40m QIL Cryo_Lab CTN SUS_Lab TCS_Lab OMC_Lab CRIME_Lab FEA ENG_Labs OptContFac Mariner WBEEShop
  40m Log  Not logged in ELOG logo
Entry  Mon Aug 10 22:18:19 2015, Ignacio, Update, IOO, Ready to do some online mode cleaner subtraction stsx.pngperformance.pngfilter.zip
    Reply  Tue Aug 11 11:30:19 2015, Ignacio, Update, IOO, SISO (T240-X) FF of MCL MCL.pngMCTRANS.pngYARM.png
       Reply  Wed Aug 12 01:32:18 2015, Ignacio, Update, IOO, Improved SISO (T240-X) FF of MCL filter.pngstsx.pngmclonline.pngyarmonline.pngsub.png
    Reply  Wed Aug 19 01:45:10 2015, Ignacio, Update, IOO, Doubly Improved SISO (T240-X) FF of MCL 6x
       Reply  Thu Aug 20 11:09:10 2015, rana, Update, IOO, some points about seismic FF 
       Reply  Thu Aug 27 01:41:41 2015, Ignacio, Update, IOO, Triply Improved SISO (T240-X) FF of MCL 1.png2.png4.png5.png3.png
          Reply  Fri Aug 28 00:59:55 2015, Ignacio, Update, IOO, Final SISO FF Wiener Filter for MCL bode.png2.png3.png4.png6.png
Message ID: 11519     Entry time: Thu Aug 20 11:09:10 2015     In reply to: 11516
Author: rana 
Type: Update 
Category: IOO 
Subject: some points about seismic FF 
  • When plotting the subtraction performance, we mainly care about the 0.5 - 10 Hz band, so we care about the RMS in this band. Don't integrate over the whole band.
  • When calculating the Wiener filter, you must use the pre-weighting so as to not let the Wiener residual be dominated by the out of band signals. We don't want the filter to try to do anything outside of the 0.5 - 10 Hz band.
  • Somehow, we want to assign a penalty for the filter to have high frequency gain. We do NOT want to slap on an ad-hoc low pass filter. The point of the Wiener filtering is to make the optimum.
  • What is the reason for the poor filter performance from 0.5 - 2 Hz ? If we use the frequency domain (Dmass) subtraction technique, we can do better, so there's some inefficiency in this process.
  • we're getting too much of the 3 Hz stack mode coupling into MCL. I think this means that our damping filters should be using RG around the suspension eigenmodes rather than just simple velocity damping. We had this years ago, but it caused some weird interaction with the angular loops...to be puzzled out.
ELOG V3.1.3-