J. Kissel, J. Warner

In summary -- in the "windy" configuration that includes the BRS (described below) we can achieve better performance, as measured by the local T240s, than when just switching to 90 [mHz] blends at almost all frequencies. This is especially true in the 20 to 80 [mHz] band, where ALS lock acquisition is limited by end-station VCO range. In this study, the wind is a consistent 10-20 [mph]. Not necessarily "high," so we'll still need more data during those conditions to confirm if there's an upper limit to performance improvement, i.e. if / when the BRS saturates and it's corrective signal becomes detrimental to the platform motion.

Discussion & Details
---------

I took some more data with H1 ISI ETMX in several configurations, looking to supplement Krishna's data on the performance impact of the BRS during 5-10 [mph] winds (see LHO aLOG 16465), this time with a confirmed consistent wind of 10-20 [mph] and a more-typical ground translation from the microseism. Further, since Krishna's study, Jim, Hugh, and I have made significant improvements to the BSC-ISI performance from beating down unexpected noise sources (see e.g. LHO aLOG 17702, Integration Issue 1004, LHO aLOG 16818, etc.) the results are considerably less confusing. This also serves to show what we get when we'll eventually automate switching between these configurations (see LHO aLOG 17639).

I compare three different configurations of the ISI's ST1 X DOF sensor correction and blend filters, over a short, 2-hour window where X-End wind has a consistent, minute-trend, mean of 10 [mph], and a consistent, minute-trend, max of 15-20 [mph]. The three configurations of the ISI's ST1 X DOF are as follows:
(1) Nominal -- 45 mHz blend; DeRosa's 0.43 Hz only, narrow-band, sensor correction (NB SC); GND T240 alone used for sensor correction, no BRS
(2) Windy when BRS doesn't work -- 90 mHz blend; DeRosa's 0.43 Hz only sensor correction (NB SC); GND T240 alone used for sensor correction, no BRS
(3) Windy with a functional BRS -- 90 mHz blend; Mittleman's broad-band, low-frequency, sensor correction (BBLF SC); Tilt is subtracted from the GND T240 with the BRS, and the super sensor is used for sensor correction.

Recall that plots comparing the X direction 45 and 90 mHz blend filters can be found in LHO aLOG 17595.

Comments:
- Look at 2015-04-07_H1ISIETMX_SensorBlend_Config_Comp_X.pdf first. This shows the X direction.
   PG1: Performance of configuration (1) has a significant amount of gain peaking (about a factor of 4) from the 45 mHz displacement sensor blend filter from 25 to 80 [mHz], as expected. This the sacrifice made to get the awesome performance between 0.1 [Hz] and 0.4 [Hz]. When we move the blend frequency up to 90 mHz, to configuration (2), the factor-of-four gain peaking moves up as well to 40 to 120 [mHz]. However, this shift up an overall RMS displacement reduction of about a factor of 5. In doing so, we lose a factor of 10 in performance between 0.1 [Hz] and 0.4 [Hz], and also some loss between 1 to 10 [Hz]. In configuration (3), The BRS+STS super sensor coupled with the broad-band sensor correction allows us to claw some of that performance back, *and* reduce the gain peaking to essentially zero.
        The trouble with interpreting ASDs is that the contain in coherent noise of the platform as well. Of course (save where the measurement is readout noise limited) this is the real motion of the platform, but it's less easy to see what's going on. Hence, 
   PG2: Comparing the coherent, linear transfer functions in each state, we see much more clearly what's going on: configuration (1) has x4-5 gain peaking between 25 and 80 [mHz]. One might even argue that we could tune the narrow-band sensor correction better, because it's performance is best at 0.35 [Hz] instead of 0.43 [Hz]. When we switch to configuration (2), the performance follows the change in displacement sensor filter, as expected. Finally, in configuration (3) we're basically blending in the tilt-free, inertial ground super sensor at 20-30 [mHz]. So we win back all of the noise introduced when the noisy displacement sensor is used out to high frequency, and in the 0.1 to 0.4 [Hz] band, we're only at most a factor of 2 to 3 away from the best nominal configuration. In fact, the performance is *even* better than the nominal configuration between 0.4 and 2 [Hz].

- Now look at 2015-04-07_H1ISIETMX_SensorBlend_Config_Comp_X_SensCorrSignal_ASD.pdf
    Unfortunately, we don't record the pre-sensor corrected, calibrated, Cartesian displacement sensors. In fact, there isn't even a test point for these channels. As such, there isn't a good way to compare the performance of the NB SC filter and the BBLF SC filter, using the CPS. As a proxy, I took a look at the output of the sensor correction path, just *before* it's subtracted from the CPS. The two signals should be equivalent to the CPS in the band that we use the sensor correction modulo a sign which doesn't affect the ASD. We can see that Configuration (3) has the *least* amount of sensor correction request below 0.1 [Hz], because the BRS has subtracted out the tilt from the GND T240 of this region.

- 2015-04-07_H1ISIETMX_SensorBlend_Config_Comp_RY.pdf shows the RY direction.
    PG1: Shown merely to demonstrate that the ground tilt ASD was essentially the same for all three of these measurement configurations, as was the residual tilt of ST1.
    PG2&3: Shown because I can -- the transfer function between ground tilt and platform tilt. Because we've reduced the HEPI pump servo noise (II 1004) we see now that the platform's RY motion is limited by and coherent with ground RY below 0.2 [Hz] as originally expected. Good!
    PG4&5: This is the transfer function between ground tilt (RY) and platform translation (X). It's pretty scary, but I'm not sure I trust it. I'm not at all confident that DTT is able to handle calibrating the transfer function between these two correctly. Anyways, I include it, again, because I can. Aside from the magnitude, pg 5 does clearly show that platform X is coherent with ground tilt.

This data set settles the question of whether the configuration (3) is better than (1) in 10-20 [mph] winds, especially if just going to configuration (2) during these kinds of winds is enough to lock the IFO. Now we just need to perform the same study at even higher winds. Looks like this Saturday may be a good candidate!

The template and all of the plots for this entry live under
/ligo/svncommon/SeiSVN/seismic/BSC-ISI/H1/ETMX/Data/2015-04-07*