J. Kissel, for K. Izumi, and S. Ballmer Now that Stefan and Kiwamu have the first reasonably successful evening with ALS DIFF (see LHO aLOG 11759), I wanted to make an assessment of how their filter design ended up compared to my original design (see LHO aLOG 11676). I attach the frequency dependence of each stage, divided into the distribution filters, plant inversion filters. Then I attach total path (with gain filters included). There's still a little bit of confusion about the distribution of over-all gain of the system i.e - the ALS DIFF PLL's error signal, which serves as the DIFF sensor, had a some calibration filters turned OFF for some reason, leaving the output in [V] instead of the planned [um] - converting the overall SUS actuator gain to be [ct_{TST DAC} / um_{ISC IN}] has not yet been installed, leaving the actuator gain still in [ct_{TST DAC} / ct_{ISC IN}] so it would take some work to understand the over gain of the loop and/or with which UGF Stefan and Kiwamu ended up. Contrasting the creation of the super actuator between SB + KI vs. mine: - They've used the blend filters I've installed identically, so no mystery there. - For the plant inversion, they've diverted from my design in two ways: - A much more accurate inversion of the resonant forest for each stage, resulting in high Q filters with long impulse responses -- the price paid for phase accuracy - They've inverted the resonant forest for the TST stage, creating a resultant very high-Q resonant feature in the inversion filter to balance the zero in the TST to TST plant at 0.87 [Hz]. I did not invert the TST stage, and managed to scrounge up enough phase for the *model* to be stable (but this means little compared with reality.) - Instead, they've turned the TST (and each subsequent stage above) into a 1 [Hz], Q = 1, single pendulum. Unfortunately, since the ALS DIFF lock isn't too terribly stable, we can't get more than a few frequency data points to confirm stability, but we'll try again tonight. I think the reasons why they were saturating with the overall unity gain frequency only at ~1.2 [Hz] (as opposed to my model of 15 [Hz]) is because of the mystery factor of 4 missing from the TST mass drive (see LHO aLOG 11676), and because we still don't have ETMY up and running because of ESD woes (yet to be aLOGged) -- i.e. a factor of 8 in drive strength that's missing. According to all the modeling done thus far and now a measurement, we really need it! Next up -- find the factor of four in the ESD and fix it! ------------------- Relevant Configuration Record (gain = 1.0 if not mentioned): H1:ALS-C_DIFF_PLL_CTRL FM3 "antiVCO" FM4 "cnts2V" G = 10.0 (LSC Input Matrix element = 1.0) H1:LSC-DARM_IN FM1 "ALSDIFF" FM2 "invPlant" FM3 "Boost" G = 10.0 (LSC Output Matrix element = 1.0) H1:SUS-ETMX_L3_ISCINF_L (these were the extra last minute filters added to try and manipulate the overall super-actuator from saturating) FM6 "p5z50" FM7 "p5z50" H1:SUS-ETMX_M0_LOCK_L FM1 "invL2L1" (part one of plant inversion filter) FM2 "invL2L2" (part two of plant inversion filter) FM3 "top/tst" (gain only filter to compensate for the gain ratio) FM5 "blend LP" H1:SUS-ETMX_L1_LOCK_L FM6 "invL2LNEW" (part one of plant inversion filter) FM7 "patch" (part two of plant inversion filter) FM8 "LISOfit" (part three of plant inversion filter) FM9 "uim/tst" (gain only filter to compensate for the gain ratio) FM3 "blendHP" (UIM-TOP HP complement) FM4 "blendLP" (UIM-TST LP complement) H1:SUS-ETMX_L3_LOCK_L FM6 "MatchedinvL2L" (part one of plant inversion filter) FM7 "patch" (part two of plant inversion filter) FM4 "blendHP-M0" (TST-TOP HP complement) FM5 "blendHP-L1" (TST-UIM HP complement) H1:SUS-ETMX_L3_LOCK_BIAS_INMON = 125000 [cts] No linearization algorithm installed.