RFI and Asynchronous Pulse Blanking in the MHz Band at Arecibo

Transcription

1 RFI and Asynchronous Pulse Blanking in the MHz Band at Arecibo Steve Ellingson and Grant Hampson November, 2002 List of Figures MHz in three 50-MHz-wide swaths (APB off). The three bands observed (centered on 55 MHz, 00 MHz, and 50 MHz) are shown on the same frequency axis. In each case, the top curve is max hold (on 42 ms linear averages), and the bottom curve is linear average MHz: Mean and max-hold spectra. Top Panel: APB off; Bottom panel: APB on MHz: Time-frequency plots. Top Panel: APB off; Bottom panel: APB on MHz: Time domain total power. Top: APB off; Bottom: APB on. The cause of the glitch at the end of the APB-on trace is unknown; perhaps observer-directed calibration activity? Note: the data for each curve was collected at different times (but within a few minutes of each other) MHz: Mean and max-hold spectra. Top Panel: APB off; Bottom panel: APB on MHz: Time-frequency plots. Top Panel: APB off; Bottom panel: APB on MHz: Time domain total power. Top: APB off; Bottom: APB on. Note: the data for each curve was collected at different times (but within a few minutes of each other) MHz: Mean and max-hold spectra. Top Panel: APB off; Bottom panel: APB on MHz: Time-frequency plots. Top Panel: APB off; Bottom panel: APB on The Ohio State University, ElectroScience Laboratory, 20 Kinnear Road, Columbus, OH 43210, USA. ellingson.1@osu.edu. 1

2 MHz: Time domain total power. Top: APB off; Bottom: APB on. Note: the data for each curve was collected at different times (but within a few minutes of each other) Introduction On Nov 3, 2002, we interfaced a custom back end (developed under our NASA IIP project [1]) to the Arecibo Observatory in order to do some piggyback mode observations with J. Cordes (Cornell), R. Bhat (Haystack), and M. McLaughlin (Jodrell Bank) [2]. The purpose of this session was to record examples of RFI (radars, in particular) in astronomical data (in this case, pulsars). Both coherently-sampled time series and integrated spectra were collected. This report documents our measurements of integrated spectrum in three 50-MHz bands covering MHz. Additionally, we also tried out the IIP s asynchronous pulse blanker (APB), which is intended to mitigate pulsed RFI from radars. 2 Instrumentation Cordes et.al. were observing a 100-MHz-wide swath centered at 75 MHz using the Arecibo L-Wide receiver with the Wideband Arecibo Pulsar Processor (WAPP) back end. The WAPP accepts its input from an IF fan-out in the receiver room. This IF is MHz (approximately), wherein the 75 MHz sky frequency maps to 250 MHz (not spectrally reversed) in the IF. To interface the IIP system, another port from this IF fan-out (specifically, the front panel jack on Rack 4, Unit ) was attenuated by 24 db, upconverted using a Mini-Circuits ZP-5 mixer with an Agilent HP8648C synthesizer as an LO, and input to the IIP L-band downconverter just prior to it s image rejection filter. The LO was tuned to put the desired sky center frequency at 20 MHz in the input to the IIP system. Thus: f sky = [2345 MHz] f LO. (1) The IIP system does a low-side downconversion to an IF center frequency of 0 MHz and samples at 200 MSPS using 10 bits. (Note: One (of two) of the IIP 2

3 system s Mini-Circuits ZFL-500HLN amplifiers preceding the A/D was removed for gain leveling purposes.) Because the analog IF is in the second Nyquist zone of the A/D, the digital passband is centered at 50 MHz and is spectrally reversed. The IIP s Digital IF (DIF) FPGA module downconverts this to 0 Hz (so now the samples are complex-valued), filters to 50 MHz bandwidth, decimates by 2, and then upconverts to a center frequency of +25 MHz (still complex). The data emerges from the DIF module in -bit I + -bit Q format at 100 MSPS. The IIP system is reconfigurable in both the hardware and software sense. For the measurements presented here, the system was configured as an FFT spectrometer. In this mode, time samples from the DIF are processed through a length-1k complex FFT, which has an effective duty cycle of 19% (i.e., 19% of the data is FFTed, and the rest is lost). A triangular window is applied before the FFT. The FFT output is then processed through a spectral domain processor (SDP) FPGA module which computes magnitude-squared of each spectrum and computes a linear power average over 4096 spectra. This gives an effective integration time of ms for an actual observation time of about 221 ms. 254 of these integrated spectra are collected in a FIFO, and then the experiment terminates. The resulting data set is saved on a PC as a single 4MB file consisting of (254 integrated spectra) (1K bins) ( bits/sample), and represents about (221 ms) 254 = 56.1 s in real time, and 10.7 s in terms of data collected. All data collected is freely available; contact the authors for distribution information. An asynchronous pulse blanker (APB) [1] is also available in the IIP system. The APB module maintains a running estimate of the mean and variance of the sample magnitudes. Whenever a sample magnitude greater than a threshold number of standard deviations from the mean is detected, the APB blanks (sets to zero) a block of samples beginning from predetermined period before the triggering sample, through and hopefully including any multipath components associated with the detected pulse. We were also able to collect coherent voltage time series data by reconfiguring the IIP system firmware; see [2] for details. This data will be presented elsewhere. 3

4 Some APB operating parameters are set by the user. In this case, the following default parameters were in effect: Trigger on sample magnitudes greater than 9.5σ above than the mean. Start blanking 100 samples (1 µs) in advance of triggering sample. Blanking period is 100 µs long. Wait at least 44 µs between triggers (attempting to prevent multiple triggers on the same pulse). No attempt was made to optimize these parameters based on the observed data; in fact, we are certain that these choices are not suitable for some radars observed at Arecibo. Nevertheless, we couldn t resist the temptation to spend our last few minutes of telescope time to see what would happen if we tried using the APB with it s default parameters anyway. 3 Results A summary of observations is shown in Figure 1. Measurements were taken at three center frequencies: 55 MHz, 00 MHz, and 50 MHz. Each measurement is 50 MHz wide, and represents (as noted above) 10.7 s of integration in 42 ms segments evenly distributed over 56 s in real time. Also shown in Figure 1 is the max hold of the 42-ms spectra. The max hold is computed by taking for each frequency bin the maximum value observed in that bin over the course of the experiment. Max hold spectra are useful for revealing bursty signals (especially radars) which tend to be suppressed in deep integrations due to their low duty cycle. Figure 2 (top panel) shows the same result for the 55 MHz measurement only. A few bursty signals with bandwidth on the order of 1 MHz are evident (look for peaks in the max hold without corresponding peaks in the average); some or all of these are radars. The signal near 51 MHz is remarkable in that it is both modulated and persistent; it could be an intermodulation product associated with VHF or UHF band broadcasting. 4

5 The bottom panel of Figure 2 shows the same result taken a few minutes later, and with APB turned on. The difference is not dramatic, although a few signals have disappeared from the max hold spectrum, so it is clear that some pulses are being removed. Also, the linear average spectrum has dropped by about 0.25 db relative to the APB-off result; from this we infer that about 5% of the samples are being blanked. Figure 3 shows the same data, but now as joint time-frequency plots. Figure 4 shows total power in the 50-MHz passband as a function of time. Comparison of the APB-off and APB-on traces suggest that the APB-on trace is dominated by statistically-stationary noise; whereas the APB-off trace is dominated by something other than noise, yielding a smooth curve. This also seems to suggest that the APB is having a positive, if subtle, effect. Figures 5 7 show the same analysis applied at 00 MHz. This swath of spectrum contains a very strong radar, operating near 90 MHz. A rotation period of 10 s is evident from Figures 6 and 7. When at its strongest (presumably, when oriented toward Arecibo), the pulses are strong enough to either clip or compress (it s hard to tell which, given the available data) the instrumentation, generating harmonics which manifest themselves as signals spaced every 10 MHz across the passband. It is also not possible to tell from this data whether it is the Arecibo or the IIP receiver which has gone non-linear; it might be both. However, it is interesting to note that the APB is quite effective at removing the harmonics, even though the 90 MHz radar is not completely suppressed. In the time domain, it appears that the APB has blanked enough to noticeably reduce the noise floor during the beam on periods (look for the divets ). As before, about 5% of the samples have been blanked. Finally, we note the presence of a second strong radar (especially apparent in Figure 7) as well as another persistent signal, this time at 05 MHz. Interesting to note in Figure 6 is that the 90 MHz radar is visible in virtually every integration, regardless of whether the beam is pointed at Arecibo or not. Figures 8 10 show the same analysis applied at 50 MHz. This swath of spectrum contains the very strong San Juan ATC radar, transmitting at 30 MHz and 50 MHz. A rotation period of s is evident from Figures 6 and 7. The be- 5

6 havior is very similar to that of the 90 MHz radar described above: When this radar is at it s strongest, the pulses are strong enough to generate harmonics spaced every 10 MHz across the passband. In this case as well it appears that the APB dramatically reduces the impact of the radar, although there is still clearly room for improvement. Finally, we note the presence of another persistent signal, this time at 40 MHz. Acknowledgments Thanks to J. Cordes, R. Bhat, and M. McLaughlin for permitting us to piggyback on their observations; and to P. Perillat and L. Wray for advice and technical support. References [1] G. Hampson and S. Ellingson, Modular Digital Back Ends for Microwave Radiometry, Nov 1, swe/ska1/naic02.pdf. [2] S. Ellingson, RFI Monitoring Arecibo (TXT), Nov 3, swe/ska1/ap0203.txt. 6

7 PSD (db/100khz) Sky Frequency (MHz) Figure 1: MHz in three 50-MHz-wide swaths (APB off). The three bands observed (centered on 55 MHz, 00 MHz, and 50 MHz) are shown on the same frequency axis. In each case, the top curve is max hold (on 42 ms linear averages), and the bottom curve is linear average. 7

8 17 PSD (db/100khz) Sky Frequency (MHz) 17 PSD (db/100khz) Sky Frequency (MHz) Figure 2: MHz: Mean and max-hold spectra. Top Panel: APB off; Bottom panel: APB on. 8

9 Figure 3: MHz: Time-frequency9 plots. Top Panel: APB off; Bottom panel: APB on.

10 18 17 Total Power (linear units) Time (s) Figure 4: MHz: Time domain total power. Top: APB off; Bottom: APB on. The cause of the glitch at the end of the APB-on trace is unknown; perhaps observer-directed calibration activity? Note: the data for each curve was collected at different times (but within a few minutes of each other). 10

11 17 PSD (db/100khz) Sky Frequency (MHz) 17 PSD (db/100khz) Sky Frequency (MHz) Figure 5: MHz: Mean and max-hold spectra. Top Panel: APB off; Bottom panel: APB on.

12 Figure 6: MHz: Time-frequency plots. Top Panel: APB off; Bottom panel: APB on.

13 18 17 Total Power (linear units) Time (s) Figure 7: MHz: Time domain total power. Top: APB off; Bottom: APB on. Note: the data for each curve was collected at different times (but within a few minutes of each other).

14 17 PSD (db/100khz) Sky Frequency (MHz) 17 PSD (db/100khz) Sky Frequency (MHz) Figure 8: MHz: Mean and max-hold spectra. Top Panel: APB off; Bottom panel: APB on.

15 Figure 9: MHz: Time-frequency plots. Top Panel: APB off; Bottom panel: APB on.

16 18 17 Total Power (linear units) Time (s) Figure 10: MHz: Time domain total power. Top: APB off; Bottom: APB on. Note: the data for each curve was collected at different times (but within a few minutes of each other).