This morning we adjusted the 0.4mm roll out of the MC2 lowest stage via rolling the optic in the sling. We were able to take the roll error from 0.4mm on each side to 0.003mm error, well within the 0.25mm guideline.
We reset the M3 stage AOSEMs and are now back to running TFs of MC2.
For the record, I've done the following: ~ $ cd /opt/rtcds/userapps/release/sus/ sus $ svn up The "common" folder was what I was specifically looking to make sure was up to date (and now the folder has all the new quad changes, as well as the CD_STATE_MACHINE updates), but the update also received a whole bunch of (other) site specific stuff.
We received a text message from the alarm handler this morning regarding a signal associated with GV16 at the XMID station. After reviewing the signal in question we realized that this signal had fallen to zero on July 20 from a nominal 1ma (scale of 1-10 mA)
This looks like the normal failure mode of the annulus ion pump power supply and will be investigated and hopefully repaired (depending on replacement parts) next week.
Signal is HVE-MX:GV16_II339
The tripleteststand had the One-stop cable and cards (between the computer and the I/O chassis) replaced yesterday in an effort to solve the mystery of the models freezing and the cards in the I/O chassis becoming invisible. The x1sushxts27 model was started Wednesday 7/25 at 14:06 PDT and is still running this morning.
The END X volume was vented yesterday so work may proceed at BSC9. During turbo pump startup we found that the QDP80 was rotating the wrong way. This was corrected and the turbo was spun up as a precaution.
I started pumping the LVEA vertex volume yesterday. Roughing was paused overnight and will continue today. Pressure is currently 60 torr.
We looked into the PLL servo. First we checked the temperature feedback signals to the laser. They did seem ok, although the green had a 1mV offset (although not being written to it). We disconnected the green temp feedback.
The IR laser crystal temperature feedback was ok, but only after we managed to reset it to zero. Even the EPICS display was not showing all the digits (we got an intergrator gain of 1e-6). The EPICS screen displays voltage, which is good to know.
We had to adjust the temprature on the laser to find the beatnote, it was set back to 42.1 deg C. The knob on the front panel is very 'coarse', so the beatnote was within 10 MHz from the nominal 39.7 MHz. We engaged the temperature servo and we saw the beatnote move towards the 39.7 MHz. The Phase-Frequency Discriminator has a flat gain when the beatnote is far away from the LO frequency (39.7 MHz). Although not a problem, with low gain you do wait for it:) You can speed it up by engaging the CMB-A servo, which feedback to the PZT. Matter of fact, the temp signal is a plain copy of the PZT signal, which goes through the Beckhoff system and has it sown little servo.
We looked on an oscilloscope at the temperature feedback signal. There are a few ripples in the signal at ~1 micro second, ~5 microseconds and ~2 milli seconds. I think we should add an analog 10 Hz low pass filter at the output to remove these. The temperature feedback has a bandwidth of up ~0.5 Hz anyway.
Looking at the PZT feedback signal. It has ~1.5Vpp 'oscillations' with a ~10s period appearing and then fading away. This is of course the 'equivalent' noise the laser tries to follow. The PZT frequency to volts conversion is ~3 MHz/V, so that indicates that the laser frequency fluctuates by ~4.5 MHz!
Looking a the I-mon from the Phase-Freq Discriminator it is flat when locked (and around the 39.7 MHz). When further away the signal looks like 'pulse-widht-modulated' with an average voltage which correspond to how far away the beatnote is and is expected. I don't know how far the beatnote needs to drift, at which point the frequency discriminator takes over from the phase discriminator. All-in-all the I-mon doesn't seem to be bad or noisy.
We looked into the PLL which seemed to be ok (see other entry 3601). We didn't pay much attention to the RefCav yet, as we want to to make the cavity lock again. To make this happen we had to reset the ETMY HEPI, lucky Fabrice was here to help. The output on ISC BLEND_RZ was off. All is well.
We have fringes again, and the REFL_PWR_MON seems to be at its previous healthy ~11000 counts level. When engaging the cavity locking (CMB-B), it seemed to be ok until is starts the oscillation at 10-20 sec period. I turned the first boost filter off which didn't seem to help much. The VCO frequency drifts upto 39.78 MHz (from 39.603 MHz nominal). It almost looks like there is a sign wrong, it locks, fluctuates around 0 then creeps to the -10V limit. When you look at the VCO frequency it keep slowly climbing. At the same time the cavity power drops. Hmm, maybe still something in the PLL servo is not right.
We have two new indicator on the OAT screen, 'RefCav Trans (V)' and 'Beatnote at (Hz)'. The first one is nominal at 1.1 V (has a green box when above 0.8V, and a red box when <0.4V), and indicates if the RefCav is locked. The Beatnote shows the VCO frequency in Hz, nominal at 39.603 MHz.
I left the cavity unlocked, but the PLL engaged.
A good .snap time is Wed July 25 2012, 20:00h (and half hour from now).
Could the 'pulse-width-modulation' effect (lack of a better discribtion), and the drift of the beatnote away from the LO, be the reason that the laser is behaving strange?
The H2 dust monitor alarm went off. Not sure what button to press, so I pressed silince for one hour ...
At 7:08 the tumble weeds are being gathered.
J. Kissel, T. Vo Now that the optical lever signal chains are functional, it's straight forward to include Thomas' calibration into the signal chain and grab a spectra. Attached are the results for ETMY and ITMY. I'm not sure if I believe the amplitude of the results though. Even though Fabrice has been making excellent progress with H2 ISI ITMY, I don't think we're down to the sub- nrad / rtHz at 0.5 Hz. We'll go through the calibration more thoroughly tomorrow, and see where I've screwed up. Anyways, for posterity, my calibration is as follows (which, although Thomas only has official calibration for ETMY, I used the same calibration for both assuming the signal chains and lever arms are roughly the same): DOF ADC Gain Oplev Calib. (order of mag.) Pitch: 40/2^16 [V/ct] * 1666 [urad/V] * 1e-6 [rad/urad] = 1.0168e-6 [rad/ct] Yaw: 40/2^16 [v/ct] * 1727 [urad/V] * 1e-6 [rad/urad] = 1.0541e-6 [rad/ct]
Blend Filters were refined on ITMY. The current state is: 750 mHz in Stage 1 RX and RY 250 mHz in Stage 2 All dofs, and Stage 1 X, Z, RZ 100 mHz on Stage 1 Y The plot attached shows the latest isolation results (Red is Damped, Blue is Isolated). We are converging toward a configuration that we should be able to copy/paste on other units. Otherwise, I was having to turn the loops in the low blend configuration. The "boost" filters were making glitches in the actuators output, tripping the watch dogs. I modified the output compensation filters to filter out the glitches. The turn on/off process is now smooth and do not trip the watchdogs anymore.
J. Kissel, T. Vo, R. McCarthy After installing the new QPD.mdl library part and getting the channels reflected onto the the QUAD MEDM screens, we found nonsense coming out after the normalisation step of the signal processing. Here, "nonsense" is defined as Input to the normalization: Pin = -2400 Yin = -2900 SUM = -10700 Output of the normalization: Pout = Pin / SUM = 2.4e+9 = nonsense Yout = Yin / SUM = 2.9e+9 = nonsense Given that this library part is working quite well for the rest of the LIGO world that's using it, it implies that there's something screwy going on with our signal chain. We then tried turning off the signal input and inserting offsets with small (off order 1), known values to provide simple tests to see if the normalization was functional. With just offsets, the signals make sense. DAH! After a little more head scratching and a few more rounds of guess and check, I realized the difference between the input signals and the offsets I was trying: the raw input signals are negative (which Thomas informs me is expected from the electronics). *ACHOO!* Inverting the sign in the SEG (input filters), all works great. My guess is that the RCG Saturation Part, which has its limits set to [1e-6,1e12] treats the negative numbers as the lower limit, and divides the signal by 1e-6. (Hence 2400 / 2.4e9 = 1e-6). An analog signal chain red-herring that led to some interesting discoveries: We'd suspected initially that there was some high-frequency content saturating the ADC (that made it not visible in EPICS variables). A spectra of the raw segment channels, up to 7 kHz, later revealed no such features. However, while we still suspected this was true, we dove into the whitening chassis just to better understand the signal chain (since it's the only thing between the QPD and the AA chassis). We discovered that, though originally designed to have one, the ISC whitening board (D1001530), which is internal to the Oplev Whitening Chassis (D1100013), did NOT have its daughter board that fixes the state of the typically-digitally-controlled, switchable-state filter board. Mohana had some extra, and will ship some over night. Once those are in place, we can play around with the whitening and gain to find a good state in which we can leave it.
Attached are plots of dust counts > .5 microns in particles per cubic foot.
Today from ~ 3:00 pm to 5:20 pm local time I excited the TOP stage of the Y arm suspensions (first ITMY, then later ETMY) in order to gather peak height data to help with diagonalizing the OSEM sensor readbacks. Power spectra were recorded with damping at the top stage turned OFF, while an excitation of white noise was simultaneously injected into all 6 degrees of freedom of the top stage. The XML templates are saved in the suspension directories: /ligo/svncommon/SusSVN/sus/trunk/QUAD/H2/ETMY/SAGL2/2012-07-25_top_actuation_ETMY_4times_power.xml /ligo/svncommon/SusSVN/sus/trunk/QUAD/H2/ITMY/SAGL2/2012-07-25_top_actuation_test.xml I also tried to drive the L2 stage of ITMY with damping OFF at the top stage, but failed to get any coherence in this power spectrum measurement.
Dave and I finished installing the models, and MEDM screens, for HAM3.
I built the HAM3 testing folder arborescence, filled HAM3 matrices and started trying the scripts we will use for the in-chamber testing.
In-field cables arrived today. Richard and his crew started installing them.
HAM3 balancing is planned tomorrow. This balancing will be performed with the BNC outpus of the CPSs' ADE boxes.
Various activities - rebooting of SUS models for QUADs in the morning, another in the afternoon - Thomas Vo to End-Y station - noisy work in the LVEA from ~1:00PM - 3:00PM - leak detecting work in the Optics Lab
I have taken the calibration data and analyzed it. The plot of the linear response curve is attached.
I used a polynomial fitting program in python to get these numbers. This should give you the conversion from volts to micro-radians.
Slope Y-intercept
Pitch [ 1.01902201e+06 8.42886118e+00]
Yaw [ 9.83122943e+05 -1.19329010e+01]
Please hold off on using these calibration parameters, it looks like the SUS readout channels that map out the quadrants on ETMY and ITMY were not correct and I will need to re-calculate if this is true. Updates soon to come.
The QPD mapping did prove to be different than what was previously shown on the MEDM screens. Here are the new values for the linear response curve. The units are micro-radians/volts. The lever arm is taken to be 6 meters still Pitch [ 1727.67131855 23.41350328] Yaw [ 1666.80728788 -33.14694715] It is relevant to address that a positive pitch on the ETMY quad suspension MEDM on the m0 stage will yield a higher intensity on the two lower quadrants of the optical lever QPD, giving a negative pitch as viewed facing the QPD. Also, a positive yaw in the quad suspension MEDM will yield a higher intensity on the two left quadrants of the optical lever QPD, giving a negative yaw as viewed facing the QPD. For alignment purposes this will be important to note. The next step is to take measurements over the course of at least 2-6 hours and measure the drift as a function of time.
I copy and pasted wrongly into the screen the pitch and yaw are reversed, this is signal(1/m) vs translation(micro-radians) which gives units of micro-radians*meters.
Slope Y-intercept
Pitch [1666.80728788 -33.14694715]
Yaw 1727.67131855 23.41350328]
Apologies
July 24, 2012 Norna's email exchange of roll tolerance requirements:
Colleagues
I found the email exchange I had with Dave R about clocking, see below.
1.5 mrad corresponds to a height mismatch of ~ 0.2 mm (0.1 mm up on one side and 0.1 mm down on other, with 75 mm radius). This value was not a concern to him as you see. Betsy's 0.4 mm up on one side and down on other gives a clocking of ~ 5 mrad. So someone (Luke?) should check the Zemax to see if 5 mrad (for example) is OK and then we should get a target value captured in a design requirement doc.
However Matt's practical advice and experience of getting it to 0.25mm sounds good for now.
Norna
TFs from before the roll fix showed some small cross-coupling between T and L. We fixed the roll of the optic and the modes are now degenerate as expected. Plots and more details coming.