After having surveyed OSEM sensor performance at LLO at the end of O3B (for QUAD suspensions see LLO aLOG entry 52578, for HSTS suspensions see LLO aLOG entry 52570, for HTTS IM and HLTS suspensions see LLO aLOG entry 52562, for TMTS OMCS and BS suspensions see LLO aLOG entry 52557), since OSEM power spectra measurements had already been taken for LHO at the end of O3B courtesy of Rahul to verify suspension health following the recent large M6.5 Idaho earthquake (see LHO aLOG 55857) I've taken the opportunity to compare current performance with reference measurements taken several years ago during aLIGO installation and suspension acceptance testing (see plots below). - TMTS, both TMSY and TMSX power spectra are clear of issues and consistent with reference measurements. - OPOS, the LHO OPO is a recent addition so is compared with the LLO OPO, which raises no concerns, but for a very slightly elevated AOSEM sensor noise on the V2 channel. - OFIS, again the LHO OFI is a recent addition so is compared with the LLO OFI, performances look consistent, but for elevated noise on the SD AOSEM which appears to only be present with damping loops are engaged, this should be verified. - BS suspension measurements raise no concerns. - HLTS, both PR3 and SR3 are consistent with previous measurements. - HTTS, OMs are consistent with previous measurements, no reference measurements had been taken for the RMs or new ZMs so they are compared with each other and raise no concerns but for very slightly elevated noise on the ZM1 LL BOSEM channel (also note that RMs don't have an ISI compared to the ZMs). - IMs do not have a reference after being incorporated into SUS a while ago, therefore, they are compared with each other, and raise no noise concerns, but for a slightly elevated IM3 LL AOSEM channel.
Stuart (LLO), Rahul
We investigated the elevated noise issue on the SD AOSEM for LHO-OFI suspension as reported by Stuart in the above alog (56019), I am attaching figure (figure1.pdf) for reference. Firstly, I re-did the osem spectra measurement to rule out any glitch or data issue, and the results were no different (Damping ON is noisier than Damping OFF), see here. This demonstrates that the OSEM and electronics chain for the LHO OFIS SD channel are absolutely OK, which means most likely the issues could reside with damping loops re-injecting noise. Stuart commented that OFIS suspension has the least mature damping loops, since it was a recent addition, and the LLO OFIS power spectra shows that it is indeed possible to damp without introducing too much sensor noise. Therefore, he recommended to investigate damping loops for the LHO OFIS and compare them to LLO OFIS.
Using “medm_lho” command I was able to get into LLO sitemap and medm screen, to check which damping filters are engaged for LLO OFI and the gains used. The latest OFI filters are checked into the CDS userapps svn,
LLO: /opt/rtcds/userapps/release/sus/l1/filterfiles/L1SUSOPO.txt
LHO: /opt/rtcds/userapps/release/sus/h1/filterfiles/H1SUSOPO.txt
I compared the OFI filters banks (LHO vs LLO) on foton and found that the only difference is in the gain (in filter bank) for Length DoF (LHO = 72.0412, LLO = 46.02), see figure (Filter_gain_Ldof.png) attached. The Length DoF damping loops actuate on the noisy SD OSEM and could be re-injecting noise. To check this, I made the SD channel OSEM noise floor worse by changing the L-DoF damping gain. At the same time, I monitored the damping loop output drive signal with ndscope to check nothing rings up. I ran DTT to check on H1:SUS-OFI_M1_OSEMINF_SD_OUT_DQ channel to watch in real-time the impact of the damping gain changes (tweaked from 2.0 to 0.05, while the nominal is 1.0), which is shown here. Next, I took the ASD of OFI sus for each of those gain changes, the results of which is attached in the pdf file (ASD_OFI). These two plots clearly shows that having more damping gain re-injects more noise and raises the sensor noise.
Later the gain was reset to the nominal 1.00.
Conclusion: A reduced damping gain (for L DoF) of 0.05 (page 5 of ASD_OFI.pdf) that it still effective in damping the suspension modes without re-injecting noise could be a good compromise. We would like to recommend this fix, and if Jenne or Jeff. K thinks that this is important then I would implement it right away. However, if we want to keep the IFO configuration exactly the same while the IFO is re-locked (when we return to phase 2), then I am also fine with that.
Stuart (LLO), Rahul
Based on J. Kissel recommendations/questions I further investigated some of the issues regarding the damping loop for OFI. Given below are the results of the investigation.
Firstly, I can confirm that the damping loop for the OFI looks to be “Good enough” thing where the gain of the filter rises as f well out in the 50Hz region. This is shown in the DoF_bundle.pdf (with 8 plots) attached below. Secondly, on comparing the damping filters between LHO and LLO (as see previously), I found that the filter design was similar for all 3 DoF, the only difference was seen in the filter gain for L DoF (LHO = 4000 and LLO = 200), as shown in gain_L.pdf.
Next, I took a transfer function of the OFI by varying the gain from 0.05, 1.0 (nominal black line), 2.0, 5 and 10 to verify how effective is the damping, the results are attached below. In the above alog we had confirmed that the damping gain of 0.05 reduces the noise re-injection, however the transfer function results looks terrible - the shapes and resonant peaks are nowhere close to the reference black line (which is for nominal gain 1.0). However, the increase in gain does not appear to have much impact on the actual mode peak being damped.
Since the filter gain for LHO was found to be higher than LLO, I reduced the gain by a factor of 2 (new filter gain = 2000) and took a DTT of H1:SUS-OFI_M1_OSEMINF_SD_OUT_DQ channel. The attached plot shows that the reduced filter gain has lower magnitude and this can be compared with the plot shown in alog 56125 (in this plot the damping gain was varied from 0.05 to 2). To confirm if this works, I will re-take the ASD spectra with the reduced filter gain, followed by TF measurements.
Attached below are the results for Amplitude spectral density for OFI (SD OSEMINF), taken with a new filter gain of 2000 (nominal is 4000). The results shows that the sensor noise is reduced even when damping is ON.
Also attached is the transfer function results for L DoF. The shape of the tf and the resosnant peak is coincident for both the cases, i.e new filter gain (blue line) and nominal (red line).
Hence, our suggestion/conclusion would be to reduce the filter gain for L DoF to a new value of 2000, which will reduce noise re-injection in the OFI damping loop.
