(Kiwamu, Stefan)

Triggered by Anamaria (https://alog.ligo-la.caltech.edu/aLOG/index.php?callRep=8984) 
we looked into whether we see the same excess noise when the demodulators are driven by 135MHz.

Short answer: yes - it seems to be a characteristic intrinsic to the demod board design.

Long answer:

1) We switched the LO of the 135MHz and 27MHz. The noise stayed with the 135MHz (i.e. both demod boxes act identical).
2) We verified the demodulation gain on the 135MHz board: 6.0mVpkk of RF into the PD input produced 33mVpkk at the monitoring point. That's 14.8dB gain, or a factor of 5.5.
3) The digital gain after this is x2 (single to differential on the demod board) x 10^(45/20) (whitening gain) x 2^16cts/V = 583kcts/V. We measured about 560kcts/V. We put its inverse - 1.8e-6V/ct - as calibration factor into the input filter modules.
4) Plot 1 shows the terminated demod noise, referred to the demod mon point, for the 135MHz and the 27MHz demodulators. To compare to Anamaria's numbers, divide by 5.5 to refer to the demod input.
5) Plot 2 shows the same thing, but taken in analog at the demod monitoring point. The noise agrees. We also verified that there are no outrageous high frequency signals - the largest signal is not surprisingly at 135MHz, with -63dBm at the monitoring point. There is also a 410MHz signal with -73dBm.
6) We also verified that the LO signals look clean - there were no competing peaks. We also took a hi-res spectrum around the carrier - at 100Hz offset from the carrier we have 95dBc/Hz for the 135MHz (-99dBc/Hz for the 27MHz), plots 3-6.

In summary , we cant find anything wrong with the LO signal or the sensing chain - the problem is most likely intrinsic to the design.