---

What we did:

       Spectra taken 24mar21  showed large 64MHz ripples when the all deflection was used to remove the IF bandpass.  On
7apr21  a  2 minute on,off position  switch was done on casA. Plotting (spcPosOn-spcPosOff)/spcPosOff showed no ripple.
Plotting spcCalOff/(spcCalOn-spcCalOff) for data taken after the position switch still showed the ripple.
I also computed the sefd using a flux value o 620 Jy for  casA (from chris salter).

When  processing the spectra:

• there were 7 172.032  MHZ bands each with 8192 channels. The spectra were spaced by 164 MHz.
• I averaged each of the 120 second scans.
  
• the bands were interpolated to a continuous band with 25Khz resolution
• When plotting, a median filter of length 5 channels was used to get rid of DC.
  

The plots show the position switch and calon,off spectra (.ps) (.pdf)

• Page 1: on,off position switching
• Top:  (on -off )/ off
  
• black polA, red polB
• The vertical scale has units of Tsys
• The junk on the left  comes from the mock bandpass having no power (the IF filter cut it off).
• There is no 64MHz ripple seen
• the dip at 8750  MHz may be a resonance in the receiver.
• Bottom: SEFD using casA
• I used 620 Jy  for the flux of casA at xband
• The sefd is   (Tsys/Tsrc)* srcFlux
• Notes no the SEFD:
• casA is extended. (chris salter in his srclist says it's about 5')
• the xband beam is around 10'
• I didn't do any corrections  for the source being extended. A point source would  give a better value
• I found a 21' offset in the elevation encoder (relative to the model made back in 2016.. maybe the encoder jumped during one of our earthquakes?). .. anyway, i came pretty close to optimizing the correction on this source, but a new pointing model might give a little better signal.
• Page 2: calon, caloff spectra
• After the position switch, i did a 10 second calOn,calOff  in the off source position.
• the off source was 35 amin from the on.. so we were well away from the source.
• Top:  calOn - calOff in spectrometer units
• You can see the  7mock bands of  172MHz 
• Bottom:  CalOff/(calOn - calOff)
• this shows the ripple that was seen on 24mar21. It is worse in polb.

What we were measuring.

• Assume we can break the signal path into 3 sections where each section has a constant temperature value and a frequency dependence:
• Tsrc .. assume this is flat..
  
• Tsky, Gstw(f)   tsky is sky and ground. let Gstw come from reflections on the structure or in the horn
  
• Tcal, Gcal(f)    cal including the coupler
• Trcv  Grcv(f)   this is every thing after the cal injection
• Position  switch:
• Ton = ((Tsrc  + Tsky)*Gstw + Trcv)*Grcv
• Toff = ( Tsky*Gstw + Trcv)*Grcv
• (on-off)/off =  (Tsrc*Gstw)/ (Tsky*Gstw + Trcv)
• Since Trcv/Trcv is about 9/100.  expand the denominator
• (on-off)/off  = Gstw*(Tsrc/Trcv) * ( 1 -  Gstw* Tsky/Trcv)
• So any frequency dependence would likely come before the  cal coupler in Gstw
  
• CalOn,off
• CalOn= (Tsky*Gstw + Tcal*Gcal + Trcv)*Grcv
• calOff = (Tsky*Gstw + Trcv)*Grcv
• (calOn-calOff)/calOff = (Tcal*Gcal)/(Tsky*Gstw + Trcv). expanding the denominator gives
• (calOn-calOff)/calOff = Gcal*(Tcal/Trcv) * ( 1 - Gstw*Tsky/Trcv)
• In this case the frequency dependence would come from the Gcal and Gstw
• Since Gstw came out small when we did the pos on/off , the cal lines or coupler must be generating the frequency dependence in the (calOn-calOff)/calOff..