Intro:

• 7 mocks each covering 170 Mhz.
•  centered at 4070,4220,4370,4520,4670,4820,4970.
• 51 usec spectra, 32 channels, 16 bit recording, no blanking
• Joanna pushed the data through the prsfitstofil (converting ao psrfits to sigproc filter bank dataformat).
• She then processed the data.

resolution: 18may16: dipanjan contacted me, and said that they had found a baselining problem in their processing..

The problem:

    Joanna reported dips in the folded profile before and after the pulse.
She provided a an example (.tiff)

• Pulsar B1237+25, 128 usec sampling,32 channels for each 170 Mhz band, and 7 mocks covering 4-5 ghz.

Dipanjan also sent a plot showing the problem from 18 and 20 sep2015 (.pdf)

• Page 1: B1929+10 18sep15.. black is total power.. no dips.
• Page 2: B1933+16  18sep15.. black is total power.. dips up to 10% of pulse covering up to 1/3 of the period.

Looking at the data:

• I took the data for B1933+16 on 18sep15 (this data showed dips up to 10% of the pulse stretching over 1/3 of the period.
• To process the data :
• input the psrfits spectra from the files number 200,201.( 774 seconds of data..)
• I just used my idl routines. I did not convert the data to sigproc filter band format.
  
• compute the total power in polA,B separately
•  dedisperion was not done.
• fold each total power time series.
• I took the polyco.dat file that they had on monster0 and  computed the period for the middle of the data set.
• I ended up using 0.35876777 seconds for the period, and 1000 phase bin channels (.358 ms/phasebin)
• The period varied by .03 usecs over this period.
• Using the central period should have broadened the pulse by less than 40 usecs (less than 1 phase bin channel).

The plots show the folded data (.ps)  (.pdf):

• Page 1: folded profiles for the 7 bands
• black is polA, red is polB jo
• Page 2: blowup to get a closer look at the baseline.
  
• the peak is about .04 Tsys.
• The vertical scale is +/- .001 Tsys.. or about +/-2.5% of the peak
• No dips are seen
• Page 3: over plot the 7 bands on the same plot.
• top: entire period.
• bottom: blowup  channels 200-400 (20% of the period).
• You can see the time shift because dedispersion was not done.
• the dm is about 158.. the measured shift is close to what is expected.

Results:

• No dips are seen in total power data folded (with no dedisperion).
• An individual 170 Mhz band should be smeared by less than 2ms..
  
• this is less than 1% of the period..
• the 10% smearing data seen in  joanna and dipanjan's data are not seen here..
• It looks like the data is ok.. It must be a processing problem:
• psrfitstofilt or in the processing to get the folded spectra.

processing: x101/160516/joannacbandmocks.pro