Andres, Sheila
After doing the PSL checks, Andres pointed out to me that the PMC power was low (1.1W reflected, 5.9 W trans). We unlocked, and saw that the total power in reflection was 12Watts, so this must be the PMC trans PD problem again. I went to have a look at the cable that Keita and company connected to the PD, (alog 9338), and saw that it was connected to the scope with a 50 Ohm input impedance. It's set to high impedance now and the readback is fine, I gues we need to be carefull about that scope, does it set itself to 50 Ohm when it is power cycled for some reason?
An ACCES Relay Card (PCIe-IIRO-8) has been added to slot 9 (Bus 1-1) of the I/O chassis for x1susex. This is to be used for automated testing of the hardware watch dog chassis.
The GDS software has been updated to gds-2.16.9. This includes fixes for the following bugs/enhancements: 299, 328, 389, 428 - Any combination of start, stop, and BW that result in an fft of 512 points fails, due to how the fftw library is used by diag. Fix by changing the way the fftw plan is generated, specifying FFTW_MEASURE instead of FFTW_ESTIMATE. 497 - Add option to foton to allow specifying a SFM name along with a filter file name. See foton man page for details on the "-s" option. 500 - Removed Kaiser Window option from diaggui, it was never implemented. 511 - Changed foton to report that no design string exists for a filter if that is the case. Previously this was reported as a different misleading problem. 519 - Allow IP broadcast for awg, tp discovery to be specified by an environment variable (LIGO_RT_BCAST) instead of assuming the broadcast address of the workstations. All these changes have been built from gds/branch/gds-2.16.9. This affects Linux and Mac OS X workstations. Existing diag or diaggui sessions should continue to work using the older software, but at some point we need to make network switch modifications that will cause the older software to stop working. Advanced warning will be given.
At BSC 1&2 looking at CPSs, this done now.
Working on ACB in West Bay.
She's heading there now.
Alexa and I are going to the ISC racks near the PSL. Also, I went to the X end for a few minutes.
Because in this after noon I happened to be around the ISC rack with an RF source in hand, I measured and adjusted the IQ balance of the 3f demodulator which had been refurnished recently. I did the same process as was done a month ago (see alog 9100).
Here are the new settings for it:
Note that I didn't change the gains for the RF27 I and Q paths because the gain difference was smaller than 1%.
Stefan, Yuta, Kiwamu,
We could lock the sideband-resonant PRMI for more than 8 minutes tonight. The angular fluctuation seemed better because of the L2P(Y) decoupling filters.
The locking is becoming reliable. We still need to investigate what causes the lock loss events.
Here are what we did:
Tonight I found the ISI BS and ISI ITMY tripped. When trying to put them pack to their previous operational mode (state 1 of ITMY, state 2 for BS), they repeatedly tripped. We decided to work without the two ISIs tonight.
Kiwamu, Stefan, We did the PRM M3 pitch L2P decoupling tonight. We followed that same strategy as for the PR2 (alog 9584): - We paid the most attention to the highest peaks in the L2P transfer function. - We fitted L2P and P2P independently to get good initial values for a fit of the ratio. - We had a harder time with the L2P (compared to the PR2) - in particular I couldn't get the phase at 1Hz right. - The result is only Plot 1: transfer functions: - blue: measured L2P TF - green: measured P2P TF - red: L2P/P2P (from measured data) - black: fit (note that we paid more attention to the frequencies between 0.5Hz and 1Hz, because they matter most (see L2P) - Purple: fitted filter times measured P2P. Plot 2: decoupling at 0.75Hz about a factor of 6 better. Plot 3: decoupling at 0.95Hz about a factor of 4 better.
Looks like a typo in the title for this entry; should read:
PRM M3 pitch drivealign work
[Jeff, Paul]
I had a look through the calibrated length and frequency paths for the IMC (IMC-X and IMC-F respectively) to check that the calibration is still in order. A description of all the filters in the IMC-X and IMC-F paths can be found in Giacomo's entry 5945.
In fact, some things were amiss in the IMC_X path:
IMC-X_M1 path:
Filters that were engaged: to_m
Filters that are now engaged: dc_cal, wresg1, wresg2, to_m
MC2 M1 feedback is not currently engaged, so this may not matter right now. However, I've engaged the filters that calibrate the IMC-X_M1 signal to M3 motion. wresg1 and wresg2 model the suspension TF from M1 actuation to M3 actuation with the resg damping filters engaged (as they are currently). If and when M1 stage feedback is applied, the corresponding IMC-X channel should be calibrated now.
IMC-X_M2 path:
Filters that were engaged: dc_cal, white, sus_d1, sus_d2, to_m
Filters that are now engaged: dc_cal, wresg, to_m
The sus_d1 and sus_d2 filters are used to model the suspension TF without the resg filter in the M1 L damping path turned on. Because we are now using the resg damping I have switched the sus_d1 and sus_d2 filters off and engaged the wresg filter, which is used to model the suspension TF with the resg filter in the M1 L damping path. The white filter is a filter which was implemented in L1 in an attempt to avoid digitization noise above 50Hz (since the signal is so small there). I don't think we want to have this filter engaged, since it has to be accounted for later in data analysis, and I think we can ignore IMC_X so far above the L to F crossover. I've therefore disengaged this filter. This isn't the first time this one has come back to haunt us: see e.g 5311 or 5452.
IMC-X_M3 path:
Filters that were engaged: dc_cal, white, sus_d, to_m
Filters that are now engaged: dc_cal, wresg, to_m
Same story as for the M2 path.
I took a snapshot of all IMC-X filter banks.
IMC-F path:
Filters that were engaged: cts2V, InvGenFilt, VCO, FtoL
Filters that are now engaged: cts2V, InvGenFilt, VCO, FtoL
The only filter that changed since Giacomo's entry 5945 is the FtoL filter being engaged now instead of the tokHz filter. This means that the IMC_F signal is now calibrated in meters, presumably for ease of comparison with IMC_X. I checked the VCO gain filter: it is still the 536,604 Hz/V value, agreeing with Kiwamu's number from entry 5556 after accounting for the double pass of the AOM.
I had at four separate times (first plot) after these changes were made to see how H1:IMC-X_DQ looked after it didn't appear comparable to LLO last week (see alog 9373). I've also attached a more recent plot of the same channel at LLO (second plot). The same channel still looks quite different between sites and H1:IMC-X_DQ still seems quite featureless in comparison to LLO. A rather naive question...are they meant to look similar?
As Laura pointed out, the calibrations for L1 and H1 data for these two paths were quite different. I've now gone ahead and changed the H1 calibration path to match the L1 settings.
IMC-X
M1 filters: dc_cal, white, wresg1, wresg2, to_um
M2 filters: dc_cal, white, wresg, to_um
M3 filters: dc_cal, white, wresg, to_um
This means that the IMC-X path is now calibrated in micrometers and it does have the whitening filter engaged. This means the whitening filter (two zeros at 0.2Hz and two poles at 1kHz) should be accounted for in post-processing.
IMC-F
Filters: cts2V, InvGenFilt, VCO, tokHz
This means that the IMC-F path is now calibrated in kHz.
I've saved a snapshot of all 4 filter bank states.
Arnaud, Sebastien
We've kept investigatiing on the 0.5Hz resonance issue that we've been having on ISI-ITMX (see first alog here). Reminder: we see a peak at 0.5Hz on all the sensors of the ISI when it is controlled.
After more tests, we realized that this peak appears on all the sensors only when the T240s are in the blend (see plot attached).
Plus, by looking at the T240-X signal, you can clearly see this peak at 0.5Hz. It is even more obvious when the noise floor is reduced (ISI controlled).
Thus the issue seems to come from the T240s. We'll continue the investigation tomorrow and find which T240 is bad and how to fix it.
The last plot seems to show harmonics at 1Hz, 1.5HZ(?) and 2Hz. Is this 0.5Hz sharp? If so, this could indicate a problem with the DAQ timing.
Good point Daniel. We have to double check how the other sensors behave. We'll check that tomorrow.
Exactly 0.5Hz (and its harmonics).
Still no resolutions to the problems with HAM4 ISI transfer functions. I took more DTT tf's today, and they all look a little better than the Matlab tf's. The attached plot shows the 3 vertical pods. The 3 thinner lines are the Matlab tf's (which, though not shown, have been repeated 3 times, all 3 times look pretty similar) and the 3 thicker lines (red, green and blue) are the same sensors with a DTT whitenoise tf.
I should add that the H3 GS-13 looks bad, no matter what. It's just hard to see against the general awfulness of the vertical GS-13's. The first attached image shows the DTT tf in the .5 hz to 5 hz range on the H3 GS13. There should be a taller, pointier peak at ~1hz, which is now kind of squashed. Second image is the power spectra for all GS13s, H3 doesn't look exactly the same as the others at higher frequencies, but not bad. All the other sensors look okay.
Sheila, Daniel, Stefan
I tuned up the beat note alignment this morning.
With the arm locked we have 88uW total on the COMM_A_LF detector, 42uW from the arm and 47uW from the SHG. This gives us a beatnote of 28mV pp, with the PFD RF power readback saying -35dBm.
The Comm PLL locks for a few seconds and runs out of range (due to fluctuations at the microseism frequency). One solution would be to feedback to the ETM from the green laser at the microseim frequency ( right now we only have very low frequency feedback to HEPI), this would probably be nice in the end, because the PLL would lock nicely and we need to figure out out how to do this for HIFOXY anyway. Another solution would be to engage COMM with the PLL partially locking, which does still give a useable error signal, which would be quicker to implement.
We remembered that we can just use an ezcaservo script to feed the COMM_PLL_CTRLMON signal back to offset of the the mode cleaner VCO,
ezcaservo -s 0 -g -1 -f 0.3 -r H1:ALS-C_COMM_PLL_CTRLMON H1:IMC-VCO_TUNEOFS
We then proceeded to measure the open loop transfer function of the PLL servo. The settings were:
The unity gain frequency is at 55 kHz with a phase margin of 40º.
With the same configurations as above I measured the power spectrum out of the PFD IMON while the PLL was locked.
work now complete