Goal: Use mips from pp minbias events to set absolute gains for EEMC towers, SMD strips, and all pre- and post-shower tiles for the 2006 run.

Basic Method:

1. Use 1.8M pp minbias events from day 89 - runs 7089008, 7089013, 7089014, 7089015, 7089021, 7089022, 7089024, 7089025, 7089091, 7090008
2. From raw (ADC-ped) spectra, identify dead or "problem" channels in all detector components, mask these out (OR'd with EEMC status tables)
3. Fit slopes of ungated SMD histograms: 3-4% statistical error in slopes ® sets strip-to-strip relative gains to ~6% accuracy
4. For 1st pass: estimate absolute SMD strip gains by global matching to last year (needed ~10% increase in conversion factor from 2005 Cu-Cu to 2006 pp minbias)
5. Identify mips event-by-event via (000xx000 || 000x000) hit signature in both planes, ADC above ped in all other layers ® "mip ID"
6. Fit Gaussian/Landau function to towers and pre/post tiles gated by mip ID, use known geometry / energy loss to set absolute gains (see Jan's note for details)

   Compare 2006 absolute tower gains to those obtained using same method in 2005, after correcting for known HV changes

   Compare 2006 tower gains to slopes measured in 2007 for Au-Au, within each eta bin

7. SMD 2nd pass: sum sets of 12 consecutive SMD strips gated by mip ID to set absolute gains, check for eta dependence
8. Check that with final SMD and pre/post shower gains, deduced tower gains are stable

A)  Dead / unusable channels

The total number of channels that needed to be masked out for this analysis, and for which no gains were determined, was very small.  These channels are either not included in the EEMC gains database (for towers) or have gains and errors entries of -2 and 0 (for all MAPMT / PIX channels).  A list of the SMD channels that were masked out in the first-pass analysis can be viewed here.  After recovering many of the low-statistics strips, the final list of all channels (towers, pre/post, SMD) that could not be analyzed, and a brief description of the problem encountered, can be found here.

B)  Relative yields of SMD strips, using slopes

Most strip spectra are very clean.  Typical spectrum is shown below:

ADC pedestal and single photoelectron peak are clearly visible.  Determination of slope was fairly insensitive to choice of fitting range ® channels 40-140 used here.

Complete sets of fitted spectra, and tables of individual strip gains and errors, are available here.  From checking plots by eye, searching gain tables for anomalies, we conclude that only 17 strips out of ~7000 are not usable, beyond the 40 'known' unconnected strips (16 associated with `notches,' 24 from strip number 288).  Using 1st-pass gains, no masking, most sectors look very good in terms of uniformity:

C)  1st pass SMD - absolute gains

By comparing to last year's final SMD results, after absolute gains had been set using the technique described on Jan's web page (see items 11 and 12), we find a systematic difference of about 10% in the slopes obtained.  Given that the 2005 results were based on Cu-Cu minbias events, while in 2006 we are using pp minbias (different occupancies, different luminosities), this does not seem unreasonable.  To account for this, we have simply increased the (empirically determined) ADC ® energy conversion factor by 10%. Comparison of database gains for 2006 to those for 2005, for two typical planes:

Actual 2006 gains for plane 05U (to show range of strip-to-strip variations)

Note:  These intermediate gain values should generally not be used for analysis, now that final SMD gains for 2006 are completed!  The 1st pass values are loaded in the EEMC database with a 'flavor' of  slope2006 ; if used, log file should contain entries of  "prelim, invSlope from day89, absScale match to 2005, for 2006 pp, Scott1".  For final gain values (see discussion below), flavor = ofl, and log entry should be   "final 2006 p+p, using MIP UxV 7vetoXX7veto, from day 89, Scott".

With SMD gains set at roughly 10% level, and with problem channels (SMD strips, towers, and pre/post layers) masked out, check for events with a "mip ID".  In each of the 12 sectors:

• search for all strips with deposited energy < 0.5 MeV
• require 1 or 2 adjacent strips to fire, with at least 7 non-firing strips on each side (e.g. 0000000xx0000000) ® "mip candidate"
• form intersections of mip candidates between the U and V planes, test if this position is inside tower fiducial cut (<0.7 cm from tower boundary)
• if so:  for each of the four 'layers' (pre1, pre2, tower, post), increment histogram if the other three layers are all above pedestal ® "mip ID"

D)  Absolute pre- and post-shower gains

Because passage of a minimum ionizing particle will result in only a small number of photoelectrons (~3) being collected for these components, spectra were fit using a Landau distribution.  Absolute gains were determined assuming that the calculated energy deposition would correspond to an ADC value of 1.46 x MPV (most probable value) found in Landau fit.

Complete sets of histograms and fits, and tables of all final gains as loaded into the database, can be accessed below:

preshower-1	spectra and fits	gain tables
preshower-2	spectra and fits	gain tables
postshower	spectra and fits	gain tables

E)  Absolute tower gains

To determine the tower gains as accurately as possible, more care was taken in the fitting and QA procedures, relative to those used for the pre- and post-shower layers. Specifically:

• Fits to the raw tower spectra were used to determine pedestal 'shapes' for each tower, which were then scaled and subtracted from the mip-gated spectra
• The fitting function used was a combination Gaussian + Landau distribution, chosen to best reproduce the observed tower response
• A more sophisticated algorithm, reflecting the previous two items, was used to estimate the uncertainties in the fit parameters
• To recover as much useful information as possible, several towers received individual attention ...
• Three towers (04TB07, 06TA07, 08TB07) are known to have their '4'-bit stuck on, so these histograms were binned over four consecutive ADC channels prior to being fit
• One tower (08TC11) had a 'dead' post-shower layer.  Though masked out, this can distort the spectral shape, so slope comparisons to neighboring towers were used to set the gain
• Four towers could not be recovered at all - see details here
• Every fit was checked by eye, and the fit parameters scanned / sorted to identify additional anomalies

Complete sets of histograms and fits, and final gain tables, can be accessed below:

EEMC towers

spectra and fits

gain tables

The absolute gains deduced via these procedures for all 720 towers are compared to their "ideal" values (E_T = 60 GeV) for each eta bin in the figure below:

F)  Absolute tower gains - comparison to other results

• Change minimum # of smd strips required to not fire from 7 to 11 in definition of "mip candidate" (see above), study effect on tower gains
• Compare current gains to those extracted from similar mip analysis for Run 5, after accounting for known HV changes
• Compare current gains to tower spectra slope analysis for Run 7, within each eta bin