• put alfa on the turret and adjust the bbm gain to get 30 counts in the a/d converters (using pnet --sig). Leave these gain levels fixed for the entire test. The gain levels for the low band were:
  

  pol	bm0	bm1	bm2	bm3	bm4	bm5	bm6
  A	10	9	11	11	10	11	13
  B	11	9	11	12	12	10	4

  Bm 6B was not working so ignore it. The high band was not used since we only had 1 frequency to inject. bbm1 was not used since that synth was used to generate the 830 Mhz signal.
  
• Remove the 30 db pad from the directional coupler were the sine wave is injected into the system (with the noise source). disconnect the noise source so we only have the sine wave.
  
• Set the synth to 830 Mhz and mix it to 250 with a 1 Ghz lo. This places the birdie at 170 Mhz. It is 5 Mhz from the center of the low band (175 Mhz).
• Set pshift to 0x1ff5, dshift=8, integration = .1 secs, len=8192,
  
• Increase the synth amplitude and run the pdev spect to see where the polyphase filter bank starts to overflow.
  

• A 830MHz Sine wave was injected via the if2noise source input (there was only a sinewave, the noise was disconnected).
• This sinewave was mixed to with a 1Ghz lo to 170MHz.
• The power was measured using the if2 power meter sitting on beam 5.
  
• This signal was then sent to the bbmixers via the normal path.
• The signal was only injected into the low band. Beam 1 had no signal since that synthesizer was used to generate the birdie.
• Before the sine wave was injected, the gain levels in the bbm were adjusted to give a 30 count rms on input noise. This is the nominal setting for  using the spectrometer so the compression levels should reflect what occurs during normal observing.

• Each plot has 4 traces: black(polAI), red(polAQ), green(polBI), blue(polBQ).
• Page 1: A/D rms (in counts) vs bbm input power. We got to an rms of 1400 with close to 1dbm input.
• Page 2: base band mixer compression vs input power.
  
• A linear fit was done to 20*alog10(atodrms) vs bbm Input power. The last two points were not used in the fit.
• The data -fit was plotted to see where the compression set in.
• The residuals start to fall off above -1 dbm At 1dbm it is close to 1db compression.
• -1 dbm is flagged for reference
  
• Page 3: A blowup of the Peak A/D counts vs input power.
• this shows how much of the digitizer range we use before compression sets in.
• at -1dbm the a/d peak values are 1400 to 1700. This is compression free.
• at 1dbm the a/d peak values are 1800 to 2047. This is near 1db compression.

Conclusions:

• Using a sine wave the bbm starts to compress around -1 dbm with 1 db compression around 1dbm
• The A/D's have a peak value of 1400 to 1700 counts before compression starts. The A/D values for 1db compression is 1800 to 2047.
• We could remove 6db of attenuation from the bbm output stage. This would increase the compression free range of the A/D but it doesn't look like it will gain us much.
  

<- page up
home_~phil