Summary
PRMI locked!!
The MICH loop was reworked, the PRCL gain was raised, and the REFLAIR demodulation phases were tuned. The guardian has been updated to the new nominal.
Loop adjustment
After aligning BS and PRM using PRX and MICH, I first tried to lock with ISC_DRMI as it was, changing nothing. It never caught after about 20 minutes of waiting in total, split between yesterday and today.
So I went back to the MICH loop. With PRM misaligned and MICH locked, I measured the open loop gain. The UGF was at 5 Hz and the phase margin sat around 20 deg across 5–10 Hz. So, I decided to modify the MICH loop and push UGF up to 10 Hz with reasonable phase margine.
The largest phase eater was the ELP35 in FM2 of BS_M3_LOCK. I loaded an ELP200 into FM3, which was free, and used it in place of the ELP35 in FM2. That loosened the phase roll-off enough to give about 40 deg of margin at 10 Hz.
Back to PRMI. MICH1 and MICH2 also carried several low-pass filters, so I switched all of them off, and it locked within a few minutes. Once locked, all of the low-passes could be switched back on without losing lock.
Repeating this, it locked within a few minutes every time.
With the lock holding I measured the MICH and PRCL loop gains. MICH showed gain bump above 10 Hz, and the PRCL UGF turned out to be only 20 Hz. Doubling the PRCL loop gain brought its UGF up to 55 Hz, and the MICH gain bump went away. This was the most stable state I found, and the time to reacquire came down to 1–2 minutes, so I took it as the nominal and edited the guardian at this point.
Servo change summary
Filter changes
| Filter bank | Module | Change |
|---|---|---|
| MICH1 | FM5 (LP80) | off for acquisition, on after lock |
| MICH2 | FM2 (LP50) | off for acquisition, on after lock |
| BS M3 LOCK | FM2 (ELP35) → FM3 (ELP200) | permanent; ELP200 newly loaded into the empty FM3 |
| REFLAIR_A_RF9 input matrix | — | set to 3.5, doubling the PRCL loop gain |
Guardian edits
| What | File | Line |
|---|---|---|
| MICH1 FM5 | lscparams.py | 530 |
| MICH2 FM2 | ISC_DRMI.py | 219 |
| BS M3 LOCK, ELP35 → ELP200 | ISC_DRMI.py | 298, 523, 585 |
| REFLAIR_A_RF9 input matrix | lscparams.py | 718 |
| Engage the low-passes after lock | ISC_DRMI.py | 628–630 |
REFLAIR demodulation phase
Last, I tuned the REFLAIR demodulation phases. A 50 Hz excitation of amplitude 3 was injected into PRCL1 ERR, and the phases were adjusted to minimise the peak in RF9_Q and RF45_Q. The I peak heights did not change much.
| demod phase [deg] | Q peak height | |
|---|---|---|
| RF45 | -176 → -171 | 0.1 → barely visible |
| RF9 | -29 → -34 | 0.1 → barely visible |
OLG
A final open loop gain measurement gave:
| Loop | UGF | Phase margin |
|---|---|---|
| MICH | 10 Hz | 50 deg |
| PRCL | 50 Hz | 50 deg |
Both open loop TFs are attached.
I do not understand why the PRCLK gain was low by a factor of 2, but Elenna noted a similar gain loss in March: 89567, this didn't go into the guardian at the time but is probably the same gain loss that Masayuki put into the PRMI input matrix yesterday.