Craig, Evan

Dhruva said he could not reconcile the apparent carrier power as reported by AS_C_NSUM and OMC_DCPD_SUM, and indeed it seems like there's a factor (of ~40%?) missing when comparing the signals at the lowest PCAL Y line frequency (the coherence of each channel with PCAL is nearly 1 here). Attachment is the transfer function in W/W, according to the current filter settings. 0.8 A/W assumed for OMC DCPD transimpedance.

I also updated the ASC-OMC_A_NSUM calibration factor from 0.16 to 0.3562, and ASC-OMC_B_NSUM from 0.16 to 0.3326 (from alog 62213).

With an OM1 transmission of 720ppm and another 50:50 beamspliter for the AS AIR beam, we get an ASC_AS_C_NSUM calibration factor of 1/1000*2/720E-6 = 2.78. Keita measured 2.972.

With a beam sampler of 0.66% in front of the OMC (see e.g. here) and another 50:50 splitter for ASC-OMC_A and ASC-OMC_B, we get a calibration factor of 1/1000*2/0.0066 = 0.303. Keita measured 0.3562 and 0.3326. The response of these QPDs is around 0.7A/W rather than the 0.8A/W (see T1500060) that we assumed in each segment's filter module. This would bring the calibration into fairly good agreement.

Applying Keita's correction (2.97 / 1.39) to the AS-C NSUM during O3b, it seems that the ratio DCPD SUM / AS-C NSUM at the PCAL Y frequency was more like 0.85 W/W. During O3b the power into the IMC was 37 W. I looked at a lock stretch (first attachment, 2022–11–05) where Dan Brown was varying the CO2 X power (52992) and although the overall AS port power changes noticeably, the ratio DCPD SUM / AS-C NSUM at the PCAL Y frequency does not change. If there is a genuine throughput issue here, this suggests it is not going to be fixed by altering the SRC mode content.

Looking at a more recent lock stretch (second attachment, 2022–07–21) from when Nancy and Sheila were adding error point offsets to the OMC alignment loops (64147), there appears to be some effect on the DCPD / AS-C ratio at the PCAL frequency.

The phase of DCPD SUM / AS-C NSUM from the recent lock on 2022–08–17 was −0.7°. From the O3b lock on 2022–11–05 with CO2 X testing, it varied from +0.5° to +2.1°.

We were wondering if the high frequency 410.3 Hz PCAL line sees the same carrier ratio for OMC DCPD SUM / AS_C as the low frequency 17.1 Hz line.

It does.

I measured the 410.3 Hz line during a time Dhruv and I put all the PCALY modulated power into that line, for around 17 minutes starting at 1343434807.
I managed to get 0.95 coherence in AS_C with a binwidth of 0.025 Hz and 27 averages.

calibrated OMC DCPD / AS_C ratio = 0.616 W/W

Code is at https://git.ligo.org/craig-cahillane/IFO/-/blob/master/ASC/scripts/asc_as_c_omc_dcpd_sum_high_pcal_line.py

From the attached plot, it seems to me that the DCPD_SUM/ASC-AS_C_NSUM at one of the calibration lines I picked was ~0.86 in O3a and O3b, somewhat down to ~0.84 in 2020/Aug/13 and then started degrading from there. This could be just the times I randomly picked for measurement, but anyway this is what I got for the times I picked.

This could be alignment, but if it's just the alignment it's a huge alignment error, there could be other things going on. We know that this cannot be something inside the OMC because the OMC finesse was fine earlier this year.

What is the unknown loss obtained for O3b squeezing? Is ~14% potential loss (O3a and O3b from this plot) consistent with that? How about ~38% (now)?

I randomly picked some points in time before the recent calibration change in ASC-AS_C_NSUM (left) and one after (right). All traces in left are after Daniel put the calibration of 1.39ct/W for ASC-AS_C, so I scaled the TF by a factor of 1.39/2.972. (There were times when the whitening gain changed to 0dB instead of 18dB but was compensated by turning off "-18dB" filter, so, no net effect.) In the right panel is after the calibration was changed to 2.972. I used h*nu/e/1000~1.2e-3 for both left and right to convert mA/W to W/W where h is Planck's constant, nu is the frequency of 1064nm light (~300THz) and e is the elementary charge. Each TF uses ~10 minutes of data.

You say "this cannot be something inside the OMC". Can the degradation of the high QE PDs theoretically possible as the cause of the additional loss?

Also:
If the loss is before the OMC and universal for all the modes, we must be able to see it as the reduction of the OMC QPD total.
If the loss is mode dependent or mode related, like the change of the mode matching, this ratio is not a good indicator.
 

I looked at a lock stretch after Daniel updated the ASC-OMC QPD calibrations. These QPDs report that more than 90% of the PCAL-modulated carrier entering HAM6 (as witnessed by AS_C) makes it to the OMC breadboard (QPD A says 99%, QPD B says 94%). Meanwhile the amount of this light making it to the DCPDs is only 63%. The responses of each DCPD are nearly equal (DCPD A / AS-C NSUM = 0.32 W/W, DCPD B / AS-C NSUM = 0.31 W/W).

😱 and it means that... the loss is likely to be

• OMC alignment
• OMC mode matching
• OMC loss (although the finesse ~400 was already checked)
• PD QE decrease (in a same way between two PDs)

ASC-OMC_A_SUM and B_SUM cannot/shouldn't provide new/different information from ASC-AS_C_SUM as the former two (OMC QPDs) were calibrated by the latter (alog 62213 and references therein).

Strictly speaking, we don't know (yet), if the value after the OMC is too low, the value before is too high, or a combination thereof. There is ~4x more carrier light before the OMC (mostly HOMs) and ~20x more light in the 45MHz RF sidebands.

Vicky has started a squeezer loss budgets here: Wiki, I added a few lines for HAM6 losses. The known losses accounted for there add up to 12%.  From the measured squeezing Dhruva estimates 32.8% total loss, leaving 24% unknown loss, which has to include SQZ alignment and mode matching as well as OMC losses.