J. Kissel, L. Dartez
%%%%% Executive Summary %%%%%%
Remeasured OMC DCPD electronics chain electronics, including compensation, post Jan/Feb OMC swap. There's a small, 0.3% drop in magnitude below 25 Hz. The first line of suspicion is that the environmental or electrical conditions surrounding the new style of transimpedance amplifier, even though the circuit and enclosure itself hasn't changed, but the investigation has just started.
%%%%% More Info %%%%%%
As y'all know, we swapped out the OMC in Jan / Feb 2024 (see highlights in LHO:75529).
That means we have brand new gravitational wave DCPDs. However, it's *only* the DCPDs that have changed in the GW path. Remember, as of O4, the PD's transimpedance amplifier (TIA) is now inside a separate podded electronics box () that encloses a brand-new style of TIA (see T1900676 and G2200551). This need not -- and hasn't -- changed with the swap, where it used to need be changed because the TIA was built in to DCPDs in pre-O4 OMCs.
So, in principle, we've "just" disconnected the old PDs, and reconnected the new PDs, to the same electronics. As such we don't *expect* the frequency response of the signal paths to change.
However, Keita reports, for the first time in history, that there're no electrical issues with the OMC sensors after the OMC swap in January (LHO:75684). While there have not been issues with the DCPDs themselves, per se, recall, for example, problems in the past including issues with shorts to electrical ground of the OMC's PZTs (IIET:12445). Keita did report that, during this Jan/Feb 2024 vent that he found and mitigated some grounding issues with the preamp though -- the 3 MHz SQZ side-band pick off of the gravitational wave DCPDs had shown some signs of electrical short to ground. Quoted from LHO:75684:
Inside the chamber on the in-vac preamp, the DB25 shell is connected to the preamp body (which is isolated from the ISI via PEEK spacer). At first DB25 shell and the preamp body was shorted to the ISI table, but this turns out to be via 3MHz cable ultimately connected to the in-air chassis. As soon as both of the 3MHz cables were discunnected from the in-air chassis, preamp body as well as the DB25 shell weren't conducting to the ISI table any more.
I interpret this to mean that there's a *potential* that the electrical grounding on board the OMC and in the GW signal path of the TIA *has* changed from "there used to be an issue" to "now there is no issue."
So with uber-careful, precision calibration group hat on, I repeated the remote, full-chain measurements of the OMC DCPD GW path -- including the digital compensation for their frequency response -- that I took on July 11 2023 -- see LHO:71225.
Attached are the results -- the magnitude of the transfer function -- for DCPD A and DCPD B. There are three traces:
- The former measurement with the previous OMC DCPDs, on 2023 Jul 11.
- The first measurement with the new OMC DCPDs connected, on 2024 Feb 22 (last week Thursday)
- The second measurement with the new OMC DCPDs connected, on 2024 Feb 26 (yesterday, 4 days later)
We do see some small change ([3.05e6 / 3.04e6 - 1]*100 = 0.3% reduction) in the magnitude below about 25 [Hz].
Preliminary investigations cover a few things that might cause this. Because of where the "wiggle of change" is happening at 25 [Hz] -- right at the RLC complex poles, I immediately suspect the environmental sensitivity of giant ~2.4 [Henry] inductors and/or the electrical grounding situation surrounding the TIA.
Regarding the environmental situation:
- The OMC and HAM6 are back mostly at ultra high vacuum (~1e-6 [Torr], when its typically 1e-7 [Torr]) :: (so, any physical distortions of the enclosure that would change the geometry of inductor should be similar)
- The TIA has been powered on for several days even prior to the 2024 Feb 22 measurement :: (so the dominant thermal load on the circuit -- the bias voltage -- should have had time to thermalize), and
- LVEA temperatures are stable, albiet 2 [deg C] cooler :: (I'm not sure if a 2 [deg C] in the external environment will have such an impact on the PDs)
Of course, it's an odd coincidence that both DCPDs chain response changed in the same direction and magnitude -- maybe this is a clue.
The fact that the 2024-Feb-22 and 2024-Feb-26 measurements agree indicate that:
- The change is stable across a few days, implying that
- The TIA circuit has been on for a while, and circuit is thermalized
Also attached are trends of these environmental conditions during the 2023-Jul-11 and both 2024-Feb measurements.
Also also attached are the two relevant MEDM screens showing the OMC DCPD A0 filter bank configuration during the DCPD A measurement (OMC DCPD B0 is the same), and the Beckhoff switch states for the excitation relay in the TIA and the whitening gain relay in the whitening chassis.
%%%%% What's next? %%%%%%
(1) Ask Keita / Koji / Ali some details about the DCPD chain that I've missed having been out.
(a) Are you sure you plugged in the transmitted PD into DCPD A and the reflected PD into DCPD B, the configuration we'd had with the previous OMC?
(b) When were the electronics powered on?
(c) Can you confirm that other than the DCPDs and the cable connecting them to the TIA, no electronics have changed?
(2) Using the same remote measurement, configure the system to measure the TIA response by itself to see if there's a change and if so if it matches this overall chain change.
(3) If (2) doesn't work, use the remote measurement tool to measure the TIA and the Whitening together, take the ratio of (3)/(2) to see if the whitening chassis response has somehow changed.
(4) If the answers to (1), (2), or (3) don't solve the mystery, or provide a path forward, then we ask "does this *matter*?" A reminder -- any change in the frequency dependence of the OMC DCPD GW path electronics that's not compensated is immediate and direct systematic error in the overall DARM calibration if not accounted for. So the question is does 0.3% error below 25Hz matter, or is it beneath the uncertainty on the systematic error in calibration that's present already for other reasons? To answer this question, we'll resurrect code from G2200551, LHO:67018, and LHO:67114 which creates an estimate of the impact on the calibration's *response* function systematic error, i.e. creating an "eta_R."
(5) If the resulting estimate of eta_R is big compared with the rest of systematic error budget, then it matters, and we're left no other course of action than to out to the HAM6 ISC racks with our trusty SR785, remeasure the analog electronics from scratch, fit the data, and update the compensation filters a la LHO:68167.
