Terra, Sheila
Tonight we had trouble engaging the ASC again.
- We still have trouble turning up the gain in DHARD, and didn't get a chance to look at this tonight. We did try reverting the ETMX notches but that doesn't solve the problem.
- We also had trouble with our MICH error point being wrong, which might have been related to TCS being off durring maintence. Now that TCS has been on for several hours the MICH loop seems to be working OK.
- We had to restore OMC QPD offsets that weren't in SDF before today's boots. We are now back in DC readout.
- Once we reached DC readout we manually tried the CARM gain redistribution that we had commented out of the ANALOG_CARM state yesterday, and we did not see glitches reappear in the MCL drive. We left this uncommented in the guardian, so we can try this in the morning.
- Jenne reset the soft offsets for our new alignment, and we also got rid of the POP A yaw offset.
Losing optical gain in POP X
We rang up what we think is a PR3 bounce mode when engaging the ASC the same way as last night. We found that we could avoid ringing this mode up by keeping the PRC2 gain low (digital gains of -500). Right before the OMC damage/vent, the POP X path was reworked and the optical gain seemed suspiciously low.
Tonight we found that the optical gain has decreased even more. Terra changed the demod phase by dithering PR3 pit (500 counts to M3) and rotated the phase positive 65 degrees, (Q1, Q2, Q3, Q4 from 55, 53, 54, 51 to 120, 118, 119, 116 ) to maximize the signal in I (minimize the Q signal). The 2 attached figures show Terra's before and after OLG measurements (excitation gain of 50), both with Jenne's gain of -5000, showing a 10dB increase in optical gain which is about what we expected based on the dither amplitude change.
After optimizing the phase, we did not see the 28 Hz mode get rung up, but this seems to come and go because we also didn't see it yesterday. We quickly tried moving L2 on the POP X path, while watching the amplitude of the PR3 dither line in the POP X signal. We moved the lens about 4 inches closer to POP X and about 3 inches further away, and didn't find any location that had more signal for PR3 so we replaced it as we found it.
We are going to leave the IFO locked in DC readout 2 Watts with the request set to down so that it will not try to relock. The noise is bad as expected.
POPX whitening gain is 0dB but should be odd, see alog 26307. FRS 6057 filed.
The whitening gain on POP X was changed from a gain step of 7 (21 dB) to 0 (0dB) on August 12th. This whitening chassis has a problem and we must use odd gain settings, or else it will return an error and not set the gains equally on all quadratns, as Keita and Hang noted 26307
The change in gain probably happened during a beckhoff restart for the shutter code, but we could have been saved from this problem by SDF. I cannot find a record for these whitening chassis in any SDF table.
Also, this does not explain the drop in gain that Jenne saw, which happened before the whitening settings changed.
The stuck whitening gain bit is the LSB of the Q3 channel. In the past this was typically an indication of a cable problem (short).
Sheila Daniel Terra
Connected the AM laser to the POP X head, and saw that we have very similar response in the electronics to what Evan measured in 27069
we had 3.3 mW out of the AM laser with a whitening gain of 21 dB, used -40 dBm of RF drive at 45.501150 MHz. We saw about 600 counts on each quadrant (except quadrant 3 which had 350 counts and also the least amount of DC light because of way the laser was mis centered on the diode).
We saw that there are rather large offsets when we changed the whitening gain, so Daniel reset the offsets. The large offsets might have contributed to problems last night, along with confusion about the whitening gain.
Also, we remembered that a factor of 6.7 of the mystery gain loss was due to adding a beamsplitter and forgetting to comensate for it on July 11.
(Edit: Actually, Haocun and I did remember to correct for this gain change, we just compensated for it in the digital loop gain. )
So to summarize:
loops were intially commisioned with a whitening gain of 21, a digital gain of -21, a 1 Hz ugf, and electronics gain similar to what we have now. (late may)
Edit: loops were originally commisioned with a filter gain of -200 for pit, -0.1 in the input matrix, an analog gain of 21 dB, and the WFS head electronics performing in a way simlar to what we have now. This is when the reference that I think Jenne used was saved, and within a few days the pit input matrix was reduced by a factor of 2.
Edit: Around June 16th, we had difficulty staying locked when these loops were engaged, which was noted in the alog. Terra and I just looked at trends of the filter gains, and it seems like we also reduced the digital gain from -220 to -3.4 although this was not noted in the alog. This, together with the input matrix change explains most of the missing gain that Jenne found.
On July 11th I forgot to compensate for the beamsplitter causing a gain reduction of 6.7 that no one noticed.
On July 26th, Evan and Keita relocated POP X and Jenne noticed that the digital gain had to be increased by a factor of 250 (or 500 for yaw) to keep the ugf the same.
August 12th the whitening gain was reduced to 0 dB from 21 dB by mistake in a beckhoff reboot.
August 16th Terra and I noticed this further reduction in gain, which is explained by the whitening gain. We also changed the demod phase which increased the gain by about 10 dB. We checked that small movements of the L2 don't change the optical gain much, and moving it by a few inches can decrease the signal.
So, we are missing about a factor of 40 gain, which we cannot explain with electronics.
In the end only a factor of 2 of Jenne's gain change in unexplained. It seems that we have had stable high power locks with both the high gain and low gain settings for PRC2, so we can decide which we want to use. We also should have a factor of 3 increase in gain because of the phasing Terra and I did.
More complicated than that.
|
Whitening (dB) |
POPX digital gain before rotation |
Input matrix | PRC2_P_GAIN |
BS |
Overall gain relative to original |
alog | |
| Originally | 33 | 1 | -1 | -220 | none | NA | |
| May 24 ~1:02 | 33 | 1 | -0.05 | -220 | none | 0.5 | |
| Jun. 17 | 33 | 1 | -0.05 | -3 | none | 6.8E-3 | |
| Jun. 22 ~noon | 21 |
2.8 |
-0.05 | -3 | none | 4.8E-3 | 27901 |
| Jul. 11-12 | 21 | 2.8 | -0.05 | -21 | inserted | 5.0E-3 | 28324 |
| Jul. 27 ~4:20 | 21 | 2.8 | -0.05 | -5000 | inserted | 1.2 | 28666 |
No mystery optical/electronic gain reduction any more. Maybe a factor of 1.2 came from the rework on the table.
It's not clear to me why the PRC2 filter gain was reduced by a huge amount on Jun. 17 but I haven't searched through alog.
Typo in the above table, originally the input matrix was -0.1, not -1.