Test Spectra

Home Page | Test Spectra Page

Why do we need Test spectra?

In most branches of analytical science methods are checked and monitored using reference materials. When purchased these are accompanied by a certificate which specifies the concentration, with an associated uncertainty, for a number of chemical entities.These will have been analysed by at least two different methods of analysis and provide benchmarks against which methods and laboratories can be assessed. Indeed, gamma-ray spectroscopists plying their trade in the service of neutron activation analysis would utilize such reference materials routinely. The corollary for testing gamma-ray spectra would be to have available reference spectra which contained peaks at known positions containing known numbers of counts. It would be unrealistic perhaps to expect one spectrum to test all aspects of spectrum analysis but a small number of spectra should suffice. Surprisingly, although the concept is a hardly a difficult one, there are no test spectra generally available which are acceptable to users, software vendors and regulatory bodies as a whole.

The ideal test spectrum would have both large peaks and small peaks on both high and low continuum backgrounds to test the ability of the software to find and measure accurately peaks in a range of spectrum environments without reporting spurious peaks. There would need to be a test for doublet resolution, with a range of peak separations and peak ratios. In order to be assured that the analysis program would not be confused by spectral artifacts such as backscatter peaks and Compton edges, these should also figure somewhere within the spectra.

In principle, all calculations of radioactivity reduce to a comparison between standard and sample peak areas. This might be directly, as in neutron activation analysis, or indirectly via an efficiency calibration curve. It follows, therefore, that an analysis program need not produce absolute peak area estimates as long as the program is consistent and the area estimate is indeed proportional to the true peak area. If an analysis program gives areas which are consistently 5% low then the efficiency calibration derived using the program will be 5% low; the error will cancel out when the final calculations are performed. A set of test spectra should, therefore, include a standard or reference spectrum containing well defined peaks to which peaks in all other test spectra can be related. This will best simulate the actual manner in which spectra are used.

At the present time, the only easily available test spectra are those produced by the I.A.E.A. and distributed in 1976 as part of a blind intercomparison exercise, now available as reference spectra. The International Electrotechnical Commission (IEC) did, at one point, seek to produce a series of spectra. As far as I am aware only prototype spectra were generated and these are not generally available. The spectra used by Sanderson for his 1992 re-evaluation of software are also available. Menno Blauww, on behalf of the IAEA, has also produced as eries of spectra fullfilling most of the requirements above. Links from the Test Spectra Home Page discuss all of these spectra in detail.

There is, of course, the option of creating your own test spectra. SpecMaker, available on this site, will allow you to do just that.

All files on this site available for downloading are in the public domain, or distributed by permission of the copyright holders.

What about the rest of the program?

The spectra I discussed above are intended to test only the peak search and measurement parts of the spectrum analysis program. There is a need to check the remainder of the program. There are three options - all of which should be taken.

1) In fact, it is relatively easy to do that once an understanding of the operation of the program is gained. Simple spreadsheets could be used to confirm that, given particular peak areas, the program generated the correct activity value and uncertainties. In fact, doing that is a very valuable exercise - it is not uncommon to find out that the program manual is 'being economical with the truth'!

2) Participate in Intercomparison Exercises and analysis Certified Reference Material. If you can consistently get the 'right answer', you can have some confidence that, whatever the detail of the peak measurement, your system is satisfactory. Remember, if you don't get the right answers, you MUST find out why and find a way of correcting your analysis.

3) There are two particular sets of test spectra that have been created to allow spectrum analysis programs to be checked from start to finish. These are analogous to the chemical reference materials refered to above. Spectra have been produced by Menno Blauww, again for the IAEA, and NPL in the UK

Appendices Data Example Spectra Links Online Course Spreadsheets Test Spectra

This website is sponsored by Nuclear Training Services Ltd. and created by NTSNet
Copyright © 2007 Dr Gordon Gilmore. Problems with this page? Contact