mock time domain sampling of pulsars

last mod:23mar11
Sections:
11sep11 : 1713+0747  at lband, 430, 327
23mar11:1937+21 lband, 430 30 minutes each. saw pulsar.
29jan11: psr0301+19 at lband and 327
21jan11: 1937+21  lband : didn't see anything.

  Intro:

  Time domain sampling using the mock spectrometer can be done by using cima (11sep11). This was tried on various days with different pulsars to check the data.

11sep11: psr 1713+0747 at lbw,430,327 (top)

    p2627 used the mock spectrometers to take time domain data on 1713+0747 at lband,430, and 327 MHz.
Data was taken on 11,12,xx,xx sep11. The lbw data from 11sep11 is shown here as well as the 327 data.
Note: There was a problem with the data acquisition for the first few days. The mock spectrometer had been configured to byte swap the data out of the mock box for 32 bit data. This ended up switching polA and polB and well as reversing the I,Q samples. This was corrected 20sep11. I updated the idl software to check for the byteswap parameter in the header to check and correct this.

Lband data:

The setup was:

  • They started with 7 10 MHz bands. They switched to 6 bms when bm 4 had trouble:
    • bm0 = 1335 MHz
    • bm1 = 1385 MHz
    • bm2 = 1435 MHz
    • bm3 = 1485 MHz
    • bm4 = 1510 MHz
    • bm5 = 1535 MHz
    • bm6 = 1666 MHz
  • The mock spectrometers have some 60 hz and power supply birdies around DC of the digitizer. To get rid of this, we shifted the baseband input freq to the mocks up by 9.996 MHz and then used the digital mixer to bring the requested frequency back to the center of the band. So Dc of the digitizer was outside the final band.

Processing:

  • File 100 (the first file of lband on 11sep11 was processed). The za for this file was 10.6 deg.
  • the parameters used for the processing of the files was:
    • dm=15.9899
    • period:0.00457056044 secs
      • The period came from tempo using the start time of file 100
      • I folded at a number of adjacent periods and they all made the pulse wider.
    • Dispersion smearing was .288 to .56 milliseconds across the 10 Mhz band.
    • The optimum fftlen was 54 to 75 channels. I used an fftlength of 64.
      • the smearing over 1 channel was around 7 usecs
  • For each band the data was input, transformed, dedispered, and then folded into 100 phase  bins.
  • The pulses were rotated to put the pulse in the center of the period
  • bins 0..40 were used to compute the noise rms. The data was then scaled to sigmas and plotted.


    The plots show the folded data for the first 50 seconds (file .00100) (.ps) (.pdf):
  • Page 1: 1335, 1385 MHz ..
  • Page 2: 1435,1485 MHz
  • Page 3: 1510, 1535 MHz
  • Page 4: 1666:
Summary:
  • The pulsar was seen  at all bands (except 1385Mhz)
    • The peak ranged from 25 to 120 sigmas in 50 seconds of data.
  • The pulse is width pretty close to waht it's supposed to be (after i got the sign of the shifts for the dedispersion correct).
    • using a 64 pnt fft, the dispersion smearing in 1 chan is about 7 usecs.

327 data:

The setup was:

  • 4 10 MHz bands were used: 
    • bm0 = 315 MHz
    • bm1 = 325 MHz
    • bm2 = 335 MHz
    • bm3 = 345 MHz
  • To get rid of the rfi at dc, we shifted the 327 rfband down by 10 Mhz, and then shifted back up by 10Mhz.

Processing:

  • Files 0,1, and 2 were procesed. The za for these files started at za=19.6 deg.
  • the parameters used for the processing of the files was:
    • dm=15.9899
    • period:0.00457055804
      • The period came from tempo using the start time of file 000
    • Dispersion smearing was 42.67 to 32.48 millisecs across the 10 Mhz bands.
    • The optimum fftlen was 657 to 573  channels. I used an fftlength of 512.
      • the smearing over 1 channel was around 83  usecs
  • For each band the data was input, transformed, dedispered, and then folded into 100 phase  bins.
  • The pulses were rotated to put the pulse in the center of the period
  • bins 0..40 were used to compute the noise rms. The data was then scaled to sigmas and plotted.
The plots show the first 3 files:
The pulses look smeared. Don't know whether it is the dedisperion, period, or what..


processing:x101/110911/lbwfoldinp,lbwfoldplt.pro


23mar11: psr 1937+21 at lbw and 327. (top)

    Data was taken on the millisecond pulsar (1937+21) on 23mar11 7:30 AM (ast) using lband and 430 MHz.
The setup was:
  • lband:
    • linear pol, 172.032/17 MHz bandwidth, cfr=1410, 1370 hipass filter in place
    • 10 ghz upconverter used.
    • band is flipped in frequency at the a/d converter.
    • Files: p2613.20110323.b0s1g0.00000 - 00033.pdev
    • dir: /share/pdata1/pdev/
  • 430 rcvr:
    • circular pol, 172.032/17 MHz bandwidth. cfr=430, 422 to 442 filter in place
    • 750 narrow band mixer used.
    • band is not flipped at the a/d converter.
    • Files: p2613.20110323.b0s1g0.00100 - 00133.pdev
    • dir: /share/pdata1/pdev
  • Mock setup:
    • bandwidth divisor 17 --> bandwidthMhz=172.032/17.
    • 8 bits complex sampling
    • adjpower to give 350 counts for 1 sig (in the 12 bits).
    • take 30 minutes of data with lband and 30 minutes of data at 430.

Checking the data:

    50 seconds of the lband data was transformed, dedispersed, and then folded to see if there was any signal.
The parameters used were:
  • dm=71.0398
  • dmSmearing across band: (202./1410)^3*dm*bw=2.11378 millisecs
  • optimalFftLen=sqr(bwMhz^2*(202/cfrMhz)^3*dm*1e3) = 146 channels
    • used 128 channels for transforms.
  • Dedisperse using some idl code i stole from sigproc (probably still has a few bugs).
  • fold the data:
    • use 100 phase bins
    • try 200 trial periods about the period in the atnf catalog (30 years old).
      • Pick period with largest signal to noise
    • period:
      • 0.001557806472448817D secs from atnf catalog (30 years old?)
      • 0.0015577073961810D  secs (period used)
        • i folded the data trying 100 periods close to the atnf period and then used the one that maximized the signali
        • after finding the peak i retested using periodStep=.2usecs * (i=-50 to 50) and then took the max.
The plot show 50 seconds of 1937+21 transformed, de dedispersed and folded:
processing:x101/110323/foldlbw.pro, fold430.pro

29jan11: psr 0301+19 at lbw and 327. (top)

    Data was taken on 29jan11 tracking the pulsar 0301+19. It's specs are:
  • period:1.387584446262 seconds
  • dm: 15.737
  • s1400:3. mJy, s430:27 mJy.
The setup was similar to the 21jan11 except that the 12 bit rms was set higher (350 to 400) giving an 8bit rms around 20-25.

The data reduction:
  • input data and compute spectra, and dedisperse using the same idl routines at 21jan11
  • 1410MHz:dmsmearing:.47 milllisecs over 10 MHz b w, use 64 length fft.
  • 327MHz: dmsmearing 37.5 milliseconds over 10 MHz bw, use 512 length fft.
  • Input and process 1 file (52 seconds of data) for each band.
     
    The plots show the reduced data:

21jan11 1937+21 with dan stinebring: (top)

    On 21jan11 p2613 tried taking time domain data on the millisecond pulser 1937+21. Data was taken at 327, 430, and 1410 MHz. The setup procedure was:
  • use cima to setup the iflo, backend, and track the pulsar
  • Switch to the pnetcl --line program to setup the mock spectrometers for dual pol, 8bit, decimate by 17 sampling.
  • adjust the mock power for an 12bit rms of around 160 (giving an 8bit rms of about 10).
  • Take data for about 20 minutes at each band.
Nothing was seen when the first file of the 1410 MHz file was reduced (about 52 seconds of data).  The 327 and 430 data were taken with the tiedowns out of position so they weren't reduced.

Summary: (top)

  • psr0301+19 was seen after dedispersion and folding at 327 and 1410 MHz.
  • The 327 dedispersion check shows that the dedispersion code is making the pulse narrower and stronger (so it may not be correct but it is increasing the pulse strength.
  • Not sure why the 21jan11 data saw nothing?
    •  maybe we weren't tracking correctly?
    • the dedisperion code still has some bugs with millisecond pulsars and large dm?


<- page up
home_~phil