40m QIL Cryo_Lab CTN SUS_Lab TCS_Lab OMC_Lab CRIME_Lab FEA ENG_Labs OptContFac Mariner WBEEShop
  Cryo Lab eLog, Page 2 of 59  Not logged in ELOG logo
ID Date Author Type Category Subject
  2889   Tue May 3 16:34:19 2022 ChrisDailyProgressPSOMAcurrent driver connections and cantilever acoustic damping

I opened the LIGO current driver chassis for the north laser, to check its PCB revision. It is v5 and will need modifications similar to those applied to the south laser's driver.

Cymac ADC channels 8-11 (zero indexed) were connected to the DAQ monitor signals on the south laser's driver. New filter modules X1:OMA-SLD_CATHODE, X1:OMA-SLD_HF_IMON, X1:OMA-SLD_LF_IMON were added to the x1oma model for access to these signals. Only the Imon signals are useful at the moment; the cathode voltage monitor overflows the ADC range.

While updating the model, I added a CTRL subsystem, intended for slow offloading of the control signal onto the current driver's LF input and/or the laser temperature. (See attachments 1 and 2.) These offloading paths can be triggered by the TRANS_MON signal.

Cymac DAC channel 4 (zero indexed) was connected to the LF current input, and the loop was closed to suppress the control signal offset (settings in attachment 3).

The cavity transmission was found to be intermittently glitchy (see attachment 4). Glitches seem to get excited along with the cantilever motion, but sometimes vanish for long stretches of time. We should probably check that the laser is not near a mode hopping region, cavity HOMs are not near degeneracy, etc.

Finally, as an experiment in acoustically damping the cantilever, I connected DAC channel 5 (CTRL_TEMP output) to the stereo speakers through an SR560 (gain=1; stereo volume=10). With a bandpass around the cantilever resonance and some tweaking of gain and phase settings (attachment 5), it was possible to knock the 44.5 Hz peak down by a modest factor (attachment 6, damping=red, no damping=green reference trace). The second harmonic was well suppressed also, but there was little visible effect elsewhere in the spectrum.

MEDM screens still have to be updated for the model changes.

Attachment 1: x1oma.png
Attachment 2: x1oma_ctrl.png
Attachment 3: ctrl_current.png
Attachment 4: signals.png
Attachment 5: ctrl_temp.png
Attachment 6: damp_dtt.png
  2888   Tue May 3 15:34:15 2022 aaronDailyProgressPSOMAloop TFs and noise spectra

I updated the calibrated noise budget based on Shruti's data above.

Shruti: Attachment 2 is the same using the LB error mon point

Attachment 1: CalibratedNoiseBudget.pdf
Attachment 2: CalibratedNoiseBudget.pdf
  2887   Tue May 3 09:51:42 2022 shrutiDailyProgressPSOMAloop TFs and noise spectra

Loop transfer functions

[Attachment 1]

  • For excitation at the 'A' port of the LB box, I measured the LB monitor (right after excitation), PDH signal (after the excitation travels the loop), and the control signal (after the LB).
  • The phase coherence of the PDH signal was quite low for excitations below 130 mV. So I measured the spectra multiple times with different excitations and plotted all of them together - some traces are less noisy but in the new somewhat automated data acquistion scheme I plotted everything together.  Below 1 kHz I had to use over 200 mVpp. The higher excitation is probably because the new HF drive has a mA/V coeff 100x smaller (0.2 mA/V iirc) than the ITC502 (20 mA/V).
  • For the next measurement I would probably try to excite at a different location because increasing the excitation further here causes the loop to lose lock

PDH calibration

I tried getting the data from the oscilloscope using the same script as earlier but saw 'connection refused' or 'connection timed out' each time. I did verify with `nmap` that the address I was using was correct. Not sure what was happening. Therefore, for the noise calibration I used the previous values.

python tek_tools.py 3 ~/cryo_lab/data/PDH/cal 220502.csv

Noise spectra

Attachment 2 - calibrated using the spectra measured at the PDH signal location

Attachment 3 - the raw voltage noise traces at the PDH, LB monitor, and Control points

Attachment 4 - Comparison of the noise spectra. Dashed - using PDH point, Bold - Using LB error mon, Dotted - previously measured noise.

Attachment 5 - The forest of peaks around and above a kHz is also now not as visible in the spectra measured with diaggui with the cavity locked. We certainly did see it earlier, not sure if we have any data saved

Overall, the noise looks much lower than the previous measurement and does not seem to have the forest of peaks. We will further update this plot with the other terms in our budget shortly and try to make sense of it.

I will add the pomona box to get additional gain at the cantilever resonance to try to further dampen the noise at 45 Hz and the higher harmonics of the cantilever



  1. Why is there a discrepancy between what is measured at the LB error mon and the PDH error signal (at the output of the mixer+LPF)? There is only a unity gain buffer in between. Does this mean the input is saturated? [to test] (05-Mar-22: realized PDH error signal was not measured from the right port)
  2. What is the strange peak at 900 kHz in the spectra? Also seen in laser frequency noise [ref elog 2886]
Attachment 1: LoopModel.pdf
Attachment 2: Noise.pdf
Attachment 3: Spectra20220502.pdf
Attachment 4: Noise.pdf
Attachment 5: noisespectra1.pdf
  2886   Fri Apr 29 11:21:51 2022 shrutiUpdatePSOMASwitch current driver to custom low noise

Using the delay line setup with a digital PID slow control on the Moku:Pro in multi-instrument mode, I first calibrated the delay line by offset locking and measuring the beat frequency with small offsets, and then measured the noise spectra at the delay line with no offset. The script I used to analysze the results is here.

Then I ran the noise calibration script and plotted this instead of the previously measured data [Attachment 1].

Attachment 2 contains a comparison of the old and new laser noises measured with the North-South beat. The north laser is still powered by the ITC510 which probably is why the forest of peaks still exist.


Attachment 1: CalibratedNoiseBudget.pdf
Attachment 2: laser_noise_comparison.pdf
  2885   Thu Apr 28 14:44:28 2022 aaronUpdatePSOMASwitch current driver to custom low noise

[shruti, aaron]

PDH calibration

Later in the afternoon, we again measured the PDH error signal sweep driving with 2V from the function generator at 1.3 kHz. I saved the trace in cryo_lab/data/PDH/cal/220428_PDHcal.csv.

The best fit parameters are

  • Slope of the error signal is 0.33 V/MHz, or 2.99 MHz/V
  • Fitted cavity length is 30.1 cm, and the FSR is 998 MHz
  • Fitted transmission of the cavity is 1.0%
  • The relative RF and LO phase at the mixer is 198 degrees (nearly out of phase, off by 18 degrees)

Dark noise measurement

We blocked the beam and measured the noise on the PDH error signal using an SR785. We sent the PDH error signal from the error mon point on the LB servo box (same location we used for the calibration above) directly to the spectrum analyzer.

The new curve for photodiode dark noise is at 1 Hz/rtHz above a few kHz, and the curve lies below our previous estimate for the frequency noise of the laser -- progress!

Calibration pipeline

I'm making a file calpar.yml in our scripts/calibrate_noise directory. This file will hold the key variables for generating the noise budget, including file names and paths, cavity parameters, and which PDH calibration to use for various measurements.

Beat note

Shruti measured the frequency noise on the beat note in the same manner as above. More details and results to follow.

Shot noise

We didn't measure the shot noise, but we estimated to see the shot noise above the observed photodiode dark noise would require about 3 mW of power at 1550nm on the photodiode. We can then scale the shot noise by the actual power level with the cavity locked.

Attachment 1: fitted_full.pdf
Attachment 2: CalibratedNoiseBudget.pdf
  2884   Thu Apr 28 11:19:24 2022 shrutiUpdatePSOMASwitch current driver to custom low noise

[chris, shruti, aaron]

 We turned down the current on the Thorlabs ITC502, rack mounted the current driver updated by Chris to D1200719 - v7, and used that to driver the south laser.

We adjusted the coarse knob to 4,90 to see roughly the same refl DC voltage as measured on the scope (which should translate to what I thought was ~135 mA from looking at the 'ANALOG OUT' on the ITC502 but seems to be <100 mA by Chris' measurement). The ITC 502 is still used as the temperature controller, the settings of which were not changed at this stage.

Then, we tried to lock the cavity. The ITC502 HF driver coefficient was 20 mA/V and the LIGO custom one is 0.2 mA/V. This required us to make few changes to the LB1005 settings and its output electronics. With a knob setting on the LB of 6,60 and nothing else in the output chain to the HF drive input of the driver (we removed the attenuator, 50 Ohm terminator that was teed, no pomona filter was used), we were able to lock the cavity for a very long time ( >5 min). The PDH signal noise when locked was ~150 mV pk-pk, dominated by the 45 Hz oscillations, relative to a pk-pk signal of 600 mV (Not quite sure of this value since the PDH signal was locked with a large offset  of ~-1 V in the output current and may have been at a different lock point than what we saw -- would need the slow temperature control to zero this).




  2883   Thu Apr 28 06:12:14 2022 ChrisElectronicsPSOMACurrent driver repair/upgrade

One of the D1200719 current driver boards (S1500270) was in the EE shop for repair after some unexplained current fluctuations were seen.

Previously I measured its output at different knob settings, and saw saturation when driving a 25 Ω load. The saturation happens when the low frequency current source block reaches its voltage limit. (This wasn’t seen in Zach’s previous test, probably because a lower-impedance load was used.)

The board has now been upgraded from v4 to v7, to fix high frequency oscillations and crossover issues. An annotated schematic is uploaded to the board’s DCC traveler.

A few tests were made to check the basic functionality:

  • A 25 Ω dummy load was connected for testing (note: this component must be rated for high power dissipation)
  • The unit’s nominal current draw from the supplies was about 200 mA
  • The response of the coarse and fine adjust knobs was measured. (Saturation occurs above a coarse knob setting of 7, when driving the 25 Ω load)
  • Noise spectra were measured on the SR780 analyzer at the HF and total current monitor outputs, with a coarse knob setting of 4. (Both were nearly white, ~10 nV/rtHz HF and ~20 nV/rtHz total)
  • A wideband transfer function from HF modulation input to HF monitor output was measured, using 4395A analyzer + ZSC-2-1 splitter + HAT-3+ attenuator on the differential input (which was driven single-ended)
  • The transfer function from the DAC input to the total current monitor output was checked (showing the expected 40 dB/decade filtering above 10 Hz)
Attachment 1: knobs.pdf
Attachment 2: spectra.pdf
Attachment 3: AG4395A_25-04-2022_170840.pdf
Attachment 4: data.zip
  2882   Tue Apr 26 16:18:11 2022 aaronNoise HuntingNoise Budgetcalibrating PDH error signal

I'm grabbing the pdh error signal from the oscilloscope over ethernet (labutils/tektronix/tek-grabber.py), and will calibrate the PDH error signal.

Just from examining the oscilloscope trace, the sideband zero crossings appear to be about 1.1 ms apart, while the DC resonance error signal is 708 mV tall and 56 us wide (measured with the cursor at the peaks). I think the slope of the error signal passing through resonance is related to the peak-to-peak measurements by

\frac{\partial \epsilon}{\partial f} = 2 \frac{\Delta V_\mathrm{pkpk}}{\Delta t_\mathrm{pkpk}} \frac{\Delta t_\mathrm{sideband}}{\Delta f_{\mathrm{sideband}}}

This is based on Eric Black's introduction to PDH stabilization, namely

D\equiv -\frac{8\sqrt{P_cP_s}}{\delta \nu} in equation 4.2, where D is the frequency discriminant and \delta \nu the linewidth.

And from Black's Fig 7, the PDH error signal's peak-to-peak height is 4\sqrt{P_cP_s}

Since the sideband frequency is 33.59 MHz, this implies the slope of the error signal is 0.41 V / MHz, or 2.4 MHz / V. I might be off by a factor of \sqrt{2}, depending on the definition of the linewidth in Black. Anyway I'll check this against the fit from PDH_calibrate.ipynb. In the end, looks like the by-eye estimate nearly agrees with the fitted result.

In PDH_calibrate.ipynb, I added a parameter to characterize the relative phase between the LO and RF signals. I think I've implemented this correctly, and find a 13 degree detuning of the LO from the RF signal at the mixer.

[stymied again by the lab ethernet and other distractions. Not seeing the oscilloscope show up on nmap, possibly just a bad cable. anyway will pick it up tomorrow morning]

update 4/27/22:

Simple debugging on the oscilloscope frontend let me grab the PDH trace via ethernet.

I tinkered around with PDH_calibrate to track down various factors of few and pi. Since the script fits the entire PDH error signal (swept through all three resonances), I can extract a couple of extra parameters that should be measured indepently for sanity:

  • The round trip cavity transmission (really 1-R) is 0.01, implying a finesse of 308.
  • The 'best fit' cavity length is 32 cm and the FSR 944 MHz

The best fit slope of the error signal is -0.46 V/MHz or 2.2 MHz/V

Attachment 1: fitted_full.pdf
  2879   Tue Apr 26 14:28:51 2022 aaronThings to BuyOpticshigh reflectors in lab

We have been wondering about the availability of high reflecting optics for the PSOMA cavity, especially the input / output coupler at 45 degrees incidence. I located about 25 of the coastline mirrors measured by Johannes in 2018 with an HR reflectivity of about 99.994% and AR reflectivity 0.04%. Along with the 1m, 180 ppm curved end mirror we have been using, the coastline mirrors should allow us to achieve a cavity finesse in excess of 10,000. I suggest we swap in these mirrors as soon as the noise budget is at all sensible.

If we require high reflectors at normal incidence, we may need to place another order, or check the properties of some of the other custom optics on our shelf.

  2878   Mon Apr 25 13:54:03 2022 shrutiNoise HuntingNoise Budgetcalibrating PDH error signal

After checking that the cavity locks, around the same temperature and current set-point, sweeping the current resulted in something that looked closer to a PDH signa [Attachment 1]. I think adjusting the RF phase wrt the LO before the mixer, by changing the relative length of the cables (10 m ~ 1 wavelength at 30 MHz; the modulation is at 33.59 MHz) should get the signal to look more like a PDH demodulated signal with a larger linear range around resonance.

I adjusted the cables between the OCXO and EOM and did manage to recover a proper looking PDH signal [Attachment 3]. Given that there were not many available cables to play with it is probably still not optimal.

I also increased the gain on the LB1005 slightly from 5.7 to 5.9 knob setting value.


Attachment 1: IMG_1620.pdf
Attachment 2: IMG_1619.pdf
Attachment 3: IMG_1622.pdf
Attachment 4: IMG_1623.pdf
  2877   Thu Apr 21 18:13:36 2022 Ian MacMillanMiscEquipment LoanAccelerometer/seismometer loan Wilcoxon 731A

Returned. See this elog


I borrowed a Wilcoxon Seismometer Model number 731A from the cryo lab. I am also going to borrow an amp/power source for it. Currently at the 40m


  2876   Wed Apr 20 15:48:07 2022 aaronNoise HuntingNoise Budgetcalibrating PDH error signal

I'm calibrating PDH error signal first to avoid recording data of 'dark noise' with no PDH signal like yesterday

I first locked the S laser to the cavity using the LB servo box in the usual configuration. I drove the 'sweep' input of the LB box with a 3.14 kHz triangle wave from a function generator, and monitor the PDH error signal at the 'error monitor' port of the LB box using an oscilloscope. I set the amplitude from the function generator at 1 Vpp, and tuned the sweep amplitude on the LB box until the sweep took the laser through both sideband resonance but not the +- FSR resonances. Then, I adjusted the timescale on the oscilloscope to capture one full sweep through both resonances, and triggered 'single shot' until I captured all three resonances.

I didn't get a clean error signal -- instead I'm always seeing a flat region as the DC frequency sweeps through resonance. Attachment 1 shows the PDH error signal in purple (ch3), refl in blue (ch 2), and trans in yellow (ch1).

I thought maybe the LB servo was holding the frequency on resonance (despite the servo being disengaged), but turning the servo gain down to 0 does not affect the error signal trace. Maybe I should watch the error signal before the unity gain buffer amplifier that precedes the LB box's error monitor? Probably someone (maybe myself in a previous elog) has seen this, so I'll ask around.

Attachment 1: 9B21DA6F-14BA-453D-93BB-C6C8803CBCA0.jpeg
  2875   Wed Apr 20 14:27:41 2022 aaronLab InfrastructureGeneralstarting up after power outage

I'm turning on the lab electronics after the planned outage this morning.

  • Turned on following electronics.
    • cymac1, gaston, spirou, and cominaux
    • Sorensen DC supplies, and the GPS-30300 supplying 8 VDC. With all other electronics listed below powered on, the Sorensens supply +18 V with 2.7 A, and -18 V with 1.0 A.
    • AA, AI, and acromag chassis.
    • PSOMA rack Rio laser drivers, and engaged both TEC
    • PSOMA REFL, TRANS, and beat note photodiodes
    • PSOMA table oscilloscopes and TV
    • moku
    • PSOMA enclosure HEPA FFU (the one unit that fits on the enclosure)
  • Ran standard cymac startup script
    • rtreset
  • Confirmed status of digital system
    • systemctl status on cominaux is 'running'
    • testpoints on the realtime system are green, except the IOP timing indicator (red as usual) and the DAQ_DETAIL GPS indicators (white boxed as usual). See attachment 1.
    • The temperature and particle count channels were dead... in fact they haven't been recording for almost 6 months no.
      • I restarted rts-edc on cymac1, but the channels still read 0 despite the digital display on the particle counter reporting nonzero values
      • I noticed that the USB hub interfacing the particle counter and a few other devices to cominaux had no indicator lights on. I tried replacing its DC power supply, but still no response.
      • Finally, I bypassed the hub and plugged the particle counter directly into cominaux. After another 'systemctl restart rts-edc' on cymac1, the particle counts are now being recorded to frames. (note that only the particle counter temperature channel is recording; the AD590 temperature monitor is not being supplied with power)
Attachment 1: Screenshot_from_2022-04-20_14-41-11.png
  2874   Tue Apr 19 14:50:00 2022 aaronNoise HuntingNoise Budgetphotodiode dark noise measurement

Things still look pretty weird. Today I'll measure the photodiode dark noise (the previous curve was theoretical, based on the known transimpedance of the 1811 and properties of the PDH electronics).

Measurement steps:

  1. Measure the PDH error signal with beam blocked, sending the output of the PDH box directly to SR785.
    • blocked beam with a razor blade dump
    • Disconnected the IF output from the PDH mixer/lowpass from the LB box input, and instead connected it to the SR785 input (1 MOhm)
    • Removed the RF power splitter and second PDH mixer from the system. These electronics are for locking both lasers to the cavity, but we want to understand the noise for a single laser locked to the cavity first. The electronics are still on the table and straightforward to set up.
    • I then recorded the power spectrum using labutils/netgpibdata/SRmeasureWideSP.py. The parameters .yml file is in cryo_lab/data/PD_dark (along with the measurement results).
      • The noise has a 1/f corner near 40 Hz, then is flat until about 20 kHz when it increases in a series of peaks. Since it's uncalibrated, I'll say no more on the overall noise level than that there is a discontinuity in the spectrum as recorded (possibly indicating a range issue for part of the measurement?).
      • When I went to turn off the lab electronics (below), I found that the OCXO preamplifier box was already off from when Shruti and I were rearranging some cables last time... so, today's measurement was a complete waste of time crying
  2. Calibrate the PDH error signal -- did not complete today
    1. Drive the laser current above 1 kHz and just enough frequency deviation that the full PDH error signal and both sideband resonances are visible on an oscilloscope trace
    2. Use [calibratePDH.ipynb?] to determine the slope of the PDH error signal near the DC resonance. The x-axis is converted from 'seconds' to 'Hz' by assuming the distance between the sideband resonances is twice the sideband frequency.


There will be a planned electrical outage tomorrow morning at 6 am. I turned off the following electronics before leaving the lab:

  • All laser diodes (2 Rio lasers on the PSOMA table)
  • All laser current and TEC drivers (2 Thorlabs drivers on the PSOMA rack, 2 custom drivers on the cryo cavs rack)
  • cominaux, spirous, gaston, and cymac1
  • moku pro
  • RF amplifier used for delay line measurement
  • both Sorensen DC supplies (after turning off the electronics they power)
  • Acromag chassis, and AA and AI chassis for the fast DAQ
  • Marconi, spectrum analyzers, SR560s, function generators, oscilloscopes, LB servo box
  • CCTV
  2873   Tue Apr 19 12:07:46 2022 shrutiLab InfrastructurePSOMAPSOMA table enclosure assembly

[Aaron, Shruti]

We managed to install one of the HEPA filter units on our enclosure today, the other one could not be installed because it did not have enough clearance of the overhead tubelight.

  2872   Fri Apr 15 16:25:00 2022 ranaNoise HuntingNoise Budgetlaser frequency noise measurement

The ITC 502 has an analog input modulation impedance of 10 kOhm, according to the manual. How did you set up the 16 dB attenuation?

  2871   Thu Apr 14 13:45:08 2022 aaronNoise HuntingNoise Budgetlaser frequency noise measurement

[shruti, aaron]

We are measuring the frequency noise on the beat note between the north and south lasers. The level of observed frequency noise on the beat note should tell us whether the calibration error is in our total measured frequency noise curve or current noise curve.


  • We used the Agilent 4395A to observe the North-South beat note and confirm we are sending 10 dBm to the delay line box
  • We then tuned the beat note to near 43 MHz and nulled the output of the delay line box
  • We are monitoring the (SR560, G=1 buffered) output of the delay line box using an oscilloscope and the Moku:Pro
    • With Moku:Pro's multi-instrument mode, we use moku's PID controller to apply feedback to the temperature tune port of the ITC510 driving the north laser and hold the beat note on the delay line null. (attachment 2)
    • Simultaneously, we record the power spectrum of the delay line box output. We saved the trace of the spectral density from the delay line output at several frequency ranges to produce the frequency noise curve for the beat note attached
    • To calibrate the delay line, we drive the ITC502 mod in port (20 mA/V) with 0.01 Vpp at 3.14 kHz passed through 16 dB attenuation. The delay line output has a 1.01 mVpp peak at the drive frequency. Assuming the laser diode response is 150 MHz/mA, this means the delay line responsivity is 4.7 MHz / mV.
      • Shruti also measured the delay line responsivity by introducing a small offset from null in the temperature control loop, and observing the difference in beat note frequency on the spectrum analyzer. She got a similar responsivity calibration (within a factor of 2) *

Unfortunately today's measurement did not agree with any of the previous measurements on the noise budget. We'll next try re-measuring the transfer functions and frequency noise of the locked cavity.

* Shruti: I added an offset to the digital PID control on the Moku:Pro at \pm0.5 mV from the locking point which was at -0.4 mV to zero the delay line using a simple Integrator with a UGF of 312 mHz. At the two points the peak of the beat note was at 44.43 MHz and 42.42 MHz respectively resulting in a calibration factor \frac{44.15-42.42}{1} = 1.73 \textrm{ MHz/mV} .


Attachment 1: CalibratedNoiseBudget.pdf
Attachment 2: IMG_16928781279A-1.JPEG
  2870   Wed Apr 13 11:10:28 2022 aaronNoise HuntingNoise BudgetPSOMA noise budget

New measurements

  • Current noise
    • Measuring current noise following the procedures in elog 2737.
      • I'm sensing the voltage across the laser diode from the DB9 breakout to an SR560 (G=100, AC-coupled, 100 kHz lowpass filter with 6 dB/oct rolloff) to an SR785
      • Updated the noise budget accordingly (see attachment 1), but must still have a calibration error because there's still a disagreement with the measured frequency noise
      • Not sure why the current noise turns up at high frequency for both ITC502 and ITC510
    • Diode sensing resistance
      • Near the operating point for the South laser diode (8 kOhm TEC resistance, 117 mA LD current), driving the ITC502 with 1 Vpp at its current modulation input results in .134 Vrms fluctuations across the laser diode as measured at the DB9 breakout by a fluke multimeter in AC voltage mode
        • The driver's modulation coefficient is 20 mA/V, which means the resistance of the South Rio laser diode near this operating point is 19 Ohms
      • Near the operating point for the North laser diode (7.5 kOhm TEC resistance, 143 mA LD current), dirving the ITC510 with 0.1 Vpp at its current modulation input results in 0.073 Vrms fluctuations
        • The driver's modulation coefficient is 100 mA/V, so the resistance of the N Rio laser diode near this operating point is 21 Ohms


  • added new DB9 cable in fiber box, eliminating one DB9 adapter in the laser diode current drive path
Attachment 1: CalibratedNoiseBudget20220411.pdf
  2869   Tue Apr 12 15:00:39 2022 Ian MacMillanMiscEquipment LoanAccelerometer/seismometer loan Wilcoxon 731A

I borrowed a Wilcoxon Seismometer Model number 731A from the cryo lab. I am also going to borrow an amp/power source for it. Currently at the 40m

Attachment 1: IMG_1558.jpeg
  2868   Tue Apr 12 12:16:58 2022 AaronDailyProgressSi fabBasic cantilever scribe and break

This morning, I used the Dynatex scriber-breaker at the KNI to make three silicon cantilevers 7cm x 1cm x 300um. These are similar dimensions to those used in Zach's thesis. I used a fresh from the shipping box (no RCA, no PECVD) Boron doped 3" wafer. The wafer broke along the crystal axis while I was loading it into the scriber, and two of the 7cm cantilevers include a small section of the curved wafer edge. This should be acceptable for now, and could be improved by using 4" wafers.

The scribe and break process takes under an hour.

Attachment 1: 51649EDB-A7BC-49DC-AD53-F95345413D0E.jpeg
Attachment 2: 2741BA36-BF23-4289-9BBC-6054A10A3ECC.jpeg
  2867   Mon Apr 11 16:30:12 2022 aaronNoise HuntingNoise BudgetPSOMA noise budget

We'd like to progressively improve the noise performance of the PSOMA cavity by updating a noise budget when we make changes. Attached is the noise budget we have now, which I'll be updating throughout the week. The budget contains rough estimates of the following contributions, and caveats or notes are listed below each source. Full calculations are on git.ligo.org in NoiseSpectraCalibration.ipynb

  • Frequency noise of the Rio lasers
    • Laser frequency noise is that estimated in elog 2807
    • This estimate only extends to about 100 Hz, since that was the target of the previous current noise measurement. We can of course repeat the three corner hat measurement, use the beat note between the two Rio lasers as a conservative estimate of the single-laser frequency noise, or find a different elog of Johannes' or Zach's that extends to higher frequency.
    • Given this, I've just used 20 kHz/rtHz at 1 Hz with a 1/f dependence extended in both directions for the Rio laser frequency noise
  • Photodiode dark noise
    • Photodiode dark noise is not measured, but calculated using known properties of the Newport 1811 photodiode: 2.5 pA/rtHz current noise across 40,000 Ohm transimpedance, resulting in 100 nV/rtHz voltage noise at the output of the photodiode.
    • The voltage noise out of the photodiode is passed through the PDH electronics, including a level 17 mixer and lowpass filter that together introduce up to 6 dB of loss. This results in 100 uV/rtHz noise on the PDH signal due to photodiode dark noise.
    • Finally, the voltage noise out of the PDH electronics is referred back to laser frequency noise using a cavity response of 123 mV/MHz and cavity pole at 2.26 MHz per elog 2846. The DC noise level due to photodiode dark noise is 800 Hz/rtHz, and rises like f^2 above the cavity pole.
  • Shot noise
    • I estimated that the incident laser power on the cavity is 200 uW and that we have 75% mode matching, but didn't locate a record of laser power at the time that the frequency noise was measured. I'm estimated the DC power incident on the photodiode with the cavity locked at 50 uW.
    • The responsivity of the photodiode at 1550 nm is about 1 A/W, and the noise spectral density from Schottky S=2e|I| (where e is the electron charge). The shot noise is propagated through the same PDH and cavity responses as the PD dark noise.
  • Current drive noise
    • Laser current noise is just wrong
    • The measured frequency noise is for the South laser diode driven by an ITC502. The current plotted is for the ITC510 driver as measured one year ago. The noise of the ITC510 is known to be higher than ITC502, but even then the reported current noise is too high to be believed. The current noise measurement is quick, so I'll repeat it this week and update that curve.
  • Seismic noise
    • Seismic noise is a very rough estimate. I used x=10^{-8}\mathrm{m/ } \sqrt{\mathrm{Hz}} * \frac{1+(f_\mathrm{cantilever}/f)^2}{f^2}
    • I'm not totally confident about this expression, but it's due to 1/f^2 dependence for ground motion highpassed through a cavity with a single suspended optic
  • Total of estimated noise contributions
    • I've ommitted the total noise curve at this time... I think it's not useful at this time as several of the noise conributions lie above the measured noise curve.
  • Measured frequency noise
    • The measured frequency noise is that reported in elog 2846

There are obviously major discrepancies between the measured noise and the estimated noise contributions. I'll work on correcting the errors in either the frequency noise measurement and calibration or the estimates this week.

Attachment 1: CalibratedNoiseBudget20220411.pdf
  2866   Wed Mar 30 16:18:58 2022 shrutiDailyProgressPSOMAMixing signal and pump (north and south)

I fixed the fiber connections, re-organized the fibers a little, and finally was able to see a little less than 1.5 mW of power contribution from each laser at the output of the fiber-to-free space coupler.

PM 50-50 BS convention also suggests that the diagram we had in the previous configuration was wrong.

North-> REV REFL (red)

South-> FWD IN (white)

50% output taken from FWD TRANS (white)



  2865   Tue Mar 29 16:03:40 2022 shrutiDailyProgressPSOMAMixing signal and pump (north and south)

[shruti, aaron]


Our next step is to try and lock both lasers to the PSOMA cantilever. For this we changed the setup to mix the two beams in fiber after modulating them at slightly different frequencies. We changed it from 'old setup' to 'new setup' shown in Attachment 1. When we turned on the south laser after making this change we only saw 4 uW when we expected 2 mW. Will troubleshoot this tomorrow.


Before changing the setup to the 'new setup' we also tested the configuration Aaron had it earlier in elog 2864 and saw that the power levels were what we expected: 

  • North laser power at 90% output: 4.8 mW
  • Power after the EOM in north path: 2.5 mW
  • 10% input->90% output of BS: 280 uW
  • Adding south EOM output to 90% input: 3.8 mW
  • Without north: 3.5 mW

The north EOM does not seem to be problematic.


Attachment 1: addpump_.pdf
  2864   Thu Mar 10 13:43:20 2022 aaronMiscLasermode matching and slow servo, three corner hat revisited

I'd like to improve the mode matching. Since it takes a while, I'll also set up a slow servo on the TEC controller to hold the laser near resonance.

  • Slow temperature control loop
    • We already had a cable carrying the relevant fast signals to the AA chassis (PDH error signal, which I connected to the LB box error monitor, after the summing amplifier; the PDH control signal, which I connected to the LB box Aux output; and REFL MON and TRANS MON, both of which are monitored by an oscilloscope on the table before going to AA chassis).
    • The monitor channels are reaching the AA chassis, but not reflected in cymac. It's been a while and there's a TIM error, so I restarted the models by running 'rtreset' on cymac1. Afterwards, mostly green indicators on X1IOP_GDS_TP and X0DAQ_DETAIL... but white boxes on the GPS timing signal (attachment 1; has been this way for at least a year). I checked the simulink model and realized I was just driving the wrong port of the AA chassis; quick fix.
    • Eventually remembered how to run the temperature control loop:
      • My calc channels weren't calculating, not sure why. Needed to systemctl restart SoftIOC.service from cominaux (also restarted modbusIOC.service for good measure).
      • Set the PID parameters as follows... it would be nice to set these up to initialize automatically. All unlisted values (like the proportional gain constant) were left at 0 or are updating calc channels.
        • channel name value
          X1:OMA-SLD_TEMP_TIMESTEP 0.1
          X1:OMA-SLD_TEMP_TT 0.1
          X1:OMA-SLD_TEMP_TF 0.1
          X1:OMA-SLD_TEMP_KI 0.1
          X1:OMA-SLD_TEMP_EN 1
      • in a tmux session on spirou, go to scripts/temp_control and run the python script PIDLocker_cavcontrol.py with the config file PIDConfig_SLDTemp.ini
      • I also updated the epics definition of X1:OMA-ERC_MON_RATIO with the new 'dark' values of TRANS_MON and REFL_MON
  • mode matching
    • With the temperature control loop running, I roughly doubled the mode matching efficiency (as estimated by the ratio of transmitted / reflected light) by adjusting the input steering optics
  • Three corner hat: Since I couldn't diagnose the full N laser beam path, I decided to throw together a synchronous three corner hat measurement
    • connected a bunch of fibers. Cleaned all fiber tips before (re)connecting
      • Moved the 99-1% fiber BS from the cryo cavs table into the PSOMA fiber box
      • Ran a 10 m patch cable from cryo cavs table to the PSOMA rack and into the fiber box. Where the cable crosses the room, I put it in a cable wrap (same model as the one carrying fiber from the rack to the PSOMA table).
      • The patch cable takes the Teraxion beam directly from the laser to the 1% coupled port of the 99-1% BS
      • I disconnected the 50-50 BS mixing N and S PSOMA lasers from the 1611 beat note PD, and instead sent the N-S mixed output to the 99% coupled port of the 99-1 BS. The output of the 99-1% BS now carries PSOMA N, PSOMA S, and Teraxion lasers, and I sent this beam into the 1611 PD.
    • Set up the Moku Pro in multi-instrument mode with an oscilloscope, spectrum analyzer, and two phasemeters
    • Turned the lasers on one at a time and check that I'm not saturating the 1611 PD... I'm up to 8 mW (of 10 mW saturation), so I'm seeing some higher order RF beats in the spectrum, but close enough for today.
    • Used the Agilent spectrum analyzer to scan around the PSOMA N and S lasers until I get all three beat notes on the Moku's spectrum analyzer. I identify the beat notes by scanning N or S lasers and seeing which pair moves.
    • Set up one phasemeter to track the two of the beats, and the other phasemeter to track the third. I used a 10 kHz bandwidth to make sure the beat note doesn't move out of band during the measurement, but didn't play around with lower BW.
    • Set one phasemeter to acquire data at 19.1 kHz with a 15s delay, then quickly switch over to the second phasemeter and start its data acquisition at 19.1 kHz about 15 s later. It would be nice if one could start recording data on all instruments at once, and I'm not sure what the final data files will look like, hopefully I can sync them -- maybe the data logger instrument does this?
    • Exported the files. I'll check them out on my laptop at my convenience.
  • Returning system to previous state: Shruti requested I leave the cavity as before, so I'm returning it to the transfer function measuring state with only one laser in the system.
    • swapped a couple fibers
      • Turn off the Teraxion laser, but I'm leaving it connected to the fiber box for convenience. Also turn of PSOMA lasers temporarily, and return them to the previous temperature and current settings.
      • Replace the 99-1% combiner output with the 50-50 mixer output, so only the NS beat note is sent to the 1611
      • The beam sent to the optical table was the 90-10% combiner output containing the North and South beams after the EOM. I replaced this with the output of the S EOM (containing the S laser only, with sidebands)
    • Swapped a few electronics in the PDH loop
      • Removed the splitter and second PDH mixer box from the loop. The REFL PD's AC output once again goes directly to the S laser's PDH mixer box.
      • Replaced the Pomona filter box with a 10 dB attenuator at the LB box input
    • replaced the monitor channels
      • Moku Pro is monitoring pickoff of the amplified NS beat note from the 1611 on channel A, LB box error mon on channel B, current control signal on channel C, and the buffered delay line phasemeter output on channel D.
      • 1611 beat note is sent to an RF amplifier, then the delay line box
      • Turned on the 'clicking' oscilloscope
    • LB box returned to previous state: gain knob at 5.3, pure integrator mode
    • turned off the temperature control loop and set TEMP_TUNE to 0
    • Confirmed the cavity still locks and the NS beat note is still under 100 MHz
Attachment 1: Screenshot_from_2022-03-10_14-26-10.png
  2863   Wed Mar 9 14:20:26 2022 aaronDailyProgressElectronicsupdates to fiber patch cables, realignment, N laser in path, relock
  • Replaced the patch cable carrying PSOMA South laser from the fiber box to the table with 2x P3-1550PM-FC-10 cables (one for S, one for N), and wrapped them in the protective cable cover. Clamped the cables with rubber on the table and fastened the flexible cable wrap to the optical table and enclosure. Then had to slightly realign the fiber launch until the 00 mode locks again.
    • Also replaced the 10 dB attenuator at the LB box input with a 6 dB attenuator, to make up for the 50-50 RF power splitter I introduced before the PDH mixers yesterday
  • Completed the N laser path: 90% of pickoff BS to N EOM, then to [10% path of a 90-10 fiber splitter... asking around for a 50-50 splitter, unless I'm missing something the 10% path doesn't even register
    • cleaned any fiber tips before connecting
    • clamped with rubber the EOM
    • Found a 90-10 splitter off the shelf, but we're out of spare fiber component mounting trays so I left it in the plastic packaging and secured with Kapton tape
  • Locking activities
    • Reducing the laser power (due to the RF power splitter, and the fiber splitter that mixes N and S beams) obviously pushes us to higher open loop gains... but we're close to the phase margin, so that doesn't come easily
    • I added Shruti's loop compensating filter box before the LB servo input, which allowed me to increase the LB servo gain knob to 6.2 (up from 5.3, on a log scale) before the loop oscillates... although, there's still not enough gain to completely suppress the 40 Hz oscillation from the cantilever's fundamental mode. I also removed the 6 dB attenuator and incidentally swapped from (-B) input to (+A) input on the LB box.
      • It's probably also worth improving the mode matching again. A couple times I locked on to a 1-0 mode.
      • Also played around with the LFGL but noticed no qualitative difference in the system (especially REFL DC) with LFGL off or on.
    • Plan is to use the Moku Lab as a PID controller lieu of a second LB servo box. However, turning on the N laser doesn't even register on the REFL photodiode... this is surprising, the PD should definitely be sensitive enough to see the N laser even at 10% of full power (should be 400 uW).
      • Is this an alignment issue? Wrong part of the laser's T-I curve? Something wrong earlier in one of the fiber components (seems likely)? I couldn't locate the Thorlabs power meter, so had trouble testing these. Has anyone borrowed it? Might just be looking in the wrong places, but usually it's on the optics table.
  • Misc
    • The second oscilloscope on the optics table was making an annoying switching sound, so I turned it off.
  2862   Tue Mar 8 13:45:20 2022 aaronDailyProgressElectronicssecond PDH box and EOM

Entered the lab around 1:30pm to measure the frequency noise of the PDH error signal with the cavity locked. But I found some things had changed. Seems the system was set up to measure transfer functions. And, someone started setting up a monitor in the enclosure. Cool, but what happened?

I made a second homodyne mixer box (using ZFM-3H-S+ followed by SLP-5). I split off (ZFRSC-123-S+) the REFL signal to send half the RF to this second box, and connected the oscillator channel 2 output to the new LO in. I also placed the our second EOM in the fiber box, but haven't connected the fibers. In this stage, I'd like to mix the two lasers in fiber to avoid aligning a second beam in free space... but would need a spare 50-50 beamsplitter to do so without rearranging the order of fiber components.

  2861   Thu Mar 3 13:09:25 2022 aaronNoise HuntingLasersome noise budgets maybe

I came to measure to measure the laser frequency noise budget and see what I can learn.

I was having trouble finding a lock despite the beamspot flashing. Eventually switched the PDH signal to the (-B) input of the LB box (was on +A), and the cavity immediately locked.

not much else, I figured out what was going on in the notebooks in scripts/calibrate_noise

  2860   Tue Mar 1 17:08:47 2022 shrutiLab InfrastructurePSOMAPSOMA table enclosure assembly

[Aaron, Shruti]

We installed all acrylic panels on the enclosure today [Attachment 1]. Doing this required adjusting nearly all the lower horizontal bars multiple times to get the doors to fit perfectly, and hammering down the rails.

The last part of the enclosure building is to install the HEPA filter units. We wanted to do that today but were not able to because the units are missing power cords (need to order).

Attachment 1: WorkNB_Y3Q2.pdf
  2859   Wed Feb 23 17:23:35 2022 ChrisElectronicsPSOMAcurrent driver status

The current driver now in the EE shop for troubleshooting (chassis S1500267/PCB S1500270) looks to be a D1200719-v3 PCB that has been modded to v4. The mod to v4 is also documented in the DCC traveler.

I repeated Zach’s 2017 coarse current adjust knob response test with this unit driving a 25 Ω dummy load, and got slightly different results. There’s a bigger offset at the zero knob setting, and saturation sets in above a knob setting of ~7.

It may be that Zach followed through on his stated plan to mod the circuit with a voltage limiter, like the one added in v7 of the schematic.

Next steps are to check the board for signs of this possible mod, and see if the observed misbehavior can be reproduced.

Other mods in v6/v7 of the circuit address high frequency oscillations and crossover issues, and we may want to implement those as well.

Attachment 1: coarseknob.png
  2858   Tue Feb 22 11:51:55 2022 shrutiNoise HuntingLaserlaser transfer function and frequency noise

In the loop model I had posted earlier a part of the discrepancy in the calibration of the delay line can be explained because I had not used a buffer at its output. After adding an SR 560 buffer while the calibration is closer than expected, there is still some discrepancy even in the DC MHz/V conversion which I am still investigating.

Nonetheless, I have attached the latest laser TF in Attachment 1 and using the same calibration, a measured laser frequency noise spectrum in Attachment 2.

Attachment 1: LaserTF.pdf
Attachment 2: LaserFreqNoise.pdf
  2857   Fri Feb 18 10:43:18 2022 aaronDailyProgressPSOMAenclosure rails, laser noise

[aaron, shruti]

Yesterday (Thursday Feb 17), I installed the rails for the PSOMA enclosure's acrylic panels. Meanwhile, Shruti used the delay line box to measure the frequency noise of the NxS beat note. Also measured the open loop transfer function from current modulation voltage to beat note frequency. Adding an SR560 as a unity gain buffer between the delay line box and the moku brought the measured curve closer to the expectation, at least at low frequency. I think the next step is to repeat the direct measurement of current noise from the ITC502 and ITC510 so we can put it on the same plot as the laser frequency noise.

(Probably Shruti is working on an update about this, but I wanted to document it before I forget)

  2856   Fri Feb 18 10:40:04 2022 aaronHowToElectronicshow to Moku on wifi

Instructions from liquid instruments (via Rana) on controlling Moku from the ipad while it is connected to the internet for data uploads:

  1. hook up Moku to ethernet via the wire. Make sure the Ethernet lights are blinking. Moku can be DHCP, or whatever.
  2. Connect the ipad to the cryo-fi wifi network so that iPad and Moku are on the same cryo network. i..e do not connect ipad to the 'Moku-XXX' access point.
  3. Should be able to coontrol Moku from iPad and also connect to internet

  2855   Tue Feb 15 16:56:50 2022 ranaUpdateNoise BudgetUpdated loop model

yes - this is looking great. I think all of the important terms are there.

It looks like something may be iffy with the placement of the poles/zeros. Usually when the magnitude slopes down, the phase also goes down. i.e. 1/f means the phase should be -90 deg, not +90 deg. Sometimes this is a real physical situation like in the RIO TF, but for most electronics, cavities, etc. its just a syntax error and you have to make sure the poles and zeros are in the left-half of the complex plane. Might be syntax weirdness in some python packages.

  2854   Tue Feb 15 11:21:18 2022 shrutiUpdateNoise BudgetUpdated loop model

Loop Model

  • The script used to generate the plot in Attachment 1 can be found here and the data is in the data repository.
  • The different elements of the loop  I have considered are:
    • Laser driver [mA/V]
    • Laser [Hz/V]
    • Cavity [V/Hz]
    • RFPD, mixer and low pass filter [V/V]
    • Attenuator: 10 dB [V/V]
    • Servo LB1005 [V/V]
    • Pomona filter [V/V]
  • I made separate measurements of the following:
    • Laser and driver combined: I used this script and Aaron's delay line setup. The script shows both the driver and laser separately, but I have shown them together here in the pink trace. The measurement of the driver separately was a gross estimate since I used the 'ANALOG OUT' port of the driver instead of its actual output which needed me to turn off the laser (which I can do at a later time if needed).
    • LB1005: I measured its transfer function by the method outlined in *Attachment 2* to obtain the green solid line in the graph. But in the estimates of the open loop TF I have used the analytic function (green dashed line) of 1/f with 21 dB gain at 3 kHz which corresponds to the knob setting, since they agree roughly in magnitude and phase to upto almost 1 MHz.


OLTF direct measurement (solid grey)

  • This is the open loop transfer function that I directly measured earlier on 7th Feb


OLTF estimated 1 (dashed black)

  • Here I considered only a constant magnitude for the laser and driver
    • Laser driver 20 mA/V
    • Laser 150 MHz/mA
  •  For the cavity, I used information from the measured PDH signal, and considered it to be a single pole low pass with gain 132 mV/MHz and pole at 2.3 MHz
  • I only considered the low pass filter (ignoring mixer and RFPD) and approximated it as \frac{1}{f/f_c + 1} where f_c = 5 MHz
  • The attenuator was a constant 10 dB attenuation with no phase change (With DC coupling, and 50 Ohms input and output I measured this to be true enough)
  • For the LB1005, I used the analytic form shown by the green dashed line
  • I have also used the analytic form of the designed filter shown by the purple dashed line


OLTF estimated 2 (solid black)

  • The only change in this from 'OLTF estimated 1' is that I have added the phase information of the laser and driver transfer function I had measured. I kept the magnitude of this transfer function to be the same constant calibration factor used in 'OLTF estimate 1'.
  • Using this, the phase of the model matches the measured one quite well.


OLTF estimated 3 (dotted black)

  • Everything the same as in 'OLTF estimated 1' except for the laser and driver transfer function
  • I used both the magnitude and phase of this measurement (shown in pink on the plot).
  • I have to re-check the delay line calibration because there is a constant factor discrepency between this and the measured OLTF for the most part
Attachment 1: LoopModel.pdf
Attachment 2: LB_meas_setup.jpeg
  2853   Thu Feb 10 14:12:56 2022 aaronDailyProgressLaserlaser transfer function

[shruti, aaron]

Shruti set up a transfer function measurement using Anchal's 'phasemeter TF' script: while the moku phasemeter monitors the beat note frequency, the moku output drives one of the laser currents with a sine wave. By stepping the driving frequency, one can measure the transfer function from voltage at the laser driver 'mod in' port to laser frequency.

The script worked fine, but the measurement did not. By manually setting up the same measurement at a fixed frequency through the ipad interface, we found that the beat note frequency is drifting too much for the phasemeter to keep up. With even small excitations to the current drive (>5 mV at mod in), the frequency moves too much for the phasemeter's loop to keep up, and the phasemeter loses lock.

Instead, we are setting up the delay line frequency discriminator box. We will directly measure the transfer function from laser current drive mod in to the output of the discriminator box. To get the transfer function of just the current driver and laser, we will need to independently measure the transfer function of the frequency discriminator (including RF amplifier and delay line), but that's no problem.

After setting up the DLFD, we tuned the beat frequency to observe the max, min, and null response of the delay line. The peak/trough of the discriminator signal are at 150 MHz and 77 MHz. The delay line output is nulled at 118 MHz. The noise on the beat note (with no additional modulation applied) causes 56 mV fluctuations pk-pk on the delay line output. We used the moku output to drive the laser current mod in, and found that a 5 mV excitation at 1 kHz was visible above the noise but still not using much of the discriminator's full range (not very quantitative on whether we're completely in the linear regime, since the noise on the beat note is already quite large).

  2852   Wed Feb 9 16:25:54 2022 ranaUpdateNoise BudgetUpdated pomona filter, loop model

For the loop mode, I suggest plotting all of the components as separate traces in the Bode plot first. e.g. cavity, driver, mixer LP, LB box, etc. Then we can see which of the components make sense. For the magnitude, each one will have different units, but you can indicate those in the legend (i.e. W/m, Amps/V, Hz/Amp, etc).

When doing the swep sine TF, I suggest going from 10 Hz - 10 MHz, and setting the Moku to use more cycles, so that the low frequency coherence is increased. You can also stitch together a few different measurements: this allows to use different magnitude at different frequencies. For 10-1000 Hz, it may be easier to use the CDS system since that makes it easy to use varying amplitude v frequency.

  2851   Mon Feb 7 11:55:07 2022 shrutiUpdateNoise BudgetUpdated pomona filter, loop model

Attachments 1 and 2: Pomona Filter

  • I was able to lock the cavity at the same gain with the updated pomona filter circuit (shown in Attachment 1) and no 10 dB attenuator before the current driver.
  • This new filter has a pole a little above the cantilever resonance (~50 Hz) and zero at 1.5 kHz in order to remove the occurrence of the many peaks seen earlier between 1-10 kHz separated by roughly this cantilever resonance frequency.
  • The resistors were increased in order to incorporate a ~10dB attenuation and remove the Mini-Circuits attenuator which was at its output. The Mini-Circuits attenuator was not properly impedance matched.


Attachment 3: Loop Model first attempt

  • I tried measuring loop transfer functions and noise spectra once again but my choice of injection of the small signal (LB servo's -B port) resulted in the lack of coherence in the measurements at lower frequencies, hence I tried to roughly estimate it. Code found here
  • I have yet to include the phase information of the laser and current driver, which may seem to matter significantly above 1 kHz
  • The magnitude at lower frequencies don't seem to match. Could this be because the cantilever's optomechanical response is quite high and therefore modifies the error signal?
  • I also forgot to include the low pass filter after the PDH demodulation mixer which is at 5 MHz


Attachment 1: filter.jpeg
Attachment 2: FilterTF.pdf
Attachment 3: LoopModel.pdf
  2850   Wed Feb 2 11:08:47 2022 shrutiUpdateNoise BudgetMeasurement and calibration of noise spectra

Attachment 1: New open loop transfer function

  • After adding the pomona box filter to the loop, I tried locking again. The max gain I was able to set corresponded to a UGF of 13 kHz -- I was unable to recover the previous UGF of 22 kHz.
  • I think the phase plot makes sense with the addition of another zero and pole, but I'm not sure if the phase margin is to be measured from +180 degrees in this case?
  • Also, I had to do this with the 10 dB attenuator before the modulation input of the current driver. Even with adjusting the servo offsets the modulation without the attenuator was high enough to reach the current limit for the diode nearly every time.
  • Attachment 4 contains estimates of the resistors inside the Mini-Circuits 10dB attenuator using measurements provided by Rana. Attachment 5 contains the filter TF measured with and without the 10dB attenuator at its output - the 45 Hz pole is now at 250 Hz which is not what we want
  • The solution is probably to modify the pomona box with higher series resistance, and lower capacitance, and use the box without an attenuator at its output.

Attachment 2:

This contains the calibrated noise spectra without accounting for loop suppression for before (blue) and after (red) the pomona box filter was added. Since the servo gain is not as high as required and filter pole is much higher than we want this may not be a valid comparison.


Attachment 3: Corrected calibration for noise spectra

Chris pointed out that I was not inverting for the loop suppression correctly in the previous calibration.

After now incorporating the phase of the open loop transfer function and using the right magnitude, the closed and open loop spectra match at frequencies above the UGF (which was 22 kHz in this case - before adding the pomona box filter)

The script I used for generating the plot can be found here and also here with results.



Attachment 1: OLTF_20220201.pdf
Attachment 2: CalibratedNoiseSpectra.pdf
Attachment 3: CalibratedNoiseSpectra20220127_corrected.pdf
Attachment 4: attn10dB.jpeg
Attachment 5: FilterTF.pdf
  2849   Mon Jan 31 19:03:43 2022 shrutiUpdateControl Systemcavity locking loop shape - pomona box TF

[rana, shruti]

Today we measured the transfer function of the pomona box I had made earlier using a tee at the output port and splitting it between a 10kOhm resistor [to simulate the modulation input of the ITC 502 current driver] and the input of the Moku kept at 1 MOhm impedance and DC coupled. The output amplitude was 700 mV with 50 Ohm impedance. for the measurement

I used the script found in labutils to quickly fit the zero and pole to the measured TF. The pole is 45 Hz, and zero at 1.5 kHz, which is more or less what was expected.

Attachment 1: FilterTF.pdf
Attachment 2: data_and_analysis.zip
  2848   Fri Jan 28 10:38:21 2022 ranaSummaryPSOMACalculating LN2 requirements for PSOMA Cryostat

The heat capacity of aluminum is ~1 J / (g K) or 1000 J/(kg * K) in SI units. We might have about 10 kg of stuff in the chamber, so thats 10000 J * 200 K = 2 MJ.

That would correspond to a boiling off of ~10-15 L of LN2 if we go all the way to 100 K, but only half of that if we just use it to get from 300 to 200 K.

Since the Debye temperature of most common metals is ~400 K, we know the heat capacity of Aluminium will drop by a factor of ~2-3 by the time it gets to ~150K (due to the freeze out of some of the high frequency phonon modes), so 2 MJ would be an overestimate.

This assumes that the radiative heat load is negligible. If our shielding has an emissivity of 0.03, the radiative heat load from the vac chamber onto the cold stuff will be ~10 W, or ~1 MJ in 24 hours.


[Aaron, Stephen]

Aaron preliminarily estimated the load for the PSOMA experiment to be 40 MJ. The first nitrogen boil off calculation for 40 MJ of cooling energy is a LN2 volume of ~250 L, as follows:

  • The heat of vaporization of liquid nitrogen is 199 kJ/kg = .199 MJ/kg
  • The density of liquid nitrogen at atmospheric pressure is .806 kg/L
  • Volume_LN2 = Cooling_Energy / ( Heat_of_Vaporization / Density_LN2 )
  2847   Thu Jan 27 21:24:33 2022 ChrisComputingDAQauto backup to LDAS failing

Timing signals have been restored and the models on the cymac are running again.

Timing for the cryo lab CyMAC is supplied by a Silicon Labs 5340 eval board. It needs to be initialized by a Raspberry Pi. As far as I can tell, the root cause of this problem is that the Pi has not been configured to run the necessary initialization script automatically on startupsurprise. The other issue is that we have two Pis for timing in the lab (the second one controls the GPS receiver), and at first I had them mixed up.crying

The timing board connects to the Pi over the I2C bus. The command i2cdetect -y 1, on the correct Pi, confirmed a device with address 77 was present. This is an uninitialized timing board. After running the initialization script (./si5340init.py Si5340-RevD-RTSCLOCK-Registers.txt, found in /opt/rtcds/tst/x1/target/timing), the I2C address updated to 75, representing an initialized timing board. Then models on cymac1 could start and run normally.

To be fixed next time: run the init script automatically on startup, give the two Pis descriptive hostnames to make it clear which one controls what, and put the timing board and its Pi into a proper enclosure.


  2846   Thu Jan 27 14:40:19 2022 shrutiUpdateNoise BudgetMeasurement and calibration of noise spectra

Calibrating the noise budget

  1. I measured the noise spectra at the control point (after the servo) and the error point (before the servo and input attenuator) using the Moku. See these (loop, setup) diagrams for a better description. For the previous measurement (elog 2775 and elog 2776) we had chosen the ambiguous units of dBm/Hz, but this time we used the noise in dBV/rt Hz which does not require the impedance. I took separate measurements for the different frequency regions and stitched the data together. The raw noise spectra at both these points are shown in Attachment 1.
  2. Using the PDH signal on the oscilloscope I estimated the cavity pole frequency using the same process outlined in the table here. I found the cavity pole frequency to be 2.3 MHz and the cavity response to be 132 mV/Hz -- the two quantities required to estimate the cavity transfer function (modeled as a single pole low pass filter). [update 27-Jan-22: Added Attachment 3 with the crude cavity pole and response measurements]
  3. I also used the open loop transfer function measured earlier to correct for the loop suppression
  4. I used the updated noise calibration notebook that I uploaded here and the data (uploaded here).
  5. Attachment 2 shows a calibrated noise spectra using only the noise measured at the PDH error point. The red curves is the measured noise in-loop calibrated to the frequency noise at cavity input, the blue curves incorporate the inverted loop suppression factor. The lighter curves show the previously measured noise spectra without the cantilever.

All this was measured without adding the pomona box filter which would be my next step.


Attachment 1: Spectra20220127.pdf
Attachment 2: CalibratedNoiseSpectra20220127.pdf
Attachment 3: CavityMeasurements.pdf
  2845   Wed Jan 26 17:55:52 2022 StephenSummaryPSOMACalculating LN2 requirements for PSOMA Cryostat

[Aaron, Stephen]

Aaron preliminarily estimated the load for the PSOMA experiment to be 40 MJ. The first nitrogen boil off calculation for 40 MJ of cooling energy is a LN2 volume of ~250 L, as follows:

  • The heat of vaporization of liquid nitrogen is 199 kJ/kg = .199 MJ/kg
  • The density of liquid nitrogen at atmospheric pressure is .806 kg/L
  • Volume_LN2 = Cooling_Energy / ( Heat_of_Vaporization / Density_LN2 )

Thoughts on making a better estimate of current QIL Cryostat thermal load, to inform the PSOMA thermal load estimate.

  • I think Radhika and Aaron are doing some estimates based on QIL Cryostat cooldown curves, but I'll do some eyeballing below in the meanjtime.
  • Collecting cooldown parameters
    • Eyeballing a representative cooldown curve the QIL Cryostat coldhead temp seems to pass from room temp to 60K in about 2 hours.
    • Eyeballing the capacity map from Sumitomo, there seems to be an average power of 70 W for those first 2 hours, then an average power of 35 W (crossover at ~50K) for the remaining duration.
      • We have about a 36 hour conductive cooldown period to reach steadyish state at below 80K.
      • We have about a 24 hour conductive cooldown period to reach 100K (looking at Inner Shield temperature, dominated by coldplate thermal mass)
  • 36 hour cooldown to 80K steady state:
    • 2 hours at 70 W and 34 hours at 35 W gives a total cooldown load of 4.8 MJ. We'll make this the current estimate for the QIL Cryostat cooldown, with ongoing load of 0.12 MJ every hour.
    • This would require an LN2 volume of 19.4 L, with ongoing consumption of .48 L/hr.
  • 24 hour cooldown to 100K:
    • 2 hours at 70 W and 22 hours at ~40 W (higher average power than above case) gives a total cooldown load of 3.7 MJ. In the case of PSOMA measurements at 123 K, this is roughly the point at which the LN2 would cease to be used and the cryocooler would take over.
    • This would require an LN2 volume of 14.9 L.
  • 64 hour period such as a weekend:
    • 2 hours at 70W and 62 hours at 35 W would make a total cooldown load of 8.3 MJ, with those first 2 hours costing about 0.5 MJ (allowing a refill to make a dent in the required capacity.
    • The required LN2 volume to sustain through 64 hours would be 50L, and for reference the largest LN2 volume offered by IRLabs standard is 8 L.

  2843   Wed Jan 26 17:28:53 2022 shrutiUpdateControl Systemcavity locking loop shape

Some observations while measuring the noise spectra [Rana, Shruti]:

  • The cantilever mode at 45 Hz is the dominant source of noise in the measured noise spectra. Attachment 1 shows the noise spectrum at lower frequencies and Attachment 3 shows this signal in the PDH signal, and cavity transmission and reflection.
  • The contribution of the noise of the second order (90 Hz) is roughly 35 dBV smaller than the fundamental, and the third harmonic at 135 Hz seems virtually non-existent visually.
  • There are present many peaks from 1 kHz  to tens of kHz. The smaller spacing between them seems to be around the cantilever resonance. (Attachment 2). Although very similar looking peaks do seem to be present in noise spectra measured before we added the cantilever (such as in elog 2776).

In order to suppress these noisy peaks, following the suggestion in elog 2838, I made the pomona box filter shown in attachment 4. The R_{in} is the output impedance of the LB1005 servo and R_{out} = 10 k\Omega is the input impedance of the analog modulation port of the ITC502 current driver being used.

[Update 31st Jan 22] R_{in} is shown incorrectly in Attachment 4 - it is supposed to be in series, not in parallel to the remaining circuit.

Attachment 1: ctrl_pdh_5kHz_Screenshot.png
Attachment 2: ctrl_pdh_10kHz_Screenshot.png
Attachment 3: scope_red.pdf
Attachment 4: filter_red.png
Attachment 5: ctrl_pdh_100kHz_Screenshot.png
  2842   Mon Jan 24 12:37:27 2022 ChrisComputingDAQauto backup to LDAS failing

The cymac crashed. I restarted it so backups could resume. No models are running yet—there may be an issue with the timing signals.


Message from dan Kozak:

The cryo lab rsyncs are failing:

        ldas-cit# rsync rsync://cymac1.caltech.edu:2045/ldasaccess/trend/minute/
        @ERROR: max connections (4) reached -- try again later
        rsync error: error starting client-server protocol (code 5) at main.c(1658) [Receiver=3.1.3]

You might check to see if there are hung rsync processes using up the
connections.  Also, 4 is a _very_ small maximum, so you might want to
edit /etc/rsyncd.conf to have the line

        max connections = 64

Let me know when you've had a chance to sort this and I'll try again.

I've alerted ChrisW, so he can coordinate w Dan


  2841   Mon Jan 24 10:24:49 2022 shrutiUpdateControl Systemcavity locking loop shape

While I played around with locking the cavity I was not able to see the cantilever mode dominate the PDH signal at most settings. The only times I did see it was sometimes when the cavity was losing lock after being locked for a while.


Aaron and I previously saw that the 40 Hz cantilever mode was using up most of the PDH signal's linear range while in lock.

This means that the loop has just enough gain to stay locked. Increasing the gain seems to make it unstable, so we want to lower the UGF.

There were some current and temperature settings at which I was not able to lock the cavity with the same overall gain (4.40 on the knob = -6 dB which I did not change) but at the following two settings I was able to lock the cavity with the residual noise being ~30% the PDH pk-pk:

  • I ~ 114 mA, T=8.193 kOhms. PDH pk-pk was 370 mV and the locked residual pk-pk was 130 mV
  • I ~ 130 mA, T=8.356 kOhms. PDH pk-pk was 400 mV and the locked residual was 154 mV

I measured the current using the 'ANALOG OUT' port of the ITC 502 on an oscilloscope using the conversion factor of 20 mA/V.

At the second setting mentioned above I also quickly measured the open loop transfer function similar to elog 2761  and elog 2759 (shown in Attachment 1 and data saved here). Since the gain was lower this time, the UGF was a smaller value of 22 kHz with a larger phase margin.

Attachment 1: OLTF_fit.pdf
  2840   Mon Jan 17 19:15:37 2022 ranaComputingDAQauto backup to LDAS failing

Message from dan Kozak:

The cryo lab rsyncs are failing:

        ldas-cit# rsync rsync://cymac1.caltech.edu:2045/ldasaccess/trend/minute/
        @ERROR: max connections (4) reached -- try again later
        rsync error: error starting client-server protocol (code 5) at main.c(1658) [Receiver=3.1.3]

You might check to see if there are hung rsync processes using up the
connections.  Also, 4 is a _very_ small maximum, so you might want to
edit /etc/rsyncd.conf to have the line

        max connections = 64

Let me know when you've had a chance to sort this and I'll try again.

I've alerted ChrisW, so he can coordinate w Dan

  2839   Wed Jan 5 15:08:33 2022 Radhika, AaronDailyProgressPSOMAPSOMA cryostat model

The scripts for the PSOMA cryostat model can be found in mariner40/CryoEngineering/ (can be moved to somewhere more fitting). There are 2 scripts: PSOMAstatParams.ipynb (for specifiying physical system parameters) and PSOMAstatCooldownEstimation.ipynb (for simulating the model). I used the tentative parameter values and geometry found here: https://dcc.ligo.org/LIGO-D2100770.

Attachment 1 shows the resulting cooldown curves. The system is being cooled by a LN2 resevoir fixed at 77K. The model has been handed off to Aaron for testing and tweaking of parameters.

Attachment 1: PSOMAstat_cooldown_model.pdf
  2838   Mon Dec 27 05:38:56 2021 ranaUpdateControl Systemcavity locking loop shape

Aaron and I previously saw that the 40 Hz cantilever mode was using up most of the PDH signal's linear range while in lock.

This means that the loop has just enough gain to stay locked. Increasing the gain seems to make it unstable, so we want to lower the UGF.

So perhaps a 3x reduction in the UGF and a 10x increase in the gain at 40 Hz? To do this we want to shape the loop with a pole:zero boost having a 30x gain at DC relative to the high frequency part.

We can do this by making a pole zero section in a Pomona box with a pole at 40 Hz and a zero at 1200 Hz. This should be stable, since I expect our UGF is > 3 kHz.

To figure out the component values, we need to now the output impedance of the LB box and the input impedance of the current driver. Any ideas?

  2837   Fri Dec 17 18:46:55 2021 aaronSummaryPSOMAPSOMA calibration parameters, minor modifications

[rana, aaron]

We investigated further the 144 kHz oscillation. The line is not visible in the beat note power spectrum (between N and S lasers). However, when we block the beam at the REFL PD, there is a -120 dBV/rtHz peak at 144 kHz. Because the full range of the PDH signal is about 0.5 V, but the line is about 1000x lower (~40 uV), we're not too worried about it for now.

Next, we checked out the cavity lock. The gain was set too high, such that the servo frequently locked onto a weakly resonant mode, while in the primary mode the laser maintained lock but the PDH signal saturated its range. When we lowered the gain, the 41 Hz cantilever resonance appeared not only in the control signal but in the PDH error signal, indicating that the loop gain is insufficient at low frequency for the laser to track the cavity. We need to change the loop shape to give us enough phase margin to increase the gain at low frequency.

With the gain knob set to 5.08, the cavity maintained a stable lock despite insufficient low frequency gain (at least near 40 Hz). While locked, both the PDH signal and the cavity transmission had a 41 Hz oscillation. This indicates that there is a DC offset in our lock setpoint. A fluctuation at frequency f_0 in the error signal should cause an f_0^2 fluctuation in the transmission signal on resonance; off resonance, the leading order term is instead f_0.

ELOG V3.1.3-