This is a follow up of 14209.
The trouble with the whitening was due to sub-standard quality of the DB37 connector shell of the binary IO cable.
The connector doesn't fully seat even if you tighten the connector screws, you need to physically press the shell hard towards the chassis to properly seat it. Once seated, the problem is gone until you give the cable one hard jiggling.
Once I removed the shell from one of the cables, it was easy to fully seat the connector and the whitening worked correctly even when I jiggled the cable multiple times. I removed the shell from all EY WFS BIO cables (but not from QPD BIO as they are working).
We have many of these connectors, but DB37 seems to be most problematic because the cable is very thick and gives the connector some good torque depending on how it's angled.
This type of DB37 connector shell has another problem, which is that the cable is so thick it's sometimes difficult to correctly put the shell halves together, as was documented by 12795.
At the very least, these shells need to be replaced with something more reliable. Or cables remade.
Gabriele, Peter, Dave
Restarted h1psliss with new model. Inputs and Outputs of the PSLISS common block connected to ADC/DAC at top level. After the model restart the DAQ was restarted.
J. Kissel, A. Pele, R. DeRosa (c.f. LHO aLOG 14086) After having spoken with the LLO team, Ryan informed me that he *did* in fact have to take considerable effort to re-arrange his sensor-correction STS2 readout system after initial install before he could use it. Regrettably, he hasn't written up any aLOGs or documentation on *how* his system is connected. After a few more hours in the EE room, with new understanding of what was originally intended, and following cables through zipties and cable racks, I now understand that indeed the bulk of the problem is in the front-end models, AND that the HAM45 rack cables were just down-right disconnected at their AA chassis inputs for some reason. I've plugged them back in, and as suggested by Ben, in the order that the HAM16 wiring diagram (D1101584) suggests, such the DuoTone channels remain un-over-written, i.e. now ALL HAM ISIs and HAM HPIs have the following cable arrangment: ADC_0_24 STS A X ADC_0_25 STS A Y ADC_0_26 STS A Z ADC_1_24 STS B X ADC_1_25 STS B Y ADC_1_26 STS B Z ADC_1_24 STS C X ADC_1_25 STS C Y ADC_1_26 STS C Z As such, I don't think I need any more analog changes to the system, but it's time to dig into the top-levels of the front end models, now that I understand how the analog system is wired up. I attach a system drawing which I intend to follow, which is now a part of ECR E1400386, II 942, all cited in Work Permit 4876. I'll begin work as soon as I'm authorized, and Ben will begin updating the drawings accordingly.
Aleaxa, Kiwamu, Jenne, Sheila
We worked on DRMI tonight, it is locking at 1Watt within a few minutes. We are using gains of 22 for PRCL, 40 for MICH, and -400 for SRCL. We are using the PSL-POWER_SCALE_OFFSET, so the gain settings and trigger settings in the gaurdian are now set for 1 Watt input power, but we could scale them to 10 Watts. We have been locking sometimes without the 27 Hz notches in M2 of PRM and SRM, and munually turning them on after it locks (this is needed to prevent the suspensions from ringing up). We are using the guardian for everything else, including the offloading to M2.
We started to measure the relative gains between 1F and 3F ( we think we need a gain of 3.2 in the input matrix to swtich PRCL to REFL 27 I) but quickly realized that we don' t have enough signal in REFL 135. It seems like we need our 45 MHz amplifier installed before we can move on to 3 F.
Times that DRMI was locked:
3:25 UTC october 1st - 4:21 UTC
5:10 UTC (maybe a few minutes earlier). This is still locked, Kiwamu is now scanning the OMC.
To do list:
When the DRMI was locked, I scanned the OMC in order to doublecheck that we were on the right operating point by looking at the upper and lower 45 MHz sidebands.
It looks we are on the right operating points -- both upper and lower 45 MHz are equally prominent while the carrier and 9 MHz sidebands are suppressed at the dark port. Good.
The attached is a plot of the OMC scan. The x axis is converted into MHz of the laser frequency by using carrier's and 9MHz sideband's peak locations. This coarse calibration is done by using another set of data taken at the time when I had a single bounce beam. (Note that I did the single bounce measurement right after I measured the DRMI scan by intentionally unlocking the DRMI for this purpose). The red curve represents that with the single bounce beam and the blue is the one with the DRMI locked. In adittion to the calibration of the x-axis, I was watching the OMC trans camera and I knew that the highest peaks in the blue curve are all 00-modes.
As shown in the plot, both upper and lower 45 MHz sidebands are prominent when the DRMI is locked and there is no significant imbalance in their amplitudes. So we locked the DRMI on a right operating point.
Also, as shown in the red curve, the modulation depth at 45 MHz is low and consistent with that measured by Dan a couple of days ago (alog 14196).
Also we noticed that SRCL was still hopping. Tonight it was pretty clear that the hopping was induced by angular fluctuations in MICH. By using the ITMX oplev loop in pitch, we could reduce the number of hoppings.
Currently it seems to be dominated by horizontal motion of some optics which we could not identify by looking at the oplev signals. The attached is time series of PRC and SRC buildup, observed by POPAIR_RF18 and ASAIR_RF90. When ASAIR_RF90 goes down, SRC tends to hop to the other mode (or perhaps it is just a short glitch and there is no stable mode). Also, it is visible in the screenshot that the PRC and SRC build up fluctuate in a coherent way.
I started it back up with the previous channel list. It looks like it must be a problem with one of the channels that Dave tried to add. The error it logged was: Sep 30 14:23:51 h1conlog2 conlog: ../conlog.cpp: 301: process_cac_messages: MySQL Exception: Error: Out of range value for column 'value' at row 1: Error code: 1264: SQLState: 22003: Exiting. This would seem to imply that the value for one of the process variables was out of the range of the data type in the database. I'll try to narrow it down tomorrow.
See the linearity results attached.
The Range Of Motion results is unable to check (with the current script) H1-, V1+, & V4+ due to current offset on the IPS. This needs to be reworked.
Maintenance Day Summary of CDS code Status:
Model code with local modifications with respect to SVN repository:
Alexa, Kiwamu, Jim, Dave
h1ecatx1 EPICS channels froze at 11:04PDT. The EPICS IOC was stil running, but all data was frozen and white-invalid. We restarted the computer and Alexa restarted the EPICS IOCs for SYS and PLC[1-3].
The ISS second loop photodiodes are very well coherent with the IMC WFS B yaw signal, as visible in the first plot.
This WFS is used for both DOF1 and DOF2 of the IMC alignment. I added an offset to the DOF1 yaw error signal. This improved significantly the intensity noise at almost all frequencies, see the second plot.
The thrid plot is a spectogram of the intensity noise as a function of time. The bottom trace shows the IMC DOF1 error signal. It is clear that a mean value of 250 counts gives lower intensity noise. It is also clear that intensity noise is highly non stationary, even with a good IMC offset.
The fourth plot shows the same story, but this time intensity noise is characterized with a trace showing the band-limited RMS between 100 and 200 Hz. Using the best IMC alignment offset reduces the noise in the band by a factor ~10. The last plot is a zoom into the last minuts of the test. It is clear that the residual fluctuations of intensity noise are related to the residual angular motion of the IMC.
In summary, we can improve a lot the intensity noise at the output of the IMC by choosing a better alignment position, and by improving the accuracy of the angular control loops.
NOTE: Today there is a tour of the Hanford site for LIGO staff from around lunch time - 5pm.
Day's Activities
Leaving for Tour at 11:30 (Jim W & Dave will help cover the afternoon)
This morning I reset Counters for following HEPI systems: HAM2, HAM3, HAM4, BSC3(itmx), & BSC3 (BS).
Note: BSC3 & HAM3 continue to have saturations though.
WP 4869 The user environment has been changed to set up the awgstream environment properly. The CERN ROOT package has been replaced with a version that supports python. The gds software (diaggui, foton, etc) has been updated to gds-2.16.12.2, which has a number of bug fixes and a few enhancements: * python scripting capability for awg and foton have been added. * Cut/paste added to the Calibration dialog box in diaggui (and foton). * Long channel names are no longer ignored in the Calibration dialog box. * Fix bug preventing diaggui from reliably making continuous measurements. * Unsigned 32 bit integer data support for diaggui time series. * Foton legacy write option has been removed. * Fix foton bugs for file path option, add "-o" option for non-gui file processing. * Fix foton bug which caused crashes when invoked from diaggui.
Sudarshan, Jim and Dave
We power cycled h1iscey as part of the investigation to fix some dodgy ADC channels, we were not successful.
sequence:
take h1iscey out of dolphin, stop models, power down CPU, power down IOChassis, power down PEM AA chassis.
power up AA, power up IOChassis, power up CPU.
Sudarshan, Gabriele
To try to reduce as much as possible the beam pointing sensitivity of the ISS photodiodes, we wrote a python script that reads the PD signals and demodulate them with the excitation sent to IM3. In this way we could send a 12 Hz, 1000 cts sinusoid in yaw and a 17 Hz, 1000 counts sinusoid in pitch, and look at time domain traces of the demodulated signals, as shown in the first attached plot. We moved a bit the picomotors, and we could improve quite a lot the situation. As shown in the second plot, right now we almost can't see the sinusoids in the PD signals, since they're below the noise background. Here are therefore upper limtis on the dP/P/dx:
Photodiode | Pitch [1/m] | Yaw [1/m] |
PD5 | 18 | 105 |
PD6 | 18 | 64 |
PD7 | 20 | 107 |
PD8 | 21 | 79 |
The optimal position we found is not exactly with the QPD centered (both x and y are around 0.3). Maybe we could improve this, but for the moment we guess the situation is good enough.
Summary:
Some whitening settings for EY green WFSA I3, Q3 and WFSB Q2 channel don't work. It's probably the whitening chassis itself as the whitening request and the readback agree with each other.
For now I'm leaving both of the chassis in place as there are some usable settings, but note that these guys have a history of many troubles due to chassis and crappy cablings (12159, 12138, 12127).
Details 1:
For WFSA I3 and WFSB Q2, the measured whitening gain doesn't match the request and the readback (attached).
You can see that in both cases one of four stages (+3dB, +6dB, +12dB and +24dB) is failing. It's the 12dB gain stage for WFSA I3 and the 6dB stage for WFSB Q2.
These were measured injecting 20mVpp signal at 100Hz using a function generator and a breakout board.
Details 2:
For WFSA Q3, the third whitening filter doesn't turn on.
For now:
I set the gain to +27dB and turned all filters off.
Update: It was crappy connector shell.
https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=14243
QUAD 06 (Q6) Phase 1B transfer function plots are attached. We had a hard time obtaining good coherence in the Transverse TF, so it is a bit hashy. Will try again.
Most notably is that, like Q8, the second pitch mode frequency is unexpectedly pushed upward on the main chain. Recall, we never found the mechanism to fix it on Q8. Interestingly, both the Q8 and Q6 assemblies are of the same batch of wires and are fresh builds, but by 2 different assembly teams, and on 2 different solid stack/test stand units. Q8 is an ETM type of QUAD while Q6 is an ITM QUAD, but both main chains have the same pendulum parameters - both are detailed in the 'wireloop' model.
The Q6 data is plotted as QUADTST.
We've checked that all wire diameters are as per the specs and that the wire segment clamps are seated properly on the masses. We've also checked that the wire segments have been assembled with the proper assymetry as per specs (looking for something obvious).
Attached are pix of this unit, in case someone wants to look at them. To me, they look just like the last few QUADs we've built, including Q8.
Maybe this is a long shot, but we've exhausted all the simple causes...could the top wire be the wrong material? If the modulus of elasticity was higher, within a factor of 2 from where it is supposed to be, that would explain this strange pitch mode.
One way to test this is to measure the violin modes of the topmost wire in situ and see if it is right. Or maybe more simply, cut some wire from this wire stock, hang some wieght off of it, and measure its violin mode.
The correct 1.1 mm diameter wire should have a violin mode of
frequency in Hz = sqrt(tension/0.0067)/(2*L)
where 0.0067 is the mass per unit length.
For example tungsten has a modulus about 2 times higher than what we are supposed to have. If for whatever reason we ended up with a tungsten wire, it would have an in-situ violin mode in the low 200s of Hz, rather than the 332 Hz spec (much denser than the usual piano wire).
Or even more simply, you could weigh some length of wire. The piano wire should be something close to 7 g/m. If you get different value from that, then the wire is the wrong material.
To confirm Brett's latest suggest regarding the wrong wire: We have 2 rolls of 1.1mm diameter top wire here at LHO which could have possibly been used for QUAD builds. Both are labeled as the correct stuff. We weighed a 1m segment from each spool. One measures 7.1g, the other measures 7.3g.
To be continued...
Another sanity check:
The Top Mass blade sets used for these 3 pitch-problematic QUADs are as follows:
Q6 - SET 10
Q8 - SET 8 - although I can't find the actual records
Q9 - SET 2
Q7 - SET 7 - still to be tested, unknown pitch frequency TFs
The SETs go from SET 1 being the most STIFF to SET 16 being the most SOFT. So, the sets we are using for the 3IFO QUADs are somewhat scattered or in the middle of the pack. They are not all clustered at the soft end, nor all at the stiff end...
And here's the spectra of this Q6. Note, the lowest stage (L2) does not have flags during the all-metal Phase 1 assembly, so the spectra plots of L2 are junk.
And now attached are a damped TF from each R0 and M0. As we all have noted in SUS - damped TFs on Phase 1 test stands are not useful since the damping is a function of the code on the out-dated test stands and the loops are not tuned very well. Long story short, there is a little bit of damping evident, given whatever filters and gains are loaded, and we can see healthy excitations run through the suspension so all seems well with damping capabilities of Q6.