Well, I guess we've only got two orders of magnitude to go there.
I'm looking at the Marconi manual now, it looks like the external modulation input is 1Vrms = displayed set deviation (see page 1-7). I think the sqrt(2) we were using before was to get to Vpp (i.e. the voltage to freq you would get with a simple DC offset). I'm squinting at your calibratedBeatnoteSpectrum.py code now, it looks like it makes a sqrt(2) adjustment. The PSD is already in Vrms so maybe no adjustment is needed. Sadly, removing this would move the spectrum higher.
Also looking at calibratedBeatnoteSpectrum.py it looks like your using the closed loop transfer function G/(1-G) to infer the BN from the actuation signal pick off rather than the suppression function 1/(1-G). Is this want you meant to do? We were discussing this before and were pretty sure that because the readout is after the 'injection point' of the beat note then we wanted to device by suppression function. Correcting by the CLTF means that at higher gain (G>10) the CLTF-->1; with a suppression function it will provide a 1/f slope correction.
I don't see a SR560 gain in there either, but maybe you put the correction in somewhere else when you stitched the spectrum? Or there's a twiddle factor I missed? The gain setting was x200 (23 dB), if that's not included then that would make a big difference. Or maybe I've missed something and its all rolled into the OLG.
We might need make a schematic tomorrow of the PLL
Also, I tried to untar the tarball below, had trouble. Not sure if its my machine or not.
I have added an October spectrum to the noise budget, it's the first plot below. Click here for the Bokeh version. Sorry it's not on nodus yet.
The Marconi actuator strength was at 10 kHz/Vpp. The PLL OLG used is shown in a plot below. The phase locked loop open loop gain unity gain frequency (PLL OLG UGF) is ~30 kHz.
Not much has changed in the last two months, as we've been running around like chickens with their heads cut off accomplishing little in the way of science but much in the way of making that future science possible. Still, the hard stuff lies in front of us.
Getting down to the predicted coatings thermal noise must be our top priority for the next two months.
This means we need to understand what limits us, i.e. make additions to this noise budget where the total expected and measured agree in some sense. One curve that is obviously wrong is the PLL osc. noise curve in grey. At 10 kHz/Vpp we are ~15 times worse than that curve indicates on this noise budget.
FSS boxes must be better understood to remove suspected low frequency noise injection and avoid saturation of the high gain EOM paths.
EOMs must be looked at to see if thermally stable and the EOM drivers are okay.
Thermal drift of the cavity MUST be understood and eliminated. Perhaps the different laser power resonating in the cavities is causing this. Measurements are currently underway.
Think about adding PMCs or venting, but we must understand WHY we need to do these time-consuming things before we do them. Also good if we can model our expected improvement. If we can get coatings thermal noise without this, all the better.