How to check the crystal spectra

Check both the Hitmap and the Zoomed Hitmap for EMC hardware showing up. If there is something striking, this may require actions from powercycling an AOB powersupply up to opening the detector to replace something.

AOB rectangle of 3x4 crystals

IOC 6 horizontally adjacent AOBs

crate 10 vertically adjacent IOCs
Check for extremely bad crystals. These may need to be masked, if persistent for a couple of days. They are the ones shown in red or blue. You can see the actual spectra by clicking "4" or the "z" right next to it.

Information to the plots

You may find "background information" useful.

Hitmaps
The hitmaps need about 6 Million events to look nice. Otherwise the spectra become rocky and dips/peaks are all over the place. During normal operation it takes about 3 days to get fresh a 6 Million, this is why a sliding window technique is used. Masked channels may fade out or in. The number of events entering the hitmap is not displayed currently.
The badness of a crystal is calculated by combining its severity, cutoff deviation, number of zero dips and the length of the longest zero dip.
Severity
The dip/peak finder calculates a "severity" for each crystal. The higher its value the worse the spectrum.
Cutoff Deviation
This plot shows the deviation of the cutoff of the spectrum from the ring average in terms of standard deviations. This makes no sense for zoomed spectra.
Number of Zero Dips A zero dip is one (or more connected) bins which have no entries. The plot shows how many of such dips a spectrum has.
Length of Zero Dips
The length of the longest Zero Dip of the spectrum in terms of bins. Zoomed spectra have 100 bins, unzoomed have 200.
Example Spectra (0z 1z 2z 3z 4z)
For each badness up to 12 example spectra are shown. Badness 0 means good spectra. The examples are the first 12 spectra of given badness that come across if you run through the hitmaps like (1,0) to (79,0) to (2,0) to (2,79) and so on. You can check here if the fit produces reasonable results.

Background Information

The goal of this monitoring tool is to point out unusual features of the EMC crystal energy spectra. These spectra come in two classes, unzoomed (0-10GeV) and zoomed (0-100MeV), take a look at the example spectra.

Unusual features are mainly dips and peaks, which can occur for example if the 4 different ranges of the CARE chip are not calibrated properly. At second hand it is possible that the whole crystal is miscalibrated, in that case the spectrum cutoff deviates from the ring cutoff. At last, there can be bins which have exactly zero entries which are sort of a special case of the usual dips.

The crystal spectra are analyzed by fitting an empirical PDF to them. When the fit is done, the algorhythm looks for bins that deviate significantly from the fitted curve. These are considered dips and peaks. Since zero dips are hard to be treated by the fit (they don't enter the likelihood), it is searched for them in a second step. The ring cutoff is calculated in a straight forward way.

Each analysis provides quality parameters which are combined to a rough measure of the "badness" of a crystal. The badness ranges from 0 (=good) to 4 (=extremely bad).

Although this procedure is fairly automated right now, a large amount of tuning of the parameters was necessary. For instance, the fit highly depends on its start parameters. So it may happen that the fit fails and therefore a whole lot of (fake) dips/peaks are found. If this is persistent over a couple of days, the parameters may be adjusted by me (moritz.karbach@slac.stanford.edu).