Note: I tried to attach the plot as a pdf but there is something wrong with backend processing this particular pdf and I'm getting a 502 error. Have attached as a zip and a png
I thought it would be a good idea to document the current state of the BN spectrum before switching out the NF1811 post cavity BN detector for Koji's newly modified resonant detector. The new detector has an area of 2 mm x 2 mm, much bigger than the 0.3 mm x 0.3 mm of the current configuration. The larger PD means we can increase the beam size and hopefully reduce the scatter in the vicinity of the beat setup.
To compare apples with apples, I refloated the table. Also, I turned off the hepa fans around the lab. The current BN is 26 MHz and PID locked using the pre-cavity BN detector (Kp = -0.005, Ki = -0.000001). It still hadn't settled from some disturbances earlier in the day, so was oscillating around ±200 kHz range about the set point. I didn't have time to wait, so I took the measurement with Marconi slope set to 10 kHz. PLL bandwidth was ~16 kHz with SR560 gain of 100 and Marconi on slope 10 kHz/V. In this range previous benchmarks of Marconi noise (see PSL:1588) would suggest that the PLL noise limit is 30 mHz/rtHz. Therefore, PLL shouldn't be the limiting noise at the moment, notwithstanding PD dark noise (which isn't really quantified in the noise budget).
When I took the first measurement I found that the some of the scatter features had been improved since the last measurement (see attached figure razor blade in trace). The noise floor seemed to be around 0.2 Hz/rtHz with a roll up at lower frequencies. I realized, however, that I had left the razor blade profiler from the previous day's measurement (PSL:2172) in place, just above the beam. I retook the measurement (see figure razor blade out trace). There were a number of peaks that popped up broadly around 70 Hz, 180 Hz and 400 Hz. Its likely that the razor blade was deflecting and dumping some of the light that has been causing these bumps in our spectrum.
I was a little suspicious of how flat the spectrum was above the characteristic scatter hump (up to 200 Hz). In three separate measurements I blocked the beam from the BN detector, 50 Ω terminated the PLL at the BN input and then 50 Ω terminated at the ADC input . These traces are also included in the figure below. It shows that the apparent limit to sensitivity is the dark noise of the NF1811 photodetector.
It should be noted that the power has been attenuated onto the NF1811 from ~ 1 mW by 3.0 OD neutral density filter down to 1 µW. Any power much above this generates harmonics, that we are trying to avoid. So it seems that with improved dumping around the NF1811 detector we can reduce a number of mechanical resonances scatter peaks but that there is a limit to the amount of signal to noise because we are likely to saturate the RF AC path stages inside the NF1811. Im not sure if this is the right way to calculated but dark noise (NEP) is 2.5 pW/rtHz with 0.75 A/W respositivity and 40kΩ gain PD gain the output noise of the detector is 75 nV/rtHz, For a 16 kHz loop bandwidth (preamp gain 100, VCO slope 10 kHz) the total gain, not including the mixer, is 1e6 which would place the noise floor at about 0.1 Hz/rtHz. I need to work through exactly the right way to calculate this that treats the noise in a physically correct way at the mixer.
So it seems that this measurement is getting closer to our previous best measurements. Data and plotting is attached in a zip below as well as being committed to the git.ligo.org ctn_labdata repository.