Regrettably, The ETM ring heater change appears to have caused a fast PI ringup and lockloss at around 9:20 pm this evening.

Attached is Sheila's picture of the OMC DCPDs (second PNG)

I reverted the EX RH settings now, so the transient is not too crazy for Robert in the morning.

This PI mode was at 10431.4 Hz. It is plausible that this is the same troublesome EY PI mode that has been seen in the past at 10430.5 Hz (50209). The ~1 Hz frequency difference can be explained by a ~1 K shift in test mass temperature, since the elastic modulus of fused silica has a fractional temperature shift of about 200 ppm/K.

Watching the 10.4 kHz region of DARM during the ring heating, there is evidently an evolution in the coupling of some technical noise at twice the arm transverse mode spacing (it is similarly seen at the transverse mode spacing itself).

During this test, directly before the PI lockloss, I had measured the DARM plant in order to see if there was any change in the DARM spring.
There was not.  

It seems that the ETM ring heater change does not really affect the SRCL offset, or the corner in general.

You can see some change in the ETMX measured spherical power (orange trace) after the ring heater change but it increases. I would have expected spherical power to decrease with more EX ring heater. Strange, the TOTAL_PIXEL_VALUE is decreasing at this time which mean that the amount of light getting back to the CCD is changing, zoomed out plot here. It is hard to tell if there is still some change happening at the lockloss time.

Attached is a movie showing the 64 kHz OMC DCPD channel spectrum during the power up, thermalisation, and then this differential ring heater change. It focuses around the 2nd order mode frequencies in the arms, 10~11 kHz. This image highlights the main features we are looking at the in the movie.

Here are some thoughts, observations:

• Why do we see these peaks in DARM?
  • Not clear why, but the HOM resonances could be getting excited by some laser noise at high frequency, as the HOM resonances are different in each arm this excites couples differentially to AS.
• After thermalisation there is approximately a 400Hz difference between the X and Y modes. As this is the 2nd order mode resonances, the mode-separation frequency difference between the arms is 200Hz
• From plot we can see that having the mode-separation frequency equal between the arms is necessary but not sufficient to say we are mode-matching the arm eigenmodes
  • This shows that if we take some curvature off of one mirror, we can put it on the other and keep the same mode-separation frequency - as we were doing in tests in previous weeks
  • Assuming there are only small RoC differences, the best case with a 200Hz mode-separation frequency is the arms are about 0.1% mismatched, but could be worse.
• How much do we expect the arm mode separation frequency to change during a power-up?
  • From FEA models, the surface RoC should change by -30uD/W of absorbed power
  • From Galaxy pages, the coating absorptions are about 0.5ppm in the ITM and 0.3ppm in the ETM. Assuming 380kW in the arms there should be 0.14W and 0.11W in the ITM and ETM respectively
  • Using the these absorbed powers and the spherical power actuation of uniform heating, going from cold RoCs to absorbing the above powers should result in around 500Hz of mode-separation frequency change in the arm
  • From the start of the movie we can see 4 distinct peaks from the 4 HG 2nd order modes from each arm (why these are so separated I do not know)
    • The average of the Y arm modes starts at ~9900Hz and after thermalising gets to ~10400Hz: a 500Hz change
    • The average of the X arm modes starts at ~10150Hz and ends up at 10750Hz: a 600Hz change
    • So order of magnitude, what we see here is sensible and what we would expect. Overall the X arm looks like it is absorbing more, which due to the ITM point absorbers isn't surprising.
• During the differential ring heater change we see the X and Y arm modes move closer together. We can identify which modes belong to which arm as we know which direction each ETM RH was changed:
  • Increasing the RH power => decreases the mode separation frequency
  • Decreasing the RH power => increases the mode separation frequency
  • As the right most peaks move lower in frequency and we are turn up ETMX RH, we can infer that the HOM peaks at higher frequencies are from the X arm
• When we decrease the ETMY RH we increase the Y arm mode separation frequency, this moves the two Y arm HOMs up further into the 10.4kHz acoustic modes and eventually causes a parametric instability to ring up and we lose lock.
• We also see that this same Y arm HOM peak also excites the 10.2 kHz acoustic mode. We either moved through it quick enough it didn't have long enough to ring up, or the PI gain isn't large enough to be a problem.
• It seems to be the same Y arm HOM which excites the 10.2 and 10.4 kHz modes

As posted in a previous comment in this alog the frequency/intensity noise coupling reduces during this differential RH test. From the movie we can see this RH change brings the X and Y HOMs closer together, which should mean we are bringing the arm eigenmodes closer together therefore reducing any differential coupling path for the laser noise.

Wow Dan, that is the most text book movie of how PI tunes in/out of modes. You could even see a 10200 mode get transiently excited as the HOM flew past, fortunately it moved beyond that mode before it got more excited.

This is the first time I am seeing HOMs in the OMC output. We don't get that at LLO, but we have a different (slightly older) OMC chassis, which you just changed at LHO very recently. Do you directly see any other HOMs at say ~5kHz, 15kHz or higher? (if you don't, that would indicate mode matching issues with the 20/02 modes) [edit: I see the 15kHz ones only in this recent alog]

Looks to me like a frontal ring heater can really help in micromangaing RoC in a shorter time frame to avoid these lock losses [link1, link2].

At what point will LHO recommision the ESD PI path? Seems to me thats the only sure fire way of handling this.

Even the old omcpi model had a 64KHz channel, H1:OMC-PI_DCPD_64KHZ_AHF, that got recorded to disk. This hasn't changed. What has changed is the availibility of a 512KHz test point, H1:OMC-PI_DOWNCONV_SIG_OUT.

For the current state of the PI models see alog 67402.

You can see some 1st order mode movements around 5kHz, movie attached. I couldn't see anything interesting happening around 15kHz.

The Y 1st order peaks are easily seen and do not look as astigmatic as the 2nd order modes. The X arm modes are not as large and harder to see. Occasionally these HOM peaks jump up a bit and become visible. I would guess this coincides with some arm alignment change, maybe a glitch at this stage during the lock.

When thermalised, the Y arm is sitting at ~5200 Hz and the X arm ~5400 Hz which also agrees wiht the 2nd order mode peaks.