Georgia, Ansel, Ben, Sheila, Daniel, Fil and others (https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=65143 and linked logs) have developed evidence suggesting that ISC Common at EX is one of the important components in the generation of the 95 Hz peak in DARM, though it is possibly also associated with other Beckhoff components ( https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=65620 ). Yesterday I used a magnetometer to verify that the 95 Hz peak was largest near this chassis. I opened the chassis and found a particular location near an EK1110 where the magnetic field was about 3 orders of magnitude larger than at the nearby ebay SUS magnetometer that we have been using (see figure). I used an upside down Dust Off spray to gently cool the EK1110 in order to test whether a clock chip inside played a role. The 95 Hz peak didn’t move so I don't know how/where it is produced.
So instead I tried reducing the coupling by reducing the resistance between the chassis and ground. The grounding clips I added seemed to work: the 95 Hz peak and the lower frequency junk was significantly reduced in the ESD power supply monitor and increased in the ground current monitor that I had installed (spectra in figure). I connected/disconnected the alligator clips in several cycles (blue/red respectively in the spectra) to confirm that the change was not coincidental.
While decreasing the noise on the power supply and increasing the current to ground seems like it should help, high-sensitivity DARM was unavailable before I left. So if DARM is worse rather than better, just disconnect the 3 alligator clips on the chasis (see figure) to restore the grounding to nominal. In any case, it should be an interesting test of the coupling mechanism.
For the record, this is the ISC Common Beckhoff chassis lives in the "remote" ISC rack in the X VEA, which is close to, if not immediately adjacent with, the EX-SUS-R1 rack which houses the ESD low- and high- voltage drivers. The drawing for the rack, D1001459, shows that this chassis spans the rack heights U17 - U19. The drawing for the chassis itself is D1002961. The only systems-level drawing that *should* cover how adjacent these are to each other would be "Rack & Cable Tray Layout, X-End, H1 & H2" D1201291, but that doesn't exist, and points to the "Rack & Cable Tray Layout, Y-End, H1 & H2" D1100024 drawing, which (a) has a mis-print where the block that represents the rack is on top of the label, so the label is illegible, and (b) doesn't reflect the locations of the racks at either the X end or Y end.
We have DARM back on ETMX today and the 95 Hz does look better. The 60 Hz also seems lower.
I made some plots comparing a time on Nov 18 (19th in UTC) and the lock today. I was only able to get 50 averages from today's lock without hitting a glitch or measurement (red traces).
First plot shows the broadband DARM. On Friday (yellow) there were many lines coherent with the EX ESD power monitor (bottom left) and the susrack magnetometer (bottom right), as well as the prominent 95 Hz. In today's lock (red) many lines are gone or have reduced coherence with DARM, while the noise in the magnetometer is actually worse (?). There's a broadband improvement in DARM which might be causing our increased range. Sheila suggested we might want to chop the alligator clips connecteed and disconnected. (Clarification to Jeff's comment: this chassis is in the ebay, not the VEA, and quite far from the ESD electronics).
Second attachment is zoomed in around 25-50Hz, several of the peaks that are gone from DARM were broad peaks coherent with the ESD pwoer monitors , and are weaker but visible in the magnetometer.
Third attachment is zoomed in around the 60Hz. The shoulders look the same but the line is reduced.
Edit: I went to EX tonight and chopped connecting and disconnecting the alicator clips from the ISC common chassis. Craig will post the GPS times. I looked at the BLRMS and the range during this test. Fuchsia is when I disconnected the extra grounding clips and yellow is when I reconnected them. The chop is visible in BLRMS_4 (red) (60Hz - 100Hz, not surprising since we know the 95 Hz goes away), and also in BLRMS_3 (green, 35Hz - 60Hz). During the tests there were a few glitches and the 28Hz line was drifting up and down, polluting BLRMS2 (orange, 25-30Hz). There were also glitches in darm every time I disconnected or reconnected an aligator clip. The glitches made the BNS range a bit hard to parse (bottom right), but it looks like the extra grounding connections are giving us an extra ~4Mpc of BNS range. Craig made some median average plots which show a broadband improvement to DARM while the clips are connected.
Georgia went out to End-X and ran three on-off tests with the alligator clips:1353286129 - 3 aligator clips disconnected (30) 1353286894 - 3 aligator clips reconnected (31) 1353287622 - 3 aligator clips disconnected (32) 1353288496 - 3 aligater clips reconnected (33 GLITCH) 1353289545 - 3 aligators disconnected (34) 1353290594 - 3 aligators reconnectedI did a median-average analysis of the first two test times with a 1 Hz binwidth FFT to analyze the broadband ASDs and coherence with the ESD power supply monitor (H1:PEM-EX_ADC_0_19_OUT_DQ) and the EX magnetometer (H1:PEM-EX_MAG_EBAY_SUSRACK_Y_DQ). We can get loads of averages this way to try to resolve the low level of coherence these channels have with DARM. With median-averaging we don't have to worry about the glitches, although the median-averaged coherence will decline a bit slower than mean-averaged coherence rate of 1/number_of_averages. Short answer: Attaching the alligator clips consistently reduces broadband DARM noise Additionally, the coherence of DARM with the magnetometer goes down. I have posted two PDFs groups of plots. In both PDFs, the blue lines are the ASDs and coherences of channels with the alligator clips disconnected. The orange lines are with the alligator clips connected. Plot 2 from each PDF shows OAF CLEAN DARM. Clips connected (orange) DARM is consistently lower than clips disconnected (blue) DARM. Plot 6 from each PDF shows coherence between DARM and EX magnetometer. Clips connected (orange) coherence is consistently lower than clips disconnected (blue) coherence. I only processed the first two sets of time (1353286129 to 1353286829) and (1353287622 to 1353288472). The third time, we moved to NOMINAL_LOW_NOISE too soon after the clips we reconnected. The calibration was consisted this whole test (set by Jeff and Louis in alog 65973). Code ----- The code for this on CDS computers is in/ligo/gitcommon/labutils/nds2utils_exampleIt's also online at https://git.ligo.org/aligo_commissioning/labutils/-/tree/master/nds2utils_example
We are suspicious of the power supply (H1:PEM-EX_ADC_0_19_OUT_DQ) not being coherent with DARM at 95 Hz in the above plots. I checked the median-averaging by making a plot of using mean-averaging during the first clips disconnected/connected test above. The results are the same for both cases: 1) DARM broadband 30-50 Hz noise is lower with the clips connected. 2) Magnetometer coherence with DARM is lower from 20-100 Hz. 3) EX power supply is not coherent with DARM, as resolved by 1399 averages (1 Hz binwidth and 50% overlap).