Georgia, Sheila, Danny, Marie, Jenne, Arnaud, Adam
This morning we added some cut offs to the DC3+4 P loops (AS centering). The first two attachments show the open loop and closed loop gain of DC3 P, before our change, there was a small phase wiggle in the closed loop response around 1 Hz, which would have an impact on the DHARD P loop gain (similar to what we saw for SRC2P yesterday 46713) . The third attachment shows the cut off that was added after those measurements were taken.
We then made some measurements of the DHARD loop with and without a DARM offset. The fourth attachment shows that even with no DARM offset, we still have a feature in the transfer function at 1 Hz, but we are not seeing the huge gain changes with DARM offset changes that we had earlier in the week. 46676
We powered up slowly on RF DARM (because of a very high violin mode we didn't go to DC readout), And measured the DHARD loop at a couple of powers. Then we looked at the boost called 10W boost in the DAHRD filter (here), which was a simple pole zero boost in mid October, but was turned into a series of resG's (third attachment) (46027, and 46149). We reverted to the old boost filter, which is less of a point design, but we made the boost a bit gentler by reducing the zero to 1 Hz.
These resG boosts were motivated by trying to suppress an instability at 0.47 Hz, but were possibly also creating an instability at 0.47 Hz. When we powered up to 25W with out the boost10W filter (really a series of resGs), we did see an instability at around 0.47 Hz. We removed a resG at 0.47 Hz from DSOFT P, and increased it's gain by a factor of two based on some modeling that Marie did, which helped us be stable at 25W. We powered up to 30W and saw the instability again. We turned off the radiation pressure compensation for the soft loops and powered up to 30W stably. A measurement of DHARD P with the power at 30W is here
changes in the guardian:
- DC3+4 pit filters aren't set by guardian, but they are changed to include the cut off.
- We are no longer using the filters that are a series of resG's that were called boost10W in CHARD P, DHARD P, or CHARD Y. We haven't yet removed it from DHARD Y
- We removed a resG filter from DSOFT P, and added a factor of 2 more gain
- In DRMI ASC we lowered the gain of PRC2 Y by a factor of 10, since it was causing an oscillation at 21 Hz.
Not in the guardian:
We tried putting the SRC2P in the guardian with the lower bandwidth when first engaged, but we lost lock when the dither loops were engaging. This could have been unrelated, but it seemed that the SRC2P loop was too slow to follow the dithers, so we have taken it out. The way we have been doing this change by hand is to turn the gain down to 0.1, turn off the HSTS compensation filter, turn on the 0.7 Hz cut off, and ramp the gain to 10.
I thought naively that I could apply my script from llo to tune the ASC open loops. Here is a comparison of a measurement of DSOFT P last night at 20W input and a model of the open loop gain. Although the model does predict a feature close to 0.4 Hz in the open loop gain, there is a large discrepancy with the measurement (see figure 1 for the open loop gain and figure 2 for the loop suppression). We are missing a measurement of the peak at 0.5 Hz due to a dtt error, but visually extrapolating the measurement suggest that it is at a higher frequency than expected (around 0.55 Hz). The phase is also quite far off.
We will keep investigate mistakes in the model.
