Details and Issues of Implementing HF Channel Simulators Which Process Baseband Audio Signals

1
Details and Issues of Implementing HF Channel
Simulators Which Process Baseband Audio Signals

• John Nieto
• Harris Corporation
• RF Communications Division

2
Presentation Overview

• Motivation
• Channel Simulator Elements
• Summary / Conclusions

3
Motivation

• In order to reduce the difference in performance
  measurements of HF modems using channel
  simulators implemented by different
  organizations, a very detailed channel simulator
  specification must be developed
• Many presentations have focused on the fading
  characteristics of HF channel simulators, but
  even without fading, some channel simulators have
  not yielded same results

4
Channel Simulator Elements
Frequency Offset
Fading Process
Power Calibration
Frequency Offset
Fading Process
A/D
Tx Filter
Hilbert
Delay
D/A
AGC
Rx Filter

CCIR 322 Noise
Jammer Suite
Gaussian Noise
Core Functions
Added Features
5
Channel Simulator Elements

• A/D (Analog to Digital conversion)
• Sampling rate of channel simulator gt 8000
  samples per second (higher is better and provides
  more multipath resolution, but more MIPS
  required)
• Channel simulator A/D should provide 12-16 bits
  of resolution
• Baseband signal level at input to A/D should be
  as high as possible without saturating A/D
• Channel simulator should provide warnings when
  signal is too high or too low into A/D to avoid
  invalid tests
• A/D sample clock stability should be lt /- 500
  PPM

6
Channel Simulator Elements

• TX Filter
• Simulates TX radio filter
• A frequency mask must be provided so that radio
  filters designed for different channel simulators
  provide same results (see STANAG 4203)
• Can be source of significant performance
  difference if filter not designed properly

7
Channel Simulator Elements

• Hilbert Filter
• Used to generate complex signal from a real
  signal
• Since a finite filter response is necessary, the
  parameters of the filter should be provided by
  channel simulator specification
• Passband of filter
• 300 Hz to (Fs/2 300) Hz
• RMS of filter
• 40 dB or higher
• Filter length should be even numbers to reduce
  MIPS (i.e. every other tap is 0)

8
Channel Simulator Elements

• Watterson Model used to generate fading and
  multipath
• Fading Process generated by filtering complex
  Gaussian noise through a Gaussian shaped filter
  where Doppler spread is equal to twice the
  standard deviation (2?).
• Multipath modeled by tap-delay line
• All paths should be normalized by summing the
  desired powers for each path and setting total
  power equal to 1.0 (for proper SNR calibration)
• For additional details, please view previous HFIA
  presentations on this topic

9
Channel Simulator Elements

• Gaussian Noise
• A test should be required to verify proper noise
  properties
• For example, with channel simulator outputting
  only noise (no TX/RX filters), generate
  Probability Density Function (PDF) of noise and
  compare to theoretical Gaussian shape
• Tails of distribution are important when
  measuring small BER points

10
Channel Simulator Elements

• Power Calibration
• In order for a valid Signal to Noise Ratio (SNR)
  measurement, channel simulator must be calibrated
• Input to channel simulator must be measured for a
  long enough period of time to guarantee a
  repeatable and representative signal power
  measurement (maybe 5 minutes)
• Measurement must account for losses through
  filters and other channel simulator elements to
  avoid improper calibration
• Calibration should be done prior to test period

11
Channel Simulator Elements

• RX Filter
• Simulates RX radio filter
• A frequency mask must be provided so that radio
  filters designed for different channel simulators
  provide same results (see STANAG 4203)
• Can be source of significant performance
  difference if filter not designed properly

12
Channel Simulator Elements

• D/A (Digital to Analog conversion)
• Sampling rate of channel simulator gt 8000
  samples per second (higher is better and provides
  more multipath resolution, but more MIPS
  required)
• Channel simulator D/A should provide 12-16 bits
  of resolution
• A/D sample clock stability should be lt /- 500
  PPM
• Output of channel simulator must be properly set
  to avoid underflows/overflows prior to D/A
• When jammers are added, great care is required to
  avoid overflows/underflows prior to D/A

13
Channel Simulator Elements

• Additional elements of a channel simulator not
  discussed in this presentation
• Frequency Offset
• Frequency rate of change
• CCIR 322 Noise
• AGC
• Jammers

14
Summary / Conclusions

• Channel simulator specification should describe
  minimum requirements for
• Resolution of A/D and D/A
• Sample rates
• Clock stability
• Frequency masks for TX, RX and Hilbert filters
• Specification should include values of parameters
  required by most design filter programs
• SNR Power calibration
• Noise process (including test procedure)
• Fading process (including test procedure)

15
Summary / Conclusions

• To avoid multiple organizations developing
  identical channel simulator elements, maybe a
  library of functions should be created and
  maintained by HFIA
• After proper testing, library can be provided to
  all members of HFIA for use in their channel
  simulators, etc