Transcript Ultra-Wide-Band Ground Bounce Range Antenna

RFI Mitigation
Steve Ellingson
Virginia Polytechnic Institute & State University
“Frontiers of Astronomy with the World’s Largest Radio Telescope” Meeting
September 13, 2007
RFI Problems
• Ourselves (not the topic of this talk, though…)
• Aviation Systems (UHF, L-, S-, and X-Band)
• DME, Air Surveillance Radar, …
• Navigation Satellites (L-Band)
• GPS (Three 20 MHz channels), GLONASS (16xx MHz)
• Communication Satellites (L-, C-, and K-Band)
• Iridium (16xx MHz), INMARSAT, etc.
• TV: NTSC (analog) is becoming ATSC (digital, worse)
• Harmonic and IM products of FM, TV, mobile telephony, etc…
• “Crud”
Lines of Defense
• Regulation / Frequency Coordination
• Avoidance
Avoiding contaminated frequency bands, Scheduling to avoid
satellites
• Analog Filtering / High Dynamic Range Receivers
• “Pre-detection” Signal Processing
Excision: Detection & Blanking, Adaptive Filtering, Null Forming
Canceling: -- “look through”; e.g., Model-and-subtract methods
Real-Time
Techniques
• “Post-detection” Signal Processing
Post-correlation & Cross-spectral techniques
• Post observation: Data Editing, Anti-Coincidence
Focus of
this talk
“Emerging” Techniques
• Radar Pulse Blanking
• Actually, “old school”; e.g., Arecibo pulse-pattern-synchronous blanker
• Many others have explored this; useful to some extent right now
• Effectiveness limited by detection sensitivity, multipath spread,
“mangled” (semi-correlated multipath) pulses – room for improvement
• More General: Time-Frequency Blanking
• Useful esp. against “crud” if Dt x Dn can be 1 ms x 1 kHz or better
•
• Spatial Nulling (Arrays)
• Especially vs. satellites (ATA, some FPA concepts)
• Canceling
• “Reference antenna” and “Model & subtract” methods – even for radar
• The low frequency “renaissance” is driving the testing /
implementation of all kinds of techniques
L-Band Radar Blanking & Canceling
Real time mitigation of the notorious
1330/1350 MHz radar @ Arecibo
using a digital receiver
Ellingson & Hampson (2003), ApJS, 147, 167.
200
MSPS
A/Ds
Other good examples:
Zheng et al. (2003), AJ, 126, 1588.
Zheng et al. (2005), AJ, 129, 2933.
Dong, Jeffs & Fisher (2005), Radio Sci., 40, RS5S04.
Against mobile telephony:
I/F to
PC
SDP,
1K
Integrate FFT
I/Q Conv., LPF,
Pulse Blanker
Before: Radar pulses
corrupt spectrum
Implemented completely in
Altera Stratix FPGAs
Various papers by Leshem, van der Veen & Boonstra
After: Radar pulses excised
(~4% of the data is blanked)
•
Time window blanking is
hard to beat, if you can
tolerate the gaps and loss
of integration time
•
Typically limited by pulse
detection performance.
Mitigation of Iridum – Blanking vs. Nulling
Argus
• Array of 24 spiral antenna elements
• Tsys ~ 215 °K per element
• 1200-1700 MHz Tuning Range
Ellingson, Hampson & Childers (2007), IEEE Trans. Ant &
Prop., in press.
Top:
Middle:
Bottom:
RFI mit off
Nulling
Blanking
Detector:
Total power pulse
W = 8 ms
b = 10s at PMF output
Nulling:
Projecting out estimated
spatial signature of burst
Cancel 56 ms window
Start 16 ms before trigger
No data loss
Time
Series
Blanking:
Rank
Detector
Pulse
Detector
More info: Proc. RFI2004
http://www.ece.vt.edu/swe/rfi2004
Blank 56 ms window
Start 16 ms before trigger
~ 20% of data is blanked
Matched
Filter
Output
PSD calculation:
Dn = 100 Hz
Dt = 10 ms
 = 58.3 s
Blanking: Better Performance;
Nulling: No Data Loss.
“Model & Subtract” Method vs. GLONASS
Australia Telescope Compact
Array (ATCA)
Narrabri, NSW
Observations of OH Maser
IRAS 1731-33
Corrupted by weak (-20 dB SNR)
RFI from the Russian GLONASS
satellite system
Technique achieves > 20 dB canceling of the
GLONASS interferer as follows:
•
•
•
•
GLONASS signal is tracked (in the same
fashion as a GLONASS receiver),
Instrumental responses are adaptively
estimated
A noise-free model of the RFI is synthesized
RFI model is coherently subtracted from the
telescope output
Ellingson, Bunton, and Bell (2001), ApJS., 135, 87.
Digital TV (ATSC)
Ch 3
ATSC
MJD
Craig County VA – Oct 2005
Preliminary Work on ATSC Model & Subtract
* Work of Kyehun Lee, VT
Mitigation of Broadcast FM?
•
Strong source of RFI in 88-108 MHz (U.S.);
surrogate for a very broad class of difficult-tohandle RFI across the VHF & UHF bands
•
Bandwidth ~ 200 kHz
•
Baseband is analog audio + many other
components, total ~75 kHz: Processing gain!
•
Simpler version used to convey audio in NTSC
•
Prone to multipath; especially apparent in weak
signal areas
A Canceller for Broadcast FM
Architecture
•
Analyze band; determine
# of signals & form coarse
estimate of associated
center frequencies
•
Extract carriers one at a
time, demodulate,
estimate model
parameters
•
Reconstruct noise-free
version using extracted
model parameters
•
Subtract synthesized
carriers from telescope
output.
Estimation Block
Looks complicated, but is only slightly
more complex than a high-performance
commercial FM receiver.
This version does not account for
channel characteristics, such as
mutlipath.
* Work of Kyehun Lee, VT
Broadcast FM Canceller: Demonstration
Complete
parametric
model
Simple
“Chirp”
Model
Before / After
Simulated signal
(no channel effects)
Before / After
Off-the-air signal
(includes channel effects)
Toxicity to
simulated spectral line
([0..1]; 1 is perfectly safe)
Somewhat toxic: Current algorithm suppresses uncorrelated in-band
spectral content (i.e., underlying radio astronomy) by about 40%.
Note detailed model knowledge helps a lot with this.
Prospects good for further improvement, especially with site diversity:
Using information from sites closer to the transmitter.
* Work of Kyehun Lee, VT
For More Information

Recent summary paper for ITU on mitigation techniques (see Lewis or
Ellingson) & references

RFI2004 on-line proceedings / Radio Science Special Section
http://www.ece.vt.edu/swe/rfi2004

A.-J. Boonstra, Ph.D. Dissertation, T.U. Delft

Non-technical discussion: Ellingson (2004), “RFI Mitigation and the
SKA,” Exp. Astronomy, 17, 261. Reprinted in The Square Kilometre
Array: An Engineering Perspective, P.J. Hall (ed.), Springer, 2005.