Reset the following CDS errors: H1SUSETMY TIM, ADC H1SUSETMY TIM H1SUSTMSY IPC H1IOPSEIEY IPC H1ISIETMY IPC H1ALSEY IPC H1ISCEY IPC H1IOPSUSEX ADC H1IOPSEIEX IPC H1ALSEX IPC H1ISCEX IPC
Evan, Stefan, Sheila, Matt, Lisa The official goal of tonight was to make measurements for the ultimate noise budget plot that Evan is about to produce (work still in progress). The unofficial goal was to improve the sensitivity between 50 - 100 Hz, and investigate the noise that is obsessing Stefan these days . More details will follow tomorrow, but after Tuesday maintenance, three earthquakes and some in principle innocuous but effectively disruptive measurements, we finally managed to relock robustly and make a couple of (alleged) improvements (1/1.5 in Valera's scale of awesomeness): - despite some recent MICH feed-forward tuning, MICH still had some residual coherence (< 0.3) with DARM between 50 and 100 Hz. We saw room for improving the current cut-off, so Matt added another filter to the LSC-MICH filter bank (FM9, EBS50) which only loses a few degrees of phase at 10 Hz, but provides about 10 dB of attenuation between 50 and 100 Hz. The coherence is now reduced < 0.1 above 50 Hz; - starting by staring at some recent Bruco reports which showed some suspicious (even if small) coherence with the ASC AS DC signals below 100 Hz, we ended up (re)discovering that the current WFS centering loops ASC_AS_A / AS_AS_B --> OM1 / OM2 have a simple cut-off at 100 Hz (CLP100 in H1ASC-DC4/DC3 P/Y). The existing LP8 filter was not stable, so now there is a new cut-off at 20 Hz (ELF20). Both these filters have been tested a couple of times during the entire locking sequence, so we leave them on. We leave the interferometer locked undisturbed ~ 2:15 AM (July 29 9:15 UTC). P.S.: For the morning crew, as Sheila reported , the locking sequence doesn't currently work automatically to low noise because the ETMX ESDs are not turning off.
Another task is to automate some CARM gain redistribution in the guardian.
Right now in full lock the CMB common gain is −22 dB, the fast gain is +7 dB, and the LSC-MCL gain is 240 ct/ct. We can reduce the input-referred noise of the SNB+CMB by a factor of 2 by changing these values to −15 dB, 0 dB, and 107 ct/ct.
[In fact, from a noise perspective it would probably be even better to do a little more redistribution by removing some gain from the IMC ao path (currently it's −2 dB) and adding it to the CMB common gain.]
Since the IFO was still locked when I came in, at around 8:06AM local time I started manually reallocating the CM and MC board gain as Evan suggested:
[CM common, CM fast, LSC MCL, MC AO] =
originally [-22, 7, 240, -2]
then to [-15, 0, 107, -2]
then to [-12, 0, 107, -5]
Finished at around 8:11 AM.
[update] After running in this state for ~30min, the IFO lost lock during the morning meeting. I manually reverted the CM board and MC board gains back as I wasn't sure which part of these is NOT touched by the guardian when locking.
The OM dither lines (around 1.7kHz) are not being used, so I turned them down by a factor of 10.
H1:OMC-ASC_P1_CLOCK_GAIN = 0.1
I'll leave them on to make sure that we don't run into DAC bit-noise problems (e.g., no HF signal gathers the DAC noise around DC, which is not fun).
Occasionally, we ring up the BS butterfly mode (2.5kHz), the BS violins (300 Hz) and something at 41Hz. What is that thing at 41Hz (which is always present, though usually not so big), and how are we exciting the other BS modes?
Is the line at 75Hz from the ring heater fans, as in LLO?
The highest Roll mode of the HSTS is modeled to be at 40.3 [Hz] -- I suspect SR2 or SRM as you drive it for lock acquisition, or they could have been rung up during the low-frequency sawtooth waves from the earth we saw last night.
At 75 Hz there is coherence with magnetometers and microphones at the Y end. So I guess it's the ring heater.
We noticed that we can't really turn off the ETMX ESD. We aren't sure why.
This causes the guardian to hang, in the state LOWNOISE_ESD_ETMY because it checks that the ESD is off. This can be worked around using manual, but the operator has to be carefull not to skip steps if we force the gaurdian to move on.
Not sure why this is. I was able to cycle the power to the High Voltage ESD Amp multiple times from the Binary MEDM screen. I was also able to turn each channel On and Off individually from this screen as verified at the output of the unit not just the read-back.
We need to update the channel that is used to turn the ESD on and off. I verified again this morning that the Binary switch does in fact work.
I second what Richard says -- the Beckhoff binary signal that *was* used to control the HV driver on ETMX has now been replaced by similar "fast front-end" binary I/O that's at ETMY. So, for each quadrant, to turn off the high voltage, set the EPICs records
H1:SUS-ETMX_BIO_L3_${QUADRANT}_VOLTAGE_SW to 0 and
H1:SUS-ETMX_BIO_L3_${QUADRANT}_HVDISCONNECT_SW to 0.
This has been true since Richard cabled up the LVLN driver yesterday afternoon (see LHO aLOG 19994).
Sheila, Stefan We lost the interferometer a couple of times in TR_CARM, but where able to track it down to a test point data access error (through hcdu.avg). (We used this to zero the offsets for the arms.) For some unexplained reason the test point for LSC-TR_X_QPD_B_SUM_IN1 was returning zero instead of the actual arm power numbers. We verified this with data viewer - but just after we convinced ourself that this problem was real, the data started to flow again. Sure enough the script worked again...
Hang, Matt
(This is Matt, writing for Hang since I accidentally lost his log entry.)
Over the last couple of days Hang generaized his A2L decoupling scripts (see entry 19767) to make an all-purpose python function for general diagonalizing. The code is:
/opt/rtcds/userapps/release/isc/common/scripts/decoup/deMod_deCoup.py
with the example application code in deMod_deCoup_demo.py which uses the new functionallity to decouple the ITMX length drive from YAW on the optical lever.
The next task will be update the A2L code to use this function, but tha twill wait for another day.
J. Kissel, for the Relocking / Commissioning Teams A progress update (picking up from when the future looked bleak), so I don't have to write a giant log at the end. (And to be honest, I'm not staying.) Executive summary thus far: We're still struggling with the restoration settings after a major front-end computer outtage (like an RCG upgrade). Not because we aren't restoring to a good time, but it's that the "good time" (which has, thus far, been "the last time we were locked in low noise") is not the right time for some settings, especially those that are part of lock-acquisition. These settings will slowly but surely rear their head, and we just have to code them into the Guardian as we go, because they are almost always NOT monitored by SDF because they're part of filter banks under guardian control. Also, big earthquakes are a huge hindrance to recovery from a rough maintenance day. We should should schedule those for Friday nights. Here's what happened in the after noon / early evening: 13:00 All models have finished there upgrade and are up and running Quickly see that HAM SUS IOP output is OK Quickly see that IPC errors from SEI are gone Richard immediately to ETMX to finish cabling up ETMY ESD LVLN Driver 13:10 Discover IMC WFS front end model has been mistakenly named ASCIMC on the top-level making all MEDM channels go white (because they're now called H1:ASCIMC-... instead of H1:IMC-...) 13:20 All SEI and SUS "recovered" (i.e. brought to ALIGNED and FULLY ISOLATED) 13:30 PR3 oplev commissioning begins h1ascimc front end model naming bug fixed, back up and running Discover ITMs are getting a HUGE signal blasted in from ASC DAQ restart 13:40 Find ASC loops that are blasting the ITMS, loops turned OFF and history cleared. << LESSON LEARNED -- We should Run the ISC_LOCK DOWN script after a full computer restart 14:00 Discovered ETMX ESD wiring had not been updated for the new LVLN ESD wiring (LHO aLOG 20003) Resume wiring up ETMX ESD LVLN Driver 14:20 IMC recovered briefly after trouble with *some* unknown setting has been globally BURT restored 14:22 IMC lost again, because ISS is railing. 14:25 Jim and Dave head to EY to update BIOS settings on fast front-end (For the record, we've chosen NOT to revert the EY fast front-end to slow front-end, we've JUST upgraded the BIOS settings) 14:40 IMC recovered << LESSON LEARNED earlier global burt restore restored to a time of full IFO lock, which had the ISS 2nd loop engaged. Without full IFO, it doesn't make sense to have the 2nd loop ON. 2nd loop turned OFF, IMC OK. EX LVLN Driver Wiring Complete EX Charge measurements launched 15:00 Jim and Dave finish ETMY front-end machine BIO upgrades Found New ISI wiener filters are ON (as is standard for any new filter bank) and just feeding STS signals straight to the platform, causing platforms to be very noisy. 15:15 Optical Lever work completes. 15:20 ETMY SUS and SEI fully recovered Begin charge measurements at EY to confirm SUS health Keita begins IMC WFS plant measurement for new DOF5 to reduce 200-300 Hz intensity noise 16:00 Charge on ETMY done -- but discover drive is saturating. (Found out later it was because of the move of the "sumComp" filter from DRIVEALIGN to COILOUTF that was performed yesterday evening [[no aLOG]]) Begin Initial alignment 17:05 Found SNR for AS 45 Q (during initial alignment of SRY) was really low, even with high-power into the IFO. << LESSON LEARNED Discovered there's a -160 dB filter that's ON for full IFO low-noise state, which we *don't* want on during initial alignment. This is *not* included in any Guardian's state request. This *should* be included in both the IFO_ALIGN guardian and the IFO DOWN state. 17:40 Initial alignment complete, beginning full IFO lock acquisition attempt (Need some hand tweaking of the BS by Evan) 17:45 We had *just* reached some stages of the CARM offset reduction for the first time and then Magnitude 5.9 Earthquake - 29km S of Acandi, Colombia 17:50 Charge measurements resume (we figure out the saturation issue mentioned above) Matt and Hang do some L2P / L2Y measurements on ITMY 18:40 Resume locking 19:10 Make it up to DARM_WFS (not even to RESONANCE where the FULL IFO is at least RF locked on ETMX) Found ETMY M0 Bounce mode damping filters were not set correctly, ringing up EY's Bounce Mode terribly -- blamed filters disappearing (debunked) -- blamed improperly engaged guardian state (debunked) It was really, that the last global BURT restore was done *before* the EY BIOS upgrade was finished. So we didn't BURT enough! << LESSON LEARNED These filters should *also* be forced into the right configuration, since they are NOT monitored by the SDF system (because there are other parts of the filter module that ARE controled by guardian) 19:45 While having resolved the Bounce Mode damping problem, Magnitude 6.3 Earthquake - 71km SSW of Redoubt Volcano, Alaska Goodnight everybody. "Things will *definitely* be better in ULTRA LIGO."
Daniel gave me the test rig for the AM stabilized EOM drivers that we should be receiving from Caltech this week. This allowed me to test that the Beckhoff controls and readbacks work as expected. I also made a summary screen (ISC_CUST_EOMDRIVER.adl) of these readbacks and controls, which is accessible from the LSC dropdown menu on the sitemap.
The chassis is labeled "Corner 6", and has 2 unconnected connectors labeled "EOM Driver A" and EOM Driver B".
The "A" connector controls the 45 MHz channels, and the "B" connector controls the 9 MHz channels.
The controls and readbacks performed the same for both channels, so I'll only write out the list once.
| Voltage input with voltage calibrator | RF Set Mon readback value |
| 0.1 V | -13 |
| 1 V | 7 |
| 3 V | 16.6 |
| 5 V | 21.0 |
| 7 V | 24.0 |
I need to investigate the situation with the "Excitation Enable" switch, but other than that we should be ready to go when the EOM driver arrives, if we decide to install it.
The RF setpoint goes from 4dBm (lowest setting) to 27dBm (highest setting) in steps of 0.2dB. The binary should start at zero for the lowest setting and increase by 1 for each step. This is a PLC problem.
[Jenne, StefanB, Cheryl]
We were having some trouble getting the IMC WFS to converge (WFS came on, looked okay for a while, then started dragging the MC transmitted power down), so we implemented and tested the new global burt restore state, accessible from the ISC_LOCK guardian.
In the end, the specific problem with the IMC WFS was that the ISC input filters on the M3 stage of the MC mirrors (definitely MC1, maybe others) had gain of zero, so the WFS signals weren't getting through to the optics' outputs.
The solution we created, which should solve all kinds of problems, is a guardian state that does a global burt restore to a recent set of autoburt snapshots. This state has a date and time hard-coded in the guardian script, although it is easily change-able if we find a more preferable time. Currently, it is restoring to the autoburts of 28 July 2015, 07:10. To select this state, you need to go to "manual", since there are no edges to get there from any other state. When it has finished the restores, it will immediately go to and run the Down state.
The list of snapshots that are restored is:
Perhaps though, we should just restore *every* snapshot that is captured by autoburt?
J. Kissel, J. Driggers Use this feature with caution, as we found yesterday evening during the rest of recovery (LHO aLOG 20011), BURT restoring to a single time with the IFO was at low-noise isn't necessarily the right time to BURT restore to, especially until we've caught all of the settings that have fallen between the guardian and SDF cracks.
I really don't like this as a solution to whatever problem you're trying to solve. Is there something wrong with the SDF system that is causing you to resort to such a sledgehammer approach?
Added 429 channels. Removed 240 channels.
For Maintenance this morning I needed to break the IFO lock, so recorded the steps used today.
- Besty, Jeff, TJ, Cheryl
The reason that requesting ISC_LOCK to down doesn't work anyore is because I made down not a goto state. This means that there is no path from NOMINAL LOW NOISE to DOWN. (We were having problems with unintentially going to DOWN.) So when you request DOWN, the guardian doesn't actually go there. To force it to do that, you can go to manual mode and select DOWN.
I understand more now, and was not aware that you had changed that on purpose.
No suggestion for a change, just reporting what I was asked to try and the results I got.
Serial cables and new software infrastructure was installed to read the status information of the GPS clocks in the end stations. The CNS II clocks are using the Motorola binary format at 9600 baud. We read the @@Ha message once a second. The GPS receiver is reported as a Synergy SSR-M8T. This seems to be an u-blox LEA-M8T in an M12+ form factor and Motorola protocol emulation. Medm screens are available from the auto-generated list.
The duplicated PSL_ERROR channels were also eliminated.
We observed broadband coherence of OMC_DC_SUM with ASC_AS_C_LF_SUM and ASC_A_RF36_PIT. We made some numbers and plots, using the 64kHz version of the channels.
First the measurements we made on OCXO oscillator:
- ASC_AS_C sees a RIN of about 5e-7/rtHz above 100Hz (either from H1:ASC-AS_C_SUM_OUT_DQ or from H1:IOP-ASC0_MADC6_TP_CH11). The same is true for its segment 1.
- The calculated shot noise RIN at 20mA (quantum efficiency 0.87) detected is 4.0e-9/rtHz.
- The 4.0e-9/rtHz agrees with DCPD_NULL_OUT_DQ's prediction (8.0e-8 mA/rtHz/20mA).
- DCPD_SUM_OUT_DQ sees a slightly elevated RIN of 4.6e-9/rtHz (9.2e-8 mA/rtHz/20mA).
- The RIN in DCPDA (H1:IOP-LSC0_MADC0_TP_CH12, corrected for the whitening) is about 5.9e-8 mA/rtHz, or RIN = 5.9e-9/rtHz at 20mA/2diodes (~15pm DARM offset)...
- ...or about 3.3e-8 mA/rtHz or 1.2e-8/rtHz at 5.7mA/2diodes (~8pm DARM offset).
- ASC-AS_C_SEG1 (H1:IOP-ASC0_MADC6_TP_CH11) and OMC-DCPD_A (H1:IOP-LSC0_MADC0_TP_CH12) shows a coherence of 0.053 at 20mA, suggesting a white noise floor a factor of 0.23 below shot noise.
- At 5.7mA the same coherence is about 0.13, i.e. the white noise floor is a factor of 0.39 below shot noise.
- These two measurements are in plot 1.
- Taking the last two statements together, we predict a coherent noise of
- 5.9e-8 mA/rtHz *0.23 = 1.4e-8 mA/rtHz at 20mA/2diodes (~15pm DARM offset) (RIN of coherent noise = 1.4e-9/rtHz) - The pure shot noise part is thus 5.7e-8 mA/rtHz
- 3.3e-8 mA/rtHz *0.39 = 1.3e-8 mA/rtHz at 5.7mA/2diodes (~8pm DARM offset) (RIN of coherent noise = 4.5e-9/rtHz) - The pure shot noise part is thus 3.0e-8 mA/rtHz.
- AS_C calibration:
- 200V/W (see alog 15431)
- quantum efficiency 0.8 (see alog 15431)
- 0.25% of the HAM 6 light (see alog 15431)
- We have 39200cts in the AS_C_SUM. Thus we have
- 39200cts / (1638.4cts/V) * 10^(-36/40) (whitening) / (200V/W) = 1.89mW and AS_C. (shot noi
- 1.89mW/0.025 = 76mW entering HAM6. I.e. we have slightly more sideband power than carrier power (Carrier: 27mW in OMC transmission).
- Shot noise level on AS_C_SUM is at 2.0e-8 mA/rtHz, corresponding to a RIN of 1.6e-8/rtHz. I.e. the coherent noise seen at 5e-7/rtHz is high above the shot noise. Dark noise TBD.
- The light entering HAM 6 has a white noise of 5e-7/rtHz*76mW = 3.8e-5 mW/rtHz
Bottom line:
-We have ~1.4e-8mA/rtHz, or 1.9e-8mW/rtHz of coherent white noise on each DCPD.
-It corresponds to 3.8e-5mW/rtHz before the OMC, i.e. the the OMC seems to attenuate this component by 2000.
-This noise stays at the same level (in mW/rtHz) for different DCPD offsets.
Next, we switched back to the IFR for testing. plot 2 shows the same coherences (all at 5.7mA / 8pm DARM offset), but on the IFR. Interestingly now AS_C and AS_A_RF36 start seeing different noise below 2kHz. We convinced our selfs that the higher excess noise seen in AS_A_RF36 is indeed oscillator phase noise from the IFR - so that is clearly out of the picture once of the OCXO. (Evan will shortly log the oscillator phase noise predictions.)
64k Channel list:
H1:IOP-LSC0_MADC0_TP_CH12: OMC-DCPD_A (used in plot)
H1:IOP-LSC0_MADC0_TP_CH13: OMC-DCPD_B
H1:IOP-LSC0_MADC1_TP_CH20: REFLAIR_A_RF9_Q
H1:IOP-LSC0_MADC1_TP_CH21: REFLAIR_A_RF9_I
H1:IOP-LSC0_MADC1_TP_CH22: REFLAIR_A_RF45_Q
H1:IOP-LSC0_MADC1_TP_CH23: REFLAIR_A_RF45_I
H1:IOP-LSC0_MADC1_TP_CH28: REFL_A_RF9_Q
H1:IOP-LSC0_MADC1_TP_CH29: REFL_A_RF9_I
H1:IOP-LSC0_MADC1_TP_CH30: REFL_A_RF45_Q
H1:IOP-LSC0_MADC1_TP_CH31: REFL_A_RF45_I
H1:IOP-ASC0_MADC4_TP_CH8: ASC-AS_A_RF36_I1
H1:IOP-ASC0_MADC4_TP_CH9: ASC-AS_A_RF36_Q1
H1:IOP-ASC0_MADC4_TP_CH10: ASC-AS_A_RF36_I2
H1:IOP-ASC0_MADC4_TP_CH11: ASC-AS_A_RF36_Q2
H1:IOP-ASC0_MADC4_TP_CH12: ASC-AS_A_RF36_I3
H1:IOP-ASC0_MADC4_TP_CH13: ASC-AS_A_RF36_Q3 (used in plot)
H1:IOP-ASC0_MADC4_TP_CH14: ASC-AS_A_RF36_I4
H1:IOP-ASC0_MADC4_TP_CH15: ASC-AS_A_RF36_Q4
H1:IOP-ASC0_MADC6_TP_CH11: ASC-AS_C_SEG1 (used in plot)
H1:IOP-ASC0_MADC6_TP_CH10: ASC-AS_C_SEG2
H1:IOP-ASC0_MADC6_TP_CH9: ASC-AS_C_SEG3
H1:IOP-ASC0_MADC6_TP_CH8: ASC-AS_C_SEG4
Some more estimation - this time for frequency noise: - Shot noise on the refl diodes is given by Pshot=sqrt(2*h*nu*Pr_lock) - The cavity sensing function is P_9_pk = 4*Gam9*P0 * dNu(f)/(f_p + i*f), where P0 would be the carrier power incident on the PD without the IFO. - from this we can estimate a frequency (phase) noise of about 8e-11 rad/rtHz. Gam9=0.219; %alog15874 PSL_low=2; %W Pr_nolock_low=13.7e-3; %W PSL_lock=24; Pr_lock=3.5e-3; %W IMCt=0.88; att=Pr_nolock_low/(PSL_low*IMCt); P0=PSL_lock*IMCt*att; inlockdrop=Pr_lock/(P0); Pshot=sqrt(2*h*nu*Pr_lock); dphi=Pshot/P0/4/pi/Gam9;
For reference, I ran the numbers on where we would expect the sidebands to show a resonance feature. I used the following values: RITM=1939.3m RETM=2241.54m L=3994.485m Checking accidental sideband resonances in the arm cavities: Resonance condition: fres = FSR * (q + (l+m+1)*fTM/FSR) Free Spectral Range (FSR) : 37.5258 kHz Transverse Mode Spacing (fTM): 32.4297 kHz Checking f1 sideband: q=242 l+m=0 Freq. diff. = 18.2284 kHz q=242 l+m=0 Freq. from antiresonant = 0.534516 kHz q=242 l+m=1 Freq. diff. = 14.2013 kHz q=241 l+m=1 Freq. from antiresonant = 4.56162 kHz q=241 l+m=2 Freq. diff. = 9.10514 kHz q=-242 l+m=0 Freq. diff. = 18.2284 kHz q=-243 l+m=0 Freq. from antiresonant = 0.534516 kHz q=-243 l+m=1 Freq. diff. = 13.1322 kHz q=-244 l+m=1 Freq. from antiresonant = 5.63065 kHz q=-244 l+m=2 Freq. diff. = 8.0361 kHz Checking f2 sideband: q=1212 l+m=0 Freq. diff. = 16.0903 kHz q=1212 l+m=0 Freq. from antiresonant = 2.67258 kHz q=1212 l+m=1 Freq. diff. = 16.3393 kHz q=1211 l+m=1 Freq. from antiresonant = 2.42356 kHz q=1211 l+m=2 Freq. diff. = 11.2432 kHz q=-1212 l+m=0 Freq. diff. = 16.0903 kHz q=-1213 l+m=0 Freq. from antiresonant = 2.67258 kHz q=-1213 l+m=1 Freq. diff. = 10.9942 kHz q=-1214 l+m=1 Freq. from antiresonant = 7.76872 kHz q=-1214 l+m=2 Freq. diff. = 5.89804 kHz
Evan, Matt, Lisa We did one more test for the broadband coherence noise: Common mode gain +3dB vs -3dB We see no chnge in the broadband level of the noise below 10000Hz. However, we do see an FSS gain oscillation at 7320Hz showing up in the OMC_DCPD_SUM - but not in AS_C_LF or AS_A_RF36 - in fact that coherence has adip where we get the frequency noise oscillation. This strongly suggests that our broadband noise is NOT frequency noise. Evan also took the frequency noise transfer function - a preliminary analysis here also confirms: the frequency noise should be significantly below the O(1e-8mA/rtHz) noise level we see.
Note that the higher order mode estimates above were made using a slightly wrong modulation frequency. Updated estimates for the correct modulation frequency are attached to alog 20147
- ASC-AS_C GETS 2.5% of the HAM 6 light (see alog 15431) (NOT 0.25%)
Actually AS_C gets 400ppm of the light entering HAM6 -- the OM1 mirror was swapped from 5% transmission to 800ppm transmission in early April. See alog:17738.