#79 digifil does not produce output of the appropriate size

[Ryan Lynch]

Description: I am trying to use digifil to produce sigproc filterbank
data from data in the GUPPI RAW format. The program runs to
completion but the output is not of the correct size and the
calculated scan length is far too long. The raw data have 256 coarse
channels and a native sampling time of 1.28 us. In the examples
below, the actual scan length should be about 60 seconds. Note that
this may be related to the problem described below.

Command: digifil -b 8 -d 4 -F 256:D -D 11.217356 -K -o guppi_58626_J1125+7819_0013.fil -t 6 ../guppi_58626_J1125+7819_0013.0000.raw

Output:

digifil: using coherent dedispersion
digifil: using convolving filterbank

Result of running 'readfile guppi_58626_J1125+7819_0013.fil':

Assuming the data is a SIGPROC filterbank file.

1: From the SIGPROC filterbank file 'guppi_58626_J1125+7819_0013.fil':

              Telescope = GBT
            Source Name = J1125+7819
        Obs Date String = 2019-05-23T11:27:53.0025
         MJD start time = 58626.47769678830809
               RA J2000 = 11:25:00.1920
         RA J2000 (deg) = 171.2508         
              Dec J2000 = 78:18:59.7600
        Dec J2000 (deg) = 78.3166          
              Tracking? = True
          Azimuth (deg) = 0
       Zenith Ang (deg) = 0
        Number of polns = 2 (summed)
     Polarization order = 
       Sample time (us) = 46.08            
     Central freq (MHz) = 1500.390625      
      Low channel (MHz) = 1400.78125       
     High channel (MHz) = 1600             
    Channel width (MHz) = 0.78125          
     Number of channels = 256
  Total Bandwidth (MHz) = 200              
                   Beam = 0 of 1
        Beam FWHM (deg) = 0.137
     Spectra per subint = 2400
        Starting subint = 1191
       Subints per file = 285015082
       Spectra per file = 27521824
  Time per subint (sec) = 0.110592
    Time per file (sec) = 1268.20564992
        bits per sample = 8
      bytes per spectra = 1024
    samples per spectra = 1024
       bytes per subint = 2457600
     samples per subint = 2457600
            zero offset = 0                
       Invert the band? = False
   bytes in file header = 351

Note that the time per file is incorrect. This may be related to the next ticket I'm about to submit (I think it will be #80).

[Ryan Lynch]

I looked into this a bit more and I think it was a consequence of two things. One was the downsampling factor which has now been addressed as part of ticket #80. The other was actually an issue with PRESTO not interpreting the filterbank file correctly when it contains more than one polarization product.

However, if I try to create a filterbank file using digifil and the -d 1 option, I get an error:

Error::stack
dsp::SingleThread::run
dsp::SingleThread::run
dsp::Transformation[Detection]::operation
dsp::Detection::transformation
dsp::DataSeries::get_udatptr()
Error::InvalidState
Error::message
Your ipol (1) was >= npol (1)

If I use -d 4 or -d 2, I get no error. The input data are definitely complex voltages from two polarizations channels, and I verified that the GUPPI header blocks specifiy NPOL=4 (i.e. real and complex for both X and Y) for all data blocks.

[Andrew Jameson]

Hi Ryan,

I believe that the error with the "-d 1" option was fixed in a recent commit to the DSPSR trunk. Would you be able to run this test again and report back if this last lingering problem is now resolved?

Cheers,

Andrew

[Ryan Lynch]

This seems to be producing results of the appropriate length with the -d 1 option now. Thanks!