Neutron coincidence counting with digital signal processing

ARTICLE IN PRESS

Nuclear Instruments and Methods in Physics Research A 608 (2009) 316–327

Contents lists available at ScienceDirect

Nuclear Instruments and Methods inPhysics Research A

0168-90

doi:10.1

� Corr

E-m

journal homepage: www.elsevier.com/locate/nima

Neutron coincidence counting with digital signal processing

Janos Bagi e, Luc Dechamp a, Pascal Dransart a, Zdzislaw Dzbikowicz a, Jean-Luc Dufour b,Ludwig Holzleitner a, Joseph Huszti e, Marc Looman d, Montserrat Marin Ferrer a,Thierry Lambert b, Paolo Peerani a,�, Jamie Rackham f, Martyn Swinhoe c, Steve Tobin c,Anne-Laure Weber b, Mark Wilson f

a European Commission, Joint Research Centre, IPSC—Ispra, VA, Italyb Institut de Radioprotection et Suret�e Nucl�eaire—Fontenay-aux-Roses, Francec N-1, Safeguards Science and Technology Group, LANL—Los Alamos, NM, USAd Consulenze Tecniche—Cocquio Trevisago, VA, Italye Institute of Isotopes (IKI)—Budapest, Hungaryf VT Nuclear Services—Sellafield, Seascale, UK

a r t i c l e i n f o

Article history:

Received 17 March 2009

Received in revised form

7 July 2009

Accepted 8 July 2009Available online 28 July 2009

Keywords:

Monte Carlo codes

Neutron detectors

Neutron coincidence counting

Nuclear safeguards

Calibration of neutron counters

02/$ - see front matter & 2009 Elsevier B.V. A

016/j.nima.2009.07.029

esponding author. Tel.: +39 0332 785625; fax

ail address: [email protected] (P. Peerani).

a b s t r a c t

Neutron coincidence counting is a widely adopted nondestructive assay (NDA) technique used in

nuclear safeguards to measure the mass of nuclear material in samples. Nowadays, most neutron-

counting systems are based on the original-shift-register technology, like the (ordinary or multiplicity)

Shift-Register Analyser. The analogue signal from the He-3 tubes is processed by an amplifier/single

channel analyser (SCA) producing a train of TTL pulses that are fed into an electronic unit that performs

the time- correlation analysis. Following the suggestion of the main inspection authorities (IAEA,

Euratom and the French Ministry of Industry), several research laboratories have started to study and

develop prototypes of neutron-counting systems with PC-based processing. Collaboration in this field

among JRC, IRSN and LANL has been established within the framework of the ESARDA-NDA working

group. Joint testing campaigns have been performed in the JRC PERLA laboratory, using different

equipment provided by the three partners. One area of development is the use of high-speed PCs and

pulse acquisition electronics that provide a time stamp (LIST-Mode Acquisition) for every digital pulse.

The time stamp data can be processed directly during acquisition or saved on a hard disk. The latter

method has the advantage that measurement data can be analysed with different values for parameters

like predelay and gate width, without repeating the acquisition. Other useful diagnostic information,

such as die-away time and dead time, can also be extracted from this stored data. A second area is the

development of ‘‘virtual instruments.’’ These devices, in which the pulse-processing system can be

embedded in the neutron counter itself and sends counting data to a PC, can give increased data-

acquisition speeds. Either or both of these developments could give rise to the next generation of

instrumentation for improved practical neutron-correlation measurements. The paper will describe the

rationale for changing to the new technology, give an overview of the hardware and software tools

available today and a feedback of the experience gained in the first tests. Associated with the

experimental tests, the ESARDA-NDA working group is also performing an intercomparison benchmark

exercise on the analysis software for pulse processing.

& 2009 Elsevier B.V. All rights reserved.

1. Introduction

For more than 30 years, the measurement of the mass ofspecial nuclear materials has been done in a nondestructive wayusing neutron coincidence counting (NCC) or, more recently,neutron multiplicity counting. These techniques rely on theconcept that in plutonium, the spontaneous fission rate is

ll rights reserved.

: +39 0332 785072.

proportional to the mass (passive NCC) and in uranium theinduced fission rate is proportional to the 235U mass (active NCC).Because neutrons are also generated by other competitivephenomena, such as cosmic ray interaction with surroundingmaterials or (alpha,n) reactions in the sample, the discriminationof fission (producing several correlated neutrons) from otherreactions (producing single uncorrelated neutrons) is based on thecapability to analyse the time correlation between neutron pulses.

Time correlation analysis is traditionally performed usingdedicated electronics. The analogue signal from the He-3 tubesis processed by amplifier/single channel analyser (SCA), producing

www.elsevier.com/locate/nima

dx.doi.org/10.1016/j.nima.2009.07.029

mailto:[email protected]

ARTICLE IN PRESS

J. Bagi et al. / Nuclear Instruments and Methods in Physics Research A 608 (2009) 316–327 317

a train of TTL pulses that are fed into a neutron analyser based onthe shift-register principle. This kind of equipment has workedremarkably well through at least three decades, but is starting toshow some limitations. One of the major defaults of current shiftregisters is the inability to deal with very high count rates (eventhough a new generation of faster 50-MHz registers is comingnow on the market), but other features, such as the possibility totake into account in the analysis the position of the detectedevent, would be required in order to improve the neutron-counting technique.

For these reasons the IAEA and Euratom have expressed theneed to investigate new avenues for a novel concept of neutronelectronics that would override the limitation of shift-registertechnology. The continuous improvement in computer hardware,in terms of both computing speed and mass storage capabilities,has induced a recent trend in many experimental applications(not limited to nuclear physics). Instead of processing signalsthrough electronic modules, the current tendency is to acquireand store large quantities of raw digital information and thenprocess it through software.

The ESARDA-NDA working group, a forum collecting world-wide specialists on NDA measurements applied in safeguards, hasaccepted the challenge and instigated projects to investigate, testand eventually implement some alternatives for a digital signalprocessing of neutron coincidence counters. These projectsaddress the developments of the two major components: thedigital electronics and the processing software.

In this paper, we will present some recent results producedwithin these projects. In the field of electronics, four laboratorieshave compared the performances of four different digital acquisi-tion systems in laboratory campaigns performed in the PERLAlaboratory at Ispra. The results of these tests are described inSection 3 (Sections 3.1–3.4). Section 3.5 also contains a descrip-tion of a few commercial systems implementing digital pulseprocessing, but Section 3.6 gives a quick preliminary suggestion ofa development based on data analysis for the improvement ofdead-time correction algorithms.

In the field of testing and validation of software for thetreatment of digital pulse trains, the nondestructive assay (NDA)working group has launched two benchmarks for comparing thedifferent software to analyse LIST-mode pulse trains. This isdescribed in chapter 4.

2. New concepts for digital signal processing

There are two different approaches currently under study, andboth of them have been analysed in the PERLA campaign underidentical conditions (using the same samples): the LIST mode andthe Multiplicity Shift Register in ‘‘virtual instrument’’ mode.

2.1. LIST-Mode Acquisition and processing

The LIST-mode concept is based on a time-stamping technique.It is not new. It has been used for a long time, in particular forhigh-energy physics experiments. What is new is that computerhardware has improved to such an extent that a typical PCnotebook used to control inspection equipment in the field nowhas the capability to record and analyse this kind of data. Theprinciple of operation is that after each detection event, the timeat which the detection happened is sent to the PC (often theelectronics provide the time elapsed between two successiveevents, but this can be easily converted to an absolute timesequence). Optionally, the information of the channel where thedetection happened can be added in order to use the locationinformation when needed. The result of the acquisition is a

(possibly large) ‘‘pulse train’’ file containing all the informationfor each individual detection event (typically time and channelnumber).

The pulse train can then be analysed off-line with processingsoftware that simulates the operation of a shift register (predelay,opening of the real and accidental (R+A) gate, scaler counting,delayed gate for accidentals, storage of multiplicity distributions)and to compute all the physical parameters needed: Singles,Doubles, Triples and associated uncertainty. The pulse train can beanalysed with many different input parameters from a singlemeasurement. It can also be analysed to give individual channelcount rates, the pulse interval distribution on single channels(to measure the dead time) and compute the Rossi-a timedistribution.

One of the interests of pulse time-stamping is to be able toprocess the signal by changing one or several measurementparameters (predelay, gate width, etc). With a classical mode ofacquisition based on a shift register, it will be much more difficultto obtain comparable information about the quality of the result.With an acquisition mode based on dated events, it becomespossible in a short time with a post-processing treatment todetermine and optimise characteristics of the measurement cell(die-away time, gate width, etc). By calculating the neutron die-away time in the system, one can optimise the acquisitionparameters and ensure the quality of a result even in cases ofmeasurements with a high (alpha, n) reaction rate.

One particular advantage of LIST mode appears when instru-ments are calibrated. Calibration exercises with carefully preparedreference standards are expensive and time-consuming forinspectors and operators of nuclear facilities. If the calibrationdata are recorded in LIST mode, it is possible to create, retro-spectively, the calibration curve for any combination of detectorchannels, provided that individual outputs are available for eachchannel. If, during operation, one detector channel fails, it ispossible to recompute, from the original calibration data, thecalibration curve for the remaining channels. This is not possiblefor the data from most conventional neutron detectors, becausethe data from the individual channels of these detectors areusually OR-ed together before they are processed in the shiftregister.

2.2. The virtual instrument approach

The virtual instrument is a module that can consist eitherof a neutron analyser integrated in the neutron counter or of aPC-controlled independent black box or an acquisition card in thePC. The Virtual Instrument through a communication port of thePC (RS232, USB, Ethernet) can emulate the front panel (the displayand knobs) of the real instrument in order to change theconfiguration and collect the data. It is basically a digital cardthat reproduces directly all the functionalities of the shift register.It directly integrates the counts within the different gates andtransmits directly the Singles, Doubles and Triples to theacquisition PC.

The virtual instrument approach presents advantages anddisadvantages with respect to the LIST mode. The majordisadvantage of the virtual is that because it simulates a shiftregister, the operating characteristics such as gate widths andpredelay are preset and fixed, but in the LIST mode, theoperational parameters are simulated by the software with thebenefit that the same pulse train can be reanalysed using differentinstrumental setups.

On the other hand, the main advantage is that the datatransmitted to the PC are already the integrated counts in eachtime step, meaning that the computing effort is minimal requiring

ARTICLE IN PRESS

J. Bagi et al. / Nuclear Instruments and Methods in Physics Research A 608 (2009) 316–327318

only an averaging of the values; therefore, the results are providedimmediately at the end of the acquisition. The LIST mode, asimplemented in most applications, requires an off-line post-processing of the pulse trains, introducing a delay in theavailability of the result, even though real-time processing ispossible in some cases. In case of high number of events (longacquisitions or high count rates) the size of the pulse file can bevery large and the processing time increases exponentially withthe counting rate (some initial analyses took up to several hours).Also the file dimensions can become an issue needing frequentbackup on external storage devices.

3. Experiments with new electronics for neutron multiplicitycounting

During two experimental campaigns in the PERLA laboratory atJRC, three different configurations prepared by three differentinstitutions (IRSN, LANL and JRC) have been tested. A largenumber of measurements have been done, but the comparisonfocuses on six of them that were repeated with the three setupson the same samples:

�

Figthe

elec

a low-intensity californium source,
� a high-intensity californium source, � a small (20 g) Pu metal sample, � a small (50 g) Pu oxide sample, � a large (1000 g) Pu oxide sample and � a large MOX sample.
The three laboratories used different counters and differentdigital electronics, but in all cases the measurements wererepeated with the novel digital system and with a traditionalShift-Register Analyser in order to compare the results andvalidate the performances of the novel electronics.

3.1. The MEDAS card and the TRIDEN software developed by

CESIGMA and IRSN

The IRSN setup consisted of an Active Well CoincidenceCounter (AWCC) operated in passive mode and in fast configura-tion with Cd liners (Fig. 1). All the acquisitions were repeatedusing a conventional neutron analyser (a CANBERRA 2150) andthe so-called Multi-Event Data Acquisition System (MEDAS) data-acquisition card connected to the TTL output of the AWCC.

The system MEDAS is an electronic card, designed by thecompany CESIGMA, to acquire logical signals coming from any

. 1. Complete hardware set-up including the AWCC neutron counter (center),

MEDAS card with its junction box (left) and the standard acquisition

tronics based on several models of Shift-Register Analysers (right).

measurement system in several modes: counting, multiscalecounting or time-stamping mode [1]. The system is able to dateand count physical events of input frequency up to 10 MHzsimultaneously on 32 inputs with a maximum accuracy of 25 ns.The capacity of recording depends then on the memory available.Typically, count rates of about 150,000 counts/s create a 100-Mbbinary file for 100-s measurement time. The MEDAS output binaryfile is converted to ASCII for further processing.

A software, named TRIDEN (Treatment of Dated pulses comingfrom Neutron Emitters), has been developed by IRSN to processthe digital pulse train issued from MEDAS and compute countrates: Singles (S), Doubles (D) and Triples (T), taking into accountthe correct setup of the detection cell (gate width, predelay andlong delay). This software basically consists of the following:

�
Simulating the operation of a Multiplicity Shift Registerstarting from the first pulse detected up to the one detectedsome milliseconds before the end of the measurement; eachone opens, after a predelay, a first-time gate to collect thenumber of real+accidental (R+A) pulses detected and, followinga long delay, a second time gate to collect the number ofaccidentals pulses (A). � Calculating the neutron count rates of interest: S, D and T from
the multiplicity distributions in the ‘‘R+A’’ and ‘‘A’’ gates(TRIDEN offers also the possibility to split an initial binary filein several parts in which it calculates S, D and T and theassociated standard deviation).
� Constructing the Rossi-a distribution in order to check
measurement parameters as the predelay and the die-awaytime.
� Deriving from the point model equations [2], knowing the
isotopic composition, the mass of plutonium resulting from abasic passive neutron-counting measurement.

During the experimental campaigns at PERLA Laboratory, thesame procedure was applied for all the sources or samplesmeasured with the AWCC, namely 10 acquisitions of 100 s eachwere performed successively with both systems: the AWCCconnected to a 2150 electronic shift register driven by the INCCsoftware, then the AWCC connected to the MEDAS system, usingTRIDEN software with the same parameters as the ones integratedin INCC (predelay of 4.5ms, gate width of 64ms, long delay of1024ms, no dead-time corrections) to analyse the resulting pulsetrains.

Table 1a shows a comparison between the Singles (S), Doubles(D) and Triples (T) estimated using the following two systems:AWCC+2150+INCC and AWCC+MEDAS+TRIDEN, for each of the sixsamples introduced above. For each count rate, the table gives firstthe relative discrepancy between MEDAS and 2150 (taking 2150 asa reference), then the relative uncertainty given by TRIDEN(standard deviation observed on the distribution of the countrates on 10 acquisitions of 100 s), and finally the relativeuncertainty given by INCC (issued from theoretical equations [2]).

The S, D and T rates coming from MEDAS are very consistentwith the ones issued from the 2150 electronics presented herewithout any dead-time corrections. The uncertainties calculatedfrom the repetition of the measurement are quite small, especiallyfor the Single and the Doubles, indicating a good reproducibility ofthe count rates. The relative discrepancy observed on S and D

rates, while taking as reference the 2150 electronics, is lower than1%. For the Triples rates, the counting statistics are lower, and thisdirectly impacts the scattering around the mean value of the 10counting cycles. The discrepancy on the Triples is higher than theone observed on S and D rates but remains lower than thecounting uncertainty in the majority of the cases except for

ARTICLE IN PRESS

Table 1aComparison of Singles, Doubles and Triples issued from MEDAS+TRIDEN versus 2150+INCC with AWCC. Measurement standard deviations are also reported.

Reference Comparison MEDAS/2150 without multiplicity dead-time correction: d ¼ 0 ns

DS (%) rS Triden (%) rS Incc (%) DD (%) rD Triden (%) rD Incc (%) DT (%) rT Triden (%) rT Incc (%)

Small Cf �0.1 0.0 0.1 0.2 0.2 0.2 0.1 0.6 0.7

Large Cf 0.0 0.0 0.0 0.1 0.2 0.2 �11.4 4.4 4.0

Small Pu metal 0.1 0.0 0.2 1.7 0.4 0.4 0.1 2.1 1.2

Small Pu oxide 0.0 0.0 0.1 �0.5 0.4 0.4 �1.9 1.9 2.6

Large Pu oxide 0.0 0.0 0.0 0.2 0.3 0.2 �3.8 7.1 7.0

Large MOX 0.0 0.0 0.0 �0.4 0.3 0.5 �8.3 3.9 6.4

Fig. 2. Die-away time calculated from TRIDEN software using small Cf source.

Table 1bComparison: S, D, T issued from MEDAS vs 2150 with AWCC with multiplicity dead-time corrections.

Reference With multiplicity dead-time correction, d ¼ 170 ns for the 2150 and d ¼ 188 ns for MEDAS

DS (%) rS Triden (%) rS Incc (%) DD (%) rD Triden (%) rD Incc (%) DT (%) rT Triden (%) rT Incc (%)

Small Cf �0.3 0.0 0.1 0.2 0.2 0.2 �0.3 0.9 0.7

Large Cf 0.0 0.0 0.0 0.1 0.2 0.2 0.1 2.2 2.2

Small Pu metal 0.1 0.0 0.2 1.7 0.4 0.4 3.1 1.5 2.2

Small Pu oxide 0.0 0.0 0.1 �0.5 0.4 0.4 �1.5 1.9 2.6

Large Pu oxide 0.0 0.0 0.0 0.2 0.3 0.2 2.3 4.6 4.7

Large MOX 0.0 0.0 0.0 �0.4 0.3 0.5 �6.1 3.5 5.8


samples having a high neutron emission rate. One explanation tothis discrepancy is to consider that the multiplicity dead-timeparameter d is different for each electronic device shown inTable 1b. The list mode electronics allows the reconstruction ofthe Rossi-a distribution and the verification, from one acquisition,of measurement parameters as the neutron die-away time asshown in Figs. 2 and 3. The curve also shows that the predelayapplied to the MEDAS card is not optimised. Indeed, the predelayoptimum value for the AWCC is about 2ms while its adjustmentwas set to 4.5ms, which is the usual setup for AWCC usingconventional shift-register electronics.

3.2. The LMMM data-acquisition system developed at LANL

The LANL setup consisted of an Epithermal Neutron Multi-plicity Counter (ENMC) [3]. The detection efficiency is around 64%.The ENMC was modified to provide both a conventional output toan AMSR and a differential signal for each of the 27 Amptek A111preamplifier/discriminators (Fig. 4).

The AMSR was used in parallel with a list-mode data-acquisition system developed by LANL. It consists of a unit(LMMM ¼ LIST-Mode Multiplicity Module) that receives the 27differential pulses and records the arrival time of each. The LISTMode Counter core consists of pulse detection logic for up to 64external sensor inputs, a buffering FIFO, and a timestamp unit, andis written in VHDL. This core is loaded into a Field ProgrammableGate Array (FPGA) chip and interfaces to a Power PC (PPC405) coreembedded into the FPGA. The PPC runs an embedded microkernelat 400 MHz that collects the list-mode data into blocks that aresent to the Windows PC by Ethernet through UDP socket protocol.The bin size or time resolution can be set as low as 10 ns, but isnormally set at 100 ns.

An application running on the Windows PC receives andstores the packets in binary files. This application is used tocontrol the acquisition time. During these experiments, the data-acquisition was run continuously with a100-s acquisition time.The data-acquisition rate is currently limited to 100 kHz. Thislimitation is caused by the code that loads the data onto theEthernet.

The resulting binary files are read by a Visual Basic programthat calculates Singles and Doubles and can optionally create anASCII pulse file. These pulse files are identical in format to thepulse files used in the ESARDA multiplicity benchmark exercise[4] and can therefore be read with the software used in thatexercise. This is done when Triples rates are required.

Table 2 shows the results from the ENMC for the samplesconsidered above. The numbers in the table are not dead-time orbackground corrected. A comparison between the results showsthat the agreement in the Singles is 0.06% or better. The list modeDoubles are systematically 0.5% lower than the AMSR doubles.This is caused by the digital synchronization of the pulses.In the AMSR, the clock interval is 0.25ms, which effectivelyincreases the gate by up to this amount—roughly half of theinterval. In the LMMM the clock interval is 0.1ms; therefore, therelative gate lengths are 24.125/24.05 ¼ 1.003. The effect is similarin principle for the Triples where we see approximately a 1%difference between the AMSR and LMMM results.

Because these measurements were carried out, the instrumenthas been improved. Laboratory tests show that the instrument can

ARTICLE IN PRESS

Fig. 3. Rossi-a curve and die-away time calculated from TRIDEN software results using large Cf source.

Fig. 4. Epithermal Neutron Multiplicity Counter with LIST-Mode Multiplicity

Module.


acquire 600k counts/s using an external Ethernet connection and2 million counts/s to the 20-Gbyte internal memory.

3.3. The MI-PTA card and other developments at JRC

The JRC setup consisted of a Scrap Neutron MultiplicityCounter (SNMC) [5]. The digital acquisition in the SNMC is nota LIST mode, but it integrates virtual instrument electronics, theMI-PTA (Multi Input-Pulse Train Analysis) card [6] (Fig. 5).

The electronic system (MI-PTA) can be either used togetherwith a standard instrument such as the shift register or replacedwith a direct process of the digital pulse train by PC. The MI-PTAhas been designed to perform a variety of pulse-train analysesindependently or combined with a PC to which the system isconnected through a high-speed USB2.0 connection. This systemhas been developed with a maximum of 128 input channels,which can handle count rates in excess of 1 million counts/swithout count losses. Pulse data can be exchanged between theMI-PTA and PC at the rate of 480 Mb/s allowing real-time analysison the PC. The system consists of a number of base units, each

with 8 TTL-compatible inputs. Each input channel has a dedicated3-bit counter, which is much faster than the input pulsefrequency. Counter states are fed through double-stage flip–flopsto avoid metastability. The DSP compares actual counter stateswith previous states to calculate the number of pulses that arrivedduring each acquisition cycle (1ms). One unit, configured as amaster, is connected to the PC. As many as 15 slave units can beconnected to the master through a High Speed Serial Bus (in theSNMC, a master and five slaves are needed to manage the 44amplifiers). An additional feature of the system is that it is soft-configurable: a program running on the PC can automaticallymonitor a single input at the time. This is very useful forcontrolling the optimum working conditions of counter tubesand amplifiers. For instance, a failure of a single channel can bedetected easily even in a system with more than 100 countertubes, as it is with the case with the SNMC.

The SNMC has also a standard BNC output for a TTL signal;therefore, it can also be used with traditional neutron analysers,such as AMSR or JSR-14. Table 3 shows the comparisons of themeasurements of the six samples with the SNMC usingacquisitions by the digital MI-PTA output and PC softwareanalysis with respect to the Shift-Register Analyser (JSR-14).

Apart from the above-described ‘‘virtual instrument’’ card, JRCis in parallel also developing a LIST-Mode Acquisition program,based on the use of FPGA technology. As for the microprocessor,the digital circuit elements decreased in physical size, resulting inincreasingly complex systems. The FPGA constructors providesoftware tools to take care of the complexity and leave to thedeveloper the creative part of their designs. The programming ofthe FPGA is done using a logic circuit diagram or a speciallanguage called Hardware Description Language (HDL). JRCstarted to investigate the use of FPGA for new developments suchas the ‘‘LIST-Mode Acquisition,’’ ‘‘LIST Mode Analysis’’ and in thefuture ‘‘Monte Carlo simulation.’’

The ‘‘LIST-Mode Acquisition’’ is based on a 200 MHz Finite StateMachines (FSM) that take care of all the synchronisation signals ofthe Acquisition Unit (Clock, Read/Write and Latch). A 16-bitintelligent OR is the Input Unit, which, combined with a 50-MHzcounter and a 32-bit output Latch, composes the whole Acquisi-tion Unit. FSM have numerous advantages: they can be applied tomany areas leading to the development of faster or secure digitalhardware systems. FSM are used in many software developmentsand play a significant role in FPGA design and development.

ARTICLE IN PRESS

Fig. 5. Scrap Neutron Multiplicity Counter with embedded MI-PTA cards.

Table 2(a–c) AMSR and list mode results for the ENMC.

(a)S (AMSR) Error S (LMMM) Error Delta S (%)

Small Cf 2525 0.9 2525 1 0.000

Large Cf (i) 305,434 27

Small Pu metal 1536 1.5 1536 1.7 0.000

Small Pu oxide 14,861 4.5 14,852 4.6 0.061

Large Pu oxide (i) 299,526 16

Large MOX (ii) 57,136 6.4 57,155 17 �0.033

(b)D (AMSR) Error D (LMMM) Error Delta D (%)

Small Cf 1465 0.7 1472 0.9 �0.478

Large Cf (i) 180,276 118

Small Pu metal 504.5 1.1 507.3 1.2 �0.555

Small Pu oxide 3475 3.5 3483 3.7 �0.230


Large MOX (ii) 12,438 11 12,482 41 �0.354

(c)T (AMSR) Error T (LMMM) Error Delta T (%)

Small Cf 504 0.6 508.8 0.8 �0.952

Large Cf (i) 53,168 211

Small Pu metal 110.5 0.9 111.6 0.9 �0.995

Small Pu oxide 831.4 2.6 833.7 3.4 �0.277


Large MOX (ii) 2556 11 2625 52 �2.700

(i) Note that the count rate was too large to transfer the data to the Windows PC.

(ii) The AMSR measurement time was 2000 s and the LMMM measurement time was 285 s, which is why the uncertainties are larger.


The ‘‘LIST Mode Analysis’’ is the intelligent part, the dataacquired by ‘‘LIST-Mode Acquisition’’ unit will be analysed andproduce the output results such as Duration, counts in R+A scaler,counts in A scaler, Singles rate, Doubles rate, Triples rate, R+A

multiplicities, A multiplicities, Time Interval Distribution (TID).This will be done using a C-to-VHDL converter to compile theroutines written in C for the FPGA; this software-to-hardware

technology enables FPGA-based accelerated computing and there-fore the integration of the C code into hardware. To date, thecompiled C code has been tested on an Intel-based computer inorder to test the validity of the code. The results obtained on a PCIntel Xeon 1.6 GHz are shown in Section 4.

The following step will consist of validating the C Code on anFPGA; the number of ‘‘logic blocks’’ in the FPGA may vary a lotbetween the different models and producers; therefore, we needto test our code in several steps, adding more capabilities at eachstep. The last step will be to merge both systems, the ‘‘LIST-ModeAcquisition’’ and the ‘‘LIST Mode Analysis,’’ in order to have anautonomous hardware able to analyse LIST Mode data ‘‘on the fly’’and to output the results. Those results could be compliant to thefuture Analytical Device Integration (ADI) standards that use the‘‘Unified Architecture’’ from the OPen Connectivity (OPC) Founda-tion; OPC is open connectivity through open standards. OPC-Unified Architecture (UA) is a platform-neutral standard, designedto support complex data types and object models, has broadindustry support and achieves high-speed data transfers.

3.4. The PTR data-acquisition system developed at IKI

IKI developed different kinds of digital processing units forneutron coincidence counting. They consist of an FPGA-basedexternal unit and PC software running under Windows.

One unit is essentially a virtual shift register. The mostprocessor-intensive part of the function is realised in thehardware and only an RS232 line is needed to the PC.

ARTICLE IN PRESS

Fig. 6. PTR-02 developed at IKI.

Table 3(a–c) Comparison of MI-PTA vs JSR-14 multiplicity shift register with SNMC.

(a)S (JSR-14) Error S (MI-PTA) Error Delta S (%)

Small Cf 2377 2 2377 3 0.026

Large Cf 290,952 24 290,944 24 0.003

Small Pu metal 1502 2 1503 2 �0.019

Small Pu oxide 14,413 6 14,410 5 0.018

Large Pu oxide 290,081 18 290,072 19 0.003

Large MOX 55,999 7 56,004 8 �0.009

(b)D (JSR-14) Error D (MI-PTA) Error Delta D (%)

Small Cf 1295 3 1292 3 0.202

Large Cf 156,024 135 155,225 179 0.514

Small Pu metal 451 2 450 2 0.188

Small Pu oxide 3062 5 3066 7 �0.150

Large Pu oxide 79,180 167 78,721 160 0.583

Large MOX 11,089 20 11,032 19 0.520

(c)T (JSR-14) Error T (MI-PTA) Error Delta T (%)

Small Cf 415 3 415 3 0.159

Large Cf 39,826 503 39,813 472 0.033

Small Pu metal 96 5 95 5 0.424

Small Pu oxide 686 7 687 6 �0.155

Large Pu oxide 25,353 604 25,445 360 �0.361

Large MOX 2117 32 2118 22 �0.034


The Pulse Train Reader (PTR) list-mode system consists of a16-input FPGA card running at 100 MHz clock frequency andreader and evaluation software (see Fig. 6). The hardware countsthe clock pulses between consecutive detector pulses. Pulsesarriving at the same time at different inputs get a count of zero.The detection logic handles 16 input lines and can capture up to50 million pulses/s without losses. The follow-up values are readthrough a USB line by the reader program and written in standardinteger format to a hard disk. The maximal impulse rate of thesystem is limited at present by the USB connection to 2 millionpulses/s.

The saved LIST Mode files are evaluated off-line after datacollection. The Singles, Doubles and Triples values are calculatedby a very fast program called Neutron5. Predelay, gate length andthe delay between the two windows can be set. The evaluation ofa usual list-mode file with a count rate of 105 counts/s takes only afew seconds on an average PC or laptop.

There is a separate program for calculating the Rossi-adistribution. Both programs can be started also from within thedata reader software.

IKI also developed a system for replaying pulse trainsregistered with the list-mode hardware. This is useful for trainingpurposes because a wide variety of sources can be replayed.

The performance of the PTR hardware and the software wasinvestigated by comparing the results of a conventional Multi-plicity Shift Register (JSR-14) and the evaluation of the registeredlist-mode files. The neutron signals were coming from a JCC-31detector. For compatibility only one input line of the PTRhardware was used.

The test measurements were carried out with a CanberraJCC-31 neutron coincidence counter borrowed from the IAEA. TheJSR-14 and the pulse-train recorder were parallel connected to thecoincidence counter measuring the same sample at the sametime. The set predelay was 4.5ms; the gate length was 64ms.

Several measurements were made to compare the Singles,Doubles and Triples calculated by the IKI evaluation softwarefrom the pulse trains recorded by the IKI list-mode device andthe results from JSR-14. For this test, AmLi, and Cf and PuBeneutron sources were used. The measurements’ results are listedin Table 4.

3.5. Developments at VT nuclear services

VT Nuclear Services have designed, developed and patented [9]a neutron-counting technology based on the time-stampingof pulses from the neutron detectors. This counting technologywas developed to permit high-speed processing of the neutron-detector pulses, enabling the high count rate throughputs (up to 1million counts processed/s) expected with such a high-efficiencyneutron-counting chamber.

A further requirement was to provide improved noiseimmunity for the counting systems, permitting it to be used inindustrial facilities.

ARTICLE IN PRESS

Table 4Comparison of PTR vs JSR-14 with IKI setup on JCC-31 counter.

Source Mes. duration (s) JSR-14 IKI-PTR Difference (%)

Count rate (cps) SD Count rate (cps) SD

AmLi 1�60,000 S (cps) 13,293.1 1.5 13,293.0 0.5 0.0008

D (cps) 2.6 1.5 2.4 0.6 8.3

T (cps) �0.4 1.1 �0.5 2 20

Cf-3 1�1200 S (cps) 5110.7 2.5 5110.6 2.1 0.002

D (cps) 906.7 2.7 908.9 2.4 0.24

T (cps) 92.8 1.4 93.3 4.3 0.5

Cf-1 1�3000 S (cps) 26,575 4 26,574 4 0.004

D (cps) 4663.2 7.9 4676.7 6.7 0.3

T (cps) 414 71 417 36 0.7

PuBe-425 3�2000 S (cps) 327,329 14 327,324 10 0.002

D (cps) 2106 116 2110 104 0.2

T (cps) �510 141 �524 126 3

PuBe-701 9�1000 S (cps) 727,787 53 727,773 54 0.002

D (cps) 10,083 157 10,097 158 0.14

T (cps) �2473 1181 -2476 1081 0.14

Fig. 7. Time-stamping Electronics System.

Fig. 8. Neutron electronics overview.


To this end, commercially available 3He neutron detectors wereequipped with proprietary designed slim-line head amplifiers thatwere coupled directly onto the ends of the detectors. This closecoupling ensures minimal noise pickup between the detection andamplification stages. These are simple low-cost units with noserviceable components and as such are designed to be throw-away units in case of failure. The amplified signals from up to eightsuch amplifiers are then routed to a proprietary design hub unit.

Each hub unit provides both high- and low-voltage powersupplies to the amplifiers and detectors, thereby reducing thecomplexity of the amplifier designs. The hub also provides allsignal processing (pulse shaping and noise discrimination) for thesignals from eight neutron detectors and assigns an address to thepulse that identifies the hub/neutron detector combination wherethe pulse originated.

This digital address data are then sent through a high-speedfibre optic link to a data-acquisition computer where the addressis retrieved by a timestamper unit that appends the data with‘‘time of arrival’’ data. In this manner, the detector that generatedthe pulse (i.e., detected the neutron) and the time of arrival of thisdata at the timestamper are recorded and used to determine theneutron-counting rates. Fig. 7 shows the neutron electronicscomponents, head amplifier, hub and time-stamping card.

The time stamp data are analysed within software to generatemultiplicity frequency histograms that are then processed withinthe system software to perform either total neutron-counting,coincidence counting or multiplicity analysis. The softwarepermits detectors to be analysed individually or within user-configurable groups so that the total neutron count rate from thedetectors in the inner and outer rings can be calculated alongsidethe multiplicity count rates from the full system. Furthermore, thecount rates from individual detectors can also be processed tocheck for failure of detectors, etc. during the actual drummeasurement (Fig. 8).

Other options for monitoring of the system configuration andparameters are possible through hubs. These facilities include thelogging of applied high voltage and discriminator thresholds foreach of the neutron detectors through outputs at the hubdiagnostic port.

This patented timestamper technology [9] has alreadybeen incorporated in neutron products installed in nuclearfacilities across the world since 2004 (Fig. 9). The TRU-Ds Drum

ARTICLE IN PRESS

Fig. 9. TRU-D Drum, PIMS and DISPIM products.


is a high-efficiency multiplicity counting system utilising thebenefits of software-based multiplicity counting. The Fisstrak:PIMS system combines 23 hubs distributed throughout areprocessing plant to provide real-time monitoring of theprocess for material control and safeguards purposes. The TRU-Ds DISPIM system benefits from the reduced cable architecture toallow in-situ coincidence measurements of installed plantequipment.

The advantages of using digital neutron pulse trains withPC-based acquisition software over conventional TTL pulses andneutron-counting hardware are well understood. The flexibilityafforded by software-based acquisition systems and the scal-ability of the neutron-counting hardware make such systemssuitable for many radiometric applications [10].

3.6. Dead-time correction using LIST-mode technique

Modern multichannel list-mode counters provide increasedpossibilities for neutron data analysis. In addition to the classicalmultiplicity distribution, further statistics on the arrival time ofneutrons, channel number, cross- and autocorrelation andtriggered multiplicity can be collected. One new analysis is theuse of the observed differences in coincidence rates betweendifferent channels to perform a dead-time correction to thecomplete multiplicity distribution.

This method has already been tested, using both simulationsfrom JRC and LANL and real data measured at LANL. For themoment a rather basic evaluation method of the data is used. Theaccuracy of the correction depends on the count rate per channeland can for sure be improved to higher count rates using a moresophisticated evaluation. It has been found that good results canbe provided for a count rate of about 150,000 counts/s using32 channels, provided that the count rates per channel areabout equal.

This method provides several advantages. The two mainadvantages are clearly that, because the Singles, Doubles, Triples,Quads, etc. are calculated later from the multiplicity distribution,this method will be suitable to correct for any such multiplicity.Secondly the correction is based on the collected data and so noprevious dead-time measurement is necessary. Furthermore thismethod does not use a classical predelay; therefore, the increasedgate-fraction leads, in principle, to the increased quality of theresults. This correction technique could be implemented in thehardware of the instrument itself.

4. Benchmarks on the processing software

In 2003, the ESARDA-NDA working group has launched abenchmark in order to compare the different algorithms andcodes used in the simulation of neutron multiplicity counters. Inorder to derive the maximum of information and at the same timeto allow a large participation, the working group decided to splitthe exercise in two parts with two participation levels: a fullsimulation exercise where participants were asked to compute thecount rates, starting from the basic technical specifications, and/ora partial exercise involving the processing of the pulse trainsproduced by a single laboratory. The results of participantsperforming the entire exercise allowed making a comparisonamong the different Monte Carlo codes for the simulation ofneutron multiplicity counters. The results of the partial exercisehelp to test the available algorithms for pulse-train analysis and toderive some important information about the models applied fordead-time correction. The results of the first and second phase ofthe ESARDA Multiplicity Benchmark have been published in aspecial issue of the ESARDA Bulletin [4], and a summary waspresented and published in the proceedings of the 47th INMMAnnual Meeting [7].

All cases run in the first two phases of the benchmark weretotally theoretical. Therefore, the conclusions derived had to beconsidered as a relative behaviour of the different models,techniques and codes. Notwithstanding the satisfactory conclu-sion that all the algorithms developed by the different partici-pants in the first two phases and used to analyse the pulse trains

ARTICLE IN PRESS

Table 5Results from experimental pulse train processing and comparison with multiplicity shift register measurements.

Counting time Singles rate S abs. unc. Doubles rate D abs. unc. Triples rate T abs. unc. Time

Case 1: Cf low intensityMeasured 1247.87 1.58 380.78 0.84 66.65 0.48Participant 1 1053.32 1244.42 4.49 382.26 2.94 67.09 2.23 3.4

Participant 2 1050.00 1244.44 1.43 381.70 0.77 66.39 0.57 1.0

Participant 3 1053.32 1244.34 8.99 381.88 4.85 66.72 3.99 0.5

Participant 4 1243.94 1.57 381.90 1.14 66.74 0.75 0.7

Participant 5 1040.00 1244.30 4.37 382.07 3.87 72.52 2.74 8.2

Participant 6 1053.24 1244.40 4.48 381.26 2.94 66.23 2.12 2.0

Participant 7 1053.32 1244.37 4.50 381.80 2.82 66.67 1.98 2.8

Participant 8 954.90 1243.90 1.10 382.60 1.20 67.30 2.00 1.0

Participant 9 1246.26 0.01 381.43 0.67 66.72 0.43 6.1

Participant 10 1053.20 1244.70 4.61 381.32 3.50 66.37 2.51 0.9

Participant 11 1053.25 1244.39 1.42 382.23 0.92 67.07 0.71 0.8

Rel. stand. dev. 0.001 0.001 0.027

Case 2: Cf high intensityMeasured 149,378.13 7.30 43,373.58 33.22 3695.29 71.23Participant 1 1000.39 149,364.24 45.84 43,522.80 178.00 3142.29 352.84 4030.0

Participant 2 1000.00 149,360.00 15.22 43,454.00 66.91 3310.80 161.28 34.0

Participant 3 1000.39 149,362.55 91.73 43,522.68 463.24 3403.39 1102.66 689.3

Participant 4 149,364.27 16.74 43,470.35 77.52 3266.27 189.99 45.3

Participant 5 990.00 149,364.15 43.03 43,492.82 245.88 3785.99 455.87 972.1

Participant 6 1000.34 149,364.22 45.80 43,525.16 286.15 3353.66 469.35 4989.0

Participant 7 1000.38 149,364.26 45.87 43,513.48 251.58 3328.03 644.65 218.0

Participant 8 999.06 149,364.00 12.00 43,543.00 65.00 3334.00 646.00 17.0

Participant 9 149,420.47 0.04 43,486.79 67.35 3272.29 144.95 3079.4

Participant 10 1000.30 149,364.63 59.18 43,532.54 244.70 3554.16 627.31 1075.0

Participant 11 1000.34 149,364.21 14.49 43,522.92 55.69 3145.53 112.13 113.0

Rel. stand. dev. 0.000 0.001 0.050

Case 3: Pu metalMeasured 760.288 1.123 129.141 0.566 14.099 0.303Participant 1 1269.51 761.16 3.34 130.84 1.68 14.16 0.68 2.5

Participant 2 1270.00 761.15 1.06 130.62 0.46 13.95 0.20 0.0

Participant 3 1269.51 761.12 6.68 131.07 3.05 14.20 1.29 0.3

Participant 4 760.10 1.03 130.42 0.59 14.00 0.28 0.6

Participant 5 1000.00 760.10 3.45 130.45 1.57 15.50 0.88 3.4

Participant 6 1269.42 761.17 3.34 130.62 1.55 14.13 0.48 2.0

Participant 7 1269.51 761.14 3.34 130.86 1.66 14.23 0.72 2.4

Participant 8 1108.76 760.50 0.90 130.70 0.60 14.40 0.80 1.0

Participant 9 761.29 0.00 131.35 0.52 14.52 0.22 3.6

Participant 10 1269.40 761.33 3.09 130.63 1.52 14.16 0.62 1.1

Participant 11 1269.41 761.16 1.06 130.84 0.53 14.16 0.22 0.8

Rel. stand. dev. 0.001 0.002 0.030

Case 4: Pu oxide small massMeasured 7353.24 4.43 912.29 3.76 109.32 2.83Participant 1 1070.66 7345.31 11.26 913.10 8.62 110.78 6.96 28.3

Participant 2 1070.00 7345.30 3.56 906.72 3.94 106.56 2.55 3.0

Participant 3 1070.66 7345.18 22.51 910.12 19.35 110.11 16.44 4.1

Participant 4 7345.65 3.09 906.43 3.66 107.00 2.35 2.2

Participant 5 1000.00 7345.65 12.16 910.25 10.89 117.06 7.27 29.9

Participant 6 1070.62 7345.30 11.28 908.96 9.20 108.53 6.17 37.0

Participant 7 1070.66 7345.27 11.25 906.70 6.84 109.61 7.89 11.6

Participant 8 1053.90 7345.40 2.60 907.40 3.30 110.00 7.70 1.0

Participant 9 7354.01 0.02 907.40 3.94 107.23 2.04 39.5

Participant 10 1070.60 7345.47 9.24 908.95 10.54 108.76 6.44 8.0

Participant 11 1070.61 7345.29 3.57 913.10 2.73 110.78 2.21 3.7

Rel. stand. dev. 0.000 0.003 0.026

Case 5: Pu oxide large massMeasured 142,661.62 16.19 20,873.04 40.18 2518.91 175.20Participant 1 1001.80 142,611.58 33.83 20,940.06 255.57 2459.96 702.97 3700.0

Participant 2 1000.00 142,610.00 10.85 20,909.00 63.13 2432.50 142.18 33.0

Participant 3 1001.80 142,609.95 67.82 20,949.26 328.42 2473.05 1103.36 630.0

Participant 4 142,611.03 14.21 20,913.85 63.36 2420.11 141.24 44.1

Participant 5 1000.00 142,611.03 33.80 20,931.81 237.37 2911.00 743.75 907.5

Participant 6 1001.76 142,611.49 34.04 20,925.37 209.17 2473.20 431.78 4536.0

Participant 7 1001.79 142,611.59 33.92 20,973.49 108.53 2562.83 566.41 208.0

Participant 8 1000.84 142,611.00 12.00 20,987.00 56.00 2568.00 572.00 16.0

Participant 9 142,610.24 0.04 20,912.41 67.88 2422.44 171.50 2783.6

Participant 10 1001.80 142,612.12 38.28 20,934.08 183.00 2462.23 460.18 916.4

Participant 11 1001.75 142,611.47 10.76 20,939.33 80.74 2458.78 223.22 104.0

Rel. stand. dev. 0.000 0.001 0.056


ARTICLE IN PRESS

Table 5 (continued )

Counting time Singles rate S abs. unc. Doubles rate D abs. unc. Triples rate T abs. unc. Time

Case 6: MOX sampleMeasured 27,662.76 5.23 3063.97 15.55 301.84 19.31Participant 1 1018.88 27,658.27 17.35 3082.96 35.38 292.67 47.84 200.0

Participant 2 1020.00 27,658.00 5.50 3053.40 9.83 276.40 10.37 6.0

Participant 3 1018.88 27,657.91 34.66 3059.32 37.29 303.23 80.36 31.3

Participant 4 27,658.42 6.25 3053.55 11.82 274.67 11.99 7.8

Participant 5 1000.00 27,658.43 18.47 3073.31 35.94 310.16 44.15 118.4

Participant 6 1018.84 27,658.23 17.33 3075.91 26.28 283.85 36.96 224.0

Participant 7 1018.87 27,658.24 17.33 3071.09 22.21 294.35 38.00 40.0

Participant 8 912.89 27,661.30 5.50 3071.90 10.60 281.50 37.40 3.0

Participant 9 27,660.13 0.01 3052.67 10.19 276.78 10.74 212.5

Participant 10 1018.80 27,658.48 21.07 3075.86 35.71 283.15 35.56 52.1

Participant 11 1018.84 27,658.22 5.48 3082.92 11.18 292.69 15.15 14.2

Rel. stand. dev. 0.000 0.004 0.038

Fig. 10. (a–f) Plots of ratios of computed values (pulse-train analysis) vs measured values form Multiplicity Shift Register.


have proven to be satisfactory, the working group felt that anextension to real experimental cases would have added asupplementary value to the exercise.

Therefore, it was decided to use the experimental campaigndescribed in Section 3 for a continuation of the 2003 benchmark.Again it contained a first step (third phase) devoted to full MonteCarlo simulation and a second one (fourth phase) for theintercomparison of software for an analysis of LIST-mode files.For this benchmark, we selected the six measurements performedwith the IRSN setup (AWCC with MEDAS card). The LIST-ModeAcquisitions (for each case, 10 repeated measurements of 100 seach) have been stored in binary pulse-train files and distributed

to the participants who have tested their software computingSingles, Doubles and Triples rates and associated uncertainties.These were compared to the reference: the S, D and T ratesobtained with an analogue-electronic acquisition (AMSR). Theresults of this second benchmark are currently under publicationon the ESARDA Bulletin. We will summarise here the mostimportant outcomes from the fourth phase, the most relevant oneto the topic of this paper; for more details, we refer to thebenchmark final report [8].

All the results are reported in tabular form in Table 5 andgraphically in Fig. 10a–f . Generally, the agreement among thedifferent processing codes is extremely good: negligible

ARTICLE IN PRESS


deviations in Singles (less than 0.1%), agreement within 0.1–0.4%in Doubles. Nevertheless, a dispersion of up to 4% in Triples isvisible, indicating that the way to compute them is not totallyhomogeneous.

The values can also be compared to the measured S, D, T, with aMultiplicity Shift Register. Indeed, it is one of the scopes of theexercise to assess the capability of LIST-Mode Acquisition tocorrectly collect the measured data in view of a possible alternatetechnology in data-acquisition for neutron-counting applications.Indeed, the results show that data acquired with the time-stampingdata-acquisition card and processed with all the tested codes agreewith the multiplicity shift-register data within the statisticaluncertainties. We should bear in mind that the shift-registermeasurements and the LIST-mode measurements were done withthe same experimental setup, but sequentially in time. This meansthat they do not refer exactly to the very same pulse train, but totwo sequential pulse trains acquired in identical conditions. So wecan only conclude that they coincide within the statisticaluncertainty and that no systematic deviations have been revealed.

One outstanding feature of these results is the quoted absoluteerror. The values appear to vary by an order of magnitude fromgroup to group. This is an important issue that again should bestudied by a more detailed comparison of the calculation methodsand even definitions of uncertainty used by the different groups.The results could be compared to the values from the shift-register electronics and theoretical values.

A further point of interest is the processing time required thatvaries by more than an order of magnitude. This will be partly dueto the computer processing power available, but there may also betips and tricks that could benefit the safeguards community, if, aswe expect, there is widespread future use of list-mode data.

Overall, the results of this exercise show that all participantsare capable of good performance for practical purposes. However,comparison of the methods used by the different groups shouldallow the establishment of more robust analysis techniques withmore reliable error estimates.

5. Conclusions

The data collected for neutron coincidence counting usingLIST-mode contains much more useful information than classical

shift-register acquisition. Every measurement has informationsuch as die-away time, Rossi-a distribution and dead timeembedded within it. This can be used to verify the state of healthof the detector at the time of the measurement and can allowfurther analysis for diagnostic, quality control and optimisationpurposes. There is also the benefit that LIST-mode format ofcalibration data allows an instrument to be re-calibrated after thefailure of a particular channel.

The results presented in this paper show that digital acquisi-tion is rapidly approaching maturity. All the systems testedperform in a way totally comparable to shift register acquisition atlow and medium count rates. Some systems present limitations atvery high count rates, but this can easily be improved in the nearfuture. This work has shown that LIST-mode data collection andanalysis are practical for in-field application using typicalcomputers without the need for high-performance computing orextraordinarily large storage space.

References

[1] A.L. Weber, T. Lambert, P. Funk, Passive neutron multiplicity counting:simulations and measurements, in: Proceedings of the 45th INMM AnnualMeeting, Nashville, TN, 17–20 July 2006.

[2] B. Harker, M. Krick, INCC software: users manual, Los Alamos LaboratoryReport, LA-UR-99-1291, November 2001.

[3] H.O. Menlove, C.D. Rael, K.E. Kroncke, K.J. DeAguero, Manual for theepithermal neutron multiplicity detector (ENMC) for measurement of impureMOX and plutonium samples, Los Alamos Laboratory Report, LA-14088, May2004.

[4] P. Peerani, M. Swinhoe, ESARDA NDA Working Group Multiplicity BenchmarkExercise—Final Report, ESARDA Bulletin, no. 34, June 2006.

[5] M. Marin Ferrer, P. Peerani, M. Looman, L. Dechamp, Nucl. Instr. and Meth. A574 (2007) 297.

[6] M. Looman, M. Marin Ferrer, B. Pedersen, P. Peerani, Novel components forneutron counting, in: Proceedings of the 25th ESARDA Symposium, Stock-holm (S), 13–15 May, 2003.

[7] P. Peerani, M. Swinhoe, Results of the ESARDA multiplicity benchmarkexercise, in: Procedures of the 45th INMM Annual Meeting, Nashville, TN,17–20 July 2006.

[8] P. Peerani, M. Swinhoe, A.L. Weber, L.G. Evans, ESARDA multiplicity bench-mark exercise—Phase III and IV—Final Report, ESARDA Bulletin, to bepublished.

[9] Improvements in and relating to signal handling and processing, InternationalPublication Number WO 00/67044 (granted as US6912485, EP1175627,KR679707, CN1175281), Patent priority date May 4, 1999.

[10] Improvements in and relating to monitoring of radioactive emissions(published as EP1707992, US2006219518, JP2006284589).

Neutron coincidence counting with digital signal processing

Documents