hugh.radkins@LIGO.ORG - posted 12:34, Wednesday 04 February 2015 - last comment - 17:30, Wednesday 04 February 2015(16466)
Changing HEPI Pump Servo to lower PID Loop reduces Sensor Noise Coupling to Most DOFs
I lowered the PID parameters earlier. Here is the comparison: All coherences seen at 4am today (PID:35 & 0.45) are reduced a good amount (just a slight reduction on BS RZ) when the PID parameters are lowered to P=20, I = 0.07.
See the attached where the Dashed reference coherences are from 4am local today and the current Solid traces are at 8am with the lowered parameters.
Except for higher frequencies of the pringle coherences on most platforms, the Dashed coherences are much higher than the Solid traces. The BS RZ looks like an exception where the coherence only drops a bit.
xml 2015-02-4_H1HPI_PumpControllerNoise.xml is in /ligo/svncommon/SeiSVN/seismic/HEPI/H1/Common/
Images attached to this report
Comments related to this report
Even more confusing, I remeasure the ASD of the pressure sensor, and the loop suppression -- as shown by the ASD -- also gets modified (see first attachment), again differently than the model of the plant predicts (see 16447). This indicates that the HEPI Pump Servo box is doing *something* to it's error point, but that error point *is not* the correctly reported pressure noise because we've proven that the open loop noise looks identical to sensors disconnected noise (see LHO aLOG 16426). Our next suspicion on the controller side of things is that the EPICs calc-record is not turning the EPICs requested P, I, and D into a control filter with the same units as reported on the screen. We'll Suspecting DTT nonsense, I've also pulled 4000-second, time-series data from the frames directly into matlab using the nds2 client, using times reported in LHO aLOG 16426. I then calculated the ASD "offline" using asd2.m, to confirm that DTT doesn't screw up the FFT of an EPICs channel. The results are identical to DTT's results. Next up: - dive into the EPICs calc record to be sure that P I D parameters are getting turned into the controller we expect. - measure the ASD of pressure sensor signals in analog, before they go into the HEPI Pump Servo, using an SR785. They say the only we to really get to know a system is to fix it when it's broken. Welp, I'm certainly getting to know this system...
Non-image files attached to this comment
