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  Thu Aug 1 12:12:35 2013, gautam, Update, CDS, New model for endtable PZTs C1ASX_GDS_TP.png
    Reply  Thu Aug 1 13:09:17 2013, gautam, Update, CDS, New model for endtable PZTs-procedure 
    Reply  Thu Aug 1 15:06:59 2013, jamie, Update, CDS, New model for endtable PZTs 
       Reply  Thu Aug 1 15:28:44 2013, gautam, Update, CDS, New model for endtable PZTs-problem solved 
    Reply  Thu Aug 1 20:58:56 2013, gautam, Update, CDS, New model for endtable PZTs-MEDM Screens setup MAIN_SCREEN.pdfMATRICES.pdfLOCKINS.pdfCONTROL_FILTERS.pdf
       Reply  Mon Aug 5 18:18:32 2013, gautam, Update, CDS, Choosing LO Amplitudes and Frequencies  GTR_Power_Spectrum.pdf
          Reply  Mon Aug 5 19:10:01 2013, Koji, Update, CDS, Choosing LO Amplitudes and Frequencies  
             Reply  Tue Aug 6 16:36:51 2013, gautam, Update, CDS, Choosing LO Amplitudes and Frequencies-revised powerSpec0806.pdf
                Reply  Wed Aug 7 23:40:49 2013, gautam, Update, CDS, X-End Green ASS - A first update time-series.pdf
                   Reply  Sat Aug 10 05:53:51 2013, gautam, Update, CDS, X-End Green ASS - Roundup New_ASX_MEDM_MAIN.pdf
                      Reply  Mon Aug 12 12:57:59 2013, Jenne, Update, CDS, X-End Green ASS - Roundup 
                         Reply  Mon Aug 12 13:30:33 2013, Jamie, Update, CDS, X-End Green ASS - Roundup 
Message ID: 8993     Entry time: Sat Aug 10 05:53:51 2013     In reply to: 8983     Reply to this: 8995
Author: gautam 
Type: Update 
Category: CDS 
Subject: X-End Green ASS - Roundup 

Over the last three days, I've had the interferometer to test and optimize the ASX Servo. Based on what I have seen, I think the conclusion is that with the current parameters, the servo does its job provided the input pointing set up at the endtable with the coarse adjustment knobs is reasonably good. Once the cavity is aligned and IR transmission maximized using ASS, I have been able to get the green transmission up to 0.8 which is close to the best we had pre-vent. I have not been elogging regularly over the last few days, so this one is going to be a longish one.

Major changes made:

  1. The SIMULINK model has been modified to accommodate an option to dither the cavity mirrors and not the PZT mirrors. Details are as follows:
    • I have sent the LO signals (CLK,SIN and COS) from the ASS model to the ASX model via the RFM model. Appropriate changes were made to all these three models, and recompiling and restarting the models was done without issue. The SIN and COS signals are used to demodulate green transmission at the dither frequencies. ***The CLK signal is not required to be sent between models as it is not being used by ASX (I turn the dither ON using the channels already set up for ASS). I realised this a little late, and at present the ASS and RFM models are compiled such that the CLK signal is also sent from ASS to RFM. This can be removed, thus freeing up 4 unnecessary inter-process communication channels. Also, I am not too sure if this is relevant, but the maximum computation time of both the RFM and ASX models seem to have gone up after I added these inter-process communication links.***
    • The rest of this part of the servo is a replica of the part where PZT mirrors are dithered. At present the servo output is the sum of its two branches (PZT mirror dither branch and cavity mirror dither branch) which works fine under the assumption that at any one time, only one arm will run. Ideally, the summing block should be replaced by a switch. However, when I tried (in an earlier attempt to include the cavity dither) to do this and restart the model, c1iscex crashed, and so I decided against using the switch block for this trial.
    • The control signal generated using green transmission demodulated at the ETM dither frequencies are used to actuate on M1 while the ITM ones are used to actuate on M2. Of course, by setting the appropriate off-diagonal elements in the output matrix, this can be modified as desired.
  2. The main MEDM screen has been updated to reflect the new additions to the SIMULINK model. Screenshot is attached. The picture isn't entirely accurate as the monitor channels in the upper row actually show the servo output + slider output. This needs to be changed in the model, and a new set of monitors need to be added to the MEDM screen. In the end, we require four sets of monitor-points in the model: PZT dither servo output, cavity dither servo output, sum of these with any offset from the PZT sliders, and the sum of the latter with the dither signal (this is what eventually goes to the PZT mirrors while the dither is on).
  3. I added scripts to the MEDM screen that turn the PZT mirror dither servo on and off. Note that when you want to run a new script on an MEDM screen using medmrun, you need to change the permissions of the file by going to the path where your script is located and running chmod 755 <name of script>. Manasa has updated the same on the wiki.

 Details of tests runs:

For the most part, I have been trying to optimize the PZT mirror dither servo. To this end, I did the following:

  • Went to the X-end and fixed the input pointing which was not optimal. Manasa first aligned the arm and ran ASS to maximize the IR transmission. I then used the coarse adjustment knobs on the mirror mounts to get the green transmission up to ~0.6.
  • I then set the following parameters in the servo (these are all in the script, path to which is /opt/rtcds/caltech/c1/scripts/ASX):
    1. LO frequencies of 10, 19, 34 and 39 Hz respectively for M1 PIT, M1 YAW, M2 PIT and M2 YAW.
    2. LO amplitudes of 75 for all the four degrees of freedom (determined by using PZT calibration to see what amplitude would couple 10% of power into the first higher-order-mode assuming a perfectly aligned beam to start with.
    3. LIA BP filters centered at the above frequencies with 2Hz passband on either side.
    4. LIA LP filters with corner frequency 0.5 Hz.
    5. LIA Signal filter bank gain set to 100 for all degrees of freedom.
    6. LIA Demod I phase filter bank gain set to 5 for all degrees of freedom.
    7. Control filter gains to 1 for all degrees of freedom (control filters are all integrators).
    8. Demod phase set to 0 for all degrees of freedom. I did not really try to optimize this but the servo seems to be doing reasonably well even with this setting.
    9. Overall servo gain to 1 (the servo worked well when I increased this to 5 as well, but became unstable when I increased it further).
  • I ran the servo. Observations were as follows:
    • Having fixed the input pointing to get green transmission up to ~0.6, the servo was able to improve it to ~0.8, which is the best we had after hours spent at the X-end prior to the vent.
    • Given a good input pointing, we can use the PZT mirrors to lock to 00 mode from some misaligned state using either the sliders, or by leaving the servo on, and helping it out at the points where it gets 'stuck' in some higher mode using either the sliders or by toggling the shutter.
    • In order to recover green transmission of ~0.8, it was often necessary to first run ASS to optimize the IR transmission. Otherwise, green-transmission saturates at ~0.6 or 0.4 depending on the misalignment of the arm cavity mirrors. The servo was unable to change the input pointing enough to deal with overly misaligned cavity mirrors. 
    • The servo is sometimes capable of bringing about mode-hopping from a higher order mode to a lower one, though this is not always the case as the PDH lock is sometimes too strong, in which case I toggled the shutter after which the servo kicked in.
    • I tested the servo under as many different conditions as I could. For instance, having left the green shutter open overnight, I saw that the transmission had fallen from 0.8 (which was what we saw on Thursday night) to ~0.4 on Friday morning. Running the servo got the transmission up to 0.6. I then asked Manasa to run ASS, (while leaving the ASX servo on), after which point the green transmission went up to 0.8. Sometimes, the servo locks to a 'bad' 00 mdoe, where the transmission saturates at ~0.2, but toggling the shutter fixes this most of the time.

Attempt to measure transfer function:

One of the things that came up during my presentation was how fast the loop was capable of responding. I was able to get a quantitative idea of this by playing around with the overall servo gain. Initially, it took ~30 seconds for the servo to get the transmission up to its peak value, with a servo gain of 1. When I ramped this up to 5, the response was much faster, with the peak transmission being reached in ~5seconds. 


I wanted to get a more quantitative picture, and hence tried to measure the transfer function by first injecting an excitation into the 'SIG' filter-bank in the demodulation stage. However, coherence between the IN1 and IN2 signals was very poor for all the amplitude configurations I tried. At Jenne's suggestion, I tried injecting the excitation at the control-filters stage, but found no improvement. Perhaps the amplitude envelope was wrong or the measurement technique has to be rethought. 

 Misc remarks:

  1. M1 is the first steering mirror and M2 is the second one (right before the beam enters the arm cavity).
  2. Though I have added the cavity dither feature to the model, I was not able to optimize this servo. Some calculations need to be done to get an estimate of the output matrix, after which the filter gains etc can be optimized.
  3. Today, I cleaned up my temporary setup at the SP table to calibrate the PZTs. Most of the hardware for the Y-end is now in the tupperware box. The QPD and laser have been restored to the optical bench next to MC2 where I found them. The second KEPCO HV supply which I had set up has now been installed at 1Y4 in anticipation of the PZT mirrors at the Y-endtables. It is currently powered OFF.
  4. Performance plots to follow as I have not pulled the data out yet.
  5. I had bought a cake from chandler today in an effort to clear my meal plan, but in the rush in the afternoon, completely forgot about it. It is in the fridge, and is strawberry tart, hope it tastes good.


 New MEDM screen:


ELOG V3.1.3-