Evaluation of Binary PhaseCoded Pulse Compression Schemes Using a TimeSeries Weather Radar Simulator

1
Evaluation of Binary Phase-Coded Pulse
Compression Schemes Using a Time-Series Weather
Radar SimulatorT.A. Alberts1, P.B. Chilson1,
B.L. Cheong1, R.D. Palmer1, M. Xue1,2
33rd International Conference on Radar Meteorology

• 1 School of Meteorology, University of Oklahoma,
  Norman, OK, U.S.A
• 2 Center for Analysis and Prediction of Storms,
  Norman, OK, U.S.A.

2
Overview

• Motivation
• Pulse Compression Concepts
• Phase Coding
• Performance Metrics
• Time-Series Weather Radar Simulator Description
• Simulation Methodology
• Results
• Summary and Future Work

3
Motivation

• Current activities focus on fielding phased array
  radar (PAR) systems
• Pulse compression (PC) becomes relevant for PAR
  systems

• Research situational applicability of PC systems
  using Time-Series Weather Radar Simulator

4
Pulse Compression Concepts

• Pulse Compression used to enhance detectability
  and range resolution
• Signals encoded through modulation of the signal
  phase or frequency
• Resolution improvement determined by signal
  bandwidth
• Amplitude increase calculated by Time-Bandwidth
  (BT) product

5
Phase Coding Performance

• Distributed nature of weather implies that
  weather from other ranges will contaminate
  results of desired range
• Minimize Integrated Sidelobe Level (ISL)
• Barker Code Commonly used phase code

6
Time-Series Weather Radar Simulator (TSWRS)

• ARPS (Advanced Regional Prediction System) model
  output initializes simulator
• Tornadic supercell
• 5x4 km domain size (25 x 25 x 20 m resolution)
• 1 second data intervals
• Scatterers distributed randomly throughout
  domain
• Properties determined by position
• Interpolation with 2 data sets
• I Q time series data composed
• Covariance processing produces estimates of
  equivalent reflectivity factor, radial velocity,
  and spectral width

7
Pulse Compression Modifications to Simulator

• Phase coding applied just prior to sampling
• Composite signal summed and decoded via filter
• Decoded data processed as normal

8
Simulator Output Standard Resolution vs 13-bit PC
8 km
13 km
9
Profiles of Reflectivity and Radial Velocity

• Good agreement except where rapid changes occur
• Reduced ISL will produce more accurate results
• Simplest way is to increase code length BUT.
• Longer codes have less tolerance for velocity
  changes
• Typically overcome by using banks of filters
  tuned to different velocities (Doppler shifts)

SNR 70 dB
10
Errors in Reflectivity and Radial Velocity

• Reflectivity errors correspond to areas of high
  gradients in reflectivity
• Better ISL performance reduces this effect
• Velocity errors tend to occur where velocity
  changes are large over a short distance (phase
  shift large)
• Aliasing also a problem

SNR 70 dB
11
Code Length Effect on Improving RMSE

• Increasing code length most effective on Z
  estimates
• SNR trend due to increase of noise floor above
  signal
• Vr and SPW exhibit a weaker trend
• Large RMSE due to Vr a few, large errors

12
SNR Effect on Reflectivity Errors
30 dB
10 dB
70 dB
50 dB
13
Summary

• Binary phase coding has been successfully
  incorporated and demonstrated on the TSWRS
• Results indicate that strong gradients in
  reflectivity and velocity results in estimation
  errors as expected
• Effect of range sidelobes inherent to pulse
  compression
• Mitigation by increasing code length shown but
  benefit varies for each parameter estimate
• Currently limited to short code lengths

14
Future Work

• Incorporate sidelobe suppression filters into
  simulator
• Least-squares
• Adaptive
• Expand code suite
• Longer phase codes will require significant
  memory
• Polyphase, PRN, MLS
• Explore frequency modulation
• Overcome small domain issues to evaluate longer
  code
• Currently limited to approximately 5 km in range
• Memory intensive process
• Track performance over longer periods of time
• Total data set spans 5 minutes

15
Questions?
16
BACKUP SLIDES
17
Spectral Width Results

• Agreement generally within 1 m/s
• Large error coincides with location of tornado