Home

Intercomparison of software packages

1. M M Default report Yes Yes many Yes Yes Yes Yes Yes Yes Yes very Yes very Yes Yes options complete complete Can the user define No Yes No Yes No No Yes No AFulloran Yes with many Yes No Yes with many the analysis report Abbreviated different different type of reports options options Even are offered modules But the units to be options are reported can be predefined and standardise can not be changed Possibilities on the Print or Print or saved Print or Print or Print or ASCII Print and Printed or ASCII or Print and or Print or Printed ASCII Print to Output of the results ASCII into a file saving intoa ASCII file An file easy to ASCII files saved into a Sampo90 save and savedintoa Only file orto the ASCII format file ASCII option for MS import Several file Both format The ASCII separated ASCII screen only 15 provided as format Word file ASCII files propnetary short report 15 by commas format also mode no creation ofa well possible type 15 also are kept as format CAM very suitable ideally for provided ASCII database is available track keepers files of 1mporting into importing them format possible as well ASCII any spreadsheet into spreadsheet format Easy to export 86 Others ActAn Gamma Track
2. 3 2 The measuring set cce dii ed ua m mE 3 3 The measured spectra sce Up 3 4 Observed properties of the measured spectra 3 5 Manipulations of the measured spectra 22 2 1 1 3 6 Determination of the reference peak positions and areas 4 THE METHOD b E ERU 4 1 Data handling SECO ere ee rrer ere ree eee er ere rr ere eer eee rr irre reer reese eee ree ese Tere ree ee re eee reer ees 4 2 Peak area re normalization a nn 4 3 Computation of standardized residuals and reduced sums of squares CHAPTER 5 WHAT DO THE NUMBERS MEAN a CHAPTER 6 THE CANDIDATES Version 2 6 2 GammaTrack version 1 3 Release 2 6 3 Gamma Plus version 1 02 0 20 400 0000 6 4 Gamma Vision version 2 3 6 5 Gamma W version 17 08 6 6 Ganaas version 3 So red
3. ITT On line help Yes Yes Yes No Yes No Yes No Yes Yes Yes but No very limited Availability of Yes but not Yes Yes partially No No No Yes Yes Yes No No context sensitivity complete help Is there an updated There s some Yes some Yes Manual Yes Yes well Yes User s Yes Very Yes Yes updated Yes well Yes Yes Poorly and efficient User s documentatio parts are 1s complete updated Manual Not extensive and and well wniten and Sometimes implemented Manual n but not a inconsistent and update updated well presented presented and efficient not very User s but not very written clear Manual well written Some guides are misleading or confuse Is there any other type Examples No Yes the one No Examples No Yes The Yes samples No Yes some Yes some of documentation files describing files isotope files Different example example files provided Nuclide library is very application files Navigator limited specific data libranes workbook on gamma spectrometry can be provided upon request Telephone help line E mail Not known Not known Yes Yes Telephone Yes E mail Yes Telephone Not known Yes Telephone Yes but No Not known or Internet Telephone help line and E and FTP help line help line E depending of address including a help line mail address for Internet E mail mail and WWW the dealer WWW location for WWW site exchange of and WWW site site help info Comments The Ganaas software provides an E mai
4. g user settings OSQ g default settings ActAn 075 1 125 15 175 2 225 01 1 10 100 1000 hits addin1 hits straight reduced sum of squares 4 Figure 35 Resolving power of the analysis Figure 36 X values related to false hits programs indicating deconvolution correctness match in peak position was extended from 0 2 x FWHM to 12x FWHM covering the range of the doublet separations in the spectrum From Figure 34 where some results for one of the analysis programs are shown t is clear how crucial this criterion 15 even though it s an extreme case At high values the peak position uncertainties reported by the analysis program are no longer relevant unrecognized doublets match both components in the reference list and as described in Chapter 4 these components are merged so that total doublet areas are compared To establish how well the analysis programs could distinguish doublets from singlets their resolving power was defined as the number of high peak hits in the addini spectrum divided by the same the straight spectrum For this ratio 2 would be a perfect score The x7 value for false hits in the 111 spectrum indicates that doublets are being mistaken for multiplets To eliminate the effect of too small reported position uncertainties the data were determined using a 10x FWHM matching criterion The results are shown in Figure 35 and Figure 36 An analysis program that performs residual searches in the fitti
5. 19 63 88 26 3 8 22 21 20 69 3 1 22 39 120 14 1166 45 14 33 9 8 1 4 14 1161 33 1 3 33 73 In smallonlow 15 E 15 14 14 14 17 14 16 15 11 1 2 0 9 07 12 28 20 12 2 3 0 9 X 12 22 1 1 09 22 54 44135 3 6 10 all matches N 81 76 77 82 77 105 96 101 89 80 14 23 poe 15 1 4 1 4 2 6 30 5 275 212 2 6 1 4 2 15 50 1784 9 8 16 4 1 5 1 4 5 2 56 5 9 3 9 0 4 6 14 pum X mee er annihilation misses 1236 251 209 172 87 93 211 190 526 199 167 5 6 s 1994 418 811 21 6 28 5 6 5 5 71 7 23 3 88 7 7 false hits j 10 15 13 20 12 total 1179 185 19 289 337 292 256 182 190 ns 293 337 1292 1257 3 8 103 5 170 4 32 9 64 8 20 7 5 8 37 104 6 56 1 174 224 Z Z gt lt 2 2 9L Table 5 GammaTrack Default settings La ast dap oai oie T qe Tm high peak 55 oe p 46 58 57 56 I Yi 23 13 7 75 6 15 7 51 7 12 3 2 25 156 3 253 55 1 is 1019 ies at 522 2 21 small on high 3 9 5 6 P F yt 5 25 a d 25 small on low ra 25 m 4 m m ds d 7 3 236 212 gt S d as a a a all matches da pr m T pm a is 140 1 Pr 53 0 ru E x 82 3 192 1 55 5 52 4 d T buc ic pum annihilation Ee e re e e e e a misses 118 258 254 268 251 196 111
6. QQ t9 O 9 2 oooooooo Figure 8 Section of z score table showing a miss at 2112 5 keV a reasonable doublet fit at 2120 0 and 2121 2 keV and an erroneous doublet fit at 2205 6 en 2206 8 keV resulting in a false hit at 2207 3 keV Z scores or standardized residuals were computed as described above for each peak 1n each test for each type of spectra section from the comparison program CMPSPEC output 15 shown in Figure 8 From the Z scores reduced sums of squares 7 were computed for different categories of peaks If no re normalization had been performed the 7 for any category was computed as the sum of the squares of the z scores in the corresponding category divided by the number of peaks in that 15 category If re normalization had been applied the number of peaks in the category minus one was used in the division for the hit categories The categories and their definitions were 1 Annihilation peak Any peak closer than 3 keV to 511 keV 2 High peaks non annihilation hits for which the ratio of reference peak area and reference peak uncertainty 1s larger than 10 1 4 2060 gt 10 3 Small peaks on high continuum and small peaks on low continuum Hits for which the peak area was less more than 0 55 respectively where 15 the area of the continuum under the peak Knowing that the reference uncertaintie
7. Counts of ig Verrem ait L X arm N 30009 3252 2500 3760 4000 Numer Figure 4 detailed section of spectrum 500000 containing a doublet peak test spectra were generated from this same type of Ra spectra as shown in Figure 1 to 4 through adding after shifting see below The only spectrum with different characteristics of the peaks is spectrum SD0034 MCA where a very strong source of was added in order to deteriorate the peak shapes of the peaks through random pileup In Figure 5 the same section as in Figure 4 is shown from the distorted peak shape spectrum SD0034 MCA for comparison ju D D 0 ct 3 1 N lt 1 k n Figure 5 A detailed section of spectrum SD034 in which peak shapes are distorted due to random coincidences with low energy photons 3 4 Observed properties of the measured spectra The first series of spectra SD0000 MCA through SD0019 MCA was measured in order to test the stability of the system and to generate reference data which may be used for later intercomparison Several peaks out of the 20 spectra were evaluated and the peak positions as well as FWHM values were determined in order to test for the stability of the system In Figure 6 we show the positions of a peak ca 1900 counts peak area in the 20 spectra and Figure 7 displays the fitted FWHM values for the s
8. X35 2 enn Figure 16 Perfect deconvolution of the multiplet in the spectrum 10n1 High peak on the left side of a small peak with ratios 10 to 1 In the following figure y is equal to 1 74 but it is not understandable what is the meaning of the 4 SIGMA value 3 distort log REGION 10 i LEFT 726 CH 1 k RIGHT 744 CH TRY 1 CHSQ 6 110724 4xSIGMA LIMIT inc 2 885618 CHSQ C4S1G Figure 17 bad background shape obtained during the automatic processing of the distorted spectrum 37 straight asc j H straight log REGION 4 i eet 727 CH RIGHT 747 CH TRY i j CHSQ 2 342663 j i 4X SIGMA LIMIT i 2 885618 CHSQ 4xSIG JEEP 0 811841 1 2 S o fpes BES SL 3295 f Figure 18 A very nice deconvolution of the closed doublet at 294 295 channels in the STRAIGHT spectra 6 9 OSQ Professional version 6 3 release 1 by APTEC Engineering Ltd OSQ Professional 15 a MS Windows based gamma ray spectrum analysis program developed the basis of a MCA emulation software As a data acquisition device it works only with Aptec cards It comes a single 3 5 diskette and six different types of Manuals Upon request well written and comprehensive Gamma ray Spectroscopy Workbook can be provided as well 0 The installation is simple and efficie
9. software does not have any on line help Routines analysis in batch mode are possible even with pre defined set of operational parameters for each batch A separate optional software WinnerScan be provided to work in conjunction with Winner Gamma for automatic scanning of radioisotope activities or peak areas vs time 46 CHAPTER 7 OVERVIEW In this Chapter an overview of the possibilities and available options for each program are summarised They are organised in areas of interest for different end users Some comments are given below each table More detailed information on the particular aspects of each program 1s given in the previous Chapter 47 Hardware ActAn Gamma Gamma Gamma GammaW Ganaas GeniePC Hypermet OSQ Sampo90 Span Winner Track Plus Viston PC Professional Gamma ER IgM A NP AT a AI I SS TT Pm d Minimum CPU 80286 80286 80386 80386 8088 80286 80386 80386 80386 80286 80286 80386 required C oprocessor Yes Yes Yes Not Not but Not but Yes Yes Yes Yes Yes Not but required recommended recommended recommended Minimum disk 1 Mbyte 3 Mbytes 2 33 Mbytes 5 Mbytes 2 Mbytes 3 Mbytes Vanable 500 Kb 213Mb without 7 Mbytes 500 Kb for 2 Mbytes space examples depending on programs 300 samples spectra the software the options to Kb examples 500 Kb for be installed examples Basic Installation app 6 Mbytes Operating system DOS version DOS version DOS 4 0 or MS Windows DOS
10. These figures indicate that the total areas of the doublets mistaken for singlets as well as constituent Total Area Estimation Statistical Control Total Areas ADD1N1 spectrum ADD1N1 spectrum Inter Winner Interwinner Span GammaPus Sampo90 GammaTrac Hypermet T Hypermet ne s 4 a user settings aaa user settings GeniePC g defaut settings GeniePC g default settings 1 10 100 1000 0 1 1 10 100 1000 reduced sum of squares reduced sum of squares Figure 37 X values based on the reference Figure 38 X values based on the uncertainties uncertainties for the programs in their two reported by the programs in their two modes for the add1nl1 spectrum modes for the addini spectrum determined with a 1 x FWHM match determined with a 1 x FWHM match criterion criterion areas of correctly recognized doublets are determined well by Hypermet PC GammaVision GeniePC and Span These programs also demonstrate to be in reasonable statistical control in this respect 69 CHAPTER 9 CONCLUSIONS First of all after this intercomparison exercise we can conclude that most of the gamma ray spectrum analysis programs are designed using advanced and powerful graphical and user interfaces Many provide keyboard shortcuts mouse operation and graphical capabilities which leads to a more efficient operation and control of the spectrum analysis Outstanding user
11. P v Espen H Nullens F Adams Nucl Instr and Meth 142 1977 243 250 M Blaauw Nucl Instr Meth A333 1993 548 M Blaauw Nucl Instr Meth A353 1994 269 71 APPENDIX I TABLES OF RESULTS 73 Tabie 3 ActAn User seing mb m mm pm pu D Dm high peak N mm 48 47 55 46 du 5 2 5 3 1 6 0 104 1 4 2 1 i 5 6 3 7 41 2 small on high 16 16 22 4 I 2 9 3 2 1 1 I ne 4 5 E 5 I small on low 18 10 14 19 18 p ei 15 2 2 1 1 8 ed A 124 36 a 0 16 9 s i 289 7 71 5 4 0 1 K 6 a 18 0 1703 400 3 all matches 108 100 91 72 84 103 104 95 83 9 x i 0 10 7 554 2 8 6 2 5 4 5 d 7 28 5 53 84 33 15 6 30 1 160 7 14 9 86 4 4 31 2 411 106 9 16 7 50 3 90 7 position X UN uoa pops Te annihilation misses N 97 103 2 12 207 228 86 214 200 221 199 167 false hits N TE it 15 21 39 10 28 X is 5 P 0 1 9 609 9 31 8 1162 6 132 3 243 4 123 18 7 25 3 total N 185 205 329 337 308 265 1191 209 1319 322 343 304 278 SL Table 4 GammaPlus Default settings p Umi mr mr mur su Dus ID Tr qe 47 49 47 47 24 8 43 14 65 54 47 314 2 9 1 4 12 0 3 9 i 18 15 111 13 16 16 15 20 16 EV 17
12. capability 7 output of the results are always written to the same file in the current directory named PEAKANAL REP undocumented feature 60 There is no DOS ERROR trap capability build into the code Every time when printer got out of paper or operating system reported an error the program reported nonsense The documentation is concise but gives complete information from the user s point of view Detailed description of the used algorithms for area and uncertainties calculations is missing f It has a very helpful flow diagram to show how user should proceed to use the software 6 12 Inter Winner WinnerGamma version 3 42 The Winner Gamma software package is a MS Windows ver 3 1 or higher based gamma ray spectrum analysis program developed on the basis of a MCA emulation software f As a data acquisition device it can control and operates most of the commercially available MCA cards It also has the built in option for emulation of an acquisition system The installation is quite simple and requires the presence of the protection key as well as to operate the software It also can be installed in three EU languages English French or German 44 Although it operates most of the MCA cards the software can not convert different spectrum formats To read the IAEA ASCII format a special program was provided by the supplier but it was not very flexible and efficient e g spectrum to be read must always have the same
13. make a similar assumption that the vanance of closely neighboring peak shape functions is the same overlapping or closely neighboring peaks are analysed using the same FWHM value for the functions This violation of the above restrictions 1s considered permissible as long as the shifts are small with respect to the FWHM of peaks at low energies in the spectrum It was decided to restrict these shifts to less than twice the FWHM values at low energy which 1s less than 4 channels This will lead to peak separations of ca 1 5 FWHM at low energies and ca 0 5 FWHM for peaks above 2 MeV It should be noted that the adding of a shifted spectrum will lead to unphysical distortions in the first and last channels of the resulting spectrum These channels may not be used for any kind of determinations or quantification It was decided that several series of spectra should be measured out of which the reference spectra for the intercomparison are generated The radioactive source or sources to be used should emit a complex photon spectrum which 1s measured as a gamma ray spectrum with many peaks It 1s desirable that the spectrum a priori contains clearly separated singlet peaks as well as multiplets Radioactive sources available for the measurement were 40 kBq of 225 Am which generates a spectrum with lines up to 334 keV 70 kBq of 227Ac which generates a spectrum with lines up to 1109 keV 160 kBq of 226Ra and progeny which generates a spectrum wi
14. name and be in a fixed sub directory Before running the program the Setup program should be started 1n order to configure the hardware in use and the software options There 15 no provision for viewing the x axis channel numbers mode 69 software seems to work only for 4K spectra or it might be a problem in the conversion algorithm During the intercomparison spectra with 8K were normally read until the 4098th channel after which random data were stored into the next channels The Winner Gamma software sports a nice and efficient interface making use of all provisions of Windows operating system Colors are user defined and it uses multiple window environment However the cursor the spectra s misleading Actually there 1s no provision of the cursor The often so called cursor information channel and channels content 1s given while moving the mouse pointer over the spectrum but even no marker 15 provided for exact visualization of the current position The software provides an automatic mode of energy and resolution calibration In the resolution calibration the range of an interval min and max for the variation of the FWHM be numerically entered and them the software on automatic mode will chose the proper function to fit the experimental points 0 mce feature in this program s while you are in the E calibration all the known calibration peaks are automatically identified and displayed by c
15. 251 234 260 235 192 aes pec P false hits T3 ae fao m pog 1 ce de nc total 174 325 315 334 256 73 1182 327 306 333 301 255 DIM P 10 5 43 0 14 9 26 9 ae 9 3 7 9 73 9 39 4 21 4 14 0 8 0 Z x Z Z Z LL Table 6 Gamma Vision Default settings User settings 3nl ee high peak N 47 44 47 45 47 44 47 45 45 44 46 X1 0 8 E 473 5 45 5 149 4 8 0 8 0 8 2 4717 45 6 149 5 6 8 x 0 11702 582 574 152 as T 582 127 4 132 small on high i 6 D 6 23 usa a Luna small on low 5 T 5 A m m p A 4 m m ae 24 as as ws a E 256 a 1346 a all matches m WP pn eT pm ae 7 1 ae 956 9 50 3 147 5 iu E 956 9 50 3 147 5 I Lis e p annihilation ST eet misses 105 105 209 239 235 188 104 105 269 239 235 188 mse ox ue ums neo nes ma 9o 32 wa neo nos zu 9o false hits 24 it Hi T i 15 24 11 11 VE 5 ee ge Le ee 179 193 338 317 342 308 267 179 193 338 318 342 309 1267 x Z gt Z total Z gt lt Z 8L Table 7 Gamma W User settings settings p n 462 288 zs 10 3 3 m P 10 1 4 2 0 6 0 5 9 5 2
16. 3 8 7 29 2 7 8 72 3 5 45 4 6 9 27 9 total xxu c8 Table 11 InterWinner WinnerGamma Default settings dist 41 38 38 37 38 36 36 73 241 5 72 3 852 9 4 1 0 0 6 0 6 20 9 5 11751 1353 233 7 23 4 10 6 3 8 52 87 6 Es 13 21 16 15 19 17 9 3 1 7 4 9 22 1 2 1 3 aa uk me 323 562 E an User 16 1 6 36 7 small on low 11 10 F m 15 1s p ad s nu A 2 2 14 213 23 190 254 223 227 1 5 19 1 23 263 Er 110 3 425 8 L1 16 65 73 65 5 pm oe ES 1 8 13 x x Es d 828 6 93 5 m 9 2 eee 5 75 0 F 17 9 34 3 77 3 positon X 46 98 148 _ 102 70 82 annihilation i LL misses N 68 I 67 176 155 187 169 129 66 67 141 142 151 156 127 false hits N 47 5 427 EN 35 48 Eg 35 ES 25 Ba ns 2 23 8 Table 12 OSQ Professional Default settings User settings stra dit ini 183 1081 int00 stra dist 3m high peak pe p 43 47 45 45 p p 47 71 52 601 8 129 4 138 5 15 9 6 4 3 9 12199 D 5 155 159 3 13 9 small on high p 21 w 20i 25 m 3 2 141 152 119 26 40 9 132 small on low 33 15 m A 25 124 a E E S as 9 0 o 82 all matches Dos xi ids is 114 2 35 is 137 8 11 4 3 ies 60 5 position X 20 706 Hia
17. Menu Menu operation Yes Yes Yes Yes Yes Yes also Yes Yes Yes Yes Yes Yes command operation Can the menu options No No No Yes No No No No Yes But partially No Yes Choice between large and short be user defined or changed type of menus Graphical representation of spectra Yes Yes Yes Yes Yes Yes Yes Yes dotted Yes full spectra Yes Yes Yes dotted or and lines or reduce Joined lines representation ROI as bars fitted peaks Yes not Yes Yes Pressing Yes Yes ditterent Yes Fitted Yes Although Yes dotted Yes As well as Yes Yes In the Yes Fitted documented info peak excellent ways Excellent spectra requires to be only Fitted baseline under the ROI mode peaks how option representation points in diff the region peaks peak fitting Doublets in colours interactive peak presented as can be seen different fitting mode filled coloured against colors Displays fits spectra points Continuum and restduals under the on a separate peaks window IS ActAn Gamma Gamma Gamma GammaW Ganaas GeniePC Hypermet OSQ Sampo90 Span Winner Track Plus Vision PC Plus Gamma Residuals Yes No Yes Yes Yes Yes Partially Yes Documented Residuals plotted but does not only for work mu tiplets Energy calibration Yes Yes Yes but Yes Yes Yes Yes No Yes Yes Yes Yes Can be difference in eV Vs over the Energy spectrum Resolution calibration Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes Can b
18. Phone 615 482 4411 Toll free phone 800 251 9750 E mail WWW http www egginc com 86 Gamma W Ganaas GeniePC HypermetPC WinnerGamma 17 08 1995 3 3 1995 2 2 1995 4 00 1995 3 42 1995 Dr Westmeier GmbH Moellner Weg 32 D 3557 Ebsdorfergrund Moelln GERMANY Fax 49 64 24 49 94 Phone 49 64 24 43 23 E mail Westmeier t online de Physics Section RIPC IAEA Wagramerst 5 P O Box 100 Vienna AUSTRIA Fax 43 1 20607 Phone 7743 1 206021706 E mail T Dolan iaea org WWW http wwrw iaea or at Canberra Industries Inc 800 Research Parkway Meriden CT 06450 USA Fax 203 235 1347 Phone 203 238 2351 E mail mmoslinger canberra com WWW http www canberra com Nuclear Physics Department Institute of Isotopes KFKI XII Konkoloy Thege u 29 33 H 1525 Budapest POB 77 HUNGARY Fax 36 1 275 4349 Phone 36 1 275 4349 E mail MOLNAR iserv iki kfki hu EURISYS MESURES Reseau Eurisys Siege social ZA de l Observatorie 4 Avenue des Fr nes F 78067 St Quentin Yvelines CEDEX FRANCE Phone 771 30 14 57 78 Fax 1 30 14 57 80 87 OSQ Professional Sampo90 Span 88 6 3 releasel 1995 5 1 APTEC Engineering Ltd East 50B Caldari Road Concord Ontario CANADA 4N8 Phone 4 905 660 5373 Fax 905 660 9693 E mail WWW http www aptec inc com 80 Helsinki University of Technology Department of Tec
19. THE TEST SPECTRA The so called Test Spectra are set of gamma ray spectra that served as bases and common spectra analysed with the different software packages during the intercomparison They were acquired experimentally following most of the recommendations made by an Advisory Group Meeting AGM held Vienna December 1994 at IAEA and reported Final Report 11 During this AGM the need for new reference spectra was emphasised to be used for the intercomparison of different y ray spectrum analysis software packages and for quality assurance and quality control purposes It was stated that these new test or reference spectra should be real and measured spectra which should therefore contain no model dependent peculiarities nor any unphysical characteristics A subsequent meeting would then have the task to generate such spectra to be used for a software intercomparison n another action 3 1 Definition of the task It was agreed that the reference set of spectra are to be measured with a high punty germanium detector which contains a p type crystal The detector shall have a relative efficiency of not more than 20 order to supply test spectra which do not significantly deviate from the spectra measured many of the nuclear analytical and research laboratories The resolution FWHM at 1332 5 keV of the detector should be good around 1 8 to 2 0 keV and the crystal should exhibit no major defects which might show
20. The interactive peak fit 1s a separate procedure method that can be either associated to the Peak Area group or run after the Sum Peak Fit algonthm The first option 1s recommended the manual but the second option allows to combine results from both methods The setup of this method allows to place filters so that not all the peaks have to be interactively analysed programs allows the manual insertion or removal of individual peaks During the intercomparison when the default fit was not satisfactory the fit was modified either by adding peaks to the multiplets or changing the fitting region Sum Non Linear LSQ Fit Setup rPeak Area Region p795 Critical level test Start channe 150 Z fixed FWHM Stop channel 8000 7 EjuUse fixed tait parameter CjFit singlets Continuum 4 Channels C Show in MCA View Chameis FWHM Reject zero area peaks Continuum function Step Residual Search TjPerform Search 42000 Minimum separationdFWHME 1 ROI Limits Determination Max Num FWHMs between peaks 14 Num FWHMs for tett timit 14 Max Hum FWHMs for righttimit 4 T The fittng parameters be changed using the Peak function for each region A few other parameters can be changed to change the fitting region and number of peaks with
21. and bad area calculation are present It arises from wrong deconvolution of a very significant peak at 609 3 keV In this case that peak has been wrongly splitted two peaks 609 2 and 609 3 keV Therefore the program was penalized twice once for making a FALSE HIT at 609 2 keV and second because a wrong reported value for the area of highly significant peak at 609 3 has been calculated In this case a significant contribution 1s made to both Z scores The CMPSPEC program also reported the following statements after every table of analysis of the results 17 TRUE MATCHES Number of matches for high peaks 46 related chisqr for areas and reported uncertainty 0 8 and for reported areas with reference uncertainty 1 2 Let s have a quick look at their meanings The total number of matches reported peaks in coincidence with reference peaks within permissible interval of variance for peaks with high significance 420 gt 10 15 reported Ideally the chi square y values of unity should be expected for the reported uncertainty case In this example a value of 0 8 was obtained The mark notifies of a significant deviation of calculated x from the expected value unity according to the y distribution with 46 degrees of freedom The value of 1 2 was obtained when the Z score values were calculated using the reference uncertainty This difference 1 2 0 8 means that the program 15 slightly overestimating uncertainties fo
22. available for proper operation We have found that more than 600 Kbytes RAM provides sufficient space to operate properly the software This strong requirement on memory implies some limitations to DOS users since many drivers and TSR programs have to be disabled from RAM while working with Ganaas Ganaas 1s easy operated by its graphical user interface The software has been designed a modular way It consists of the following modules Set up module Organizes the different programs their locations and sub directones t also set up the graphics and communication facilities e Parameter Setup This program set and stores the parameters for the spectrum analysis It includes different calibration procedures and results in a file PAR which 15 called later for the fitting procedure e Analysis Program It allows the selection of the input parameter files spectrum files and performs the spectral analysis and displays the results Activity Analysis Calculates measured activities of the radioactive isotopes ACR e Concentration Determination this module s used for NAA calculations It can be done using fundamental parameters calculation or by the comparison method against samples corrections It also includes a module for Fast Neutron activation analysis calculations with solving routine for peak and reaction interference Correction factors can be also taken into all above mentioned stages e Utilities Modules inclu
23. be built in since high energy tailing distorts all peaks in the same way 8 5 Doublet peak area and uncertainty determination The programs cannot be expected to determine the peak areas in a spectrum containing only doublets as well as in the straight spectrum Nevertheless they should be in statistical control under all circumstances The quality of the area and its uncertainty estimation for the case of the 111 spectrum is shown 1n Figure 31 and Figure 32 in the same way as for the straight spectrum in Figure 29 and Figure 28 Area Estimation ADD1N1 spectrum Statistical Control ADD1N1 spectrum InterWinner InterWinner Span Span E g user settings GammaBus i g default settings GammaPus Sampo90 Sampo90 1 GammeTrac 2 GammaTrac g default settings Hypermet Hypermet GeniePC GeniePC Ganaas GammeVision E oso ActAn 1000 1 10 100 1000 10000 1 10 100 reduced sum of squares reduced sum of squares Figure 31 X values based on the uncertainties reported by the programs in their two modes for the add1n1 spectrum Figure 32 values based the reference uncertainties for the programs in their two modes for the addini spectrum As shown in Figure 31 none of the programs prove to be in statistical control All programs underestimate the uncertaintie
24. gamma ray spectrum of and progeny were used in the determination of the peak areas and positions With each analysis a list was made of difficult peaks For GAMMA W these peaks are located at energies 275 keV 280 keV 295 keV 352 keV 511 keV 543 keV 703 keV 741 keV 786 keV 1253 keV 1694 keV and 1937 keV For SAMPO these peaks are at channels 737 785 880 1220 1280 1528 1864 1975 and 2074 After these analysis runs the lists of results were compared and matched The results obtained for the clearly visible peaks even the ones labeled as difficult were good agreement within statistics except for the doublet at 295 keV that 1s resolved differently by the two programs This peak 15 therefore integrated for total area and labeled as unresolvable in the reference list For all peaks where the agreement 15 good the unweighted averages of the areas and positions reported by the two programs were taken as the reference values As uncertainty of the reference values the larger of the 1 standard deviation uncertainties reported by the two programs were taken Many noisy peaks were found by one program but not by the other If these peaks have an uncertainty exceeding 20 they will be be invisible in the test spectra with only 2000 seconds counting time and were therefore removed from the reference list In the cases where the uncertainty 15 less than 20 the spectrum was inspected visually for the presence of the
25. heavily referred Up to the judgement of the experimenter or user Finally a fully interactive analysis was performed for the STRAIGHT spectrum using as initial values of the parameters the optimum values found in the third step During the interactive analysis all the possibilities for improving the results in each region of interest were used This depends very much on the program being tested as some programs allow to change many items add or delete peaks change the fitting regions background models etc while other programs allow only to add or delete peaks Certainly the quality of the results 1n this test are more subjective than others and also will depend on the flexibility of the program and the accuracy with which each program interprets the operator commands A last test was devised and applied to some of the programs In this test the STRAIGHT spectrum was used and list of all the peaks of spectrum was provided to the operator This time the participants were asked to use the interactive features of the programs order to reproduce the list of references as accurately as possible Due to the problems described above this last test was not considered successful as no significant 1mprovement in the results could be achieved for the tested programs the results for each type of test were correspondingly coded The analysis and processing of the results were done only after all tests were performed CHAPTER 3
26. install program s not able to create a sub directory together with parents e g c gamtests gt The main program can be started at the DOS command prompt as gt and it accepts pipeline options Some of the pipeline parameters are essential for the proper operation of the program gt We have experience that this method for mitialization 1s quite inconvenient as the user needs to remember the command line parameters for example that typing gt K will allow erase of spectra during acquisition The program only accepts data formatted in the proprietary format types spt or spm Even ASCII formatted data can be read by the program tf The GammaTrack supports a very convenient and easy to operate energy calibration procedure From the ROIs in the spectra the program automatically finds the tentatively standard energies The calibration curve may be displayed together with the calibration coefficients User defied lines can be entered from the keyboard as well The FWHM calibration is performed simultaneously However the program does not treat properly the 511 keV peak If this peak is marked for calibration t results an erroneous FWHM calibration This could be corrected only by deleting the 511 keV peak from the list The drawback of this automated calibration 1s that it forces the user to use the calibration parameters and function that the program provides as the result of the calibration procedure For insta
27. it is clear form the figures that a program like Gamma W is performing better in this respect than e g InterWinner WinnerGamma 8 2 Peak shape model dependency of absolute peak area determination As described in Chapter 4 during the intercomparison a re normalization factor has been calculated for each program in order to compensate any possible bias in the peak area determination due to the different models employed in each program In Figure 27 the re normalization factors and their uncertainties for the peak areas determined by the analysis programs are shown 1 006 1 004 zw T E e 2 eg 1000 e e I 0 998 e d 0996 O 0994 gt lt 0992 e e 0990 0 988 i E E 5 Z z 8 E 8 8 5 8 E a O z B wn 8 o lt 2 4 a 5 e E E 5 S E E 8 8 65 5 Figure 27 Renormalization factors and uncertainties for the analysis programs It has been said that peak areas in y ray spectra can get defined differently when applying different peak shape models This would mean that one analysis program could yield other areas than the next as long it was self consistent The results from any determination would come out the same as long as the program was used both for calibration and measurement a bias would cancel out This is
28. or Bad are not given 22 Chi 42 09 12 48 48 147 A parameter Crit Level 1s calculated for each peak as explained in the listing how this parameter should be used or interpreted is however not mentioned It was found usefull and mce capabilities built in this software e g library editor among others but they are not within the scope of this intercomparison GammaTrack s DOS gamma ray spectrum analysis program developed on the basis of a MCA emulation software As a data acquisition device it works only with Oxford Nucleus cards Hardware requirements are not stringent except for memory available to DOS Running this program requires then that the computer is configured to get the maximum memory to DOS by placing the proper statements and devices 1n the CONFIG SYS and AUTOEXEC BAT files On many systems the support for networking sound cards or extra devices will have to be removed order to successfully operate GammaTrack A hard key protection key type Sentinel has to be connected before beginning the installation The installation program INSTALL EXE 1s quite simple although it lacks of some flexibility It prompts for a directory to install GammaTrack which is created if not existing During the installation compressed files are extracted from the diskette using the LHA program After installation is completed a browser displays a readme file for updated information 7 The
29. or right edge The zooming in and out does not keep the cursor centered and can have a disorienting effect For the cursor position channel number energy and channel contents are displayed The cursor is moved with the left and right cursor keys As identification of the spectrum under study the filename is shown on the screen The user cannot set the colors used by the program ff Fit results energy shape and efficiency curves can all be shown graphically with optional display of residuals although the program does not use multiple windows 28 Peak area determination 15 performed by first performing a peak search then subtractng the background from the regions to fit and finally fitting the peaks One peak fitting function is available A Gaussian with low energy tail ft Three background functions can be used Linear linear with an error function under the peak and Bremmsstrahlung The error function 1s determined in the process of subtracting the background from the region to fit from the rest of the spectrum on the high energy side of the spectrum The fitting of the peak 1s done by a non linear least squares algonthm described by Slavic 16 2 No exception 1s made for the 511 keV peak The spectrum can be analysed 1n automated mode of the user s definition using macros and or batch mode In interactive mode the user can suggest peak positions and set the region to fit but in the fitting process peaks may be dropped agai
30. position determination from the simpler problem of energy calibration the conversion from channel to energy was performed by a program that applied the same energy calibration to the output of all programs allowing for different methods of numbering of channels some programs start counting at channel 1 others at 0 Also 1t was found that for the added spectra a slightly different energy calibration 1s optimal than for the straight and distort spectra For both groups a quadratic energy calibration was applied For the added spectra the relation between peak position p and energy E used 1s given by E 2 78 0 396952p 4 Ox10 p and for the straight and distort spectra E 2 78 0 396952p 7 0x10 p The same conversion program yielded output in a standard format simular to the format of the reference lists containing peak positions in terms of energy and peak areas both with their absolute 1 standard deviation uncertainties This step was not entirely trivial because only a few programs did report uncertainties in the peak positions 1 Gamma W Hypermet PC and Sampo90 For all the other programs which reported energies with two or more digits after the dot implying uncertainties of less that 0 01 keV these uncertainties were set to 0 01 keV Also not all programs reported 1 standard deviation uncertainties The reported uncertainties were converted to one standard deviation absolute uncertainties acco
31. sample holder at 178 mm distance from the detector endcap The set up was closed with the 50 mm thick lead cover and no modifications to the system were allowed during all measurements with the 2 source Before starting the first measurement the components were set to reach thermal equilibrium for one hour with all voltages applied The spectra shown in the Table 1 were measured under the above conditions in the course of the following four days 3 3 The measured spectra The relevant characteristics of the measured spectra are displayed in the following figures 599500 MCA Poss EE 2 1 50775 E xs e E S29 969 505 true y Figure 1 First half of spectrum SD0000 MCA t 1 d 1C Figure 2 Second half of spectrum 500000 It is seen from Figs 1 4 that the spectrum contains a large number of peaks most of which are clearly separated singlets and only a few small doublet peaks The very closely situated peaks around channels number 3500 see Figure 4 as well as other smaller doublets are well suited to create complex quadruplets through adding of spectra after shifting them by a few channels u 9 SBOOCOC MOS Zeota ls Figure 3 A detailed section of spectrum SD0000 containing singlet peaks 50200 MUS Oste s T zoo ea qa
32. significant false peak was reported by the program Usually this situation arises from the erroneous splitting of a significant peak during the analysis TOTALS Number of regarded peaks 172 related chisqr for areas 9 2 CONSTANTS USED Threshold energy 100 00 Criteria for energy matching 1 E2 lt 2 sqrt sqr dE1 sqr dE2 E1 E2 lt 0 5 FWHMCE1 Criterion for high significance A ref err gt 10 Criterion for high continuum 3 0 net gross Criterion for annihilaton peak 511 lt 3 0 As shown here the analysis program CMPSPEC is also reporting the criteria and parameters used Now lets gives some Hints and Tips for which qualifiers you should look at 1n order to find your golden gamma ray spectrum analysis program O Looking for a general purpose gamma ray spectrum analysis program A general purpose gamma ray spectrum analysis program should detect normal peaks all possible types of gamma ray spectra and report them with the correct uncertainty Then in our test we should look for a program that would gives a x values statistical control meaning without an ugly mark for the following categories Number of non 511 matches all together 59 related chisqr for areas and reported uncertainty 1 2 and for reported areas with reference uncertainty 1 4 This category includes all matches together high peaks small peaks small on high and low baseline If ther
33. so E 2s pes 32 9 xa ud mE pn MANNA misses N 110 120 255 239 272 235 194 91 102 209 1208 1228 1172 X 8 9 12 9 x 74 3 110 0 32 2 10 0 6 1 7 6 223 25 5 68 9 7 9 false hits N 12 13 26 11 X 4148 15 scio 123 829 73 8 3 a UE total N 170 182 317 299 330 293 1251 182 196 319 301 336 252 i mo 2a x ues ws m2 fare v8 Table 13 Sampo 90 Default settings User settings 48 46 p ee e 46 86 i 67 69 47 27 0 20 0 54 7 5 10 9 2 6 gt 2 22 1 23 3 37 5 small on high 29 27 28 m ds F 24 du ii 74 47 41 24 E 50 il EN 93 23 u 479 163 1185 206 17 smali on low 19 18 18 x 08 v5 is ve a a 3 5 48 3 22 06 RE 12 a 06 15 76 1313 27 all matches n 1116 117 101 193 ui pn d 16 2 6 0 68 13 3 72 x y S 323 7 13 9 20 2 10 5 3 39 8 29 0 19 5 2213 200 m 2 3 8774 loni 3 d 4 misses N n3 us 4 248 280 196 113 115 193 175 708 188 154 X 97 2 560 928 ET 9 5 9 7 E 123 98 8 false hits N 44 4 2 170 177 353 annihilation E 2 8 Table 14 Span Default settings User settings TOT ee TOT T N 151 51 43 41 41 67 0 7 10 9 4 6 7 3 0 1 9 1 1 10 6 I X2 1 0 22 102 17 9 7 1 2 8 2 0 462 small on high 14 17 15 17 11 25 15 28 24 27 23 2
34. up as peak shape distortions 1n the measured spectra The measurements of spectra should be performed a temperature stabilised environment where major vanations of electronic properties due to variations of the ambient temperature be excluded The electronic components used should be commercially available units which are neither specially configured nor modified in any way The setting of the amplifier and ADC should be such that the energy range going up to ca 3000 keV 15 measured in an 8k 8192 channels spectrum In Some tested software proved to be extremely reluctant to perform the fitting the way the operator wanted 4 contrast to the recommendation of the former AGM the experts this meeting have decided to measure 8k spectra rather than 16k spectra The reason for this deviation from former proposals 1s that the commercial market mostly supplies ADCs with a conversion gain of 8k and only new units allow to digitise onto 14 bit numbers In addition many existing MCA units do not support 16k spectra Displaying an energy range of ca 30 keV to ca 3000 keV on 8k channels the resolution function FWHM vs channel number 1s expected to span the range from FWHM 2 channels at the low energy end to FWHM 6 channels at high energies This would allow to test also the ability of codes to treat very narrow peaks in a correct way The following basic rules pertinent to this task are defined e Spectra must contain only real measure
35. were matched to begin with For each peak in the list of reference values a partner in the list of reported values was looked for z decision that a match had been found was taken if t eed to within their quadratically summed uncertainties or if the distance between the peaks was less than 0 5 FWHM The values for this refereed FWHM were taken from the commonly established calibration curve see Chapter 2 as determined for the 40 000 seconds counting time spectrum If no match was found in the reference list the reported peak was considered to be a false hit and only the second z score 3 could be computed using zero both as the reference area and as its uncertainty e Ifthe reported area was missing t was considered a miss and only the first z score 2 could be computed With such a method implemented missing a noisy peak or reporting a false with high uncertainty in the area do not result 1n high z scores and grants a fear punctuation for the program The same tests were designed for multiplets The added reference spectra contain doublets with separations varying from 0 4 to 1 2 x FWHM and with different area ratios e For this test 1f two peaks the reference list matched one peak in the analysis program output 1e 1f both reference peaks were located within 0 5 x FWHM of the analysis result the two reference peaks were merged before the computation of z scores In such a way it 1s possible to test a
36. 0 seconds spectra This generated multiplets with known peak area ratios of 3 1 10 1 and 100 1 with respect to the peaks the 2000 seconds spectrum spectrum where the peak shape is distorted through random coincidences with photons coming from another highly active source has been measured as well This source completely distorted the spectrum the low energy regime but the only significant influence at higher energies was through random coincidences alone resulting high energy tailing of all peaks 3 2 The measuring set up The measurements were performed room no B09ZK92 1 m the B building of the IAEA headquarters The room 1s situated in the centre of the building it has no windows and the air condition keeps 115 environment conditions very stable The following units were used for the test Semiconductor Detector p type coaxaal one open end closed end facing the window Canberra model number GC1318 7500 High voltage 5000 Voits Relative efficiency 12 6 Resolution at 1332 5 keV 1 72 keV Resolution at 122 keV 700 eV 473 30 htres dewar with vertical dipstick Preamplifier Canberra model 2001C High Voltage BIAS Canberra model 3125 operated at 4500 Volts Linear Amplifier Tennelec TC244 with the following settings Coarse gam 20 Fine gain 1 000 Peaking time 8 microseconds Pulse shape Gaussian BLR manually set symmetric using an oscilloscope for control Input polarity negative Output unipola
37. 2 00 or DOS 3 1 or OS 2 210 DOS 31 or Windows 3 ior DOS 33 or DOS31or Windows 3 1 required 3 1 or later 5 0 or later later 31 or later later later later later later later later or later Minimum RAM 590 Kb 610 Kb 4 Mbytes 8 4096 KB 512 KB about 600 KB 8 Mbytes 300 Kb 8 Mbytes 505 Kb 600 Kb 8Mb strong Mbytes strong 16 Mbytes recommended if estimation recommended requirement recommended Tequirement recommended operating under Win 95 Graphics cards Minimum required EGA EGA EGA VGA EGA CGA VGA VGA VGA EGA EGA VGA Modes that can be EGA EGA VGA CGA EGA VGA and EGA VGA or CGA EGA Any driver VGA SVGA SVGA Graphics VGA SVGA VGA VGA SVGA used effectively VGA SVGA and SVGA VGA and LCD SVGA better VGA supported by accelerator SVGA Graphics VGA OS 2 recommended accelerator recommended Color Monitor Required Yes Yes Not but Yes Not but Not but Yes Yes Not but Not but Yes Not but recommended recommended recommended recommended recommended recommended Used Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Mouse support Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes Yes Keyboard operation Yes Yes Yes Yes Yes Partially No No Yes Yes Yes Yes only possible Software protection None Parallel port Parallel port None Parallel port key None Parallel Port Parallel port key Software will Parallel port Special Parallel port or key key Key necessary operate normally key format of PCMCIA key only for without any the ne
38. 3 9 all matches 1100 105 39 3 19 9 11 8 5 7 x 7 8 29 9 12 3 104 6 124 22 4 16 8 15 0 36 4 11 0 19 EZEN annihilation te ee te misses gt ior 211 193 211 185 94 102 210 196 234 214 180 false hits N 28 14 27 13 total N 168 197 25 285 327 285 251 168 196 328 292 329 289 249 X 4 4 15 4 61 9 37 4 22 8 12 2 7 6 4 2 38 0 53 2 9 0 11 8 7 0 7 9 6L Table 8 Ganaas Default settings User settings dit int 3m 101 Hie uen 48 47 53 52 52 51 47 52 51 0 4 0 5 8 6 10 7 4 6 a 0 6 47 2 8 15 35 376 2 1003 502 45 6 1100 small on high N If 10 10 10 10 11 10 14 40 98 37 0 36 66 34 X2 30 196 172 13 0 112 19 4 18 9 12 2 small on low EN all matches position X 51 31 annihilation ite misses N 105 231 250 239 247 184 1100 105 230 250 241 297 184 abei N total N 169 184 a 325 x 314 169 184 301 322 E 315 08 Table 9 GeniePC Default settings User im 3m iP high peak N 45 42 35 43 39 uM 59 61 56 47 X 13 16 629 6 1260 1830 14 9 s is 44 4 2 0 9 X 51 6l 1225 1 92 8 225 0 16 4 29 176 1792 20 small on high 2 1 1 2 2 16 I8 40
39. 36 62 63 232 3 5 d 29 a ni 25 19 ES EE E 28 9 9 74 22 settings small on low E 12 12 2 5 ive 25 8 48 37 31 08 L 102 25s 5 0 d EE 122 49 56 ail matches 47 48 93 1746 nt 21 214 6 15 6 13 8 annihilation X sen 9 misses N lu hus 292 260 292 252 204 87 93 237 198 234 208 174 X 52 9 94 3 26 5 12 7 57 46 4 8 0 22 9 10 6 8 0 false hits N 19 30 20 22 17 15 24 x Dh 9 10 2 EE ied 314 0 19 4 19 5 3 8 total 1179 330 307 333 300 252 1187 209 340 309 344 1307 275 I8 Table 10 Hypermet PC Default settings User settings Default settings 2 high peak N 47 47 59 54 55 52 46 48 69 55 56 53 x _ 12 28 1 2 16 5 x 01 249 A PAGE 1 22 9 1 prs 4 442 small on high 23 23 21 d a i du ER 52 2 7 4 1 28 zu m 68 5 71 3 6 194 7 14 13 16 0 2 2 3 7 52 0 9 i 36 6 ine 71 1169 2 1 all matches N 60 61 69 1110 p 93 89 12 21 254 ie a 17 3 122 26 2 56 1739 35 9 ian 29 4 20 8 I position X n arja 81 22 67 eo annihilation RE EE misses Ho 109 257 226 1258 227 189 80 196 176 229 195 163 8 9 is 56 10 5 9 5 z 5 3 553 93 E false hits 13 05 on 6 1 305 170 185 327 295 329 295 252 180 318 289 E bee 6 2 47 6 50
40. Channels C With Low End Channels 0 High End Channels cance Centroid Calculated C Between Gross Half Max 2 Between Net Half Max C Between Net Tenth Max C Over Whole ROI Centroid Weighting C Net Chan Net abs Net Chan Figure 19 Aptec OSQ Professional Control parameters of the peak search The program can report activities and calculate Minimum Detectable Activities MDA for peaks which have not been found 39 8 Start Time ROIEG Centroid 186 38 Gross 46133 Activity 0 0 00 cps Ra 226 FWHM 6 94Net 17713 err 0 00 1 76 Bkgnd 28420 Figure 20 Aptec s OSQ Professional Fitting of the 186 keV peak OSQ Professional provides a built in utility for nuclear data library edition It also provides some specific applications data libraries f The user can define his report by selecting a number of options Units in the reports be standardised The help on line is concise although some times too short 47 manuals are excellent Well written and edited with plenty of figures examples Tips and hints ft OSQ Professional provide to the user many small but nice features e g when the program starts it always loads the last analysed spectrum or when looking at any particular region of the spectrum and you look even load into another spectrum it will go to that region automatically even exp
41. EA CALIB1 CEN Peak shape calibration ACTAN IAEA CALIB2 SHP Date of processing 12 11 95 09 16 59 True time 33 333m Live time 33 333m Area error lt 99 0 Background Iteration Cycles 15 Sensitivity for peak search 2 8 Sensitivity for closed multiplet search 8 Maximum Deviation of FWHM from calibration 20 Maximum iteration number for peak search in residuals 1 Analysing from channel 256 to channel 8095 Crit Level Net Area sqrt 2 Background Peak classification Normal Small S Residual R Bad fitted F No Centroid Ey Net Area Background FWHM Ch kev counts counts keV lF 462 6 186 45 181681 245 1 3 24532 0 83 28S 499 4 201 01 378t 175 46 3 15102 0 84 3 5 574 5 230 81 352 157 44 5 12102 0 87 4 F 593 8 238 46 270 154 56 9 11653 0 87 5 603 3 242 25 28303 259 0 9 18394 0 88 The program properly keeps records of the job by listing all the files that were used for energy and shape calibration and all the parameters that were set for peak search For each peak quite complete information is given the peak position in channels and keV the net peak area and the standard deviation also in percentage the number of counts in the continuum under the peak the FWHM and the x value A qualification Normal Small Residual or Bad fitted is given to each peak Residual clearly indicates that peak was found during residual search The definitions of Small Normal
42. Gamma Plus Gamma Vision GammaW Ganaas GeniePC Hypermet PC 050 Professional Sampo90 Span Winner Gamma Spectrum utilities spectrum format Only from No Program Yes several Yes major Yes major Yes Vast Yes major Yes but Yes major Yes major Yes major No Butit conversion ASCII might be commercial commercial choice commercial limited commercial commercia commercial can operate requested formats formats including formats formats 1 formats formats major from the major Commercial supplier commercial MCA cards formats Can the program No Yes Smooth Yes Yes Sum Yes Partially Yes No No No Yes perform spectrum and strip Smoothing stnp and Smooth using Smoothing and Smoothing manipulation addition smooth file format stnpping add Subtraction conversion subtract and option multiply by multiplication const Add const Calculations Activities calculations Yes Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes Can the program Yes buta Yes No No Yes Two Yes but the Yes The Yes Yes Yes No calculate template has modules for report has to be isotope concentrations to be Thermal and programmed library is very programmed Fast NAA are using the report limited included language Report of detection Yes LLDare Not known several Partially MDA are Yes as Yes Not reported Not reported limits calculated formulas concentrations calculated by Minimum according but Not according to provided are reported as the l
43. IAEA TECDOC 1011 Intercomparison of gamma ray analysis software packages as 32 T INTERNATIONAL ATOMIC ENERGY AGENCY 1998 The IAEA does not normally maintain stocks of reports in this series However microfiche copies of these reports can be obtained from INIS Clearinghouse International Atomic Energy Agency Wagramerstrasse 5 P O Box 100 A 1400 Vienna Austria Orders should be accompanied by prepayment of Austrian Schillings 100 in the form of a cheque or in the form of IAEA microfiche service coupons which may be ordered separately from the INIS Clearinghouse The originating Section of this publication in the IAEA was Physics Section International Atomic Energy Agency Wagramer Strasse 5 P O Box 100 A 1400 Vienna Austria INTERCOMPARISON OF GAMMA RAY ANALYSIS SOFTWARE PACKAGES IAEA VIENNA 1998 IAEA TECDOC 101 1 ISSN 1011 4289 IAEA 1998 Printed by the IAEA in Austria April 1998 FOREWORD Whth the advances personal computers software packages started to play a key role the control acquisition and validation of the data in nuclear experiments In particular because of the wide range of applications of gamma ray spectrometry gamma ray spectrum analysis software packages are among the most used software any nuclear laboratory They are being used in such important applications as environmental studies low level monitoring neutron activation analysis acce
44. MPSPEC which gives an ASCII file illustrated in Figure 8 This output table contents Energy and uncertainty for Reference as well as for reported processed spectrum e Net Peak Area and uncertainty for Reference as well as for reported processed spectrum Z scores calculated with Reference and Reported uncertainties Let s take a look at some of the cases that may happen just to explain the meaning of our results Here a table of the CMPSPEC program s output 1S presented Here uncertainties the energy column were omitted in order to simplify our discussion 16 Reported measured data En rea Z score Case No 1 The Area reported by the program 1s 0 That means this peak at 106 7 keV was not detected by the program In this case the Z score 1s calculated from the reference spectrum value alone Zet If the peak m the reference spectrum has large uncertainty therefore the contribution to the Ze will be small as in this case Meaning this peak s not so important m terms of information that might be extracted from So penalisation for missing t should not be large In the opposite case the program 15 substantially penalized with a high Zer value because a significant peak has been missed And the case 1s counted for the missed peaks category Case No 2 Both areas are different from zero In this case the program found a true peak at 186 2 keV This category was classified as Matches The Z
45. THE CANDIDATES In this Chapter a short description of each program s given For certam aspects of the operation of the programs more detail information 15 provided complementing the overview description presented in the Chapter 7 Advantages and inconveniences are also pointed out whenever possible Faulty operations encountered during the intercomparison are reported as well For each program a picture of the fitted 186 keV peak 1s shown ActAn 1s a gamma spectrum and neutron activation analysis software developed at the Centre of Applied Studies for Nuclear Development Havana Cuba The development of this program has been partially financed by an Agency Research Contract No 5014 The functions are high resolution gamma ray spectrum analysis and instrumental neutron activation analysis A nuclear data library editor 1s also available The design goal was to have a program that could do a fair job on slow and small PC s 1e 286 PC s running DOS with no extra memory The package consists of one 3 25 HD installation floppy and a 20 page User s Manual 20 Installation is done by running the batch file INSTALL BAT that resides on the installation floppy with as argument the drive where the software needs to be installed e g A gt INSTALL The batch file will create a directory ACTAN and copy all files that make up the package The next thing to do is to include the directory in the PATH and set the environment varia
46. THEIR PRODUCERS Ras OE Tee DO vl ER TRES 86 CONTRIBUTORS TO DRAFTING AND REVI IEW 89 Symbols used In this document a number of symbols are adopted 1n order to catch the attention and provide a fast guide to the reader They are as follows tf Advantages issues Disadvantages 9 Improper performance or error encountered during the operation of the software This 15 more known as a Bug m the program Tips or Hints on a certain issue For Notes The two first are purely based on the subjective opinion of the participants and authors of this report A companion diskette with the set of test spectra and the programs for processing the test results 1s attached CHAPTER 1 INTRODUCTION With the advent of powerful computers and structured programming languages the software has reached an important level as the logical controller at different stages from a single instrument to an entire computer controlled experiment That is also the case for software packages in nuclear instruments and experiments Today software in nuclear techniques and experiments controls the instrument s operation performs the data acquisition storage and their validation and analysis In particular because of the wide range of applications of gamma ray spectrometry software packages in this field are among the most used There is a vast selection of commercia
47. The user can limit the reported peaks by entering its error limit that is peaks with an relative error in the peak area larger than this value will not be listed in the report 3 Sensitivity factor for peak search The less this value is the more sensitive will a peak search Only the optimum value of 5 is specified in the manual 4 Maximum peak number This option sets the maximum peak number and allocates memory space for saving the parameters of the peaks The maximum peak number should be between 200 and 512 You can select to subtract the background contribution calculated from the another spectra from the results of the NAA calculation for instance This option was not used in this exercise The Integral Peak Technique which produces step background line was used In this way the baseline can be subtracted from the raw spectral data before the nonlinear fit is done Residual fit is not shown on the screen although there is a graphic region dedicated to this task 43 There is not proper handling of the 511 peak The code usually combines Total Peak Area method for singlets with nonlinear squares fit to calculate net peak area and uncertainties for multiplets The nonlinear fit can be done using either Gaussian shape or experimental peak shape previously stored in the calibration run Figure 23 Span Fitting of the 186 keV peak SPAN sports a batch processing capability as well as automatic energy re calibration
48. acy in uncertainties calculation for simple spectra Check the followings 1n the STRAIGHT spectra test The chi square of the HIGH PEAKS MATCHES for STRAIGHT spectra using reported and reference uncertainties should be approximately equal If 1t 1s not the case then two cases are possible 1 Overestimation of the reported errors by the program X lt X ret 2 Underestumation of the reported errors by the program x gt x ret Looking for a program with high multiplet resolving power You should use ADDINI ADD1N3 ADD3N1 and ADDIONI spectra for the test The complexity of spectra s growing from the ADDINI to ADDION spectra therefore you can trace the behavior of the program as intensity ratios of the multiplet peaks are increasing this case a good performing program will result chi square using reported uncertainty around 10 in the HIGH PEAKS MATCHES category This value is obtained because usually some doublets not deconvoluted therefore a larger value of chi square for MISSED PEAKS 15 obtained The SMALL PEAKS chi square value usually 1s in statistical control for these spectra Test of my program for different capabilities on relatively simple spectra Always use the STRAIGHT spectra test Many other examples can be brought but the ones already given are sufficient for you to assemble the test of your choice Now that everything has been explained let us introduce our candidates CHAPTER 6
49. ame peak It is seen that the variation of the peak position is on the order of 0 1 keV whereas the resolution of this small peak varies quite considerably in the beginning of the experiment Both quantities the peak position and the resolution become more stable towards the end of the experiment which is about 16 hours after the measuring system was powered Stability of peakpositions Stability of FWHM Peakpos tions 16 FWHM of peak channels Spectrum number Spectrum number Figure 6 Positions of a peak in 20 Figure 7 FWHM of a peak in 20 successively measured successively measured spectra spectra 10 3 5 Manipulations of the measured spectra After considering the negligible instability of the measuring set up as observed the first spectra it was decided to sum up these 20 spectra 1n order to create a master spectrum which 1s used to extract reference data for the peak positions and peak areas The position values were used in the intercomparison exercise for automatic analysis routines as absolute reference values The FWHM values are not recommended for use any intercomparison because the values are dependent on the model used the fitting process The stability of the FWHM values however should not be dependent on the model and it can be used to determine the stability of the system Both the position and the FWHM values for the peaks are given in terms of channels which 15 a spectromet
50. ameter HW Doublets are detected by comparing the fit with the measured spectrum parameter Rmax If the difference between FWHM of the peak and the calibration FWHM is less than DEV FWHM then the peak FWHM is used otherwise the FWHM from calibration is used allowing for deconvolution of closed multiplets The program can only add one peak per residual search iteration 7 The actual peak area determination is not documented The baseline is removed from the spectrum using an iterative filter A peak shape is calculated based on the peak shapes stored during the calibration step This numerical function is fitted to the measured spectrum For this method to be valid the peaks used for calculating the shape should have very good statistics at least 10 times more counts than the spectra to analyse otherwise the calculation of uncertainties in the fitted data will be biased A consequence of using this method is that if the calibration spectrum is analysed zero values for Chi squares will be obtained 21 Figure 9 ActAn Fitting of the 186 keV peak For each spectrum analysed a file with the same name and extension SRC is created containing all information that can be displayed or listed An ASCII file with extension SRP contains the analysis results in a readable form A typical example is shown below ActAn v2 5 Peak Search and Area Calculation Report Spectrum C ACTAN IAEA STRAIGHT DAT Energy calibration ACTAN IA
51. anal Chem 124 1988 467 Calibration and Use of Germanium Detectors for the Measurement of Gamma Ray Emission Rates of Radionuclides N42 14 1991 American National Standards Institute New York M J Koskelo Mercier Nucl Instr and Meth A299 1990 318 Mercier Koskelo J Radioanal Nucl Chem 160 1992 233 Koskelo Mercier J Radioanal Nucl Chem 193 1995 211 Final Report of the AGM on software for nuclear spectroscopy December 1994 Rep 5 95 IAEA Vienna Internal Report M Blaauw V Osorio Fernandez W Westmeier Submitted to Nucl Instr meth W Westmeier Benutzerhandbuch des Gamma Spektrenanalyseprogramms GAMMA W Gesellschaft f r Kernspektrometrie mbH Ebsdorfergrund Molln Germany 1990 P A Aarnio et al Nucl Instrum Methods 219 1984 173 U Reus W Westmeier Atomic Data and Nuclear Data Tables 29 1983 Slavic Nucl Instrum Methods 134 1976 285 B Jaeckel W Westmeier P Patzelt Nucl Instrum Methods A261 1987 543 G W Phillips K W Marlow Nucl Instr Meth 137 1976 525 See also NRL Memorandum Report 3198 Naval Research Laboratory Washington DC January 1976 R Gunnink Nucl Instrum Methods A299 1990 372 376 R Gunnink J Niday Computerised Quantitative Analysis of Gamma Ray Spectrometry Univ of California Rep UCRL 51061 11 3 1972 Wang Liyu Nucl Elect Detect Techn China Vol 6 1986b 356 361
52. and it if you were at expanded mode before 6 10 Sampo 90 version 3 6 The Sampo90 program is a well known gamma ray spectrum analysis program It is a copyright of P Aarnio et al Helsinki Univ of Technology although several dealers are distributing it The copy obtained for this intercomparison was operated under DOS ver 3 3 or higher It needs a 7 Mbytes disk space The installation is simple and effective using the restore command but the user has to modify config sys and autoexec bat files by himself J It has no mouse support although it gives a full user operation through the keystrokes 40 ff Sampo90 software has a built in a conversion utility from the Canberra S100 Accuspec ORTEC ADCAM The Nucleus PCA spectrum formats It has an extremely useful capability to 1mport user defined ASCII files 69 Sampo90 does not accept all valid DOS filenames e g filenames with characters in them cannot be entered Sampo90 s user interface 1s reasonably friendly and fast Input of numbers 15 a little cumbersome if cursor keys are used to select command Even after using the program for days the right moment to enter numbers was often missed In some cases the same information has to be entered several times Also the different parts of the program are unaware of each other For example if 8192 ASCII spectrum s to be analyzed the number has to be entered 3 times Graphs of peak fits
53. area standard Relative area Relative determine by and relative background Options Relative peak area determination fitting Itis standard deviation deviation uncertainty uncertainty fitting or uncertainty n area Net Gross uncertainty reported not deviation reported in reported in integration Method of reported background How Uncertainty 15 documented reported in times sigma sigma 1 2 or Level of uncertainty 15 Uncertainty 15 relative uncertainty 15 relative and in how the other sigma in 1 20r3o 1n 30 reported not reported relative absolute calculated is values uncertainty uncertainties not obtained can be etc documented chosen 96 Peak identification ActAn Gamma Gamma Gamma GammaW Ganaas GentePC Hypermet 050 Sampo90 Span Winner Track Plus Vision PC Professional Gamma ON SE AAI s k kak CIE SL ELISE sk TGS SEE EEL OO NL ORE TEE SOL EES DEO IS LET AIT OL IIE ELE NLL DOE LLL APE Does the program Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes performs radionuclide including identification gt detection of possible interference Reports of the Printed and Printed only Printed Printed or Printed and Printed or Printed or Printed and saved Printed Printed Printed or identification results saved into a file orsaved mto and or saved saved into a saved into a saved into Saved into a into a file ASCII saved into ONLY No saved in
54. bilities of the programs ActAn Gamina Gamma Gamma GammaW Ganaas GeniePC Hypermet 050 Sampo90 Span Winner Track Plus Vision PC Plus Gamma rr ps TI EERE M AEQ EAI T GET I I RE LE ER TE NU cc EE LT EN Zoom Option Yes ROIand Yes As Expand Yes In Yes ROIs Yes a little Yes ROI Yes As Yes As Yes User defined Yes Yes Itcanbe Yes User available fitted peaks facility ROIs expand mode fitted peaks disonenting fitted peaks Expand facility expand option defined by defined peak fits Zoom the lower part and residuals ROI fitted and residuals ROIs and any controlled by the user user defined but screen shows 5 Expand peaks and part of the the user limited 64 entire too residuals spectra Zoom channels on X spectrum the can be user axis upper part the defined Itis expanded not possible to region display fits or residuals in the main program Y scaling modes Lin Auto and Lin Log Auto Lin Log Lin Log Lin Log Lin Log Lin Log Lin Log 544 Lin Log Sqrt Log and Lin Log Lin Log Manual No and Manual No Auto and Manual and Manual 541 Manual Manual and Auto Manual and manual Auto only Auto and Sqrt Manual Sqrt Sqr Manual No Auto No No Auto and and Auto Auto No Sqrt Manual No and Auto Sqrt Sqn scaling Sqrt scaling User defined colors Yes Yes Yes Yes No No Colors Yes No Yes Yes No Yes can be disable only
55. ble ACTAN to that directory e g PATH C ACTAN SET ACTAN C ACTAN The installation is very simple and primitive do not allow to change target sub directories etc but it works fine When the program starts the Main Screen is displayed From this main screen functions can be accessed via menu s and sub menu s A number of functions are available trough the function keys The three main functions are e Gamma Spectrum Analysis e Instrumental Neutron Activation Analysis e Nuclear Data Library Editor ff The menu system be used quite efficiently both with the mouse or the keyboard During spectrum analysis or whenever ActAn displays graphical data the so called Graphical User Interface is used which again can be operated by mouse or keyboard documentation on the actual spectrum analysis procedures is rather limited The parameters that affect the peak search and doublet resolution are briefly discussed in the user s manual The following parameters can be adjusted for peak search e Error limit to display peak search results Peaks with a relative error in the peak area larger than this value will not be listed in the report e Sensitivity for peak search HW range 2 7 7 e Sensitivity for closed multiplets Rmax range 0 10 e Maximum deviation of FWHM from calibration DEV_FWHM Maximum number of residual search iterations Initial peak search is based on a square wave correlator par
56. calibration needed by the software as well as for quantification purposes onginally measured as SD0020 MCA DISTORT ASC a spectrum measured for 2000 sec real time distorted by the presence of the source The resulting peak shape distortion can be used to test the stability of analysis programs in circumstances where calibration and measurement do not agree originally measured as SD0034 MCA a multiplet generated spectrum where the peak area ratios are 11 expected to be 1 1 within statistics ADDIN3 ASC a multiplet generated spectrum where the peak area ratios are expected to 1 3 within statistics ADD3NI ASC a multiplet generated spectrum where the peak area ratios are expected to be 3 1 within statistics ADDIONI ASC a multiplet generated spectrum where the peak area ratios are expected to be 10 1 within statistics ADDINIOO ASC a multiplet generated spectrum where the peak area ratios are expected to be 1 100 within statistics Due to the shift of either 3 or 3 channels the added spectra will exhibit good separation of peaks at low energy 1 5 FWHM whereas the separation at high energies 1s very poor ca 0 5 FWHM and many of the high energy doublets may be indiscernible as such The wide range of peak separations and peak area ratios obtained allows for thorough tests of the ability of analysis programs to separate doublets automatically with some user interference or with methods of definit
57. contents of the manual are duplicating the online help file And it describes the operation of the version 1 2 24 6 3 Gamma Plus version 1 02 0 Gamma Plus is a DOS based MCA emulation and gamma ray spectrum analysis software As a data acquisition device it works only with Silena type of MCA acquisition add on cards The installation should be quite simpie see comments below through the install program The software comes in 3 5 diskette and its operation is protected by the presence of a hard key in the parallel port Although claimed to be simple the intercomparison exercise failed to install several times this software Most of the problems were related to the type of drivers or hardware devices installed in our computers We found out that GammaPlus is not compatible with Quantum LPS540S SCSI hard disk drive Problems on its operation were encountered in computers running compression drivers for hard disk particularly Stacker ver 4 00 An incompatibility between GAMMAPLUS and 386 is noted The following command line must be present for proper operation of the software DEVICE C DOS EMM386 EXE FRAME NONE GammaPlus can read only spectra in Silena format which is well documented in the Appendix B of the user s Manual f The software has nice and fast user interface implementation It also properly uses the multiples windows facilities Energy calibration can be done automatically I
58. d data without any synthetic or model dependent modifications Peaks in the spectra must have peak areas which span the range from gt 50000 counts very large to below the detection limit e Manipulations of spectra must not modify the correct statistical nature of the counts per channel distribution 1 e Spectra must not be scaled Spectra must not be subtracted Spectra may be added e Manipulations of spectra must not modify intrinsic dependencies of the detector used such as eg the resolution function or the Peak to Compton ratio P C as a function of the peak energy e Numbers to be quantified must be obtained such way that no system model dependent properties will influence them For the generation of multiplet peaks with known ratios of the peak areas however one has to violate some of the above restrictions It was decided that multiplets should be generated through the adding of two spectra which were measured under absolutely identical conditions but with different real counting times Thus the ratio of the areas of the same peak is well known through the ratio of the counting times The multiplet nature should be generated in such a way that one spectrum 15 shifted by a small number of channels before making the adding procedure In this way well characterised multiplets can be generated preserving at the same time the exact nature of the resolution function Most if not all spectrum analysis programs however
59. de two libraries and the library manager and editor P SPEDAC is a module used for reformatting the spectra from different formats to the required input format of the Ganaas program Ganaas needs about 3 Mbytes hard disk space for the installation of all the programs After installation the path has to be set to include the sub directory Spectral data from a source format 1s converted by the program P SPEDAC to a target format required as the input of the spectrum analysis program the following logical sequences Select format of spectrum to be converted from source Select format of spectrum to convert to target Select directory of filenames of all source spectra Select directory of target spectra Once the source 1s selected in the next stage the following sequences are executed 30 Read the spectrum in the source format Display the spectrum for visual inspection Save the spectrum in the target format STOP This procedure can be performed in a batch mode for up to 12 spectra files 0 Almost all commercially available formats are implemented in the file conversion module of the Ganaas Energy calibration FWHM calibration is easy to perform in Ganaas After it is done the table of points or the graphs with corresponding residuals can be seen The efficiency calibration is powerful in Ganaas It can be done by entering the data points and then selecting a polynomial of different degrees
60. di I 13 8 4 0 i 19 35 36 88 23 22 z 4 24 7 26 6 2 0 3 8 62 9 4 153 1 8 small on low 11 5 12 12 11 10 P 0 1 9 A 3 2 5 2 21 7 5 1 2 14 2 7 1 4 2 3 x x 244 114 5 1 33 4 129 9 all matches 5 82 74 2 6 1 6 s is ru EM i 12 7 19 8 position 50 annihilation postion X EI et wa misses N 70 175 156 184 166 146 163 134 165 150 fs X a 282 8 0 26 4 12 7 22 6 17 3 7 9 false hits N 25 3 30 43 51 E X A 5 4 2 3 1 5 4 2 9 3 1 3 2 3 4 total N e 259 250 270 259 228 180 280 269 305 300 270 et ERE ERE ERE E ERR high peak APPENDIXII LIST OF TESTED SOFTWARE PACKAGES AND THEIR PRODUCERS 2 5 1995 Centro de Estudios Aplicados al Desarrollo Nuclear CEADEN Calle 30 No 502 e 5ta y 7ma Miramar Playa Havana CUBA Fax 53 7 Phone 53 7 22 14 22 E mail ceaden cenia ac cu GammaPlus 1 02 0 1995 SILENA S P A Via Firenze 3 1 20063 Cernusco s N ITALY Fax Int 39 0 2 92 106 331 Phone Int 39 0 2 92 106 293 E mail GammaTrack 1 3 release 2 Oxford Instruments Inc Nuclear Measurement Group P O Box 2560 601 Oak Ridge Turnpike Oak Ridge Tenessee 37831 2560 USA Fax 615 483 5891 Phone 615 483 8405 Toll free phone 1 800 769 93673 E mail GammaVision 2 3 1995 EG amp G ORTEC 100 Midland Road Oak Ridge TN 37831 0895 USA Fax 615 483 2177
61. e s any problem you find a near the reported x value then you should check for which type of category of peaks the code 1s given incorrect values If this will not be your case you can have that program 1n the list of favorites Looking for a program with outstanding peak search capabilities Check the following values in the STRAIGHT spectra test Chi squares of MISSES This value should be near to one A high value means that significant peaks from the reference spectrum were missed e Chi squares of FALSE HITS This value should be also near to one A high value of chi square in this case means that some significant false peak was reported Since usually this situation arises from the wrong splitting of the significant peak you should look carefully at it If this 1s not the case this qualifier can give you a hint for a good peak finder program e The number of HIGH PEAKS MATCHES should be equal to 48 8K STRAIGHT spectrum Looking for a program with accurate peak area calculation in simple spectra Look to following values 1n the STRAIGHT spectra test 19 The number of HIGH PEAKS MATCHES should be equal to 48 8K STRAIGHT spectrum but with a value of chi square for reference uncertainties near to one to be in statistical control A high value of chi square for reference uncertainties this case would mean that area calculation of the singlet peaks 1s incorrect Looking for a program with outstanding accur
62. e superimposed over the spectrum Efficiency calibration Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Multiple windows Only Yes Yes but Yes No No Yes Yes No Only split for Yes No Yes Each available for zoom mainly for zoom individual showing diff window can parts of spect be configured Info The independentl windows can y not be resized Additional Yes Yes Almost Yes Id TT Yes Headers Yes Spectrum headers TT Yes very Yes Almost Yes TT LT 145 Yes but Yes Almost Yes Channel information displayed ful header LT TT LT ID LE LI LT counts complete all information dates gross very all Counts ch counts energy information of available in counts Preset limited information and energy energy peak position parameters and the header of Time etc available in TT and LT If position header the file the headerof ROI selected provided the file info on ROI 15 displayed Spectrum cursor Available Yes Yes slow and Yes Yes slow Yes good Yes slow Yes using Yes Yes acceleration Yes Yes but can Partially a acceleration accelerated movement and movementover mouse mode available be moved kind of cursor mode movement Key using arrow accelerated spectra Key accelerated Key and mouse using mouse info 15 available and mouse keys movement and mouse movement controlled only provided Key and controlled Accelerated Key and controlled Key and Keystrokes for while moving mouse movement mousc mousc fast movement the
63. e User s manuals well written and organized However the manual does not document the proprietary spectra format spc Contrary to the old format chn fully documented 6 5 Gamma W Version 17 08 by Dr Westmei ff The software has a powerful conversion utility program that converts most available commercial spectra format into ASCII format which is the default format for Gamma W According to the manual Gamma W is a high sensitivity high precision code for the analysis of gamma ray spectra from Ge Li and HPGe detectors Peak areas are determined through the fitting of mathematical shape functions to experimental counts after the subtraction of the analytical background distribution The figure below shows the graphical fit control on screen at 186 keV of STRAIGHT spectrum in optimised automatic method Gamma W is a DOS application We have found that more than 512 Kbytes RAM provides sufficient space to operate the software properly E 186 3 8 f 17530 9 8 9 88 kel Figure 13 GammaW Fitting of the 186 keV peak The spectrum can be displayed the x axis region controlled with functions keys the y axis with cursor keys The latter can be logarithmic or linear but the user has no control over the offset in the case of linear y axis Some inconveniences in the graphical user interface were found For example the graph does not scroll horizontally if the cursor pushes against the left
64. e the confidence interval for the number of degrees of freedom obtained 6 peaks Therefore in the case of small peaks on low background the uncertainty calculation in the program 1s accurate Number of non 511 matches all together 59 related chisqr for areas and reported uncertainty 1 2 and for reported areas with reference uncertainty 1 4 and the chisqr for their positions 15 0 This 15 a summary of the chi squares for all peaks total of matches 59 As can be seen the results are very good The program resulting this 15 certainly in very good statistical control It calculates net peak area as well as uncertainties with very good accuracy for all cases high peaks small peaks in the low and in the high background FITTING THE 511 keV PEAK Number of peaks found there 2 related chisqr 160 3 This 15 the test for the treatment of the annihilation 511 keV peak The above example 1s not the case of a good determination of this peak because it has been splitted into two peaks t This 1s the average of the squares of the Z scores values a category 18 MISSES AND FALSE HITS Number of misses 110 related chisqr 9 9 The total number of missed peaks 15 evaluated for a program A high value of x 1s reporting the fact that some significant peaks were missed Number of false hits 1 related chisqr 99 1 The total number of false hits s given as well A high value of x in this case 15 telling that some
65. efficiency shape and energy calibration are available The shape calibration 15 the only one without a residual plot The energy plot has a unique and beautiful feature It indicates increased uncertainty in energy when extrapolating outside the range of the calibration points 47 For shape and energy calibration the program can automatically find and fit suitable peaks for the shape calibration The user can then modify the list to exclude odd peaks An exception can be made for the 511 keV peak A very nice possibility of automatic shape calibration where each peak the program proposes to use for the calibration 1s shown to the user The user can accept or reject the peak and if the peak 15 too close to others the program offers the correct advice not to use the peak The default significance level for the peaks in this procedure 15 only 4 that should be a higher value like 10 20 E2 12 10 Qe P M k aS 2 3 4 5 6 7 771000 1500 2906 2500 2060 3500 4000 Figure 21 Sampo90 Energy calibration graph In almost all Sampo90 graphs residuals are shown Spectrum analysis consists of peak search followed by a fitting procedure The range of channels to be fitted can be indicated but the search 1s not radionuclide library oriented The peak fitting function is the standard two tailed Gaussian a linear or parabolic background wit
66. ented the program screen input for different parameters of fitting and logical operations s ugly and very inconvenient 609 Dunng the intercomparison a problem was found m relation to the capacity of the LOG file It was seen that doing fitting for whole spectra which has many peaks LOG file becomes too big Once the file reaches the 64K capacity the program hangs up 36 During the intercomparison the program gave Floating Point Errors during its operation It was found not very stable in operation yet There is no automatic recognition of the 511 keV peak you can do only manual fit using variable FWHM it is explicitly recommended in that way in the User s Manual There is an incorrect definition of the SIGMA LIMIT parameter which affects the cnteria of goodness of fit y7 4 SIGMA_LIMIT A linear sum of the Gaussian and a Gaussian with exponential tail on the left side is used as peak shape function A step function folded with the Gaussian peak and a similarly folded tail function resembling a sharp Compton edge are added to the usual first or second degree polynomial baseline function Below are show some pictures illustrating some instances and performance of this program 2 addloni ton I PPT mea 2 LH my 2 3 YAXX11 LIMIT x xx44xY lt gu y3Uu 0 775727 b en
67. er a short description of the spectra 1s relevant this part of the report The list of test spectra 1s as follows CALIB ASC calibration spectrum containing Co 57 Cs 137 Na 22 Mn 54 and Co 60 STRAIGHT ASC Ra 226 progeny spectrum counted for 2000 s DISTORT ASC same as above but counted in the presence of 241 source to induce high energy tailing ADDIN1 ASC Sum of two Ra 226 spectra one of them shifted by 3 channels to the nght ADD3N1 ASC Sum of two Ra 226 spectra one counted for 2000 s one for 667 s the second shifted to the right by 3 channels ADDIN3 ASC Sum of two Ra 226 spectra one counted for 2000 s one for 667 s the second shifted to the left by 3 channels ADDIONI ASC Sum of two Ra 226 spectra one counted for 2000 s one for 200 s the second shifted to the nght by 3 channels ADD1N100 ASC Sum of two Ra 226 spectra one counted for 2000 s one for 20 s the second shifted to the left by 3 channels The participants in the workshop had the responsibility of testing either or two programs each from the 12 programs that were available for the test All the participants were considered experienced y ray spectrometrists They also represented a wide range of users from developed laboratones to spectrometrists from developing countries Among the participants there were also experienced programmers and persons with experience 1n statistical analysis of data All the programs were tested as received
68. et data usc GE eaa uN sss 6 7 GeniePC version 2 2 ssecsececcccsssscccsesecneccseusessesseesscenecasecsnecceseateseessuecases 6 8 Hypermet PC version 4 00 oco nad ensi Uus 6 9 OSQ Professional version 6 3 release 1 6 10 Sampo 90 version 3 6 1n aste 6 11 Span version 5 1 6 12 InterWinner WinnerGamma version 3 42 222 222222222 1 001 0 CHAPTER 7 OVERVIEW eesecepeeostoasonescsssseseososceassoevsvessaceeovpotneevecevoseeceoeoovoecessesecoocepevooeo 13 13 15 15 16 CHAPTER 8 RESULTS AND DISCUSSIONS eh 61 8 1 Peak detection ability eet dana plan mesi 61 8 2 Peak shape model dependency of absolute peak determination 62 8 3 Interpretation of the x values asiento tardis sed eet eases tees 63 8 4 Singlet peak area and uncertainty determination 64 8 5 Doublet peak area and uncertainty determination 66 CHAPTER CONCLUSIONS oiu EF D Ra Pe E 70 REFERENCES itu a ou eot ue es eu e IUE 7 APPENDIX I TABLES OF 73 APPENDIX II LIST OF TESTED SOFTWARE PACKAGES AND
69. format HELP Provides a short help text Unfortunately FILECNVT does not indicate plain ASCII which actually corresponds to the TOOLKIT option so the conversion of the test spectra to an intermediate format was performed Finally a program to convert the spectra from ASCII format to cnf format was prepared using the OS 2 REXX batch language which provides a convenient way of transferring the data as file names can be selected from directory lists In summary Genie PC provides facilities to read other file formats but this facility 15 not conveniently implemented in the package 47 energy calibration procedure was found to be quite straight forward effective and powerful Display capabilities and multiples windows are properly used for the convenience of the user It 1s also possible to load an already existing calibration parameters The Markers button loads the centroid of a peak between two markers If the data set has been previously calibrated t 15 possible to load the old data using the Populate button T The Automatic mode of calibration was found to be quite convenient and efficient for simple and fast calibrations 0 user can select the degree of the polynomials The shape calibration is done in a simular way parallel to the energy calibration An additional parameter as low Tail 1s given and plotted which illustrate the shape of the peaks in the whole range of spectrum energy J Note that the eq
70. hnical Physics Rakentajanaukio 2 FIN 02 150 Espoo FINLAND Fax 358 0 451 3195 Phone 1358 0 451 3191 E mail aarnio hut fi China Institute of Atomic Energy P O Box 275 29 Beijing China 102413 Fax 86 10 9357102 Phone 86 10 9357102 CONTRIBUTORS TO DRAFTING AND REVIEW BERNASCONI G BLAAUW M CAPOTE Noy R van ESPEN P FAZINIC S MANH D H MOLLA N Islam OSORIO V WESTMEIER W Seibersdorf Laboratories International Atomic Energy Agency Seibersdorf Austria Interfaculty Reactor Institute University of Technology Delft Netherlands Centro de Estudios Aplicados al Desarrollo Nuclear CEADEN Havana Cuba Micro and Trace Analysis Center MiTAC University of Antwerp Antwerp Belgium International Atomic Energy Agency Nuclear Physics Institute Dalat Viet Nam Institute of Nuclear Science and Technology Bangladesh Atomic Energy Commission Dhaka Bangladesh International Atomic Energy Agency Dr Westmeier GmbH Gesellschaft f r Kernspektrometrie M lln Germany 89
71. hout a Compton step under the peak In the fitting procedure the chi square 1s minimized f separate peak shape can be used for the 511 keV peak 1 e a wider Gaussian ft Spectrum analysis can be fully automatic fully interactive or in between 41 ff The significance thresholds for search and fit as all parameters involved are clearly defined in the manual and can be set by the user Peaks can be inserted or deleted by the user even before fitting should a peak be missed in the search as well as during manual fitting Peak fit for singlets 1s linear by default No uncertainty in peak position is determined In the case of multiplets or singlets if so desired non linear fitting is applied Positions are varied as well as peak area all other shape parameters are fixed The user does not have the option to free the tail or width parameters Peaks are fitted together 1f separated by 4 or 6 FWHM manual inconsistency The fit area is determined automatically With a separate program not included in the standard package concentrations in NAA can also be calculated using the comparative method Figure 22 Sampo90 Fitting of the 186 keV peak with the corresponding plot of the residuals Interference corrections when interpreting the spectrum are performed using least squares methods t There 15 context sensitive help available in the program Error messages are clear Peak positions areas and their uncer
72. ibrary Detectable to documented documented Curn s including 3 sqr sd area driven peak Activity formulae formula Cum s one when it is lt search and area or user LDL algonthms defined using peak erosion method Quality Assurance and Quality Control How the program Operational Operational Yes tosome ANSINI330 Parameters files Hasacomplete Yes Operational Full track No info Full track of complies with Quality parameters parameters extent and N42 14 are kept for and full QA parameters parameters are of provided on operational Assurance save into the are stored Operational reproducibility QC system are stored stored into operational search parameters file Results are into the file parameters of results but built in for into files extensive parameters identificatio database compared with Results of are stored not program do not reporting the headers n Little can be a cumulative analysis are all use them for results Headers can be QA and track of used created distnbution statistically QC All parameters saved included QC is built parameters QA and QC factor controlled into reports Almost no 65 ActAn How can all input data Menu and and calculations dialogue boxes traced Values are stored in files and can be re called Gamma Track Menu and dialogue boxes Most of the operational parameters are stored into the file Gamma Plus Some operational parameters are s
73. igh values can result Therefore a high X value can represent a group of nearly perfectly distributed z scores i e a mean value of 0 and a mean of squares of 1 containing such a single high z score At the same time as described in the Chapter 4 we are using two kinds of y values e the first based on z scores accounting for both the reference and the reported uncertainties Z defined in formulae 3 in Chapter 4 the second on z scores computed from reference uncertainties only Ze as defined in formulae 2 Chapter 4 This two qualifiers can point out different meanings 1 the first X indicates the level of statistical control of the program as described above the second indicates how well the peak areas themselves were determined regardless of the reported uncertainties 63 8 4 Singlet peak area and uncertainty determination Since only the straight and the distort spectra contain singlet peaks the analysis results obtained from these spectra are relevant for the quality of singlet area determination of the programs The ANSI standard describes the testing of the independence of peak area determination from the baseline level This aspect is not explicitly tested here However if a program performs well in the test described here it follows that it satisfies the ANSI criterion If it performs badly one of the possible causes could be such a dependence Statistical Co
74. ing and viewing the results and the residuals of the fits tt Calibrations are easy and efficiently implemented It also includes an automatic peak detection for calibrations In principle only one data point 1s needed 26 The efficiency calibration has built in several mathematical functions for two intervals e above the knee point as interpolation linear quadratic or polynomials of several user defined degrees bellow the knee interpolation linear and quadratic functions Polynomials are at most expressed log E The baseline function is calculated differently depending of the type of the peak For singlets one of he following options can be chosen automatic average over five points three points value or minimum For multiplets a choice between stepped or parabolic type baseline functions Gaussian peak shapes are used for fitting multiplets while peak numerical integration 1s used for singlets Figure 12 GammaVision Fitting of the 186 keV peak It should be realized that the way GammaVision performs the spectrum analysis 1s unique and some how different to the usual methods GammaVision will deconvolute multiplets when distinct centroids are determined either by the peak search or by the library In the case of overlapping peaks multiplets which the peak routine cannot locate centroids the spectral data this reduces to using only constituent peaks that are 1n the gamma ray LIBRARY as deconvolutio
75. interfaces were found in GeniePC GammaVision and OSQ Professional Most of the programs incorporate utility modules e g spectrum format conversion which extends their usefulness and applicability Some of the programs have built in scripts macros application oriented libraries and other features that were found very useful for predefined tasks routine analyses etc Example of this is the batch analysis script provided by Sampo 90 and GeniePC among others Almost all the programs provide easy to use and powerful methods for energy peak shape and efficiency calibration It was found that most of the programs integrate into their basic spectrum analysis routines a number of calculation programs or modules which leads to more precise and correct calculations for element s concentration Examples are nuclear data libraries and library editors efficiency calculations and extrapolations geometry corrections peak interference solvers etc the programs made good use of present capabilities of modern PC s Spectrum deconvolution and calculations are done rapidly and the user can inspect the results over the spectra using powerful graphical capabilities During the intercomparison it was found that all the tested programs reported almost the same peak areas within 196 independently of their used peak shape The expected peak shape model dependency was not found in practice In the applied tests singlet peak areas are determined
76. inties Value Value Ze 1 2 2 Org In such a way fair scores will characterized the different tests grving proper weight to misses false hits incorrectly quoted uncertainties etc Therefore for our specific case the z scores 15 the differences between reported values and reference values divided by their own uncertainties In those cases where both a reported area and a reference area were available hits two such z scores could be computed 1 A z score related to the quality of area determination based the uncertainties the reference files as ref 2 Z ref 2 2 4206 Orf 2 and a z score related to the statistical control of the analysis program based on both the reference uncertainty and the uncertainty reported by the analysis program 13 3 2 2 Orr where and Ap are the reference and reported peak area and o and their uncertainties respectively Since 20 1s the ratio of the counting times of the reference spectrum and the test spectra see Chapter 3 20 Was considered the optimum uncertainty to be reported by the analysis programs Such z scores are expected to be normally distributed with a zero mean and a unity standard deviation 1e z scores higher than 2 or lower than 2 indicates that something 15 wrong at the 0 05 level In each comparison run the reference and reported list of peaks
77. ion whichever 1s possible or necessary 3 6 Determination of reference peak positions and areas As a result of the methods used to obtain the test spectra no independent and unbiased method of sufficient precision and accuracy 1s available to determine the peak areas and positions in the test spectra It was decided that careful analysis of the summed spectrum with a total counting time of 40 000 sec by two experts using two analysis programs in as interactive mode as possible yields both positions and energies with the required precision and accuracy to compute good values for the expected positions and areas the test spectra The spectra measured at 40 000 sec have much worse statistics than the one measured at 2 000 sec and the proposed method 15 therefore valid It was decided to use the programs GAMMA W 13 and SAMPO 90 14 because of previous expertise on them of the staff performing the task During analysis logs were kept on each decision taken In both programs the shape and energy cahbrations were made using the spectrum to be analysed itself and the gamma ray catalogue of Westmeier 15 The peaks used for this purpose were at 74 8 keV 77 1 keV 186 0 keV 295 2 keV 351 9 keV 609 3 keV and 1120 3 keV Second order energy calibration was used GAMMA W uses also second order fitting of the FWHM channel dependency SAMPO uses the linear relation between the square root of the peak width and the channel number Knowledge of the
78. l and FTP address for the exchange of information and user s help but it 15 not documented in the User s Manual It is only mentioned in a README text file included in the first installation diskette CHAPTER 8 RESULTS AND DISCUSSIONS In this Chapter the results of the intercomparison are presented and briefly discussed Only the most relevant graphics has been shown here The full report on the results is summarized through tables 1 to 12 in Appendix I to this report In these tables X1 and X2 refer to the values computed using the uncertainty reported by the program and the reference uncertainty respectively Doublets of the annihilation peak are unphysical and the corresponding results are therefore omitted from the tables Empty columns in these tables correspond to incomplete test or lack of reliable data This might be the case of incomplete analysis runs not performed by mistake during the intercomparison These data are therefore missing from the tables 8 1 Peak detection ability The ability of the programs to detect singlet peaks which 15 the first aspect to be tested according to the ANSI standard can be judged from the number of detected small peaks on high background Peak detection Peak discrimination STRAIGHT spectrum STRAIGHT spectrum InterWinner x InterWinner Span GammaPus GammaPus Sampo90 a T T 90 GammaTrac setinge pa
79. l and open domain software packages that has been designed for the efficient and correct analysis of high resolution gamma ray spectra It is the aim of this intercomparison to test and describe the abilities of 12 such software packages In the past similar intercomparisons have been made and reported in the literature 1 6 They all deal in one way or the other with topics like e precision of the programs finding peaks large small ones in low baseline small ones on the Compton edge etc e precision finding multiplets accuracy and correctness performing energy FWHM and efficiency calibration and many others They all rely on different sets of test spectra 2 3 which in many cases did not represent real measured spectra and lacked statistical correctness On the other hand the ANSI standard 42 12 section 8 7 specifies methods to verify the performance of such software packages The basic concepts of this standard were applied to selected programs by Koskelo 8 10 After an Advisory Group meeting AGM organised by the IAEA it was recommended to perform an intercomparison of gamma ray spectrum analysis software based on a new set of test spectra 11 This time the main objective of the proposed intercomparison was to focus on aspects such as the ability of the programs to determine the peak areas and the peak area uncertainties e peak positions and the peak position uncertainties the statis
80. lerator based nuclear analytical techniques and a number of medical applications On the other hand such software packages are expensive Therefore it would be of general common interest to have precise and specific mformation available on the present status and performance of advanced gamma ray spectrum analysis software The IAEA undertook an intercomparison exercise to review available software This TECDOC describes the methods used in the intercomparison exercise characterizes the software packages reviewed and presents the results Only direct results are given without any recommendation for a particular software or method for gamma ray spectrum analysis Since this report covers a wide selection of commercially and non commercially available software and reports their performance under different circumstances it reflects to some extent the present state of the art 1n the field of gamma ray spectrometry analysis software This TECDOC will be useful to a wide range of persons including university students technical staff domg gamma ray spectrometry software programmers scientists interested technical aspects of data analysis gamma ray spectrometry software operators and executive directors or project managers who might be mvolved setting up a project in this field or involved the process of purchasing equipment and software The IAEA 15 grateful to Blaauw for his work on the drafting and editing of the manusc
81. licated particularly 1f the computer 1s intended to operate a network 0 The mstallation of Genie PC and its additional components 15 quite simple and flexible The computer has to be rebooted after installing the Spectroscopy Assistant However it 1s not necessary to reboot the computer after installing the advanced options In order to run any of the utilities that compose this package the user must run the Virtual Data Manager VDM first This program stays resident and works as a virtual machine that performs I O The program stores the spectral information 1n a proprietary binary format that takes the extension cnf and 1s called datasource A utility called FILECNVT converts spectra from various formats into GemePC data sources also referred as CAM files in the manuals This utility 15 implemented as OS 2 command line program has to be run from the OS 2 command box and 1s cumbersome to use alone as for converting each spectra However it 1s convenient to use batch files The command to convert a spectrum is gt FILECNVT lt INPUT FILE NAME gt OUTPUT FILE NAME QUALIFIER The allowed qualifiers S100 Canberra S100 cards SPECTRAN Canberra Spectran AT GAMMA Canberra Gamma At 32 TOOLKIT Canberra Senes 35 Toolkit SAMPO Micro Sampo 90 ORTEC Ortec file type chn or spe ND6S Nuclear Data 6 senes MCAs NUCLEUS Nucleus cards Intertechnique now Eurisys Meas type
82. licking at the points the display the user can chose to include it in the calibration and the available elements and their corresponding energy are displayed automatically besides the mouse pointer for the user selection Efficiency calibration 15 done using a In E or 1 E polynomials with an user defined degree The plot of the efficiency also reports uncertainties Real time Channel Energy keV Content a FWHM Bross a Isotope Figure 24 Eurisys Winner Gamma Fitting of the 186 keV peak 45 The analysis of the spectra 1s based first on the peak search and then on the peak fit peak search and fitting can be controlled by the operator but 1s limited to the variation of the peak sensitivity factor only Activities can be also calculated and reported Standard units can be user pre defined for reporting activities 7 Residuals of the fitting procedure are not displayed The software allows the comparison of several spectra It has implemented some spectrum manipulation features like change of the contents for an specific channel dangerous smooth with user defined type of smooth polynom strip compress multiply by a constant add one or several spectra subtract spectrum The received User s manual was very limited No information at all on specific issues like peak area determination and search peak fitting etc
83. loppy disks into the hard disk after creating the needed directories No peak shape calibration 1s needed Shape parameters for peaks and background are found by a non linear fit for each separate region the spectrum using an initial first trial value for the peak width Dynamic variations of the parameters throughout the spectrum are automatically adjusted As minimum input using default parameters this code requires only information for two peaks approximate FWHM and exact energy These two peaks should be selected as distant as possible This information 15 enough to carry out energy and automatic shape calibration There 1s a LOG file for the whole fitting procedure where all fitting operations and parameters are recorded and can be seen analysed at any time It 1s possible even to see the whole fitting sequence for each multiplet for instance if you are doing three different fits im one region you can see graphically the results for each fit one after another fF code has very easy and user friendly way of adding deleting peaks n any region of the spectra You can add or delete peaks using channel position or energy The program marks the BAD FITTED regions according to x criteria Although the HELP option appears in the mam screen menu no Help ts available at all in this version 1t might be 1mplemented the future Q common graphical display of Energy vs Channel and FWHM vs Channel are not implem
84. me time 0 Any calibraton is easy and efficiently done OSQ Professional Energy and FWHM calibrations can be done automated mode were peaks from the spectra are automatically 1dentified and energy values are extracted from the provided nuclear data library an specific one for calibration 1s provided called 1sotope lib Nevertheless the user still has the option to change delete or add any additional point or to change the reported calibration parameters Efficiency calibration 1s also convenient and efficient Functions like Polynomaals of Ln E 1 E and cubic spline are given for the choice of the operator Polynomial s degrees be chosen as well The peak search and fitting procedures are controlled by the user through a wide choice of parameters Figure 19 shows available parameters Possible peak centroids are located by filtering the spectrum with a third order derivative A zero crossing method together with criteria of statistical confidence are used to detect peaks and precisely locate peak positions However residuals values plotted only when multiplets are fitted Peak Calculation Setup for 50 Basa Bkgnd Line C Straight Line Least Square Fit Degree 1 C2 C3 Ca 5 7 In Cnts 1 T7 In Chan 1 Use Method 2 ee r Confidence MDA Sigma Error Sigma P i 3 Peak Search C Partial Start Max Peak Error 96 ROI Bkgnd End Points with 5 End
85. mouse controlled provided controlled controlled provided mouse controlled as well cS Which information it displays ActAn channel or energy and counts at position Gamma Track channel or energy and counts at position Gamma Plus channel or energy and counts at position Gamma Viston channel energy and counts at position GammaW channel energy and counts at position Ganaas channel energy and counts at position GeniePC Several Channel energy counts markers etc Hypermet PC channel energy and counts at position OSQ Plus channel energy and counts at position Sampo90 channel energy and channel content Span channel energy and counts at position Winner Gamma channel energy and counts at position ES Calibration Procedures ActAn Gamma Gamma Gamma GammaW GeniePC Hypermet OSQ Sampo90 Span Winner Track Plus Viston PC Plus Gamma Mimmum and Min 2 Min 2 not Min 3 and Min 2 and Min 1 Linear caltbration 1 User defined 2 Min and Maximum number of Max 3 limited for Max 20 11 User defined 5 Quadratic 3 parameters used for E Automatic Max polynormal E Automatic Depending function Automatic calibration mode calib also mode of on the using three mode of provided provided but as calibration function exp ports calibration especial option provided as Min provided Mi
86. n by the program The fitted peaks are reported with positions energies fitted areas and uncertainties There are two kinds of reports The first 1s kind of log file that will contain everything computed in a session the second a file containing only peak energies areas and their uncertainties The user has some control over the log file output format The log file can be directed to a printer directly to a file output files are plam ASCII The peak width calibration function s a second order polynomial just like the energy calibration function The program offers a vast selection of mathematical functions for efficiency calibration Among the most effective 1s a log log polynomial with any number of parameters up to 10 or the Jaeckel amp Westmeier curve as published the literature 17 ft program can manipulate spectra and can run automatic analysis batch mode The program identify radionuclides and report the activities once the efficiency calibration 15 performed Activities are computed from each identified peak Also concentrations can be computed from the activities for INAA based on thermal and epithermal flux along with irradiation times 6 6 sion 3 3 by Physics Section IAEA Vienna The gamma ray and neutron activation analysis software GANAAS package has been developed by IAEA Physics Section under auspices of its Technical Cooperation TC programme with the main objective
87. n fits the peak s found and reports the position or energy the fitting region the FWHM and the net peak area 49554295 ATE ee Y Figure 11 Silena GammaPlus Fitting of the 186 keV peak Overview of the graphical user interface The fit 1s shown graphically a small window a linear scale however without the possibility to change the y axis 7 No uncertainties in any of the obtained values or any fit quality criteria such as chi square 15 reported The parameters that control the fit are the same as those used for the analysis function not possible to add or remove individual peaks The GammaPlus can perform some spectrum manipulation like stripping smoothing and normalization to factor The user s manual contains quite a number of mistakes and misprints 6 4 Gamma Vision version 2 3 USA The software GammaVision 1s an MCA emulation software together with a built in environment for high resolution gamma ray spectrum analysis software It comes in a set of installation diskettes and two user s manuals It has also a protection key through the parallel port This program operates under MS Windows 3 1 or later versions and had been designed under a powerful and easy to use graphical user interface 0 software can be operated from a network 47 graphical capabilities and the powerful user interface provide extremely user friendly and convenient way of operation particularly for fitt
88. n process the spectrum data produced by Nucleus Ortec Canberra or Silena data acquisition systems During the intercomparison Span program fail to read 8K spectra It got wrong numbers after 4400 channels This looks like a memory allocation problem f documentation on the spectrum analysis procedures is rather limited The parameters that affect the peak search and doublet resolution are briefly discussed in the user s manual The actual peak area determination is not documented The baseline is removed from the spectrum before the area calculation is done A peak shape is calculated based on the peak shapes stored during the calibration step after which experimental peak shapes table is created or using the Gaussian function depending of the user choice The fitting procedure can be controlled by several parameters Among others the following parameters can be adjusted for peak search 1 Peak Analysis Method There are two approaches involved in the code the Total Peak Area TPA and Gaussian Function Fitting GFF methods The limit of identification for closed doublet of equal intensities is about 0 5 times the FWHM of the components User could select either the TPA method or GFF method for all peaks both singlets and multiplets These two methods can also be selected by computer automatically using the option AUTO that means the TPA method is used for singlets and the GFF method for every multiplet 2 Error Limit
89. n this case the software search the presence of standards peaks for calibration The energy values are taken from the provided isotope data library The user has also the possibility to enter or erase any peak from the automatic calibration Energy values can be user defined as well The same goes for FWHM calibration Peak shape calibration is done as an asymmetry calibration calibrations can be plotted f program detects and skips automatically the 511 keV from calibration procedures The spectrum analysis is done first by peak search and then the peak areas are calculated by fitting the peaks The program uses the following formula for uncertainties Error Z c Area where Z is the confidence level and o the standard deviation The analysis options in the software provides among others Area Calculation which is peak area calculation in a ROI using classical background subtraction Peak Location which is peak search followed by fitting Several parameters can be set in order to control the peak search and the fitting procedures as for instance Peak search criteria number of times above o of the baseline for a peak to be accepted Confidence level the Z value to be used as described above Gaussian filter sensitivity for multiplets search on residuals etc 25 f utility Info Peak was found useful It can be called from the mam menu and performs a peak search around the current cursor position and the
90. n tools In other words GammaVision deconvolution operates in the following way Any peak region which has more one library entry or actual detected spectral peaks which would overlap 1s deconvoluted If certain criteria are not met the deconvolution may be aborted Any peak envelope for which no deconvolution 1s done but which has a FWHM or FH25M which 1s in excess of 1 2 times the calibrated value 1s flagged with a warning symbol Obviously this method has some advantages but also a number of disadvantages and 1t 1s evident that such methodology requires a provision of good nuclear data hbranes and relies on an accurate peak position routine For creation of the specific libranes the GammaVision comes with a separate software package called Navigator a type of Visual Nuclear Data Base representing the chart of the nuclides and their most prominent characteristics Out of this software the user can make his own specific application nuclear data hbrary It also can be edited through a provided library editor There 1s full compatibility between the gamma ray analysis software nuclear data library manager and editor and the application Nuchide Navigator Detection limits are calculated and can be chosen from 10 different types of formulas some with doubtful physical meaning 27 f There is no on line help provision in the software Activities and concentrations are calculated if all corresponding data exist Th
91. nalysis programs determining the total area of doublets with small separation For test of peak positions the same was also done 1f such two reference peaks were located within the position uncertainty reported by the analysis program Finally all the computed z scores were squared added and the total divided by the number of peaks 1n order to obtain reduced sums of squares as the final test results For the implementation of above described method a program CMPSPEC was written to perform these tasks As input this program read two files The file containing the reference data and the file containing the tested program s output To obtain the latter the original output results of the different programs had to be converted to a uniform format containing for each peak detected the peak position and absolute one standard deviation uncertainty the peak area and its absolute one standard deviation uncertainty By the program this report we are going to refer as a hit to an event when a reported value matched within the expected deviation a reference value A miss will be the opposite A false hit when a reported value 1s not in coincidence with the reference value 14 4 2 Peak area re normalization The first task performed by the CMPSPEC program was the determination of the normalization factors to be used for the correction of all peak areas for each program to compensate for possible bias 1n peak areas determination This s
92. nce in this intercomparison exercise we needed a first order energy calibration function which will result in only two calibration parameters In order to obtain that with Gamma Track we were force to use only two experimental points The dangerous doing this 1s that the FWHM and shape parameters had to be extracted from only two peaks 23 We have found out that calibration parameters can not be transferred from one spectrum to another The problem arises from the fact that any calibration data in the header of the spectrum takes precedence over data in calibration files The order of this precedence cannot be overridden 69 Also a menu function called Merge calibration data does not work if the spectrum is not previously calibrated and also fails if the calibration has been saved reporting a division by 0 error The fitting procedure can be controlled by several parameters The main sensitivity factor and here higher numbers mean lower sensitivity the number of peak insertion passes and the peak insertion sensitivity factor number of background channels for the evaluation of the baseline under the peak Help File Analyze Cale Setu Options Mode Quit 136K AMA A PST A an WI NNNM Id Straight Spectrum Area roi 2 17762 Scale 16K Peak centroid 186 331 Net peak area 17762 1 8 Display 64 Chn 462 keU 186 47 Cts 9843 Normalised Residuals 1 peak Fi help I
93. net z ms Hypermet i ND HS kuona PETER SHOE Ganaas Ganaas GammeW GammaVision ActAn 0 10 20 30 40 30 20 10 0 10 20 peaks peaks Figure 25 Numbers of detected small Figure 26 Difference between number of peaks on high background detected small peaks on high background and number of false hits 61 listed in the tables These numbers are shown in Figure 25 The number of misses and the related value are also related to this ability but less discriminating Most programs allow the user to set a peak search sensitivity parameter directly influencing the detection ability and the data in this figure therefore mostly reflect the setting of this parameter In an intercomparison a more interesting aspect of program performance is the ability to detect small peaks and not to detect spurious peaks i e to have a small number of false hits In Figure 26 the difference between the numbers of detected small peaks on high background and the numbers of false hits are shown As we have stated in the Introduction a detection of a spurious peak is not necessarily a disadvantage of the program as long as its peak area is reported with a high uncertainty as indicated by the 7 value for the false hits At the same time it should be pointed out that spurious peaks can result from unavoidable imperfection of the peak search algorithm as well as from incorrect multiplet deconvolutions Taking this into consideration
94. ng process may miss multiplet components or detect spurious multiplet components The lower the residual search threshold the less components will be missed but the more spurious components will be found From the graphs it 15 seen that the threshold was set to low for Sampo 90 and InterWinner by the user the 68 deconvolution power is very large but so is the X value for false hits The trade off between the two aspects turns out not to be the same for all analysis programs For example the trade off has turned out less favorable for Gamma W than for InterWinner As shown by Koskelo 10 and implemented earlier at IRI 23 24 an alternative for multiplet deconvolution is the determination of total area followed by the determination of the constituent peak areas in the interpretation stage This method can only work well if the total areas of the doublets in this test can be determined correctly by the analysis programs Also the programs should report a peak position uncertainty maybe multiplet range would be a better term in these cases that can be used as a search window in the y ray catalogue Since they don t search windows of 1 keV or 1 x FWHM are common practice unnecessarily complicating the interpretation process To establish the quality of the total area determination the results from the 1 x FWHM match criterion statistical comparison runs were used for Figure 37 and Figure 38 otherwise identical to Figure 31 and Figure 32
95. nimum and 2 Min 2 Min 1 not Min 3 and Min 2 and Min Only Resolution cal User defined Three Max and Maximum number of 3 limited for Max 20 11 one parameter performed Max 5 order expenmental Min 3 A parameters used for Automatic Max 1 used for automatic using Automatic Depending peaks kind of Resolution calibration mode shape calib two parameters mode of on the to perform automatic provided which i5 calibration function FWHM mode performed provided calibration provided automatically Minimum and Min 2 Dependingon not Min 3 and 1 Min 2 Depending of Depending User defined Depending Min 1 Maximum number of the function limited for Max 60 Maximum type of function of type of order of type of Max 12 but parameters used for Max depending of function Depending function depending on Eff calib available data on the the exp data points function Functions available Several Log of cubic Divided in Several Several Several Selectable degree Ln E Interpolation Selectable Polynomial for Eff calibration Polynomial polynomials two regions Excellent Polynomtals of Polynomialsof of ortho polynomials diff Types degree of function of function of and inverse interpolation Choice and Ln e snd l E Ln 1 normalised of different polynomials ortho Ln E or of Ln E or at exponential linear capabilities of different of different polynomials to degrees Ln E po normalised 1 8 func
96. ns F8 insert Del remove peak SPACE fill on off Esc finish Figure 10 Oxford Gamma Track Fitting of the 186 keV peak 0 The fitting operation for a simple spectrum in the default mode using default parameters was found to be fast and effective except for the region of 511 keV As seen from Figure 10 the splitted windows are properly used The programs also display the residuals Another improper operation of GammaTrack was found during the intercomparison and is related to the difficulty to save the analysis results in a file The Peak Search menu has only the options to display on screen and print If the unfortunate user chooses to print but no printer is connected he will be caught in a catch 22 situation between Retry and Ignore whose only door out is rebooting or killing the DOS session if running under windows with the loss of all the results 1 way to save the results into a file creates some confusion The file name where results should be saved has to be set in the Automatic analysis parameters as report device also the report template has to be set as T The implementation of the reports of the results was found to be very flexible and powerful through the template definitions of reports but sometimes complicated The software complies with quality assurance and quality control rules The software comes with a 266 page Operation s Manual Most of the
97. nt It gives the choice of full or custom installation It creates sub directories if needed and it prompts for any replacement of a system or program files for the user choice The software is eventually not protected but it will print save any spectrum analysis only with the presence of an APTEC MCA card This might not be a nice option for a user having purchased the package who wants to analyse gamma ray spectra in a computer where an APTEC MCA is not installed f It can import spectrum files from many commercial and non commercial companies or type of formats including VAX format 69 IAEA spectrum format conversion failed In addition IAEA type of format is listed with default file extension which is wrong ff The OSQ Professional has an outstanding user interface It is colorful convenient and fast Mouse operation is easy and most essential operations can be controlled by key strokes Almost a wide spectrum information can be displayed In addition to the spectrum ID the preset time true time live time dead time as TT LT TT in 96 gross count counts sec start and stop times start and stop dates and user s Id are displayed too 2 Although the programs window can be splitted horizontally or vertically for seeing another region of the spectra or a zoomed peak full implementation of multiple windows is absent 38 Meaning several spectra or more than two parts of same spectra can not be viewed at the sa
98. ntrol Area Estimation STRAIGHT spectrum STRAIGHT spectrum Inter Winner i hnter Winner om g user settings Span default settings GammaPus Sampo90 GammaTrac Hypermet GeniePC Ganaas Gammaw GammavVision 0 1 1 10 100 1 10 100 reduced sum of squares reduced sum of squares Figure 28 x values based on the Figure 29 values based on reference uncertainties reported by the uncertainties for each program programs in their two modes for in its two modes the straight spectrum In Figure 29 the 7 values based on the reference uncertainties are plotted for the programs in their different modes It can be seen from the graph that the performance of most programs remained more or less the same when the change from default to user settings was made In most cases the user settings lead to slightly better results This implies that the default settings of the analysis programs are suitable for the analysis of singlet peaks might indicate that the information offered by the corresponding programs to judge the quality of the results can be improved In Figure 28 the x values based on the uncertainties reported by the programs have been plotted Although it can be seen that most x values differ from unity significantly implying that more strengthening and development is needed in this aspect in the future however it also can be concluded that all programs except one demonst
99. old 125 100 100 2 x o 8 5 75 g o o o 198 5 Numberofhits 5 10 I Number of false hits 5 X of false hits 50 Se X of hits reference uncertamties E c 8 a 25 4 s 3s a m m 0 2 0 4 06 08 1 0 12 Match criterion FWHM Figure 34 y based on reference uncertainties and the total number of matches as a function of the peak position match criterion in terms of FWHM Lines were drawn to guide the eye 67 To investigate the problems with the addln1 spectrum the output of the statistical comparison program was studied and 1 was found that those cases where a doublet has been mistaken for a singlet most of the programs report a singlet that does not match either of the doublet components to within 0 5 x FWHM or the reported position uncertainty as illustrated Figure 33 This 1s most likely to happen on the low energy side of the spectrum where the doublet separation 15 12x FWHM Additional runs of the comparison program were performed where the criterion for Resolving Power False Hits STRAIGHT and ADD1N1 spectrum ADD1N1 spectrum Inter Winner InterWinner Span Span E GammaPus GammaPlus Sampo90 Sampo90 GammaTrac GammaTrac Hypermet Hypermet GeniePC GeniePC idi o T Gamrmav ision
100. on PC Professional Gamma background integral of the multiplets Linear and no documented peak Stepped or function parabolic provided as functions well Special treatment of No No No No No Yes Yes No No Yes No No the 511 keV peak possible to fitit manually outside the FWHM calibration Automated mode of Yes Yes with Yes Yes Yes Yes Yes Very Yes Yes Yes very No Yes analysis with defaults reasonable good good parameters available good performance performance performance at spectrum for simple of singlets spectra Threshold for peak Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes search User selected Yes but not Yesand well Yesand well Yes Yes Yes but not Yes well Yes But Yes Very Yes Very Yes but not Yes Not threshold peak search so well documented documented effectively well documented input expert well well documented documented parameter and well documented explained documented mode badly documented documented No documented documented recommended value provided Force manual peak No Yes No Yes Yes No Yes Yes Partially by Yes No No insertion mode inserting implementation manually a ROI Can a peak be added No Yes No Yes No No Yes butextra Yes No Yes No No or removed manually software has multiplet 7 requested Peak area and Net area Net peak Net peak Net peak Net peak area Net peak The are is Netpeakarea Net Gross User defined Net peak area Net Gross standard deviation obtained from area and area standard
101. on of concentrations in terms of Neutron Activation Analysis for both thermal and fast neutrons 1s provided Peak interference 1s also implemented while working in this mode T software does not provide help on line and the user s manual was found to be not complete and not updated 6 7 GeniePC version 2 2 GeniePC 1s a networking data acquisition and gamma spectroscopy system that runs under the IBM OS 2 operating system The package has 2 different parts e Basic spectroscopy for data collection calibration of spectra and simple analysis which includes automatic peak deconvolution 3 diskettes e Advanced options this s provided 2 additional diskettes titled gamma analysis and interactive peak fit This design 2 parts has some drawbacks particularly all the gamma analysis modules with the exception of the interactive peak fit operate a highly non interactive way Results are not updated in the spectrum display windows unless the spectrum data set 15 manually reloaded Even performing the energy calibration requires to run a new program which however in this case allows to take the ROIs selected the spectrum No access to library energies 1s available during calibration For the test we used a computer where OS 2 Warp version 3 was already installed The installation of OS 2 will therefore not be discussed however previous experience shows that installing OS 2 can be quite comp
102. on original diskettes as provided by the manufacturers together with protection keys if there was any and the full documentation as t 1s distributed commercial basis or the documentation as received from the manufacturer for the participation in the intercomparison No special made software was accepted for the intercomparison Each participant was asked to spend at least 4 hours familiarising himself with the manuals before attempting to install the corresponding software 1n his assigned computer The computer capabilities and configuration were also carefully done order to match or exceed the requirements of the manufacturers As a separate exercise programs were tested in differ drivers e g double space drivers like Stacker running with many TSR programs in RAM SCSI drivers etc This was done only with the purpose to test the stability of the operatio See Contributors to Drafting and Review for complete list of participants See Appendix for complete list of tested software packages and their providers 2 After the programs were installed in the respective computers first energy and shape calibration was performed using the spectrum CALIB and a reference list of peaks as follows Channel Energy keV 301 122 06 1281 511 00 1661 661 66 2097 834 84 295 1173 24 3207 1274 54 3353 1332 50 The reference lists specify peak positions in terms of energy To separate the problem of peak
103. peaks In all these 12 cases the peak was observed visually and thus not removed from the list The area and uncertainty reported by the program that did detect the peak were taken as reference values The last problem 15 the 511 keV peak Its area was decided to be determined by integration The resulting reference list was subsequently used to produce the corresponding reference lists for the test spectra by applying the identical shifts and dividing both areas and their absolute uncertainties by the ratio between the counting times of the test spectra and the reference spectrum CHAPTER 4 THE METHOD To evaluate the performance of the tested software and to represent the results within the scope of the objectives explamed Chapter 1 Introduction a method for processing of the results was adopted Here it 1s explained detail The aim of the tests for each program were quality of the peak area determination quality of the estimation of the uncertainty same for the peak position determination e the ability of the program to detect peaks and to properly weight them e and ability to resolve multiplets All above for different type of gamma ray spectra 4 1 Data handling The overall intercomparison of the results was based on the statistical qualifier z scores or standardized residuals This 1s the differences of the magnitude reported against the one referred divided by their own uncerta
104. quite well by all programs except GammatTrack and all programs are in reasonable statistical control with respect to the peak areas positions and their uncertainties Very good results were obtained with Span GammaPlus Hypermet PC Ganaas OSQ Professional Sampo90 and Gamma W in no particular order Doublets at separations below 1 2 x FWHM often are mistaken for a singlet by all programs and reported with much too small peak position uncertainties even if the settings of the program allow it to perform residual searches and add multiplet components in the fitting procedure With respect to resolving power as defined in this test and quality of area determination ActAn Gamma W and Sampo90 yielded the least bad results Only after the consequences of peak position uncertainty underestimation were removed by a change in the test procedure Ganaas Hypermet PC GammaVision GeniePC and Span proved to determine the constituent peak areas or the total peak areas in case of an unrecognized doublet quite well and also proved to be in statistical control The results obtained with user defined settings are usually better than the results obtained with the default settings In those instances where this is not the case the analysis program apparently can be improved with respect to the information offered to the user to judge the quality of the results e g plots of residual patterns The report of the analysis results has been improved by almost all
105. r Pileup rejector off ADC Canberra model 8077 with the following settings Range and Gain 8K Offset 0 Peak detect AUTO PHA mode No coincidence condition MCA Canberra S 100 plug in card in PC with an Intel 25 MHz 80486 CPU Resulting in a set of spectra as follows Table 1 Set of measured spectra and their specifications Spectrum Real time Remarks name in seconds SD0000 MCA 2000 Automated counting with Canberra 5100 software through SD0019 MCA SD0020 MCA 2000 For comparison if needed This spectrum was measured 6 hours after termination of the above automatic task SD0021 MCA 667 SD0022 MCA 200 SD0023 MCA 20 SD0024 MCA 2000 SD0025 MCA 2000 Another automatic task through SD0029 MCA SD0030 MCA 2000 Same set up as before but 37 kBq of additional Co on the endcap 500031 2000 Same as before but with several additional sources present SD0032 MCA 6000 A set of 37 kBq each calibration sources measured at 169 mm distance from the endcap SD0033 MCA 40000 The 226Ra source at 178 mm distance from the endcap High voltage turned down to 2500 Volts SD0034 MCA 2000 The 226Ra source at 178 mm distance from the endcap High voltage at 4500 Volts An additional 241 source with ca 10 mCi at ca 280 mm from the endcap The detector head was situated inside a small castle with 50 mm average lead thickness The inside of the castle was covered completely with 0 5 mm of copper The source was placed in an acrylic
106. r high significance peaks that 15 the calculated uncertainty 1s higher than reference one therefore a lower value of related chi square 1s obtained Another case for a different category Number of matches for small peaks on high continuum 7 related chisqr for areas and reported uncertainty 2 7 and for reported areas with reference uncertainty 2 4 The number of matches corresponding to SMALL PEAKS category lying on the high background 3 0 Net_Area lt GROSS are reported Agam here the mark denotes that the deviation of the calculated chi square from the expected value unity 15 significant according to the chi square distribution The value of 2 4 was obtained when the Z values were calculated using the reference uncertainty The difference between Chi Squares 2 7 2 4 1s small thus meaning that this case of small peaks on high background the uncertainty calculation of the program 1s accurate Number of matches for small peaks on low continuum 6 related chisqr for areas and reported uncertainty 2 2 and for reported areas with reference uncertainty 2 1 The number of matches corresponding to SMALL PEAKS lying on the low background 3 0 Net Area gt GROSS are reported In this example a value of 2 2 was obtained for x reported by the program The value of 2 1 was obtained when the Z score values were calculated using the reference uncertainty The difference 2 2 2 1 1s very small Besides both chi squared values insid
107. rate to be m reasonable statistical control influenced to some degree by the user settings From the statistical control and the quality of the area determinations an uncertainty overestimation factor f can be computed using f ref 5 XL The resulting values are shown in Figure 30 Most programs turn out to overestimate the uncertainties they report relative to the optimum uncertainties from the reference list Relative to the actual discrepancies between reported areas and reference areas however the reported uncertainties are reasonable as Figure 28 illustrates Uncertainty Estimation STRAIGHT spectrum InterWinner Span WS user setings g default settings overestimation factor Figure 30 Uncertainty overestimation factors f Even though most programs do not report peak position uncertainties but merely imply them to be 0 01 keV with two digits after the dot the statistical control in this respect is quite good as can be seen the tables 511 keV peak area is determined correctly only by Ganaas GeniePC and Inter Winner even though a 511 keV peak was present in the calibration spectrum 65 The distort spectrum was analysed well by GammaPlus Gamma Vision and Gamma W None of the other programs apparently more sensitive to the change in peak shape reported to the user that such a change had occurred In principle this feature could
108. rding to the definitions given n the documentation of the programs The second stage of the comparison was to analyse the STRAIGHT and multiplet ADD spectra in what we called the Automatic Mode using the calibration above mentioned This means a non interactive analysis based on the default parameters for spectrum analysis provided by the producer of each program or 1n the absence of default parameters the values suggested 1n the documentation of each program n the Getting Started or equivalent chapter The idea was to run each of the programs in the same way that an inexperienced end user or beginner would be expected to do so The results were saved as ASCII files and processed by the comparison program described latter this chapter The third stage was to analyse the same spectra again in non interactive mode but this time using the best possible set of parameters for analysis The only restriction being that energy and shape calibration data should come from the test spectrum This part of the test tries to emulate as much as possible the results that an expert user would get from a large set of spectra when an interactive analysis would not be possible The STRAIGHT spectrum was used to optimise the different analysis parameters by experimenting with different values and checking the results The same parameters were applied to the ADD spectra To chose the best possible set of parameters the documentation provided was of course
109. ric quantity and does not require any secondary calibration It 15 known that different computer programs define channels numbers differently starting to count either from zero or one For the generation of spectra with multiplets the following spectra were used SD0024 MCA until SD0029 MCA 6 spectra measured for 2000 sec each SD0021 MCA One spectrum measured for 667 sec SD0022 MCA One spectrum measured for 200 sec SD0023 MCA One spectrum measured for 20 sec Doublet spectra were generated through shifting a spectrum by either 3 channels 3 channels and adding it to another spectrum channel by channel The following added spectra were generated Table 2 Generated spectra The newly generated spectra were sidus m an ASCH format as one channel content per line The first channel of each spectrum contains a counting time which 1s the sum of live counting times of the two mdividual spectra which were summed together the second channel contains the real time It should be noted that this counting time has limited spectrometric significance The final set of test spectra was conformed as follows CALIB ASC a calibration spectrum taken from weak point sources ca 37 kBq of Co Cs Mn and Co Small contributions from other natural radionuclides are also detected 1n this spectrum originally measured as SD0031 MCA STRAIGHT ASC a plain 226Ra spectrum measured for 2000 sec real time This spectrum may be used for any kind of
110. ript S Fazinic and Osorio of the Division of Physical and Chemical Sciences were the IAEA officers responsible for this publication EDITORIAL NOTE In preparing this publication for press staff of the IAEA have made up the pages from the original manuscript s The views expressed do not necessarily reflect those of the IAEA the governments of the nominating Member States or the nominating organizations Throughout the text names of Member States are retained as they were when the text was compiled The use of particular designations of countries or territories does not imply any judgement by the publisher the IAEA as to the legal status of such countries or territories of their authorities and institutions or of the delimitation of their boundaries The mention of names of specific companies or products whether or not indicated as registered does not imply any intention to infringe proprietary rights nor should it be construed as an endorsement or recommendation on the part of the IAEA CONTENTS CHAPTER I INTRODUCTION erecta ce prit Dena aS ua ua aaa CHAPTER 2 THE TASK wqossaeesnosoeqoss 2 1 Procedure for testing the spectrometry programs CHAPTER 3 THE TEST SPECTRA 3 1 Definition of the 1
111. s in the peak areas they report However some programs are less over confident than others Most programs perform better both with respect to peak area and to uncertainty with the user settings In some cases the user settings result in a change in y7 values of more than one order of magnitude With the default settings the corresponding programs only 66 integrate peaks e g GeniePC OSQ Professional or do not perform a residual search e g Sampo 90 GammaVision The reason that the GammaVision results do not improve with user settings is that this program will deconvolute multiplets if first the constituent peaks are in its gamma ray library see Chapter 6 or second detected by the peak search algorithm It was decided that supplying this library to the program would give it an unfair advantage over the other programs Such library based multiplet deconvolution potentially could yield very stable results as shown in the past for X ray spectrum analysis v Espen 22 The remaining programs that yield more or less equally good results with default and user settings 98593 492 0 0 98593 492 Figure 33 Section of z score table showing a doublet mistaken for a singlet reported with such a position uncertainty that the reported singlet matches neither of the doublet components always fit and deconvolute peaks but the user may extert some control over the process by setting parameters such as a residual peak search thresh
112. s were determined from a spectrum with a counting time 20 times longer than the reference spectra b was estimated from the reference peak area and its uncertainty using _ 200 A 2 b 4 e match All peaks belonging to the previous three categories i e high peaks small peaks on high and small peaks on low continuum e Misses Peaks in the reference list not matched the reported list of the programs for of above categories e False hits Peaks in the reported list of the programs not matched the reference list Total previous categories except for the annihilation peak A separate X was computed for peak position from the differences between reference peak position and program output peak position and their uncertainties in terms of energy analogous to the x for peak areas CHAPTER 5 WHAT DO THE NUMBERS MEAN Now that all the methodology has been explained and the tasks to be done were described probably you got confused on the ocean of areas uncertainties z scores and reduced x So lets have a look to all of these using examples and some data Before anything keep n mind the formulas described the previous Chapter This are formulas 2 3 mainly Take a piece of paper and write them and have them front of you So after processing the different test spectra we obtained a set of output files from the tested program This set 1s processed by the program C
113. score 1s calculated using both the reported uncertainty Z and the uncertainty from the reference spectrum 2 The fact that values for Z scores are nonzero means that the reported and reference areas are not the same As long as the difference between the reported area and the reference one are in statistical control the calculated Z scores should be low average value 0 standard deviation 1 Otherwise the program 15 penalised because of an erroneous peak area calculation Case No 3 The area for this peak in the reference spectrum 15 0 That means at 230 6 keV was considered not to be any peak n the reference spectrum This category was classified as False Hits If the reported uncertainty of the program for that false hit peak 15 large then the contribution to the Zp will remain low Illustrating one of the main ideas of the intercomparison see Introduction Chapter that 15 1 is not as bad whether the programs detects or not a peak as much as it detects it with 1ts corresponding uncertainties That 15 telling to the user look there might be or not a peak 1 s up to you to decide and be careful extracting any information from it In the opposite case the program that finds a doubtful peak and reports t with great conviction small uncertainties should be heavily penalised because of such bad habit This case gives no contribution to the Zer value Cases No 4 and 5 This 15 an interesting example where a false
114. table function the manual tail on the side side Are user defined peak No No The No No No No But No No No No No shape functions asymmetry of altemative possible peaks 1s fitted fitting as functions are polynomial provided the degree of it can be changed Algonthm for fitting Uses stored Non linear fit Least square Least square Least square Non linear fit Non linear Least square In case Non linear fit No peak shapes poss fitting fitting fitung using Fit fitung singlets linear information from the Modified Marquardt fit default available calibration to Marquardt algorithm Non linear fit fit the peaks Chi squares are minimised Baseline functions A Parameter Not known For singlets In each region Smoothed Various Step function Linear or Linear or Total Peak No combination less step Not automatic 5 is calculated Step options folded with Least square parabolic Area Method information of iterative function documented 3 points accordingly function Not provided the Gaussian fit of up to 5 based on a available smoothing calculated at all average documented Default is a degrecs Step function and a step from scaled For step function Not well 10 where the number of baseline counts in each channel is analytically determined from the experimental number of peak counts in all higher channels SS Act n Gamma Gamma Gamma GammaW Ganaas GeniePC Hypermet OSQ Sampo90 Span Winner Track Plus Visi
115. tainties can be reported in the end The report can be user defined and can contain any parameter used in the program It has an ASCII format and can be printed or written to file The graphs can be sent to a printer directly If the user wishes to create a report suited for it QA related information can be printed in the report f User s Manual according to the authors not complete was found to be well organized with excellent readability containing tutorial section to start procedures section for more advanced operations commands section as reference Algorithms are clearly explained and access to parameters pointed out It just lacks an Index section to be an excellent manual 42 6 11 Span version 5 1 China SPAN v5 1 is the multipurpose software package 21 used for gamma ray spectrum analysis and neutron activation analysis developed at the China Institute of Atomic Energy The package consists of one 3 25 HD installation floppy and a 54 page user s manual The software was designed to run on a standard PC computer under the DOS operating system Installation is difficult probably due to the disk protection scheme used by the code 60 Moreover automatic change of the autoexec bat and config sys files not always work We were DX 25 which used STACKER disk compression software Installation was repeated three times before we finally got a working version of the code software ca
116. tep was needed because of the of different methods for peak area determination used by the programs For this task the following steps were performed e Thereference list and the measured list were matched e Then using all hits for the case of the straight spectrum a weighted average and its uncertainty of the ratios of reference peak areas and program output peak areas were determined As its uncertainty the internal standard uncertainty of the mean was taken The weights used were the imnversed squares of the uncertainties in the area ratios computed from reference and reported uncertainty e Peak area ratios differing from umty by more than 0 1 assumed to be the results of incorrect deconvolutions were excluded from the average Even though this uncertainty 1s the smallest that could have been taken within the rules of error propagation the ratios found were all statistically unity as can be seen in the Figure 27 Since none of the re normalization factors found for the straight spectrum deviated from unity significantly as will be discussed later the re normalization factors were applied to the program output peak areas only in the case of the distort spectrum where the peak areas themselves were expected to be biased due to dead time For this spectrum a dead time relative to the straight spectrum of app 9 was thus established 4 3 Computation of standardized residuals and reduced sums of squares t
117. th lines up to 2448 keV sources are sealed sources which have passed the prescribed wipe tests 226Ra source was also tested for the emanation of radon and no significant outgassing has been found It is finally agreed that the test spectra should be measured from the 226Ra source at a large distance between the source and the detector endcap The spectrum measurement scheme was set up n the following way 20 spectra were measured for 2000 seconds real time each from which was determined the stabihty of the measuring set up The sum spectrum of these 20 was used to generate the set of known peak positions and relative peak areas calibration spectrum from commercially available radioactive standards sources was acquired which have a simple photon emission pattern This spectrum can be used for any kind of calibration purpose required from the analysis programs The calibration spectrum was decided not cover the whole range of energies as found in the test spectra to force the testing of the physical meaningfulness of any calibration curves determined and used by the analysis programs set of spectra was measured for 2000 seconds real time each which was used later to perform the summing of multiplets No spectrum was used twice the summing procedure e Several spectra for known real times of 667 seconds 200 seconds and 20 seconds were acquired which were added with 3 channel shift to one of the above 200
118. the programs Many of them had implemented the multiple choices or user defined type of reports Excellent implementations were found in GeniePC Sampo90 OSQ Professional and GammaVision Even though some of the programs obviously need more improvement than others no program emerges as the best from this intercomparison The user may select the program that is the most suitable for the specific y ray spectra to be analyzed or the most flexible if different kinds of spectra are to be analyzed 70 1 2 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 REFERENCES H Houtermans K Schaerf F Rechel K Debertin Intercomparison of High Precision Gamma ray Spectrometry Internal Report G 2 IAEA Vienna 1980 Parr et al The IAEA intercomparison of methods for processing 1 gamma ray spectra Presented at Conference of American Nuclear Society on Computers in Activation Analysis and Gamma ray Spectroscopy Mayaguez Puerto Rico May 1978 ARCAL II Report on a workshop on gamma ray analysis software ARCAL II Internal Report May 1991 Costa Rica Sanderson Evaluation of Commercial IBM PC Software for the Analysis of Low Level Environmental Gamma Ray Spectra International Committee for Radionuclide Metrology W renlingen Switzerland 1987 L H Christensen K Heydorn J Radioanal Chem 113 1987 19 K Heydorn L H Christensen J Radio
119. their initial positions When all the fits have been made for each region then the results can be saved to the corresponding cnf file A list of the results can be obtained using a Report procedure C It 1s interesting to point out that during the manual or interactive analysis of the STRAIGHT spectrum some peaks that appear to be incorrectly fitted 1ndependently of the values of the parameters without changing the analysis algorithms were found The images bellow illustrates these cases 34 0 Peak Lu rc Zoom nzoom Setup 0 Fit i Filter 67 Plot Report In this case a peak smaller than the reality was fitted see residuals Adding another peak next to the onginal one did not improve the situation When it was attempted to fit this peak again individually by using the energy 1280 keV as filter the program refused producing an error as seen in Figure 15 Error dii After this errors t was necessary to reboot E OS 2 m order to get normal operation again Number 35425 Apparently the interactive fit procedure remamed active in the background making OS 2 unstable Figure 15 error code produced when trying to insert a manual peak The programs of this package are well documented as well as the form
120. tical control and stability of reported results The idea is It is not so important how successful the program is in finding or not a small peak in different situations as much as how the found peak is reported and how stable is that finding Meaning if a program finds a small peak on a large baseline it is very difficult to judge its existence in this case it is more important to have it reported with corresponding large uncertainty A program that missed such a small peak should not be penalised as much as in the previous intercomparisons because such a peak is doubtful and the amount of information that it might carry can be poorly extracted The above mentioned phrase also means that it is important that results are correctly reported in statistical meaning and that a program should operate and give stable results These were the syllabus and objectives of our task In this report the task methods and results of the intercomparison are presented in order to assist the potential users of such software and to stimulate the development of even better y ray spectrum analysis software A companion diskette is attached to this TECDOC It contains the set of test spectra and the software for processing the test results CHAPTER 2 THE TASK 2 1 Procedure for testing the spectrometry programs The first stage of the intercomparison 15 the preparation of the test spectra This 1s described in detail in the next chapter Howev
121. tion quadratic and degrees degrees the data Polynomial And SORT polynomials polynomial Parametrisation of 1 E at pol for three Bellow function ts diff degrees energy interpolation provided for Cubic Tegions linear and calculation of Spline quadratic eff using one exp point 129 Spectrum analysis ActAn Gamma Gamma GafiimaW Ganaas GeniePC OSQ Sampo90 Spas Winner Track Plus Viston 4 Professional U 222 N MEA 2 15 the spectrum Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes analysis peak area determination performed by peak search Is the spectrum No Yes Yes Yes No No Yes No Yes No No Yes but you analysis peak area have to buy determination the nuclides possible by specific library option radionuclide determination Fitting functions Parabolic Gaussian Gaussian Gaussian Gaussian full Gaussian Gaussian Linear sum of Gaussian with Gaussian with Gaussian or No available interpolation according to with a tail energy with an when is Gaussian and tail two tailing on fitting based information of an the manual distribution exponential set to Gaussian parabolic or on available experimental with a fifth low energy and Gaussian with linear experimental peak shape power low tail Not with exponential continuous peak shape tables energy tailing described in exponential tail on the left
122. to ASCII not to a file into a file file ASCII file ASCII ASCII file file Both easy to import ASCII file save no a file easy to import propnetary ASCII format easy to easy to import propnetary ASCII ASCII since it contains format only possible import format format CAM format format special files of ASCII supported characters format files are provided Easy to export Nuclear Data Libraries ActAn gt Gamma Gamma Gamma Gamma W Ganaas GeniePC Hypermet OSQ Sampo90 Spar Winner Track plus Vision PC Professional Gamma RP OE A AOA SSS cn Is there isotope Yes Yes Yes Yes Yes Yes Yes Yes but very Yes Yes Yes Yes but as nuclear data library limited optional item available Can the nuclear data Yes An Yes Library Yes Yes see Yes Yes Yes Ithasa Yes Butno Yes Yes Yes Yes An useful library modified automated editor 15 notes very powerful Inside utihty to hbrary editor edited or created by way of integrated library editor perform it is 15 provided the user entering peak into the main provided information 15 program Editions to the provided library should be done with an extemal ASCII editor Provision of specific Yes NAA No No Yes Yes Yes NAA Yes upon No Yes several No No Yes several apphcation onented request isotope library LS Reports ActAn Gamma Gamma Gamma GammaW Ganaas GeniePC Hypermet OSQ Span Winner Track Plus Vision PC Professional Gamma
123. to assist TC projects and nuclear analytical laboratories starting services based nuclear analytical techniques as well as to provide a common based and comprehensive software for several IAEA training courses and other training activities GANAAS software package 15 oriented towards the analysis of high resolution gamma ray spectra activities and element concentration calculations for Neutron Activation Analysis NAA 5 Both Thermal and Fast Neutron Activation Analysis 29 The Ganaas software 1s flexible and easily manageable The program allows intensive dialogue with the operators This 1s more applicable while doing calibrations t mstallation 15 quite simple and effective It 15 completely driven by a specialized installation program This program allows to install the main software called essential module and gives the user the opportunity to mstall only those modules that he will use If necessary 1t creates new sub directories and file structure The installation package also offers a set of examples programs Although it works fine the installation procedure lacks of some flexibility e g the name of the target sub directory 1s fixed to Ganaas and can not be changed 1 Although very simple to operate the installation program s not documented Only a brief explanation is provided in a text file included in the first installation diskette Ganaas 1s a DOS operated software It has an strong requirement on RAM
124. tore in final file or report Menus and dialogues can change or display the different options Gamma Vislon Menu and dialogue boxes Values are stored in files and re called by software when requested GammaW through menus and files Ganaas several parameters files are kept for reproducibility of results GeniePC are stored into a file and a database is created A full set of all operational parameters are stored into a files and a comprehensive database corresponding to the experimental set up i created Hypermet PC Menus and dialogue boxes Spectrum parameters are stored into results files 050 Professional They can be viewed or changed any time menu and dialogue boxes Extensive headers are provided and can be called at any time They can be included reports Sampo90 into the program All parameters are kept and can be reported on the user s wish Span info on calibration used See above Winner Gamma 15 built into the program Menu and dialogue boxes Parameters values are stored in file and results report if chosen option A database can be created 09 Help and Documentation ActAn Gamma Gamma Gamma GammaW Ganaas GeniePC Hypermet OSQ Sampo90 Span Winner Track Plus Vision PC Professional Gamma SO ____ __ _ _ _ _ _________ _ PERE lt
125. true as long as coincidence summing does not play a part 62 A consistent bias in peak area determination would affect the apparent peak efficiency of a detector But where the peak area of a sum peak is given by the product of the photopeak efficiencies for the contributing photons it would not be given by the product of the apparent photopeak efficiencies the area determination bias factor would come in only once not as many times as there are contributing photons As shown in Figure 27 the programs tested yield unbiased peak areas within the statistics of the test the only outliner at 0 01 being GammaVision as well as within a reasonable range of 1 No problems need therefore be expected when corrections for coincidence summing are to be computed from efficiency curves determined using one of these programs remainder of the intercomparison some of the statistical comparison runs were performed after peak area renormalization It was found that the results only changed marginally 8 3 Interpretation of the values The values computed in this intercomparison can be interpreted as reduced chi squared values A value of unity indicates statistical control a value less than unity overestimation of uncertainties and a value larger than unity underestimation of uncertainties The z scores underlying the x values however are not normally distributed in those cases where incorrect deconvolution occurs very h
126. tting and peak search parameters that the user might select these parameters not well documented in the User s Manual t The report of the spectrum analysis is very complete it gives the peak positions in terms of channels and energy the corresponding FWHM calculated area and its uncertainty and the quality of the fit The results of the analysis can be seen in an ASCII scrolling table or visualized in the entire or individuals peaks The spectrum data are presented in a histogram form in white The fitted curves are displayed in red and the fitted backgrounds are in green colors Each peak can be seen together with the residuals The residual values appear below the fitted peaks providing impression about the goodness of the fit However color scheme of the program can not be user defined or changed the operational and calibration parameters together with the spectral data and spectrum results are stored in different files It would be more convenient for tracebility purposes to have them all in one single file 7 The program does not provide the option of manual insertion deletion of peaks Ganaas provides a nuclear data library editor with the use of which the user can extract his own library from the master library 31 Ganaas suports a wide range of calculation modules and options For example it can calculate activities and concentrations as well in an interactive or automated mode Calculati
127. twork protection but distribution operation prints results diskette 6 Gamma Gamma Gamma GammaW Ganaas GeniePC Hypermet OSQ Sampo90 Span Winner Track Plus Viston PC Professional Gamma only with the Maximum presence of an of three Aptec MCA installations card allowed Comments Almost all the software packages do not strictly require a color monitor for their operations but the fact that areas fitting results residuals and many other information are given in different colors implies the requirement of color device for best performance and view the case of the software based on Windows OS the required amount of RAM 15 bounded to the requirements of the operating system itself Thus here the recommended amount of RAM for a reasonable fast operation of the software 15 given Some times the microprocessor requirements for a software of those of its operating system the minimum microprocessor requirement for software packages runnmng under MS Windows 3 1 are those minimum hardware requirements for the MS Windows itself in this case microprocessor 386 15 the minimum option Programs running under DOS hike GammaTrack and Ganaas have a strong requirement of free RAM available the failure to meet that requirement will end m non operational software For those programs sometimes it 15 compulsory to have DOS RAM free as much as possible limiting the capabilities of running TSR programs 06 Graphical capa
128. uations and data points are reported but uncertainties in peak positions are shown only in graphical way No uncertainty of the coefficients 1s reported If the calibration report is produced using the report utility together with the appropriate template and section then the uncertainty in the channel position 1s reported but again not the uncertainty in coefficients or in calculated energies The analysis of a spectrum GemePC consist of various methods which run a predefined order These methods can be selected by the user and are classified in groups like Peak Search Area Fit and Report Other groups as Acquisition Calibrate Area Correction baseline subtraction Nuclide Identification etc are also provided All the procedures necessary for analysis can be set into an Analysis Sequence which may be prepared with the Analysis Sequence option of the Edit menu The software provides full control of the analysis by setting several parameters as shown 1n Figure 14 All settings of the parameters are well documented and logically accessible through screens and menu options f GemiePC provides a flexible type of reports for the results The contains of the reports can be user defined The type of the report can be chosen with the help of templates Although the report using the standard template 1s quite complete 17 GemiePC provides the possibility of an interactive peak fitting 33
129. ulas used in the calculations The documentation of the programs that constitute Genie PC 1s quite good although perhaps 15 too extensive consisting of various thick manuals Fortunately due to the high quality of the user interface and the context sensitive online help available it 15 not strictly necessary to read the documentation for using the program 35 6 8 Hypermet PC version 4 00 dapest Hungary Hypermet PC v4 0 1s the PC version of the original HYPERMET code 18 developed by Philips amp Marlow in 1976 at Naval Research Laboratory Washington DC USA for automatic and efficient analysis of the complex gamma spectra The current version has a short preliminary User s manual 29 pages which 15 represents the minimum survival guide for operating the code in automatic mode It includes the following chapters Introduction e Starting HYPERMET e Initialisation of the Hypermet PC parameters e Reading files changing the display appearance Fittmg spectra and managing fit results e Additional options e The Nuclide ID option 22 User s Manual s really very preliminary and does not cover explain all the possibilities of the software For instance there was found a very poor almost none explanation of fitting options EXPERT MENU which 15 extremely important for the orientation of the user Installation went smoothly but it 1s manually done The user should copy himself the program files from f
130. user defined Most of polynomials are based on the log E log E functions 0 As well as the usual method for efficiency calibration in Ganaas there is a new feature in this package that make it unique compare to the others software It is the provision of a possibility to calculate the efficiency of the detector based on only one experimental point and the characteristics of the detector itself that the user has to enter before hand Results of the efficiency calculated in such a way have been proven to be very effective and convenient in some cases were there is not sufficient data or time to measure the efficiency curve The calculation method is based on information published by Gunnink 19 The main spectrum analysis routine of Ganaas is based on a PC adapted version of the well known program GAMANAL 20 In comparison with the other similar software its most characteristic feature is the way Ganaas treats the continuum in the gamma ray spectrum Instead of using a polynomial function together with the peak functions in a relatively narrow energy region the Ganaas first determine the continuum background for the entire spectrum identifying at the same time the non peak regions Then when each peak is dealt with the continuum beneath it is interpolated with a smoothed steep function The peak shape function is a Gaussian with a tailing function on its low energy side The Ganaas offers control for the spectrum analysis through several fi

Intercomparison of software packages

Contents

Download Pdf Manuals

Related Search

Related Contents