Intro:

• 800 MHz rcvr
• mock spectrometer using 5 bands with bandwidths:172,57,28,14,7 MHz bw.
• center freq changed for different sources
  
• 1 sec integrations. 300 seconds on,off source, 60 secs on,off calibrate, and cal on,off.
• I used the 172Mhz band limiting data to 700-800 Mhz.
  

Looking at the data for J001018+1433

• Images were flattened by dividing by the median bandpass.
  
• dyanmic spectra 300 secs off source polA (.gif)
  
• dynamic spectra 300 secs off source polB (.gif)
• You can see more high frequency ripple in polA (around 1 MHz ). There are also some wider bands that move to the left.

• Top: SourceDif and Calibrator diff (both normalized to unity across the band).
• black polA source, red polB Source. Green is the pola, polB for the calibrator.
• You can see jumps in the polA source difference (black).
• 2nd: (srcOn-srcOff)/(calibOn-calibOff)
  
• black is pola, red is polb.
  
• The ripples in polA stick out.
• 3rd: standard (on-off)/off processing for source
• 4th: standard (on-off)/off processing for calibrator.
• Note that the 20 MHz ripple is not cancelled by the on,off position switch on the strong calibrator.
  

• Page 1: average bandpass for 300 seconds off source.
• blank pola, red PolB
• Page 2: frequency content of bandpass ripple and how it changes.
  
• the 300 seconds of data was split into 5 60 second averge spectra.
• the 1st average spectra was then divided into minutes 2..5.
• the fft was then computed for these 4 spectra: abs(fft(1minuteavg/avgminute0)).
• This is the acf of the spectra
• The data is plotted vs delay in usecs.
• Top PolA, bottom polB
• black is minute 2, red=minute3, green=minute4, blue minute 4.
• the spike at 1 usecs is the 1 mhz standing wave dish to platform.
• polA shows the 1 usec spike, polB doesn't.
  
• For the 1 usec spike to appear, the standing wave phase must change its phase.
  
• If the phase was constant, then dividing by the first minute should cancel it.
  

Summary:

• After  double position switch processing, polA baseline is worse than polB
• The 20 MHz bandpass ripple does not cancel on on/off processing if the on is a strong source.
• PolA has 1 MHz ripple that doesn't cancel over a 5 minute integration. polB is more stable.
  
• I processes a second source. It also showed ripple in polA processed data.