This is an expansion on my quick log from last night (alog 67686). For this current lock, I have once again set our ETMX bias at the value that gives the lowest EMI coupling (as determined in alog 67075) by following the by-hand procedure I'll sketch out below.
I think we'd like to run with ETMX's ESD bias at the value (+128 V) that Robert has found gives the lowest coupling from EMI. To set that up, last night I
- Held the guardian at Transition_from_ETMX so that we are not using EX for actuation.
- Ramped (over 60 sec) the ETMX bias offset to be max positive (rather than our current max negative). I also flipped the sign on the ETMX L3 Drivealign L2L filter bank.
- I let the IFO transition using the guardian back to ETMX.
- I measured the DARM open loop gain (using as a reference the OLG from the time that Sheila was checking bias values back in Jan - alog 66814).
- I adjusted the Drivealign L2L gain to match the DARM olg reference. This is making the assumption that the only thing different between then and now is the actuation strength of ETMX, which I adjusted digitially to match.
- I took several steps in the direction of lower bias:
- Ramped the bias down, while simultaneously ramping the drivealign gain up by a matching factor (aka a rough guess).
- Measured a few points of the DARM OLG, calculated what the new drivealign gain should be to make the olg match much more closely
- Put in that updated drivealign gain.
- Repeat, starting again with next bias step, until I reached a bias of +128V.
- Since I matched the DARM olg by adjusting the L3 actuator (although, to a January reference, and not one freshly taken, so that could have introduced some errors), I don't think that the calibration should have changed significantly, so I did not do any calibration fudging (which Jeff had pointed out we should be careful of in alog 67322's comment)
That done, I also checked whether this lower ETMX bias value had obvious increased low frequency noise as compared to the full bias value. Sheila had found in alog 66814 that the negative biases did have such an increase in noise at low frequency. However, repeating that test yesterday with the positive biases I did not see evidence of increased noise.
I also checked that the ITM biases were already at the minimum values that are suggested in alog 67627, of just leaving them both at their maximally negative values.
I also changed the ETMY bias so that the output was at the suggested +115 V. lscparams and ISC_LOCK are supposed to be setting it to that value, but they put in a bias offset of +1.9, whereas to actually get +115V I had to change it to -3.0. After Camilla is finished with charge measurements, I'll change that lscparams value to -3.
This lock this morning, Austin held the guardian for me at Transition from ETMX, and I just ramped the bias voltage to the +128 V place and typed in (without doing any OLG measurements) the drivealign gain I found yesterday. I think that we should have the guardian locking us on the positive voltage side, and then at Transition_from_ETMX have it lower the bias and raise the drivealign gain before we transition back to ETMX. I will note however that using the same bias offset value (+3.0) was today giving a slightly lower measured voltage by ~5 V. I don't yet know how important this is, so I turned the bias offset up to +3.3 to get +128V measured.
Attachments:
(1) DARM olg, after having matched the drivealign gain. Red points are measurement last night, all the refs are from Sheila's mid-Jan work.
(2) Cleaned DELTAL_EXTERNAL showing that I'm not seeing an obvious increase in low frequency noise when changing between full positive bias and ~1/4 positive bias.
(3) mini table showing the measured ETMX bias and corresponding L3 drivealign L2L gain that made the DARM olg match the references.
ISC_LOCK's prep_for_locking should now reset the EX bias to the full negative bias (so the guardian doesn't wait around thinking that the bias is off, and so we can relock with our full bias), and then in transition_from_etmx, while we're temporarily not using EX, when it turns the bias back on, it'll turn it on to the low EMI coupling value, and put in a matching gain to drivealign_L2L.
Times from those spectra:
full bias = 1 Mar 2023 00:54:28
1/4 bias = 1 Mar 2023 00:30:02
bandwidth = 0.0469, 50 avgs
Today, we seemed to have a lot of elevated noise at low frequency.
Once we were relocked (and a second time today saw that apparent elevated noise), I ramped the ETMX bias up to full (ramping down the L3 L2L drivealign gain at the same time), and the elevated noise went away. I note, however, that the shape of the elevated noise was awfully close to the shape that we see while the gains are ramping, which made me suspicious of perhaps the drivealign gain not being set correctly today to match up with the 1/4 bias value.
While still at full bias, I took a few points of a DARM open loop gain measurement, and saw that it was very close to the references from Jan. I then ramped back to 1/4 bias, measured DARM olg, and found that it was much too high. I needed to reduce my drivealign gain to 177.55 ( = 206.85 * 0.86) to make the DARM olg match up with the references. When I put that in, the spectrum on the wall looked much more normal.
On the one hand, the elevated noise looked similar between our two locks today, so perhaps something is different today vs. yesterday (like the SRCL offset), but that should be a new permanent thing. But it's also possible that we need to measure and set the drivealign gain each lock, which isn't ideal. We'll watch this over the next few days and locks to see.
Tagging CAL. Again, because this might mean that running at this lower bias.
It appears the drivealign gain value of 177.55 didn't get saved in the guardian. I have just edited it. It is in LSCparams as "ETMX_GND_MIN_DriveAlign_gain" and it is set during the "transition_from_etmx" state