Presented at the 46th Annual meeting of the INMM, Phoenix, AZ, July 10-14, 2005 |
Stephen Croft1, Eloisa Alvarez2, Debbie Curtis2, Bob D McElroy1, Colin G Wilkins2, Malcolm R Wormald2 and Brian Young1
1Canberra Industries, Inc., 800 Research Parkway, Meriden, CT, 06450. USA.
2Canberra Harwell Ltd., Building 528.10 Unit 1, Harwell International Business Centre,
Didcot, Oxfordshire, OX11 0TA, UK.
ABSTRACT
We describe the method applied within Canberra’s waste assay software,
NDA2000, for estimating the Minimum Detectable Activity (MDA) following a Passive
Neutron Coincidence Counter Assay. In order to clarify the approach and aid
the explanation we illustrate its use by giving numerical examples. The MDA
algorithm shows explicitly how design and operational parameters can be used
to influence the detection limit. This enables the key factors to be identified
and balanced in a structured way, when establishing the compromises between
often conflicting requirements at the start of instrument design or measurement
programs that target a given performance. We extend the discussion beyond the
usual confines of ideal statistical counting distributions. Finally, we briefly
comment on the role of Advanced Waste Analysis (AWA) algorithms available within
NDA2000 involving statistical filtering and multiplicity truncation in relation
to stabilizing the MDA estimates and overall system performance in real-life
measurement situations.
INTRODUCTION
Fundamental to the design and application of non destructive assay instruments
for the assay of nuclear materials is the question of how small a quantity
can be reliably detected (in this context activity or mass is implicitly used
interchangeably - the two being simply related). The term “Minimum Detectable
Activity”, or simply MDA, is most commonly used to refer to this quantity.
The term “Lower Limit of Detection” or LLD, or depending on context “Minimum
Detectable Concentration”, MDC may also be used. The MDA is often used
as one of the many Figures of Merit (FoM) by which different systems may be
compared. To be useful in this sense, however, it must be suitably defined
in terms of probability theory and quoted under standard measurement conditions
- both of which must be carefully stated. To ensure a fair comparison between
counter designs, all of the underlying assumptions must be clearly stated.
In this work we discuss passive detection limits in passive neutron coincidence
counting. Extension to totals neutron counting is trivial. Active detection
limits are more involved as has been discussed elsewhere [1-3] yet many of
the points we wish to make are generic to all methods.
