RE: TLD Background subtraction

[Bob Flood <bflood@SLAC.Stanford.EDU>]

Scottie's ideas work because they are based on a very common characteristic
- most of the people we monitor don't get any dose. If you badge 1,000
people and 200 of them show a non-zero dose (above your reporting
threshold), then you have about 800 people helping you monitor background.

Think of it this way: if you can run your normal processing calculations to
yield readings WITHOUT BACKGROUND SUBTRACTED (this is very important) and
then plot these data in a histogram, the graph should look very much like a
normal (Gaussian) distribution except there should be a "bulge" on the
higher side of the distribution because some folks actually do get dose,
and the high side should tail off much farther away from the centroid of
the distribution because of real occupational doses. But the centroid of
the distribution should equal the average background for the monitoring
period, and the width of the distribution will give you a visual indication
of your system's background measurment variability. By the way, if you
cannot produce the data without background subtracted but your system shows
both positive and negative net values, the same graph can be done, but the
centroid will be centered on zero if your background subtraction is
accurate and will be somewhere else if it isn't.

If you cannot conveniently plot the data but can examine net
background-corrected values in a report, you should see about as many
negative numbers as positive numbers when background is handled accurately.
A preponderance of one or the other is in indication of a potential problem.

---------
Bob Flood
Dosimetry Group Leader
Stanford Linear Accelerator Center
(650) 926-3793
bflood@slac.stanford.edu

************************************************************************
The RADSAFE Frequently Asked Questions list, archives and subscription
information can be accessed at http://www.ehs.uiuc.edu/~rad/radsafe.html