Performance plots later. Lvl2 will be quick but later too.
9:20 am Jeff B and Andres R to X-End lab area, retrive 3-IFO TMS components.
9:35 am Danny S to CS VEA, West bay area, quad work.
12:58 pm Karen to Y-end VEA, cleaning.
FIRE DRILL.
1:25 pm Danny S to CS VEA, West bay area, quad work.
1:28 pm Richard M and Patrick T from CS control room, Beckhoff work.
1:30 pm Jeff B and Andres R to Y-End.
1:45 pm Rick S and Peter K to CS VEA, pull cables from CER to HAM2 area.
3:14 pm Danny S to CS VEA, West bay area, quad work.
3:29 pm Travis S to CS VEA, west bay area, quad work.
4:06 pm Greg G to CS VEA, TCS work.
4:12 pm John W and Bubba G to Y-End VEA, walk/inspection.
PSL Check: 9/29/2014
Laser Status:
PMC:
FSS:
ISS:
Now with added "damped" plots. Note, the damping loops on the electronics test stand are hodge podge and so damping was poor for some regions of many loops. As well, like I mentioned in earlier logs, the coherence of this in-air QUAD is poor at lower frequencies. I spent some time trying to work out better excitation filtering/boosting but to no avail. Damping works on both M0 and R0 chains of Q6.
Attached below is a comparison of undamped and damped Phase 1b QUAD06 TFs, which are also compared to QUADs of similar construction. Summary: As already noted above, damping loops are in no way optimised on this test-stand, however, damping on all DOFs on both chains of QUAD06 can be observed. The most egregious damping behaviour occurs on the R DOF of the reaction chain. It should be noted that, since the undamped TF for this DOF appears clear, this indicates that issue is most likely filter configuration related when attempting to engage damping loops. Thus alleviating any concerns. All data, plots and scripts have been committed to the sus svn.
This morning I adjusted the x-arm alignment to obtain green locking. First I misaligned ETMX, and adjusted TMSX using the ITMX baffle PDs. See table below for configuration:
Old Average (alog 13741) |
Target PD1 | Target PD4 | new Average | |
TMSX (P,Y) | (-23.8, -320.1) | (-58.1, -292.7) | (9.6, -354.9) | (-24.25, -323.8) |
NOTE: he baffle PDs read 2.4V at 0dB gain.
Then, using the ETMX camera I centered the beam on ETMX by adjusting the ITMX alignment. I found ITMX (P,Y) = (74.7, -8.2). Finally, I maximized the flashes by aligning ETMX. H1:ALS-X-TR_A_LF_OUT reached about 0.85 cnts. With this alignment we were able to lock the green beam to the arm. The alignments are saved to the guardian.
The dither alignment in yaw helped bring the counts up to about 1. The pitch dither made things worse.
It seems that what we see on the ISS second loop photodiodes is real intensity noise in transmission of the IMC. Indeed, as visible in the attached plot, the signal on all ISS PDs has the same shape and is very well coherent with other monitors of the IMC transmitted and reflected powers.
So, as suggested in my previous entry, the RIN at the IMC output is very large, 1e-5 at 10 Hz and 1e-6 at few hundreds Hz.
To investigate the origin of this broadband excess of noise, I had a look at some angular channels related to the IMC.
As visible in the third attached plot, the IMC transmission shows some coherence with the IMC WFS signals and with the periscope accelerometer. In particular:
There is some broadband coherence also, so maybe most of the noise floor is due to angular motion of the beam, converted to intensity noise by the IMC.
Jenne, Kiwamu,
We aligned the SHG and ALS comm paths on ISCT1 this morning. After the alignment, we obtained the following power leves:
(SHG path alignment)
The beam going through the crystal had been tiled downward -- it was too high on the incident side and too low on the output side of the crystal. We tweaked a 1" steering mirror in front of the bottom periscope mirror to correct it. Also the beam was off in the horizontal direction on the first 2" lens and therfore we shifted the position of this lens to have the beam centered on the lens. We noticed that the HWP after the 2" lens was a bit too low. However, since it was not terrible too low, we left it as it was.
Then we steered a 1" mirror before the crystal to level the beam tilt. It looked like a lens and PBS after the crystal were too high by a bit. Since the first lens strongly deflects the beam, it was difficult to center the beam on both optics. We compromised the beam pointing such that it is not clipped on either of the optics.
We then steered the crystal mount in order to match the crystal to the beam pointing. A screw knob at center bottom was already all the way in and it seemed we needed to go further more, but we could get a green power of 1.2 mW in this configuration and we decided not to touch it any more. We optimized the other three knobs to maximize the green power.
(ALS comm path alignment)
We touched the following optics:
- ALS-PBS1
- ALS-M12
- ALS-BS3
- ALS-BS4
- ALS-BS5
- ALS-M6
in order to optimize the beatnote setup. Since the alignment of the X arm was not stable, we did not really optimize the beat note power. But it the maximum we saw was about 800 mV p-p observed with a scope terminated with a 50 Ohm.
(Green power monitor's feedthrough was bad)
We did not see signals in the green power monitor. We eventually found that this was due to a bad feedthrough (or perhaps the cable was not plugged all the way in). We changed the feedthrough from the top one to the bottom one and it solved the issue. Now the signal is acquired to an ADC. Good.
Install Work
Richard: cabling for gig-E cameras
cont to build quad (finding parts)
Acceptance Reviews: starting installation acceptance process (following LLO's lead). This is a sizeable documentation endeavor.
Commissioning
Jenne up visiting for the week.
The regular ~1nm variations in ETM coating thickness cast a diffraction ring onto baffles that may produce excess noise through modulation of retro-reflected light (https://dcc.ligo.org/T1300354). A test of whether this scattering noise may limit our sensitivity could be made by shaking a beam tube baffle in the region where the maximum light power falls on the baffles, about 2375 m from the vertex. Because only one or a couple out of many baffles would be affected by shaking and because the LLO sensitivity is still about 3 orders of magnitude away from the goal at the most troubling beam tube resonance (~14 Hz), the increase in motion over normal would have to be at least 4 orders of magnitude for the present LLO sensitivity.
In order to see if I could increase motion of the beam tube by this much I made a pusher system and tried it out here at LHO. Figure 1 shows the voice coil shaker and the coupling rod that has interchangeable springs to optimize force. A universal joint is required to minimize non-axial forces on the voice coil plunger and the coupling rod. The system was powered by a 150 W inverter in an Uplander LT van. Figure 2 shows the relative positions of, at the left, the accelerometers, in the middle, the shaker, and, on the right, a beam tube enclosure door.
Figure 3 shows that the shaker increased the axial beam tube motion at the accelerometer by 4 orders of magnitude at 14 Hz. The noise floor for the shaking injection is higher than for non-shaking because I had to reduce the gain by 10. The injection is not as monochromatic as I would like. This version 2 is better than the first version, but there are still peaks injected at higher frequencies from slight rattling of bolts, springs and other components in the rod. Nevertheless, it looks like it could be used for a test as long as we are focusing on direct (vs. upconversion) coupling.
model restarts logged for Sun 28/Sep/2014
2014_09_28 06:27 h1fw1
2014_09_28 17:53 h1fw1
unexpected restarts of h1fw1
Last night I used the good alignment of the OMC from the dither loops to sweep the cavity with a single-bounce beam from ITMX and ITMY. The result is that I think the modulation depth mystery is gone (Gamma2 = 0.077, for the 45MHz sidebands; in hindsight my earlier calculation was wrong) and we see a difference in the mode-matching to the OMC between ITMX and ITMY that matches what we expect from the as-built optical parameters. So, maybe life makes sense again...
===== Modulation Depth =====
The short story is that I mis-identified the peaks from the 45MHz sidebands in the measurement from Sep 19th. Probably I was looking at the 1st and 4th order carrier modes; with the good alignment the 1st order mode is much smaller, and there's no way to mistake that the 4th order peak doesn't have a matching peak at the other end of the FSR. The two plots attached show the cavity sweeps of ITMX and ITMY; the peaks used to calculate the modulation depth are marked with black crosses. There are five sweeps for each optic. The peak heights are averaged and used to calculate J1(Gamma), and the value of Gamma that returns J1(Gamma) is calculated using the scipy.special.jn() function. The result is:
Gamma1 (9MHz) = 0.211
Gamma2 (45MHz) = 0.077
These results are consistent for the upper and lower sidebands of each modulation frequency, and are the same for ITMX and ITMY. The variation from one sweep to the next is small (usually less than 0.001). Also, this measurement of the modulation depth for the 45MHz sidebands agrees with a measurement from last December. No modulation mystery after all?
(I hope that Koji will check these results with his more-sophisticated peak-calibration code. If you are comparing my plots to his, note that my data use up-going sweeps of the PZT, while his are down-going (for example here), and so the order of the peaks is flipped between the two. The latest (Sep 28th) sweep data are here.)
===== Mode Matching =====
Using the mode scans we can make an estimate of the mode-matching of the single bounce beam into OMC. Following Koji's calculation I use the peak height of the 2nd-order carrier mode to make a rough estimate of the mode overlap, using the expression MM = 1 - (CR_TEM20/CR_TEM00). This calculation is rough, in that it ignores the power in the higher order modes, and just uses the peak height; I will try to include the HOMs and integrate the peaks later, but the answer should be the same to within a percent or two. For each ITM we get:
ITMX = 0.884
ITMY = 0.913
The magnitude of the mode mismatch, and the difference between the ITMs, is in agreement with predictions from the beam-propagation. Good alignment matters! I suppose the next step will be to check that we can improve things with TCS. (Again, someone should make an independent calculation using this sweep data...)
===== Other Stuff =====
This afternoon I wrote a short slider offload script to use with the alignment guardians that Sheila & Gabriele commissioned yesterday.
The script is ~userapps/release/asc/common/scripts/offload_M1_LOCK_ALIGN.py. It moves the outputs of the M1 DRIVEALIGN matrix to the alignment sliders, taking care to check the calibration gain that comes after the sliders on some optics. It should work for most any optic that has an M1 stage; I tried it on the PRM and SRM and it did the trick. The syntax is, for example:
./offload_M1_LOCK_ALIGN.py -o PRM
Not sure how to best get this into the ISC_DOF guardian. Don't forget to save those aligned slider values!
After touching up the alignment of PRX, SRY, and MICH I was able to get the DRMI to lock at 1W a few times, for a few minutes. The gain settings were:
MICH: 30
PRCL: 22
SRCL: -400
...and FM8 (ELP70) engaged on the MICH bank rather than FM9 (ELP40), although I didn't have a chance to measure the MICH loop to see if it made a difference. A loop measurement of PRCL is attached; the red curve is today, the black curve is a reference.
no restarts reported
Gabriele, Dan, Sheila
We did several things today and last night...
First we worked on focusing the cameras on the input arm, ETMX, and both ITMs. Now we have a much clearer view of our interferometer. I also copied the code to align TMSX using the GigE camera from the IAS guardian into the ALS guardian, and tried it out with the cavity locked on green. This kind of works, although the camera centroid is noisy, the noise roughly coresponds to 3 urad of TMS angle. We probably want to get a mask going on the camera to see if that helps.
We are using refl WFS for SRY and PRX. We also attempted this for MICH on the dark fringe, but the signal was too noisy to work well. For all the asc loops, we used FM1+2. For PRX we used REFL B 9 with a gain of -0.0003 for Pitch and 0.003 for Yaw. For SRY we used REFL A 9 with a gain of 0.03 for both pitch and yaw. All of these states are now added to the guardian currently called ISC_DOF ( we will think of a better name). This currently has the states that were in the DRMI guardian that are actually used for initial alingment, plus the arm IR locking states. The graph for this guardian is attached.
After running though the parts of the inital laignment sequence that we have working so far, we locked PRMI at 10 Watts, and steped the power down using the rotation stage, manually correcting the power in DC PWR IN using the offset, and increaseing our gains. We locked 2.8 Watts with a PRCL gain of 50.4 and a MICH gain of 10, and at 1 Watt with MICH at 110 and PRCL at 22 (we measured the loops and adjusted the gains to keep our MICH ugf around 12 Hz and PRCL around 80Hz, suprisingly the PRCL gain had to be increased by less than a factor of 10 to keep the same gain with a factor of 10 decrease in the power.)
We then took some guesses at the gains to use for DRMI at 1 Watt based on Anamaria's document, what we had originally used, and what we needed for PRMI with 1 Watt. After some adjustment, we locked with the parameters shown in the screen shot. DRMI was locked on the sideband at 1 Watt from about 1:57 UTC September 28th to 2:21 when we tried to change a filter and it dropped lock.
The DRMI acquired lock a few more times over the next couple of hours. During one of these I collected loop measurements of the MICH and SRCL loops. The gain settings were:
MICH: 30.0
PRCL: 22.0
SRCL: -200.0
The MICH UGF was 10Hz with 30deg of phase margin (red trace in DRMI_MICH_loop.pdf). Gabriele saw that the ELP40 filter in the LSC MICH bank was eating a lot of phase around 10deg so he installed an ELP70, but I wasn't able to catch a lock with this filter engaged.
The SRCL UGF is 20Hz and it looks like the gain could be increased by a factor of two.
We had measured PRCL during an earlier lock and the UGF was the same as the reference (70Hz wth 40deg).
The SRM and the BS get whacked really hard during the acquisition for DRMI, and I think one of the limitations to robust locking is the L2P noise that is injected when the DRMI tries to grab a fringe. The picture attached shows a trend around the time when the MICH, PRCL, and SRCL loops were triggered. The BS oplev in PIT has a big excursion after the longitudinal actuation is triggered by the fringe crossing (the peak of >20 counts in POPAIR_B).
Yesterday afternoon we reconnected the ADC to the whitened signals for PD5-8. Now the status is the following:
I wanted to compare the RIN meaured by the first loop photodiodes PDA and PDB with was is measured by the second loop photodiodes. However, looking into the way PDA_REL and PDB_REL are reconstructed, I found a couple of strange things, and I'm quite convinced that the estimated RIN was wrong. Peter King computed the whitening transfer function of the PDA and PDB box. Comparing this with what was implemented in the PD?_CALI_AC, I found a quite large discrepancy I could not understand. Therefore I implemented an inverse of the expected whitening, which is good between 1 Hz and few kHz, as shown in the first plot. I also found that the low pass filter used in the PD?_CALI_DC was giving me a strange notch at about 10 Hz in the PD?_REL signals, so I changed it into a simple low pass, as visible in the second plot. For both the AC and DC parts, the old filters are still there.
The second loop photodiode PD5-8 outputs are dewhitened as explained in a previous entry. The output is calibrated in V, and the DC level that we measured before changing the ADc connection is about 3 V.
The third plot shows the RIN as visible in both the first and second loop photodiodes when the first ISS loop is open. In particular, we see that the second loop photodiodes are in good agreement with the first loop photodiodes, at least above 100 Hz. This is a cross check that the calibration in terms of RIN is, if not correct, at least self consistent.
The fourth plot shows what happens when the first loop is closed. As expected the out of loop PDB shows a flat RIN that I guess should correspond to the shot noise level, even though I don't know right now how much pwoer is getting into the diode, so I can't teel if the absolute level is correct. Looking at the second loop photodiodes, it is clear that we have a huge excess of noise with respect to the level obtained at the output of the first loop. At this point we can't tell if this is real intensity noise, or just jitter noise that couples to the second loop PDs. Recall that we didn't optimize at all the coupling of jitter to RIN for the second loop photodiodes. For sure, there's still a lot of work to do...
model restarts logged for Fri 26/Sep/2014
2014_09_26 13:05 h1dc0
2014_09_26 13:07 h1broadcast0
2014_09_26 13:07 h1fw0
2014_09_26 13:07 h1fw1
2014_09_26 13:07 h1nds0
2014_09_26 13:07 h1nds1
2014_09_26 23:12 h1fw1
unexpected restart of h1fw1. DAQ restart to include TCS HWS channels and new DMT chan list.
No working green WFS with centering yet.
1. WFSB segment 3 demod signal is a factor of 10 weaker than the others when the beam is centered.
Giving the WFS centering servo a YAW offset (+0.4 in YAW so the beam stays at -0.4) would recover the balance. But at that point we're already almost railing MCL PZT.
2. Difference between good alignment (H1:ALS-C_TRY_A_LF_OUT_DQ=870-900) and not-terrible (smaller than 800) is big on WFS centering.
We made the centering servo work when the alignment is not that bad, but later when I manually aligned the arm, the servo started railing.
So, one of the problems seems to be that we don't have much range for MCL PZT mirror.
The screenshot shows the demod phase setting and RF WFS error signal, and you can see that the WFSB seg3 signal is very small. The reason why WFSB spectra look washed out is because the WFS centering kept railing during the measurement (sorry).
Next step would be to go to the end station, manually releave the PZTs, and repeat the measurements. I'll leave the centering servo disabled for now.
Attached see the TFs--Comparing these to the most recent in the matlab SVN (August 2013,) these look just fine. There is a new little extra wiggle on some dofs at 0.7 & 1hz that I'd say are the table top components.