RFI Identification and Mitigation Using …RFI Identification and Mitigation Using Simultaneous Dual...
Transcript of RFI Identification and Mitigation Using …RFI Identification and Mitigation Using Simultaneous Dual...
RFI Identification and Mitigation Using Simultaneous Dual Station Observations
Ramesh BhatMIT Haystack Observatory
J. Cordes (NAIC/Cornell), S. Chatterjee (NRAO), J. Lazio (NRL)
Arecibo-Greenbank ObservationsGoals of the observations:Demonstrator observations for SKA-like instrumentsDevelopment of RFI ident’n and mitigation algorithmsSome science (eg. Giant pulse studies)
Data taken on:The Crab PulsarM33 (several beams)UGC galaxies
ARECIBO Multi-WAPP 4 x 100 MHz
1120-1520 MHz
GREENBANKSPIGOT800 MHz
1100-1900 MHz
Identical resolutions in time, freq.
T_samp=82 us, deltaF=0.78 MHz
Identification and Mitigation of RFI: Techniques & Applications
• Filtering, Thresholding & Flagging• Spectroscopy (e.g. HI emission from galaxies)• Time series (e.g. Giant pulse detection)• Transient seeking:
– Matched filtering– “friends-of-friends”
No magic bullets or a universal method!
Data in the time-frequency plane -- i.e. Dynamic spectra
Fast-sampled (e.g. 10 – 100 us), high spectral resolution (~1 K channels)
RFI Environment: L-band data
WAPP and SPIGOT bands: short integrations (deltaT= 1 sec, deltaF=0.78 MHz)
RFI Environment: L-band data
WAPP and SPIGOT bands: short integrations (deltaT= 1 sec, deltaF=0.78 MHz)
RFI Identification: Dual-station data
RFI Identification and Rejection SchemeRaw Data (time-frequency plane)
Apply 2-d filtering to capture RFIAverage all the
data, interpolate where necessary
Filtering and thresholding(1d and 2d) in stages: 1st, 2nd, and 3rd
RFI Identification: Filtering Scheme
Apply a 2-d median filtering to the time-freq data (sequence of 1-s spectra)
RFI Identification: Filtering &Thresholding
RFI Excision: Thresholding & Flaging
RFI Excision: clean and flag’d data
21cm line emission from the Galaxy 1420.4 MHz
H I emission from UGC 2339, exp’dat ~1346.5 MHz
RFI Identification: dual-site data
GREENBANK
ARECIBO
RFI Identification: Time Series
CRAB Pulsar
GBT/SPIGOT
L-band Data
(1100-1900 MHz)
Impulsive RFI or giant pulses?
Transient “event” detection schemeDynamic spectra data
Full resolution data in time and frequency
~100 us, ~1 K chans
Dedispersion: single DM or many DMsTime series hanning, smoothing (optionally), thresholding, grouping
Compare the event lists from two sites, or two sub-bands; Examine the t-fsignature, strength
Dispersion: Pulsed signals
Dedispersion
single DM (for known objects)
many trial DMs (GP search)
undedispersed (DM=0) data
Dedispersed time series of the Crab pulsar
Giant pulse from the Crab pulsar (Arecibo)
Transient (“Events”) Detection Scheme
RFI?
Giant Pulses
GBT Crab pulsar data (L band)
Characterize “events” by- signal strength- location- spatial extent
RFI vs “real” events: Filtering Schemeusing multiple sub-bands
Bursty, strong RFI can potentially mask even the strongest giant pulses!
Use of multiple sub-bands to filter out RFI-rich vs. “clean” bands
No. of events detected is a strong function of RFI-contamination in the band
4 WAPPs, 1120-1520 MHz, 100 MHz chunks
Detection of “real” events : sub-band approach
S1 AND S2
S1 AND S3
S1 AND S4
S1
S1
S2
S3
S4
Uncorrelated events? vs. effects of ISS?
Detection of real “events”: Dual-station approach
GBT/SPIGOT
AO/WAPPsExtend the analysis to ‘event lists’ from the two sites (between the respective sub-bands, or more complex criteria)
Application to M33 data
ARECIBO (1320-1420 MHz)
GREENBANK (1300-1400 MHz)
GBT beam ~3 times AO beam
AO gain ~ 5 times GBT gain
Data taken several beam areas on M33
Dedispersion/event detection over a range of DMs
Exp’d DM toward M33 ~ 60-80 pc/cc (NE2001)
Possible giant-pulse detetctions from Arecibo 430 MHz observations (McLaughlin & Cordes 2003)
Slide courtesy: K. Becker
Slide courtesy: K. Becker
Summary
• RFI identification/mitigation in 2d t-f data• Applications to data from AO and the GBT • Different techniques for RFI identification• Filtering, Thresholding, flag’ng schemes for
fast-sampled, high res’n data in time-freq. • Transient “event” detection schemes • Dual-site, sub-bands approach to filter out “real”
events from impulsive RFI in time series data• Still work in progress…..