QA/QC for NaI Gamma Spec -Reply

[FRAMEP@ORAU.GOV]

Tim Paul wrote in part:

>We're in the process of updating our QA/QC procedures for our NaI
>based counting system.... Recently, we've had a suggestion to
>monitor dead-time and state actions to be taken if dead time is too
>high.  The trouble is, I'm not sure where to set the action level or
>even a good way to determine this action level. Any help you could
>provide in this area would be appreciated.

A few thoughts rather than a clear answer: the system's MCA estimates
dead time and corrects for it. The problem is that this estimate has
an error, and the larger the dead time, the higher the estimate's
error. On top of this error might be the errors asociated with the
pulse pile up that occurs at high count rates because of random
summing. 

I have never seen this written down, but the rule of thumb that I'm
familiar with is to keep the dead time under 10%. In fact, I've heard
of one lab where they intentionally keep the dead time around 5%. A
means to evaluate this type of error, more or less, has been
described in  "Tests to Measure the Performance of a Germanium Gamma
Ray Spectrometer and its Analysis Software" by Gehrke in
Radioactivity and Radiochemistry Vol 1, No 2, 1990 and ANSI N42.14
1978 "Calibration and Usage of Germanium Detectors for Measurement of
Gamma ray Emission of Radionuclides". Although these are written
specifically for germanium systems, the technique should be
applicable to NaI systems too. The ANSI document was updated in 1991.
Although I haven't seen the update, I'd bet it decsribes essentially
the same thing:

A medium activity Co-60 source is counted long enough to get over
100,000 counts in the 1332 peak.  Then a hot low-energy gamma emitter
(<400 keV eg Cd-109/C0-57) is placed on top of the Co-60. The count
is repeated, this time the total countrate and dead time is up. The
difference in the 1332 counts reflects the error introduced by the
higher dead time, and pileup.  

It might also be possible to estimate the dead time error in the
following way: count long enough to get 100,000 counts in a peak,
Cs-137's 662 keV would do. The count should have plenty of dead time
and the dead time should be due to the presence of at least one
other, high activity, gamma emitter( eg Co-60). Then increase the LLD
and/or decrease the ULD so that the MCA is ignoring some counts
outside the peak of interest. This should drop the dead time.  Do
this repeatedly to get the peak areas at a variety of different dead
times. There won't be any gain shifts during the test and the changes
in counts reflect dead time estimation errors only i.e. pileup errors
are not an issue. 

When analyzing samples, it would usually be preferable to move the
source further away than play with the discriminators or other stuff
if the dead time is too high. Hope this helps a little.

Paul Frame
Professional Training Programs
ORISE
framep@orau.gov