Last night Gautam walked us through the algorithm used to lock PRFPMI. We tried it several times with the PSL HEPA filter off between 10:00 pm July 7th to 1:00 am July 8th. None of our attempts were successful. In between, we tried to do the locking with old IMC settings as well, but it did not change the result for us. In most attempts, the arms would start to resonate with PRMI with about 200 times the power than without power recycling while the arms are still controlled by ALS beatnote. The handover of lock controls "CARM+DARM locked to ALS beatnote" to "Main laser + IMC locked to the CARM+DARM" would always fail. More specifically, we were seeing that as soon as we hand over the DC control of CARM from ALS beatnote to IR by feeding back to MC2, the lock would inevitably fail before the rest of the high-frequency control can be transferred over.
Nonetheless, Paco and I got a good demo of how to do PRFPMI locking if the need appears. With more practice and attempts, we should be able to achieve the lock at some point in the future. The issues in handover could be due to any of the following:
More insights or suggestions are welcome.
Note; An earthquake came around lunch time and tripped all watchdogs. Most suspensions were recovered without issues, but ITMX appeared to be stuck. We tried the shaking procedure, but after this we couldn't restore the XARM lock. From alignment, we tried optimizing the TRX but we only got up to ~0.5 and ASS wouldn't work as usual. In the end the issue was that we had forgotten to enable the LL coil output so after we did this, we managed to recover the XARM.
I wanted to put my optomechanical instability hypothesis to the test. So I decided to cut the input power to the IMC by ~half and try locking the PRFPMI. However, this did not improve the stability of the buildup in the arm cavities, while the control was solely on the ALS error signal.
Basically, with some tweaks to loop gains, it worked, see Attachment #1. Note that the lower right axis shows the IMC transmission and is ~7500 cts, vs the nominal ~15,000 cts.
Cutting the input power did not have the effect I hoped it would. Basically, I was hoping to zero the optical CARM offset while the IFO was entirely under ALS control, and have the arm transmission be stable (or at least, stay in the linear regime of REFL11). However, the observation was that the IFO did the usual "buzzing" in and out of the linear regime. Right now, this is not at all a problem - once the IR error signal is blended in, and DC control authority is transferred to that signal, the lock acquisition can proceed just fine. And I guess it is cool that we can lock the IFO at ~half the input power, something to keep in mind when we have the remote controlled waveplate, maybe we always want to lock at the lowest power possible such that optomechanical transients are not a problem.
I also don't think this test directly disputes my claim that the residual CARM noise when the arm cavities are under purely ALS control is smaller than the CARM linewidth.
What does this mean for my hypothesis? I still think it is valid, maybe the power has to be cut even further for the optomechanics to not be a problem. In Finesse (see Attachment #2), with 0.3 W input power to the back of the PRM, and with best guesses for the 40m optical losses in the PRC and arms, I still see that considerable phase can be eaten up due to the optomechanical resonance around ~100 Hz, which is where the digital CARM loop UGF is. So I guess it isn't entirely unreasonable that the instability didn't go away?
After this work, I undid all the changes I made for the low power lock test. I confirmed that IMC locking, POX/POY locking, and the dither alignment systems all function as expected after I reverted the system.
Following Koji's suggestion, I decided to investigate the relation between my Finesse model and the measured data.
For easy reference, here is the loss plot again:
Sticking with the model, I used the freedom Finesse offers me to stick in photodiodes wherever I desire, to monitor the circulating power in the PRC directly, and also REFLDC. Note that REFLDC goes to 0 because I am using Finesse's amplitude detector at the carrier frequency for the 00 mode only.
Both the above plots essentially show the same information, except the X axis is different. So my model tells me that I should expect the point of critical coupling to be when the average arm loss is ~100ppm, corresponding to a PRG of ~17 as suggested by my model.
Eric has already put up a scatter plot, but I reproduce another from a fresh lock tonight. The data shown here corresponds to the IFO initially being in the 'buzzing' state where the arms are still under ALS control and we are turning up the REFL gain - then engaging the QPD ASC really takes us to high powers. The three regimes are visible in the data. I show here data sampled at 16 Hz, but the qualitative shape of the scatter does not change even with the full data. As an aside, today I saw the transmission hit ~425!
I have plotted the scatter between TRX and REFL DC, but if I were to plot the scatter between POP DC and REFL DC, the shape looks similar - specifically, there is an 'upturn' in the REFL DC values in an area similar to that seen in the above scatter plot. POP DC is a proxy for the PRG, and I confirmed that for the above dataset, there is a monotonic, linear relationship between TRX and POPDC, so I think it is legitimate to compare the plot on the RHS in the row directly above, to the plot from the Finesse model one row further up. In the data, REFL DC seems to hit a minimum around TRX=320. Assuming a PRM transmission of 5.5%, TRX of 320 corresponds to a PRG of 17.5, which is in the ballpark of the region the model tells us to expect it to be. Based on this, I conclude the following:
In other news, I wanted to try and do the sensing matrix measurements which we neglected to do yesterday. I turned on the notches in CARM, DARM, PRCL and MICH, and then tuned the LO amplitudes until I saw a peak in the error signal for that particular DOF with peak height a factor of >10 above the noise floor. The LO amplitudes I used are
There should be about 15 minutes of good data. More impressively, the lock tonight lasted 1 hour (see Attachment #6, unfortunately FB crashed in between). Last night we lost lock while trying to transition control to 1f signals and tonight, I believe a P.C. drive excursion of the kind we are used to seeing was responsible for the lockloss, so the PRFPMI seems pretty stable.
With regards to the step in the lock acquisition sequence where the REFL gain is turned up, I found in my (4) attempts tonight that I had most success when I adjusted the CARM A slider while turning up the REFL gain to offload the load on the CARM B servo. Of course, this may mean nothing...
The response of the PRFPMI length degrees of freedom as measured in the LSC PDs was characterized. Two visualizations are in Attachment #1 and Attachment #2.
[Jenne, Rana, EricQ, Diego]
Tonight we worked on getting the IFO back in a working status after the break, and then tried some locking.
We left the IFO uncontrolled and in a "flashy" state so that tomorrow we can look into the "back-flashing to the MC" hypothesis.
Two plots from tonight:
Lock loss. Based on the fact that it looked like the DARM servo was running away, Rana posited an effective sign flip in the DARM loop, perhaps due to a parasitic angular feedback mechanism.
While Jenne was probing the IFO at lower powers, we noticed a sudden jump in ASDC. Found the GPS time and fed it to the lockloss plotter. Seems fairly evident that some sudden ETMX motion was to blame. (~2urad kick in yaw)
As a warm-up after the holidays, before the real locking began, I installed 1064nm bandpass filters in front of the transmission QPDs to eliminate the stray green light that is there.
The Yend had threads epoxied to it, so that end should be good. Steve is going to repeat that for the Xend QPD at some point. Right now, the filter is just on a lens mount about 2cm away from the PD box aperture, since that's as close as I could get it.
Also, while I was at the Xend, I noticed that the transmission camera is gone. I assume that it was in the way of Manasa's fiber work, and that it'll get put back somehow, sometime. She elogged that she had removed it, but I mistakenly thought that it was already replaced. We don't use that camera much, so I'm not worried.
Take-away for the night: We need to do some more fine-tuning of the PRCL and MICH loops when we have arm resonance.
Koji sat with me for the first part of the night, and we looked back at the data from last week (elog 10727), as well as some fresh data from tonight. Looking at the spectra, we noticed that last week, and early in the evening today, I had a fairly broad peak centered around ~51Hz. We are not at all sure where this is coming from. The PRMI was locked on REFL 33 I&Q, and CARM and DARM were both on ALS comm and diff. This peak would repeat-ably come and go when I changed the CARM offset. At high arm powers (above a few tens? I don't know where exactly), the peak would show up. Move off resonance, and the peak goes away. However, later in the night, after an IFO realignment, I wasn't able to reproduce this effect. So. We aren't sure where it comes from, but it is visible only in the CARM spectra, so there's some definite feedback funny business going on.
Anyhow, after that, since I couldn't reproduce it, I went on to trying to hold the PRMI at high arm powers, but wasn't so successful. I would reduce the CARM offset, and instead of a 50Hz peak, I would get broadband noise in the PRMI error signals, that would eventually also couple in to the CARM (but not DARM) error signal, and I would lose PRMI lock. I measured the PRCL and MICH transfer functions while the arms were at some few units of power, and found that while MICH was fine, PRCL was losing too much phase at 100Hz, so I took away the FM3 boost. This helped, but not enough. I had 1's in the triggering matrix for TRX and TRY to both PRCL and MICH, so that even if POP22 went low, if the arms were still locked then the PRMI wouldn't lose lock unnecessarily, but I was still having trouble. In an effort to get around this, I transitioned PRMI over to REFL 165 I&Q.
While the arms were held around powers of 2ish, I readjusted the REFL 165 demod phase. I found it set to 150 deg, but 75 deg is better for PRMI locking with the arms. For either acquiring or transitioning from REFL33, I would use REFL165I * -1.5 for PRCL, and REFL 165Q * 0.75 for MICH. (Actually, I was using -2 for REFL165I->PRCL, and +0.9 for REFL165Q->MICH, but I had to lower the servo gains, so doing some a posteriori math gives me -1.5 and +0.75 for what my matrix elements should have been, if I wanted to leave my servo gains at 2.4 for MICH and -0.02 for PRCL.) I don't always acquire on REFL165, and if it's taking a while I'll go back to putting 1's in the REFL33 I&Q matrix elements and then make the transition.
With PRMI on REFL 165 I&Q, I no longer had any trouble keeping the PRMI locked at arbitrarily high arm powers. I was still using 1*POP22I + 1*TRX + 1*TRY for triggering PRCL and MICH. My thresholds were 50 up, 0.1 down. The idea is that even if POP goes low (which we've seen about halfway up the CARM resonance), if we're getting some power recycling and the arms are above 1ish, then that means that the PRMI is still locked and we shouldn't un-trigger anything. I didn't try switching over to POP110 for triggering, because POP22 was working fine.
Earlier in the night, Koji and I had seen brief linear regions in POX and POY, as well as some of the REFL signals when we passed quickly through the CARM resonance. I don't have plots of these, but they should be easy to reproduce tomorrow night. Koji tried a few times to blend in some POY to the CARM error signal, but we were not ever successful with that. But, since we can see the PDH-y looking regions, there may be some hope, especially if Q tells us about his super secret new CESAR plan.
Okay, I'm clearly too tired to be writing, but here are some plots. The message from these is that the PRMI loops are causing us to fluctuate wildly in arm transmission power. We should fix this, since it won't go away by getting off of ALS. The plots are from a time when I had the PRMI locked on REFL165, and CARM and DARM were still on ALS comm and diff. All 3 of these colored plots have the same x-axis. They should really be one giant stacked plot.
Also, bonus plot of a time when the arm powers went almost to 200:
Just to get our day started right, we tweaked up the alignment of the Ygreen to the Yarm (after IR alignment), and also touched up the X beatnote alignment on the PSL table. Ran the LSC offsets script, and then started locking.
All of the locking tonight has been based on CARM and DARM held on ALS comm/diff, and PRMI held on REFL165. Today, CARM was actuated using MC2. No special reason for the switch from ETMs. The AS port is noticeably darker when using REFL165 instead of REFL33.
Around 12:33am(ish), we were able to hold the arms at powers of about 100, for almost a minute. The fluctuations were at least 50% of that value, but the average was pretty high. Exciting.
Q and I tried a few times to engage the AO path while the arms were held at these high powers. Q hopefully remembers what the gain and sign values were where we lost lock. We didn't pursue this very far, since I was seeing the 50Hz oscillation that Koji and I saw the other day. I increased the CARM gain from 6 to 10, and that seemed to help significantly. Also, messing with the PRMI loops a bit helped. Q increased the pole frequency in FM 5 for both MICH and PRCL from 2k to 3k. While he had Foton open, he made sure that all of the LSC DoF filters use the z:p notation.
I then did a few trials of trying to transition CARM over to normalized REFL11I. Now that I'm typing, it occurs to me that I should have checked REFL11's demod phase. Ooops. Anyhow, using the phase that was in there, I turned on a cal line pushing on ETMs CARM, and found that using -0.002*REFL11I / (TRX + TRY) was the right set of elements. I also put an offset of 0.05 into the CARM CESAR RF place, and started moving. I tried several times, but never got past about 30% normalized REFL11 and 70% ALS comm.
During these trials, Q and I worked also on tweaking up the PRMI lock. As mentioned last night, PRCL FM3 eats too much phase (~30deg at 100Hz!), so I don't turn that on ever. But, I do turn on FM1 (which is new tonight), FM2, 6, 8 and 9. FM8 is a flat gain of 0.6 that I use so I can have higher gain to make acquisition faster, but immediately turn the gain down to keep the loop in the center of the phase bubble. MICH needed a lowpass, so in addition to FM2, I am now also triggering FM 8, which is a 400Hz lowpass that was already in there.
Now, my MICH gain is 2.4, with +0.75*REFL165Q, and PRCL gain is -0.02 with -3*REFL165I. Triggering for both MICH and PRCL is 1*POP22I + 5*TRX with 50 up, 0.1 down.
In my latest set of locks, I have been losing lock semi-regularly due to a 100Hz oscillation in either the PRCL or MICH loops. If I watch the spectra, most times I take a step in CARM offset reduction, I get a broad peak in both the MICH and PRCL error signals. Most of the time, I stay locked, and the oscillation dies away. Sometimes though it is large enough to put me out of lock. I'm not sure yet where this is coming from, but I think it's the next thing that needs fixing.
Here is a shot of the spectra just as one of these 100Hz oscillations shows up. The dashed traces are the nominal error signals when I'm sitting at some CARM offset, and the solid traces are just after a step has been made. The glitch is only happening in the PRMI, not CARM and DARM.
We attempted some of the same old CARM offset reduction tonight, but from the other direction. (We have no direct knowledge of which is the spring and which is the anti-spring side)
We we able to get to, and sit at, arm powers on the order of 5. Really, we kind of wanted just to push things to try and inform our current ideas of what our limiting factor is, so as to appropriately expend our efforts.
We took many digital CARM OLTFs at different offsets; it never really looked like a burgeoning pole was about to make things unstable. The low frequency OLTF data had bad SNR, so it wasn't clear if we were losing gain there. We weren't at arm powers where we would expect the DC transmission curve to flatten out yet, from simulations (which is above a few tens).
My impression from at least our last lock loss was a DARM excursion. However, using the DRMI won't get rid of the second two points.
Other thoughts from talking with Rana earlier:
Also, Q and I squished on the suspension connectors earlier tonight. MC2 was going wonky, which we feared might be because we were in that area working on Chiara earlier. Then, after squishing the MC connectors, the PRM started misbehaving, so we went and gave all the corner suspension connectors another squish. No suspension glitching problems since then.
Also, we found the PRM OL off and turned it back on. The ETMY was swinging a lot after lock loss, so we set its SUSPOS damping gain to match the ETMX and it stopped swinging so much.
Next up: more of the same, make this sequence more stable, turn on CARM OSC and watch the LOCKI outputs while we slowly ramp between signals.
Also, what should be the sign of the CARM offset ???
There are several things at this point that we know we need to look into:
* Simulate an arm sweep, up to many orders of the sidebands, to see how close to the carrier resonance any sideband resonances might be. If something like the 4th order sideband resonates, and then beats with a 1st order sideband, is that signal big enough to disturb our 3f locking of the PRMI / DRMI? We want to be holding the arms off resonance with ALS closer to the carrier than any "important" sideband resonances (where the definition of "important" is still undetermined). (Simulation)
I have done a sweep of CARM, while looking at the fields inside of one arm (I've chosen the Xarm), to see where any resonances might be, that could be causing us trouble in keeping the PRMI locked as we bring the arms into resonance.
Since Gabriele pointed out to me that we're using the 3x55MHz signal for locking, we should be most concerned about resonances of the higher orders of 55, and not of 11. So, on this plot, I have up to the 6th order 55 MHz sidebands, which are 332 MHz. Although the Matlab default color chart has wrapped around, it's clear that the carrier is the carrier, and the +4f2, which is the same blue, is not the giant central peak. So, it's kind of clear which trace is which, even though the legend colors are degenerate. Also, the main point that I want to show here is that there is nothing going on near the carrier, with any relevant amplitude. The nearest things are the plus and minus 55 MHz sidebands themselves, and they're more than 50 nm away from the carrier.
Recalling from elog 9122, the PRFPMI and DRFPMI linewidths are about 40pm. 50pm away from the resonant point is ~1/10 the power, and 100pm away from the resonant point is ~1/100 the power. So, 50 nm is a looooong ways away.
Just for kicks, here is a plot of all the resonances of the 1f and 2f modulation frequencies, up to 30*f1, which is the same 6*f2:
The resonances which are "close" to the carrier are the 9th order 11 MHz sidebands, and they're 280pm from the carrier, so twice as far as we need to be, to get our arm powers to ~1/100 of the maximum, and, they're a factor of ~1e4 smaller than the carrier.
The interferometer can nearly be locked again. I was unable to fully hand off control from ALS-->RF, I suspect I may be using the wrong sign on the AO path (or some such other sub-optimal CM board settings). I'll hook up the SR785 and take some TFs tomorrow, that should give more insight into what's what. With the arms held off resonance, the PRMI acquires lock nearly instantly (REFL165 I for PRCL, REFL165 Q for MICH), and can stay locked nearly indefinitely, which is what I need so I can get the RF lock going. However the sensing matrix (for vertex DoFs, arms held off resonance) still makes no sense to me. The MICH loop has ~50 Hz UGF and the PRCL loop ~150 Hz. I think the MICH loop shape can be optimized a little for better low frequency suppression, but this isn't the show-stopper at the moment. For record-keeping, the ALS performance was excellent and other subsystems were nominal tonight.
Pity really, I was hoping to make it much further tonight. I think I'll have to go back to the high BW POX/POY lock, and also check out the conversion efficiency / noise of the daughter board on the REFL11 demod board. Compared to before my work on the RF source, the demod phase for the PRMI lock using REFL11 as an error signal has basically necessitated a change of the digital demod phase by 180 degrees - so I made the appropriate polarity changes in the CM_SLOW and AO paths (the assumption is that CARM in REFL11 would require the same change in digital demod phase, and I think this is a reasonable assumption - indeed, with the arm powers somewhat stable ~100, if I look at the PDH signal in REFL11 I and Q, it does seem to show up largely in the I quadrature (pre digital phase rotation). Anyway, with so many weird effects (wonky PRM suspension, strange PRMI sensing etc etc, who knows what's going on. This will take a systematic effort.
I defer the electronics characterization for the daytime (if I feel like I need it tomorrow I'll do it, else. Koji has said he can do it on Friday).
I was unable to fully hand off control from ALS-->RF, I suspect I may be using the wrong sign on the AO path (or some such other sub-optimal CM board settings). I'll hook up the SR785 and take some TFs tomorrow, that should give more insight into what's what.
I need to think a little bit about the ASC commissioning strategy. On the positive side
Things to think about:
I am inclined to believe that the arm cavity losses are such that the IFO is overcoupled. Some calculations, validated with Finesse modeling also suggest that there isn't a sign change for the CARM error signal when the IFO goes from being undercoupled to overcoupled, but I may have made a mistake here?
Thoughts from others?
After the QPD fix, both arms report consistent buildup - see Attachment #1. The peak values touch ~250, corresponding to a PRG of ~13. The IFO becomes critically coupled at PRG=15. I am finding that the 3f signal offsets are changing as a function of the CARM offset, and this could be responsible for the lock breaking as I approach 0 CARM offset. I found that I could maintain a more stable and deterministic transition to zero CARM offset by dynamically adjusting the 3f PRCL error signal offset to keep the REFL11 signal approximately at 0. Some shaking seems to have commenced so I am breaking for now.
Note that I find scattered throughout the elog references to a similar problem of the PRMI losing lock as the CARM offset is reduced, e.g. here. But haven't stumbled across what the resolution was, the PRFPMI could be locked pretty easily in 2015 I remember.
Today Angelina and I looked at the PRM OL with an eye towards installing a 2nd QPD. We want to try out using 2 QPDs for a single optic to see if theres a way to make a linear combination of them to reduce the sensitivity to jitter of the HeNe laser or acoustic noise on the table.
The power supply for the HeNe was gone, so I took one from the SP table.
There are WAY too many optics in use to get the beam from the HeNe into the vacuum and then back out. What we want is 1 steering mirror after the laser and then 1 steering mirror before the QPD. Even though there are rumors that this is impossible, I checked today and in fact it is very, very possible.
More optics = more noise = bad.
The returning spot diameter on the qpd ~10 mm. In order to reduce the spot size I moved the f 1145 mm lens toward the PRM ~ 25 cm. The spot size was reduced to ~8 mm, 3200 counts.
I'll try to find an other lens tomorrow.
Atm 1, PRM oplev inward path with 2 lens solution: 14 cm gap between F 1145 and F 1545 mm lenses.
Atm 2, The PRM beam size 3 mm and the beam quality is still bad. The BS path only needed alignment.
The PRM sus gains checked OK
All other suspension oplev gains setting were checked out OK
We've aligned the guiderod and wire standoff to the PRM, each partly. They have both been aligned to the correct distance above the scribe lines, but they have not yet been centered forward/backward along the thickness of the optic. So, we're working on it...
2" G&H mirror is installed on a DLC mount just in front of the BS. I had to remove one of the 4 BS dog clamps, so we must put it back when we are finished with this test.
I aligned the G&H mirror such that the reflected beam is overlapped with the incident beam, and I aligned the PRM such that the regular REFL beam is retro-reflected. This is the same as getting the beam bouncing off the PRM back to the G&H to be overlapped.
I then saw flashes of the cavity, when I held a card with a hole in the cavity, so the beam was going through a small aperture in the card, but I still saw flashes. I was not able to see flashes on the IR card transmitted through the G&H mirror.
I also cannot see any flashes or scattered light on the face of PR2 camera.
I do, however, see flashes on the face of the PRM. Movie saved, will post soonly.
Light is coming out of REFL on the AS table, but it's clipped somewhere....needs investigation/work before we can lock.
I also didn't see anything at the POP port with a card, but I'm hopeful that perhaps with a camera I'll see something.
Dang it. I didn't confirm that the movie was good, just that it was there. It's corrupted or something, and won't play. I'll just have to make a new movie today after I realign the cavity.
* Put 2" G&H mirror into BS chamber, in front of BS.
* Align it, lock cavity using an existing REFL PD.
* Align POP setup so I can use POP camera to take image of transmitted cavity mode, and actually take that image.
* Take image of face of PR2.
* Measure finesse of cavity using POP, or a Thorlabs PD at POP (looking at transmission through PR2) by scanning PRM, and infer cavity gain....compare with values in elog 7905.
* If time / inclination allow, take beam scan measurements of the REFL port.
I will not be able to do as was done in elog 6421 to look at the beam size at POP for non-resonating beams. I expect ~0.1uW of light at POP in the non-resonant case: 100mW * 5.5% * 20ppm = 0.11microwatts.
Jenne and I noticed high pitch sound from our acoustic interferometer noise diagnostic system.
The frequency of this narrow band noise was 1256Hz, which is enough close to twice of the PRM violin mode freq.
After putting notch filter at 1256+/-25Hz at the violin filters, the noise is gone. Just in case I copied the same filters to all of the test masses.
Later, I found that the 4th violin modes are excited. Additional notch filters were added to "vio3" filter bank to mitigate the oscillation.
I have turned off the 3.2Hz res gains in the PRC ASC loops, since those seem to make the loops unstable.
Right now the pitch gain is -0.001, with FM1,3,9 on. Yaw gain is -0.004, with FM1,3,9 on.
Pitch gain can't increase by factor of 2 without oscillating.
I tried to take transfer functions, but I think the ASC situation is really confusing, since I have OSEM damping, oplev damping, and this POP QPD damping on the PRM. It's hard to get coherence without knocking the PRC out of lock, and it keeps looking like my gain is 0dB, with a phase of 0 degrees, from ~1 Hz to ~10 Hz. Outside that range I haven't gotten any coherence. Moral of the story is, I'm kind of puzzled.
Anyhow, as it is right now, the ASC helps a bit, but not a whole lot. I increased the trigger ON value, so that it shouldn't kick the PRM so much. I wish that I had implemented a delay in the trigger, but I'm not in the mood to mess with the simulink diagrams right now.
Ignoring the OSEM damping loops, the oplev servo loops make it so that the POP ASC loops do not see a simple pendulum plant, but instead see the closed loop response. Since the filter in the OL bank is proportional to f, this means that the open loop gain (OLG):
Which means that the CLG that the ASC sees is going to dip below unity in the band where the OL is on. For example, if the OL loop has a UGF of 5 Hz, it also has a lower UGF of ~0.15 Hz, which means that the ASC needs to know about this modified plant in this band.
For i/eLIGO, we dealt with this in this way: anti-OL in iLIGO
A series of measurements / calculations for the PRM ASC characterization and servo design
1) Actuator characterization
The actuator responses of the PRM in pitch and yaw were measured (attachment figure 1). I believed the calibration of the oplev QPD to be
1 count/urad. The oplev servo loops were turned off at the FM inputs, and the filter banks were turned off so that the response has the open
loop transfer function except for the servo filter.
The measured transfer functions were fitted with LISO. The LISO results (c.f. the source codes) were shown in the figure. The responses also
include the 60Hz comb filter present in the input filters. The responses are well approximated by the single pendulum with f0 of 0.6-0.8 and q of 3.5 and 6.3.
From this measurement, the actuator responses of the PRM at DC are estimated to be 2.2 urad/cnt and 1.8 urad/cnt in pitch and yaw, respectively.
2) Sensor response of the POP QPD
As we already know how the actuators respond, the QPD optical gain can be characterized by measuring the actuator response of the QPD
(attachment figure 2). The QPD signals are such noisy that the response above 1Hz can't be measured with sufficient coherence. Below 1Hz,
the response is well represented by the actuator response measured with the oplev. From this measurement, the optical gains of the QPD
with respect to the PRM motion are 650 cnt/urad and 350 cnt/urad.
3) Open loop transfer function of the current ASC servo
By combining the above information with the servo setting of the servo filters, the open loop transfer functions of the PRM QPD ASC loops
were estimated (attachment figure 3). Actually the expected suppression of the fluctuation is poor. The yaw loop seems to have
too low gain, but in fact increasing gain is not so beneficial as there is no reasonable phase margin at higher frequency.
With the estimated openloop transfer functions and the measured free-running angular fluctuation, the suppressed angular spectra can be
estimated (attachment figure 4). This tells us that the suppression of the angular noise at around 3Hz is not sufficient in both pitch and yaw.
As there is no mechanical resonance in the actuator response at the frequency, intentional placement of poles and zeros in the servo filter is necessary.
4) Newly designed ASC filter
Here is the new design of the QPD ASC servo (attachment figure 5). The target upper UGF is 10Hz with the phase margin of 50 to 60deg.
The servo is AC coupled so that we still can tweak the alignment of the mirror.
As this servo is conditionally stable, at first we should close the loops with stable filter and then some boosts should be turned on.
Estimated suppressed fluctuation is shown in the attachment figure 6. We can see that the fluctuation was made well white between 0.5Hz to 10Hz.
The filter design is shown as follows:
FM1: zero at 0Hz, pole at 2000Hz, gain at 2000Hz = 2000
zero: f: 0.5Hz q: 1 / 4.5Hz, q: 1 / f: 1Hz, q: 3
pole: f: 2Hz q: 3 / f: 2.7Hz, q: 2 / f: 1Hz, q: 15
FM9: (HF Roll-off)
pole: f: 40Hz q: 1.7
Servo gain: -0.028
FM1: zero at 0Hz, pole at 2000Hz, gain at 2000Hz = 2000
zero: f: 0.7Hz q: 2 / 3Hz, q: 7 / f: 2Hz, q: 6
pole: f: 1.02Hz q: 10 / f: 4.5Hz, q: 0.8 / f: 1.5Hz, q: 10
FM9: (HF Roll-off)
pole: f: 40Hz q: 1.7
Servo gain: -0.0132
Yesterday afternoon Paco and I measured the PRM L2P transfer function. We drove C1:SUS-PRM_LSC_EXC with a white noise in the 0-10 Hz band (effectively a white, longitudinal force applied to the suspension) and read out the pitch response in C1:SUS-PRM_OL_PIT_OUT. The local damping was left on during the measurement. Each FFT segment in our measurement is 32 sec and we used 8 non-overlapping segments for each measurement. The empirically determined results are also compared with the Fisher matrix estimation (similar to elog:16373).
Fig. 1 shows one example of the measured L2P transfer function. The gray traces are measurement data and shaded region the corresponding uncertainty. The olive trace is the best fit model.
Note that for a single-stage suspension, the ideal L2P TF should have two zeros at DC and two pairs of complex poles for the length and pitch resonances, respectively. We found the two resonances at around 1 Hz from the fitting as expected. However, the zeros were not at DC as the ideal, theoretical model suggested. Instead, we found a pair of right-half plane zeros in order to explain the measurement results. If we cast such a pair of right-half plane zeros into (f, Q) pair, it means a negative value of Q. This means the system does not have the minimum phase delay and suggests some dirty cross-coupling exists, which might not be surprising.
Fig. 2 compares the distribution of the fitting results for 4 different measurements (4 red crosses) and the analytical error estimation obtained using the Fisher matrix (the gray contours; the inner one is the 1-sigma region and the outer one the 3-sigma region). The Fisher matrix appears to underestimate the scattering from this experiment, yet it does capture the correlation between different parameters (the frequencies and quality factors of the two resonances).
One caveat though is that the fitting routine is not especially robust. We used the vectfit routine w/ human intervening to get some initial guesses of the model. We then used a standard scipy least-sq routine to find the maximal likelihood estimator of the restricted model (with fixed number of zeros and poles; here 2 complex zeros and 4 complex poles). The initial guess for the scipy routine was obtained from the vectfit model.
Fig. 3 shows how we may shape our excitation PSD to maximize the Fisher information while keeping the RMS force applied to the PRM suspension fixed. In this case the result is very intuitive. We simply concentrate our drive around the resonance at ~ 1 Hz, focusing on locations where we initially have good SNR. So at least code is not suggesting something crazy...
Fig. 4 then shows how the new uncertainty (3-sigma contours) should change as we optimize our excitation. Basically one iteration (from gray to olive) is sufficient here.
We will find a time very recently to repeat the measurement with the optimized injection spectrum.
Edit 7.30pm: I have managed to recover Y-arm in air locking, and the transmission is up at ~0.6 again which is what we were seeing prior to touching anything on the BS-PRM table, so it looks like the tip-tilt has not gone badly astray... I have also restored the Satellite boxes so that both PRM and SRM have their designated boxes
Tonight, and during last week's locking, we noticed something intermittently kicking the PRM. I've determined that PRM's LR OSEM is problematic again. The signal is coming in and out, which kicks the OSEM damping loops. I've had the watchdog tripped for a little bit, and here's the last ten minutes of the free swinging OSEM signal:
Here's the hour trend of the PRM OSEMS over the last 7 days a plot of just LR since the fix on the 9th of September.
It looks like it started misbehaving again on the evening of the 5th, which was right when we were trying to lock... Did we somehow jostle the suspension hard enough to knock the foil cap back into a bad spot?
It started here
Perhaps the problem is electrical? The attached plot shows a downward trend for the LR sensor output over the past 20 days that is not visible in any of the other 4 sensor signals. The Al foil was shorting the electrical contacts for nearly 2 months, so perhaps some part of the driver circuit needs to be replaced? If so a Satellite Box swap should tell us more, I will switch the PRM and SRM satellite boxes. It could also be a dying LED on the OSEM itself I suppose. If we are accessing the chamber, we should come up with a more robust insulating cap solution for the OSEMs rather than this hacky Al foil + kapton arrangement.
The PRM and SRM Satellite boxes have been switched for the time being. I had to adjust some of the damping loop gains for both PRM and SRM and also the PRM input matrix to achieve stable damping as the PRM Satellite box has a Side sensor which reads out 0-10V as opposed to the 0-2V that is usually the case. Furthermore, the output of the LR sensor going into the input matrix has been turned off.
Looks like what were PRM problems are now seen in the SRM channels, while PRM itself seems well behaved. This supports the hypothesis that the satellite box is problematic, rather than any in-vacuum shenanigans.
Eric noted in this elog that when this problem was first noticed, switching Satellite boxes didn't seem to fix the problem. I think that the original problem was that the Al foil shorted the contacts on the back of the OSEM. Presumably, running the current driver with (close to) 0 load over 2 months damaged that part of the Satellite box circuitry, which lead to the subsequent observations of glitchy behaviour after the pumpdown. Which begs the question - what is the quick fix? Do we try swapping out the LM6321 in the LR LED current driver stage?
GV Edit Nov 2 2016: According to Rana, the load of the high speed current buffer LM6321 is 20 ohms (13 from the coil, and 7 from the wires between the Sat. Box and the coil). So, while the Al foil was shorting the coil, the buffer would still have seen at least 7 ohms of load resistance, not quite a short circuit. Moreover, the schematic suggests that that the kind of overvoltage protection scheme suggested in page 6 on the LM6321 datasheet has been employed. So it is becoming harder to believe that the problem lies with the output buffer. In any case, we have procured 20 of these discontinued ICs for debugging should we need them, and Steve is looking to buy some more. Ben Abbot will come by later in the afternoon to try and help us debug.
I've restored the gains to their old values, and measured the loop TFs.
[Suresh / Kiwamu]
We tried adjusting the OSEMs on PRM, but we didn't complete it due to a malfunction on the coils.
The UL and LL coils are not working correctly, the forces are weak.
Tomorrow we will look into the satellite box, which is one of the suspects.
During the adjustment we found that the POS excitation force was unequal in each sensor.
At the beginning we thought it's because of the difference of the sensitivity in each OSEM due to the bad OSEM orientations.
However it turned out that it comes from the actual force imbalance on each coil.
We checked the force of each coil by putting an offset (-2000 cnts) in each output digital filter and looked at the OSEM signals in time series.
The UL and LL coils are too weak and the responses are almost buried in the noise of the OSEMs in time series.
We briefly checked some analog electronics and found the DAC, AI board and deWhitening board were healthy.
We were able to see the right amount of voltage from the monitor pin on the front panel of the coil driver.
So something downstream are suspicious, including the satellite box, feedthrough and coils.
- - -
Although the coil issue, it could be worth trying to check the input matrix.
Adjustment of the PRM OSEMs are done. The coils turned out to be healthy.
The malfunction was fixed. It was because the UL OSEM was too deeply inserted and barely touching the AR surface of the mirror.
+ Excited POS at 6.5 Hz with an amplitude of 3000 cnts by the LOCKIN oscillator.
+ Looked at the signal of each sensor in frequency domain.
+ Maximized the excitation peak for each of the four face OSEMs by rotating them.
+ Minimized the excitation peak in the SIDE signal by rotating it.
+ Adjusted the OSEM translational position so that they are in the midpoint of the OSEM range.
(POS sensitivity check)
From the view point of the matrix inversion, one thing we want to have is the equally sensitive face sensors and insensitive SIDE OSEM to the POS motion.
To check the success level of today's PRM adjusment, I ran swept sine measurements to take the transfer function from POS to each sensor.
The plots below are the results. The first figure is the one measured before the adjustment and the second plot is the one after the adjustment.
As shown in the plot, before the adjustment the sensitivity of OSEMs were very different and the SIDE OSEM is quite sensitive to the POS motion.
So PRM used be in an extremely bad situation.
After the adjustment, the plot became much better.
The four face sensors have almost the same sensitivity (within factor of 3) and the SIDE is quite insensitive to the POS motion.
Steve fixed the PRM oplev pointing. I turned on the loops and measured the OLG, then set the pitch and yaw gains such that the upper UGF was ~8Hz (motivated by Jenne's loop design in ELOG 9401)
I then measured the oplev spectra of the optics as they were aligned for PRMI. (OSEMs on, oplevs on, LSC off, and ASC off)
Next, Jenne and I need to fix the ASC loop such that it properly accounts for the oplev loop.
In an effort to stop the PRC from wiggling around so much, I recentered the POP QPD after maximizing the POPDC power when locked on carrier. The beam was basically off the QPD in yaw, and at half-range in pitch.
For the PRM, I aligned it until the arm flashes were maximized and the REFL camera showed a centered spot with dips happening during the arm pops. AS port was more messy since the Michelson alignment wasn't perfect, but the spots were both near the center of the cam and the SRM alignment maximized the wangy fringiness of the image as well as the angry cat meow sounds that the full IFO makes as heard through the DAFI (listening to POX).
On Monday, Osamu should be back and can help with doors and then alignment recovery and locking.
[Gautam, Steve, Johannes]
We put on the remaining heavy doors on the chambers (ITMY, ITMX,ETMX, in this order) this morning. On the ITMY and ETMX tables we placed old OpLev steering mirrors that are clean and baked as witness plates such that may one day provide some insight into dust accumulation on optics.
With the heavy doors on we confirmed that we were still able to lock both IFO arms and used the dither scripts to optimize the alignment. Following that we centered all OpLevs and aligned the X and Y green beams.
Radhika reported that the PRM UL OSEM PD is not responding. This PD has been identified to have a shorting problem, but the short existed only at the bias pin of the PD. We disconnected the bias voltages and the PD was working with no (=0V) bias.
It seems that it lost the signal about 8 days ago and the signal intermittently appeared and disappeared.
I suggested to Radhika to remove the Al foil suspecting the other pin of the PD is not shorting.
While looking over Koji's shoulder earlier, I noticed the big peak in the PRM yaw spectrum (and I was starting to get annoyed by the hum....the fibox is so useful in motivating tasks that otherwise get looked over!)
I used DTT's peak find feature (cursor tab, enable both cursors, select Peak X/Y as your 'statistic', set the 2 cursors to be on either side of the desired peak) to find the frequency of the PRM's violin mode. It is 627.75 Hz. I adjusted FM5 of the C1:SUS-PRM_LSC filter bank (the "violin" filter) to be centered around this frequency, with the start and stop freqs +\- 4Hz. I plotted the filter linearly in frequency to ensure that my target freq was not too close to either side of the bandstop. After loading and engaging the new filter, the hum slowly started to go away.
Note, for posterity: The bandstop used to be centered around ~645 Hz or so. I assume this is a copy-and-paste situation, where we hadn't gone through to check the exact frequency for each optic.
It seems that the PRM violin mode freqs shifted from 625-ish to 640Hz.
The peaks rang up because of the servo.
Once the notch freq was shifted to 640Hz, the violin mode started to decay.
Short summary of my Sat. Box. debugging activities over the last few days. Recall that the SRM Sat. Box has been plugged into the PRM suspension for a while now, while the SRM has just been hanging out with no electrical connections to its OSEMs.
As Steve mentioned, I had plugged in Ben's extremely useful tester box (I have added these to the 40m Electronics document sub-tree on the DCC) into the PRM Sat. Box and connected it to the CDS system over the weekend for observation. The problematic channel is LR. Judging by Steve's 2 day summary plots, LR looks fine. There is some unexplained behavior in the UR channel - but this is different from the glitchy behaviour we have seen in the LR channel in the past. Moreover, subsequent debugging activities did not suggest anything obviously wrong with this channel. So no changes were made to UR. I then pulled out the PRM sat.box for further diagnostics, and also, for comparison, the SRM sat. box which has been hooked up to the PRM suspension as we know this has been working without any issues.
Tracing out the voltages through the LED current driver circuit for the individual channels, and comparing the performance between PRM and SRM sat. boxes, I narrowed the problem down to a fault in either the LT1125CSW Quad Op-Amp IC or the LM6321M current driver IC in the LR channel. Specifically, I suspected the output of U3A (see Attachment #1) to be saturated, while all the other channels were fine. Looking at the spectrum at various points in the circuit with an SR785, I could not find significant difference between channels, or indeed, between the PRM/SRM boxes (up to 100kHz). So I decided to swap out both these ICs. Just replacing the OpAmp IC did not have any effect on the performance. But after swapping out the current buffer as well, the outputs of U3A and U11 matched those of the other channels. It is not clear to me what the mode of failure was, or if the problem is really fixed. I also checked to make sure that it was indeed the ICs that had failed, and not the various resistors/capacitors in the signal path. I have plugged in the PRM sat. box + tester box setup back into our CDS data acquisition for observation over a couple of days, but hopefully this does the job... I will update further details over the coming days.
I have restored control to PRM suspensions via the working SRM sat. box. The PRM Sat. Box and tester box are sitting near the BS/PRM chamber in the same configuration as Steve posted in his earlier elog for further diagnostics...
GV Edit 2230 hrs 7Nov2016: The signs from the last 6 hours has been good - see the attached minute trend plot. Usually, the glitches tend to show up in this sort of time frame. I am not quite ready to call the problem solved just yet, but I have restored the connections to the SRM suspension (the PRM and SRM Sat. Boxes are still switched). I've also briefly checked the SRM alignment, and am able to lock the DRMI, but the lock doesn't hold for more than a few seconds. I am leaving further investigations for tomorrow, let's see how the Sat. Box does overnight.
Looks like the PRM Sat. Box is now okay, no evidence of the kind of glitchy behaviour we are used to seeing in any of the 5 channels.
Now that we have increased the range of the AA to +/- 10 V I have increased the PRM side OSEM transimpedance from 29 kV/A to 161 kV/A by changing the R64 in the satellite box. The first attached plot shows the ADC input spectrum before and after the change with analog whitening turned off. The PD voltage readback went up from 0.75 to 4.2 V. The second attached plot shows the sensor, ADC, and projected shot noise with analog whitening turned on and compensated digitally. The ADC calibration is 20 V/ 32768 cts. The PRM damping loops are currently disabled.
I checked for oscillation by looking at the monitor point at the whitening board. There was no obvious oscillation on a scope - the signal was 20 mV p-p on 1 us scale which was very similar to the LL channel.