Title: Analysis of Adaptive Array Algorithm Performance for Satellite Interference Cancellation in Radio Astronomy
1Analysis of Adaptive Array Algorithm Performance
for Satellite Interference Cancellation in Radio
Astronomy
- Lisha Li, Brian D. Jeffs, Andrew Poulsen, and
Karl Warnick - Brigham Young University
- XXVII URSI General Assembly 2002
2Summary
- GLONASS, Iridium, ground-based radars, etc.
create overwhelming interference in important
bands. - Adaptive beamforming/array processing algorithms
are promising for interference cancellation, but
are not well characterized in radio astronomy
environment. - Low SNR, very sparse arrays, very high gain
elements. - We study five algorithms for a small telescope
array. - A real-time MSC, LMS filter is implemented.
3The Algorithms
LCMV Linear constraint minimum
variance GSC Generalized sidelobe canceller MSC
Multiple sidelobe canceller
MSNR Max. signal to noise ratio SPSN Subspace
projection spatial nulling
4Algorithm Applicability
- Adaptive beamforming LCMV, GSC, MSC, MSNR
- For more compact arrays or sub-arrays.
- Single channel output array performs as single
high gain telescope (like GBT). - Candidate for SKA sub-arrays.
- Array nulling SPSN, modified MSC
- For large imaging arrays.
- Output is full array, usable with synthesis
correlator.
5Calibrated Array SINR Improvement Comparisons
- SINR at feeds is -60 dB, SNR -14 dB.
- Average interference reduction of 45 dB.
- Source is at zenith.
- 8 MHz processing bandwidth.
- MSC is least affected by grating lobes.
- Grating lobes are a big problem for small array.
-
-
-
-
-
-
-
-
6SINR Performance with Calibration Errors
- Circular complex Gaussian calibration error, mean
1, variance 0.01 - Relative performance among algorithms similar to
calibrated case. - Significant SINR improvements still, but larger
variation.
7LCMV Null Placement and Mainlobe Distortion
- LCMV for 3 element VSA, tight 15 ft. spacing.
- Interferer is in 3m dish mainlobe, and array
grating lobe. - Null placement and cancellation are good.
- Array mainlobe is distorted.
8A Real-time MSC, LMS Filter Adaptive Cancellation
Experiment
Interference Cancelled Signal
Primary Antenna
Periodogram PSD Estimator
dn
en
Ss(wk)
-
xn
yn
hn
Reference (Auxiliary) antenna
m
Periodogram PSD Estimator
Si(wk)
9VSA Antenna Array in Adaptive Canceling Experiment
- Reference antenna aimed at interference /
photographer. - Right dish is primary channel, aimed at
Cassiopeia.
10Real-time LMS Filter Parameters
- 13 full complex adaptive FIR filter taps.
- 500 kHz processing bandwidth, I-Q baseband.
- Complex LMS filter update algorithm.
- VSA 3m dish antennas used for signal and
reference. - Primary antenna steered to Cassiopeia.
- Reference antenna steered to roof mounted
interference source.
11Real-time LMS Filter Parameters (cont.)
- Interferer is F.M. sweep modulation,
- 100 kHz BW
- Carrier centered at 1420.66 MHz
- -62 dBm at dipole 95 ft. from receiver dish.
- 1024 bin (500 Hz per bin) periodogram spectral
estimate computed in real-time. - Integrate and download every 2s.
12Advantages of Real-time Cancellation
- Operates on raw pre-correlator sampled data.
- Can be inserted as a transparent front-end
process in an existing telescope system. - No long-duration, high data rate recording
needed. - Useful if time-sample outputs are desired, not
integrations. - Post processing adaptive filtering requires huge
data storage. - DSP hardware (programmable and FPGA) are now fast
enough to support desired bandwidths.
13VSA Test Platform Receiver
- Analog receiver (foreground)
- 4 channels
- 16 MHz bandwidth
- 59 K total system noise
- DSP array processor in (background)
- 4 channels, 65 MHz A/D.
- Digital Receiver front-end.
- 4 TMS320C6201 floating-point processors.
14DSP Detail
- 4-200 MHz processors with digital receiver
front-ends. - In real-time, performs
- Complex baseband, band select, decimate, filter.
- Two channel 1024 point periodogram and
accumulate. - 13 complex tap FIR LMS adaptive filter. (four
multiplies per tap).
15Hydrogen Line and Interference Signals
FM interference signal seen by reference antenna.
Real-time PSD Estimate
Hydrogen signal from Cassiopeia seen by primary
antenna. 28 minute PSD integration. Automatic
tracking.
16Real-time Cancellation Results
Signal and Interference Seen by the primary
antenna.
LMS MSC adaptive canceller output. Real-time
result. 30 min. integration PSD.
17Noise Floor Calibration Signal(using RF absorber
in feed)
18Conclusions
- Adaptive beamforming algorithms are promising for
compact arrays, sub-arrays, array feeds. - MSC is most robust,
- No grating lobe problems.
- Useful for both beamforming and imaging arrays.
- The real-time LMS MSC was very successful, should
work in many environments. - Next step test real-time MSC with GBT.
19Hydrogen Line Signal, Cassiopeia
20FM Interferenceat Reference Receiver
21Signal plus Interference at Primary Receiver
22Adaptive Canceller Output