Keita Daniel
With the most recent modifications we were finally able to engage the new outer loop ISS. The gain slider was at 10 dB with 2 W input power. We increased the power to 50 W, while running the feedforward correction for the AC coupling loop. This gave us an ISS with a ugf around 300 Hz. Increasing the gain further will require more loop shaping to avoid internal saturations.
The first attached plot shows the PSD at 50 W with the servo engaged, whereas the black reference trace is with the servo off.
The second attached plot shows the power up transition—indicating that the feedforward works pretty well.
We noticed is that the power fluctuations at longer time scales become larger, when we tried to adjust the offset into the first loop. The attached plot shows that fluctuations grow, when we approach 4% diffraction power, but are fine, when we are in the 2% to 3% range. Strange.
Morning meeting
- We had a nice lock last nice.
- PSL chiller swap planned
Activities (all time in UTC)
14:45 Chris in LVEA collecting ext cords not in use (1-2 hrs -- safety audit work)
15:47 ISS work (Daniel, Keita)
16:36 Fil + Elizabeth to MY
16:49 Fil + Elizabeth back
Fil to LVEA woring by PSL rack (ISS chasis pull out)
~17:15 ISS work done. PSL begin chiller swap
17:33 Rebooting framewriter1 (Dave, Jim)
18:01 framewriter1 back, restarting ISS model + DAQ
18:14 Fil going back to LVEA
18:33 PSL done
18:35 Jason to LVEA toggling noise eater
18:36 Fil out
19:09 Jason resetting noise eater (again)
21:07 Dick, Betsy to MY
21:20 Sheila to LVEA HAM6 area
21:40 Betsy+Dick back
22:02 Gerado to CP3
22:06 Fil to LVEA by PSL rack (ISS chasis work)
22:15 oaf ADC/DAC error. Dave restarted the model in time to save TCS.
22:36 Fil back
22:39 Dick to CER
OpLev trends from the past week. FAMIS #4692
Everything looks normal here.
Fil Keita Daniel
At 50 W input power we are now saturating the amplifier stage, since there is too much whitening upfront and the compensation is after the gain. We moved the 20 Hz/380 Hz compensation to the transimpedance board to give us more headroom. Continuing from alog 29527 and alog 29293 we changed:
Some changes were made to the front-end model to change the polarity of the 0.1 Hz switches and to add filter modules to the PD readbacks. This allows adding anti-whitening filters. Since the AC readbacks now have a DC gain of 1, we don't need the dedicated DC readouts anymore.
Attached is the transfer function of one of the transimpedance PD chains.
Gain reallocation once again:
This should allow us to increase the gain by up to 30 dB and reach an ugf of maybe 10 kHz.
After finding that the ISS 2nd loop pitch and yaw signals are swapped, and inverted wrt im4 trans (my first look at qpds downstream of the IMC, alog 29542), I've looked at four qpds downstream of the IMC for signal consistancy - IM4 Trans, ISS 2nd Loop, POPA, and POPB.
What I found:
chart showing the current orientations:
pitch | yaw | |
IM4 Trans | 1 | 1 |
ISS 2nd loop | -yaw | -pitch |
POPA | -1 | 1 |
POPB | ? | 1 |
Attached is a two hour second-trend plot of the DAC drive for the TCS ITMX temperature setpoint and the chiller water flow rate (which shows when the chiller tripped). The data is from Wednesday night, 10pm to midnight PDT.
At 22:42 PDT the IOP stopped driving the DAC outputs (presumably due to a FIFO error) with the user supplied signals and instead outputted zero volts.
At 23:32 PDT the chlller tripped (+50mins later)
Nutsinee says that the nominal driving temperature is around 20C, the trip point is 15C, and the maximum temperature rate-of-change is 0.1C/minute. The transition from 20C to 15C would then take 50 minutes, which agrees with the timing seen on Wednesday.
The plot shows that after the chiller tripped, the control room was alerted of the problem, the front end was restarted and the chiller was reset within 10 minutes.
Clearly if the verbal alert system warned the control room at the time the DAC output went to zero, in this failure mode we will have plenty of time to restart the OAF IOP model and recover chiller operations.
P. King, J. Bartlett, J. Oberling
We swapped the PSL chillers for the backup in an effort to isolate the recent laser tripping issues. The laser is now back up and running.
Operating hours for the chillers, as reported by the chillers themselves and the chiller MEDM screen, are:
Chiller | MEDM | |
Crystal | 148.75 | 33317.00 |
Diode | 86.25 | 33560.00 |
There is a difference between the chillers and MEDM because the MEDM screen gets its info from the PSL Beckhoff PC, which uses its own internal counter to track the chiller hours instead of the hours recorded by the chillers themselves; we currently do not know a safe way to reset this Beckhoff counter.
I wrote a simple interface to download data from the Tektronix network/web-enabled oscilloscopes (TDS3034B and (presumably) the like). You can use it to download a single data file, or download data periodically:
jameson.rollins@opsws8:~ 0$ readlink -f $(which tektronix)
/opt/rtcds/userapps/trunk/sys/common/src/python/tektronix.py
jameson.rollins@opsws8:~ 0$ tektronix -h
usage: tektronix [-h] <command> ...
Interact with Tektronix oscilloscope over the network.
positional arguments:
<command>
download Download data from scope periodicially
optional arguments:
-h, --help show this help message and exit
Use environment variable HOST to specify network address.
(default: 10.22.10.51)
jameson.rollins@opsws8:~ 0$ tektronix download -h
usage: tektronix download [-h] [-c CHANNEL] [outdir] [count] [period]
Download data from scope periodicially
Data will be saved in a two-column (time,voltage), comma-separated CSV
file named "<time>.dat", where <time> is an ISO-formatted timestamp of
the local time of the data snapshot.
NOTE: data takes about 30 seconds to download, so period should not be
less than 30 seconds.
positional arguments:
outdir directory to store timestamped data ['.']
count number of downloads (0=Inf) [1]
period time between downloads in seconds [60]
optional arguments:
-h, --help show this help message and exit
-c CHANNEL, --channel CHANNEL
channel to download ['ch1']
jameson.rollins@opsws8:~ 0$ tektronix download
output directory: /ligo/home/jameson.rollins
fetching 2016-09-09T11:16:31.747247.dat ...done (in 29.959158s).
jameson.rollins@opsws8:~ 0$
For the current OSA set up, we are using scope2 which has an IP address of 10.22.10.51
WP6146, fix BIO bug, add PD filters.
Daniel, Keita, Dave:
h1psliss mode was restarted, followed by a DAQ restart.
model restarts logged for Thu 08/Sep/2016
2016_09_08 08:25 h1calcs
2016_09_08 08:25 h1iopoaf0
2016_09_08 08:25 h1ngn
2016_09_08 08:25 h1oaf
2016_09_08 08:25 h1odcmaster
2016_09_08 08:25 h1pemcs
2016_09_08 08:25 h1susprocpi
2016_09_08 08:25 h1tcscs
Restart of h1oaf0 models (specifically the iop) to clear a re-occurance of a dac-fifo error.
Came into the Control Room to find that the laser had tripped. It was up and running an hour ago when I checked it from my office. Same problem as before. The error bit for the external flow sensors was not set in Beckhoff, so the problem is not with the ADC terminals. The flow sensors are pretty close to brand new. Conclusion is that it's the chiller that's glitching, for whatever reason.
In scrolling through the various chiller menus, I did not see any obvious error messages other than the ubiquitous flow sensor 1 error message. However I did make a note that the number of operating hours for the pump (9577) is different than the operating hours reported by the chiller (10490). Not sure why.
Kiwamu, Stefan
We had two long-lock attempts tonight. During both we tracked the the REFL optical gain to frequency noise, as well as the ASC optical gain to an IM4 pitch dither.
The first one we started from a 20W lock that had more CO2 central heating on the X compensation plate (see Kiwamu's log).During this first lock we again saw the sign of the REFL45_I signal to frequency noise flip. This happened about 15min after reaching 50W.
The second lock started with the regular TCS settings. We reached 50W, and reduced the 45 modulation index by 2dB. Guardian now uses the now BS ASC input matrix, only using AS_A_36 with a gain of 1.5 (PIT and YAW).
In this setup we were able to hold the lock for 1h20min.
Attached is a gain tracking plot for that 2nd lock. Note that there is a number of changes in REFL gains that will affect out ASC system:
- While the power is ramped up there seems to be a Gouy phase shift - with the signal in REFL_A_45 decreasing, and the signal in REFL_B_45 increasing.
- However, after that both 45 ASC signals show almost no change in gain.
- On the other hand, the two 9MHz signals drop their gain by about a factor of 3.
- Also note the episode around -50min - we manually tried to tweak IM4 and PR3, because we were afraid that the input matrix wasn't putting us to a good spot any more.
- Over the 80min the sideband 2-omega kept dropping slightly, but was fine otherwise. Towards the end we notced the AS_90 get rattier. We suspect that this is either due to the BS being slightly misaligned at DC, or due to some ASC loops significantly loosing gain.
Bottom line:
- It still seems possible to get the interferometyer running indefinitely at 50W, but we have to spend some more time understanding how the ASC gains change.
- But this still leaves some pretty uggly higher order modes in the interferometer - I have no idea what that will do to our noise couplings.
LEGEND for plot:
H1:LSC-LOCKIN_1_DEMOD_4_I_OUTPUT LSC: laser frequency to REFL_9_I (blue)
H1:LSC-LOCKIN_1_DEMOD_3_I_OUTPUT LSC: laser frequency to REFL_45_I (dark red brown)
H1:ASC-LOCKIN_OSC1_DEMOD1_I_OUTPUT ASC-REFL_A_RF45_I_PIT gain (IM4 dither) (orange)
H1:ASC-LOCKIN_OSC1_DEMOD2_I_OUTPUT ASC-REFL_B_RF45_I_PIT gain (IM4 dither) (yellow)
H1:ASC-LOCKIN_OSC1_DEMOD3_I_OUTPUT ASC-REFL_A_RF9_I_PIT gain (IM4 dither) (light blue)
H1:ASC-LOCKIN_OSC1_DEMOD4_I_OUTPUT ASC-REFL_B_RF9_I_PIT gain (IM4 dither) (dark blue)
PS:
Some of the mysteriously unstable loops earlier today were due to the DOWN script now setting back all the gain changes related to the RF modulation depth reduction. This was fixed tonight.
At around 3:00 local, I increased the ring heater power on ITMY from 0.5 W (0.25 W each segment) to 1.5 W (0.75 W each segment). This has to be cancelled by applying approximately an additional 400 mW on CO2Y.
Jenne, Matt, Patrick, Sheila, Stefan, Kiwamu,
We tried two different differential CO2 settings today at 20W to see if we can improve the imbalanced RF sidebands (29535). One of the two tests was already reported by Jenne et al. on 29553.
According to the tests, the addition of a 10 uD single-pass substrate defocus to the X compensation plate seems to balance the amplitude of the upper and lower 45 MHz sidebands.
[Settings]
The interferometer arrived at a 20 W lock at 22:53 local. We then added a 800 mW of laser power on to CO2X, resulting in a 1 W CO2 power at 22:55:40. As opposed to the previous test (29535), we did not change CO2Y this time and therefore it stayed at 0 W.
The measurement finished at 23:15 and we set CO2X to 700 mW. We then increased the PSL power to 50 W while maintaining the same lock.
[Results]
As expected from the previous test, the imbalance of the 45 MHz RF sidebands, as seen by the optical spectrum analyzer, was improved by this action. After roughly 20 min, we reached a point where the upper and lower sidebands almost had the same amplitude. HWSX was functioning at the beginning of the test (for 10 minutes or so), but then unfortunately started showing some meaningless signals. We have no idea why. Nevertheless, we were able to extrapolate where we would be in terms of the substrate defocus based on the initial rise. It looks like that an additional 20 uD double-pass (= 10 uD single pass) is the good defocus value for ITMX in order to get balanced upper and lower 45 MHz sidebands. A 10 uD single-pass defocus corresponds to a 400 mW CO2 laser on the X compensation plate in equilibrium [assuming a 25 uD/W actuation coefficient (single-pass), 28799].
The attached is a screenshot of OSA outputs in which scans from different time are overlaid. It is clear that the imbalance improved as time elapsed.
On the other hand, the frequency noise coupling became worse by a factor of a few comparing to the initial coupling value at 900 Hz. It did not experience a sign change. Not sure what this means.
[Sheila, Kiwamu, Matt, Jenne, etc]
We changed the TCS CO2 settings after we had thermalized at 20 W for a while, and were watching several metrics. Here I'll talk about the frequency line at 900 Hz, and Sheila will append her OSA plots as a function of time.
I injected a line at 900 Hz into H1:LSC-EXTRA_AO_2_EXC, which goes to H1:LSC-REFL_SERVO_COMEXC. I demodulated Refl 9 and Refl 45 like Stefan did last night (alog 29509), and also added another demodulator of OMC DCPDs. For each of these, I ran a zservo to adjust the demod phase to keep the Q signal at zero.
In the attached plot, bright red is the PSL power, and yellow and taupe are the X and Y TCS CO2 powers. Purple, cyan and brick red are Q-phase signals, so can be ignored. Green is REFL 9, blue is REFL 45. Orange is OMC DCPDs, indicating the coupling of frequency noise to DARM. Note that the range for blue, green and orange are all symmetric about zero, so while it looks like the frequency coupling is decreasing, in fact it is increasing. Sorry for a poor choice of demod phase.
After about 40 min at 20 W, the CO2 powers were stepped differentially. X went from 0.2 W to 0 W, while Y went from 0 W to 0.2 W. When this happens, you can see that the frequency coupling to DARM gets worse even faster than it had been while thermalizing with the usual TCS settings at 20W. The REFL 45 and REFL 9 traces don't change in character when we alter the TCS settings. Clearly this is the wrong direction for freq noise coupling, but it's nice to see that we can definitely affect it. Sheila will post her results when her plots are ready, but it looks like the sideband imbalance gets worse when we made this TCS step, so it's good that everything hangs together.
Next lock, we'll try increasing the ITMX CO2, and leave the ITMY CO2 at 0 W. This is a combination of common and differential, but it's all we can do before stepping the ring heaters later tonight.