Single channel spike in center of correlator band passes

sep 2002

    The correlator has a weak spike in the central channel caused by the digital filter clocks of the next lowest bandwidth. If you have Nchannels  in the spectra, then the spike will be in channel:
 
  • Nchannels/2+1 if bandpass has not been flipped
  • Nchannels/2      if bandpass has been flipped

  • This is counting channels 1 to Nchannels. You can test if a bandpass has been flipped using the idl routine: corhflipped().

        Data was taken for 100 minutes on 05sep02 with the IF connected to the downstairs noise source. 1.5625 Mhz bandwidth and 1024 channels were used. The plots show the size of the spike relative to Tsys. The spike is located at .78125 Mhz.

    1. Fig 1 has the four sbc for polA
    2. Fig 2 shows the four sbc for polB
        The spike strength varies from sbc to sbc: .009 Tsys (25 sigma) to buried in the noise. This is reasonable since each sbc has a separate digital filter which could have different coupling to the interfering clock. The spike has not been resolved at 1.5/1024=1.5 Khz channel widths so the spike will be more evident in narrow channel widths.

        The correlator itself seems to be integrating ok. The expected  rms is .00033 Tsys  (actually it should be .00031 using 1/sqrt(b*tau*1.208)*1.04 which includes the true bin widths and the degradation from 9 level sampling). We measured an rms of .00035 using only a 2nd order polynomial fit to the baseline. Part of the extra rms is coming from the residual shape in the filters.

        There is another spike at .3373 Mhz that is the same in all 8 sbc. This implies that it is common to all sbc and both polarization's. This means it must be coming from the common noise source. The center of the IF band at the noise source is 750 Mhz and the plotted band has been flipped so the frequency of the spike is  750 + 1.5625/2-.3373 = 750.444 Mhz.
     

    processing: x101/020905/spikeCenBp.pro
    home_~phil