Investigating Run 9 Luminosity

Here is a summary of some of the details of the absolute luminosity measurement I've looked into.  The SN written by Ross nicely explains the method and results.  Below are some issues I found which impact the resulting luminosity measurement. 

1)  Barrel HT DSM threshold used:

For the vernier scan runs (as well as the majority of the regular production runs) the BHT3 threshold was DSM>31, but for some of the early production configurations the threshold was DSM>25.  The code that counts BHT3+coin events in each run was supposed to account for this by assuring in software that a BHT DSM ADC was above some threshold, but the threshold used previously was DSM>30 instead of DSM>31 like what was used in the vernier scan runs.  Only ~10% of the data was taken with this lower BHT3 threshold, and the change in threshold from 30 -> 31 is about a ~20% effect for these runs.

Conclusion:  By implementing this change in the counting of BHT3+coin for L2Wrand triggers, the integrated luminosity of the sample does drop by 2% due to these production configurations with the lower DSM threshold.

2) Awayside Sum threshold used:

As described in the SN by Ross a coincidence requirement was used to remove a lot of the background from the vernier scan runs and luminosity monitoring during the standard production runs.  The awayside requirement is that the HT DSM sum opposite in phi of the TP with the highest tower (within some opening angle in number of trigger patches) is above a threshold.  In the vernier scan analysis the requirement was that this sum be >= 98, while in the luminosity monitoring code for the regular runs the requirement was that the sum be > 98.  

Conclusion: This correction to the threshold used for monitoring the lumi causes an increase in the integrated luminosity of ~5%.

3) "K" value calculation

The calculation of the "K" factor used in the fit to the vernier scan is based on the measurements of the number of ions in the accelerator measured by the WCM just previous to the vernier scan run.  The CDEV record stored in the DB contains the WCM and DCCT intensity measurements approx. every five minutes.  According to text in proceedings from the PAC meetings by Zhangbu and Angelika, the DCCT measurement is more precise and thus should be used to calibrate the WCM.  This calibration should be done "at the end of the ramp when only bunched current can be present", since "the WCM is not sensitive to de-bunched beam".  

My interpretation of this is that the WCM bunched current used in calculating the K factor should be normalized to the DCCT measurement by the total charge ratio (DCCT/WCM) from just after the ramp, as suggested by the proceedings linked above.  The table below gives the correction to the K factor associated with each beam determined from the WCM and DCCT readings from the end of the ramp (ie. first CDEV record with beam energy=250 GeV).

Conclusion:  By scaling down the K factor for each fill the integrated luminosity decreases by ~3%.

4) BTOW dead region correction

According the status tables used in the W analysis the faction of good barrel towers are smaller than the fraction of good towers used in Table III of the SN. 

Conclusion:  By using these values for epsilon_BEMC the integrated luminosity will decrease by ~1%

Summary:

Combining sections 3) and 4) above the calculated BHT3+coin cross sections from Table III of the SN should be:

Then from sections 1) and 2) using the same thresholds in the vernier scan analysis and luminosity monitoring, the total luminosity for the dataset is  L = 13.18, which is ~2% lower than the previous result (of 13.44).