40m QIL Cryo_Lab CTN SUS_Lab TCS_Lab OMC_Lab CRIME_Lab FEA ENG_Labs OptContFac Mariner WBEEShop
  Cryo Lab eLog  Not logged in ELOG logo
Entry  Tue Oct 9 00:44:22 2018, johannes, Notes, Cryo ISS, New Transmission PDs and ISS 8x
    Reply  Thu Oct 11 01:25:43 2018, johannes, Notes, Cryo ISS, ISS Servo Characterization 7x
       Reply  Tue Oct 16 00:02:15 2018, johannes, Notes, Cryo ISS, ISS Servo Characterization planex_rin.pdf
Message ID: 2181     Entry time: Thu Oct 11 01:25:43 2018     In reply to: 2180     Reply to this: 2182
Author: johannes 
Type: Notes 
Category: Cryo ISS 
Subject: ISS Servo Characterization 

Attachment #1 (A1) shows the optical table setup until just before the beams reach the EOMs. The two lasers are focused to r=~250 um in the AOMs have their polarizations cleaned up to P by a sequence λ/4 → λ/2 → PBS. Before changing the setup such that the 0th order undeflected beams are sent into the cavities, the optical ringdown was measured to determine the mirror loss (hadn't been done before at cryo temps). The PDA10CF seen in A1 was used as a power monitor, and a second PDA10CF was placed in transmission of the cavities to record the ringdown. Attachment #2 (A2) shows one of the recorded ringdowns, its exponential fit, and the witness power decay on the pickoff PD, showing that the bandwidth of the PDA10CF (~110 MHz) does not limit the ringdown measurement.

Taking into account the 65+/- 3.5 ppm mirror transmission reported in elog 1988, the scatter+absorption loss per mirror is 4.8 +/- 3.5 ppm for EAST, and 6.2 +/- 3.5 ppm for WEST.

The Z-pair mirrors just after the AOMs were then used to instead steer the undeflected AOM output into the cavities, for which irises were placed at strategic points to make subsequent alignment easier. For modeling the ISS loop the transfer function from AOM (commercial) driver input to the Pickoff PDA10CF was measured. For this the AG4395A output was merged with a 500mV offset supplied by a DS345 using a bias-tee, and a FET probe was used to sample the AOM input directly. The two AOMs had previously both been aligned for maximum deflection efficiency, but the East path was showing significantly more phase loss than the West (55 degrees at ~120kHz vs 55 degrees at ~350 kHz ), so I moved the East AOM on its perpendicular translation stage to bring the beam closer to its transducer and cut down on the propagation delay at the expense of conversion efficiency, which helped.

Being driven at saturation ("cw" setting on the commercial), the AOMs diminish the undeflected beam by 80.9% on the East, and 88.6% on the West path. Attachment #3 (A3) shows the updated recorded responses for both paths. I fitted a model to the data which contains a time delay and multiple poles to force the rapid roll-off above 1MHz. Instead of allowing for a fit to multiple individual poles (many are needed!), I assumed a single frequency and adjusted the pole order until the magnitude decay matched nicely (which was achieved for order ~30!!!). The phase is not a great match, but for ISS modeling purposes this does not matter as much. The fits, shown in A3 as well, report delay parameters of 116 ns for East, and 102 ns for West for pole locations of 13.67 MHz and 14.47 MHz, respectively.

Next the transfer function AOM driver → TPD output was measured. Dividing out the cavity and the AOM responses yields the TFs shown in Attachment #4 (A4). Evidently the majority of the phase lag to be considered for the stabilization comes from the AOMs, and the custom TPDs only add marginally to it.

The open loop gain of the west ISS servo (without the boost engaged) with components populated as described in D1800214-v1 was measured using a fake plant to loop back to its input. The fake cavity is a passive RC circuit with a single pole at 45 kHz. Dividing the open loop gain by the fake cavity TF and multiplying it with the measured plant response results in Attachment #5 (A5). The two ISS Servo channels were virtually identical, so I'm only showing one of them. One needs to decrease the open loop gain by a factor ~4 and then some to make it stable with the real plant

Attachment #6 (A6), which shows a measurement of the ISS OLG using the SR785 (again without boost). The maximum is located at ~430 Hz, as designed. A pole-zero fit returned the following parameters:

Vanilla ISS Servo Fit
Gain 2.78
Delay 184 ns
Zero 1 0 Hz
Zero 2 45 kHz
Pole 1 311.2+305.8i Hz
Pole 2 331.4-332.6i Hz

Instead of decreasing the total gain by four, the servo shape was instead altered, lowering the peak frequency by a factor of ~4. This reduces the gain at ~100 kHz by the required amount, but doesn't sacrifice peak gain. This was achieved with the following component substitutions

Component Old value New Value
C16 4.7 nF 47 nF
C17 10 uF 22 uF
C11 4.7 nF 15 nF
C5 4.7 uF 14.7 uF

Attachment #7 (A7) shows the expected shape of the new servo, and the the post-modification measurement.

Attachment 1: aom_setup.pdf  1.028 MB  Uploaded Thu Oct 11 12:11:25 2018  | Hide | Hide all
Attachment 2: cavity_ringdowns.pdf  116 kB  Uploaded Thu Oct 11 12:34:58 2018  | Hide | Hide all
Attachment 3: aom_only.pdf  29 kB  Uploaded Thu Oct 11 12:45:51 2018  | Hide | Hide all
Attachment 4: iss_plant_no_cav.pdf  37 kB  Uploaded Thu Oct 11 13:16:23 2018  | Hide | Hide all
Attachment 5: ISS_OLG.pdf  36 kB  Uploaded Thu Oct 11 15:18:42 2018  | Hide | Hide all
Attachment 6: ISS_SR785_old.pdf  21 kB  Uploaded Thu Oct 11 15:36:39 2018  | Hide | Hide all
Attachment 7: ISS_SR785_new.pdf  26 kB  Uploaded Thu Oct 11 15:57:50 2018  | Hide | Hide all
ELOG V3.1.3-