Evan G., Darkhan T.
Summary:
Continuing to assess the observation Jeff made that there is a 1.8% discrepancy comparing PCALY_?X_PD_OUT_DQ, ?=(T,R) (where there should be no discrepancy), we investigated the long-term trend of the TX and RX PD values. To measure this, we grabbed 10-minute mean value trends of PCAL*_?X_PD_OUT_DQ where *=X or Y starting from September 11, 2015, for 110 days. Plotting the RX/TX for each end station, we can make the following observations
-
The H1 x-end Pcal has a very sad drift downward on the order of 20% over the 110 days.
-
In H1 y-end Pcal, there is a distinct change ocurring on Nov. 17 (a maintanence day!), from the 10 minute trends, the new epoch starts at GPS = 1131826740
-
The second epoch of H1 y-end Pcal may show signs of a two-mode Gaussian distribution, but only because this fits the histogram better than a single Gaussian function
Conclusion:
We can account for 0.25% of the discrepancy measured by Jeff. If we also consider that the Foton filter values for H1 y-end Pcal are incorrect by 1.5% from the beginning of the run due to a change in measured optical efficiency. The current calibration factors in the front-end filters use a measurement from May 2015 and ETMY was misaligned during this measurement. The current hypothesis is that the ETM alignment during the end station calibration measurements can affect the measurements, so this could have changed the amount of light reaching the receiver module. We believe this should resolve the 1.8% discrepancy measured recently. Separately evaluating the two epochs could give a better measure of systematic uncertainties (difference between front-end filter and the correct values) during the two epochs rather than a combined uncertainty.
These measurements should be repeated for L1.
Details:
We took the ratio RX/TX of the trends and then removed outlier data points that were laying far from 1.0. (There were outliers when the end station PDs are calibrated or when the ETMs are misaligned.)
Then we plotted trends and made histograms. From here forward we cannot say much about the RX/TX ratio for x-end because it is probably clipping and therefore unusable. The results below are for H1 y-end.
The change in RX/TX ratio is easily observed and we consider the two epochs separately. Each epoch histogram is fitted with a Gaussian distribution (not x-end because the trend is not stable) or, in the case of epoch 2 y-end, a two-mode Gaussian distribution. The trend of PCALY RX/TX calibrated PD output show that on Nov 17 (Tuesday maintenance day, GPS = 1131826740) an RxPD to TxPD ratio changed by about 0.23%. The discrepancy between the two photodetectors increased from 1.0125 to 1.0148.
Comparison of RX/WS, TX/WS and e (optical efficiency) of PCALY end-station measurements on Oct 13 and Dec 22, thankfully measured during the two epochs, give
Quantity D20151013 D20151222 PercentChange
_________ _________ _________ _____________
'e' 0.98933 0.989 0.033457
'TX / WS' -2.7427 -2.7382 0.16349
'RX / WS' -4.0004 -4.0028 -0.061419
'RX / TX' 1.4585 1.4618 -0.22527
from which we can conclude that mostly the change we see in the 110 days long calibrated RX / TX trend that happened on Nov 17 is mainly due to change in TX / WS response ratio, while the RX/WS response changes by 1/3 less and optical efficiency even smaller. We find that th RX/TX ratio from the calibration measurements give a change of 0.23% as well, confirming our trend results. In addition, the trend reveals the main culprit for the change is the TX/WS response. Quite possibly, the wedge splitter was bumped on the maintenance day resulting in a change in the splitting ratio.
