I re-did the Mathematica notebook according to the most current diagram (note to daytime self: attach .nb file!!!), and found that the denominator has changed, such that plugging in the new D=-A_refl*P_als*S_als gives the same
full-system closed loop gain of
where is the open loop gain, and the * indicates either the REFL or ALS portions of the system.
I have also plotted some things with Matlab, although I'm a little confused, and my daytime self needs to spend some more time thinking about this.
In the actuators (both for REFL and ALS), I include a pendulum, the digital anti-imaging filters that let us go from the 16kHz model to the 64kHz IOP and the analog anti-imaging filters after the DAC. Note to self: still need to include violin filters here.
For the servo gains, I copy the filters that we are using from Foton, and give them the same overall gain multiplier that is in the filter bank. For the ALS going through the CARM filter bank, this is FMs 1, 2, 3, 5, 6 with a gain of 15. For the RF (actually, POY here) going through the MC filter bank, this is FMs 4, 5, 7 with a gain of 0.08.
For the plants of each system, since this is still single arm lock, I just include a cavity pole (80kHz for ALS, 18kHz for REFL).
In the sensors (both for REFL and ALS), I include the analog anti-aliasing as well as the digital anti-aliasing to allow us to go from the 64kHz IOP to the 16kHz front end system. For the ALS I also include in the sensor the closed loop response of the phase tracker loop (H/(1-H), where H is the open loop gain of the phase tracker). For both sensors, I also include a semi-arbitrary number to make the full single-loop open loop gain have a UGF of 200Hz. In the ALS sensor, I also include a minus sign to make the full open loop gain have the correct phase.
Here I plot the open loop gains of the individual single loops, as well as the open loop gain of the full system (Hals + Hrefl - Hals*Hrefl). I feel like I must be missing a minus sign in my ALS loop, but I don't know where, and my nighttime brain doesn't want to just throw in minus signs without knowing why. That will affect how the final ALSfool (blue trace) looks, so maybe it's not really as crazy as it looks right now.
Also, I was trying to explain to myself why we are getting the shape that we are in our measurements of the cancellation (http://nodus.ligo.caltech.edu:8080/40m/11041). But, I can't. Below are the plots of the transfer functions from either point 9 or 10 (before or after the G_refl) to point 5, which is the ALS error point. The measurement in elog 11041 should correspond to the blue trace here. For these traces, the decoupling is set to just (-A_refl), although there aren't any noticeable changes in the shape if I just set D=0. If we start with the assumption that D=0, the shape and magnitude are basically identical to this plot, and then as we make D=-A_refl P_als S_als, the transfer functions both go to zero.
So. Why is it that with no decoupling, the transfer function from 10 to 5 is tiny? Why do the shapes plotted below look nothing at all like the measured cancellation shape? Daytime brain needs to think some more.
Here is an updated cartoon, with the ALS sensor explicitly shown as the beatbox times the closed loop response of the phase tracker servo.
The most important transfer functions are written on the diagram. Others can be extracted from the attached Mathematica file (which corresponds to this diagram).
For tonight's experiment, I re-installed the delay line cable and changed the attenuation to 10dB for the 55MHz modulation.
I quickly locked the PLL and checked that the modulation is the ratio of the field strength between the worst (19ns) and best
case (28ns) is 31dB, that is ~35 times reduction.
The modulation setting was reverted.
Demod phase for REFL11/33/55/165 and AS55 were reverted to the previous numbers too.
I'm playing around with the lastest ALS fool feedforward on the Yarm, and I like what I'm seeing.
First, I verified that I could reproduce the TF shapes in ELOG 11041, which I was able to do with a gain of +9.3 and FMs 5 and 6 in the FF module.
Then, I locked the arm on ALS with full bandwidth, and on POY with the settings currently used the MC module, and took their spectra as references. (I put an excitation on the arm at 443Hz to line them up to the same arbitrary units.)
Then, with ALS at its usual 100Hz UGF and boosts on, the Fool path on, and the MC FM set to trigger on/off at 0.8/0.5, I slowly brought ALS towards zero offset and saw it pop right into resonance. I then manually triggered the PDH boosts.
Here are some spectra, showing that, with the Fool path on:
Once the PDH loop is running, the ALS loop can be switched out at the CARM FM output without much of an effect beyond a small kick.
However, looking at the loop shapes, maybe I got lucky here. I took the usual injection TFs at the MC FM, the CARM FM, and at ETMY, to get the overall OLG; all of them have >0.9 coherence pretty much everywhere except the first two points.
As desired, the PDH loop looks pretty normal.
I have no intuition about how the fooled CARM loop should look, since this is even more complicated than a two-loop system.
I don't currently know what is causing the odd feature in the overall at ~50Hz, and it spooked me out when I saw the multiple UGF crossings. The only thing I could think of happening there is the pole in the ALS phase tracker boost. I turned it off, and remeasured; the feature persists...
To wrap it up, here's something I think is pretty cool. Here's what happens when I give ETMY a 1000 count position step impulse (no ramp). [Here, CARM is on ALS with G=12, but only FM5 on]
Although the arm was controlled with IR before the kick, ALS maintained control. As soon as ALS brought the arm back towards resonance, the PDH loop picked it right back up.
Some random notes:
DTT data is attached, in case it's useful to anyone!
Koji raised a good question about the step response I wrote about. Namely, if the UGF of the ALS servo is around 100Hz, we would expect it to settle with a characteristic time on the order of tens of milliseconds, not seconds, as was seen in the plot I posted.
I claim that the reason for the slow response was the fact that the feedforward FM stayed on after the kick, despite the MC filter bank being triggered off. Since it has filters that look like a oscillator at 1Hz, the ringdown or exponential decay of this filter's output in response to the large impulsive output of the PDH control signal just before being triggered down would slowly push the ALS error signal around through the feedforward path.
Given this reasoning, this should be helped by adding output triggering to the FF filter that uses the MC trigger matrix row, as I wanted to do anyways. I have now added this into the LSC model (as well as DQ at 2kHz for the MC_CTRL_FF_OUT), and the impulse response is indeed much quicker.
In the following plot, I hit ETMY with a five sample, 5000 amplitude, impulse (rather than a step, as I did yesterday). The system comes back to PDH lock after ~40ms.
We tried several times tonight to engage the Fool path with the PRFPMI. No success.
First, we locked the arms on ALS, in CARM/DARM mode, and measured the cancellation ability, to make sure that the filter shape and gain was set correctly. For the PRFPMI, it was okay using the same shape as the single arm case, but the gain was +20.0. There might be a bit more cancellation to be gained if we adjust the shape at the ~1dB level, but we're already able to get 20dB of cancellation, so we decided that would be enough to give things a try. To get this much cancellation, we set the phase tracker loops to both have 2kHz UGFs, almost exactly. We should implement a UGF servo, or the amplitude method version of that as Koji suggested ages ago, so that the phase tracker is always at the same place.
I don't think that REFL 11 is seeing as much CARM as I expect. We ended up switching over to linearized REFL55 for our attempts. When we're close to zero CARM offset, the arms are constantly flashing through resonance, and we get the loud buzzing. REFL11 doesn't seem to see any of this, even though we should be close enough to see some PDH action. REFL55 does change as we get closer to resonance, so I think it's seeing some real CARM stuff.
We tried engaging the Fool, but I don't think it did anything too useful. We need to make an estimate of what we expect our gain of the REFL loop to be - or at least the sign.
The PRMI is still not stable enough. It keeps falling out of lock when we get to high-ish arm powers. Not good. More brain power tomorrow on the modulation cancellation issue.
Perhaps if things are stable at moderate arm powers, we can use an excitation to line up the ALS vs. REFL error signals, and then watch the noises of them change as we move around in CARM offset. This should tell us when the linearized REFL signal is quiet enough that it's worth triggering and trying to transfer over.
The last lockloss tonight, there was something funny going on, that we can't explain. Even though both arms were locked on the CARM/DARM combined ALS signals, beatx doesn't see the giant oscillation that causes carm to lose lock until much later. Fool was trying to do something, but that should affect both als individal signals in the same way. Mystery.
Last Wednesday we tried PRMI 3f modulation cancellation. Under the 3f modulation cancellation, the PRMI could not be locked
with REFL signals, and the PRCL signal was just barely sufficient to lock PRCL with help of AS55Q MICH.
- The PRCL signal level in REFL33 was reduced by factor of 20 compared with the conventional modulation setting.
=> The 3f modulation cancellation does not chage the level of 11/22MHz sidebands, it is expected that REFL33 signal
has no significant change of the signal level. But it does. If we change the relative phase between the modulations
at 11 and 55MHz, the signal level is recovered by factor of 5. Therefore something related to 55MHz modulation
(55MHz x 22MHz, or 44MHz x 11MHz) was contributing more than -11MHzx22MHz.
- Under the 3f demodulation cancellation, MICH signal in the REFL ports were extremely weak and there was
no hope to use it for any feedback control.
- WIth the PRMI locking by REFL33->PRCL and AS55Q->MICH, the sensing matrix was measured. All of the REFL
ports however, showed extremely degenerate sensing matrix between MICH and PRCL.
This was enough confusing to us, and we didn't draw any useful information from these. Here are some ideas to
investigate what is happening in out optical and electrical system.
- One approach is to use as simple optical setup as possible to inspect our sensing systems. For example,
we may want to try PRX/PRY/XARM/YARM cavities to check the functions of the REFL diodes and qualitatively characterize
the sensing chain (Optical gain [W/m], noise level, demodulation phase) so that we can compare these with
an interferometer seinsing model.
- Another approach is to change the mdulation setting more freely and empirically try to find the characteristic
of our optical/electrical systems. e.g. change the relative modulation phase and/or 55MHz attenuation, and try to understand
how 11-22, 11-44, 22-55, 0-33 pairs are contributing the signal.
I have used Optickle to model the effect of changing the phase between the 11MHz and 55MHz EOMs. Also, I have looked at what modulation order is most significant (we hope it's -22*11 and -11*22).
First, so that we can compare these numbers more directly to measurements, I have included in the model the fraction of light that gets to each PD. I'm assuming that the Faraday is about 80% transmissive, but I don't know what the true number is. Here's a cartoon showing the attenuation factors.
EDIT, 26March2015, JCD: REFL path updated. See elog 11172.
To model the change in relative phase between the 11MHz and 55MHz modultions, I have held the 11MHz EOM stationary, and moved the 55MHz EOM. Since I needed an actual distance, I used a conversion factor,
The sensing matrix was measured at 143Hz. It has been corrected from mevans-meters to Newtons as the denominator.
The big thing to notice here is that the PRCL magnitude is not changing by a factor of 20. More like a factor of 2. BUT, I have not yet included the fact that Koji also reduced the 55MHz modulation amplitude by a factor of 3.
As for the mysterious degeneracy of all the REFL PDs, I think we need to take a more careful measurement. It's possible that we were seeing the real thing for REFL33, but the others don't seem to change in degeneracy with relative modulation phase.
Why does it even matter for the 3*f1 signal what is going on with the f2 modulation? Well, it appears that we are definitely being dominated by the 44*11 and 55*22 components.
To check this, I restricted Optickle to various orders of modulation (ex. up to second order only includes the [-22, -11, 0, 11, 22] MHz components), and plotted them all. The change in the signal between one trace and another is the contribution from that extra modulation order. The traces are only minutely different between orders, after the 5th order. So, since they're all overlapping with the 5th order trace anyway, I don't plot them.
EDIT: to clarify, when I said "up to X order", I meant up to that order in 11MHz sidebands. Optickle is applying the 11MHz and 55MHz modulations in the same way every time, but then I specify up to what order to include in the summation of different contributions to the field at a given port. So, for the "up to 2nd order", I only include cross terms that come from [-22, -11, 0, 11, 22] MHz. For the next order, I only include terms that come from [-33, -22, -11, 0, 11, 22, 33] MHz, etc. So, there are no 55MHz effects when I'm only including contributions up to 2nd order (since there is a maximum cross beatnote of 44MHz), but starting with 3rd order, I do start to see signals in, for example, REFL55 and AS55, since I get terms from -22*33 and -33*22. The first order in 55MHz (i.e. 55MHz*Carrier) only starts to show up when I calculate "up to 5th order" and above, since that includes [-55, -44, -33, -22, -11, 0, 11, 22, 33, 44, 55] MHz.
What happens if I reduce the 55MHz modulation depth by 10dB? Since we are dominated by 55MHz-related signals, the signal at REFL 33 goes down. The maximum change we could have seen for the REFL33 PRCL signal (difference between max of blue trace and min of orange trace) is a factor of 27.
Where are we on the x-axis of these plots, and where was the maximal cancellation place that Koji found? I need to re-check that part of the code tomorrow, to make sure that I've included all of my contributions from different components of the (field* field) matrix.
But, the moral of the story for tonight is that at least for the REFL33 signal, it's actually plausible that the optical gain went down by a factor of 20, and that the MICH and PRCL signals were degenerate. I suspect that the total cancellation place that Koji found was somewhere around 175 degrees on the x-axis of these plots and that our nominal place is around 0 deg - around there, both the magnitude and the phase situations are possible simultaneously.
About 5-10 minutes ago I just put in the modulation cancellation setup according to the recipe in http://nodus.ligo.caltech.edu:8080/40m/11032
[Jenne, EricQ, Rana]
We spent this evening measuring and thinking about our 3f signals, and the effect of the modulation cancellation.
I reinstalled the delay line box, and reduced the modulation depth of the 55MHz signal, so that we are in the state of modulation cancellation - there should be almost no power at 33MHz injected into the vacuum. I carefully tuned the demod phase of the REFL 11, 33 and 55 MHz PDs, and locked the PRMI on REFL55 I&Q. The signal in REFL 165 was very tiny, so as best as I could tell, the demod phase that Koji found last week was correct.
Here is a little record of what the demodulation phases should be, for the "nominal" and "cancellation" configurations, so that we don't have to continually use the time machine.
Also, here is the locking recipe for REFL55 I&Q in the cancellation configuration:
With this setup, we measured the sensing matrix. MICH had an excitation at 370.123 Hz with 8,000 counts to -ITMX+ITMY (this is about 0.3nm for each ITM), and PRCL had an excitation at 404.123 Hz with 50 counts to PRM. For tonight, here is a PDF of the peaks in DTT. The data is saved in /users/Templates/LSC_error_spectra/SensMat_PRMI_24Feb2015.xml.
Rana has proposed to us an idea for why the REFL 33 signal should be dominated by the 55*22 contribution, rather than -11*22. Eric is in the process of checking this out with a Mist model to see if it makes sense. Here's the gist:
Our Schnupp asymmetry is small (3.9cm, IIRC), so the transmission of the 11MHz signal out the dark port is small. This means that the finesse of the PRC for 11MHz isn't so huge. On the other hand, since 55MHz is a higher frequency, the transmission out the dark port is larger and is a closer match to the PRM transmission, so the finesse of the PRC for 55MHz is higher.
Since the magnitudes of the fields at the reflection port are not changing significantly, our PDH signals are being created by the relative phase of something which is anti-resonant (ex. carrier or 22MHz for sideband lock) vs. something which is resonant (ex. 11MHz or 55MHz). Since the finesse of the 55MHz signal is larger, the accumulated phase change is greater, so we expect a larger slope to our PDH signals that involve 55MHz as compared to those that use 11MHz.
If we are comparing the contributions between -11*22 and 55*22, they both include the anti-resonant 22MHz. So, the difference in the signal strengths comes directly from the difference in phase accumulation due to the variation in the dark port transmission.
EricQ had a thought, and while I have enough awake brain cells to report the thought, we're going to have to revisit it when more of our brains are awake. In either case, the transmission through the dark port is small compared to the transmission of the ITMs, so why don't the ITMs dominate the finesse calculation, and thus are the 11MHz and 55MHz really getting that much of a difference in finesse? To be checked out.
If the reflectivity of the front mirror is fixed (=PRM reflectivity), the finesse increases when the reflectivity of the end
mirror (=Compond mirror reflectivity) increases. i.e. 11MHz has higher finesse, 55MHz has lower finesse.
If the reflectivity of the front mirror is fixed, the amplitude gain of the cavity is higher when the reflectivity of the end mirror increases. i.e. 11MHz has higher gain, 55MHz has lower gain
The Schnupp asymmetry is definitely not an important parameter (no need for Koji to explode). It only serves to give us a bigger Q phase signal slope, but is not significant for the I phase signals.
The main anomaly is that the REFL33 optical gain (W/m) seems to have been reduced so much by the phase and amplitude adjustment of the 55 MHz modulation signal. One guess is that the true 3f signal is being made not by the (2*f1 - (-f1)) beat, but by some higher order beat. In addition to the usual RF fields produced by the 2 modulations, we must consider that the sidebands on sidebands produce intermodulation fields just after the EOM and so the fields with which we interrogate the PRMI are more complicated.
Jenne's Optickle calculation today should show us what the sensing matrix contribution is from each pair of fields, so that we can have a sensing matrix signal budget.
And now I've removed the delay line, and am in the process of reverting the demod phases, etc.
I have measured the sensing matrix for the PRMI at the REFL photodiodes for both the nominal configuration and the 33MHz cancellation configuration. The nominal configuration measurements do not compare well with those from November (http://nodus.ligo.caltech.edu:8080/40m/10701) which makes me unhappy . Both sets of nominal config reported below are from today, after tuning the demod phases and making sure the MICH and PRCL loops looked the same as yesterday (esp. overall gain). The cancellation config data is from last night.
Note that the magnitude for each photodiode is referred to its own "PD counts". Since the electronics are different for each PD, and that is not taken into account here, you cannot directly compare an element from one PD to an element from another PD. What you can do (which is most of what we need right now) is compare all the different measurements for a single photodiode.
So, what I'm apparently seeing is that the magnitudes of the sensing matrix signals that are made using 55MHz (i.e. everything but REFL11) change when we go into the cancellation configuration, but the phases of the sensing elements do not change significantly. Also, I am apparently seeing that REFL11 and REFL33 only have about 3 degrees of separation between the MICH and PRCL signals no matter what configuration is used. This doesn't make a lot of sense, since we know that we can lock robustly on REFL33I&Q (it's been sitting there happily as I write this elog), so it seems crazy that we could actually be so degenerate. Also, at the bottom of the elog that I wrote in November 2014, I show a sensing matrix where both REFL11 and REFL33 have about 45 degrees of separation between the MICH and PRCL signals.
I don't think I'm doing anything too crazy here, particularly with the phase. For a given PD and given DoF, I find the magnitude of the peaks of the I and Q signals, and just do atan2(Q-signal, I-signal)*180/pi, and those are the numbers that go in the phase columns above.
Before taking my measurements, I tuned up the demod phases for the PRMI-only case. I think REFL11 may have previously been tuned for CARM when the arms were held with ALS, but I don't really recall. Anyhow, now all 4 REFL PDs are tuned for PRMI-only.
This was done while the PRMI was locked with REFL 55 I&Q.
EDIT, 26Feb2015: Last night I mixed up the REFL11 and REFL33 new demod phases. Bold are the corrected version. Also, note that REFL33 was formerly tuned for PRCL in PRFPMI, which may be why it changed by ~10 degrees.
15.3 +/- 0.3
Here's the recipe for locking REFL 55 I&Q in the nominal modulation configuration. It's the same as the REFL33 I&Q lock that I was using today, except that for the REFL33 version, the matrix elements are both unity.
I'm working on some more modelling investigations of this whole situation. The main thing I wanted to do was to look at the complex field amplitudes / IFO reflectivity to see how the PDH signal is affected by different field components.
I still have plenty more to do, but I got a result which I though I should share. In addition to Jenne's simulation, I also see that between our "nominal" and "canceled" states as defined in Kojis ELOG 11036, there is a factor of ~20 difference in the PRCL signal in REFL33.
The plots below are kind of like "PDH Signal Budgets" of the two states.
Specifically, the reason our gain gets reduced is that, in the "canceled" state, the 44*11 and 55*22 products conspire to weaken the signal by having a slope opposite to the -11*22 type products. In contrast, in our "nominal" case, all of the products slope together.
However, this also predicts that the nominal REFL33 is more sensitive to Carrier*33 than to the signal we desire, -11*22. The only reason it ever worked seems to be the biggest contriubutor, the unexpected 44*11!
The "residual" trace is the difference of REFL33 and the sum of the field products shown, to justify that all relevant products had been included.
The simulation that produced this was set up to create 4 orders of modulation at each EOM, with 3 orders of sidebands on sidebands. The demodulation phase was taken by lining up a PRM excitation entirely along I, as we would do on the actual instrument.
MIST Simulation files attached!
So, my previous post suggested that 44*11 products might be the dominating signals in our "nominal" setup. I suppose that this could be not so bad, since it's not too unlike -11*22; the 11MHz field couples into the PRC and reflects with a rapidly changing phase around PRC resonance, and 44MHz is antiresonant, so it is a good local oscillator for REFL33.
However, it then occured to me that my previous HOM analysis only looked at the 11MHz and 55MHz sidebands.
When extending this to all of the sidebands within 55MHz, I discovered a troubling fact:
With the IFO parameters I have, the second spatial order +22MHz and fourth spatial order +44MHz fields almost exactly coresonate with the carrier in the PRFPMI! (DR, too)
If this is true, then any REFL33 signal seems to be in danger if we have an appreciable amount of these modes from, say, imperfect modematching.
On the other hand, we've been able to hold the PRMI with REFL33 when ALS is "on resonance," so I'm not sure what to think. (As a reminder, this analysis is done with analytic formulae for the complex reflectivities of the arm cavities and coupled recycling cavities as a function of CARM, spatial order and field frequency. Source is attached.)
It seems the Y arm geometry is to blame for this.
Maybe we should try to measure/confirm the Y arm g-factor...
Ok... This is what I was afraid of, and it seems true.
i.e. the relation ship of the modulations for the 3f cancellation is making the PRCL signals cancel each other.
It agrees with Anamaria's analysis that 11x44 is the strongest component in aLIGO 27MHz signal.
00x33 has the order of
11x22 has the order of
11x44 has the order of
22x55 has the order of
Therefore 11x44 is inherently the strongest contribution at 33MHz.
(And then, of couse, the signal amplitudes have additional dependences on the reflectivity
and the gain of the IFO at each freq)
If we believe this result, it may be difficult to exploit the benefit of the signals under the 3f cancellation.
We probably have to go back to the original idea of cancelling the 3f modulation by adding 3f modulation.
(i.e. Produce 33MHz signal by freq tripling, add this signal to Kiwamu's box to elliminate 3f.)
No more illegal power supply at the LSC rack
The amplifiers are now being powered by the rack power supply through fuse blocks.
To make new connections, I shutdown the +/-15 V low noise power supplies. They were turned back ON after the work.
If so, or if not but you care about the signal that passes through these amplifiers, I suggest you remove this temporary power supply and wire the power from the rack power supplies through the fuse blocks and possibly use a voltage regulator.
In 24 hours, that power supply will be disconnected and the wires snipped if they are still there.
I have clarified my elog from last night to indicate that the sensing matrix in the "33MHz cancellation" configuration was measured with the PRMI held on REFL55 I&Q.
Also, I just re-read my control room notes from yesterday, and I typed the wrong demod phases into the table last night. The elog has been edited. Most significantly, the REFL33 demod phase did not change by 70 degrees. It did change by 10 degrees, but that is likely from the fact that it was formerly tuned for PRCL in PRFPMI, and last night I tuned it for PRCL in PRMI-only.
I have adapted one of Evan's python scripts into an ipython notebook for calculating our PRMI sensing matrix - the work is ~half done.
The script gets the data from the various PD channels (like REFL33_I) and demdoulates it at the modulation frequencies. At the moment its using just the sensing channels, but with the recent addition of the SUS-LSC_OUT_DQ channels, we can demod the actuation channels as well and not have to hand code the exc amplitudes and the basolute phase. Please ignore the phase for the moment.
The attached PDF shows the demod (including lowpass) outputs for a 2 minute stretch of PRMI locked on f2. Next step is to average these numbers and make the radar plots with the error bars. The script is scripts/LSC/SensingMatrix/PRMIsensMat.ipynb and is in the SVN now.
** along the way, I noticed that the reason this notebook hasn't been working since last night is that someone sadly installed a new anaconda python distro today without telling anyone by ELOG. This new distro didn't have all the packages of the previous one. I've updated it with astropy and uncertainties packages.
I've fixed the Radar plot making part, so that's now included too. The radial direction is linear, so you can see from the smearing of the blobs that the uncertainty is represented in the graphics due to each measurement being a small semi-transparent dot. Next, we'll put the output of the statistics on the plot: mean, std, and kurtosis.
One of the things that I looked at tonight was whether or not I could hold the PRMI on REFL165 at CARM offset of 0, and it turns out that I can. Hooray. The next step was having a look to see if it is actually less noisy than the REFL33 lock.
I calibrated REFL33 and REFL165 to meters (I have the data to do the same for 11 and 55, but haven't done so yet). This way, we can directly compare the signals from each PD.
I scanned between +3 and -3 CARM digital offset (which we think is about 1nm/count while held on ALS), with a ramp time of 10 seconds. I did this several times while the PRMI was locked on both REFL33 and REFL165. Here are the gps times for 8 examples where the PRMI did not lose lock during the sweep:
Here are screen shots from the first REFL33 sweep, and the first REFL165 sweep. DTT can't print 3 plots together, so I'll have to make this nicer later. The top plot is the error signals, calibrated to meters. The middle plot is the control signals, that need to be calibrated to Newtons. The bottom plot is the arm powers, so you can see roughly where we were in the sweep.
We'd like to see a MIST simulation, or perhaps e2e, to see what the predicted disturbance is for each of the error signals during the CARM resonance. We want to make sure that the loops are engaged for all of the degrees of freedom for the simulation.
Recipes for tonight:
REFL165 sometimes has a tough time catching lock by itself, but if you add either REFL33 or REFL55 error signals to the REFL165 signals, it'll catch, and then you can just remove the extra error signals. Also, it doesn't stay locked very robustly unless you include the PRCL FM1 boost.
Here are a bunch of PDFs of time series from last night's CARM sweeps. The y-axes are all calibrated (except for the TRX/TRY, which are just normalized to single arm power, as usual) to real units - meters for the error signals, and Newtons for the control signals. The y-axes for each plot are the same on all PDFs (ex, the control signal plot in the lower left has the same range for all cases) so that it is easy to compare directly.
The most striking thing is that while the PRMI is held on REFL33, the MICH control signal saturates as we go through arm resonance. If the PRMI is held on REFL165, there is no such problem. I think we're going to have a lot more luck keeping the PRMI on REFL 165.
Plots while held on REFL 33:
Plots while held on REFL 165:
Using PRX, I remeasured the relative actuation strengths of the BS and PRM to see if the PRM correction coefficient we're using is good.
My result is that we should be using MICH -> -0.2655 x PRM + 0.5*BS.
This is very close to our current value of -0.2625 x PRM, so I don't think it will really change anything.
The reason that the BS needs to be compensated is that it really just changes the PRM->ITMX distance, lx, while leaving the PRM-ITMY distance, ly, alone. I confirmed this by locking PRY and seeing no effect on the error signal, no matter how hard I drove the BS.
I then locked PRX, and drove an 804Hz oscillation on the BS and PRM in turn, and averaged the resultant peak heights. I then tried to cancel the signal by sending the excitation with opposite signs to each mirror, according to their relative meaured strength.
In this way, I was able to get 23dB of cancellation by driving 1.0 x PRM - 0.9416 * BS.
Now, in the PRMI case, we don't want to fully decouple like this, because this kind of cancellation just leaves lx invarient, when really, we want MICH to move (lx-ly) and PRCL to move (lx+ly). So, we use half of the PRM cancellation to cancel half of the lx motion, and introduce that half motion to ly, making a good MICH signal. Thus, the right ratio is 0.5*(1.0/0.9416) = 0.531. Then, since we use BS x 0.5, we divide by two again to get 0.2655. Et voila.
Better elog tomorrow - notes for now:
REFL165 for PRMI has been "a champ" (quote from Q). We're able to sit on ALS at average arm powers of 30ish. Nice.
Some ALSfool work - measured cancellation almost as good as single arm.
One time transitioned CARM -> normalized REFL55I
Many times did DARM -> normalized AS55Q, see lots of noise at 39ish Hz - may be coupling from MICH??
Arm ASC loops helped improve dark port contrast.
Note to selfs: Need to make sure DTT templates have correct freq ordering - must be small freqs to large freqs.
A slightly more coherent elog for last night's work.
All night, we've been using REFL165 to hold the PRMI. It's working very nicely. To help it catch lock, I've set the gain in the PRCL filter bank high, and then the *0.6 filter triggers on. The carm_cm_up script now will lock the PRMI on REFL165.
We had to reset the REFL165 phase after we acquired lock - it was -91, but now is -48. I'm not sure why it changed so significantly from the PRMI-only config to the PRFPMI config.
We measured the ALS fool cancellation with the arms held off resonance, at arm powers of a few. Although, they were moving around a lot, but the measurement stayed nice and smooth. Anyhow, we get almost as good of cancellation as we saw with the single arm (after we made sure that both phase trackers had the same UGF):
We were able to partly engage it one time, but we lost lock at some point. Since the frame builder / daqd decided that that would be just the *perfect* time to crash and restart, we don't have any frame data for this time. We can see up to a few seconds before the lockloss, while we were ramping up the RF PD loop gain though, and MICH was hitting the rails. I'm not sure if that's what caused the lockloss, but it probably didn't help.
The ALS fool gain was 22, and we were using FMs 4, 6 (the pendulum and Rana's "comp1"), the same filters that were used for the single arm case. The LSC-MC filter bank gain lost lock when we got to about 5.6 (we were taking +3dB steps).
We were using REFL55I/(TRX + TRY) as our CARM RF error signal. We were using REFL55 rather than REFL11 because we were worried that REFL11 didn't look good - maybe it was saturating or something. To be looked into.
Here's the striptool that was running at the time, since we don't have frame data:
At this point, since we weren't sure what the final gain should be for the RF CARM signal, and we could sit at nice high arm powers (arm powers of 30ish correspond to CARM offsets of about 50pm), we decided to try just a straight jump over to the RF signals.
The first time around, we jumped CARM to (-0.2)*REFL55/(TRX+TRY), but we only stayed lock for 1 or 2 seconds. That was around 1:55am.
We decided that perhaps it would be good to get DARM moved over first, since it has a much wider linewidth, so the rest of the trials for the night were transitioning DARM over to (0.0006)*AS55Q/(TRX + TRY). AS55 was saturating, so we reduced its analog gain from 18dB to 9dB and re-ran the LSC offsets. The MICH noise was pretty high when we were at low CARM offset, although we noticed it more when DARM was on AS55. In particular, there is some peak just below 40Hz that is causing a whole comb of harmonics, and dominating the MICH, PRC and DARM spectra. I will try to get a snapshot of that tonight - I don't think we saved any spectra from last night. Turning off DARM's FM3 boost helped lower the MICH noise, so we think that the problem is significant coupling between the two degrees of freedom.
After the first one or two tries of getting DARM to AS55, we started engaging the arm ASC loops - they helped the dark port contrast considerably. The POP spot still moves around, but the dark port gets much darker, and is more symmetric with the ASC on.
Much of tonight was spent fighting with ETMX. This time, ASC was definitely off, there was nothing coming out of the ASC filter banks except the static output of the ASS. I tried turning off the 1000 count POS offset, but I think that made it a little worse. I ended up putting the offset back.
It's a little confusing, since it sometimes moves when there is no LSC actuation. However, it definitely moves when there is some LSC actuation. I did a test where every time I enabled the IR arm locking and caught lock, I saw a step in the SUSPIT and SUSYAW error signals. Once lock was aquired, it would settle and stay somewhere. If I unlocked the cavity, there was no "undo" step - it just stayed where it was. I wasn't letting it sit long enough to see if it spontaneously moved during this test.
Here's a plot of this test. The only button I'm touching is the LSC enable button. ASC is off, ASS is frozen (DC values exist, but no dither, no feedback). This was done when the 1000 count POS offset was off. The steps are less bad when the offset is on.
In between fighting with the ETM, I was able to do several trials with the PRFPMI.
I was playing with CARM and ALS fool.
First, I used REFL55 normalized by the sum of the transmissions as the error signal for the MC filter bank and saw that REFL11 (as an out of loop signal) got much more smooth, and centered around zero. However, I wasn't able to get the same thing with REFL11. No matter the sign I used for the MC filter bank, the IFO would squeak (some high freq gain peaking I think), and then I'd lose lock. This was true whether I used REFL11 through the common mode board or just directly into the ADC.
Just now, I did one trial of switching DARM over to AS55Q, just to grab a spectra of the MICH noise that Q and I saw yesterday.
I'm a little confused by some delay that seems to exist between the "A" and "B" error signals (right after the LSC input matrix) and the _IN1 point of the servo filters. I didn't save the measurement (bad Jenne), but there's a ~40 degree difference between DARM_A_ERR/DARM_IN2 and DARM_IN1/DARM_IN2. I don't think there should be anything there. Anyhow, it makes the DARM loop measurements look funny. If you just look at, say, DARM_B_ERR/DARM_IN2, you'll think that there's no way that the loop will be stable. However, it will actually be fine.
For tomorrow, we should take the DARM loop measurement with much less actuation. As with last night, I blew the lock by trying to measure the DARM loop.
As discussed at today's meeting, we would like to (re)measure the Arm cavity lengths to ~mm precision, and their g-factors. Any arm length mismatch affects the reflection phase of the sidebands in the PRMI, which might be one source of our woes. Also, as I mentioned in a previous elog, the g-factors influence whether our 2f sidebands are getting pulled into the interferometer or not.
These both can be done by scanning the arm on ALS and measuring the green beat frequency at each IR resonance. (Misaligning the input beam will enhance the TM10 Mode content, and let us measure its guoy phase shift)
I started working on this today, but I have measurements to do, since at the time of today's measurements, I was fooled by the limits of the ALS offset sliders that I could only scan through two FSRs. Looking back at Manasa's previous measurment (ELOG 9804), I see now that more FSRs are possible.
Ways I will try to improve the measurement:
Just for kicks, here are scans from today.
Brief elog of my activities tonight:
I was able to transition the digitial CARM control to REFL11 through the common mode board a total of one time, lock broke after a few seconds.
My suspicion was that when we did this on Monday, we unintentionally had a reasonable DARM offset, which reduced the finesse enough to let us take linear transfer functions and hop over. So, tonight, I intentionally looked at transitioning to CM_SLOW at some DARM offset. Using DARM offset of a few times 0.1 really calms the "buzzing" down, and makes it fairly straightforward to measure linear CARM sensing TFs. However, the CARM optical plant seems to change a fair amount depending on the DARM offset, in such a way that I was not able to compensate well enough to repeatedly transition.
Before I did anything else tonight, I measured the ALS noise down to 0.1 Hz, as a benchmark of how things are behaving.
With the arms locked on POX/POY, the HZ calibrated ALS channels reported
Then, with the arms CARM/DARM locked on ALS, the PDH signals reported (using a line and the HZ channels for conversion)
Not bad! I roughly estimate this to mean ~90pm RMS CARM/DARM motion. (If X was as good as Y, it would be ~50pm...)
Some things I feel are worth noting:
Tomorrow, I'll post some transfer functions of the difference between the ALS and CARM plants that I measured.
I think we've seen this in simulations, but it's a little disheartening to see in real life. AS55Q looks like it flattens out pretty significantly right around the DARM=0 point.
Right now I have the arms held on ALS (CARM=-1*MC2, DARM=2*ETMY, as Q used last night), and the PRFPMI is on REFL165I&Q. I have set CARM to be as close to zero offset as I can (so I get all the usual buzzing), and then I'm sweeping the DARM offset between +3 and -3 counts (roughly +/-3nm) with a 3 second ramp and looking at normalized AS55Q. The channel called "DARM_B_ERR" is 0.006*AS55Q/(TRX + TRY). The arm transmissions, as well as the ASDC are plotted as well - ASDC is scaled to fit on the same axes as the transmissions.
Anyhow, here's the time series of the DARM sweeps. AS55 demod phase of -55 degrees seems to give the cleanest signal (within 5deg steps); this is the same phase that we've been using all week.
This has been done before:
Arm length measurements and g-factor estimates in 2012, but only with an accuracy of ~30 cm. However, Yuta was able to get many FSRs somehow.
We played around tonight with different possible ways of transitioning DARM to normalized AS55Q. Before each try, we would use ezcaservo (or just eyeball it) to make sure that the normalized RF signals had a mean of zero, so that we knew we were pretty close to zero offset in both CARM and DARM.
We tried something that is similar in flavor to Kiwamu's self-locking technique - we summed in some normalized AS55Q to the DARM error point (using the DoF selector matrix that I created a few weeks ago), and then tried to engage a little low frequency boost. We tried several times, but we never successfully made the transition.
In the end, we just did a direct transition over to normalized AS55Q, and lost lock after several seconds. The buzzing that we hear didn't change noticeably after the transition, which indicates that most of the noise is due to CARM (which makes since, since it has a much smaller linewidth). The problem with holding DARM is that occassionally we will have a CARM fluctuation that lets the arm power dip too low, and DARM's error signal isn't valid at low arm powers. So, we need to work on getting CARM stabilized before we will have a hope of holding on to DARM.
Here's the lockloss plot from that last lock:
Also this evening, I scanned back and forth over the CARM zero crossing while locked on ALS, to see what the RF error signals looked like. Normalized REFL55 seems to have much more high frequency noise near the edges of the linear range than does REFL11. Also, the REFL 11 signal is much larger. So, what I think I want to try to do is use ALS fool to lower the CARM noise by a bit, then make the DARM transition. Then, we can come back to CARM and ramp up the gain.
With these CARM sweeps, I think that I know the relative gain and sign between ALScomm and the normalized REFL signals, and the REFL signals versus the normalized versions. I think that 100*REFL11I/(TRX+TRY) gives the same slope at the zero crossing as just plain REFL11I. Same factor of 100 is true for REFL55I. The REFL11 slope is 20,000 times larger than the ALS slope, while the REFL55 slope is -500 times the ALS slope (note that REFL55 has a minus sign). We can probably trigger the Fool on when the arm powers are above 50, and trigger off when they're below 20. For the zero crossings, the REFL55 threshold should be about 20, and the REFL11 threshold should be about 500.
I also need to re-think the triggering logic for ALSfool. We probably don't want the zero crossing logic to be able to un-trigger the lock, just in case we get an extra noise blip. So, we want to trigger on with an AND, but only trigger off if the arm powers go too low. Also, the zero crossing logic should look at the normalized error signals, not the plain signals.
We need to modify the ALSwatch logic so that it doesn't look at EPICS values for the thresholding. There may be an updated filter module that includes a saturation monitor, but otherwise we can use the saturation monitor part that is in the OSC section of CDS_PARTS. We'll set the threshold on this to match the limiter in the filter bank. Then, if the filterbank output is constantly hitting the limiter, we should run the down scripts.
This is work that I did yesterday but didn't have time to elog. Since it seems non-trivial to give ourselves ramping matrices, but we only really needed the ramping in the DoF selector matrices, I've replaced the separate _A and _B parts with full filterbanks. Recall from elog 10910 that I had given each degree of freedom's _A and _B input options an offset, an epics monitor and a test point. Now those are removed, and handled inside of the filter banks. The outputs of the filter banks sum together.
This required some screen modifications, but everything should work the same way that it did before this change. I've also changed the DAQ channels from the _A_ERR and _B_ERRs that I had hand-created to now be the _A_OUT and _B_OUT test points from the filter banks (acquired at 2048Hz).
I have not yet modified the burt snapshots for the ifo configure screen. The arms will work the same as always, since they didn't have any selector matrix stuff ever, but the rest still need tweaking.
More work from yesterday.
Rana and I had discussed on Thursday night that we probably want to be able to use the zero crossing of an error signal to trigger a servo on, but not to un-trigger it. So, now the zero crossing trigger is latched, using the power trigger to reset the latch.
Also, the input to the zero crossing trigger is the input to the MC servo, before the triggered switch. This allows us to look at the normalized error signals rather than just the non-normalized ones, if that's what we're trying to lock on. This signal is taken before the triggered switch, so that it's looking at whatever is coming out of the input matrix (including normalization).
So. If the absolute value of the MC error signal goes below the threshold, it outputs a 1, no matter what the arm power is. If the arm power is high, the power trigger also outputs a 1. These are AND-ed together, so only if both are 1 do you actually trigger the MC filter bank. If the zero crossing trigger has been set to 1, it will stay at 1 until the arm power goes low enough to untrigger the power trigger. So, even if you have a little bit of noise on the error signal and it pops above the threshold momentarily, this won't cause the servo to un-trigger.
This is implemented using a "set-reset latch". The output of the latch is the zero crossing trigger, which is AND-ed with the power trigger. This final AND-ing, in addition to doing what we want, solves the ambiguity that is inherent in SR-latches for one combination of inputs.
The trigger screen has been modified to reflect these model changes.
Here's a screenshot of the model, which includes some notes for anyone who opens the model since it's a bit confusing:
Last report on model change / screen work from yesterday.
The ALSwatch script has always been just looking at the EPICS output of the CARM and DARM filter banks, and if it saw a single saturation, it would run the down script. This was non-ideal because (a) the EPICS channels aren't the real signals, and (b) sometimes we'll hit the rails briefly and that's okay - we want to shut things down only when we're constantly saturating.
It turns out that there was a pre-existing saturation monitor part in CDS_PARTS, which I have used. There is one each looking at the output of the CARM and DARM filter banks. The threshold for what saturation means is set as an epics input. The part outputs a running count (number of saturations since the last time it was not saturated, resets each time it goes non-saturating) and a total number since the last reset (also an epics input).
(To be continued... still writing)
[Jenne, with Matt and Fujimi as witnesses]
It might be about time to throw that champagne in the fridge. Nice. Not quite close enough to talk about popping it open, but we'll want it chilled just in case...
I still haven't logged yesterday's work, and I'm still working now, so no details, but I just handed both CARM and DARM over to non-normalized RF signals, and had the arms stable at powers of about 105. I was workinig on the ETM alignment, and the power was increasing, so I think that's where the extra power will come from. I was lowering the DARM gain as I improved the alignment, because the optical gain was increasing so much. I probably just didn't do that fast enough for the last aligning, which is why I lost lock.
Anyhow, here's a plot, because I'm excited:
Exciting! How long was it?
I have in my notebook that at 9:49pm CARM was no longer using ALS as an error signal, and at 9:50pm, DARM was no longer using ALS as an error signal. It looks like I was locked for 3+ minutes after getting to RF-only signals.
The increase in power near the end of the lock stretch was me trying to improve the dark port contrast by touching the ETMX alignment. DARM was definitely oscillating as I improved the dark port contrast, so I was trying to hand-lower the gain as I worked on the alignment.
This elog will be about work that happened yesterday. I will write a reply to this with work from this evening's success.
Work started with the plan of trying ALS fool, using the new triggering scheme (elog 11114).
The PRMI was having a bit of trouble holding lock with REFL165, so we checked its demod phase. On Monday (elog 11095) we rotated the REFL165 phase from -91 deg to -48 deg while in PRFPMI configuration (I think the -91 was from PRMI-only phase setting). However, Friday night we saw that MICH was super noisy, especially when the CARM and DARM offsets were near zero. Rana rotated REFL165's phase until the MICH noise seemed to get lower (by at least an order of magnitude in the control signal), while we were at zero offset everywhere. We were not driving and looking at any lines/peaks, just the overall spectra. The final REFL165 demod phase is -80.
We tried engaging the fool path with no success.
First, Rana moved the low frequency boost in the MC filter bank from 20:1 to 0.3:0.03. This gave the whole loop at least 20 or 30 degrees of phase at all frequencies below the design UGF (a few hundred Hz? Don't quite remember). To check this, we put in a "plant" filter, and turned on the locking filter (3:3000^2) and the low freq boost and the plant, and the phase never touched 180 at any low freq. This is so that we can ramp on this filter bank's gain without having an unstable unity gain crossing anywhere. Also, I added two +10dB filters to the first two filter modules, so that we could ramp on the gain at the input rather than the output.
Last night we were actuating CARM on MC2 and DARM on the ETMs, and the MC filter bank was set to actuate on MC2. Even with super duper low gain in the MC filter bank, so that the control signal was much less than one (1) count, it would make CARM unhappy. The CARM filter bank's output was doing +/- a hundred or more counts, so why a few tenths of a count mattered, we couldn't figure out. We were using the power trigger for the MC filter bank, but not the zero-crossing trigger. Since the fool tuning was checked while actuating on the ETMs, we wonder if maybe the tuning isn't valid for MC2 actuation? Maybe there's enough of a difference between them that the fool needs to be re-tuned for MC2 actuation? Fool had the complex pair of poles at 1Hz, the "comp1" filter to give phase lag, and a gain of 22.
I think that at some point we even turned off the fool path, but left the MC path on with a little bit of gain, and the audible noise over the speakers didn't seem to change in character at all. Weird.
We ended up leaving the fool path for another time, and started working on error signal blending at the CARM filter bank input. This is pretty similar to Kiwamu's self-locking principle.
Our goal was to ramp up the gain of the RF error signal at low frequency, while letting ALS keep hold of things at higher frequencies.
CARM and DARM sweeps from earlier seemed to indicate that the RF signals are valid without normalization above transmitted powers of 50 or so, so we thought we'd give those a whirl for this error signal blending.
From doing a CARM sweep through resonance, we guessed roughly that the REFL11 (non-normalized) slope was about a factor of 10,000 larger than the ALS slope. We put a 1e-4 into the input matrix element REFL11I -> CARM_B. For some reason, REFL11 seemed to be centered around -250 counts, so we put an offset of +0.025 ( = 250*1e-4) into the CARM_B filter module to compensate for this.
Since we thought that a gain of 1 in the CARM_B filter bank would make it equal to ALS, we tried some lower gains to start with. 0.3 kicked it out of lock, so we ended up liking and using 0.15. With this low gain on, we tried turning on a low frequency boost, 20:1, but that didn't do very much. We turned that off, and instead turned on an integrator, 20:0, which totally made things better. The transmitted arm power was staying higher more of the time.
From a DARM sweep, we thought that AS55Q (non-normalized) should also have an input matrix element of 1e-4 for DARM. We gave DARM_B a gain of 0.1, which seemed good and not too high. Again, trying the gentle boost didn't do much, so we went with the integrator.
At this point, since both RF signals were being used as error signals with integrators, we declared that at least at DC we were on RF signals. Hooray!!
After this, we started increasing the CARM_B gain a little, and decreasing the CARM_A gain. When Rana finally set the CARM_A element to zero, we lost lock. We realized that this is because we didn't include a zero to compensate for the arm cavity pole, which the IR signal will see, but the ALS won't.
We decided that the plan of attack would be to get back to where we were (DC error signals on RF), and try to start engaging the AO path.
As I (very excitedly) reported in elog 11116, I was able to follow the error signal blending procedure from last night, and get CARM and DARM onto digital non-normalized RF signals. The lock held for about 3 minutes after this transition (elog 11118 has plot of this).
I was then able to script what I did (in the carm_up script), and repeat the transition. Q joined me in the control room, but we have not been able to complete the transition a third time.
Here's the sequence that worked the two times:
After those two attempts, we ran the LSC offset script, since that hadn't been done since early yesterday. We did a quick CARM sweep, and REFL11 seemed to be centered around 0 counts, so we removed the 0.025 count offset from the CARM_B filter bank.
For later attempts, we keep seeing oscillations in the lockloss plots around 50 Hz, as if we're seeing gain peaking at the low side of the phase bubble. We have tried turning off various filters at various levels of RF gain, but none of the combinations seems to be excellent. Turning off the FM6 bounce/roll filter in CARM was particularly bad (immediately lost arm transmitted power), but others weren't good either (eg turning off FM3 boost lost arm powers within a second or so). When we lose arm powers, the RF signals aren't valid, so if you don't turn them off fast enough (and ALS is still on with enough gain), you'll lose the full IFO lock. If you're fast though, you can turn off the CARM and DARM _B outputs and not have to start from scratch.
There seemed to be a very fine line to walk between not enough gain (~50Hz oscillations), and too much gain (200-300Hz oscillations). It has been pretty frustrating later in the evening. We seem to only have about 3dB of gain margin on the low side, when all the boosts are on. Not excellent.
When the RF signals had a moderate amount of gain, but ALS was still holding CARM and DARM, Q checked the phases of REFL11 and AS55 with excitation lines. He rotated AS55 from -55 deg to -30 deg (+25 deg) and REFL11 from 144 deg to 164 deg (+20 deg).
Prior to the all-digital attempts, I tried several times to turn on the AO path, without success. I think that the best that I got was 0dB on the CM board input 1 gain, +14dB on the CM board's AO gain, and -30dB on the MC board's AO gain before the mode cleaner lost lock.
I was hoping that I could get CARM entirely to RF signals, and that would make things more stable and less complicated, and I could try again to turn on the AO path, but we haven't been able to do this tonight.
A few times in the later attempts we tried turning on the UGF servos for CARM or DARM. I'm not sure if the lines kicked things out of lock, or if the UGF servos went a little crazy, or what, but we never survived for more than a few seconds after turning on the excitations.
There is a problem with the optical lever servos. I had thought I'd been seeing it ever since Q re-did the models, and now I'm pretty sure that's what's up. Q is hot on the trail of figuring out what may have changed that shouldn't have. We may want to revert to an old Foton file, and re-copy the old filters into the new filter banks just in case. The watchdog damprestore scripts have been tweaked to clear the oplev filter bank histories before turning on the oplevs, and this seems to solve the symptom of kicking the optic when oplevs are engaged.
Although we haven't been able to make the transition to RF-only a third time, I think we're getting there. Progress has certainly been made in the last 2 days!
According to the official rules, we only need 8 seconds to declare it "locked".
I wonder if the double cavity pole compensation filter for CARM was on for all the attempts yesterday? IF it looks like it will not saturate, it would be more stable to have the whitening on for REFL11 / AS55. Since on Friday, I set the REFL165 demod phase just by minimizing the MICH control signal with the arms on resonance, we ought to check out the PRMI degeneracy with the ETMs misaligned.
Speaking of signal mixing: Although we weren't able to get the carrier term cancelled in the 3*f1 signals by the relative mod phase method, I wonder if we can do it by mixing the 3*f1 and 3*f2 signals in the input matrix. Might help to keep the PRMI more stable, if that's an issue.
P.S. I have done some scripts directory / SVN cleanup. Adding some directories that were not in (like lockloss) and then removing stuff from the repo using 'svn rm --keep-local filenames' for the image and data files.
Here is a longer stretch of data, from the first RF-only lock on Saturday night. Unfortunately daqd had died about 400 seconds before the lockloss, so I can't show the RF signals coming on.
ALS was on the _A channels for CARM and DARM, so when those go to zero (about -300 seconds for CARM, and about -200 seconds for DARM), we're using RF signals only for the error signals.
CARM noise definitely improves, but holy smokes does DARM start to look good! Although, right at the end it starts to look like REFL11I is getting bigger. Not sure why, but we'll have to watch out for this.
Here's the equivalent plot for the second lock stretch. This is the one that was handled by the carm_up script. It looks like I had about 150 seconds of RF-only lock here.
DARM error is getting bigger with time jnear the end, even though I wasn't working on alignment here. Zooming in, DARM is oscillating at 16.4 Hz, which is the bounce frequency. I thought I had my bounce/roll filters on, but somehow it still got a little rung up. It just rings up to a steady state though, it's not getting huge, so I don't know that it was the cause of the lockloss.
Not much luck locking tonight; we made the RF transition to CARM numerous times, but it never lasted more than a minute or so. We were able to take a couple of loop and spectrum measurements as we transistioned.
Here are some spectra showing the noise evolution of CARM_IN1 and DARM_IN1 as we start to transition CARM to RF. We did not manage to grab spectra while CARM was RF only; we can go back in the DQ to find some data.
As we transition, our phase bubble is shrinking, which may explain our poor stability. On the following plot, I actually mistyped the legend. The cyan trace is ALL RF. I'm not sure why we have a 1/f^2 shape from 100->200Hz.
We adjusted the pole compensation frequency by looking at REFL11/ALS during a CARM swept sine measurement, the -3db/-45degree point looked more like 80Hz. Strangely, the compensated REFL11 signal appears to lag the ALS signal around the UGF. Maybe this is a loop effect?
In terms of practical improvements, I've written a script that reliably transitions from POX/POY IR lock to ALS CARM/DARM lock already on resonance. This is saving us a bunch of time. I've svn'd the new ALS script and the new carm_cm_up that uses it.
We looked into the odd oplev behavior as well. We had earlier seen what looked like railed values on the FM output medm screen (which seemed unexpected for an AC coupled loop), but dataviewer showed it was actually ringing/railing at some 10+Hz as the oplev beam fell off the QPD. The ringing continues even after the quadrant values stop crossing zero, so I think it may be the filters themselves misbehaving. Why there is new behavior here is still beyond me.
We lost a fair bit of time to a fussy mode cleaner tonight; there was a good 45 minute stretch where it refused to lock for more than a minute or so, the PC drive angriliy never falling below 5. The thing I changed when it started working was using the fast C1:IOO-MC_F channel instead of the slow C1:IOO-MC_FAST_MON as a readback for the FSS input offset; oddly there is a DC difference between the two. This has resulted in a FSS offset of ~4.2, whereas it was previously ~1.8. After this change, the PC drive fell to ~1.0 levels, and the IMC has been mostly ok.
Given our problems stabilizing the RF lock, we attempted to give the FOOL path a shot, since we now had a better idea of the neccesary REFL11 gain. In short, no luck. Every attempt to use some RF signal just disturbed the lock further. We didn't really pursue it too much after a couple of attempts showed little promise.
I have looked at the CARM and DARM RF loops, assuming the loop shapes that we've been using, and it pretty much looks like a miracle that we were ever able to make the transition. The CARM and DARM loops are very marginal.
The ALS CARM loop was already pretty close to marginal, but we lose an extra 12 degrees of phase with the REFL loop:
However, if our cavity pole compensator's zero frequency is too low, we get all of that phase back, at the sacrifice of 2dB of gain margin at both ends of the phase bubble.
I looked at an Optical simulation to check what the cavity pole frequencies are expected to be, with the losses that we've measured. In both cases, I assume the Xarm has about 150ppm of loss. The DARM cavity pole is about 4.5kHz no matter what the Yarm loss is. The CARM cavity pole is about 172 Hz if the Yarm has 500ppm of loss, or 120 Hz if the Yarm has 200ppm of loss.
In the plots below, I use a CARM cavity pole frequency of 150 Hz, to roughly split the difference.
Edit, 13Mar2015, JCD: Rana points out to me that I was using from Foton the analog design strings, without including the fact that these are actually digital filters. This means that I am missing some phase lag. Eeek.
The ALS loop includes:
The REFL loop includes:
The first plot is the case of perfectly matched cavity pole and compensating zero (150Hz, with compensator having 3kHz pole):
This next version is the case where the compensating zero is a little too low, which is the case I think we have now:
The last plot is a DARM loop. Everything is the same, except that the RF plant has a 4.5kHz pole, and no compensation:
I made very rough calibration of the DARM spectra before and after the transition for the second lock on Mar 8.
The cavity pole (expected to be 4.3kHz) was not compensated. Also the servo bump was not compensated.
While the DARM/CARM were controlled with ALS, the calibration of them are provided by the ALS phase tracker calibration.
i.e 1 degree = 19.23kHz
This means that the calibration factor is
DARM [deg] * 19.23e3 [Hz/deg] / c [m Hz] * lambda [m] * L_arm [m]
= DARM* 19.23e3/299792458*1064e-9*38.5 = 2.6e-9 *DARM [m]
Then, the feedback signal was calibrated by the suspension response (f=1Hz, Q=5)
so that the error and feedback signals can match at 100Hz.
This gave me the DC factor of 5e-8.
The spectra at 1109832200 (ALS only, even not on the resonance) and 1109832500 (after DARM/CARM transitions) were taken.
Jenne said that the whitening filters for AS55Q was not on.
No wins tonight.
I've tried playing with the shapes of the loops a little bit (mostly CARM so far), to no avail. I think I made it to CARM RF-only only one time tonight. I was able to turn on the REFL11 whitening, although I lost lock while about halfway through the DARM transition.
I tried making a double integrator instead of a single integrator for CARM_B, since that would allow me to make a complex zero pair which could help win back some phase. I also tried just straight copying FM1 from CARM into CARM_B, so that it could be turned off for the ALS part of the loop, but left on for the REFL part, but that didn't work very well. Like Rana and I saw last Friday, we really need the REFL signal to have a true integrator, to force the PDH signal toward zero, before we can complete the transition.
I moved the cavity pole compensator's zero back up to 120 Hz, since that was what had worked on Saturday night. That helped me get farther before running into gain peaking problems at ~50Hz. This is because, as seen in my simulation earlier tonight, I win back some gain margin by having the pole compensator more closely matched with the pole frequency.
I've been turning off both FM1 and FM2 in CARM and DARM. I think this is helping a lot, when I can get far enough to do so. I don't want to turn off the second boost until after I'm about 50/50 on REFL. (When I have that much REFL, with the true integrator, the PDH signal sticks to zero).
I tried once turning off the bounce/roll filters for CARM and DARM, rather than the FM1 boost, since the bounce roll filters eat lots of phase, but I got pushed off resonance. I think not having that focused boost may have made my overall RMS larger, which caused me to randomly jump too far outside of the good PDH range.
Early on in the evening, I turned off the MC2 violin filters in the ETM LSC-SUS filter banks, since I am actuating on only the ETMs tonight. However, I saw a violin mode ring up at ~642, which showed up in POX but not POY. This was causing up-conversion, beating against the 40-50Hz buzzing from the IR resonance. The MC2vio1,2 filter covers this frequency (because it's an absurdly wide notch), but the EXYvio1 filter does not. There seems to be some confusion on the wiki as to what the ETMX violin mode frequency is - it says 631 (638??). The notch that is in the EXYvio1 filter is for 631 Hz, but this is not correct. DAYTIME self: Make the MC2 violin filter smaller than 40Hz(!) wide, and move the ETMX notch up to the correct frequency. For tonight, I just turned back on the MC2 filter, and the mode has rung down.
Idea: MICH offset, or ETM misalignment, enough to keep the power recycling low-ish, so that the CARM cavity pole doesn't come down too far in frequency? Daytime brain should think about this.
Q doesn't like elogging, but he sent me this nice detailed email, so I'm copying it into the log:
I’ve locked the power recycled Y arm numerous times today, to try and find a usable AO recipe for the full locking.
Much more success tonight. I only started my tally after I got the CARM transition to work entirely by script, and I have 6 tally marks, so I probably made the CARM to RF-only transition 7 or maybe 8 times tonight in total. Unfortunately, I only successfully made the DARM transition to AS55 once. From the wall striptool, counting the number of times the transmitted power went high, I had about 40 lock trials total.
The one RF-only lock ended around 1:27am.
I think 2 things were most important in their contributions to tonight's success. I modified the bounceRoll filters in the CARM and DARM filter banks to eat less phase. Also, using Q's recipe as inspiration, I started engaging the AO path partway through the CARM transition which makes it much less delicate.
Bounce roll filter
Koji and I added a ~29Hz resonant gain in the bounce roll filter several months ago, to squish some noise that we were seeing in the CARM and DARM ALS error signals. This does a lot of the phase-eating. I'm assuming / hoping that that peak won't be present in the CARM and DARM RF error signals. But even if it is, we can deal with it later. For now, that peak is not causing so much motion that I require it. So, it's gone.
This allowed me to move the complex zero pair from 30 Hz down to 26 Hz. Overall I think this gained me about 10 degrees of phase at 100Hz, and moved the low end of the phase bubble down by about 10Hz.
Prep for REFL 11 I through the CM board and CM_SLOW
In order to use Q's recipe (elog 11138), I wanted to be able to lock CARM on REFL11 using the CM_SLOW filter bank.
I did a few sweeps through CARM resonance while holding on ALS, and determined that the REFL1 input to the CM board needed a gain of -20dB in order to match the slope of CM_SLOW_OUT to CARM_IN (ALS), leaving all of Q's other settings alone. Q had been using a REFL1 gain of 0dB for the PRY earlier today.
I needed to flip the sign in the input matrix relative to what Q had (he was using +1 in the CM_SLOW -> CARM_B, I used -1 there). To match this in the fast path, I flipped the polarity of the CM board (Q was using minus polarity, I am using positive).
The CM_SLOW filter bank had a gain of 0.000189733. I assume that Q did this so that the input matrix element could be unity. I left this number alone. It is of the same order as the plain REFL11I->CARM input matrix element of 1e-4 from Saturday night, so it seemed fine.
During my sweeps through CARM resonance, I also saw that I needed an offset to make CM_SLOW's average about 0. With the crazy gain number, I needed an offest of -475 in the CM_SLOW filter bank. As I type this though, it occurs to me that I should have put this in the CM board, since the fast path will have an offset that isn't handled. Ooops.
Trying Q's recipe for engaging AO path
I am able to get the MC2 AO gain slider up to -10dB (-7 is also okay). If I increase the digital CARM gain too much, I see gain peaking at about 800Hz, so something good is happening. (That was with a CARM_B gain of 2.0 and CARM_A gain of 0. Don't go to 2.0)
I tried once without engaging his 300:80 1/f^2 filter in the CM_SLOW filter banks to start stepping up the CM REFL1 and MC AO gains together, but I only made it 2 steps of 1dB each before I lost lock.
I tried once or twice turning on that 300:80 filter that Q said over the phone really helped his PRY locking, but it causes loop oscillations in CARM. Also, I forgot to turn it off for ~45 minutes, and it caused several locklosses. Ooops. Anyhow, this isn't the right filter for this situation.
AS55 whitening problem
Twice I tried turning on the AS55 whitening. Once, I was only partly transitioned from ALSdiff to AS55, the other time was the one time I made the full transition. It caused the lockloss from the only RF-only lock I had tonight :(
Unfortunately I don't have the time series before the whitening filters (not _DQ-ed), but you can see a giant jump in the _ERR signals when I turn on the whitening, just before the arm power dies:
The AS55 phase is -30, I has an offset of 28.2 and Q has an offset of 6.4. Both have a gain of 1. This should give us enough info to back out what the _IN1 signals looked like before I turned on the whitening if that's useful.
Other random notes
Ramp times for CARM_A, CARM_B, DARM_A and DARM_B are all 5 seconds. This is set in the carm_cm_up script.
carm_cm_up script freezes the arm ASS before it starts the IR->ALS transition, to make it more convenient to run the ASS each lockloss.
carm_cm_up script no longer has a bunch of stuff at the bottom that we're not using. It's all archived in the svn, but the remnants from things like variable finesse aren't actively useful.
carm_cm_down script turns off the CM_SLOW whitening (which gets set in the up script)
carm_cm_down script clears the history of the ETM oplevs, in case they went bad (from some near divide-by-zero action?), but the watchdog isn't tripped. This clears away all the high freq crap and lets them do their job.
FSS Slow has been larger than 0.55 all night, larger than 0.6 most of the night, and larger than 0.7 for the last bit of the night. MC seems happy.
both carm_cm_up and carm_cm_down are checked into the svn. The up script is rev 45336 and the down script is 45337.
Some offset (maybe the fact that the fast AO path had an un-compensated offset?) is pulling the arm powers down as I make the transitions:
After that, I by-hand made the DARM transition on the 6th successful scripted CARM transition, and tried to script what I did, although I was never able to complete the DARM transition again. So, starting where the recipe left off above,
Since DARM doesn't have an analog fast path, it is stuck in the delicate filter situation. I think that I should probably start using the UGF servo once the arm power is stable so that DARM stays in the middle of its phase bubble.
Rather than typing out the details of the recipe, I am attaching the up script.