Reports until 01:47, Wednesday 12 May 2021
H1 AOS
craig.cahillane@LIGO.ORG - posted 01:47, Wednesday 12 May 2021 - last comment - 14:43, Thursday 13 May 2021(58927)
Took PUM DAC high noise data; wrote script to switch stages and wait automatically
I redid the high noise PUM DAC noise test Anamaria and I tried to do a couple of days ago,
but failed due to glitches spoiling the measurements.

The idea is to run the coil drivers in the high-noise state, observe the PUM DAC noise in DARM, then estimate the power in DARM due to PUM DAC noise in the high noise state.
Then, we can apply our analog filters to the PUM DAC noise to get the medium noise state projection.

------------------------------------------------------------------------------------------------------------------------------------------

I wrote a script which automatically switches the coil driver states for us.
The script, all_quad_pum_switch.py, cycles through each quad, changing each PUM coil drivers setting to state 2 (high noise), waits 5 minutes, then back to state 1 (medium noise).
It then saves the gpstimes of all the switches.

Tomorrow I'll write a post-processing script which grabs the data at the times of the switches and a reference time,
then projects into DARM.

------------------------------------------------------------------------------------------------------------------------------------------

This code lives on the CDS computers at: 
/ligo/gitcommon/labutils/coil_drivers_state_switch

Instructions on how to run are in the git repo is at:
https://git.ligo.org/aligo_commissioning/labutils/-/tree/master/coil_drivers_state_switch
Comments related to this report
craig.cahillane@LIGO.ORG - 21:43, Wednesday 12 May 2021 (58935)
Here are the initial results.
High noise on ETMY coils seems to yield some strong impression in DARM.
Others have some small effect, but nothing serious.

Still working to make sure ETMY results are not a glitch,
and to project into DARM.
Non-image files attached to this comment
craig.cahillane@LIGO.ORG - 01:47, Thursday 13 May 2021 (58940)ISC
We manualled to LOWNOISE_COIL_DRIVERS to put all the coils in the lowest-noise state.  This seemed to work fine.  We could probably put this in the guardian.
Then we ran another PUM coils test suite using all_quad_pum_switch.py.

I projected the PUM DAC noise into DARM for both state 1 (medium noise) and 3 (low noise).
A few notes:
- The excess noise from ETMY above was a red-herring.  Using median-averaging got rid of the glitch which occurred when the ETMY was in the high noise coils state.
- In state 1, the noise at 30 Hz is projected to be around a factor of 10 below DARM.
- In state 3, the noise at 30 Hz is projected to be around a factor of 100 below DARM.

My concern for this measurement is we're just taking the variance around a PSD and multiplying it by a TF.
The best I can say for these projections is they are an upper limit, and we don't need to worry too much about PUM DAC, especially in the low noise state 3.

I'm taking a longer measurement now, to try to reduce the variance, but as Georgia notes in 58937, our LF noise changes with fluctuating alignment.
The ADS lines are off for this measurement, which should be fine on an hour timescale.
Non-image files attached to this comment
craig.cahillane@LIGO.ORG - 14:43, Thursday 13 May 2021 (58947)ISC
Redid the test projection for state 1 to state 3 last night.
This time I used 239 averages (20 minutes of data per quad) to make the excess power from DARM projections to reduce variance.
I can't really see any excess power from the PUM coils being in state 1.

This means that either
1) PUM DAC noise is far below DARM, even in the high noise state 1, or
2) the noise doesn't really change, despite what our "state" is.  

If we have time, we will try the HF dither line test from Chris's comment to this alog: alog 54727.
My understanding is Livingston saw improvement from this HF dither at 7 kHz, which upconverts the nonlinear DAC noise from LF to around the dither line.
Non-image files attached to this comment