB Modem (HC codec, long interleave) 66 ................................................................................................ Constant BER Surface for 16 kbps SSB Modem (HC codec, long interleave) 66 ................................................................................................ Constant BER Surface for 4800 bps MIL-STD-188-110A Modem 67 (un-coded) ............................................................................................... Constant BER Surface for 2400 bps STANAG 4285 Modem 67 (convolutional coder, long interleave) .................................................... Constant BER Surface for 16 kbps ISB Modem using HC Codec 69 Channel Conditions Rician Interleave Long under and .......................... 70 Configuration for `On-Air' 16 kbps Modem Experiment ....................... Hourly Average BER between Cobbett Hill and Malvern using 16 kbps ISB Modem with HC Codec (histogram shows number of 71 data points in the data set) ....................................................................... Diurnal Kilobyte Frame Delivery Statistics between Cobbett Hill 72 Codec HC kbps Modem 16 ISB Malvern with using and ....................... Performance of Extended Bandwidth HDR Waveforms in AWGN Channel

Figure 4-17

...................................................................................................

74

Performance of Extended Bandwidth HDR Waveforms in a Rician Fading Channel (one non-fading and one -6 dB Gaussian fading 75 mode) .......................................................................................................

Chapter 5 Figure Figure Figure Figure

5-1 5-2 5-3 5-4

Figure Figure Figure Figure Figure Figure Figure

5-5 5-6 5-7 5-8 5-9 5-10 5-11

Figure 5-12 Figure 5-13 Figure 5-14 Figure 5-15

From Conventional Analogue Receiver to Software Radio .................... 78 A Super-Heterodyne Receiver Architecture .......................................... . 81 Direct Conversion `Zero IF' Receiver Architecture ............................... 82 Image Suppression in a Quadrature Mixer Due to Amplitude and 83 Phase Imbalance ..................................................................................... . 84 Super-heterodyne with Zero-IF Stage..................................................... 85 Receiver IF-Sampling Conversion Single .............................................. . 86 Receiver Digital HF Wideband Diagram Block of a .............................. Block Diagram of a Wideband Digital Transmitter ................................ 87 90 Distance Separation Path Loss versus .................................................... . 91 Snapshot of Measured Signal Power in HF Band [GEC, 99] ................. Typical Narrowband Superhet HF Receiver Composite Filter 92 100] [after Pearce, Characteristics ......................................................... . 93 Reciprocal Mixing in a Receiver with Frequency Translation ............... 2ndand 3`dOrder Intermodulation Product Levels versus Two Tone 95 Input Power . ............................................................................................ 97 Wideband Direct Sampling Digital Receiver ......................................... . 98 Front End Anti-Aliasing Filter Performance Requirement .....................

x

xi

Figure 5-16 Figure 5-17 Figure 5-18 Figure 5-19 Figure 5-20 Figure 5-21 Figure 5-22 Figure 5-23 Figure Figure Figure Figure

5-24 5-25 5-26 5-27

Figure 5-28 Figure 5-29

Error in Sampling Amplitude Due to ADC Aperture Uncertainty (Jitter) 102 .................................................................................................... Predicted AD6644 SNR versus Clock Jitter for Various Analogue Input Frequencies 103 .................................................................................. Typical High Quality Local Oscillator SSB Phase Noise Specification 105 .......................................................................................... ADC Quantisation Errors Due to DNLs [after Brannon, 111] 106 .............. Architecture of High Performance AD6644 14-bit Multi-Stage ADC [Analog, 107] 106 ............................................................................... Application of Wideband Dither to Improve ADC SFDR 107 .................... Improvement in AD6644 Spurious Performance with Addition of a Dither Signal [Analog, 107] 108 .................................................................. Predicted Mean Available Dither Power Due to HF Congestion (Lower Bound) 109 ...................................................................................... Digital Down-Converter 110 ........................................................................ NCO as a Complex (Quadrature) Direct Digital Synthesiser 112 ............... 113 Practical Decimating CIC Filter - Integrator, Decimator and Comb .... Frequency Response of CIC showing Impact of Aliasing (M=100, 113 L=4, R=1) .............................................................................................. Modelled Frequency Response of CIC Filters as a function of L and R ..................................................................................................... 114 114 Dynamic Range in the Direct Sampling Digital HF Receiver ..............

Chapter 6 Figure 6-1 Figure 6-2 Figure Figure Figure Figure Figure

6-3 6-4 6-5 6-6 6-7

Figure 6-8 Figure 6-9 Figure Figure Figure Figure

6-10 6-11 6-12 6-13

Figure 6-14

121 Block Diagram of Prototype Direct Sampling Digital HF Receiver .... Measured Selectivity of Combined 28 MHz Elliptic Low Pass 123 Filter ...................................................................................................... 123 Front-End Filter Group Delay Variation ............................................... 124 Prototype Front-End Filter Impedance Matching ................................. 125 SNA586 GaAs RF Amplifier Test Circuit ............................................ 125 MMIC RF Amplifier Test Pieces .......................................................... SNA-586 GaAs RF Amplifier Characteristics Measured on 126 Network Analyser ................................................................................. 127 SNA-586 GaAs RF Amplifier Gain and Linearity Measurements ....... SNA-586 RF Amplifier 2°d and 3rd Harmonic Performance 128 (Extrapolated) ........................................................................................ Selectivity of 5thOrder Harmonic Filter (Modelled with SPICE) ........ 129 Programmable Digital Down-Converter [after Graychip, 124] ............ 131 GC4014 DDC NCO Implementation [Graychip, 124] ......................... 131 Controlled Oscillator (NCO) Spurs GC4014 Numerically 132 [Graychip, 124] ..................................................................................... Frequency Response of GC4014 CFIR Filter (3 kHz Nyquist

bandwidth) Figure 6-15 Figure 6-16 Figure 6-17 Figure 6-18

133 ............................................................................................ GC4014 DDC (3 kHz Nyquist

Modelled Performance of 133 bandwidth) ............................................................................................ Effect of Adding Dither to ADC Input Signal (Input tones are 137 20 dBFS) ............................................................................................... 137 Measured Digital Receiver Sensitivity: -158 dBm/Hz ......................... Measured Digital Receiver IMD using 2 tones at -20 dBm .................. 138

xi

xii

Figure 6-19 Figure 6-20 Figure 6-21 Figure 6-22 Figure 6-23

Measured Receiver Instantaneous Dynamic Range with dBm -15 input 139 ...................................................................................................... 62.208 MHz Sampling Clock PhaseNoise (Free Running) 140 ................. 62.208 MHz Sampling Clock Performance (VCXO locked to TCXO) 141 .................................................................................................. 62.208 MHz Sampling Clock Phase Noise (VCXO/TCXO/Ext. Standard) 141 ............................................................................................... Receiver Under-sampling Performance: -15 dBm input at 200

Chapter 7 Figure 7-1 Figure Figure Figure Figure Figure Figure Figure

7-2 7-3 7-4 7-5 7-6 7-7 7-8

Figure Figure Figure Figure Figure Figure

7-9 7-10 7-1 1 7-12 7-13 7-14

Figure 7-15 Figure 7-16 Figure 7-17 Figure 7-18 Figure Figure Figure Figure

7-19 7-20 7-21 7-22

Figure 7-23 Figure 7-24

Prototype (Serial No. 001) Digital Transceiver with Dual SHARC 147 DSP ....................................................................................................... 149 Architecture of Wideband Digital Transceiver ..................................... 153 SHARC ADSP-2106x Architecture (from [Analog, 135]) ................... 155 Dual SHARC Processor Module with 2Mx48 Shared Memory ........... 155 Dual SHARC Processing Module Interconnections ............................. 157 Digital Transceiver Configuration Architecture (Simplified) ............... Photograph of Digital Transceiver RF Sub-Systems and Interfaces..... 159 Block Diagram of Digital Transceiver Frequency Standard Sub160 System ................................................................................................... 161 Frequency Standard User Interface ....................................................... 162 Excerpt of Frequency Standard Sub-System Schematic ....................... CPLD Design for Frequency Standard Showing Control Register....... 163 CPLD Implementation of Phase-Frequency Detector (PFD) ................ 164 165 Phase-Frequency Detector (PFD) State Machine ................................. One Channel of Digital Receiver RF Font-End (Excerpt from 167 Schematic) ............................................................................................. 168 Dither Generator (Schematic Excerpt) .................................................. Block Diagram of 4-Channel Digital Receiver ASIC [Graychip, 169 124] ....................................................................................................... 170 Digital Receiver Graphical User Interface ............................................ Block Diagram of 4-Channel Digital Transmitter ASIC [Graychip, 172 146] ....................................................................................................... Digital Transmitter DAC and RF Chain (Excerpt from Schematic)..... 173 174 Interface User Graphical Transmitter Digital ....................................... Triple PCI Front End Protection and Filtering Module ........................ 176 Digital Transceiver PCB 'Stack-up' (0.062"±0.008 Finished 177 Thickness) ............................................................................................. 179 Class Diagram for Digital Transceiver (CDigitalTransceiver) ............. 180 Platform Transceiver for Digital Diagnostics ......................................

Chapter 8 Figure Figure Figure Figure Figure

8-1 8-2 8-3 8-4 8-5

187 PLX9054 Messaging 120 Implementation of using .............................. 189 ]) 13 1 [PLX, (from Queues Organisation of 120 Messaging ................ 190 Structure of Message Header ................................................................ 190 Structure of Message ID Field .............................................................. 195 UML Diagram of Asynchronous 1,0 Messaging Implementation ....... xii

VIII

Figure 8-6

Five Simultaneous ProcessesRunning on Digital Transceiver

.............

199

Chapter 9 Figure 9-1 Figure 9-2 Figure Figure Figure Figure Figure Figure Figure

9-3 9-4 9-5 9-6 9-7 9-8 9-9

Figure 9-10 Figure 9-11

Alternative Sounding Geometries 202 ......................................................... Vertical lonogram Produced by a Digisonde Pulse Compression Sounder 201 ................................................................................................. Comparison between Pulse and Chirp Sounding [Barry, 165] 205 ............. Configuration of WHISPER Receive System 208 ....................................... UML Sequence Diagram Illustrating Sounder Receiver 210 ...................... Pulse Compression Waveform Performance Metrics 211 ........................... NT Chip PN-Sequence Periodic Autocorrelation Function 213 ................... Simulated Radio Filters (80 kHz Complex Baseband) 214 ......................... 64 kchip/s PN-1023 Pulse Compression Waveform in 80 kHz Channel 215 ................................................................................................. Band-limited 64 kchip/s PN1023 Waveform CIR versus Doppler Offset 216 ..................................................................................................... Band-limited 64 kchip/s PN1023 Waveform CIR versus Doppler Offset (close in)

Figure 9-12 Figure 9-13 Figure 9-14 Figure 9-15

Figure 9-16 Figure 9-17

217 Performance of as a Function of Frequency Offset .............................. 219 Use of Windows in Calculating the Scattering Function ...................... 220 WHISPER Occupied Bandwidth (Waveform: tx1023-64r) .................. WHISPER

Back-to-back

Test: Measured Ambiguity

Response

..........

220

WHISPER Back-to-Back Test: Complex Impulse Response 221 Resolution Back-to-Back Test: Doppler Resolution (512 CIRs, Hanning Window)

Figure 9-18

.....................................................................................

216

................................................................................................

221

Centre of Measured `tx1023-64r' Waveform Power Spectrum 222 (PRF=62.5 Hz) ......................................................................................

Chapter 10 Measured Channel Scattering Function (3.9 MHz, 10 Apr 2001 226 07: 38) .................................................................................................... 10-2 Measured Channel Scattering Function after Doppler Filtering 226 (3.9 MHz 10 Apr 2001 07: 38) ............................................................... 228 10-3 Oblique lonogram (10 Apr 2001 07: 38) ............................................... MHz (3.9 Function Scattering Channel View Magnified 10-4 of 228 10 Apr 2001 07: 38) ............................................................................... 10-5 Channel Power and Phase (Radians) plotted for 2.7 ms Mode 229 38) 07: 2001 10 April MHz, (3.9 Time Measurement versus ................ 10-6 Channel Power and Phase (Radians) plotted for -4.3 ms Mode 229 07: 38) 2001 10 April MHz, (3.9 Time versus Measurement ................ Mode for (Radians) Phase Channel ms Power 10-7 plotted -6 and 230 07: 38) 2001 10 April MHz, (3.9 versus Measurement Time ................ 231 31 UT) 20: 2001 10-8 lonogram (9 April ........................................................ 10-9 Measured Scattering Function (6.7 MHz, 9 April 2001 20:30) ............ 232 10-10 Measured Scattering Function after Doppler Filtering (6.7 MHz, 9 232 April 2001 20: 30) ..................................................................................

Figure 10-1 Figure Figure Figure Figure Figure Figure Figure Figure Figure

xiii

\I\

Figure 10-11 Power and Phase Plot of 3.7 ms Mode Showing Rapid Movement of Layer (6.7 MHz, 9 April 2001 20: 30) ............................................... 233 Figure 10-12 Oblique Ionogram (10 April 2001 08: 1OUT) 234 ........................................ Figure 10-13 Measured Scattering Function (5.7 MHz, 10th April 2001 235 Figure 10-14 Measured Scattering Function After Doppler Filtering (5.7 MHz, 10thApril 2001 08: l OUT) 235 ..................................................................... Figure 10-11 Power and Phase Plot of -3 ms Modes (5.7 MHz, 10thApril 2001 236 08: 1OUT) ............................................................................................... Figure 10-11 Power and Phase Plot of -5.4 ms Spread Mode (5.7 MHz. 10th 236 April 2001 08: 1OUT ..............................................................................

Chapter 11 Figure 11-1

Frequency Agile Simulator Architecture

..............................................

241

XIV

X\

List of Abbreviations ACF

Auto-correlation function

ADC

Analogue-to-digital converter

AGC

Automatic gain control

ALC

Automatic level control

ALU

Arithmetic logic unit

API

Application programming interface

ARP

Applied Research Programme

ARQ

Automatic repeat request

ASIC

Application specific integrated circuit

ATU

Antenna Tuning Unit

BDR

Blocking dynamic range

BER

Bit error rate

BIT

Built-in test

BLOS

Beyond line-of-sight

bps

Bits-per-second

BPSK

Binary PSK

CAST

Configurable

radio

with

advanced software

technology

(EU

research

programme) CCF

Cross-correlation function

CDAA

Circularly disposed antenna array

CFIR

Compensating FIR (filter)

CIC

Cascaded integrator-comb

CIR

Complex impulse response

CRC

Communications research centre (Ottawa, Canada)

CRP

Corporate research programme

CVSD

Continuously variable slope delta-modulation

CW

Continuous wave

xv

DAC

Digital-to-analogue converter

DAMSON

Doppler and multipath sounding network (HF sounder)

dBFS

Decibels below full-scale

DCE

Data communications equipment

DDC

Digital down-converter

DERA

Defence Evaluation and Research Agency (UK MOD)

DFT

Discrete Fourier transform

DLP

Data-link protocols

DNL

Dynamic non-linearities

DMA

Direct memory access

DOD

US Department of Defence

DPP

Delay power profile

DRM

Digital radio Mondiale

DSP

Digital signal processing (or processor)

DTE

Data terminal equipment

DUC

Digital up-converter

E2PROM

Electrically erasable programmable memory

EEPROM

Electrically erasable programmable memory

FEC

Forward error correction

FFT

Fast Fourier transform

FIR

Finite impulse response (filter)

FM

Frequency modulation

FMCW

Frequency modulated continuous wave

FPGA

Field programmable gate array

FSK

Frequency shift keying

GPS

Global positioning system (US DOD satellite navigation system)

GUI

Graphical user interface

HC

Hyper-code

XVI

HDR

High data rate

HF

High frequency

120

Intelligent Input/Output

IDR

Instantaneous dynamic range

IF

Intermediate frequency (radio systems)

IIR

Infinite impulse response

IMD

Intermodulation distortion

INL

Integral non-linearity

I/O

Input/output

IRQ

Interrupt request

ISB

Independent side-band

ITS

Institute of telecommunications sciences (US Federal agency)

ITU

International telecommunications union

JTAG

Joint test action group

JTRS

Joint tactical radio system (US military software radio programme)

kbps

Kilo-bits per second

LED

Light emitting diode

LO

Local oscillator

LPF

Low pass filter

MCI

Module control interface

MFA

Message frame address

MFLOPS

Million floating point operations per second

MFSK

Multi-tone FSK

MIMO

Multiple-input, multiple-output

MMIC

Miniature microwave integrated circuit

MSPS

Mega-samples per second

NATO

North Atlantic Treaty Organisation

NF

Noise figure

zvii

111

NVIS

Near vertical incidence sky-wave

OFDM

Orthogonal frequency division multiplex

OMG

Object Management Group

OS

Operating system

PA

Power amplifier

PC

Personal Computer

PCB

Printed circuit board

PCI

Peripheral component interconnect

PFD

Phase-frequency detector

PLL

Phase locked loop

PN

Pseudo noise (sequences)

PFD

Phase-frequency detector

PFIR

Programmable FIR (filter)

PPF

Deterministic phase function

PPM

Parts-per-million

PRF

Pulse repetition frequency

PRI

Pulse repetition interval

PSK

Phase shift keying

PU

Participating Unit

QAM

Quadrature amplitude modulation

RF

Radio frequency

RAM

Random access memory

RMS

Root mean squared

RRS

Recursive running sum

RTOS

Real-time operating system

SCA

Software communications architecture (for JTRS)

SDR

Software defined radio

SFDR

Spurious free dynamic range

xviii

xix

SGL

Scatter gather list

SHARC

Super Harvard architecture (DSP processor)

SIG

Special interest group

SINAD

Signal plus noise and distortion to noise and distortion ratio

SMF

Stochastic modulating function

SMT

Surface mount technology

SNR

Signal-to-noise ratio

SRAM

Static random access memory

SSB

Single side-band

SWF

Shortwave fadeout

TCXO

Temperature compensated crystal oscillator

TDMA

Time division multiple access

TOF

Time of flight

UML

Unified modelling language

USB

Upper side-band

VCXO

Voltage controlled crystal oscillator

WHISPER

Wideband HF Ionospheric Sounder for Propagation Environment Research

\IX

Chapter 1 Introduction

1

Chapter 1.

Introduction

This thesis describes novel work in the areas of digital radio and its application to the high frequency (HF, 2-30 MHz) band. The propagation environment at these frequencies has a significant impact on the performance of radio systems (especially data communications). However, the ability to communicate information over very long ranges using ionospheric propagation paths without any intermediate infrastructure makes the use of HF attractive for many applications. In order to increase the utility of HF communications there is a strong desire to increase HF data rates. Currently data rates of up to -2400 bps can be reliably achieved in standard 3 kHz HF channel allocations.

Whilst

further increases in data rate within the confines of these

narrowband frequency allocations is likely, the use of larger bandwidths (contiguous or otherwise) appears to offer potential for much greater throughputs. This requires a greater understanding of the characteristics of wideband channels and also requires transmitting and receiving equipment capable of wideband/multi -channel operation. This thesis details research undertaken over a three year period between April 1998 and April 2001 funded by the author's employer, the Defence Evaluation and Research Agency (DERA), part of the UK Ministry of Defence. The research was undertaken as Research Communications the at and engineers a working collaboration with scientists Centre (CRC) in Ottawa, Ontario, Canada under a UK-Canadian government agreement [MOD, I] defence technology and science on collaboration on The original aim was to investigate the characteristics of wideband HF ionospheric propagation

channels

communications.

in

order

to

exploit

them

for

high

data rate (HDR)

Very early on an opportunity was identified to work with CRC to

implement and demonstrate a HDR modem providing 16 kbps in a6 kHz bandwidth. The success of this work contributed to the standardisation of extended bandwidth HDR [DOD, 2]. IOB MIL-STD-188-1 in a widely used military modem standard; waveforms Work started on the development of a new channel sounder to provide data to be used in the investigation of the wideband channel.

This involved the design of wideband

Chapter 1 Introduction

transmission and reception equipment. What started out as a means to an end became a key focus of the research as it became apparent that it would be possible to build a high performance direct-sampling digital HF radio. A new wideband digital HF transceiver (HF software radio) with such an architecture has been implemented and is presented in this thesis. This was used as a platform on which a new wideband channel sounder, known as WHISPER, has been developed. The development of WHISPER and its use to investigate the wideband (-80 kHz) HF channel is also documented.

1.1

Structure of Thesis

This chapter provides an introduction and some background to the research presented in this thesis. Chapter 2 introduces the HF propagation environment (particularly

communications).

summarised.

HF channel simulation

communications

The

effects

of

and its impact on radio systems

noise

and interference

are also

techniques that may be used to develop and test

systems are discussed. Characteristics of the wideband HF channel and

identifying models considered, applicable channel are

a number of areas in which work

is required to provide a better understanding of the processes at work.

In chapter 3 the key technologies used to implement data communications waveforms interleaving, forward including modulation and channel error correction, are reviewed equalisation.

Practical application to HF is illustrated by examining the design of a

modern narrowband serial tone waveform.

Finally techniques applicable to wideband

introduced. high throughput are communications and Chapter 4 describes a novel high data rate 16 kbps prototype modem operating in an kHz. 6 bandwidth of extended

Results from HF simulator measurements and on-air

data high limitations The rate of such testing of the modem are presented. performance for identification discussed leading the which applications to of range of an modems are benign be be on they can used reliably: notably on surface wave paths and expected to Skywave paths. Finally a number of alternative extended bandwidth waveforms are high developed that of recently proposed and their performance compared with throughput narrowband (3 kHz) waveforms. Chapter 5 focuses on the architectures and implementation of wideband and multihigh HF environment places on channel digital HF radios. The requirements that the

I

Chapter 1 Introduction

3

performance receiver design are considered. Results obtained from the characterisation of a high quality conventional narrowband HF receiver are given to establish a basis for comparison. A very high performance, direct sampling wideband digital HF receiver is proposed. Such a receiver would (conceptually) allow an arbitrary number of channels to be simultaneously received using a single RF front-end and digitiser. The design and performance of a practical receiver of this type is considered in detail. In chapter 6 the design and implementation of a new prototype direct-sampling digital HF receiver is presented along with measured performance results. These are discussed and suggestions for improvements to the prototype design are advanced Chapter 7 describes the implementation of a wideband, multi-channel digital HF transceiver, using the direct sampling architecture. The design of the transceiver, which has been specifically conceived as a highly re-configurable software defined radio system, is explained.

Descriptions of the key sub-systems including the diversity

receiver, transmitter, the digital signal processing (DSP) sub-system and the control software are given. Chapter 8 deals with a particular challenge of complex systems such as software radios the effective deployment of application software across a number of heterogeneous processors.

This chapter proposes a new asynchronous, event-based, processing

architecture which employs lightweight (low overhead) active messaging to allow processing tasks to be effectively distributed across multiple processors and across buses. The architecture and its effective implementation on the digital transceiver hardware are described. Suggestions for further developments and improvement have been made. Chapter 9 provides a brief introduction to channel sounding techniques before introducing, WHISPER, a new oblique wideband HF ionospheric sounder. The design is based low this on software radio of power pulse-compression sounder, which techniques, is introduced and its implementation as an application on the wideband digital HF transceiver described. The chapter then considers the design of sounding waveforms

suitable for

an investigation

kHz) (-80 the wideband of

channel

Finally, the Skywave results of channels. propagation characteristics of mid-latitude back-to-back measurements performed at RF are analysed to confirm the performance of the sounder.

3

Chapter 1 Introduction

4

Chapter 10 describes a short campaign of wideband measurementsthat have been made over a 170 km path in the UK during Spring 2001 using the WHISPER sounder. The results of these measurements have been analysed and confirm the ability of the sounding instrument to measure the channel scattering function and the amplitude and phase within individual modes. A number of possible directions for a more detailed analysis of the data are then suggested. Chapter 11 identifies a number of areas in which digital HF radio may be applied to advantage. Finally, chapter 12, draws some conclusions from the work presented in this thesis and identifies recommended areas for further work.

1.2

Original Work

This thesis contains several elements of original work in the areas of high throughput HF communications, high performance HF digital radio architectures and their implementation, implementation of a wideband HF channel sounder using the digital radio and in the conclusions drawn from measurements obtained during on-air trials.

the analysis of wideband channel

Specific contributions to development of

the field are identified below:

1.2.1

High Throughput

HF Data Communications Employing

Extended Bandwidth (6 kHz) Channels New waveforms have been proposed for the transmission of higher data rates in an laboratory kHz). The bandwidth (6 tests and analysis of HF of results channel extended data collected during on-air trials of these 16 kbps waveforms are presented. Analysis undertaken indicates that these new waveforms are capable of reliable operation over Skywave benign HF channels providing a sufficient received signaland wave surface to-noise ratio can be maintained.

The work presented has contributed to the

international standardisation of higher throughput waveforms (specifically US MILSTD-188-11OB). It has also demonstrated the value in exploiting wider bandwidths for HF radio applications.

4

Chapter 1 Introduction 1.2.2

j

Architectures

for Wideband

Digital Receivers and Transmitter

Exciters Architectures (software

for the implementation

radios) have been investigated.

digital HF receiver has been proposed.

A new direct-sampling

allow an arbitrary number of channels to be simultaneously

propagation,

architecture

for a

In such a receiver the entire HF band is digitised

and then all subsequent processing is undertaken digitally.

front-end and digitiser.

digital HF radios

of wideband and multi-channel

Conceptually

this would

received using a single RF

The requirements for high performance receivers due to the HF

noise and interference

performance conventional

have been established.

environment

A high

narrowband super-heterodyne receiver has been characterised

to establish a basis for comparison.

A design for a practical direct sampling receiver is

proposed and analysis presented which indicates that is now possible, for the first time, to construct a high performance receiver of this architecture.

The work shows that, with

careful design, such a receiver should attain performance

comparable with (or even

exceeding)

that of the majority

receivers.

A prototype

transceiver confirming

1.2.3

platform)

wideband

has been

of conventional direct

super-heterodyne

sampling

implemented

and

receiver its

single

channel

(part of a digital

performance

HF

characterised

the potential for such a design.

Implementation

Transceiver Digital HF Wideband of a

A new high performance wideband, multi-channel HF transceiver (HF software radio) has been implemented. It designed has been direct the and sampling architecture using implemented as a highly re-configurable software defined radio platform intended to including diverse use as a radio/modem, channel sounders and an applications support four feeding front diversity incorporates It RF channel simulator. end a two channel independent receiver channels and a four channel transmitter exciter with a single RF output.

The design is very compact, being implemented on a single peripheral

host into be interconnect (PCI) personal computer a plugged card which can component (PC).

Chapter 1 Introduction

1.2.4

6

Development and Implementation

of a High Performance

Asynchronous Event Based Processing Architecture A

lightweight

event based processing architecture for use across an array of

heterogeneous processors has been conceived, developed and implemented on the wideband digital HF transceiver platform. It uses an active messaging concept in which messagesarriving in a queue cause pre-defined processing activities to take place.

1.2.5

Design and Implementation

of a Wideband HF Channel

Sounder using the Digital HF Transceiver A new, high resolution wideband oblique HF channel sounder, which has become known as WHISPER, has been developed and implemented as an application on the wideband digital HF transceiver platform.

The system is very flexible with sounding

waveforms and their characteristics (sounding waveform, bandwidth, repetition interval file. determined by being configuration a etc)

The sounder makes use of an external

GPS receiver to provide accurate synchronisation of transmitter and receiver in order to be to time-of-flight measurements made. Pulse-compression sounding waveforms allow fidelity high have been designed to time allow sidelobes with carefully controlled kHz) HF (-80 the time channels. wideband of characteristics varying measurement of The high performance of the complete sounding system has been verified through backto-back RF tests in the laboratory.

1.2.6

Analysis of On-Air Wideband HF Channel Characterisation Measurements

The WHISPER wideband sounder has been used to make high resolution measurements UK. in km 170 the impulse southern path response on a of the time varying complex The results of these measurements have been analysed and confirm the ability of the function the instrument the amplitude and and to scattering channel measure sounding detailed for directions A individual more a number of research modes. phase within been have HF design proposed. modems, of wideband analysis, pertinent to the

6

Chapter 2 The HF Propagation Environment and Its Impact on Communications

7

Chapter 2.

The HF Propagation Environment and Its Impact on Communications

This chapter introduces the HF propagation environment and its impact on radio systems particularly communications.

For a much more comprehensive and detailed

treatment of the subject the reader is referred to the texts that have been extensively used in this chapter's preparation; [Davies, 3], [Goodman, 4] and [Maslin, 5]. HF radio propagation (2 to 30 MHz) provides both line-of-sight (LOS) and beyond lineof-sight (BLOS) coverage and can occur by a variety of mechanisms. Direct wave propagation provides true LOS communications whereas the surface wave mechanism is BLOS and especially effective over sea paths supports shorter range communications providing typical ranges of 200 km or more. For greater ranges sky wave propagation must be exploited.

2.1

Surface Wave Propagation

Surface wave (or ground wave) propagation is supported over short distances on land (perhaps up to -30 km depending on terrain) and at much greater distances over highly 5]. [Maslin, sea water conducting vertically

This mode of propagation requires the use of

is commonly and polarised antennas

communications.

line-of-sight for extended used

Curves which show field strength versus range for ground wave

6]. 368-7 [ITU, in P. ITU-R propagation are given

2.2

Sky Wave Propagation

2.2.1

Structure of the Ionosphere

Sky wave communications, which involve transmitting signals beyond the radio horizon, rely upon refraction of the signals by the earth's ionosphere [Davies, 3]. The ionosphere is a highly inhomogeneous ionised region of the earth's atmosphere lying in

7

Chatter 2 The HF Propagation Environment and Its Impact on Communications the altitude including

range 85 - 1,000 km.

ultra violet (UV),

Its presence is primarily

(loss processes) occur due to the collision of

ions and the attachment of electrons to neutral gas atoms or

electrons and positive molecules (the principal

density

due to solar radiation

x-rays and energetic charged particles all of which cause

ionisation of neutral gases. Recombination

ionisation

8

chemical reactions are given in [Davies. 3. p63] ). Different

and loss processes become predominant

and temperature

profiles

at different

as well as incident

altitudes

solar radiation)

(related to

resulting

in a

layered structure (Figure 2-1). The principal regions of the ionosphere are designated D. E and F. Some of these regions are themselves layered or structured (e. g. E, E, F1 and F2). The number of layers, their heights and their ionisation density vary with time and in space. Variations year solar cycle

occur diurnally

(Figure 2-1, Figure 2-2), seasonally, with the II

and due to changes in geomagnetic

ionosphere is particularly

complicated

The high latitude

activity.

and will often be significantly

different

to that

lower latitudes. mid and observed at

Sun

F2 4%,W

-

F1

D 40 ay it

Figure 2-1 The Earth 's Ionosphere and its Principal

2.2.2

Regions [Mslin,

5]

D-Region

km 90 from 70 for The D-region, principally responsible signal absorption. extends HF leads ionisation to day signal strong ionised. During is the solar only weakly and D-region hours sunset latitudes of low of At couple a within and absorption. mid

8

Chapter 2 The HF Propagation Environment and Its Impact on Communications absorption becomes negligible

contributing

9

to the stronger signal strengths and

increased noise experienced in the HF band at night. As the region is caused b} solar radiation it is observed to be stronger during the summer than the winter.

300

F2

F1

200 rn

Night i 100

D

109

Day

1010

10"

1012

Electron Concentration (M-3)

Figure 2-2 Typical Electron Concentration within the Ionosphere [Davies, 3]

2.2.3

E-Region

The E and F regions are predominantly responsible for sky wave reflections. The level km 110 ionisation in E-region induced the of with the peaks near an altitude of solar daytime region generally being very regular. At night only residual ionisation remains and the E-region virtually

disappears.

The E-region plays an important part in

km). (70°. During large flares highly energetic protons are

Polar Cap Absorption

by ionisation ionosphere, On they the colliding with gas cause entering released. molecules.

The resulting PCA can last for as long as several days exhibiting

in high day The the polar cap and time. the through effect starts strong absorption is linked PCA to the The sunspot events of occurrence southward. can move cycle. "

Ionospheric Storms - Whilst SIDS last for short periods of time and PCAs are lowionospheric latitude high and midaffect storms phenomena, principally a latitudes.

Ionospheric

storms include

geomagnetic storms, auroral and

by D-region be These absorption accompanied may etc. storms magnetospheric (expansion F-region E. and storms (auroral and sudden commencement), auroral These which storms, F2), diffusion of emissions. noise radio and scintillation, being particles days, charged of last the stream of a result are several may Whilst field. the by the magnetic earth's deflected towards the auroral regions impact biggest terrestrial the on maximums maximum effects are observed at solar be likely is to at solar minima. radio systems

2.2.7

Ionospheric Propagation

back directed they that are The ability of the ionosphere to refract HF radio signals so level by determined is the of into (rather space), than passing towards the earth ionisation

frequency of angle the and and present

incidence of the signals.

dependent is many upon for BLOS frequency communications Consequently, selection factors including the link geography and time-of-day.

13

Chanter 2 The HF Propagation Environment and Its Impact on Communications

Escape ray

Skip ray

14

High angle rays

Low angle rays

Critical angle, 14a Zip distance

Figure 2-5 Ray Paths as a Function of Elevation at a Single Frequency [Maslin, 5] Figure 2-5 shows how the trajectory of a single frequency signal varies as a function of by ionosphere be Low the and return to earth reflected elevation signals will elevation. for distance from launching The the sufficient critical angle antenna. at a considerable its from layer is to to to electron a proportional occur produce reflections refraction density. As the elevation of the transmitted radio wave increases it is reflected to earth between be The the received, signals can no region where at shorter and shorter ranges. transmitter and where the sky wave returns fall, is known as the skip zone. At still higher elevations the ionosphere does not refract the signal sufficiently to reflect it immediately and it penetrates further into the ionosphere and is dispersed over much launch the When distances (high the critical angle exceeds elevation ray). angle greater for the layer altogether the ray will pass through it and is termed an escaperay. When an electromagnetic wave interacts with electrons in the ionosphere the earth's (0) into the and two field to ordinary the components; split signal causes magnetic 0in interactions the Further (together resulting (X) 'O-X'). result waves extraordinary detail in discussed is This greater having much X wave a particular polarisation ellipse. in [Davies, 3, pp.226-232] and the reader is referred there for a more complete description. distance for (MUF), frequency particular The maximum usable communications at a by: is layer for approximated a specific and MUF = fo sec0

(2-1)

incidence the is 0 layer on the frequency of angle f, is the and, the critical of where reflecting layer.

Over longer distances the curvature of the earth and the vertical

14

Chapter 2 The HF Propagation Environment and Its Impact on Communications

15

ionisation profile must be taken into consideration so a correction factor, k, whose value falls between 1.0 and 1.2, is introduced [Davies, 3], [Van Valkenburg, 8]: (2-2)

MUF = kf, sec0 Figure 2-6 depicts a VOACAP from Frankfurt, circuit sky wave

[Hand, 9] monthly median propagation prediction for a Germany to London, UK.

the MUF and circuit propagation loss typically

The figure illustrates how

vary with diurnal, seasonal and sunspot

activity.

12-

b. January, SSN = 150

a. January, SSN = 10

\(1\(

\P em 4,:.

SY,

x

1401

ýIfVF

-. ýI06 -

mm 00

ý. 5.00

__

d. July, SSN = 150

c. July, SSN = 10

Figure 2-6 Predicted Propagation

Loss and MUFfor

Frankfurt-to-London

Sky wave Circuit

in differences East-West (with component) Over long distance communications paths an ionisation degrees in and structure of varying the time-of-day are encountered, resulting frequencies, higher favour propagation in the ionosphere. Daylight portions of the path This the frequencies. of lower selection favour make can time portions whereas night day/night the for frequencies so-called across communications communications suitable terminator (see Figure 2-7) a difficult task.

15

Chapter 2 The HF Propagation Environment and Its Impact on Communications

Figure ?

2.3

Position of'Dav-Night -7

Terminator ut 100Z

16

in December

NVIS Propagation

Near vertical incidence sky wave (NVIS)

km and beyond [Maslin,

to ranges of -150 platforms

supports terrain independent communications

at low-level

operating

5].

are required

Where ground stations or airborne to communicate

deep in hilly

or

mountainous areas, the NVIS mode of operation frequently provides the only means of direct communication. significant

vertical

frequency

radiation

management.

time absorption) electrically

2.4

The effective

use of NVIS

component

Achieving

can be a particular

requires antennas that provide a

(i. e. horizontally

polarised)

and careful

level (and overcoming signal an acceptable

day

when low powered transmitters

and

challenge

small antennas are utilised.

Impact of Propagation

This section introduces the principle

on Radio Waves HF that signals propagating perturb mechanisms

be have. Whilst impact describes to these they the that considered are often effects and purely detrimental

the following

discussion

also seeks to identify

how the resulting

by in the be in exploiting to system a suitable advantage signals can used changes resulting diversity in the received signals. The ionosphere is a dispersive medium leading to spreading of the pulses travelling through it.

Reflections from multiple layers combined with multi-hop propagation

be the dispersion in Time order of of may multipath and results multipath returns. Hertz Doppler (fading) of many (occasionally >_10ms), and spread several milliseconds (occasionally >_40Hz) may be observed. The severity of these effects is particularly 16

Chapter 2 The HF Propagation Environment and Its Impact on Communications

17

significant at high latitudes (above 65°, i. e. in auroral and polar cap regions, [Davies, 3], [Angling, 10]) where the ionosphere is severely affected by energetic particles arriving indirectly from the sun. Similar disturbances may also be observed at low latitudes (i. e. within 15° of the equator).

2.4.1

Attenuation

The principal causes of radio wave signal attenuation are: "

Free space Propagation Loss essentially the loss due to the spatial spread of energy.

0

Environmental

Absorption

and Ground

Reflection

Losses

Absorption -

in the

environment due to low conductivity media (e.g. terrain losses) and surface scattering. "

Reflection Losses - Imperfect reflection of signals or scattering in the ionosphere.

0

Ionospheric absorption - Absorption in the ionosphere.

"

Polarisation

Loss - An inability to pick-up all the available power at a receive antenna because it does not match the polarisation of the incoming radio wave.

2.4.2

Multipath

and Signal Dispersion

A number of mechanisms signal to be received.

cause multiple,

These are commonly

returns (modes) can be identified.

time delayed, versions of the transmitted termed multipath where distinctly

separate

The principal forms of multipath and time dispersion

are:

0

both Under a ground conditions where be (Figure 2-8a) there a will wave and sky wave component can propagate Groundwave/Skywave

Interaction

is flight (>2.5 in differential time the possible on short ms of significant relative links during the day-time). 0

Reflections from Different Layers - Even for narrow transmitter antenna vertical beam widths, signals are launched with a range of elevation angles. Radio waves for the be if a angle critical they arrive at an angle exceeding will only reflected layer to produce reflections and this is proportional to its electron density.

17

Chapter 2 The HF Propagation Environment and Its Impact on Communications

18

Multipath is generated when signals arrive at the receiver having been reflected b% different layers with different electron densities (Figure 2-8b). 0

Differing Number of Hops - Signals arrive having undergone a different number of ionospheric and ground reflections (i. e. hops), see Figure 2-8c.

0

High angle/Low angle - As previously described, the angle at which radio w, ýaves impinge on the ionosphere can result in them taking substantially different trajectories to the receive location (Figure 2-8d).

"

Mode Dispersion

-

Finite antenna beam widths and the thickness of the

ionospheric layers and their varying refractive index, causes a continuum of dispersion Typical dispersion (Figure 2-8e). time the receiver; mode returns at HF [Maslin, 5] be systems are narrowband although within a mode may -200 µs is in dramatic A this phenomenon to example of more unable resolve such effects. the presence of spread-F; when the F-region is inhomogeneous contains many irregularities. Under such conditions, and particularly at high latitudes, delay spreads of several milliseconds are possible. "

is incident the When transmitted upon signal a ionosphere it leads to the excitation of differently polarised waves termed the

Magneto-ionic

Splitting

2-8f. The Figure O-X), (together see abbreviated ordinary and the extraordinary differing to independently amounts of O-X waves then propagate and are subject is frequency is delay to The and delay, fading and attenuation. related relative to Narrowband resolve in unable are systems typically measured micro-seconds. by detected be O-Xwaves may only this small time delay and thus the presence of With fading). (flat fading wider leads to interaction single-mode their which directly the two it becomes to kHz) observe (> bandwidths possible reception -50 design. importance to system wideband returns and thus they are of

18

Chapter 2 The HF Propagation Environment and Its Impact on Communications

19

F-laver ýr

a. Groundwave/Skywave

b. Different Layers

F-laver ayer

c. Different number of hops

d. High angle / low angle

Ordinary Extra-ordinary

e. Mode Dispersion

f. Magneto-ionic splitting

Figure 2-8 Summary of the Causes of Multipath and Dispersion [after Muslin, 5]

It should be noted that it is quite possible for a number of these mechanisms to be at work simultaneously causing a complex series of multipath modes, some with in 2-9 Figure illustrated is This be dispersion, which to received. appreciable [Arthur, II]. CW FM IRIS ionogram sounder collected using an reproduces an oblique

19

Chapter 2 The HF Propagation Environment and Its Impact on Common ccnj °

r rgure

2-9

l ypical

oblique

tit

lonogram

(Malvern

to Farnborough

I /. --I] ! U-Ui-!

2O

UU)

The coherence bandwidth of a system is proportional to the reciprocal of the multipath dispersion.

In a system with a bandwidth smaller than the coherence bandwidth the

impact of multipath is observed as non-frequency selective `flat' fading. In a system fading be bandwidth bandwidth the the than observed to will coherence greater with a be frequency selective as the interaction of the multipath components varies between frequency. function interference destructive of as a constructive and

2.4.3

Doppler Effects

There are two principal mechanisms that cause Doppler effects to be applied to induced Doppler. ionospheric Doppler and platform propagating signals: Ionospheric Doppler is caused when the reflecting layer is moving in such a way that the overall path length is changing. The Doppler shift is proportional to the rate of t, travelling frequency, For v, velocity, relative the with a signal phase path. change of fD, is: light, Doppler to the shift, c, speed of close

c

20

Chapter 2 The HF Propagation Environment and Its Impact on Communications

21

Given a radio wave reflecting off a single, ionospheric layer moving vertically with speed v,, the imposed Doppler shift will be related to the angle of incidence with the reflecting layer, 0. There also arises a factor of two because both the upward and downward phase path lengths are changing: 4) fD

COS o C

In additional to purely vertical movements (such as normal diurnal changes or due to rapid changes in ionisation caused by ionospheric disturbances), ionospheric Doppler shifts can be caused in more complex situations.

Examples include off-great circle

reflections from moving irregularities, high angle sky wave propagation through an inhomogeneous ionosphere etc. Clearly, Doppler shift can also be imposed on a radio wave by the component of velocity

of the receiver relative to the transmitter (normal to the direction of

propagation). Values of radio platform induced Doppler shift are generally larger than those caused in the ionosphere particularly for fast moving platforms such as aircraft. Equation (2-3) is applicable. For example at 90 km/h (25 ms-') the maximum induced Doppler shift imposed on a 10 MHz signal would be -0.8 Hz. An aircraft travelling at 300 ms' would produce a Doppler shift of -30 Hz on a 30 MHz signal. A useful `ruleis Hertz by `One is Doppler the that a moving radio platform shift produced of-thumb' per Mach per Megahertz'.

The acquisition and tracking of signals in the presence of

is i. Doppler Doppler shift, e. rate, particularly challenging. changing The same effect as Doppler shift can also be caused by frequency offsets between radio is frequency due colloquially to errors or setting reference errors system equipments designed be for to HF is to It Doppler operate with typical systems termed radio offset. frequency Hz total the frequency offset. to ±75 combined cope with offsets of up to a

2.4.4

Fading and Doppler Spread

is in there fading is a variation of to The term which any situation generally applied by be a This work within at mechanisms time. caused may with received signal energy due 2-1 in Table identified to fading) or (intra-mode as single propagating mode interference between modes (inter-mode fading); Table 2-2.

2

Chapter 2 The HF Propagation Environment and Its Impact on Communications

22

The interference of multiple modes leads to the establishment of patterns of constructive and destructive interference repeated at the fading rate. This results in frequency selective fading. For a channel with two multipath components, which have an inter mode separation of d s, the correlation bandwidth of the channel is approximately I/d Hz [Proakis, 14, p764].

Where the fading is caused within a single mode it

generally results in non-frequency selective fading termed flat fading. Cause of Intra-Mode Fading

Fading Type

Fading Period

Comments

Small scale irregularities in F-

Flutter

10 -100 ms

T}picallý associated with SpreadF.

region. Movement

of irregularities

in

Diffraction

ionosphere.

Typically

follow a Rayleigh

distribution.

Rotation of axes of polarisation ellipse. Curvature of the reflecting

10 - 20 s

layer

Polarisation

10 - 100 s

Focusing

15 - 30 min

Absorption

60 min

Skip

Usually

Requires both 0 and X magnetoionic components.

[Davies, 3]. Time variation of ionospheric absorption. Time variation of the MUF

Has greatest impact at sunset and sunrise.

non-periodic

Table 2-1 Summary of the Causes of HF Fading within a Single Propagation

Mode

Mechanisms that cause fast fading (fade periods of less than -10 s) are of particular interest to the designers of HF data communications

have because they a major systems

impact on modulation schemes that rely on amplitude or phase stability and thus often require special adaptive processing to overcome them. Fading Type High/low

angle rays

Sky waves Ground wave/skywave Magneto-ionic

splitting

Fading Period 0.5 -2s 1-5s 2- 10 s 10 - 40 s

Table 2-2 HF Fading Due to Inter-Mode Interactions

be It Doppler is may termed Frequency dispersion of a propagating signal spread. ionosphere. fine the the of by structure a number of mechanisms associated with caused Varying degrees of off-great circle propagation can result in a signal containing a length). in the path (due phase Doppler to variations systematic shifts continuum of Experimental work by Watterson et al [Watterson, 12], [ITU, 13] suggested that midGaussian be a latitude Doppler spread in narrowband channels could modelled using

22 it

Chanter 2 The HF Propagation Environment and Its Impact on Communications

23

distribution.

However, where the reflecting layer is tilted or moving with a component of velocity perpendicular to the principal direction of propagation the resulting Doppler spread is likely not be Gaussian or even symmetrical.

2.5

Propagation Diversity

The preceding description of the perturbations suffered by trans-ionospheric radio waves indicates the difficulty of developing HF radio communications systems. However these mechanisms result in inherent diversity that can sometimes be exploited to improve the quality and availability of communications: 0

Time-of-Arrival/Frequency each mode will receiver [Proakis, gain.

Essentially

generally

Spatial

have uncorrelated

14, pp797-806] this exploits

bandwidth significantly

0

Diversity - Where a multi-mode signal is received, Doppler

spread.

these can be re-combined in-band frequency

diversity

In a Rake type

providing

diversity

where the channel

exceeds the correlation bandwidth.

Diversity

As two closely spaced receivers are moved apart the correlation between the received signals decreases (as their paths through the inhomogeneous ionosphere become significantly

different).

The correlation

distance is defined as the distance at which the correlation coefficient reduces to l/e. Useful spatial diversity can be obtained with antenna separation distances of a few wavelengths, with a spacing of >_10,,being considered to provide a high degree of de-correlation, defined as a correlation coefficient

27 - l

(3-3)

ii

bps/Hz is the normalised capacity per Hertz of bandwidth, also termed

the spectral efficiency. It can be shown that there is an absolute lower bound on Eb/NO(irrespective of the complexity

of

the modulation

and coding

below schemes employed) which

communication is not possible:

Limit n-0o

2'

dB ln(2) 0.693 or = = -1.6

(3-4)

In practice, it has proved very difficult to achieve anything close to the Shannon bound discovery Only the of with complexity. receiver achievable with even remotely iterative coding, so-called Turbo-codes, [Berrou, 531 in the mid-1990s have practical bound. Shannon to the to systems started approach close This section provides a summary of the basic modulation techniques employed to The of adaptive data application channel. radio a over communications convey In is introduced. Doppler spread equalisation to mitigate the effects of multipath and to techniques reduce is coding control to the error of use addition, consideration given bit error rates (BERs) in a received transmission to a tolerable level.

39

Chapter 3 HF Data Communications

3.2.1

Amplitude

40

Shift Keying (ASK)

In amplitude shift keying the amplitude of the carrier is modulated by the data stream to be transmitted.

In the simplest case this results in binary on-off keving (OOK) as

depicted in Figure 3-2.

A successful demodulation technique employs incoherent

detection (envelope detection, Figure 3-3) and a Law assessor[Law. 54] to establish an in based for decision the threshold energy present a number of on each symbol adaptive interference. fading impact helps This the to and of signal mitigate preceding symbols. In a more general implementation of ASK, a number of binary symbols may be mapped to a multi-level

amplitude modulation.

This will increase the spectral efficiency

is ASK In decreasing not the practice (bps/Hz) at the expense of signal's robustness. its because in poor power efficiency. of generally used, part

0101101

aq

AIR,

UVVU

II JVududuJq

Figure 3-2 On-Off Keying (OOK)

40

Chapter 3 HF Data Communications

`-º

Received Signal

41

-

pop x2Decision Variable

r(t) No -º

-º

z(t)

X2

goo Squaring function (envelope detector)

Filter matched to pulse shape Reference cos(2icft)

Figure 3-3 Non-coherent , latched Filter Structure

3.2.2

Frequency Shift Keying (FSK)

In binary frequency shift keying (FSK), as depicted in Figure 3-4, the binary data sequence to be sent is used to modulate the carrier frequency.

This was traditionally

implemented either phase continuously using a single voltage controlled oscillator (VCO) or, non-phase continuously, by switching between the output of two oscillators at the two tone frequencies, known as frequency exchange keying (FEK). FSK may be received using a number of techniques. The optimum performance in an AWGN channel is obtained utilising orthogonally spaced tones [Burr, 51] and coherent detection using two matched filters.

Two signals, x, (t) and x, (t) are orthogonal if ,

their inner product is zero: XI(t)xz (t)dt =0

(3-5)

For use on HF channels, wide deviation FSK, with non-coherent detection and a Law assessor is often utilised.

Even though non-coherent detection suffers a penalty of

increased has in AWGN, detection technique this dB compared with coherent -1 interference To in fading and mitigate against narrowband channel. robustness a frequency selective fading the signal is detected as two independent OOK signals and then a decision is made given the additional knowledge that the two are complementary.

41

Chapter 3 HF Data Communications

4

I 0101101

fl

f2

f1

f1

f2

II Vu

V

ýý

f1

V

Iý

f2

dU

lý üIV V

Figure 3-4 Frequency Shift Keying (FSK)

3.2.3

Multi-level

Frequency Shift Keying (MFSK)

A generalised multi-level extension of binary FSK (2-FSK) is MFSK in which multiple tones are utilised.

The source data is encoded into a stream of multi-bit symbols and

frequency. If `A1' tones are used, to tone a particular each resulting symbol mapped then each tone may carry 'k' bits of information, where: k =109 (M) or M=_1k z This is illustrated in Figure 3-5 for 8-ary MFSK.

(3-6)

In this case k=3 consecutive bits are

is into transmitted as one of eight possible tones. encoded a symbol which

42

Chapter 3 HF Data Communications

F F2 7fN. '001'

-, nn

a ý, 1" , 11 17J 1ý1

A1rA1A1A1A

43

--

-

1ic

1ý 1`

F7 110

F5

----

A11A1AAl

Symbol Duration, T 001'

'110,

ýý

.100,

il ýIýý IIIýIIIýýýý Iý

Iq ^I Il I"I

II

ICI

ýI+

V

ýu

U

liil

u

Iý

If

t figure 3-S M=N, M-ary Multi-Frequency

Shift Keying (AIFSK)

MFSK receivers for use on fading radio channels generally employ a bank of noncoherent matched filters (as per Figure 3-3). A non-coherent matched filter, optimised to the symbol length `T', has a frequency response which is a sinc function with nulls every l/T Hz. It can be shown that performance is maximised if the tones are spaced orthogonally if the tone spacing, df, is [Sklar, 55]: Af =-,

where n=1,?, 3,

...

Normally, unless very high levels of Doppler spread are anticipated, 'n'=l

(3-7) is chosen to

minimise the occupied spectrum (i. e. maximise spectrum efficiency): 1 Ofmin=T

(3-8)

MFSK has proven to be particularly suitable for the very robust transmission of data for disturbed including Clark, 57] 56; data [Ralphs, low HF rates over channels at For low SNRs. Doppler (high a commensurate and at spread) channels multipath and data rate a higher order MFSK waveform (M=16 or M=32 is typical) can significantly the in MFSK durations to long against 2-FSK. The protect able are symbol outperform by (as inter-symbol interference multipath). caused effect of

43

Chapter 3 HF Data Communications

3.2.4

44

Phase Shift Keying (PSK)

In phase shift keying (PSK) information is transmitted using a constant amplitude carrier, modulating its phase according to the symbols to be transmitted. In binary PSK (BPSK) the phase change is 180° (Figure 3-6b). detection offers the maximum performance.

In an AWGN channel coherent

In practical systems. particularly those

disturbed over working channels it is difficult to establish the required phase reference. In this case differential

PSK (DPSK)

can be utilised.

For DBPSK the phase is un-

altered if the next symbol is a '0' and reversed if the next symbol is a' 1'. In this case the phase reference for each symbol is the previous one (Figure 3-6c). easier to implement, than coherent BPSK. HF waveforms

Whilst this is

errors tend to come in pairs, and its performance is -3 dB worse An alternative approach, which is commonly applied in modern

which seek to maximise performance, is to take additional measures to

equalise the effects of the channel making coherent detection possible. The equalisation is discussed in detail in 3.2.8. section greater of received signals

1011

I

010010

(a) Digital Data to be transmitted

(b) Binary Binary Phase Shift Keying (DBPSK)

1I IA i' (c) Differential Binary Phase Shift Keying (DBPSK)

Figure 3-6 Binary Phase Shift Keying (BPSK)

44

Chapter 3 HF Data Communications

45

Imag

010 011

001

100

000 Real

101

-

111

110

Figure 3-7 8-PSK Constellation Multi-phase PSK (M-PSK) encodes a number of data bits to be transmitted into a symbol which maps to a particular phase. For example, in 8-PSK three information bits map to a single symbol (Figure 3-7). This increases the throughput (spectral efficiency) at some cost in the required Eb/NOto maintain a given BER (as the distance between constellation points is reduced). 4-PSK or quadrature PSK (QPSK) is a special case in that the distance between symbols in the constellation is the same as for BPSK and so, theoretically the performance is the same. A common method of implementing a QPSK modulator is illustrated in Figure 3-8.

Practical M-PSK waveforms for use at HF

require the use of equalisation to provide acceptable performance.

x, Binary data to be transmitted

100

S(t)

Serial to Parallel Convertor

-º

Q

90°

Reference cos(2 rzf,t)

Figure 3-8 QPSK/QAMModulator

45

Chapter 3 HF Data Communications

3.2.5

46

Higher Order Modulation Modulation

Quadrature Amplitude -

(QAM)

Where higher spectral efficiencies are required, combined phase and amplitude modulation can be employed, effectively still modulating a single carrier. In quadrature amplitude modulation (QAM) a series of binary digits to be transmitted is mapped to a symbol which represents a phase/amplitude combination. 16-QAM (Figure 3-9) has a theoretical

4 bps/Hz

spectral efficiency

although this

is reduced in practical

implementations by the inclusion of error correction codes and synchronisation/training data (discussed later).

Imag

1101

1100

1001. ... .....

' 1000

4ý0001 ........

0101

+

0000

0100

Real 1110

f ........

1111

1010

0010

1011 1

0011

4 ...

.

0110

0111

Figure 3-9 16-QAM Constellation The application of modulation schemes, such as QAM to HF is still relatively new and implementations to mitigate the effects of the the requires use of sophisticated receiver HF channel (including adaptive equalisation, effective error control coding). Even so, the application, particularly of higher order QAM schemes, is limited to relatively benign channels with good SNR.

3.2.6

Higher Order Modulation - Multi-Carrier

Techniques

An alternative approach to using single tone waveforms (such as QAM) to provide high spectral efficiencies is to modulate a series of carriers.

Most modern multi-carrier

(OFDM). division frequency implemented multiplexing waveforms are as orthogonal

46

Chapter 3 HF Data Communications

4

The modulation applied to each carrier varies from simple schemes such as DBPSK (e.g. Kiniplex, [Mosier, 58]) to using multi-level schemes like QAM (e. Digital Radio g. Modiale MF/HF broadcasting, [Stott, 50]). A particular advantage OFDM is that it of may be efficiently implemented using Fast Fourier Transform (FFT) filter banks as the core components of both modulator and demodulator. The implementation of OFDM provides an inherent tolerance to multipath as the symbol rate on each carrier is very low. However, additional measures do have to be taken. In particular a guard period (in effect a lengthening of the symbol period) has to be applied to reduce the intersymbol caused by multipath.

In order to allow coherent demodulation known pilot

tones and pilot symbols are inserted to allow the impact of the channel to be identified and mitigated.

Other techniques such as interleaving are utilised to mitigate against

frequency selective fading and narrowband interference.

These issues are well

described in [Burr, 511. The construction

of the transmitted

in the transmitted carriers results ratio (PAR).

signals from a series of independently signal having a significant

modulated

peak-to-mean amplitude

This is because at some instances the outputs of the individual modulators

will add coherently). by SNR, transmitter

Where, as is often the case at HF, system performance is limited power is a key factor.

In a practical system it is not untypical for

the transmitter to have to be backed off by >l0 dB to avoid saturation at peak powers. This issue is not always reflected in comparative waveform performance comparisons.

A number of techniques have been developed to reduce the impact of the PAR including 591 takes [Shepherd, which clipping adaptive and mapping symbol semi-orthogonal 60]. [Enright. fact that the occur the severest excursions rarely advantage of

3.2.7

Relative Performance of Modulation Schemes

be discussed can The relative performance of many of the modulation schemes 3-10 (based in Figure on BER in AWGN presented the plots performance compared A 61]). [Stremler, 14] additional of [Proakis, from number [Burr, 51], and analysis observations are appropriate: 0

HF fading in Performance AWGN The performance plotted is for an channel. least, QAM PSK require at in and be the of case poorer and, channels will equalisation.

47

Chapter 3 HF Data Communications

48

Higher order MFSK modulation provides improved robustness (i. e. lower Eb/No

"

operation) at the expense of reduced spectral efficiency. Conversely higher order PSK and QAM provide increased spectral efficiency (bps/Hz) at the expense of robustness. As would be expected from an inspection of the M-PSK

"

and M-QAM

constellations, at M>8 the robustness of QAM starts to exceed that of PSK.

10'

.

`---------- -- - --------------------------------

------. -. e. -`----------------'

- ---

ý`------------------------

---------------

-

-------------

102

-

2-FSK (non-coherent)

----

ý__

&FSK 16-FSK

---------------------

_...

- -'----

---------4_...;

---

--'-

--'-----

--

-------------------------------

-----------------

CT w m

w m J

----------

-

....

---------------

----------

10°

---

B-PSK 16-PSK 4-QAM 16-QAM 64-QAM

------- -- - ---------------------------------------------------------------------------------------------- -------------------------------------------

-------------

---------------------

04

._... DBPSK

--

-----------------------'-----------

_""______________

BPSWOPSK

------------------

-

----

- ---------

---------------------------

ý5

ý

10ý

---- - -- -- -----------

--------------------------------------------------

-------------

-------------

..................

-----------------------------

--------

------------

__

------1_

ý. ___i_ .. r _ ....... _____________ _

_.......

_.... -_... _____ -------

------------

-------------

--------------

------

--------------

---

-

------------

ii_i 0

5

15

10

20

25

05

Equalisation

20

25

Eb/No (dB)

Figure 3-10 BER Performance of Various Modulation

Adaptive

15

10

Eb/No (dB)

3.2.8

---------{.

------------------__---------t-----

------

---------------------------`--------............... -`---------------------------------------------',

--

---ý----'-----.....

------------------------------------------------'-----'`°--°.

10°

10'

to Mitigate

Schemes in an A WGN Channel

Multipath

and Doppler

Spread The relative delay between signals arriving at a receiver due to multipath can cause Skywave HF in the ISI, multipath may span where as such channels severe particularly many symbols.

For example, the symbol duration in a standardised 3 kHz HF

waveform such as MIL-STD-188-11OA

[US DOD, 62] is -417 is (2400 symbols/s)

frequency Further, be in 5 excess of ms. compared with the multipath which may dispersion (Doppler spread) and distortion introduced by the transmitter/receiver will in the received waveform which can adversely variations amplitude and cause phase by be This demodulation. employing equalisation. corrected may effect 48

3 HF Data Communications

4()

Input Signal xw

C. z

Cu

C,

C2

I

Tap Weight Adaptation EqualisedOutput

V1111I F.

IN.

Figure 3-11 Linear Transverse Equaliser In its simplest form a linear 'zero forcing'

equaliser can be used to filter the received

frequency H(_): inverse the the response, of channel z-transform signal with

(3-9)

C(-) _1 H(_)

Whilst it is possible to set the coefficients during the passage of unknown data ('Blind known include it is the to 'training' to coefficients allow symbols usual equalisation') in data is is be Where the training the channel not stationary repeated at a rate to set. excess of the Nyquist rate of the channel. A zero forcing equaliser can be formulated using an infinite length linear filter: qk

(k = 0)

X1 -I

CJ1

k-J

0

(k

#

(3-10)

0)

A finite impulse response (FIR) linear filter (as illustrated in Figure 3-11) approximates this: Y-C1"k-j

Ik -

j=-K

In 14]. [Proakis, is particular However, such an equaliser's performance sub-optimal fading in is when the presence of the performance of such a linear equaliser very poor Indeed, to in signal. filter the a real of absence the can significantly amplify noise is the impulse the of filter conjugate response whose ensure stability, a noise-whitening equaliser filter is sometimes employed. A better approach than the zero-forcing criteria is to minimise the mean square error (MSE), between the symbol transmitted, Ik and that detected, Ik . ,

49

Chapter 3 HF Data Communications

50

A more advanced equaliser is the non-linear decision feedback equaliser (DFE), illustrated in Figure 3-12.

The feed forward section is the same as for the linear

described above. equaliser

However, a feedback filter is added whose input is

detected symbols. previously

Functionally its aim is to remove from the present

by ISI detected the that part of caused previously symbols. estimate Input from Coherent Detector

Output Data Symbols

Feedforward Transversal

+_

Filter

} Ik

Symbol-by-symbol Detector

{I

I k

Feedback

Transversal Filter

44

Figure 3-12 Decision Feedback Equaliser (DFE)

In this case the equaliser output can be expressed as: Ik

CJVk-l l=-K1

+ýCIIk-J

(3-1?)

)=I

where Ik

is an estimate of the kth symbol as before,

Ik

is the kth detected symbol

cj

feedback K2 forward feed K1+l coefficients the and are

Given the requirement to jointly optimise coefficient sets Kj and K2 to achieve the MSE criteria: J(K,, K, ) = EIk -IkI2

(3-13)

14]: by [Proakis, filter forward feed are given the coefficients of the 0

(3-14)

where

50

Chapter 3 HF Data Communications

-;

gflj

hnh.

,»-o

51

+N0 ý, +,

i, j

(3-15)

1,0

and ho

...

hL_,

are the taps of the channel (length L);

No

is the channel noise density;

9ii

is the Kronecker delta matrix: Sj =1

for ij,

5ij =0

otherwise. The coefficients of the feedback filter can be expressed in terms of the coefficients of the feed forward filter:

Ck = J=-Ki

cI hk_ ,k

=1,2, ... ,

(3-16)

Kz

The feedback filter is able to completely cancel the ISI from previous symbols providing these have been correctly detected and that the filter length, K,, exceeds the length (total multipath dispersion). channel Since the multipath fading channels, such as Skywave HF, are not stationary practical modem implementations must be able to adapt continually and sufficiently quickly to track the changing channel.

Therefore the coefficients must be calculated using

computationally affordable, fast converging algorithms such as the Kalman Recursive Least Squares (RLS) algorithm [Hsu, 63]. There are a number of alternative adaptive equaliser structures including the optimum (but computationally

expensive) Viterbi

maximum likelihood sequence estimation

(MLSE) algorithm [Bartlett, 64], [Falconer, 65] and more efficient block decision feedback equalisers (BDFE), discussed in [Jorgenson, 69].

In both cases these

joint detection block the treat take process as a algorithms of symbols and a optimisation problem (as opposed to the DFE symbol-by-symbol approach).

Further

based FEC by improvements be the on equaliser adapting performance realised can corrected symbols.

3.2.9

Forward Error Correction (FEC) Coding

The majority of practical modems make use of FEC to provide acceptable bit error bits Essentially calculated parity these all employ the transmission of additional rates. 51

Chapter 3 HF Data Communications from the information bits to be transmitted.

52 This reduces the useful throughput but

increases robustness. This section aims to provide a brief introduction to the subject indicate the general performance that can be obtained. Error control coding is a and large subject and for detailed information in this evolving field the reader is urged to consult one of the large numbers of texts on the subject (e.g. [Lin, 70]). A code word of `n' bits is formed from 'k" information bits and n-k parity bits. The code rate, i. e. the proportion of information bits in a code word and therefore its is k/n. The bit in number of efficiency, errors a code word that can be corrected is essentially determined by the `distance' between possible code words and is clearly The be broadly FEC the to code principal rate. classes of can categorised as related `block' codes (e.g. Reed Solomon codes, Bose-Chaudhuri-Hocquenghem codes) and `convolutional' codes. Different codes have different properties. For example ReedSolomon codes have an ability to correct small bursts of errors whereas convolutional independent burst have have to that correct errors a much greater capability poor codes (random) errors.

A FEC code may either be utilised alone or, where additional

harnessing be (thus is the potentially may concatenated required, codes robustness benefits of different codes). The decoding process may be either `hard decision', where decoding is based on decisions', in demodulator, decisions 'soft the where the or, on made symbol demodulator provides the decoder with a numeric confidence for all its decisions. This likelihood determine be information to to a maximum good use put can additional decoding. This is now commonly done by forming a time series graph of all possible Viterbi block in the then (a trellis) and using a code received symbol combinations likely determine the to sequence of originally transmitted symbols. algorithm most Recently, a new class of codes, the so-called `Turbo codes' have been discovered [Berrou, 53]. These are constructed as concatenated component codes interspersed with interleavers. By utilising iterative, soft decoding employing a 'soft input-soft output' (SISO) decoder performance approaching to within a fraction of dB of the Shannon bound is possible with sufficient iterations. The performance curves for a number of different FEC schemes, operating in AWGN, 51]. [Burr, 14], in [Proakis, 3-13 Figure are presented

52

Chapter 3 HF Data Communications

1 10-

---------------

53

10-

BPSK uncoded Rate 1ß, k=7, Vderbi, Soft Decision -------- LRate 112, k=7, Vderbi, Sott Decision --_________ - Rate 113, k=7, VAerbi, Hard Decision __________-. """" Rate 112, k=7, Vderbi, Hard Decision ------------Reed Solomon (15,9) Reed Solomon (31,23) -----; - ---

-

I

_ yý-', )ý i -; - - »- - --i"

102 tc -------------------

----------------

-----

---

10-3

-----

--------------------------I -----

-------

------

a)

___

-------

--------

r ______

-ý------_'

- r-l-r--_ý-r 1-

----------------

CL, r_1_i___ 1. 1I

tII

____

----_-_-----_

----

----

J_____lt

ý_

--S

I.

-----------------

_--------

_____

--------------

__ -t-

-------------------_________--_--_--_

------'--------------------

-i----`,

F-----

_

-_

-----------------

--------------

-------

----------------

A------

--ýI

12

------

-----------------

_________________

---------------_____--_

10

---

__

I;

2468

-_

1

10.5

r--_

10-6L 0

--r

_____

r

I.

------------------

---+-----

____

----------

-------------------

----

------

104

I'

10 5

------

--------------------------

_

___---"-__-__

________ _______

ý

m ------__ _ __

--

-----------

W

-_-t: ___

-------

-----------------

------

__

_____

______ __ __

--

W

_t ____J-"

_

_

-

-- - -- ---------------------------------

103

- -----------_ ____

code, 1 deratlon code, 2 derations code, 3 derations code, 6 derations code, 18 derations

-- --------------------------------------

---

ö

104

---

______ _____

------------

ö

m

Turbo Turbo Turbo Turbo Turbo

-

___

------------------------------

-_t_J_L i--, ---

-" ---".....

j1

102 }t

BPSK uncoded Rate 12, Berrou Rate 112, Berrou Rate 12, Berrou Rate 12, Berrou Rate 1/2, Berrou

14

1I

-t}

-

---

--- -------------------

T --------------

02468

-----

10

-------

12

14

Eb/No (dB)

Eb/No (dB)

(a) Reed Solomon and Convolutional Codes Performance

(b) Turbo Code Performance Versus Number of Iterations

Figure 3-13 Performance of Various Forward Error Correction

Codes

3.2.10 Interleaving Practical FEC codes, whilst powerful for correcting information streams containing bursts large have independent (random) to of ability correct a poor errors, statistically is fading Interleaving a process of channel. errors such as might occur on a slowly block (illustrated be typically transmitted to the the a using scrambling order of symbols in Figure 3-14) or convolutional interleaver [Proakis, 14]. The interleaver 'depth' is be for HF fade duration; it this the several to to may expected chosen allow overcome becomes burst de-interleaving Following a number the of errors receiver, a seconds. at The hopefully FEC act of multiplexing a coded the correct. can of random errors which data steam onto a number of carriers for a parallel tone waveform inherently provides interference. fading from frequency narrowband and some protection selective However, an interleaver will still generally be used to overcome flat fading.

53

Chapter 3 HF Data Communications

54

12345

r--

1

1ýi

234567B9

W

nr2

n+3

m4

Tn

m5 --

2,1+1

31.2

2n13

21+4

211.5

3n

M ROWS il

Im-tlrwl

(m-1)nt2 Im-tlrw3

I

(m-1)n. 4 (m. 1)n+5

mit

n Columns

-

ý., ý.,

ý ý,

ý_ », 2

Mz

' -----

--

Figure 3-]-!

Operation of an (mxn) Block Interleaver

3.2.11 Synchronisation and Tracking Practical modem receive implementations must overcome a number of additional be Specifically, they to able to: need challenges. 0

Detect and acquire a transmission (initial synchronisation);

0

Determine and correct for Doppler shift on the transmission; and

"

Continue to track the signal being received correcting for Doppler shift and modem clock drift.

3.3

MIL-STD-188-11OA

Communications Data HF Tone Serial -a

Waveform This section concludes the discussion of modem techniques by introducing a modern HF waveform employing many of features discussed. MIL-STD-188-110A (MS-110A, [US DOD, 62]) is a serial tone waveform providing throughputs of 300 to 2400 bps based bps 75 highly is on There bps mode FEC 4800 robust a also with un-coded. and in-band spread spectrum.

MS-110A is based on the use of convolutional coding,

interleaving and adaptively equalised 8-PSK modulation format. The structure of the 3-15. in is illustrated Figure waveform

54

Chapter 3 HF Data Communications

Preamble

Data

1440 or 11520 symbols

20 symbols

(a) MIL-STD-1

88-11

OA, 300

to 1200

OA, 2400

Training 20 symbols

Data 32 symbols

and

Data

Training 20 symbols

20 symbols

_

EOM Sequence

---

bps

Preamble 1440 or 11520 symbols (b) MIL-STD-188-11

5j

4800

Training 16 symbols

Data 32 symbols

Training 16 symbols

EOM Sequence

-

bps

Figure 3-15 Structure of MIL-STD-188-1I

OA waveform

At all data rates the modem transmits at 2400 symbol/s in a bandwidth of 2.7 kHz. At data rates up to 1200 bps the training phase is 8.3 ms long and repeated at a rate of 60 probes/s. These numbers indicate the modem's ability to tolerate multipath and Doppler spread respectively [Brakemeier, The waveform

71 ].

FEC depends on data rate (see Table 3-1).

It utilises a matrix block

interleaver with three depths: zero (bypassed), short (0.6 s) and long (4.8 s). supports auto-baud; the ability

for the receiver to automatically

and interleaver settings of a transmission from information Data Rate

Bits Per 8-PSK Symbol

Effective Code Rate

identify

the data rate

in the pre-amble. FEC

Convolutional Code, Rate ' 2, k=7, repeated 4 times

150

1

300

I

'/4

Convolutional Code, Ratek=7, repeated 2 times

600

1

'/2

Convolutional Code, Rate '/2,k=7

1200

2

'/2

Convolutional Code, Rate '/2,k=7

2400

3

4800

3

Convolutional Code. Rate '/2, k=7 Un-coded

1

Table 3-1 Modulation/Coding Figure 3-16 shows a constant

It also

BER plot

[Arthur,

Parameters, for MIL-STD-188-11OA 72] of a measured commercial

Doppler spread and of multipath modem's performance envelope over a range conditions for BER=10-3. As can be seen the modem essentially operates reliably with Beyond dB SNR Hz Doppler either of ms. of to multipath and spreads of -6 -6 -10 dB indicate SNR_40 Regions limits deteriorates. these of the performance quickly SNR irrespective be the BER and available of where the met requirement could not indicates the performance limits of the waveform/equaliser.

55

Chapter 3 HF Dato Communications

6

Figure 3-16 Measured Performance of 1200 bps MIL-STD- 188-11 OA Modem (BER1O-3)

3.4

High Data Rate HF Communications

Figure 3-17 summarises the current (2001) state of progress that the international has achieved

research community increased throughputs

HF waveforms

(and includes work presented in this thesis).

users relied almost wholly communications.

in developing

Only

HF

and modems with For many years

baud FSK 75-300 waveforms on un-coded

for HF data

in the last ten years have more capable modems become

available supporting

Skywave data rates up to 2.4 kbps using sophisticated equalised

waveforms and FEC.

NATO

has led have US recently completed work which and the

to the standardisation of modems with data rates of up to 9.6 kbps [NATO, the aforementioned

waveforms

73]. All of

bandwidth kHz 3 in the standard operate

ITU

HF

allocations.

56

Chapter 3 HF Data Communications

Research to 64 kbps

iß

co Prototype 9.6 - 19.2 kbps

Now

MIL-STD- 1108

3.2 - 9.6 kbps STANAG-4539

Military Fielded 75 - 2400 bps

2000

1960s - 1980s

Narrowband 3 kHz

Extended Bandwidth 6 kHz

Bandwidth Wideband 12-100+ kHz Multi-channelling

Figure 3-17 Current Progress in Development of'HFData

Communications

(to 2001)

There are a number of potential technical directions that can be explored to increase HF data rates. These include: "

Waveforms already

higher with efficiency

highly

developed

sophisticated

waveform

latest The HF generation of waveforms are -

and make extensive

processing

use of new and recently

(e. g. DFE and DBFE

equalisers)

and FEC

techniques (e.g. Turbo codes). Whilst there are likely to be further improvements in this area large throughput gains are unlikely to be realised in the near future. 0

Diversity

Techniques

frequency or polarisation Use of spatial, -

diversity

is a

for improve known the therefore potential to technique robustness and well increased throughput.

Recent work has shown that, with the use of space-time

by is [Burr, 74], the number of throughput techniques constrained coding in the propagation medium. that transmission present are paths uncorrelated Diversity implementations require the use of multiple antenna apertures as well as multiple transmitter and/or receiver channels. 0

Wideband Waveforms - It may be expected that throughput can be increased in 52] the [Shannon. that bandwidth given proportion to the contiguous occupied HF in the wideband waveforms employed are capable of performing adequately channel.

Therefore an improved understanding of the characteristics of the

57

Chapter 3 HF Data Communications

wideband channel is required.

58 Further, equipment capable of wideband

transmission and reception is required.

For practical use suitable spectrum

allocations must be obtained. Previous work has investigated the implementation of Direct Sequence Spread Spectrum (DSSS) techniques [Dixon, 75] to HF by workers such as Milsom [Milsom, 24], Van der Perre [Van der Perre, 76] and Perry [Perry, 77]. 0

Multi-Channelling

(use of non-contiguous narrowband channels) -A related option is to increase throughput by splitting the data stream to be transmitted into a number of parallel channels and then transmitting these over a number of (potentially) non-contiguous channels [Jorgenson, 78]. This aims to provide increased bandwidth through the use of non-contiguous narrowband channels and has the advantage that it is more likely to be practical given the regulatory need for spectrum allocations.

This technique, whilst it requires the use of transmission

and reception equipment with multiple channels, may potentially utilise existing narrowband modems.

However, there is potential for increased efficiency by

adaptively exploiting differences and diversity between the narrowband channels in use. It can be concluded from the above that the principal opportunities for increasing HF data communications rates require the use of wider contiguous bandwidths or multichannel approaches.

3.5

Summary

This Chapter has introduced some of the key technologies used to implement data communications

modem

waveforms.

Waveforms

for

narrowband

HF

data

high to have been discussed techniques applicable wideband, communications and some throughput communications introduced.

In the next Chapter a new, high throughput

(16 kbps), extended bandwidth (6 kHz) modem is introduced. This modem employs interleaving, including: here introduced robust error correction, many of the techniques data high to rate of a provide adaptive equalisation and order modulation schemes 16 kbps over HF channels. In Chapter 7a new wideband, multi-channel digital transceiver (software radio) capable of supporting wideband and multi-channel operation is presented. One of its intended uses is an experimental software modem platform to develop and experiment with new data transmission techniques. 58

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

59

Chapter 4.

A High Data Rate Modem for Extended Bandwidth Channels

Considerable effort has been expended in recent years to increase data rates over narrow band HF channels as the demand for improved throughput over HF to support a variety of user applications increases. Until recently, state-of-the-art modems, incorporating waveforms such as MIL-STD-188-11OA

[US DOD, 62] and STANAG 4285 [NATO.

79] have had realistic limits of -2400 bits per second (bps) over HF sky-wave circuits. With improvements in digital signal processing and modem technology (especially the development of high performance equalisers and improved error control coding techniques) high data rate waveforms, such as those included in the forthcoming MILSTD-188-11OB [US DOD, 2] (formerly specified in Annex G of draft STANAG 5066 [NATO, 81]), are becoming practical.

Potential applications include high throughput

HF data networking and range extension for line-of-sight V/UHF radio links. This chapter describes a novel high data rate (HDR), 16 kbps prototype modem operating in an extended bandwidth of 6 kHz'.

The data rate was selected for

compatibility with extant line-of-sight radio communications systems and the operating bandwidth chosen to allow conventional independent side-band (ISB) HF radios to be utilised. Results from HF simulator measurements and on-air testing of the modem are presented.

The performance limitations of such high data rate modems will be

discussed leading to an identification of the range of applications for which they can be bandwidth HDR be Finally to expected used reliably. a number of alternative extended high introduced the that their rate of waveforms are and performance compared with 3 kHz waveforms now being standardised.

The initial concept for these extended bandwidth modem waveforms was the author's. The for implemented based proposed waveforms are on a new generation of narrowband waveforms Annex G of draft STANAG 5066. The implementation of both these `Annex G' waveforms and the The CRC. Jorgenson Mark bandwidth by Bob Johnson of new extended and variants was undertaken simulator tests, on-air trials work and analysis are all the author's own. This collaboration allowed the work presented in this chapter to be completed in a period of just some 12 weeks.

59

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

4.1

60

Waveform and Modem Processing Description

Two 16 kbps waveform variants were developed and implemented on Pentium PCs running the QNXTM [QNX, 80] real time operating system. An inexpensive sound card was used to provide the audio interface to the radio. The 16 kbps modem waveforms used in this study are developments of the high data rate waveforms specified in Annex G of STANAG 5066. The new waveforms have been designed to exploit the 6 kHz of bandwidth available in some HF channel allocations. Like the Annex G waveforms, they are serial tone waveforms which have been designed to be as efficient as possible to achieve high throughputs One of the 16 kbps variants uses the 6 kHz channel bandwidth as a contiguous single side-band (SSB) while the other has been designed for use with ISB radios and operates in an independent upper/lower side-band configuration. The only significant difference between the two implementations is in the final modulation stage. The 6 kHz SSB implementation uses a single 3300 Hz sub-carrier modulated at 4800 symbols/s while the ISB implementation employs a modulation rate of 2400 symbols/s which is applied to two ISB sub-carriers, each centred at 1800 Hz within their respective audio sidebands. The ISB implementation takes the data stream at the output of the interleaver (or bits 16-QAM is interleaver to if symbol passes employed) and alternately no codec generators for each of the sub-carriers. The frame structure used by both the SSB and ISB waveforms is shown Figure 4-1. An initial 240 symbol preamble is followed by 48 blocks of alternating data and known by followed is data 16-QAM 282 block, a data Each symbols, of consisting symbols. 204 blocks, 48 After data. known symbol 31 a symbols of mini-probe consisting of (`sync-on-data'), late facilitate is initial to acquisition preamble reinserted subset of the Doppler shift removal and sync adjustment.

60

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

61

F'::1 Initial SynchronisationPreamble 240 symbols -

I

Data block - 282 symbols

Mini-probe 31 symbols of repeated 16 symbol Frank-Heimiller polyphase code

4

Regularly re-insertedpreamble 204 symbols -

Figure 4-1 16 khps Modem U'avetorm Structure The data blocks, using the 16-QAM constellation long to provide the efficiency

points shown in Figure 4-2, are very

required for high data rates. The 16-QAM constellation

used has been designed to provide a good peak-to-average ratio while retaining the good Gray-coding

square 16-QAM constellation.

properties of the traditional

The probe segments, which follow

each of the data blocks, consist of known symbol

sequences ([Frank, 66], [Heirniller,

67], [Frank, 68]) chosen for their good correlation

properties,

and are long enough that they can be used to derive channel estimates

independent of the user data. The forward error correction coding employed, termed a hyper-code

(HC),

is a proprietary,

high

rate,

iterated

block-code

which

offers

performance comparable to that obtained with turbo-codes. An alternative version using the rate '/2 constraint

length

7 convolutional

codec employed

in STANAG

4285,

punctured to rate 15/16, has also been investigated, but offers poorer performance than that obtained with the proprietary

code.

selectable delays of approximately Alternatively,

a no interleaving

Us

In both cases, a convolutional (short) or 6.5 s (long)

interleaver with

can be employed.

is also available. option

As a consequence of the common waveform structure used in both implementations, the ISB variant offers roughly twice the delay spread handling capability of the SSB variant This Doppler occurs as a half the spread resistance. while providing approximately With the this waveform, baud 2400 baud 4800 rates. result of on-air signalling versus 61

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

62

ISB implementation offers a delay spread handling capability of the order of slightly more than 5 ms, which is comparable with most current serial tone implementations. Alternate SSB waveforms could be designed which would increase the delay spread handling capability of the SSB variant while reducing its Doppler spread tolerance. The modem incorporates an advanced adaptive equaliser developed by Jorgenson [Jorgenson, 69] to compensate for multipath and fading imparted by the HF channel.

1 s

0.5

s

s

s s

0

s a

-as

R

s

-0.5

'

s

s

1

s «

1

-0.5

0

0.5

1

Figure 4-2 Modified 16-QAM Constellation Employed in the 16 kbps Modem

4.2

16 kbps Modem Performance

4.2.1

Performance Characterisation

Simulator HF using an

The performance of the 16 kbps modem was characterised using a validated Watterson 82] [Willink 13] [ITU, F. 1487 in described ITU-R HF et at, and type simulator as data known in baseband. A a operating at special modem test mode was used which BER in the calculated at resulting sequence was generated the transmitting modem and 6 kHz bandwidth input had of The the receiving modem. a maximum simulator used For SSB for it reasons of practicality testing the modem variant. which made suitable the ISB modem was tested using two independent simulators with the same path to was fading sideband Gaussian each The applied noise and parameters set on each. (in by the the Software in author but written was time. therefore the same uncorrelated C language) to control both the simulator and the modem in order to automatically 62

Chapter 4A Hi.gh Data Rate Modem för Extended Bandti, 'idth Channels

63

characterise the modem BER over a wide range of SNR, Doppler spread and multipath conditions using the test set up shown in Figure 4-3. USB Audio Watterson HF Channel Simulator 1

USB Audio

f

16 kbps ISB Modem (Transmitter)

16 kbps ISB Modem (Receiver) LSB Audio Watterson HF Channel Simulator 2

LSB Audio

Modem Control (PC ISA Bus)

RS232 Control

Results Data File

MS-DOS Control PC

Figure 4-3 16 kbps Modem Laboratory

C haracierisalion

All the results presented are given in terms of signal to noise ratio (SNR) in a6 kHz bandwidth

unless explicitly

stated otherwise.

The Rayleigh fading imposed by the

doublein is (26) has Gaussian the terms of specified spectral profile and simulator a sided fading bandwidth.

The modem's bit error rate (BER) performance was measured under a number of HC (zero, code configurations: modem of a number standard channel conditions with interleaving) long (zero long interleaving), and unand short and convolutional code in ISB the 4-4 a interleaving). Figure (zero modem the performance of shows coded fading flat (single in 4-5, Figure mode that Gaussian a non-fading noise channel, and Good (two CCIR in The channel 1 Hz Doppler a performance with spread) channel. imposed fading Hz 0.1 0.5 difference ms and of equal power modes with a relative time time a relative (two in with CCIR Poor modes power equal on each) and channel a difference of 2 ms and 1.0 Hz fading imposed on each) are depicted in Figure 4-6 and Figure 4-7 respectively. The multipath in a CCIR Poor channel is beyond the capability bit irreducible SSB in the error rate. of variant and results an

63

Chapter 4A

High Data Rate Modem for Ex/L'/u/L'd Bandwidth Channels

64

1 HC Long } 0.1

-ý HC None -'F- Conv Long

io

w

HC Short

Conv None

0.01

Uncoded

m

0.001

0.0001 L 6

8

10

ý, _ý 12 14

1_ \

16 SNR (dB)

18

20

_-22

24

Figure 4-4 16 kbps ISB Modem Performance in a Gaussian Noise Channel.

1

0.1

is HC Long w+

0.01 HC Short -ý- HC None -

0.001

Conv Long

Conv None Uncoded 0.0001 It

Co

N

Cfl

NN

00

MM

N

SNR (dB)

Figure 4-5 16 kbps ISB Modem Performance in a Flat Fading Gaussian Norse c manner

64

Chapter 4A High Data Rute 1/(0, y fin- Extended Bandwidth Channels -,

65

1 E

HC Long HC Short 0 1 .

ö E w m

-

HC None

-

Conv Long

--

Conv None Uncoded

0.01

0.001

ý0.0001 (p

_ Op-

ON rC

VO r-

e-

00 r

C)

CN

qNNNN

00 N

Cý M

C,4 C)

q CO

(0 ()

C) 'IT

co Cl)

SNR (dB)

t figure 4-6 16 kbps ISE Modem Performance in CCIR Good Channel 1 F--

HC Long HC Short HC None

0.1 L

Conv Long Conv None I

w :L-

Uncoded

0.01

0.001

0.0001

CO

00

O

04

'q

(O

0 04

NN 04

'q

(0 NN

00

0 CO

04 M

IT C)

(0 CO

CO M

Oý

SNR (dB)

Figure 4-7 16 kbps ISB Modem Performance in CCIR Poor Channel The operating envelope of the ISB and SSB modem configurations is illustrated in the The 4-9. Figure 4-8 in Figure BER [Arthur, 72] achieved and constant plots bps 4800 MIL-STD-188-11OA single be the performance can compared with that of STANAG kHz) in 3 the data its for high and tone waveform in Figure 4-10 (chosen rate 4285 2400 bps waveform in Figure 4-11 (chosen for its similarity in structure to that of SNR a noise with to 16 a four kbps give These the normalised are plots modem). bandwidth of 3 kHz for constant BER of 10-3. 65

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

.

66

CÄ

f;

50 45 40 35 m 30 25 20 15 co 10 5

> 5 4ý 3

ti 2 10

Multipath

(ms)

\

0.5

LO

AM ýý N

0p

Doppler

Spread

(Hz)

0ö

Figure 4-8 Constant BER Surface, /br 16 kbps ISB Modem (HC codec, long interleave).

Figure 4-9 Constant BER Surface for 16 kbps SSB Modem (HC codec, long interleave).

66

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

50 45 40 35 30 25 20 15 cf) 10 5

5 4 3 2 Multipath

10

(ms)

67

0.50

0°ö°

r figure 4-1 U (- onstant

tbIK

Doppler

)urjace for

Spread (Hz)

4NUU bps MIL-J I D- I n(N-I1 UA Alodem (uncoded)

50

45 40 35 m 30 25 ' 20 15 CO 10 5 0

5 4 3\ 2 Multipath

(ms) ýMv

\

0.5

1\ 0°

_.

N

U Doppler Spread (Hz)

Figure 4-11 Constant BER Surface,for 2400 bps STANAG 4285 Modem (convolutional long interleave). coder,

67

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

4.2.2

68

Rician Channel Performance

The HF simulator test results in Figure 4-12 (plotted for a constant BER of 10--',two same power modes, one non-fading, one with 1 Hz Doppler spread applied and 2 ms between the modes), show that the 16 kbps modem performance increasesmarkedl} for Rician type channels (i. e. where there is a non-fading propagation mode in addition to Gaussian fading modes). Such channels occur most frequently at HF when a surface wave component is present. The range achievable using surface wave communications is significantly influenced by the surface conductivity, and is generally at its largest for budget Link have been ITU paths. sea calculations made using surface uninterrupted loss [ITU, 6]. curves propagation wave

Even using pessimistic figures for local

interference, sea state, etc, the calculations indicate that near error free operation at km be in 400 to a maritime environment using should readily achievable up ranges 400 W. transmitter the order of and of antennas powers omni-directional conventional Previous generations of modems (particularly

designed for types) tone surface parallel

intolerant of sky-wave components and perform poorly when they wave applications are are present.

The 16 kbps experimental

degraded by is the existence of not modem

its operating region. within multiple modes

68

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

69

vý\

50 45 40

5

35 w 30 Y 25 20 15 C

4

2 Multipath

(ms)

10 5

1o 0.5

NM 0O

6

Doppler Spread (Hz)

Figure 4-12 Constant BER Surface for 16 kbps ISB Modem using HC Codes and Long Interleave under Rician Channel Conditions

4.3

On-air Trials

4.3.1

Experimental Configuration

Trials were conducted over a 170 km, predominantly East-West, path from DERA Cobbett Hill (Cove Radio) to Malvern in the UK.

At the transmit site a 10 kW ISB

transmitter (generally operated at -2 kW) was utilised with a wide-band fan dipole fed dipole At (Malvern) `droopy' tactical a antenna the antenna. a simple receive site Marconi H2550 digital receiver, operated with independent automatic gain control (AGC) for each sideband and configured to have a fast attack and medium (--0.5 s) decay time. Only the ISB configuration of the modem was tested on-air, principally because it was the configuration that provided the combination of multipath and fading tolerance appropriate to the path. The on-air trials were automated, being synchronised at each end using Navstar Global Positioning System (GPS) satellite time sources, such that a number of modern image transfer) file transfer and (BER configurations and traffic types measurements,

69

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

_0

could be repeatedly exercised. Two frequencies were utilised: 4.8 MHz during the da and 2.8 MHz during the night changing at dawn and dusk in accordance with propagation predictions obtained from ICEPAC [Hand, 85].

o0

Malvem (Receiver)

170 km'--,, Cobbett Hill (Transmitter)

Figure 4-13 Configuration for

4.3.2

Results of On-Air

'On -Air' 16 kbps Modem Experiment

Tests

Trials data was collected over a five day period in May 1998 on the Cobbett Hill to Malvern

long both HC ISB the the and short and codec path using modem with

interleaving.

A transmitter

being kW 2 representative of the used, was power of

maximum power generally available on point-point

HF links. The data was analysed in

diurnal 4-14 Figure kilobyte blocks BER (-0.5 plot shows a one calculated. s) and the The (all BER the trial the modem provided error rates of modes). overall average over below 0.1 for in excess of 80% of the day, only performing worse than this during the

for be to some used night-time period. This performance should allow the modem digital voice applications (see section 4.6.4) where high error rates can be tolerated.

70

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

-1

An inspection of ionograms collected during the trials period, showed that the principal propagation modes present were single and multi-hop F region returns with occasional, lived, periods of E-region reflections. Ionospheric support was often limited to a short very narrow band of frequencies, especially at night (often < 0.5 MHz), and this has probably had a significant impact on the night-time results. Some form of automated frequency management could be expected to significantly improve the achieved results. 40000 30000 20000 10000 0 0.1

rn

0.01

No. data blocks

-Average 0.001 -1 0

0 0

0 0

0 0

0 0

O

o

O

O

O

0000000 0000000 r--NN Time

Figure

BER

(UTC)

Hourly Average BER between Cobbett Hill and Malvern using 16 kbps ISB -1-1-1 Modem with HC Codec (histogram shows number of data points in the data set).

The data, analysed in terms of kilobyte frame (block) delivery statistics, is summarised in Figure 4-15. The plots show the percentage of frames delivered at better than or equal to the stated BER. delivered error free while to 1900 UTC).

Over a 24 hour period 32% of all received frames were (0400 day-time just 43% figure the taking results to the rose

The results show that the modem, in its current form, could not be

(assuming data kbps for 16 an acceptable communications continuous reliably used BER criterion of n

ic/th Channels

Figure 4-17 Performance of Extended Bandwidth HDR Waveforms in a Rician Fading Channel (one non fading and one dB Gaussian fading mode). -6

4.6

Application

4.6.1

Data Communications

of Extended Bandwidth HDR Modems and Networking

The results of laboratory testing and trials of the new modem waveforms described in this chapter have shown that they may be used over benign channels where adequate be SNR be The could used effectively within an received waveforms can achieved. adaptive system employing ARQ alongside other more robust alternatives to maximise throughput under favourable conditions. The most likely application are for links that high indicate Calculations Rician that paths. are predominantly over surface wave or km distances to be on sea paths suggesting up availabilities should possible over -400 data links data HF and high BLOS communications. throughput maritime uses such as inter-networking.

4.6.2

Digital Broadcasting

The potential data rates of the experimental waveforms that have been considered may be suitable for use in high quality digital HF broadcasting. They offer an alternative

75

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels high-performance serial tone waveform technology to the OFDM waveforms presentl\

being standardisedfor systemssuchas DRM. 4.6.3

Image Transmission

A good application for high data rate HF modems is the transmission of imager%. Pictures (of the order of 160 Kbytes when compressed) were successfully transmitted image HF compression and coding techniques [Chippendale. 87] and the using robust 16 kbps modem in -90 s.

4.6.4

Range Extension for Line-of-Sight Radio Networks

A possible application of high data rate HF modems is for providing beyond line-ofsight (BLOS) range extension to V/UHF tactical communications networks (e.g. tactical military networks). therefore the traffic

Often, these existing links are relatively high BER bearers and that uses them is appropriately resilient.

An example is

continuously variable slope delta-modulation (CVSD, [Proakis, 14]) vocoders that are able to operate acceptably with BERs up to -20%. The use of 16 kbps HF for extending for CVSD technical the networks, while not an optimal solution, obviates need existing trans-coding between CVSD and low data rate vocoders such as LPC- 10 [NATO, 88] which is, of itself, a technically challenging task.

4.6.5

Maritime

Situation Awareness

Having established that extended bandwidth HDR modems can be expected to operate high is in a application well a maritime propagation environment one possible between data platforms. throughput system to allow the sharing of situation awareness Traditional military systems achieve this at HF using NATO Link-11 which has a 16, Link NATO provide line-of-sight throughput of -2.2 kbps whilst systems, such as There kbps. applications. 16 civil in many also are throughputs minimum excess of of

4.7

Modems HF HDR Bandwidth Standardisation of Extended

high of The work presented here made a direct contribution to the standardisation in MIL-STDISB) kHz 6 (i. throughput modem waveforms for extended bandwidths e.

76

Chapter 4A High Data Rate Modem for Extended Bandwidth Channels

--

188-1l OB [DOD, 2]. This standard is a major new release of an internationally applied HF modem interoperability standard.

4.8

Chapter Summary

This chapter has described a novel high data rate 16 kbps prototype HF modem. Unlike designed for in HF kHz 3 it in operation modems standard allocations extant operates an kHz 6 (either bandwidth SSB ISB). The ISB variant employs of contiguous or extended diversity between from Results HF that the signal exploits sidebands. architecture an have been The testing the and on-air of modem presented. measurements simulator high data have been limitations discussed of such rate modems suggesting performance they have application to surface wave and benign HF Skywave channels. The work in for higher throughput to the waveforms use standardisation of presented contributed fielded communications applications and demonstrated the value in exploiting ýýider bandwidths for HF radio applications.

The following chapters present new work on

digital HF radio architectures capable of supporting wideband and multi-channel transmission and reception.

77

Chapter 5 On the Specification and Design Digital HF Radios of

-N

Chapter 5.

On the Specification and Design of Digital HF Radios

For

many

years

multi-conversion

in high performance

dominant

radio communication

Recently, with the development and digital

converters

has

digitisation Sophisticated functionality

super-heterodyne

of affordable

signal processing

become

increasingly

implementations, in programmable

having

designs

receiver

systems (particularly

high performance

(DSP) technology, popular increasing

proportions

at HF).

analogue-to-digital

intermediate

[Wepman,

devices (programmable

have been

89],

frequency

[Brannon,

90].

of the radio/modem

logic devices (PLDs) and DSP

processors), are commonly termed software defined radios (SDRs). The ultimate goal is radio [Mitola,

the true software

91] in which the signal captured at the antenna is

digitised and processed entirely digitally field is developing implementations software

radios.

rapidly. that

in programmable devices. Technology in this

In reality there is almost a continuum of possible receiver

range from

The essential

conventional

analogue

features of these different

implementations

to true

classes of receiver

is

summarised in Figure 5-1.

"

Conventional -

" g x°)

LL cm c rn

Receiver:

Traditional all analogue receiver (RF to baseband) Commonly a super heterodyne architecture

Digital Receiver: -

Traditional analogue receiver RF front end Baseband or final IF narrow-band digitisation Digital signal processing (DSP) for filtering, demodulation etc Digital control of analogue sections

Software Defined Radio: Digital receiver, rapidly re-programmable to support different waveforms Waveform processing undertaken digitally, mainly in programmable devices

" True Software Radio: -

Wide-band digitisation 'close to the antenna'

-

digitally done baseband processing down-conversion, Channel selection, ASICs) FPGAs, fly' (using the 'on software, Highly re-configurable

-

Multi-mode,

-

multi-channel,

multi-band

Radio Softtivare Receiver io Analogue Figure 5-1 From Conventional

78

Chapter 5 On the Specification and Design of Digital HF Radios

79

Work presented in the previous chapter identified the need for radios capable of wideband and multi-channel operation to support higher throughput HF communications. This chapter provides a brief resume of the technology developments that are making software radios possible. The relative merits of a number of possible wideband receiver architectures are considered. Performance requirements for an HF dictated by the HF environment and the intended application, are studied in as receiver, detail. Results of work to characterise a high performance single channel some conventional super-heterodyne HF receiver are presented in Appendix C to help establish the current `state-of-the-art'.

The work described in this Chapter presents a

digitisation approach, applicable to HF, where the whole band (2new whole-band 30 MHz) is digitised and DSP algorithms are used to select, down-convert and demodulate signals of interest. Potential benefits of such an approach include: 9

reduced complexity and lower component count;

0

simultaneous reception of multiple signals;

0

programmable channel bandwidth (support for wideband channels):

0

software re-configurability; and

0

for high performance. potential

The predicted characteristics of a practical implementation of such a receiver are investigated in detail. Finally, an alternative wideband architecture employing a single the is with compared the predicted complexity/performance conversion proposed and direct sampling approach.

5.1

Applicable Technology Developments

Traditionally, a great deal of new radio communications technology was a result of developments However, are in new research support of military requirements. increasingly a result of the rapid progress being made to support the commercial sector. industry (PCS) In particular, the rapid growth of the personal communications systems base-stations small, and field develop its multi-standard new and with need to rapidly low-cost handsets is of note. Examples of the critical technologies that are enabling new radio architectures include:

79

chapter 5 On the Specification and Design of Di ital HF Radios

80

0

High performance analogue-to-digital converters (ADCs) and digital-to-anal oý`ue (DACs); converters

"

Low-cost digital up-converters (DUCs) and digital down-converters (DDCs) in the form of application specific integrated circuits (ASICs);

"

High speed, high capacity RAM based field programmable gate arras (FPGAs) and sophisticated development tools;

0

High performance general purpose DSP devices;

"

Improved RF devices (e.g. Monolithic Microwave Integrated Circuits, MMICs). and

0

New families of devices (e.g. miniature electro-mechanical structures, MEMS).

5.2

Wideband Digital Radio Architectures

There are a number of basic architectures applicable to HF receivers and also a number hybrids. of

These primarily

include super-heterodyne receivers, direct conversion

(zero-IF) receivers, single conversion (sub-harmonic) IF-sampling receivers and direct in digital Each these of are considered turn. sampling wideband receivers.

5.2.1

Conventional Super-Heterodyne Receivers

The architecture of a typical three conversion super-heterodyne (superhet) receiver is depicted in Figure 5-2. Signals applied to the antenna input are first filtered to remove tuneable They then (known a mixed with are pre-selection). as out-of-band components 81.4 frequency intermediate first (e. local g. synthesised oscillator signal to a common The selected signal is filtered, generally using crystal or designed is to bandwidth the filters, receiver to the mechanical maximum channel is (AGC) by circuit, handle. Amplification, controlled an automatic gain control in HF MHz 1.4 (e. frequency IF receiver) 2nd an is The g. to then a applied. mixed signal bandwidth. final filtering to the required where there is a further AGC amplifier and digital Narrow-band receiver baseband The final mix produces the output signal. MHz in an HF receiver).

hybrids may employ final IF or baseband digitisation.

80

Chapter 5 On the Specification

RF

Ueiigfn of Digital Hj, Radius (111(1

1st IF

(J

2nd IF

Baseband

ol

R

000.

0

L------------Synthesiser

Frequency

Standard

Figure 5-2 A Super-Heterodyne Receiver Architecture. IF stages are used because it is difficult

Multiple

to provide sufficient

(>_l 10 dB) and gain (up to 100 dB) at any single frequency without A high quality

occurring.

(typically

selectivity

signal leakage

synthesised) tuneable local oscillator

(LO) is

required to allow tuning to the frequency of interest in addition to fixed I, Os for the subsequent mixing stages. In order to prevent frequency errors all the local oscillators are normally

phase locked to a single frequency standard.

In most designs, limited

dynamic range at each stage necessitates the use of a complex AGC system to preserve receiver performance. receiver are generally

Whilst

excellent performance can be achieved, these types of

limited to receiving a single (normally

narrowband) channel at

design have high to time, and a component count. any one are challenging

5.2.2

Direct Conversion

In the direct conversion ('Zero signal directly 5-3).

to complex

IF') receiver a quadrature mixer is used to convert the

baseband where it can be amplified

and digitised (Figure

In theory this process results in perfect cancellation of the image signal.

potential advantages of this architecture simple

Digital Receiver

(Zero-IF)

filtering

requirements

in the signal chain,

in image the superhet. than suppression and easier

However for HF use, a direct conversion capable of tuning

include low complexity

The

high quality synthesiser a receiver requires

MHz) four (2-30 to octaves over close

and providing

accurate

is Such complex and offsets potential simplifications quadrature outputs. a synthesiser balance dependent the is on Further, performance critically achieved elsewhere. hybrid implemented as a single achieved within the quadrature mixer; normally imbalances will Any phase or 90° amplitude (including the component phase shifter). result in imperfect cancellation of the image.

81

Chapter 5 On the Specification and Design of Digital HF Radios

RF

m

82

Zero IF

Digital

lool, ,

40 1[

ADC Baseband Processor

0ý

Q ADC

-0-0-0010-90''

Synthesiser cos(2nf t)

Figure 5-3 Direct Conversion

'Zero IF' Receiver Architecture

It can be shown [Razavi, 92] that the image suppression (image rejection ratio. IRR) provided by a mixer with an amplitude magnitude unbalance

deviation of a phase -and

0from 90° is given by:

IRR=10lo

The implications

11_2(1)c0s0+(1+t)fl gýý 1+2(1+c)cos8+(I+c)'

dB

(j_1)

in demonstrated in imbalances amplitude and phase are of even small

the plot of IRR in Figure 5-4. Good, commercially

available, quadrature mixers might

dB 1° 0.1 balance respectively. and of typically guarantee an amplitude/phase dB of rejection. provides -40

This only

This is a particular problem for use at HF frequencies

in-band. be frequency image the may well where due This to DC amplitude is occur can Another particular problem the so-called offset. breakthrough due LO which to signal in It ADCs. or strong the can also occur offsets 94], Beach, 93; Rooyen, [van which DC hence in component a results self-mixing and lies right in the centre of the wanted complex basebandsignal.

82

Chapter 5 On the Specification and Design Digital HF Radios of

83

0

-10

40dB 2.0 dB

-20 10dß

05dß ö -30 025

ä

dB

c

J -a0

0.1 dB

E 005

dB

-50

O dB -60

70

of

i Phase

Higure )-4

Deviation

10 (Degrees)

image suppression in a Quadr(itrire direr .

Due to Aml)littrcle and Phase Imbalance

Correction techniques [e. g. Yuanbin, 95] can be applied to correct both the DC offset and mixer imbalance using a reference signal to measure the error and then to apply a correction to the IQ signal once digitised. cancellation, particularly

However it remains difficult

across wide bandwidths.

to ensure perfect

The problem of DC offset correction

is further exacerbated when receiver (or environment) motion causes a time dependency requiring continuous adaptation.

Practical direct conversion receivers have been demonstrated at HF. They are, however, fundamentally single-channel narrowband receivers.

5.2.3

Super-Heterodyne Receiver with Zero-IF Conversion

A hybrid HF receiver architecture (Figure 5-5) has been proposed by Coy el al [Coy. 96].

This employs a conventional

tuned synthesiser and heterodyne mixing stage to

fixed A be filter Ist IF to quadrature received. reach a selects the signal where a roofing down-conversion

stage, followed

required selectivity.

by quadrature

This narrowband

architecture

digitisation

has the advantage that whilst the

same IQ balance problems exist as in the direct-conversion overcome because only a single narrowband

is used to provide the

IF-frequency

receiver they are easier to be to considered. needs

A

83

m

Chapter 5 On the Specification and Design of Digital HF Radios

84

high performance digital receiver based on this architecture has been implemented b\ Coy and shown to give good results. However, it is fundamentally a single channel. narrowband receiver. RF

1st IF

Zero IF

Digital

00, ADC Baseband Processor

1100ADC 90

Synthesiser cost2af hA

Fixed LO

Figure 5-5 Super-heterodtivne with Zero-IF Stage

5.2.4

Single Conversion

Another possible architecture single conversion followed to directly down-convert ADC may be followed

IF-Sampling

Receiver

is a hybrid of the superhet receiver which relies on a

by filtering

and bandpass sampling (sub-sampling) of the IF

the IF to baseband. In a wideband multi-channel

receiver the

by a bank of digital receivers to recover a number of wanted

channels (Figure 5-6). The choice between single or multiple provide high selectivity

conversion

is primarily

images. to and reject mixer

related to the need to

In a receiver with narrowband

LO. In be tuneable a a required, requiring sampling multiple conversions will generally wideband receiver capable of whole-band digitisation,

filter both the pre-selector since

(before the mixer) and IF filter can be used to provide the required selectivity ahead of the ADC single conversion is sometimes possible.

In this case a single, fixed LO can be

employed. At least one conversion is required in higher frequency receivers (currently UHF and banddirect is limited bandwidth because and so prevents the ADC analogue above) for HF be to an If used this architecture were pass sampling at the signal frequency. direct its sampling a receiver with a single up-conversion only potential advantage over IF be RF an at applied gain can approach (discussed next) is that the majority of the frequency chosen to ensure amplifier harmonics will fall out-of-band. Disad\ anta-es

84

Chapter 5 On the Specification

and Design of Digital HF Radios

X5

include the additional analogue circuitry

required (LO. mixer and filter) and the associated non-linearities that they introduce. RF

1st IF

Digital

ADC

Digital Down Converter

Baseband Processorr

Digital Down

Baseband Processor

Converter

Digital Down Converter

--

2

Baseband Processor

Figure 5-6 Single Conversion IF-Sampling Receiver

5.2.5

Direct Sampling

A direct sampling digital dynamic

range ADC)

Wideband

Digital Receiver

receiver digitises the RF input signal directly (using a high

and then processes these signals digitally

(Figure 5-7).

The

input feeds filter antenna a which acts as a pre-selector and as the anti-aliasing filter for the receiver digitiser.

This is followed by an RF amplifier with digitally controlled gain.

As will be shown later, the overall dynamic range requirements of an HF receiver are so large (-140 dB) that it exceeds that of any currently 90+ dB).

The gain control

environment,

the receiver's

ADC is never saturated.

ADC

devices (typ.

is therefore required to ensure that, in a changing signal instantaneous dynamic

range is maximised

but that the

In a narrowband receiver AGC action is continual, primarily

tracking the power of the wanted signal (which additional unwanted modulation the adaptation

practical

in a fading channel can result in

on the received signal).

In this wideband architecture

is in response to the total energy present in the band (i. e. many

uncorrelated signals).

It would therefore be expected, in most circumstances, to require

longer). (possible timescale the or of minutes on adaptation at a much slower rate

To capture the whole HF band (2-30 MHz), the ADC digitisation rate would have to be in excess of 60 mega-samples-per-second (MSPS). All subsequent processing can be done in the digital domain including signal selection (filtering), frequency translation (down-conversion) and sample rate decimation.

The high processing rates of these

functions indicate that they should be implemented in (programmable) DDC ASICs or

85

Chapter 5 On the Specification and Design ofDigital HF Radios

86

PLDs. possibly

An arbitrary number of channels can then be extracted by using functions fed DDC from the single ADC. Great flexibility multiple over reception frequencies (fixed, sweeping or hopping), channel bandwidths, digitally applied gain. Finally, in possible. are as etc all the other architectures, general purpose DSP be for the relatively low rate basebandsignal processing (including may used processors functions such as signal demodulation, AGC etc). RF

Digital

1 ADC

t_

Digital Down Converter

Baseband Processor

º

Digital Down Converter

Baseband Processor

º2

Digital Down Converter

Baseband Processor

ºn

RF Gain Control_

Figure 5-7 Block Diagram of a Wideband HF Digital Receiver and, in particular,

ADCs and DDC ASIC

An investigation

of component

implementations

suggest that a high performance HF receiver with this architecture may

be practical for the first time.

technology

This thesis will go on to consider an HF receiver design

detail. in this with architecture much greater

In the next chapter results are given for a

prototype receiver of this type.

5.2.6

Architecture

for a Wideband Digital Transmitter Exciter

A complimentary, wideband, direct sampling architecture can be contemplated to fed to baseband Complex implement a multi-channel digital transmitter. signals are DUCs. The DUC output, interpolated up to a sampling rate appropriate for the RF RF The (DAC). resulting output frequency, is applied to a digital-to-analogue converter high A (PA). signal is amplified to allow it to drive (excite) a suitable power amplifier harmonics (aliases). DAC filter filter is to remove roll-off used as a re-construction

86

Chapter 5 On the Specification

and Design of Digital HF Radios

Baseband In 1

Baseband Processor

Digital Up Converter

Baseband In 2

Baseband Processor

Digital Up Converter DAC

Power Amplifier Baseband Processor

Baseband In n

Digital Up Converter

Figure 5-8 Block Diagram of a Wideband Digital Transmitter

5.3

HF Receiver Performance

The performance environment

requirements

investigated and are

of

Requirements

HF receivers

are largely

in the following

sub-sections.

governed

by the HF

These are compared

by high typical the provided a performance quality conventional receiver providing with implementations HF be baseline against new software radio which can evaluated. a The performance

characteristics

of a commercial

high performance super-heterodyne

HF receiver [Racal, 1041 have been measured to provide a comparison against which

described be The are measurements and results assessed. alternative architectures can in Appendix C.

5.3.1

Sensitivity

The sensitivity of a receiver is a measure of the weakest signal that can be satisfactorily limiting thermal is demodulated. This noise the to equivalent related often received and by: 97], [Fisk, input the the given system power at of

Thermal Noise Power = 10 log,,, (kTB)

dBW

(5-2)

where k

is Boltzmann's constant (1.38x10-'13J/K);

T

is the system temperature in degrees Kelvin: and

B

is the system bandwidth in (Hz).

87

Chapter 5 On the Specification and Design of Digital HF Radios

88

Noise figure (NF) is a standardised measure of a system's noise level aboýe the available thermal noise power. For a standard temperature, Ti., of 290 K the noise floor is of a system thus: NoiseFloor = NF + 10logo B -174

dBm

(5-3)

Reception in the HF band is often externally noise limited due to galactic, atmospheric is This illustrated in Figure 2-10 which shows the effective noise. man-made or different these noise sources, above thermal, in the HF band. contribution of In order to always be externally noise limited an HF receiving system, located at a quiet higher the and operating site at receive end of the band, should have a noise figure of This is equivalent to a noise floor of dBm/Hz dBm 13 or a signal -158 -1 dB in 10 SNR a standard 3 kHz bandwidth. As can be observed from the providing

>f) inter-modulation

distortion (IMD)

Suppression and SFDR

are applied to a non-linear device,

signals are generated [Fisk, 97]. The most significant

3t2 2/1-/2 f 2f:,, (3/], (2f ±f) third and the order products and are second order products j, 2f2-fi).

The third order inter-modulation

distortion

(IMD)

products, are of particular

level in increase decibel for by dB the test increase 3 because the of they every concern tones.

The input referenced third order intercept point (IP3IN) is the notional input

by level level the IMD produced the outputs as same signal products are at at which the the wanted tones. High quality conventional (+30 dBm is considered excellent).

HF receivers have an IP31Nof >_+20 dBm

By measuring the IMD (dBc), produced by two

D Appendix determined (see be IP31N (dBm), can equal power input tones of power PIN and [Kundert, 101]):

IP3I, _ IMD + P1 dBm %, 2

(5-5)

93

Chapter 5 On the Specification and Design of Digital HF Radios

94

Many manufacturers specify IMD performance in terms of the third intercept point referenced to the output of the device: IP30UT. Where the gain of a receiver is G. the intercept point referenced to the input and output of a receiver are related: IP31N =1P3OUT -G

(5-6)

Spurious free dynamic range (SFDR) is commonly defined as the signal input range from the noise floor to the largest signal level that will not generate 3`d order spurious floor. In an analogue receiver SFDR can therefore be defined that noise above products follows: as SFDR =3 {IP3,

logo BNF 174 + N -10

dB

(5--)

dBm

(5-8)

free input the maximum spurious signal level, P,,,,,,,: and P, =1 [2IP3, + 101og B+ NF -174 N io n A3

kHz bandwidth receiver with a 14 dB NF and +25 dBm IP31Nthus has a SFDR of

100 dB. During the practical work undertaken for this thesis the importance of 2"d order IMD important became increasingly They is ignored, clear. are an often performance, which because frequency higher (as in the HF to systems) opposed equipments consideration filtered be be fall in-band to they out subsequently may not able where products can (e.g. in a receiver front end). Amplifiers or mixers in conventional receivers, protected by narrowband filters, are largely immune to these 2"d order products because the frequency in they that interact them to removed are sufficiently cause signals that would are rejected by filtering.

Hence they may have high 2°d order intercept points ( IP21 ).

However components subject to wideband signals, such as front-end RF amplifiers and °d IMD 2 and products first in to order the producing mixer a receiver are vulnerable IP21 frequency. harmonics that may fall within the bandwidth of the selected receive N is given by (see Appendix D): IP21N -- PIN + IMD

dBm

(5-9)

been has analysed using The relative impact of 2d and 3`d order IMD products b\ IMD generated level products 5-13 the of Figure (5-5) shows equations and (5-9).

94

Chapter 5 On the Specification and Design o Digital HF Radios

95

two input signals as a function of the input signal level for a range of IP2/\ and IP31\ . The graph demonstrates the importance of considering both IP21, and IP3,, to achieve an acceptable overall level of performance.

High performance HF receivers

have IP21 >_+60dBm (>+70 dBm is considered typically will excellent). ,

-20 IP2IN=+40

-40

dBm

-

-IP21N=+50 dBm

--

IP21N=+60 dBm

--

IP21N=+70 dBm

---

IP21N=+80 dBm

/ /

IP31N=+0 dBm -60

o

-IP31N=+lOdBm

--

IP3IN=+20

--

ä

r-

-

-IP3IN=+30

/

dBm

r '/

/

f,

/

dBm

/

,'

-80

0

` -100

ý/

-120 ? ý5 140 lo -50

Ficure

/

-40

ý/

/

I

ýf

"' , NF=16, Receiver noise floor (3 kH z b andw idth )

A'

/// -.-

/

f

.

ý.

e 'O '

-20 -30 Two Tone Input Signal Power (dBm)

13 2"d and 3rd Order Intermodulation -5-

-10

0

Product Levels versus Two Tone Inpu, Power

Harmonic suppression (particularly of 2d and 3`d order harmonics) is also particularly important within the front-end of an HF receiver. This is becausethe HF band spans The in-band. fall MHz 15 will almost four octaves and so harmonics of signals up to The for IMD. to extent those to for harmonic suppression are equivalent requirements For design. function example is circuit harmonics of a much which are suppressed very harmonic 2"d designed be to effective provide may components such as amplifiers linearisation 102 [Dye, or p114] cancellation (e.g. using a push-pull architecture techniques).

95

Chapter 5 On the Specification and Design of Digital HF Radios 5.3.5

Practical

Impact of Receiver Intermodulation

96 Products

The discussion above has focused on the level of IMD product suppression achievable within receivers. The impact of intermodulation products on practical communications further consideration. requires some

Given a particular receiver performance, the harmful IMD products can be shown to be a of presence probabilistic function of the number of signals in the HF band with sufficient power to generate them and whether the frequencies of those signals are such that a product will be in-band of the anted transmission [Miller, 103]. Given sufficient information on the occurrence statistics of (frequency, in the HF band, such an argument can be used to determine power) signals the required dynamic range for an acceptable probability of being IMD free. The HF occupancy studies undertaken by Gott [e.g. Gott, 30] may be applicable for such an it is although not clear whether measurement data on the strength of signals analysis above a threshold of 100 µV/m (the highest threshold available in their published occupancy models) was retained.

5.3.6

Spurious Signal Products

In addition to IMD, receivers are also subject to other internally generated spurious signals such as unwanted mixer products, local oscillator leakage, harmonics etc. Typical receiver specifications call for >99% of 3 kHz channels to be free of spurious above the noise floor.

5.4

HF Transmitter

Performance Requirements

A signal to be transmitted must be adequately filtered to constrain its bandwidth: this is typically specified such that >99.9% of the power is contained in the allocated bandwidth (e.g. [ITU, 105], [NATO, 106]). Unwanted emissions (IMD, harmonics etc) for harmonic to be (minimum) outputs Typical are must also requirements minimised. in be -40 dBc and IMD the transmitting system dBc. Wideband generated noise -65 the does it sensitivity be kept level low compromise that not to must such a sufficiently dB 40 have a gain of of nearby receivers. A typical 500 W power amplifier might in input its a result would is large thermal at noise that even which sufficiently transmitter noise power of -134 dBm/Hz (-99 dBm in 3 kHz). In situations where a

96

Chapter 5 On the Specification and Design of Digital HF Radios

9-

HF transmitter exciter feeds a wideband power amplifier (PA) little or filtering be can applied and PA specifications are particularly onerous. no additional

multi-channelling

5.5

A Direct Sampling Digital HF Receiver

The following

sections presents a novel direct sampling wideband digital HF receiver and analyse the performance

architecture

expected to provide.

that a practical

implementation

may be

The next chapter presents measurements made on a prototype

digital transceiver which is presented in as a part of a wideband constructed receiver, Chapter 7. performance.

This

chapter concentrates

on the characteristics

and likely

Figure 5-14 is a diagram of the architecture investigated.

sections consider the performance

attainable

The following

of each of the key components of the receiver and

then its overall performance. Front End Filter

RF

In

Digital Attenuator

Digital Ane

Digital Control

RF Amplifier

º

ADC

Digital

Complex

Down

Baseband

Converter

Output

Digital Control

Figure 5-1-1 Wideband Direct Sampling Digital Receiver

5.6

Front End Filter

high digital the direct required The front-end filter provides the receiver with sampling lightning It secondary filtering provides Nyquist also to aliasing. prevent order

from over-voltage and over-current. protection and protection

97

Chapter 5 On the Specification and Desiew of Digital HF Radios

1st Nyquist Zone (Wanted Signal)

a

98

2nd Nyquist Zone (Aliased)

/

Useful Rejection

Frequency 0

fý

F/2

FS- fc

Figure 5-15 Front End Anti-Aliasing

F

Filter Performance Requirement

The choice of filter cut-off frequency is made such that, taking the filter's transition into consideration, no signals beyond the cut off frequency will cause in band region in (illustrated Figure 5-15). If the normalised transition ratio of the filter aliasing (lowest stop-band frequency divided by highest pass-band frequency) is CO, and making use of the symmetry of the aliasing process, then the relationship with the maximum unfrequency, f, is by: given receiver aliased F, - ýý < fc

f

(5-10)

PASS

where F,

is the digitiser sampling rate.

The minimum acceptable filter performance is therefore given by: F. w=-` -1 fl

(5-11)

The low pass filter is required to ideally provide 120 dB of attenuation for signals Whilst this limit in band the performance. receiver's outside the operating order not to is an apparently challenging requirement it is essentially the same requirement as that of limitations However, HF receiver. the image reject filter in a conventional narrowband in the maximum sampling rate of the ADC may impose an additional requirement to implement. is to band achieve a narrow transition challenging which

98

Chapter5 On the Specificationand Design of Digital HF Radios 5.7

Digitally

Controlled

99

RF Amplifier

The digitally controlled RF amplifier is required to provide sufficient gain to match the dynamic signal range with that of the ADC. It will be shown (in the next received dynamic the that range obtainable from currently available ADCs is insufficient section) to fully meet the needs of an ideal HF receiver. This limitation is analogous to the blocking performance in conventional narrowband receivers and it will be shown that (indeed equal) performance to a high quality narrowband receiver can be comparable achieved. The RF amplifier

in a wideband

receiver must have a noise figure low enough to

provide the required receiver sensitivity harmonic and IMD suppression.

and sufficient

Harmonic and 2nd order IMD is a particular concern

for HF receiver front-ends as, being multi-octave,

many of these products will fall in-

band. The proposed architecture employs a digitally by a fixed gain amplifier. amplifier

be tolerated can

linearity to provide acceptable

At HF the insertion

controlled RF attenuator followed

loss of the attenuator ahead of the

and has a number of advantages.

In a strong signal

environment, where the signal may have to be attenuated in any case (to match the ADC dynamic range), increasing front-end attenuation reduces IMD and harmonic products (increasing the effective receiver intercept points).

5.8

Analogue-to-Digital

Converter (ADC) Performance

The performance that a wideband digital receiver can achieve is highly dependent on key This the digitiser (ADC). characteristics considers the performance of the section done to digitiser quantify determine work that summarises and performance achievable HF digital in receiver be wideband the performance that can practical a achieved implementation. 14ADC high with The Analog Devices AD6644 [Analog, 107] is a new, performance in been has device This used bit precision and a maximum sampling rate of 65 MSPS. following The analysis and in the prototype digital receiver presented the next chapter. illustrated is using discussion of practically achievable digitiser performance information on the AD6644's characteristics.

99

Chapter 5 On the Specification and Design of Digital HF Radios

5.8.1

100

ADC Signal-to-Noise Ratio (SNR) Performance

When an analogue signal is sampled by a digitiser with finite quantisation inter\ als it be shown that the achievable signal-to-quantisation noise ratio (SQN R) is can readily Proakis, 108 by [e. p37]: g. given 3.22N

SQNR(dB) =10log, o 2

1.76 6.02. \' + =

ADC bits. is N For 14-bit the of number a converter, such as the AD6644, this where dB is bound SQNR; theoretical which a t86 on ADC SNR. gives that a practical ADC

device can achieve is principally

dynamic non-linearities performance, jitter. clock aperture

(DNLs)

The maximum SNR

limited by its analogue noise

in the conversion process and sampling

Equation (5-12) can be modified to include their impact [Analog.

107]:

z SNR(dB) = 1.76 - 20 10910

(2; N+

cF tý, s +v

where FA

is the analogue input frequency;

týrm. s

is the ADC sampling clock aperture jitter;

6

is the average DNL of the ADC (-0.41 for AD6644);

bjtt,

for AD6644); and in LSBs (-1.2 ADC is the equivalent RMS thermal noise

N

is the number of ADC bits.

figure: into The SNR of the ADC can be easily converted an equivalent noise logo SNR(dB) (dBm) NF1,,, (dBm) = P,. -10 ,, ßu1 (.

+ 174

(5-1-1)

where NF4(.

is the ADC noise figure; and

PEU/Isca, is the ADC full scale input power. eR15

100

Chapter 5 On the Specification and Design of Digital HF Radio s

5.8.2

101

ADC Noise Performance and Noise Figure

Equation (5-12) gave a bound on achievable SQNR determined purely b} the number of ADC quantisation intervals (SQNR,,zz86dB for a 14-bit converter). A more practical ADC SNR is the to consider the ratio between the largest input achievable of measure internally floor ADC. This may be written: the the generated noise and of signal V.

SQNR(dB) =1.76 + 20 log,,

rullscale_ADC

VHase

The internal noise generated within the AD6644 is equivalent to 1.2 least significant bits (LSBs) peak-to-peak thus the maximum achievable SQNR is: 2 14 SQNR(dB) = 1.76 + 20 log 10 2 '' =78.8dB

5.8.3

(5-1h)

Sampling Clock Jitter (Phase Noise)

When an analogue signal is sampled by an ADC any variation in the instantaneous instant will sampling

translate directly into a change in the quantised amplitude

from instant in The the in 5-16). Figure sampling (illustrated small random changes have jitter' `aperture `aperture a can and termed or uncertainty' are sample-to-sample impact ADC performance. on marked

101

Chapter 5 On the Specification

and Design of Digital

HF Radios

102

A A

Aperture Uncertainty (Jitter) At

Do-

-4

Encode Trigger

Figure 5-16 Error in SamplingAmplituie Given a sine wave of frequency,

F,

Due to ADC Aperture L icertainh" (Jitter)

its voltage is: , v=A

(5-I-)

sin(2TrF, l )

The gradient (slew rate) is given by the first derivative:

dv dt

A2, TF, cos(2;TF,t) =

(5 18)

At the nominal sampling instant, t=0, the signal slew rate is given by: Ov

dv

At

(It

A2/TFa, =

t=0

(5-19)

The error voltage at the sampling instant is the jitter, t,, multiplied by the signal slew rate

V//?/Ox At

tniix

t rFI =A2;

(5-20)

By considering a full scale input waveform it is then straight forward to "rite the jitter: RMS due to sampling theoretical SNR limitation imposed on the ADC

102

Chapter 5 On the Specification and Design of Digital HF Radios SNR = -20log,

)

t 1(2rc-,

103

dB

(5_21)

This is an important result. In particular it should be noted that the available SNR is limited not by the ADC sampling rate but by the frequency of the signal being sampled jitter. The jitter is due to that present on the total the sampling overall and sampling clock (oscillator)

and to aperture uncertainty

within

the ADC itself.

For the 14-bit

AD6644 ADC the nominal SNR is 75 dB (F5 0

GO

>1)

°O

0O

CO

MM°

yv

--Go0000 C,

'c ay

EU mVO výr.

7

1:]

zoý

UD

yi

m "b

--Q -ß ýTvCU c

lid

O Li

-NM

_a1 I_1

Sr }i

V, -5

pO ri th C:

f0 =JQ

`S

11_1 O 0

V5

CC

Receiver HF Sampling Direct Table 5-2 Predicted Performance of a

116

Chapter 5 On the Specification and Design of Digital HF Radios

5.11

An Alternative

11

Single Conversion Wideband Receiver

Architecture Given the demanding 2nd order IMD and harmonic performancerequirementin the direct a of

sampling receiver an alternative architecture «orthy of is consideration one employing a single, whole band up-conversion. Such a receiver

amplifier

its have the of majority signal gain at a wideband IF reducing the Ind order IMD would and harmonic performance requirements on that amplifier.

However. such an architecture would require additional circuit elements (mixer, IF filter) \\hich must have in linearity degrade to themselves order the receiver performance. If, in a not excellent implementation, RF is an amplifier required ahead of the mixer then its particular performance requirements are commensurate with that of the RF amplifier in the direct In such a case the single conversion approach is unlikely to provide approach. sampling benefit. any The IF frequency selected must be within the ADC's analogue bandwidth and lo\\ due SNR jitter (from to that the sampling reduction all sources) is acceptable. enough The IF following the frequency translation must be sufficiently high in frequency to designing image The both out-of-band rejection. experience of and good good provide the front-end (anti-aliasing) filter for the direct sampling receiver indicates that an IF frequency centred on -70 MHz would be practical and not be overly demanding on the sampling clock phase noise (jitter) performance. A review of the applicable literature has been undertaken to establish the performance based is the The use in on for presented analysis a receiver. such of mixers suitable use 120]). [Dexter, 119], [Cox, 118], [GEC, high (e. g. of a performance mixer 5-3, in Table is The predicted performance of a single conversion receiver shown dB. kHz)=111.3 BDR(3 dBm and namely: NF=13.4 dB, IP3IN=+22.7 dBm, IP2IN=+77.1 b) improved be In this case the performance is limited by the IF amplifier and could ith ýti amplifier an However, this require higher would using a performance alternative. give would and direct for receiver the sampling similar performance to that identified similar performance but have greater complexity. wideband whole-band digitising

for that a It is therefore concluded

HF receiver this architecture offers no significant

additionalbenefit.

117

Chapter 5 On the Specification and Design of Digital HF Radios

118

77 N

LL

e

77

ö

-r

r

-r

,ý

N

ýö 2

ov

ce

uni

rn

, ".

Op

-

rv-

v-r

r+

ýc

e

T

5

9

ýn

m

?

ti_ coc-

!

6O

m

K

1

-ya

d9

r M

ýC

_>QM?

a -e

v

'S

r

=

Y Vp

I fCý

NE

NC

-v

-o

fý

tý

C

0 0

0 0

0 0

0 0

0 0

!'T

z Ixy

xv

d 0 m y

ä

ýE

o 0

O 0

O 0

O 0

O 0

O 0

O Oý

N

N

Ö

Ö

Ö

Ö

ý

o o

0 ýn

0 o

0 0

0 0

0

CO

-Y

O

V

O

V

O o 0 c> 0 0 o ýn W O CO

0 0 o O

o 0 0 O

O 0 o O

o 0 vý O

O

O

O

O

ý

äa

6 aE

CO

ý NE dm

N

T

0 O O

0 O V

0 O O

0 ý V

0 O O

0 o0 O? N

Ö

& rn

o O

0 Q

o Q

o 0

O Q

N

O

N

O

ö

y

ä

O LLý^

x

r'7

ý,

m

e

i y

5

p

N

O

Uo

c n

Q ,

ai

0 =

Lt

r,

gc b c.

y

N

M

dE

V

0

Q

C

x

o

ll

m

r[?

G ^-i

I `m G

c0

r-

1/F Digital C'of71'ersiorl Table 5-3 Predicted Performance of a Wideband Single Receiver

118

Chapter 5 On the Specification and Design ofDigital HF Radios

5.12

119

Chapter Summary

This chapter has considered the implementation of wideband digital HF radios. The HF the that environment places on receiver design have been investigated. requirements The performance of a high performance conventional narrowband HF receiver has been basis for comparison. to establish a characterised The work presented indicates that it is now possible, for the first time, to construct a direct high performance, sampling wideband digital HF receiver. Such a receiver very would conceptually allow an arbitrary number of channels to be simultaneousl} front-end RF a single using and digitiser. With careful design of the front-end received filter and selection of a suitable RF amplifier, performance closely matching or even highest the that of performance single channel receivers commerciall\ exceeding be obtained. Whilst receivers employing front-end pre-selection (i. e. subcan available filters), have the potential to offer higher performance, the application of such octave filtering makes them inherently narrowband. It has been shown that in a direct sampling receiver, front-end linearity, particularly for the RF amplifier, is critical. As there is no frequency translation in such a receiver and band HF the since covers approaching four octaves the amplifier second order intermodulation and harmonic performance will have a major impact on the overall receiver strong signal handling performance.

The RF amplifier

performance

requirements have been calculated and a number of suitable commercial RF amplifiers have been identified showing that the proposition is practical. An alternative wideband receiver architecture employing a single, whole band, upconversion has been investigated and compared with the direct sampling approach. Sucha receiver would have the majority of its signal gain at a wideband IF reducing the 2ndorder IMD and harmonic performance requirements on that amplifier. However, filter) IF (mixer, which such an architecture would require additional circuit elements degrade the have receiver in themselves linearity to must not excellent order is this implementations architecture performance. It is believed that, in practical unlikely to offer any benefit over the direct sampling approach. has design that HF The following chapter presents a prototype direct sampling receiver beenconstructed and evaluated.

119

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver

120

Chapter 6.

Performance of a Prototype Direct Sampling Digital HF Receiver

The previous chapter considered the requirements for high performance HF receivers and examined the characteristics of practical direct sampling receiver implementations. This chapter presents the design of a prototype receiver and measured performance from laboratory a prototype Suggestions for improvements to the results obtained design are advanced. prototype The prototype digital receiver was implemented as part of a digital transceiver x6ose is described in detail in the next chapter. This chapter concentrateson the construction basic implementation and its measured characteristics.

6.1

Description of Prototype Receiver

Figure 6-1 is a block diagram of the prototype direct sampling digital receiver that has been designed and constructed. A separate front-end protection and filtering module filtering the provides overload protection and ahead of the receiver necessary anti-alias itself. In the prototype a digitally controlled RF switch allows the receiver input to be RF by is followed between different This amplifier, gain a variable switched sources. implemented as a digitally controlled attenuator followed by a fixed gain RF amplifier. A 30 MHz harmonic and noise reduction filter minimises the level of internally digitally A digitiser. controlled narrowband the generatedout-of-band signals reaching dither source (applied below 1 MHz) is included to maximise the ADC SFDR. The digitised signal is passed to a programmable DDC ASIC which selects and domnconvertsthe channel to be received.

() 121

Chapter 6 Perfoi°mance of u Prototype Direct Sampling Digital HF Receiver

Cj

tr

.. I

_

L'

_

_ v

r

L5

LL

7 -t

CL

1'

T

2

U

L

L lä

ý

v1 ii

-

j.

T_

0

_-

I

_

II 1I II I

LL =

.1

LJC

D,

-;

--

Figure 6-1 Block Diagram of Protohpe DirecOanipttrig iJigliur 111'lk LcII

121

Chapter 6 Performance of a Prototype Direct SamplinDigital

6.1.1

HF Receiver

12"

Front-End Protection and Anti-Alias Filter

Following experimentation with a number of prototype designs the final anti-aliasing filter implementation uses a dual stage high-order design. The first is a 7`horder elliptic frequency. Tuneable ferrite-cored inductors are used to 28 MHz corner LPF with a loss, pass-band corner, and transition region to be optimised. The insertion the allow LPF, 9th fixed filter, inductors for their order elliptical makes a use of air-cored second high Q. These were found to give the greatest overall attenuation and to minimise rollfact In large inherent frequencies. bandwidth the higher RF analogue the of at up for MHz AD6644) (250 in for ADC results a requirement a filter that amplifier and beyond 500 MHz. The its well measured performance of the performance maintains filter is shown in Figure 6-2 (selectivity), Figure 6-3 (group delay variation) and Figure 6-4 (input/output impedance matching). Whilst the filter selectivity was generally found to be satisfactory (approaching 120 dB) it would benefit from additional work to reduce the insertion loss at the top of the HF band and to improve its VSWR.

122

Chapter 6 Performance ofa Prototape f)/r ct SýnýtýliýýýrUirilý, l lnf Recc-ivL7

I;

0 --20 40 0 J C 0

-60

u,

-80

-

-100 -120 -140 0

10

20

30

40

50

60

70

80

90

100

Frequency (MHz) 0 -20 0) 0 C

-40 60

0 (D -80

-100 U)

-120 140 0

100

200

300

400

500

Frequency (MHz)

Figure 6-2 Measured Selectivity ofCombined 28 MHz Elliptic Low Pass Filter

0.5

N 0.4

C)

0.3

U 0.2

0.1

0.0 05

10

15

20

25

30

Frequency (MHz)

Figure 6-3 Front-End Filter Group Delay Variation

1233

Chupter 6 Performance of a Prototype Direct Sampling Digital HF Receiver

0

124

10

8

sii -10 ä)-15

4> SWR -20

2

-25

0

05

10

15

20

25

30

Frequency (MHz) 08 -5

7

S22

-10

6

m -15

5

N-20 N>

4U)

-25

3

vswR

-30 -2 -35

1

-40

0 05

10

15

20

25

30

Frequency (MHz)

Figure 6--1 Prototype Front-End Filter Impedance Alatching

6.1.2

Gallium Arsenide (GaAs) MMIC RF Amplifier with Digital Gain Control

This section discusses the performance of the GaAs miniature microwave integrated The direct in (MMIC) RF amplifier used the prototype sampling receiver. circuit 1211 [Stanford, block 5052 SNA-586 which was readily amplifier used was a type gain It linearity. NF was third had order and available, small, easy to use and good published ideal is IMD) performance not subsequently found that its second order (harmonic and for this application. '

the to it related closely The choice of the RF amplifier device and the decision to not replace are developing to investigated first means a as progressof the project as a whole. The digital receiver was An front-end. a wideband down-converter and digitiser following a conventional super-heterodyne it limited Due resources to for that purpose. early prototype had been built using the chosen amplifier of mitigation filter some to harmonic provide was decided to utilise the same device but to include a has harmonics above 15 MHz. Further work on characterising the digital receiver performance quantified the limitations due to the amplifier.

124

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver

?5

*8V (85 mA)

Cl

C2

100 nF

tO uF

T7

33R 33

C3 27 pF L1 47 uH U1

C4

MA

C5

11 IN

OU

SMA

31

0.1 uF

01

uF

N

SNA-586 NQ

figure

6-J JNASti6 GaAs RFAmpli,

er Test Circuit

The amplifier has been characterised using the circuit of Figure 6-5 constructed as a test piece (Figure 6-6).

The amplifier was measured in this configuration using an s-

parameter network analyser. The results indicate that it offers acceptable gain and matching performance for the intended HF application (Figure 6-7).

-V 4W

,a

r

ýýs,

.maß

n:.

gý

ýJ_

Figure 6-6 MMIC' Kr Amplifier iesi riece. y

harmonic IP21N its lt has been found that whilst its IP3IN was acceptable and for strong digital full capability inadequate receiver performance where to realise the Using dBm. 17.5 + signal handling. The amplifier's measured IP3IN was measured as in IP21N excess Figure 5-13 it can be seen that to realise the full SFDR this requires an 125

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver

126

for Measured dBm. the results +70 amplifier's second and third order intercept point of frequency in function Figure 6-8. Table are shown 6-1 presents the measured of a as Using these results and assuming a standard Figure 6-9 has been drawn to show at what input signal level characteristic order second

harmonic performance of the amplifier.

2nd and 3rd order harmonics would start to impact harmonically related channels. Products of order greater than three were not found to be a significant performance limitation. 5

0 -2

-22 -

4 _24 3

-6 N

pß Ti N

m a ry -26

2> -10 28

-12 -14 0

-16 0

5

10

20

15

25

_30

30

0

5

10

Frequency (MHz)

15

20

25

20

25

30

Frequency (MHz)

22

05 2

20

4 b

18

m'm 83 -10 R-12

16

12 vNi VSWR

14

gp

12

-14 -20 i -18

0 05

10

15 Frequency (MHz)

20

25

30

10 05

10

15

30

Frequency (Mw)

Network Measured Characteristics Amplifier RF GaAs on SNA-586 Figure 6-7 Analyser

126

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver

127

30

30

25

25

20

20

Co

C ä 15

15 C U --SNA-586IP2in

10

--

10

SNA-586 IP3in SNA-586 Gain

55

00 10

05

15 Frequency

20

(MHz)

30

(MHz)

Figure 6-8 SNA-586 GaAs RFAinplifier

Frequency

25

Gain und Linearity Measurements

2nd Harmonic (dBc)

3rd Harmonic (dBc)

2

-42.0

-55.0

5

-44.0

-51.1

10

-41.0

-56.0

15

-42.0

-57.0

20

-42.3

-57.1

25

-39.5

-57.3

30

-40.0

-59.0

2-30 (Mean of above)

-41.5

-56.1

dBm input) (-15 Performance Harmonic Table 6-1 SNA-586 Measured

127

Chapter 6 Performance of a Prntotipe Direct Sampling Digital HF Receiver

128

0-

-2O

-ao

a CL

-60-

d 0 a' d

-80

d c O

c -100 0 j 0Q -12

w

-140

160 -80

-70

-60

-50

-40 -30 Input Signal Power (dBm)

Figure 6-9 SNA-586 RF Amplifier 2ndand

6.1.3

Harmonic

-20

-10

0

Harmonic Performance (Extrapolated)

Filter

A fifth order 0.25 dB Chebychev filter with a corner frequency, I.. of 30 MHz has been implemented to minimise

the impact of harmonics that fall out-of-band

(particularly those due to the RF amplifier used). The filter was modelled using the SPICE analogue circuit simulation tool (Figure 6-10).

128

Chapter 6 Performance of a Prototype Direct Sampling DigiIul I/F Receiver

10.0

129

" 20*log 10 ('v(out)Y'v(in)')

-0.0

-10.0

T

U)

-20.01

-30.0-

-40.0-

-50.

-60.0 O.OM

20. OM

40. OM

60. OM

80.0M

100. OM

Frequency (Hz)

F inure 6-10 Selectivity of 5 t1 Order Harmonic Filter (Modellei I H'ith S'PIC'E.) ,

6.1.4

Digitiser (Analogue-to-Digital

Converter)

A high performance ADC, the Analog Devices AD6644 [Analog, 107], that became available (as a pre-production `X-Grade' part) at the time that this study for in direct the sampling receiver. was selected use prototype

as undertaken This is a high

The AD6644 65 MSPS. 14-bit ADC to rates up performance which supports sampling has a calculated NF of 29.8 dB and a full-scale input power of 4.8 dBm (using a transformer coupled input). This equates to an ADC IDR/BDR of 149 dB (114 dB in 3 kHz). The basic ADC SFDR is 90 dB. With the use of dither a SFDR of>110 dB can be achieved. Therefore a narrowband dither generator is included in the prototype implementation and allows a dither noise signal to be added to the input of the ADC. To maximise performance the dither is added out-of-band at frequencies below I MHz.

6.1.5

Sampling Clock Generation

implemented been has In the prototype direct sampling digital receiver a sampling clock by utilising a high quality voltage controlled crystal oscillator (VCXO) with a narrow (-15 kHz) tuning range, and phase locking this to either a temperature compensated (e. frequency a g. standard high crystal oscillator (TCXO) or a stability external ]) 12? [Connor. VRI Rubidium standard). The VCXO utilised (Connor Winfield HV54

1ý9

Chapter 6 Performance of u Prototone Direct Sampling Digital HF Receiver

130

has an RMS phase jitter of _1 MHz

-155* `Estimated

Value

Table 6-2 Prototype Widehand Digital Receiver TC'XO/L'C'XO Combined Phase Noise

6.1.6

DDC Performance

in Prototype

Wideband

Digital Receiver

The prototype direct sampling digital receiver makes use of a Graychip GC4014 DDC ASIC [Graychip, 124]. This device contains four independent DDC cores enabling down is input The to four complex mixed to signals simultaneously. reception of up baseband using a complex NCO. R=1) cascaded integrator-comb (M=16... 32k).

This is followed

It is then filtered and decimated using a 4-stage (L=4, (CIC) digital filter including a programmable decimator by a 21-tap, decimate-by-2.

low pass filter and then a 63-tap, decimate-by-two low pass filter (Figure 6-11).

FIR (CFIR) compensating

FIR (PFIR) by-four programmable or

baseband (IQ) 16-bit device The complex produces a

be to digital programmed may receiver output. The gain through each stage of the level. input for signal a given maximise the instantaneous dynamic range

I-110

Chapter 6 Performance of'a Prototype Direct Sampling Digital HF Receiver

Input

Deamate by 8-16K

Deamate by 2

Complex Baseband

Deamate by 2 or 4

14

131

16

Decimating CIC Filter

Compensating -i FIR Filter

programmable -i FIR Fitter

Format & --0' 16 Cam

I

Decimation

Mode

Tap Weights

Format Gain

Frequency

rigure

o-11 rrogrammaete

Digital Down-Converter

[after Graschip, 124J

The GC4014 NCO (Figure 6-12) maintains a 32-bit phaseaddressregister ýtihich it uses to access a 16384 value (2'`') sine/cosine look-up table to generate 16-hit output look-up finite The table word-length samples.

leads to periodic errors which manifest

themselves as spurious NCO outputs.

In an analogous manner to the techniques used to

improve ADC spurious performance,

a digital noise signal (dither) can he applied to

improve the SFDR.

F-IA_E OFFSET 22 E TS

vVC,ýG`

D T-i F. OE', E,. 4TOI 13 31-S

ISETS

E fJ=ý: D51. E L` ,HJ

TA5LE

Bý 'F

_° C-T

11,E

Figure 6-12 GC-1014DDC NCO Implementation [Gruvchip, 12-I] The GC4014 NCO performance, taken from the data sheet, is reproduced in Figure 6-13. Without dithering the worst case spurs occur at -82 dBc (Figure 6-13a) whilst. 6-13b). dBc (Figure below fall with dithering applied, the majority of spurs -105 in dBc. the one shown Without dither the worst case NCO spurs occur at -82 such as frequency the to sampling Figure 6-13c, and are due to a few frequencies that are related in the errors by small rational numbers (e.g. 3/16 * Fs). In these cases the rounding the error fashion, thereby concentrating in lookup sine/cosine table repeat a regular These worst it the spectrum. power at a single frequency, rather than spreading across by the or errors that initial minimises by phase casespurs can be avoided selecting an that data all The states Hz). sheet changing the tuning frequency by a small amount (50

131

Chapter 6 Performance of'a Prototype Direct Samling

Digital HF Receiver

132

fall below be dBc to made can with the selection of a proper initial phase or -96 spurs tuning frequency.

F

-T-

-10'IE

I 1 iºý1ý.

ý1ilRIýr4"

1ý7Ji. .

t+:

Mua. ýiLJf-14

. ,w"s'I`'i,

.'.

Dither) Spurs (no NCO GC4014 a.

IrtY ý,tý 1i'JAW wRý»i"ýr+MIýºW

b. GC4014 NCO Spurs (Dither applied)

3ý CE -32 CE

d. GC4014 NCO Peak Spur Scan (Dither applied)

c. GC4014 NCO Peak Spur Scan (no Dither)

Figure 6-13 GC401-1 Numericulhv Controlled

Oscillator

(NCO) Spurs [Grabchip, 124]

The CIC output is filtered by two stages of filtering. The first stage is the CFIR, which band CIC for the the and provides close-in selectivity. pass slope of compensates based) (ROM two coefficients of sets of

One

The be set of coefficients used selected. may

in the `normal' mode give a pass-band which is flat (0.01 dB ripple) over 100% of the final output bandwidth

band dB has 85 rejection. of of out and which

The 'narrow'

halving the band dB 10 the of expense at rejection out of mode coefficients provide >1 bandwidth. useful output

The second stage decimate-by-two

either internal ROM based coefficients, internal 80% bandwidth

filter four PFIR uses or

The filter downloaded coefficients. or externally

PFIR filter provides 80 dB of out of band image rejection and

0.03 dB peak-to-peak pass-band ripple.

132

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver

133

Figure 6-14 Frequency Responseof GC4014 CFIR Filter (3 kHz. Vi quIi.sI bandwidth) Figure 6-15a shows the overall DDC selectivity in the normal mode and Figure 6-15b in the narrow mode (as modelled in MATLAB using the GC4014 CIC, CFIR and PFIR filter specifications). The DDC implementation is designed such that the peaks in the stop band at 3.5 times the output sample rate will, after decimation, fold into the transition band from 0.4 to 0.5 of the output sample rate. This out of band power can be filtered out by either using a custom PFIR filter with a narrower pass band, by or postfiltering the DDC output. In summary the GC4014 DDC used in the prototype receiver is able to provide >110 dB selectivity within the centre of the pass band and a SFDR of >_102dB (typically 105110dB).

00 -20 --- ----------------------------------------------------------------m -40 --- ----------------------------------------------------------------Peaks alias into baseband transition region a. -60 -- --------- ----------------------------- - --------------n

-20 ---- -------------------------------------------------------------p^p -40 --- ----------------- ---------------60 --

-----------------------------

---------------------------------------------------------------

-. _. ---____---..___. -80 _____________--____---____--____--___----_.

_80 .)

------------------------

Frequency(kHz) a DDC Selectivity (CIC, Normal CFIR, Default PFIR)

--

- -------------------------------

30 20 Freauencv(kHz)

40

50

b. DDC Selectivity (CIC, Narrow CFIR, Default PFIR)

Figure 6-15 Modelled Performance of GC4014 DDC (3 kHz Avguist bandwidth)

133

('harter 6 Performance of a Protolupe Direct Sampling;Digital HF Receil er

6.2

1; -1

Predicted Performance of Prototype Digital HF Receiver

Following an examination of the major components in the digital receiver that effect has been parameter cascaded analysis a undertaken to predict the performance, has been that the receiver prototype of constructed. The results of this are performance has been repeated for inputs of 6-3 dBm in Table dBm, which and which given -113 -10 is the predicted ADC clipping level (0 dBFS). The key predicted performance from this analysis are as follows: NF=16 dB, IP3FN=+19dBm, that come parameters IP2IN=+27dBm, BDR=110 dB. The 3 kHz channel BDR is predicted to be -I 10 dB (determined by the ADC noise floor). The previous analysis indicated that the DDC dB in Hz bandwidth dB 1 in kHz) but (80 3 be IDR this of may -115 will provide an (by impacting is DDC it the the programming gain) without required wherever placed front-end performance.

134

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver

N

N

ýýý^M

yo co

[G 6

z,

--

cc

Dv

v

K

Ný

yZi

ý+

?

ý

iW



co

co

to

?

yO

O

S NM1

äm

"

Ö,

^

l0

N ý,

m

ý

d

6ý '. ýD

vvv

i

T

6,

r) .I

yv

7

c

äE

II

Ol

QQ4 e

am

v

v

T

Vý

aý

T

N°

jc -

S.

ov

-m

yE

rp'

M

.J

O

O

c

1ý

oe

ý"

V+

a

135

d

'ä

m

i

ý

-a

ýc> ýn ýn ýn

m

I

L rr, -r 6

o o0

-

E

y mC

.ý

M O

Nm

dG

M

O

l0

.

ýD

M

N

V

1m

yg

äo

ý+ oo

yJ n

0 o o

0 co m N

>A >0

O O

O 0A

O M

O O

m'

O O

a E

d

C

m

e.

'"

M

-

ý.

N

O

N

N

O M O

O

O

C öE mm >°

O

O c

_

0 O

LL ZO

B"

a a 'u

0 O

0 O

°

y

d y

=

_

NC pm

ö

y Z9

L.

_

^l

o 0

°

^

ýy

C

0 0 0 '1

ýý L7

ý

0 O

-

0 0

N

O

ý(J YJ

C

ý,

o

N.

O

O

o

0

G M

O

0

0

9

u

y D

OC

ým

Z, «ý pä

E

e p

C7

N

O VO

O Q äp

Ö

O

Li

C

L

_

_

(ISi o j M

W C

N 6, M

r

`ý V7

C

CO

z

N

Q

M

C7

V

w

V

m

G

M

ýn

[fi

2 Cn

o

ý

O

11

O

4

O

I

0

0

l0

Q

N-

r

N

d

ýO

H

ö C

_

zý

_

Q M

C o C i

I ý[J

o ä

V p

V

O

y`

p ,D

ý Q

pý C

ý p Q I+

.. Ö

a

c

J

d

r

N

Receiver HF Sampling Direct Table 6-3 Predicted Performance of Prototype Wideband

135

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver

136

Prototype Receiver Performance Measurements

6.3

Following construction the performance of the prototype receiver has also been by undertaking a series of measurements. The complex basebandsignal characterised from disk to the output of the receiver's DDC. This was then analysed was streamed The MATLAB. here results presented are from measurementsmade without the using front-end filtering and protection module. The following measurements Nýere made to basic the receiver performance: establish "

Impact of dither on ADC SFDR;

0

Receiver sensitivity (noise figure);

0

Third order intermodulation products;

0

Second order intermodulation products;

"

Harmonic performance;

"

Receiver SFDR;

0

Blocking Dynamic Range (BDR) and Instantaneous Dynamic Range (IDR):

0

62.208 MHz sampling clock phase noise; and

0

Under-sampling (sub-octave sampling) performance.

6.3.1

Impact of Dither on ADC SFDR

In order to assesthe benefit of adding dither to the input of the ADC input to improve dBFS for the ADC) signals (11.0 MHz and its linearity, two dBm (equivalent to -20 -30 11.01MHz) were applied to the receiver. Figure 6-16 shows that, with the receiver's dither generator switched on, the power of spurious signals decreasedfrom -80 dBc (by (set level dither The dBFS). 100 dBFS) to approximately -90 dBc (-110 used commanding a DAC which in turn set the gain of a voltage controlled amplifier) was improvement SFDR further increased level being no until the from the analysis improvement The predicted the was seen. results measured agree with determined empirically

presentedearlier.

136

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver Power Spectral Density, FFT Size: 1024 points

POwer Spectral Density

-20

-20

w0

40-

b0

ao

FFT Ci, o -...

137

-

IJ A0

$0

0

^ýMW

a -100

-100

12 -5

-2

-4-3

Figure 6-I6

-3

-2

x 10s

Dither

a. No added

6.3.2

-120 .4

012345 -1 Frequency(Hz)

-1

0 Frequency (Hz)

234 x 10`

b. With Dither

Lffect of Adding Dither to ADC Input Signal (Input tones are dBFS) -20

Receiver Sensitivity

With the receiver bandwidth set to 1 kHz a single RF tone was applied and its level its dB 45 floor (equivalent to an SNR of 10 dB the output was above until noise reduced in 3 kHz). The required input signal level to achieve this was dBm (see Figure -113 6-17), giving a noise floor of -158 dBm/Hz. This accorded closely with the predicted receiver sensitivity. Power Spectral

Density:

7.600 MHz, 1024-pt FFT

0

-10

m

-20

Ca

N -30

0

n

-40

o-50 a

-60

-70 -600

-400

-200

0

200

400

600

Frequency (Hz)

Sensitivity: aIim ty_ Receiver Digital Measured 6-17 Figure -1)5

117

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver

6.3.3

Third Order Intermodulation

138

Products

Using two signal generators, two -20 dBm tones were applied to the receiver at a kHz. 20 The intermodulation third of order separation product could then be clearly Fourier transform of the receiver output was taken (Figure 6-18). A the when seen dBc is of equivalent to a third order intercept point of + 19 dBm. spurious output -78 Power Spectral

Density:

7.510 MHz, 1024-pt FFT

0

-20

CO

dBc IMD -78

CO -40 N 0

-60 aN

d U)

äý

-80

0 a-

-100

-120 -5

-4

-3

-2

-1

012345 Frequency (Hz)

x 104

Figure 6-18 Measured Digital Receiver IMD using 2 tones at -20 dBm

6.3.4

Second Order Intermodulation

Products

The second order intermodulation performance of the receiver was measured using two IMD in 2nd This dBm MHz. input 7.2 6.3 MHz product order a tones resulted and at -15 2nd The dBm. be dBm +27 order Hence IP2IN to 14.5 MHz. of -57 was calculated at level input in dB because the signal change nature of these products was confirmed aI resulted in a2 dB change in the IMD level.

6.3.4.1 Harmonic Products The harmonic performance of the receiver was found to accord closely with the results 6.1.2). (see RF amplifier presentedearlier for the harmonic performance of the

138

Chapter 6 Performance ofa Prototype Direct SamplinDiDigital HF Receiver

6.3.5

139

Blocking Dynamic Range (BDR) and Instantaneous Dynamic Range (IDR)

The receiver's maximum input signal level was confirmed by applying an RF tone and increasing its power until the ADC overload indicator was activated. The pm\er dBm. do Given this the measured noise floor of to was dBm/Hz -13 required this -158 dB. BDR 145 the of a receiver gives The IDR was measured by applying a -15 dBm signal to the receiver, and adjusting the down-converter its dynamic in to the maximise range. The result of this test is gain be 6-19. As IDR is in Figure dB 115 the can seen which is less than the BDR shown by DDC (word length, limited is NCO performance etc) as the performance and Since is discussed. DDC the the output programmable it is always possible previously to adjust its gain to maximise the IDR for the signal power in the selected channel (frequency/bandwidth). Power Spectral

Density:

7.600 MHz, 1024-pt FFT

0

-20 d CO -40 >115 dB SNR aý

0

60

Ü N

a -80 0

-100

-120 -600

-400

-200

200

0 Frequency

400

600

(Hz)

Figure 6-19 Measured Receiver Instantaneous Dynamic Range with -1.) dtcm input

139

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver

6.3.6

140

Performance of 62.208 MHz Sampling Clock Generator

A prototype of the 62.208 MHz phase locked master oscillator \Nas constructed and testedby using it as the sampling clock source on the prototype receiver. A high quality to was used provide a test tone and the received signal spectrally generator signal in The Matlab. measured oscillator performance is as shown in Figure 6-20 to analysed Figure 6-22. Figure 6-20 shows the phase noise of the VCXO free-running. In Figure 6-21 it is shown phase-locked to the TCXO improving its performance. In Figure 6-22 the performance when the VCXO is phase locked to the TCXO which is, in turn, phase locked to an external high quality frequency standard (some very low level spurs, probably due to the PLL, are evident in this case). This demonstrates that near ideal performance was achieved with the close-in phase noise being determined by the 10 MHz TCXO reference oscillator and, outside of the loop bandwidth, by the 62.208 MHz VCXO. Power Spectral Density, FFT Size: 1024 points 0

-20 M

-40

-60

-80 U, ö -100 a -120

-140 -600

-400

-200

0 Frequency (Hz)

200

400

600

Figure 6-20 62.208 MHz Sampling Clock Phase Noise (Free Running)

140

Chapter 6 Performance of a Prototype Direct Sampling Digital HF Receiver Power Spectral Density, FFT Size: 1024 points

141

Power Spectral Density, FFT Size: 1024 points

00

-20

-20

R m

40Co L`

-40

T

L

-60

-60

Sö

-80 N

fn

3: -100

o -100 ý^*M1+fwyý_ -120

-120

-140 -600

-400

-200

0 Frequency

200

400

(Hz)

600

-140 -2.5

-2

-1.5

-1

0 0.5 -0.5 Frequency (Hz)

P igure 6-21 62.208 MHz Sampling Clock Performance (I'(XO

0

1

1.5

2

2.5 10'

locked to TC'XO)

Power Spectral Density, FFT Size: 1024 points

-20

-40 T -60-

-BoCL V) ö -100 0

-120

-140 -600

-400

-200

0 Frequency (Hz)

200

400

600

Figure 6-22 62.208 MHz Sampling Clock Phase Noise (VC,VO T(XO Ext. Standard)

6.3.7

Under-Sampling Performance - VHF/UHF Applications

The wideband digital receiver design is flexible and capable of wider application. The basic receiver was designed to allow signals within the full analogue bandwidth of the ADC, -250 MHz, to be utilised. The large analogue bandwidth makes it suitable for be digital Hence could the wideband under-sampling applications. receiver architecture In to frequency higher meet final in VHF order IF/digital the radios. used as and stages the Nyquist criteria the signal applied to the receiver must be externally band-pass filtered to ensure that it is not wider than Fs/2 (i. e. -31 MHz).

The under-sampling

filter harmonic (with the pre-ADC performance of the prototype receiver was evaluated removed) using a -15 dBm 200 MHz input tone and found to be close to the expected

141

Chatter 6 Performance of a Prototype Direct Sampling Digital HF Receiti er

142

(see Figure 6-23). Note that when performance theoretical under-sampling an input impact (jitter) increases the of phase noise signal with frequency (as discussed in 5.8.3). This manifests itself as a reduction in SNR. Power Spectral

Density,

0

FFT Size: 1024 points

-20

m

ö a)

-40

-60

-80

-100 a -120hyý

-140 -6000

-4000

-2000

0 Frequency

Figure 6-23 Receiver Under-sampling

6.3.8

Discussion of Prototype

2000

4000

6000

(Hz)

Performance: -15 dBm input at 200 AIH:

Direct Sampling HF Receiver

Performance The prototype receiver performance was found to accord closely to that predicted in all respects

(measured

as

NF=17 dB,

IP3 =+20 dBm,

IP2[, =+27 dBm,

BDR(3 kHz)=l 10 dB, IDR=115 dB). The receiver sensitivity, BDR and 3`d order IMD are all good and directly comparable with good conventional narrowband receivers. The IDR is excellent and typically

35 dB better than a conventional narrowband

is linear the implies of particular through IDR, receiver, receiver. gain which importance where absolute signal strength is important (e.g. in a channel sounder) or 2nd the where AGC is undesirable. However, as the performance predictions showed, in the receiver RF prototype IMD harmonic the used order amplifier and performance of fell significantly short of optimal for good strong signal handling. The performance directly the to is but related is achieved not a fundamental limitation of the architecture RF amplifier used.

142

Chapter6 PerformanceOLY PrototypeDirect SamplingDigital HF Receiver 6.4

Improving

the Performance

of the Prototype

143

Receiver

The following paragraphs consider how the performance of the digital receiver ma\ be improved.

6.4.1

Rejection of Sub-HF Frequencies

Numerous broadcasters transmit high power signals at long-wave and medium-ýýave frequencies (LF and MF). Further, the noise floor at LF/MF is significantly higher than For frequencies. intended for higher a receiver solely HF operation it would be useful at to include pre-selection filtering to reject these lower frequencies as it would reduce It is worthy of note that the prototype inherently harmonic and IMD. provides due filter to poorer and RF amplifier matching at these out-of-band additional rejection frequencies.

6.4.2

Use of a Higher Performance RF Amplifier

It is clear from both the analysis and measurement results presented that the direct is limited by linearity the the sampling receiver of prototype performance of the RF amplifier that was used.

A search of the literature and specifications for

has information high intercept on a point amplifiers provided commercially available be devices (or amplifier could a suitable number of suitable with sufficient expertise designed and build). Alternative RF amplifier configurations such as those utilising a higher operating voltage and so providing a greater linear voltage swing range, balanced push-pull designs (with inherent second order harmonic cancellation) and the use of linearisation techniques (e.g. feed-forward error cancellation) techniques are all potentially applicable. Calculations were presented in the previous chapter, assuming the use of a type QB 101 RF amplifier [Remec, 117]. The performance of this amplifier is compared with that of be it indicates This that 6-4. would the SNA-586 used in the prototype receiver in Table HF digital direct sampling possibleto construct a much higher performance whole-band deficiency the of the IMD major receiver and address the second order performance, prototype receiver.

143

Chapter 6 Performance of'a Prototype Direct Sampling Digital HF Receiver

Parameter

RF Amplifier

RF Amplifier

SNA-586

QB101

Noise Figure

5.5 dB

4.5 dB

Gain

20.5 dB

21.9 dB

Output Pl dB

+21 dBm

+31 dBm

IP1)r,

+17 dBm

+32 dBm

IP21N

+26 dBm

+83 dBm

Supply voltage

5V

24 V

Supply Current

85 mA

420 mA

fable

6.4.3

b-4

KH Amplifier

144

l'erforntunce

( 'huracteristics

Compared

Benefit of ADCs with Higher Sample Rates

The performance of the ADC has a major impact on the performance of a digital receiver. In the time since the prototype receiver was built incremental technology developments have led to the availability of ADCs with similar dynamic performance to the converter used but in higher speed grades. [Analog, 125], a derivative

The AD6645

from Analog Devices

of the AD6644, allows operation at up to 105 MSPS. The

benefit higher in direct is HF of employing sample rates a sampling receiver principal that it lessens the transition requirement in the front-end anti-aliasing filter.

Whereas

the prototype receiver allowed operation up to 28 MHz at full performance using a band higher MSPS ADC coverage and whole sampling rate would allow a -62 likely This filter design the allow a reduced order would significantly ease requirement. filter to be used and thus make it easier to achieve good pass-band ripple and matching (VSWR).

6.4.4

Improving

Digital

Down-Converter

Dynamic Range

It was previously noted that, once digitised, signal processing can be undertaken with an in GC4014 the prototype level fidelity. the The used of arbitrary performance of DDC ASIC More SFDR. dB) recent receiver provides >102 dB (typ. 105-110 implementations are able to provide slightly improved performance: >1 15 dB SFDR (e.g. [Graychip, 126]). They also have greater output word lengths (24-bit) reducing the need for gain adjustment in operation.

More latterly it has been shown that the

FPGAs becoming is implementation of DDCs with similar performance practical within

144

Chapter 6 Performance of'a Protolupe Direct SamplingDgital

HF Receiver

145

Given the performance of available ADCs, ý\ ith this level of DDC limit the the performance of a digital receiver. will not performance [Walke, 127].

6.5

Chapter Summary

A prototype wideband direct sampling digital HF receiver has been constructed and its The (measured results measured. achieved as NF=17 dB, IP3IN=+20 dBm. performance IP2IN=+27dBm, BDR(3 kHz)=110 dB and IDR=115 dB) were found to agree closely for design. In in ith the the they those predicted/specified general are accordance XN with for high HF in (discussed Chapter 5). Whilst the performance receivers requirements have implementation been improved in a number of respectsas could prototype receiver 2"d IMD it demonstrated for first (particularly that, the order performance) proposed direct-sampling digital HF receiver is a practical high time, a performance wideband, proposition. The following chapter presents the design of a wideband digital HF transceiver (HF in discussed direct HF incorporates this the sampling receiver software radio) which 10 9 Chapter Chapter digital HF transmitter and exciter. chapter and a complementary discussan application of the wideband digital transceiver as an HF channel sounder and presentresults from on-air measurements.

145

Chapter7A Wideband, Multi-Channel, HF SoftwareRadio

146

Chapter 7.

A Wideband, Multi-Channel,

HF Software Radio

This chapter describes a wideband, multi-channel,

direct sampling digital

HF

transceiver that has been designed and constructed'. It has been specifically designed as defined highly software re-configurable, a radio system. The previous chapter discussed the technical performance of its direct sampling digital HF receiver and presented measurement results. This chapter describes the overall transceiver design for which the digital transceiver system was designed include use as a multi-channel HF radio modem and as a platform on ý\hich to

and implementation.

Applications

implement a flexible HF channel sounder to allow the characterisation of the HF latter is described This in Chapter 9. use environment.

The implementation of such a complex system is a very significant piece of work. The hardware design of the digital transceiver is the author's own work. However, a number of others made designed CRC Bova Mike and In of significant contributions to realising a working system. particular, implemented the bus arbitration and local bus control logic in a CPLD. As part of this work he also implemented a number of software routines to permit communications with the board. Once the CRC the routing Huynh Minh had identified undertook key of author the placement of components, Also, a number of underand placement of the transceiver PCB under the author's supervision. in detailed the as direction graduate students under the author's made useful contributions acknowledgementsat the start of this thesis.

146

Canter 7A Wideband, Multi-Channel, HF Software Radio

,-" -V

1-

II

ýe

'Aar

yr `.

,i

AIM

t ak,

*_rr

Y pr

ýý

aA ä

""

. irl

Figure --1 Prototllpe (.Sc'ricll A'O.001) Digital I ranscewer wim J}ucIl

('

ufw'

J-) 'J)

147

Chanter 7A Wideband, Multi-Channel, HF Software Radio

7.1

148

Wideband Digital Transceiver Architecture

The digital transceiver hardware architecture (illustrated in Figure 7-2) has been designed to allow its functionality to be very largely defined through the download of PLD The has been designed as a and configurations. transceiver software application full length peripheral component interconnect (PCI) card which can be hosted in a (PC). Application computer personal software may be downloaded from conventional the host to the processors within the DSP sub-system. All interaction '\ ith the principal digital (e. transmitters large and g. receivers) (100,000 gate) occurs via a peripherals RAM based field programmable gate array (FPGA, [Brown, 128]). This device, termed be FPGA, can used to perform additional high speed or time critical the processing data hardware. It in be for to also allows paths configured as required any processing particular application.

The host can download new configurations to this FPGA to

implement application specific functionality. The principal features of the digital transceiver are summarised below and described in following in detail the sections: more 0A 0

PCI interface to allow the transceiver to be installed in a host PC; An architecture employing a local address/data bus with bus arbitration and multiple bus mastering;

0A 0A

high performance DSP sub-system module installed on a mezzanine site; flexible

architecture allowing

software configuration

and download of

application software; "A

low for phasefrequency standard sub-system responsible generating stable, frequencies; reference noise sampling clocks and other

"

4-Channel digital HF receiver with diversity RF input;

"

4-Channel digital HF transmitter exciter;

0

Digital interfaces including synchronous and

"

Built-in self test diagnostic capabilities; and

"A

interfaces; data asynchronous serial

filter front-end module. separate protection and

148

Chapter 7A Wideband, Multi-Channel, HF Software Radio

149

ää°°n

äLýCNN

ýYcrrr yjy-Eäa o ÜI

N

yjN

rL/1

(Il

o0 ýn

uNNqNcUSQ qAyJ

fA

CL

N

äa

Ää

Cq `N NR'e

EN

f

=` N°

A

'1

F

ýa

ý`ý

ý

Medea

`° Cý

ýýör_

lbw rn'uo-ze)ý, e iý

mmm

ÜU

c rv (II dR9

a f% me

m

-YqzjLý

F

`rvc

ANC

m

rI

n-iL Eý

-

ý_

°JVrIb __

i.

I (ý9 aörn _o PY

oieyg

IU

__ L

ö

OWQ

ö3c Uý 'ý §P

aN

ýR ö

N

Paey8

Ný

,dUK -

c1°ä

U

-1

4, rýý . mý

mvýýý$q 5 -'s

öa-ý

cmýýn LL wg-

a° ý

ýO ýSw

Ea ö<

TESTA EMULATION

VO PORT DATA

DATA

7

ADDR DATA

A"

PROGRAM SEUSQUENCER

IOD

PM ADDRESS B

as

1

EXTERNAL

Ion

PORT

17

22

ADDR BUS MUD

DIA ADDRESS BUS

MULTIPROCESSOR PM DATA BUS

L

INTERFACE

4E

BUS

CONNECT Px

.

DATA BUS MUx

DMDATABUS 40/32

HOST PORT

DATA REGISTER FILE fAULT1GLIEB

'a

'Op REGISTERS INEfgRY

DMA CONTROLLER

YAPPED,

SERIAL PORTS

ao81T SHIFTER

6

121

6

LIIIK PORTS

36

CONTROL. i

STATUS DATA BUFFERS

4

161

1/0 PROCESSOR

a. ADSP-2106x Super Harvard Architecture 00000 0000

0.00.00000

IOP REGISTERS 00002 0000

INTERNAL MEMORY SPACE

BANN

0 mso

NORMAL WORD ADDRESSING

DRAM

M004

0000

1OPTIONAL)

SHORT WORD ADDRESSING 0.00060000 INTERNAL MEMORY SPACE OF ADSP-2106 WITH ID=001

BANK

I

00010 0000 INTERNAL MEMORY SPACE OF ADSP-2106v WITH ID=010 DYOOl6 0000 INTERNAL MEMORY SPACE OF ADSP-2106Y WITH ID-011

MOa0DUO MEMORY INTERNAL OF ADSP-21061 WITH ID=100

MULTIPROCESSOR MEMORY SPACE

SPACE

w-z

3

M5,

EXTERNAL MEMORY SPACE

00028 0000 MEMORY INTERNAL OF ADSP-2106, WITH ID=101

BANK 2

BANK

SPACE

BANK

0x0030

0000

0.0038

0000

SIZE IS BV SELECTED MSIZE BIT FIELD OF SYSCON REGISTER

MEMORY SPACE INTERNAL OF ADSP-2106X WITH ID=110

BROADCAST

WRITE

NONBANKED

TO ALL ADSP-2106n II

OxOO3FFFFF

NORMAL WORD ADDRESSING. 32-BIT DATA WORDS 48-BIT INSTRUCTION WORDS SHORT WORD ADDRESSING 16-BIT DATA WORDS

O. FFFF FFFF

b. ADSP-2106x Memory Map

135]) [Anulog, from Architecture Figure 7-3 SHARC ADSP-2106x

7.4.2

DSP Sub-system Mezzanine Site (SHARCPAC Site)

A SHARCPAC [Bittware, 136] compatible mezzanine site is included on the transceiver board to allow the incorporation of a local, tightly coupled DSP sub-system. four have to processors. SHARCPAC modules, which are available commercially, up inclusion The a of local They are available with on-board, high speed memory. SHARCPAC compatible site allows modules with different processing/memory

153

Chapter 7A Wideband, Multi-Channel, HF Software Radio

154

installed be to meet the needs of a given application without to capabilities requiring hardware modifications to the main transceiver board. Although the SI-iARCPAC interface has been designed for use with SHARC processors it would be possible to develop a DSP sub-system module with alternative (i. e. more po\t erful) processors should the need arise. The

wideband

transceiver

board

has

been

designed with

a

split

local

(control/address/data) bus architecture so that if a SHARCPAC module is installed, it full speed even whilst other transactions are taking place on the at may operate bus local (see Figure 7-2). The two buses are only joined when transceiver's local between bus and DSP sub-system are required. The the transceiver transactions bus link function employs high speed 2-port bus switches [Pericomm, 137] in fully logic bi-directional this transceivers to as allows operation and very loýý preference dedicated for SHARCPAC judicious A latencies. the power plane site and a signal bus 5V 3.3 V DSP the the power supply voltage allows of or switch use of choice logic link board is (the transceiver the a single main change of processor modules with wholly 3.3 V).

7.5

Dual SHARC DSP Processor Module

7.5.1

Description

A DSP sub-system module (Figure 7-4). which attaches to the digital transceiver digital the developed SHARCPAC site, has been to support the anticipated uses of DSP SHARC 40 MHz dual ADSP-21060L transceiver'. This module is fitted with fast (zero 48-bit state) wait Mbits) (2 large wide shared of processorsand a quantity is DSPs SHARC for MHz the 40 The (SRAM). clock static random access memory In to SHARCPAC order board site. the from digital transceiver via the supplied a for operation) multi-processor (a reliable minimise clock skew critical requirement to generate lock this and to is signal 138] to buffer [ICS, used phase zero-delay clock the clocks

commercially four were processors two or Whilst commercial SHARPAC modules with one, for the be real-time to needed available, none had the large, high speed memory capacity expected the experience given For and this reason, implementation of applications such as the channel sounder. it SHARPAC was for site. the interface electronics that had to be gained in any case to implement the decided to also develop the DSP module described.

154

Chapter 7A Wideband, Multi-Channel, HF Software Radio

t, figure

I-"t

L/uu1

.3r1JiixL

rrucessur

iv[uuuie

155

it'im

ZIVIX-ii

murea

memOr1'

Figure 7-5 illustrates how the two SHARC processors on the processing module are interconnected and which interfaces are fed through to the digital transceiver board. Table 7-1 shows how these are routed on the digital transceiver board. LP2

LPO

LP4

LP6

LP7

LP5

LP3

LP1

Module LED I-

ItN-T,

IF

If

IF

I

L5

L4

L3

L2

I FLAGO FLAG1 ý FLAG3 ý--+

º -ý

FLAGG FLAG2 FLAG3

SPORTO

0

IRQ2

SPQR TI

IRQO

L2

IRGO

L4

L3

go L1 LO Sharc

#1

ADSP-21060 [ROO ý

interface ---

L1 'i .4 LO d Sharc

LED

FLAG1 IRO1

IRO1 FLAG1 I SPORTO Iý--ýº

ä6

L5 FLAGO FLAG2

FLAG6

FLAG3

FLAGS

FUAG7

ADSP-21060

SPORTO

IRQ2 SPQR Ti

s

IRO1

SPORT1

SPORT3

SHARC Processor Control/Address/Data Bus

Shared Memory 2Mx48 SRAM

SHARC JTAG

Interconnections Module Processing SHARC Dual Figure 7-5

155

Chapter 7A Wideband, Multi-Channel, HF Software Radio

SHARCPAC Interface

Destination on Digital Transceiver Board

LPO, LP2. LP5

Processing FPGA

LP3

Connector (for off-board connections)

LP I, LP4, LP6, LP7

Not used.

SPORTO, SPORT3

Processing FPGA

SPORTI

Connector (for off-board connections)

FLAGO, 1,3,5,6,7

I arte

7.5.2

Module Control

156

Processing FPGA and LEDs

IRQO

Bus Arbitration

IRQI

Processing FPGA

-I

CPLD, Processing FPGA

Dual 51JARC' DSP Sub-Si'stem Module Interfiices

Interface

The module control interface (MCI) is defined within the SHARCPAC specification [Bittware, 136] to allow a carrier board to determine the functionality of the module installed and then to control it. (MID)

number',

interrupts,

The MCI controls the multiprocessor identification

and resets of each SHARC

DSP.

Interrupts can also be

generated and masked using the MCI for debugging purposes. The boot mode of all non-primary

SHARC

DSPs is controlled

by the MCI.

Access to an EEPROM

containing information on the connections and functionality of the SHARCPAC module is gained through a register in the MCI. The MCI has been implemented as a set of eight is digital bus its in CPLD (MCI) Altera transceiver the which registers a small on own CPLD). bus digital interact (via to the arbitration able accessand receiver with

7.6

Digital Transceiver

Configuration

Download Software and

The digital transceiver configuration hardware architecture is illustrated in Figure 7-6. lt provides a number of mechanisms by which the system's PLDs, DSP processorsand memory can be configured.

I (from a cluster is allocated to each SHARC processor within

A multiprocessor identification (MID) to 6) and is used for inter-processor communications.

156

Chapter 7A Wideband, Multi-Channel, HF Software Radio

157

D FPG< JTAG JTAGChz,

n Bus Arbitration CPLD

Frequency Standard CPLD

EPM7512

EPM7064

L

No.

00es FPGA

PROM EPC2

F10K10 0A

FPGA Download

Bus

PCI Bridge PCI Bus PLX9054 4MB FLASH E2PROM Local Bus AddressiData Bus Link

SHARC JTAG Chain

Site)

DSP Processor #1

DSP Processor

SHARC ADSP21060

SHARC ADSP21060

#2

JTAG Port

higure /-6 Digital

7.6.1

Proces sing Site (SHARCPAC

DSP Sub-System Address/Data Bus

Configuring

I ransceiver Configuration

Transceiver

Programmable

Architecture (Simplifies! )

Logic Devices

There are three Altera PLDs in the digital transceiver and an EPC2 PROM [Altera. 1391 that, if fitted, can configure the processing FPGA from a pre-stored bit stream. These devices are all connected in series on a single JTAG [IEEE, 134] programming chain. This is the primary (and only) means of programming frequency standard CPLD and the serial PROM.

the bus arbitration CPLD, the

The EPC2 PROM was included in the

design as a back-up for the software download interface and to allow the transceiver to be configured for use without a host PC. However, the primary means of programming the processing FPGA arbitration

CPLD

is via a memory

allowing

a host

implemented interface mapped

initiated

software

download

of

bus the via the

FPGA

configuration file.

7.6.2

Software Download (DSP Sub-system)

The primary means of downloading programs and data to the DSP sub-system software is via the host which can directly access the shared memory, and in the case of SHARC the programming A of secondary means processors,the internal processor memory. (E2PROM) would PROM which is MByte processors via a4 serial electrically erasable DSP The PC. subfor a of be digital outside use allow the transceiver to pre-configured testing development in and be the used systemalso has its own JTAG chain which can of software. 157

Chapter 7A Wideband, Multi-Channel, HF Software Radio

58

Transceiver Digital Interfaces

7.7

This section summarises the external digital interfaces available from the digital transceiver platform.

Link Ports / High Speed Serial Ports

7.7.1

The following SHARC compatible high speed interfaces are brought to connectors on the transceiver board: LPO, LP1, LP2 - These interfaces which are directly connected to the processing FPGA and are nominally intended to be used as additional SHARC processorLink Ports. They may also be re-configured as SHARC compatible serial ports or for input/output. user-defined "

SHARC LP3 - This provides a link port interface which is directly routed to the SHARCPAC site. If required this link port may either be routed to another board increase high back data between LPO-2 looped to the to number of speed paths or the SHARC processors in the DSP sub-system.

"

SHARC SP1 - This provides a SHARC serial port interface which is directly routed to the SHARCPAC site.

7.7.2

Auxiliary

Digital I/O

Three un-committed digital logic lines are routed from the processing FPGA to a is lines to that is It defined. used these is signal one of connector and their use user in input the sounder channel synchronisation provide a one-pulse-per-second application described in a later chapter.

7.7.3

Serial Interfaces

The wideband transceiver includes three onboard "

interfaces: serial communications

serial RS-232 synchronous/asynchronous RS232 compatible Controller Serial Universal Integrated interface using a highly configurable Zilog

Sync/Async

the included wideband is to allow interface [Zilog, 140]. The synchronous serial data be as It bed. configured can test developed be transceiver to as a modem

158

Chanter 7A Wideband, Multi-Channel, HF Software Radio

159

(DCE) equipment allowing standard synchronous serial data communications terminal equipment (DTE) to be connected to it directly. "

Async RS-232 - RS-232 compatible asynchronous serial interface utilising a National Semiconductors Universal Asynchronous Receiver/Transmitter (UART) type PC 16550 [National, 141].

"

RS-485 - RS-485 compatible bi-directional differential signalling interface which header (via links) be to either utilise the PC16550 UART or a and configured can to interface to the FLEX FPGA for custom UART designs. The differential interface is included to allow the wideband transceiver to serial asynchronous directly control other equipments such as power amplifiers and antennas. Such an developed been digital has Harris to the transceiver to allow control a application Inc. HF power amplifier and antenna tuning unit (ATU).

This work has been

described by Chau [Chau, 142].

Frequency Standard Sub-System

Receiver RF Front-End Channel o

FRED STD InlOut

Receiver CH.B RF Input

Receiver RF Front-End Channel A

Receiver CH.A RF Input Dither Generator Tx%RxCommon RF In/Out Transmitter Exciter RF Sub-System

PTT

WWI

lillilllll

Figure 7-7 Photograph

of'Digital

I ransrel%'er to

159

Chapter 7A Widehand, Multi-Channel HF Software Radio

7.8

Frequency Standard Sub-System

7.8.1

General Description

160

The frequency standard sub-system generates the loý%phase noise (loý% jitter) sampling by digital the transmitter and diversity digital receivers. It also clocks required sources TTL is MHz 10 clock which used to generate the clocks used throughout the a transceiver's digital processing sections. A sampling clock frequency of 62.208 MHz is it is because close to the maximum rate supported by the ADCs. DDC and DUC chosen

integer is it multiple of the most popular modem symbol rates (e.g. 75. common a and 2400, l 6k, 28k8) and the 10 MHz reference. A block diagram of the frequency standard is Figure 7-8. given at sub-system Trim TCXO (PLL override)

Loop Filter

DAC

40 kHz _

10 MHz TCXO

40kHz PFD

by 250

10 MHz External Reference

Divide by 625

SEL

16 kHz PFD

Variable Divide

External Reference 1,5 or 10 MHz

44

Standard Erd External 10 MHz In

Only

Filte r

62208 MHz VCXO

62 208 MHz Output

t6z

Divide

by 3888 10 MHz Output 1

-001-

Figure 7-8 Block Diagram of'Digital

Transceiver Frequenci' Standard Si, b-,wstem

The sampling clocks are generated by a high quality 62.208 MHz voltage controlled have designed to phase a The (VCXO). are generated clocks crystal oscillator sampling Hz 100 dBc/Hz and Hz 10 at better offset, than -70 at noise of -120 The performance. dBc/Hz kHz the receiver required >_1 with commensurate at -155 is the frequency normally VCXO is phase locked to a more accurate which reference has (TCXO) a which on-board 10 MHz temperature compensated crystal oscillator dBc/Hz

frequency accuracy of better than ±4.6 parts-per-million (PPM).

The TCXO. which is

160

Chapter 7A Wideband, Multi-Channel HF Software Radio

161

frequency reference, can be phase locked to an external 1,5 or 10 MHz the transceiver frequency standard (nominally a0 dBm 5052 sinusoidal source). Alternatively the TCXO frequency may be trimmed using an on-board, software controlled. DAC to an better 0.02 PPM. The than transceiver has a single frequency standard of accuracy input/output port which can be software configured to output the TCXO 10 MHz Figure 7-9 shows the graphical user interface (GUI) for user frequency In the standard. of addition to controlling the system the GUI also control displays the lock status of the VCXO and TCXO phase locked loops (PLLs) and frequency. reference

is whether an external reference present. Frequency Standard TC m

Tire

Figure 9-6 Pulse Compression Waveform Performance tieiric. s . The Doppler frequency range for a CIR measurement is also determined by the frame being PRF is it ±(PRF/2), the the reciprocal of the PRI. The presenceof larger period; Doppler frequencies than this range will cause frequency aliasing that cannot be directly delay (i. The Doppler time) multipath e. unambiguous and ranges that can be resolved. directly therefore, are, related. simultaneously measured The Doppler resolution of a measured channel scattering function is determined by the total measurement time. Where the required output of the CIR measurement mode is the channel scattering function, consideration must be given to the useful Doppler in 173] has been [Zuckerman, is It that order to that shown resolution actually obtained. in measurethe spectral content of a stationary random process, our case a nominally Gaussianfading channel, with a given frequency resolution accuracy (i. e. to obtain a chosen degree of statistical stability

in the results) the measurement time must

This can to significantly exceed the minimum time required make a single observation. be resolved by averaging a number of scattering function measurements (incoherent integration). Unfortunately, no signal processing gain is realised and all phase information is lost. Alternatively the measurement time can be significantly increased Contiguous Fourier transform. in each which increases the number of frequency bins bins can be averaged to reduce the data to the useful Doppler resolution. for a The CIR measurement processing gain over a single pulse channel estimate, by: is generalpulse compression waveform, approximately given (NsEQ PG(dB) =10 log, 0 x Frames)

(9-7)

211

Chapter 9 Application of'Digital Radio to HF Channel Characterisation

"12

where N,,,, -,

is the number of symbols in the waveform: and

Frames

is the number of CIRs processed.

In frequency dispersive channels the achieved processing gain will reduce as a function is length. Doppler the reduced as coherence across shift waveform of

9.4.2

Requirements

for Wideband

Before selecting a sounding waveform be to established. need

Mid-Latitude

Measurements

the basic measurement ranges and resolutions

The values in Table 9-1 have been chosen «ith

a basic

be investigated: HF to the paths of understanding Parameter

Value

Transmission bandwidth

80 kHz

Multipath

15 ms

range

Delay Time Resolution Useful Doppler Range Doppler Resolution Dynamic Range

9.4.3

Design of Wideband

40 dB Table 9-1 Required CURMeasurement Pertormancc'

Sounding Waveform

The principal pulse compression sounding waveforms that have been used in WHISPER are bi-phase shift keyed (BPSK)

modulated

maximal

length pseudo noise sequences

(PN-sequences). These exist for all sequence lengths 2"-l

(m>l, mEN).

have the special property that when they are correlated cyclically, is equal to the sequence length (2"-1)

These codes

their correlation peak

9-7). Figure (see and the side-lobes are all -1

PN-sequences Binary dB. may 20log(2m-2) dynamic range of giving a useful nominal feedback taps be generated economically using a clocked shift register with a number of

[Skolnik, 160].

212

Chatter 9 Application of Digital Radio to HF Channel Characterisation

r igure Y-

213

, vT (-nip r1v-, )equence Periodic Autocourelation Function

Transmitting them back-to-back

is also useful in that it allows a 100% transmit dut`

signal power (some codes require gaps between sequences equal to

cycle, maximising

the sequence length

to preserve

their

good

correlation

properties).

Given

the

bandwidths and delay time ranges of interest PN-sequences of lengths 511 and 1023 are appropriate.

BPSK was chosen for a number of reasons. It is simply implemented at

complex baseband in both transmitter

and receiver, provides the maximum 'distance'

between different symbols (thus providing good performance in the presence of Doppler and amplitude perturbations)

and it is a nominally

constant envelope signalling scheme

which maximises the transmitted signal power. Table 9-2 summarises the structure and

characteristics of a number of sounding waveforms (including one variant compatible for comparison). sounder

with the narrowband DAMSON

In each case the chip rate is

bandwidth. the of sounding -80% Waveform

Chip-rate

(BPSK)

PN-1023 PN-1023 PN-511 Barker-13

Delay Range

(kchip/s)

(ms)

81

12.6

61.4 40.8 2.4

16.6 12.5 12.5

Multipath Resolution

(µs) -10 -15 -35 -600

Doppler Resolution

Measure Time

Processing Gain

(Hz)

(s)

(peak, dB)

±40

8192

0.01

103

70

±30

8192

0.008

136

70

±40

8192

0.01

102

67

±40

128

0.6

1.6

32

Doppler Range

No. of CIRs

(Hz)

Table 9-2 Characteristics of Various wt1IM'LK, )ounamg wave/ur!r') If a digital sequence is transmitted with no pulse shaping through a band-limited in time the filter ringing channel such as a radio system the steep edges will cause in domain. This is observed as inter-symbol interference on the received signal which b} is using a rinse This turn results in poor peak-sidelobe performance. overcome

213

Chapter 9 Application of Digital Radio to HF Channel Characterisation

214

is filter both to which chosen maximise the obtainable peak-to-sidelobe ratio shaping inevitable broadening the of the correlation peak. This problem is and yet minimise for the to of windows selection spectral analysis using the Fourier transform and, in akin fact, the same windowing functions are applicable [Harris, 175]. Ho\\ e\ er, in this be the coefficients window must used as the coefficients in a transversal application The time domain windowing function was selected after simulation of the band-defining filters transceiver generation, and signal detection signal sequence digital For filters, this the transceiver purpose processing. which are a combination of

filter.

digital finite impulse response (FIR) and cascaded integrator-comb (CIO

filters

[Hogenaur, 115], were modelled using a high order FIR f ilter ýýith the appropriate cut9-8). (Figure frequency off

0

-20-

-40-

-60-

-80-

-100-

-1201 01234567 Frequency (Hz)

x 10'

Figure 9-8 Simulated Radio Filters (80 kHz Complex Basebana)

A number of standard window

functions were synthesised and the resulting

PN1023 for sounding a 9-9 the output simulation performanceexamined. Figure shows Gaussian and window, a (rectangular with window), waveform with no pulse shaping filter 50 FIR five tap the with a 5-tap, 50 dB Chebychev window. It was concluded that An delay time sidelobes. dB Dolph-Chebychev window coefficients gave acceptable to transversal equaliser is to use a alternative approach to using fixed pulse shaping back-to-back in a minimise the error between the transmitted signal and that received calibration. The formulation is shown in Appendix F. 214

Chapter 9 Application of Digital Radio to HF Channel Characterisation

215

0 tll ýýII

-

-10

Rectangular Gaussian Chebychev

-20 ,

IM -30

1

ICI

Ill E.

0

a

-40

-50

-600.5

0.3

-0.4

-02

0 01 -01 Delay Time (ms)

02

03

04

0.5

Figure 9-9 64 kchip/s P N-1023 Pulse Compression Waveförm in 80 kHz Channel As the sounding waveform

is to be used to characterise a channel that is subject to both

time and frequency perturbations

it is necessary to ensure that the chosen sounding

sequence will perform adequately over the expected operating range. This was verified by generating the signal's specified delay time. Doppler offset

correlation

function

with frequency shifted replicas at a

A plot of the PN-1023 correlation

is presented

in Figure

function as a function of

9-10 and Figure 9-1I

(close in).

These

demonstrate how, as the Doppler frequency increases the correlation peak-to-sidelobe ratio, and to some extent the correlation

less Doppler For than decreases. offsets peak,

±10 Hz the peak to sidelobe ratio is adequate (always better than 45 dB) and the

(Figure Doppler its dB is 2 conditions correlation peak power under zero of within 9-12). Scattering function measurements can easily be corrected for the decrease in correlation peak during post processing.

215

Chapter 9 Application of Digital Radio to HF Channel Characterisation

-116

10

N figure

v-1 u tsana-limited

04 KcNrp/s

? JVI Ulf

waveform

OR

versus

Doppler

Offset

0

70

Figure 9-11 Band-limited

6-1 kchip/s PN1023 Waveform CIR versus Doppler Offset (close in)

216

Chapter 9 Application of Digital Radio to HF Channel C'hurcuctc ri cation

21

60 55 Co ö

50

ö 45 a) Q 40 320 -15

-10

05 -5 Frequency Offset (Hz)

10

15

20

-10

05 -5 Frequency Offset (Hz)

10

15

20

0 m

-0.5 0 a

-2 -20

-15

Figure 9-12 Performance of as a Function of Frequency Offset

9.4.4

Practical Waveform Implementation

Issues

The sounding waveforms are designed to be implemented on the WHISPER sounder file description To digital transceiver platform. generate a waveform which utilises the the following additional steps are undertaken. A series of waveform transmit samples (including etc) by shaping pulse all three are obtained waveforms repeated generating and the centre section captured. This ensures that the transmitted waveform samples can be repeatedwithout discontinuity to provide the ideal periodic correlation properties that gives good peak-sidelobe performance. For practical implementation reasonsthe is length that an exact waveforms are also re-sampled at design time to give a waveform integer sub-multiple of the digital transceiver clock frequency (62.208 MHz). This DDC for and the be receiver decimation integer chosen ensuresthat an exact rate can data in that there are an exact number of CIR measurements a second which allows collection at the receiver to be started on any second boundary (accurately synchronised implemented. been have WCFs that using GPS). Table 9-3 details the measurement

2»

Chapter 9 Application of Digital Radio to HF Channel Characterisation

Waveform

Chip rate

Filename

(kchip/s)

,tx 1023-81r'

Sequence Length

Sample Rate

81

1023

129.600

16-10

1x1023-64r'

64

1023

77,760

1296

1x511-32r'

32

511

48,000

600

`barker-l3'

2.4

13

2.4

13

1uare i-. ) implementation

9.5

218

Receive Digital

Re-Sampled Length

of WHISPER Sounding lýý71ejoýms

Signal Processing

Received signals are down-converted flexibility To maximise receiver.

baseband by the wideband digital

to complex

the data is saved to disk as complex sample pairs at

this stage. CIR measurements are calculated off-line signal against a template of the transmitted waveform.

by cross correlating the received Initially a scattering function is

first few from the seconds of data (256 CIRs) to allow the overall multipath generated interest be determined. Doppler to of range and

Subsequently the entire measurement is

processed to determine the sampled time, sampled delay CIRs, q (4r, 4t), over the delay range of interest (reducing the size of the data set by typically windowed and FFT'ed

to obtain

the scattering

function,

a factor of 2). This is

S(d rAt),

before Doppler

filtering (removing all frequencies beyond that of the received signals) to improve the signal to noise ratio and further decrease the size of the data set. The decimated data may then be inverse FFT'ed to return to the time domain representation, 0, (4zit).

9.5.1

Use of Windows

in Calculating

the Scattering Function

The scattering function is estimated by calculating the Fourier transform of a time series

is Fourier CIR The delay transform a periodic time, of z measurements at a specific is If the series analysed not periodic processover the series of samples transformed. features of be that may mask over this period unwanted sidelobes will generated interest. Window functions can be used to force periodicity (by multiplying the series to be transformed by a function that tapers to zero at either end). The result of using a in the 175]. [Harris, calculating number of windows, with different characteristics is trade-off a The function utilised have been scattering window examined. (principally) between the depth of the sideband suppression and broadening of the main different a on lobe. Figure 9-13 windows by a number of shows the losses imparted

218

Chapter 9 Application of'Digital Radio to HF Channel Characterisation

'719

back-to-back Where data is presented in this thesis Hanning measurement. periodic a is window utilised unless otherwise stated.

0 77 Rectangular Nanning Dolph-Chebychev X60 dBi Dolph-Chebychev i80 dBi

10

-20 -30-

-40Ca -50 ö

+ ý.

-60 li -70

-80 90 r

S, -100 -10

-8

-6

02468 -2 Doppler Frequency (Hz)

-4

10

Figure 9-13 Use of Windows in Calculating the Scattering Function

9.6

Laboratory

Measurements

The WHISPER sounder transmitter back configuration

via

to Verify Sounder Performance

back-toin a connected and receiver system were

a variable

attenuator

measurements to verify the sounder's performance. measurements made using the tx1023-64r

in

order

(effectively

make some control

The results presented here are from

waveform

(64 kchip/s PN-1023 sequence):

Figure 9-14 shows the spectrum of the sounding waveform. back-to-back scattering function

to

the ambiguity

Figure 9-15 is a plot of the function).

dBc be as expected. time waveform seen at -55 side-lobes can clearly dynamic range of the digital

The sounding The excellent

the surface the of be rest across witnessed receiver can

by limited the plot (actually chosen dB is instantaneous dynamic >90 the where range floor rather than the instrument performance). Figure 9-16 shows the achieved time sequence. the sounding of the resolution of the sounder compared with autocorrelation 15 better is than µs lt shows the time resolution that can be obtained from the sounder bandwidth. kHz 80 in an at -3 dB and -30 us at -30 dB for this 64 kchip/s waveform

219

Chapter 9 Application of Digital Radio to HF Channel Characterisation

220

Figure 9-17 demonstrates the achieved Doppler performance of the sounder for 512 CIR Hanning a using window. measurements Power Spectral Density: 10.000 MHz, 2048-pt FFT

0 -10 CD m -20 N

-30

03

°

-40

ti

`) 0- -50 U) -60 CL

-70 -80 -40

-20

-30

0 10 -10 Frequency (kHz)

20

30

40

Figure 9-1-1 WHISPER Occupied Bandwidth (Waveform: tx1023-64r)

U 20 (1)

m -40 0 -60

-60

ppp-

0 DopplerFrequency (Hz)

5 20

"0

10

Delay Time (ms)

..,

n

Figure 9-15 WHISPER Back-to-back Test: Measurea Ainvi L1Iiº 220

Digital Radio Application 9 to HF Channel Characterisation of Chanter

221

0 -

-10

ACF Transmit Waveform CIR Back-to-Back PF Test

-20 0) M ai

-30

ö

0 -40

-50

-60

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

08

1

Delay Time (ms) Figure 9-16 WHISPER Back-to-Back

Test: Complex Impulse Response Resolution

0 -10 -20 - -30 m-40 -50 -60 -70 -80 90

0

-20

10 0 -10 Doppler Frequency (Hz)

20

30

Window) Hanning CIRs, Figure 9-17 Back-to-Back Test: Doppler Resolution (512

221

Chanter 9 Application of Digital Radio to HF Channel C'ha acta 150110/7

9.7

>

Suggestions for Improvements to the WHISPER Sounder

The following suggestions are made as to enhancements to WHISPER that would be beneficial.

9.7.1

Reduced Power Spectral Density Waveforms

The current sounding waveforms

are cyclically

repeated PN-sequences. Ho ý\ e\ er. this

be that the power spectrum means will a comb with peaks spaced at the PRF periodicity (Figure 9-18). The wideband nature of this sounder means that it often has to operate as (on is termed secondary a user a non-interference what

basis to other users).

By

known (to pseudo random modulating a code apply a 0° or 180° phase change applying to each sounding waveform repeat) it would be possible to spread the energy and reduce

the waveforms power spectral density (PSD) further. Given good synchronisation and knowledge of the code, this is a process that is reversible at the receiver. A randomising 8192 PRIs that could provide up to an additional 40 dB reduction in every repeated code PSD.

0 -1018

-20-30

Q -40 -50

N a Q-60J-70-

-80 -900

-150

-100

50 0 -50 Frequency (Hz)

100

150

200

' Waveform tower )pecirum Measured 'tx1023-6-Ir Figure 9-18 Centre of (PRF- 62.5 H::)

ýY)

Chapter 9 Application of Digital Radio to HF Channel Characterisation

9.7.2

Implementation

223

of a Chirp Sounder Capability

When making HF channel measurements it is very useful to have access to real-time ionogramsto allow interesting propagation conditions to be identified and im estigated. To increase the utility of the WHISPER sounder it would be possible to exploit the digital the transceiver platform to implement nature of multi-channel a second sounder that could operate in parallel with the primary pulse compression mode. In particular it implement be to useful particularly would a Chirp sounding function. The principal be implementing would a sweeping LO. This could be done by implementing challenge logic in the digital transceiver processing FPGA code in one of two «tays. The DDC NCO could be programmed at a rate sufficiently high to cause it to appear to sweep it Alternatively would be possible for the FPGA to re-program the DDC coherently. NCO to step in frequency every N samples and then for a final stage complex frequency translation to be implemented in software after the DDC.

9.8

Chapter Summary

This chapter has documented

the development

ionospheric sounder capable of undertaking time varying

complex

impulse

response.

sounder based on software radio techniques.

of WHISPER,

a new oblique HF

wideband measurements of the channel This is a low power pulse-compression It has been implemented as an application

on the wideband HF digital transceiver described in Chapter 7. The design of sounding waveforms suitable for an investigation over mid-latitude

Skywave propagation

back-to-back measurements sounder implementation

of wideband (-80 kHz) channel characteristics been has presented. channels

in the laboratory

and quantified

have confirmed

its achieved performance.

Results of RF

the veracity

of the

The next chapter

presents results and analysis of wideband measurements made on a 170 km path in the UK during spring 2001.

223

Chapter 10 Measurement of the Wideband HF Channel using WHISPER

224

Chapter 10.

Measurement of the Wideband HF Channel using WHISPER

Considerable effort has been expended in recent years to greatly increase HF data rates for beyond line of sight (BLOS) communications as the demand for improý ed throughput to support user applications grows. For the most part work has concentrated in increasing throughput conventional narrow-band HF channels (e.g. [Jorgenson, on 176]), but some researchers (e.g. [Elvy, 177]) have also investigated using much larger bandwidths. Further, whilst wideband HF propagation has been investigated previousl` (e.g. [Wagner, 178], [Vogler, 43]), there is still no accepted channel model that is able to describe both the large scale features (such as multiple modes, Doppler shift, Doppler for modem designers, the detailed time varying properties spread,etc) and, significantly of thesechannels such as inter- and intra- mode fading statistics [Sudworth, 179]. A new wideband

HF

sounder,

known

as WHISPER,

developed

specifically

to

investigate the fine structure of wideband HF channels was introduced in the previous chapter. This chapter details a series of wideband measurements made on a 170 km mid-latitude path using this sounder to verify its performance. be able to investigate

[Watterson,

whether the Watterson

A particular aim was to 12] uncorrelated Raleigh

fading model often used to represent narrowband (>>ýýrmuýýc Characterisation C of u Appendix ýý(*;,,, -r,f/ona/ HF Receiver

Signal Generator #1

a

4 6 dB COE r> Ü

Signal Generator #2

'

2-30 MHz RF Input

Racal HF Receiver

14 MHz IF Output

FFT Analysis

6dB

Figure U-4 Cxperimenrur t- onligurculon to :ºieasure Keceiver Blocking Dm. numic Range for different test the The result of

interferer powers is reproduced in Figure C-5. The

be 14 dB figure (see above) which is a noise floor to measured was RA3701 noise of dBm/Hz. As can be seen the receiver starts to be de-sensitised ý\hen the interferer -160 increase in dBrn level and any unwanted signal this of above power a \aloe -12 reaches in Hence [3[)R is in the sensitivity. reduction receiver commensurate a results dBm dBm) is between BUR difference I1I dB (the and dB/Hz which a of 148 -12 -160

in a standard 3 kHz bandwidth. good sensitivity

(noise floor

It is therefore concluded that a digital receiver«ith

< -155 dBm) and a zero attenuation clipping level of'

blocking (i. have dBm performance. comparable) e. acceptable an will around -10

274

C Characterisation of a High Pet fugý>>,,, Co ýýýPower Spectral Density, FFT Sze

1024 points

P°

0

-10 50 dB SINAD 2

48 dB SINAD -30

-40

a°

-50

WryMNºri

ý'

-60

60

ä

Sze 1024 pyn%s

-20-

m -30

O

Scectrai Density,

'-ý

-10,

-20

T

ý7ulHFRecaýiiýcýiý

G0

-60

-70

1 Cl-

a

-80

-70 -80

-90 400 -600

400

-200

a. -110 dBm10 MHz wanted

0

200

400

-90 -600

600

Frequency (Hz)

signal, no interferer

b.

-000

-200

0 Frequency (l

200

400

)

Soo

-110 dBm10 MHz wanted signal, -10 dBm 25 MHz inte serer

PowerSpectralDensity,FFT Size: 1024points

PowerSpectralDensityFFTS¢e 1024 pwnts

'OF

-10

30 dB SINAD

38 dB SINAD --20

m

m

ý. , u1'ri*"ýYI"tly"1ýi '1'ýt+

-30 40

++^{ý f.;

w

C, -00Yr.

50 -00 60 a

-60 -70 -10 -80 80 00 r Fe"""y

C.6

L-J

D1ocn[Y[grerfurrr[ur1u

-200

'"''

interferer 25 MHz 0 dBm MHz dBm10 signal, wanted c. -110 rlgure

apt

1VWWUJUrcu

0 Frequency (Ftr)

200

400

600

d. -110 dBm10 MHz wanted signal, +10 dBm 25 MHz interferer u[

1. -t IVLlL.

IF

vulpU1

of

LCHJ

v1

AeLe(Ver

Instantaneous Dynamic Range

The receiver's instantaneous dynamic range (IDR) was characterisedby measuringthe tone input from on-channel a single the of power signal the receiver as output power the The dBm. with 0 made dBm were between measurements and was varied -120 for IF settings gain IF in repeated and mode gain receiver a manually controlled IDR indicate of C-6) a maximum between25 and 250 (no units). The results (Figure dB (depending on the gain setting). -75

27;

ix C Characterisation of a (High Pertormance ('onl,

e,.

2h

0 -10

-20 ^, -30 a, co -40 0 CL -50 a 0 -60

-70

-80

-90 ---120

-100

-80

-60 Input Power (dBm)

-40

-20

0

Figure C-6 Measured Instantaneous Dynamic Range (IDR) of RA3 -01 Receiver /r a Range of IF Gain Se'tili7 \

C.7

Image Rejection

The RA3701's

ability

to reject a strong out-of-band signal at the first mixer image

frequency was measured.

In this receiver the 1Stmixer image rejection is all due to the

receiver's front end filtering experimental configuration

ahead of the mixer. The measurement was made using the in Figure C-4. It was found that with the receiver tuned to

10 MHz, a dBm interferer applied at 92.7995 MHz produces an image signal in the -8 receiver IF pass band having an equivalent signal strength to a wanted -118 dBm signal at 10.001 MHz.

This is illustrated

in Figure C-7.

Note that, due to the receiýer's So, the receiver front end

frequency plan the 1.4 MHz

IF is inverted in frequency).

filter provides an effective

110 dB of rejection (the difference between the signal

powers).

dB 110 of filter that provides digital a receiver with an anti-aliasing the least as good as rejection in the second Nyquist zone will provide performanceat frequencies image high quality of Additional attenuation commercial receiver tested. It frequency response. in installation's will, many cases, be provided by the HF antenna design filter is concluded that the low anti-aliasing pass the performance required of It is concluded that

, 76

Characterisation High C Performance of a Conventional .. deceiver ndix n/' to equivalent achieve performance to a good conventional required recei\er is less challenging than may have initially been somewhat anticipated. Power Spectral Density, FFT Size, 1024 paints --

0

-_' ------------

1 a¢e

1024 pones

-118 dBm Wanted Signal

-10

-8 dBm Tage S 3ýa -20 m m

m

ýi%: -30

ä

r*K

ýlyýý l '4M

'o SO

-60 -

a

-70

2000

a.

-1500

0 500 -500 Frequency (Hz)

-1000

1000

1500

AO -2000

2000

dB 10.001 MHz wanted signal, no interferer -118

b

1,

1500

-1000

-500

-118 dB 10.001 MHz wanted signal -8 dBm 92 7995 MHz interferer

Figure C-7 Measured Irrage Rejection of

C.8

X50:

43-01 Receiver 1" 1lixer" .

Summary of Super-heterodyne Performance

The performance of the RA3701

front is RF end summarised in Table ('-I. receiver

The measured results are compared with the published specification I Racal, 1041. Parameter

Specification (RF Pre-amp off)

Specification

Measured

(RF Pre-amp on)

(RF Pre-amp off)

Noise figure

IP3JN IP21N Blocking Dynamic Range InstantaneousDv namic Range Image Rejection Reciprocal Mixing (1" LO Phasenoise)

1-I dB

IU dB

14 dB

+32 dBm

+'_SdBm

+30 dBm

Not specified

Not specified

+57dBm

Not specified

Not specified

113dB (3 kHz) 148dB (I Hz)

Not specified

Not specified

75 dB

>90 dB

>90 dB

110dB

ii 20 kHz dBc -96 offset 106 dBc sir 80 kHz offset

-96

dBc ii 20 kHz

d Not measure

offset dBc at 80 kHz 106 offset

Table C-1 Racal RA3701 High I'erjormance iii ýýý«-

1 ,,

Noise Figure Lineari System 1ý and ý >>encü.

Intermodulation

Appendix D. System Noise Figure and Linearity (Intermodulation)

, -ý _ý

D System Noise Figure and Lineari ýi

Introduction

D, 1

This appendix provides a summary of a number of key conceptsand equations relied body in the thesis the of relating to the sensitivity (noise floor) upon and linearit\ of devicesand systems.

Thermal Noise Power and Noise Figure

D,2

Thermal noise is generated as a result of the thermally excited random motion of free in It a conducting medium. can be shown [e.g. Stremler. 61 electrons p199] that an\ have device floor will a noise which cannot be less than that due to practical the in thermal noise power generated a matched voltage source. This thermal equivalent

is density, by: given power noise Thermal Noise Power Density =101og10(kTO)

dBIV/ 11_

(D-1)

where k

is Boltzmann's constant (1.38x10-23J/K); and

To

is the system temperature in degreesKelvin.

Any practical device will generate a higher level of noise than this. For convenience this additional noise is related to the thermal noise power at the device input b\ the noisefigure, NF: NF=1+T

(D- )

T.

where T e

is the effective noise temperature of the device.

NF is unit-less and commonly expressed in decibels.

D.3

Devices Cascaded System Noise Figure - Noise Figure of

ý\hole a as figure system of a It is useful to be able to calculate the composite noise noise The cascaded D-1). (Figure from the parameters it devices up making of the figure is given by:

ý1 ^ ýý

D System Noise Figure and Linearity (Inte

NF

NF, + = ,()l

NF2 -1 G

+

NF3-1 G, G

I

1

ý +F

where NF

is the noise figure of the nth cascadeddevice: and

(;,,

is the gain of the n'th cascadeddevice.

Note that this calculation must use the linear noise figure and gain values. A general in is that most practical systems the impact of the earliest observation stagesoil o\erall is figure the greatest. noise

IP3,,,,

º

Stage1º

Stage2º

Stage3-------

NF,

AT

G,

G,

G,

1P31

1P3,

IP;:

StageNº

11'?.

Figure D-1 Noise Figure and Intercept Point of a Si°stemof Cascaded/Devices

D.4

Harmonic

Intermodulation and

Characteristics of non-Linear

Devices Signalprocessing devices, such as amplifiers and mixers, are non-linearto at least, Once degreeand may in many cases be characterised as having a transfer function of the form: K, x(t)+KxI y(t) =

(t)+K3x`(t)+...

Kx" (1)

(D-1)

form: the If sucha device is excited by two signals simultaneously of 0) A, t + A, cos((O, cos((t),t) + x(t) =

(D-5

of The significant most thenan output will be generated including all possible products. identifies): theseare (obtained by expansion using trigonometric O)ý [A, i A, + cos(&, K, cos(co,t) +

M-61

which representthe first order terms,

280

ix D System N

Linearitv (Inte i

NF(Y)7= NF +

NFZ

NF3 XF -1 -1 + 7G, -+... G, G; G, G

-1 G,

, ýI)-.,

where NF

is the noise figure of the n'th cascadeddevice: and

G

is the gain of the n'th cascadeddevice.

must use the linear noise figure and gain \alues. A general in most practical systems the impact is that observation of the earliest stageson o\erall is figure the greatest. noise

Note that this calculation

º

Stage 1

º

Stage 2

1P3,,,.

º

Stage 3

AT,

NE.

AF,

C

C,

C,

1P3,

1P32

IP3

--_

____

StageN

\F, IP3

Figure D-1 Noise Figure and Intercept Point of cr Svstemof CaccadedDc'vccs ,

D.4

Harmonic

Intermodulation and

Characteristics of non-Linear

Devices Signal processing devices, such as amplifiers and mixers, are non-linear to at least some degree and may in many cases be characterised as having a transfer function of the form:

' K,, (t) K3x' (t) K, (t) K, +... x, + x' x(t) + y(t) = form: the If sucha device is excited by two signals simultaneously of 0) A, t + cos((t), x(t) = A, cos((i), t) +

(D-5)

of The significant most products.

thenan output will be generated including all possible identifies): theseare (obtained by expansion using trigonometric O)] [A, t + A, cos((0, K, cos(w, t) +

M

which representthe first order terms,

280

D System Noise Figure and Lineari -lJ)t)endix 'ýl

K2

'2 cos(2co, i)

K, A2 cos(2wzt + 20)

2z

(D-1

K2A, Az cos[w,t

-w2(t+q$)]

K2A1A, cos[w,t+wz(t+O)] the second order terms, and represent which 3K3A 4

cos(3w1t)

3K3AZ 4

cos(3wzt) (D-8)

3K3A12A2 cos[2w,t±co 2(t+o)1 4 3K3A1A;

4

cos[2w2(t + 0) ±w t]

the most significant third order terms. In many cases the remaining 3`dand higher order

termsare negligible and may be ignored. It can immediately be seen from the above that as the power of the input signalsis increasedthe resulting harmonics and intermodulation distortion (IMDs) productswill have different slopes according to their order. Secondorder productsfollow a square law increasing in power by 2 dB per dB increase in the input power. Third orderterms increaseby 3 dB/dB.

Hence the concept of an intercept point can be evoked to

characterisethe power level at which the intermodulation productsmll theoretically havethe same power as the fundamental outputs (Figure D-2). The interceptpoint can being intercept be referenced to the devices input or output, the output point merely factoredby the device power gain. This is a point that cannotusually be reachedasa device will normally go into compression first (where the fundamentaloutputsno is measured longerlinearly increase commonly Intercept point input the signal). with to a applied are tones. form power two, using a two tone test. In its simplest same intercept order Then. the n'th systemand the resulting IMD products are measured. ]: 101 point(referencedto the input) can readily be calculated [Kundert. sI

m Noise Figure and Linearit

Intermodulation ý`,

IMD IPn, = PIN+n_1 N

dBm

,Iv,

where PIN

is the input power of each of the two tones (dBm ); and

IMD,

is the power of an n'th order IMD product relative to a fundamental (dBc).

(P1) is defined dB point l as the point at which the output is 1 dB The compression for linear operation. that to expected reduced intercept directly being measurable, points provide a keß measureof the Despite not device Given figure. device. the free noise the dNnamic a of spurious performance be is SFDR defined free the calculated. (SFDR) readily as can spurious signal range 3`d floor intermodulationproducts the to the from at point which the order noise range floor. exceedthe noise

ýý, -ý

Appendix D System Noise Figuiv

2nd Order Intercept Point (Slope: 2 dB/dB) 3rd Order Intercept Point (Slope: 3 dB/dB) 1 dB Compression Point

1 dB

Output Saturation

a-

O Fundamental (Slope: 1 dB/dB)

Input Power (dBm)

Figure D-2 Response of a non-Linear Device showing Compression and Extrapolated Intercept Points

D.5

System Intercept Point - Cascaded Non-Linear Devices

The intercept point of a system due to that of the cascadeddevicesmakingit up (Figure D-1) is given by: G2Gi

I=1+G, "INI

'

IPIN2

+

IP/N3

G3G2Gj + IPIN4 +...

(D-10)

TOT

where "IN,

is the input intercept point of the n'th cascadeddevice;and

G

is the power gain of the n'th cascadeddevice.

2S

Figure and Linearity (Intermodulation) Noise System D Appendix

,, ý

linear for the terms to intercept is it use gain and necessary Note that point (ratherthan in A the equation. above is general decibel observation in equivalent) that , stems their later has the bi 1MD of stages impact can the performance e the t iý