Update: June-17-2010

Rewrote the masking code and I have it flagging bad strips for three reasons: Not fibered or Zero counts under fitting area, Too few counts, and No gain fit at all. All the not fibered strips match up with those that Jan has listed in his 2006 mask. It was also able to find some that Jan did not note, and with inspection of the histograms, these strips are dead. The too few counts flag is for all the fits that are fitting with less then 90 counts. This number was choosen because Jan noted some strips with low counts and the highest number of counts that was marked as 'low' was about 90. This threshold can change if need be. The no gain fit flag was added for any strip that had counts above threshold but the fit did not work. These strip histograms will most likely need visual inspection.

The attached files have been updated with the newest results and code

June-15-2010

I was able to make some code that will make a mask file in the format that Jan's document has(attached). The one part that needs to really be tuned in is what we consider a 'bad' strip. Currently I have the code taking in all the percent errors for a section/plane, getting the mean percent error, finding the standard deviation of this mean percent, and then masking out the strips that drift to far from this value(currently set at three sigma). The issue with this method is that all the higher number strips get excluded because they, on average, have percent errors 7% and above. If I make the window to big everything gets counted as a good strip and only the ones with a zero gain/error are filtered out.

1Q: Is this an ok method for finding bad strips? If not, what is a better method?

Partial A to 1Q: A better method maybe to fit a function to a percent error vs. gain number graph. Then use the fitting parameters and errors that the fit gives us to judge how far off from the fit funtion a point is and then count it as 'bad' if its too far off. This method maybe totally off base, but im not too sure how else to judge if a strip is bad above looking at its output by hand.

2Q:Is it possible to constuct these masking files like this at all? or will we need to look at each section by had and make this like Jan did?
(Personaly I would perfer to have the code to it for us)

I have attached a few things to this post. I have Jan's mask file from 2006(smdAllMaskDay89v1.txt), my mask file that my code creates(smdMaskFile2010.txt), and the code files that I used to do this(MaskFileCode.zip).