jeffrey.kissel@LIGO.ORG - posted 22:31, Wednesday 07 August 2019 - last comment - 12:57, Friday 09 August 2019(51115)
Calibration Measurements: Entire Suite Complete; Actuation Same as Expected, Sensing Same as Last Week
J. Kissel
After many hours of waiting for recovery from the TCSX laser trip, and then an Earthquake, and then remaing stuggles with global alignment vs. the lock acquisition sequence, I stubbornly stuck around until I was able to complete all of this weeks standard calibration measurements.
I attach the processed results, but I haven't have much time to think about them.
From what I can initially see:
- 2019-08-08_H1_sensingFunction_mcmcModel_vs_measurement.pdf
The low frequency shape of the sensing function is quite the same as it was last week, now much more resembling a physically detuned pro-spring.
- H1_sensingFunction_PCALXvsPCALY_referenceModel_vs_allMeasurements.pdf
The low frequency response is still visible in each PCALX and PCALY in the same fashion, indicating it's still a feature of the DARM Open Loop Gain.
- 2019-08-08_H1_sensingFunction_mcmcModel_paramCornerPlot.pdf
Even though the data now more matches the model, the MCMC still has trouble fitting the data because of the high Q of the feature coupled with too few data points and the covariance between things like optical gain and optical spring parameters. You can see this reflected in the MCMC parameter corner plot, which are full of "islands" of local minima.
- 2019-08-08_MCMCTDCFs_vs_CommishEvents.pdf
That means the trend plot of MCMC parameters should be taken with a grain of salt.
- The actuators remain quite identical to the reference measurement, which is excellent -- at least we don't have to worry about that too.
Yet again, due to IFO recovery, I was not able to get any further exploratory measurements to map out the coupling of all the different knobs we have found, or suspect are causing this feature (spot position, dhard WFS gain, src asc offsets, SR3 disc heater, ITM C02 laser heating, etc.) in the face of the new global alignment position we created on July 31 / Aug 1 2019.
However, with statistics of 2, we can at least say that the feature is now relatively consistent... maybe.
The data lives here:
Sensing Function:
/ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O3/H1/Measurements/FullIFOSensingTFs/
2019-08-08_H1_DARM_OLGTF_LF_SS_5to1100Hz_15min.xml
2019-08-08_H1_PCALY2DARMTF_LF_SS_5t1100Hz_10min.xml
2019-08-08_H1_PCALX2DARMTF_LF_SS_5t1100Hz_10min.xml
2019-08-08_H1_PCALX2DARMTF_BB.xml
2019-08-08_H1_PCALY2DARMTF_BB.xml
2019-08-08_H1_OMCDCPDSUM_to_DARMIN1.xml
Actuation Function:
/ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O3/H1/Measurements/FullIFOActuationTFs/
2019-08-08_H1SUSETMX_L1_iEXC2DARM_10min.xml
2019-08-08_H1SUSETMX_L1_PCAL2DARM_8min.xml
2019-08-08_H1SUSETMX_L2_iEXC2DARM_12min.xml
2019-08-08_H1SUSETMX_L2_PCAL2DARM_6min.xml
2019-08-08_H1SUSETMX_L3_iEXC2DARM_12min.xml
2019-08-08_H1SUSETMX_L3_PCAL2DARM_6min.xml
Processing scripts live here:
Sensing Function
/ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O3/H1/Scripts/FullIFOSensingTFs/
process_sensingmeas_20190808.py
plotMCMC_vs_GDSTDCFs.py
process_sensingmeas_manydates_wPCALX.py
Actuation Function
/ligo/svncommon/CalSVN/aligocalibration/trunk/Runs/O3/H1/Scripts/FullIFOActuationTFs/
process_actuationmeas_20190808.py
with
/ligo/svncommon/CalSVN/aligocalibration/trunk/Common/pyDARM/src/
actuation.py rev 8129
sensing.py rev 8144
Non-image files attached to this report
Comments related to this report
I'm not confident that using data below 30 Hz is the correct approach for fitting an accurate optical gain and cavity pole frequency, two parameters we think that the model actually fits correctly for and is able to track well. If I restrict the MCMC fit to the frequency range [30,5000] Hz, then we find different optical plant parameters, albeit at the cost of poor fit at low frequencies. We are not confident in the model anyway at low frequencies. Fortunately optical plant uncertainty is not a major contributor to response function uncertainty at low frequency. Below are the MCMC corner plot and model versus measurement and residuals. We can see that the residuals are much flatter and are minimized in the region of interest for frequencies greater than 30 Hz. Fit parameters are: Parameter | Quantiles (0.15, 0.50, 0.84) --------------------------------------------------------------------- Optical gain, H_c (ct/m) | 3.114e+06, 3.116e+06, 3.118e+06 Cavity pole, f_cc (Hz) | 398.9, 399.8, 400.6 Detuned SRC spring frequency, f_s (Hz) | 6.256, 6.398, 6.528 Detuned SRC spring quality factor, Q_s | 97.7, 91.55, 80.08 Residual time delay, tau_c (usec) | 0.4081, 0.9149, 1.39 ------------------------------- OR ---------------------------------- Optical gain, H_c (ct/m) | 3.116e+06 (+2464,-2273) or (+0.07908%,-0.07294%) Optical gain, H_c (mA/pm) | 4.16 (+0.00329,-0.003035) or (+0.07908%,-0.07294%) Cavity pole, f_cc (Hz) | 399.8 (+0.8064,-0.8524) or (+0.2017%,-0.2132%) Detuned SRC spring frequency, f_s (Hz) | 6.398 (+0.1299,-0.1421) or (+2.031%,-2.221%) Detuned SRC spring quality factor, Q_s | 91.55 (+639.2,-1455) or (+14.32%,-6.292%) Residual time delay, tau_c (usec) | 0.9149 (+0.4754,-0.5068) or (+51.96%,-55.39%) Optical gain is higher, cavity pole frequency is lower compared with attempting to poorly measure the low frequency deviation to our model.
Non-image files attached to this comment