Design and Demonstration of an Interference Suppressing Microwave Radiometer

1
Design and Demonstration of an Interference
Suppressing Microwave Radiometer

• IGARSS 2004
• Frequency Allocations for Remote Sensing
• Joel T. Johnson, Grant A. Hampson,
• Steven W. Ellingson,

2
Motivation

• Traditional radiometer architecture poor at
  rejecting RFI
• Low-level RFI problematic in post-processing
  difficult to distinguish from geophysical
  information
• High amplitude but low duty cycle pulsed RFI (for
  example, microsecond radar pulses out of
  millisecond integration period)
• can appear as low-level RFI
• Similarly, strong amplitude CW interferers can
  appear as low-level RFI
• RFI localized in time and/or frequency can
  potentially be suppressed by simple
  time/frequency blanking methods
• Traditional architecture can be retained by
  sampling data stream faster (0.1 to 1 msec) and
  adding analog sub-band channels increases data
  rate post-processing RFI removal, but can only
  go so far.
• Since 2002, a digital receiver based radiometer
  has been under development at Ohio State to
  implement such methods in real-time

3
Outline

• System design
• Implemented L-band prototype
• Local experiments
• Water pool observation
• Radio astronomy observations

4
Design Concept

• Traditional direct-detection radiometer
• New design
• Try to remove RFI in real time clean data can
  still be integrated to retain low data rate

Antenna
Filter
LNA
ADC
Digital Hardware
RFI
RFI Suppression/ Filtering/ Detection/Integration
ADC
Antenna
Filter
LNA
Downconvert
5
Design including RFI Removal Stages
(DIF)
Low-noise front end
Analog Downconverter
Digital Downconverter
Antenna
ADC
(APB)
(FFT)
Asynchronous Pulse Blanker
1024 point FFT
Frequency domain blanker
(not yet implemented)
(SDP)
Data Recording/ Control
Detection/ Integration
6
APB algorithm

• APB estimates mean/variance of incoming time
  domain signal a sample gt b standard deviations
  above the mean triggers blanker
• Pre-detection samples can be blanked by including
  memory in the system, NWAIT parameter sets time
  period
• Blanked samples replaced with zero calibration
  effects can be corrected by scaling average power
  appropriately
• Some FFT issues, but tests show minor

7
Frequency Domain Blanking

• Post-FFT, two types of blanking can be considered
• Time blanking of each FFT bin
• Similar to original APB, but now at higher S/N
• Implementation very similar to time-domain APB
• Cross-frequency blanking
• Requires some information on expected instrument
  passband
• Can look for rapid changes in spectrum to
  indicate narrow-band RFI
• Can also permanently blank certain bins known to
  contain RFI (for example hydrogen line emissions
  at L-band)
• Again calibration effects can be corrected by
  keeping track of the number of blanked samples
• Rapid frequency domain blanking of type 2
  perhaps not required, since narrowband
  interferers vary slowly still reduces data rate
  though

8
Outline

• System design
• Implemented prototype
• L-band local experiments
• Water pool observation
• Radio astronomy observations

9
Digital Back-End

• Prototype samples 100 MHz, includes Digital IF
  downconverter (DIF), asynchronous pulse blanker
  (APB), FFT stage, and SDP operations
• Implemented in FPGAs for algorithm flexibility
• Altera "Stratix" parts apprx 10000 LE, 260
  each
• A final prototype has been designed to combine
  processor components into one Stratix FPGA apprx
  30000 LE, 950
• Microcontroller interface via ethernet for
  setting on-chip parameters
• Possible modes
• Direct capture of time domain data, sampled every
  10 nsec
• Integration, blanker on/off, integration lengths
  0.01 to 21 msec
• Max-hold, blanker on/off

10
Three FPGA Prototype

• Modular form used for processor boards note
  microcontrollers
• EEPROM's on each card for autoprogramming of
  FPGA's on power-up

11
Outline

• System design
• Implemented prototype
• L-band local experiments
• Water pool observation
• Radio astronomy observations

12
L-band Antenna/Front End Unit

• Front end Tsys approx. 200K neglecting antenna

Temperature control is critical to maintain
internal standards rest of system
not temperature controlled
13
L-band Dual Channel Downconverter

• One channel is 1325-1375 MHz, other is
  1375-1425 MHz
• Downconverter, digital receiver, computer, and
  thermal control systems in rack inside lab
• High-compression point amplifiers used isolators
  used to reduce channel coupling

14
Terminator Test of System Stability
Terminator Spectra
After ND Stabilization
0.25 dB
15 hrs
-0.25 dB
Sensitivity vs. Integration
Total Power vs. Time
15
Water Pool Observations

• Experiments designed to demonstrate radiometric
  accuracy in the presence of interference
• Observations of a large water tank external cal
  sources are ambient absorbers and a sky reflector
• Highly accurate ground-based radiometry is tough
  due to contributions from objects not under view,
  including reflections
• Keep cal targets exactly the same size as pool to
  reduce these effects observations of pool as
  ambient temp varies also
• Initial tests in existing RFI, incl. air traffic
  control radar at 1331 MHz

16
Pool and Cal Targets
Still working toward obtaining absolutely
calibrated data Can still examine
effectiveness of blanking strategies in
uncalibrated data
17
Relative Power Variations Pool Observation
Blanker Off H pol
Blanker On H pol
240 secs
Noise Generator
Terminator
18
Sky Observations

• An alternate experiment was initiated using
  observations of the sky
• a 3 m reflector was available used same
  feed/front end
• Sky observations at declination angles up to 30
  degrees
• Expect to see cold sky plus astronomical sources
  minor atmospheric influence
• Potential for using
• cold sky plus moon
• in a calibration
• Initial results use
• software FFTs
• and integration
• low duty cycle as a
• result
• 24 hour observations
• of astronomical
• sources

19
Sky Observation Results Blanker on

• Software FFTs allow very high spectral
  resolution (0.4 kHz) sufficient to observe
  Doppler shift of neutral Hydrogen line

Hydrogen line emission around 1420 MHz
S-curve is due to Doppler shift associated
with galactic region observed
Elapsed Time (Hr)
Moon
20
Relative Power Variations Sky Observation
APB Off
APB On
.25 dB
-.25 dB
Radar contributions greatly decreased by APB
21
Conclusions

• Digital receiver prototype developed and
  currently being applied in
• L-band water pool and sky observations
• Base suppression algorithm is APB, followed by
  post-processing narrow band removal at present
  can implement spectral processing in future
  hardware as well
• Current data shows qualitative success of this
  approach, although continuing to work toward a
  final demonstration
• Goal is to demonstrate well calibrated and stable
  brightness measurements even in the presence of
  RFI
• We have also deployed this backend in aircraft
  observations at C-band, subject of next talk..