- Email: [email protected]
Data acquisition with a fast digitizer for large volume HPGe detectors
ARTICLE IN PRESS
Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305 www.elsevier.com/locate/nima
Data acquisition with a fast digitizer for large volume HPGe detectors L.C. Mihailescua,b, C. Borceaa,b, A.J.M. Plompena, a
European Commission, Joint Research Centre, Institute for Reference Materials and Measurements, B-2440 Geel, Belgium b ‘‘Horia Hulubei’’ National Institute for Physics and Nuclear Engineering, P.O. Box MG-6, 76900 Bucharest, Romania Received 21 March 2007; received in revised form 8 May 2007; accepted 8 May 2007 Available online 25 May 2007
Abstract A 12-bits and 420 MSample/s waveform digitizer was used with large volume High Purity Germanium (HPGe) detectors for measurements of gamma-ray production cross-section from inelastic neutron scattering. For these time-of-flight measurements at the GELINA white neutron source, the use of a fast digitizer significantly increases the efficiency compared with a conventional data acquisition system. First, the pulse processing time required to obtain the amplitude with high resolution is significantly reduced so that pulse pile-up with the prompt gamma-burst is eliminated for neutron-induced events. Second, an improved time response is obtained for which the amplitude and rise time dependence is strongly reduced compared to that of a conventional constant fraction discriminator. Excellent energy and time resolution is obtained with algorithms suitable for on-line signal processing, so that data storage is under control. Bench tests are presented that compare methods of signal processing. For the best method, the data acquisition system based on the fast digitizer was tested during measurements of gamma production cross-sections for 206 Pb and 208 Pb. A direct comparison was made with results obtained with conventional electronics operated in parallel. r 2007 Elsevier B.V. All rights reserved. PACS: 29.30.Kv; 28.20.v; 25.40.h; 24.10.1 Keywords: Fast digitizer; Large volume HPGe detectors; Time-of-flight; Neutron inelastic scattering
1. Introduction Precise neutron-induced cross-sections are needed for nuclear applications under study like the new generation of reactors (the Generation IV initiative) or accelerator driven systems. At GELINA there is a continuous interest for new measurements of neutron-induced cross-sections with improved energy resolution and total uncertainty. New measurements of neutron inelastic, ðn; 2ngÞ and ðn; 3ngÞ cross-sections provide one example. For this purpose, a new experimental setup was created at the 200 m flight path station [1,2] of the GELINA white neutron source for highresolution time-of-flight (t.o.f.) measurements. The measurements are based on the ðn; xngÞ-technique which involves detecting the g-rays emitted following inelastic scattering, (n,2n) and (n,3n) reactions. The gamma production cross-section is the most important quantity Corresponding author. Tel.: +32 14 571 489; fax: +32 14 571 381.
E-mail address: [email protected] (A.J.M. Plompen). 0168-9002/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2007.05.231
determined in such an experiment. It is obtained for every observed transition in the full neutron energy range, from threshold to about 20 MeV. Large volume HPGe detectors of 75–104% relative efficiency are used for the detection of the g-rays. The incident neutron energy resolution for the main transitions is determined by a time resolution of 4–8 ns over a time-span of 20 ms. The desired total uncertainty of the gamma production cross-sections of about 5% for the most intense transitions requires long acquisition times (e.g. 500 h). Therefore, improvements of the detection efficiency are important to increase the flexibility and use of this experimental setup. Despite the very good results obtained with the setup shortly described above and in more detail in Refs. [1,2], two evident limitations emerged: (i) the acquisition system based on conventional electronics has a dead time of about 10 ms which is significant as a result of the strong prompt gamma-flash of the accelerator and (ii) the slow-rise-timerejection (SRTR) function of the constant fraction discriminator (CFD) was used to achieve good time
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
resolution for the large volume HPGe detectors resulting in a strong loss of efficiency for gamma-ray energies below 500 keV. As it will be shown later, the time spectrum obtained with the Constant Fraction (CF) function has one peak with a long tail that lasts for about 700 ns for an HPGe detector with a 104% relative efficiency. This tail is due to the pulses with a slow rise time and they occur more frequently at low energies of the g-ray. The SRTR function rejects all the events in the tail. The rejected events are about 55% for the detector mentioned above in the case of a 60 Co source. At GELINA, the neutron production is based on a linear accelerator and on a natural uranium target. Electrons are accelerated up to 150 MeV by the linear accelerator and then stopped in the U target producing bremsstrahlung. The neutrons are produced through ðg; xnÞ and (g, fission) reactions on the U target. Due to this production mechanism, every neutron burst is preceded by a flash of g-rays (bremsstrahlung radiation). This gamma-flash is scattered on the sample placed at the 200 m flight path station and then detected by the HPGe detectors. The gamma-flash induced events in every HPGe detectors amount to about 10–20% of the total number of bursts. This percentage depends on the properties of the sample, the detection efficiency of the HPGe detector and the angle of observation. To avoid complicated dead time corrections due to the pile-up of the neutron-induced events and the gamma-flash induced events, all the neutron bursts in which a gamma-flash induced event was detected had to be rejected. This has been done in hardware, constructing an inhibit signal [1,3,4]. Preserving the same number of detectors, the detection efficiency can be increased if the dead time is reduced and if the use of the SRTR rejection is avoided. The solution for these two limitations was a new data acquisition system based on a fast digitizer. The difficulties in using large volume HPGe detectors arise from the large variations in the rise time and in the shape of the pulse on the rising edge. A digital signal processing algorithm can deal better with such pulses. Here, the use of a particular fast digitizer is investigated to improve the dead time and time-response of the employed HPGe detectors so that high incident neutron energy resolution together with good gamma-ray energy resolution is preserved while significantly improving the detection efficiency. This paper will present the tests for t.o.f. measurements with a commercial fast digitizer, the digital algorithms used for the signal processing and the results for neutron inelastic scattering measurements. Throughout this paper the fast digitizer will be compared with the data acquisition system based on conventional electronics. The final comparison was done during measurements of neutron inelastic scattering cross-sections for two different enriched samples of 206 Pb and 208 Pb, where the output of preamplifier from the HPGe detectors was read simultaneously by the fast digitizer and by the data acquisition based on the conventional electronics.
299
2. The fast digitizer A fast digitizer is a flash analog-to-digital converter that records samples of the input signal with a high repetition rate and transfers them for signal processing. Two properties define the performance of a fast digitizer: the number of bits for the amplitude and the sampling rate. The number of bits defines the amplitude resolution and the sampling rate defines the time resolution. For the project described here, a commercially available DC440 digitizer from Acqirist was chosen. This has 420 MSamples/s maximum sampling rate which corresponds to a minimum sampling interval of about 2.38 ns and has 12 bits for the amplitude. The full scale range of the inputs can have the following values: 0.25, 0.5, 1, 2, 5 and 10 V. The input impedance is 50 O. Every DC440 card has two input channels and a common external trigger input. This card has the trigger time interpolation function that allows a very precise determination of the trigger moment. In this way, the arrival time of the external trigger is determined within one sampling interval. The clock accuracy of the card is less than 0.2 ppm. The preamplifier output of the detector is given directly to one of the two independent input channels. The DC440 module has no possibility of onboard signal processing and in consequence the full set of samples has to be transferred to the computer. There the signal is processed online or recorded on the disk. The digitizer system used here has two DC440 cards plugged in a CompactPCIt create. The create is connected to the computer through a PCI bus. An important limitation of the DC440 module is the transfer speed through the PCI bus (up to 100 MB/s). For the ðn; xngÞ cross-section measurements at the 200 m flight path station, about 10 000 samples ð24 msÞ are needed to cover the neutron energy range of interest (500 keV up to 20 MeV). In the worst case when both cards have a trigger simultaneously, 40 000 samples have to be transferred. The transfer of these data can be done in slightly less than 1.25 ms which is the time interval between two consecutive neutron bursts (800 Hz repetition rate of GELINA). Moreover, for the typical conditions of these experiments, the average counting rate is less than 10 neutron-induced events per second so that no dead time results from the data transfer.
3. The experimental setup Prior to implementing the online data acquisition with the DC440 digitizer we tested off-line the pulse processing algorithms on data recorded with a start–stop setup and a 60 Co g-ray source. Once the best pulse processing algorithms were chosen, the electronic scheme for the online data acquisition was created. The electronic schemes used for off-line tests and for t.o.f. measurements are presented in this section.
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
300
3.1. Start–stop setup with
60
Co source
simultaneously the time and the amplitude spectra and their coincidences.
To compare precisely the time and amplitude resolutions obtained with the DC440 digitizer and with the acquisition system based on the conventional electronics, start–stop experimental setups were created: one with the fast digitizer and another with the conventional electronics. The scheme of both setups is given in Fig. 1. The same detectors and the same 60 Co source were used in both cases. In the setup with the DC440 fast digitizer (Fig. 1, left) the HPGe preamplifier output was given to one input channel of the digitizer and the split output of the plastic scintillator (Pilot-U) to the second channel from the same card. The common external trigger of the digitizer card was the output of the scintillator CFD. The plastic scintillator output was given to the second input channel only to have a check for the precision of the external trigger. For every external trigger, the digitizer recorded 10 000 samples that were saved in binary files on the computer hard-disk. In the start–stop setup with conventional electronics (Fig. 1, right) spectra were recorded with both functions of the CFD module, SRTR and CF. The delay used for the output of the scintillator CFD had a precision of a few picoseconds. This delay and the reset signal were needed for a proper functioning of the multi-hit time-to-digital converter (FTD) [5,6]. This start–stop setup allowed saving
3.2. The setup for t.o.f. measurements For the measurement of the neutron inelastic scattering cross-section with the digitizer, the electronic setup of Fig. 2 was used. The purpose of the conventional electronic modules was only to reduce the counting rate to the rate of the neutron-induced events, i.e. avoid triggering by the g-flash. For this, an external trigger was constructed to select only the events that are generated within about 19 ms after the arrival of the gamma flash. This time range allows the detection of the neutrons from about 500 keV up to 20 MeV. The trigger is the t0 signal from the accelerator which signals the arrival of the electron burst at the uranium target. This trigger is validated by the presence of a neutron-induced signal in one of the two HPGe detectors connected to one card. For every external trigger, the digitizer card records 10 000 samples and transfers them to the PC where the signals are processed online. On the disk were recorded the amplitude and the time of the pulse. An additional flag that gives the number of pulses in every signal was recorded to check the consistency of the recorded data. During the measurements with the samples of 206 Pb and 208 Pb the output signals of all the four HPGe ADC
SA
PCI bus
PC
TFA
CFD
cable delay
CFD CFD
stop reset
Pilot-U
Pilot-U
external trigger
digitizer DC440
HPGe
HPGe
input 1 input 2
precise delay
FTD start
MMPM
PC
Fig. 1. Two independent start–stop experimental setups for testing different signal processing algorithms: one with the DC440 digitizer (left) and the other with conventional electronics (right).
Fig. 2. Left: the t.o.f. setup based on the fast digitizer Acqiris DC440. One card of the digitizer accommodated two HPGe detectors. Conventional electronic modules were still used only to reduce the counting rate to the neutron induced counting rate (about 6 counts/s in average). Right: the corresponding timing of the signals.
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
detectors of the setup were split to be connected to the digitizer and to the conventional acquisition system.
4. Digital algorithms for the fast digitizer The signals recorded from the HPGe detectors (see the upper panel from Fig. 3) were transferred to the PC and then analyzed to extract the time and the amplitude of every pulse. The difficulties in using the large volume HPGe detectors are due to the large variations in the rise time and in the shape of the pulse on the rising edge. Examples of the experimental pulse shapes are shown in -12 uk N2
vk
amplitude (arb. units)
-22 D
N1 -32 0
10000
5000
10
5
0 0
5000
10000
channels (1ch=2.38ns)
Fig. 3. Up: the preamplifier output as recorded with the digitizer (initial signal vk ) compared with the resulting signal (uk ) of the recursive algorithm from Eq. (6). The uk signal has a higher amplitude compared with the initial signal vk due to the correction for the ballistic deficit. Down: the result of the complete trapezoid algorithm, resulted from the differentiation of the signal uk .
Fig. 4 for a large volume HPGe detector with 104% relative efficiency. This detector has a crystal with the closed-end coaxial shape. The edges of the crystal were rounded. The crystal length was 6.3 cm and the diameter was 8.7 cm. The pulses from the output of the detector preamplifier were recorded with the DC440 fast digitizer. The rise time of the pulses varies from less than 100 ns to about 700 ns and these variations are larger with the increase of detector volume. The time and the amplitude algorithms that were tested in the present work are described in this section. 4.1. Timing On the initial signal, vk (k is the sample index), a pulse finding algorithm was applied which identified the number of pulses. This algorithm consists of a differentiation of the signal followed by the verification that peaks in the differentiated signal exceed a certain threshold. If no pulse was found, the signal was discarded. In case pulses were found, a digital Timing Filter Amplifier (TFA) algorithm was applied to reduce the noise and to bring to zero the offset of the signal. This algorithm consists of a two step recursive expression with discrete output signal ok : t2 D D=ð2t1 Þ ð1Þ D D=ð2t2 Þ ð2Þ ok ¼ A e ok e ok (1) t2 t2 t1 t1 where t1 is the integration time constant, t2 the differentiation time constant and A the amplification (here, A ¼ 1). D is the sampling interval (2.38 ns): oð1;2Þ ¼ k
1 X ðvkl elD=t1;2 Þ.
(2)
l¼0
Each factor o1;2 k can be written recursively: D=tð1;2Þ ð1;2Þ ok . oð1;2Þ kþ1 ¼ vk þ e
(3)
With the assumption that the input signal is constant before the first sample k ¼ 1 is recorded, from Eq. (3) it results
-20
Voltage (arb. units)
301
oð1;2Þ ¼ v1 1
-25
-30 0
200
400 600 Time (ns)
800
1000
Fig. 4. Different shapes of the signal from the preamplifier output of a 104% efficiency detector as recorded with the DC440 digitizer. Rise time values up to 700 ns are observed. The corresponding energy of these pulses is about 600 keV.
1 X eD=ð2t1;2 Þ ðeD=t1;2 Þl v1 D=t1;2 l¼0
(4)
and therefore o1 ¼ 0. On the output of the TFA algorithm, we tested three different algorithms for the determination of the time: Leading Edge Threshold (LET), Extrapolated Leading Edge Threshold (ELET) and Constant Fraction Discriminator (CFD). All of them have a well-known equivalent in the analog electronics (Ref. [7]). The LET algorithm is the simplest solution for the time determination. The time of the pulse is the moment when the signal crosses a given threshold, l. The LET algorithm is sensible both to time jitter and walk effect. The ELET algorithm [7] consist of two thresholds l 1 and l 2 , with l 1 ¼ k l 2 and k usually a small integer number. The ELET algorithm is insensitive to the walk effect as long the signal rises linearly.
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
The CFD algorithm used here is the digital version of the method described in Ref. [7]. The recursive equation used to construct the CFD signal ck is ck ¼ fokd ok
with kXd
(5)
where d is the delay and f the attenuation constant. To reduce the computing time, the CFD signal ck was constructed only on a short region around the pulse of interest. The digital pulse processing allowed also adjusting the baseline before the determination of the zero-crossing point. 4.2. Amplitude determination In conventional electronics the key element for pulse amplitude determination is the spectroscopic amplifier. This provides simultaneously the proper pulse amplification to match the requirements of the ADC that follows and the pulse shaping, reducing the long tail from the preamplifier. In choosing the most appropriate shaping method, a compromise between three general concepts has to be found: signal-to-noise ratio, pile-up of the pulses and ballistic deficit. The best signal-to-noise ratio is given by the cusp-like shaping followed by the triangular shaping. To overcome the ballistic deficit a shaping with the flat-top comparable with the maximum rise time is needed. The trapezoidal shaping meets simultaneously good signal-tonoise ratio and good compensation for the ballistic deficit [8]. The basic function of the trapezoid algorithm is to convert the exponentially decaying signal from the preamplifier into a trapezoidal signal. This implies the convolution of the preamplifier signal (input signal vk for the digitizer) with different time dependent functions. Several proposals for such methods can be found in the literature (Refs. [9,10]), but the differences between them is only the order in which the convolutions are done. The algorithm described here was optimized also for a high speed calculation. In the algorithm used here, two steps can be distinguished in the construction of the trapezoidal output signal. First the preamplifier signal is corrected for the finite decay time t and converted into a step function. The correction for the finite decay time t was done using the following relation: ukþ1 ¼ uk þ ðvkþ1 v0 ÞeDt=2t ðvk v0 ÞeDt=2t
(6)
where Dt is the sampling interval. The resulted step signal uk is shown in the upper panel of Fig. 3. The second step is to differentiate the signal uk obtaining the trapezoid shape. To maximize the computing speed and using the timing information obtained already, only two averages on the uk signal are done, one to the left of the starting time (on the interval N 1 ) and another to the right (on the interval N 2 ), after a maximum rise time D (Fig. 3, up). This average gives the maximum height of the trapezoid (Fig.3, down). Moreover N 1 and N 2 can be
dynamically determined as the maximum time interval between two consecutive pulses. Prior to the application of the trapezoid algorithm it is necessary to determine the decay time, t, of the preamplifier and the signal offset, v0 . The decay time t was determined from a least square fit of the decaying part of the pulse. The fit was done once for every detector, for a large number of pulses to assure a small statistical error. A Gaussian distribution was obtained for the decay time t. The centroid of the distribution was taken. For the 104% relative efficiency detector t ¼ 47:8 ms with the full width at half maximum (FWHM) of 0:7 ms. This distribution comes mainly from the noise of the signal and from the finite time interval (number of channels) used for the fit. 5. Results and discussion 5.1. Results with the
60
Co source
The timing algorithms mentioned in Section 4.1 were tested off-line on the same sets of signals. The best results for each algorithm are shown in Fig. 5. For the TFA algorithm, an integration time t1 ¼ 10 ns and a differentiation time t2 ¼ 200 ns were used. The parameters used for the CFD algorithm were d ¼ 31 ns and f ¼ 0:25. A short delay gives good results because it determines the zerocrossing very close to the starting time of the pulse where the rise often appears nearly linear (Fig. 4). The FWHM is 6.2 ns for the CFD algorithm and 18 ns for the worst case, the LET algorithm. A threshold l ¼ 15 keV was used for the LET algorithm while for the ELET algorithm l ¼ 15 keV and k ¼ 2. The parameters of the algorithms were obtained from trials with different values. Following this comparison, the CFD algorithm was chosen for the online data acquisition. Fig. 6 shows the comparison between the time resolution obtained with the digitizer and with the conventional electronics for two detectors with different relative 600 LET ELET CFD
400 Counts
302
200
0 0
100
Fig. 5. Time resolution with the algorithms.
200 Time (ns) 60
300
400
Co source and different digital
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
303
1000 75.9 %
Digitizer SRTR CF
Digitizer SRTR CF
104 %
1000
100 Counts
Counts
100
10
10 1
0.1
1 0
100
200
300
400
500
600
Time (ns)
0
200
400
600
800
1000
Time (ns)
Fig. 6. The comparison of the time resolution with the digitizer and with the conventional electronics for detectors with different relative efficiencies: left: 75.9% and right: 104%. The two functions of the analog CFD module were used: Constant Fraction (CF) and slow-rise-time-rejection (SRTR).
400 digitizer-trapezoid conventional system
Counts
300
200
100
0 1326
1328
1330 1332 Eγ (keV)
1334
1336
Fig. 7. The amplitude spectrum of the 60 Co source around the 1.33 MeV peak with the digitizer and with the conventional electronics for a 104% relative efficiency detector.
efficiencies: 75.9% and 104%. Energies E g X150 keV were selected in all these cases, both for the digitizer and for the conventional module. The fast digitizer and the conventional electronics give the same FWHM(5.8 ns for 75.9% detector and 6.2 ns for the 104% detector). Moreover the full width at 10th maximum (FWTM) is almost the same for the digitizer and for the conventional electronics. The difference is in the tail, defined here as all the counts outside one FWTM. The CF spectrum has a large tail that extends to 400 ns for the detector with the smaller volume and to 700 ns for the larger detector This tail represents 37% and 55% of the full spectrum, respectively. The fast digitizer preserves the detection efficiency of the CF function and the tail is significantly reduced to about 21% of the number of counts. The SRTR spectrum has no tail because all events with zero-crossing times later than leading edge times are rejected. This rejection is the reason for the lower detection efficiency and is more severe for lower gamma-ray energies.
The amplitude spectrum obtained both with the digitizer and with the conventional acquisition is shown in Fig. 7 for energies around the 1.33 MeV peak of 60 Co. FWHM ¼ 2:1 keV at 1.33 MeV for both the digital and the conventional system in the case of the 104% relative efficiency detector. This resolution was obtained with the following values for the trapezoid algorithm (Section 4.2): N 1 ¼ N 2 ¼ 3:3 ms and D ¼ 1:4 ms. The value of D was chosen to be larger than the maximum observed rise time and to avoid any shape fluctuations in the pulse at the end of the rise. Different values of N 1 and N 2 were tried between 0.5 and 5 ms. With the increase of the averaging intervals N 1 and N 2 the energy resolution improves and saturates around 3 ms. A minimum of about 1 ms was needed to still have a reasonable resolution (better than 2.5 keV). From these tests it was concluded that a minimum of 2:5 ms spacing between two consecutive pulses (pile-ups) is needed in the present trapezoid algorithm to have an amplitude resolution better than 2.5 keV at 1.33 MeV. Pulses separated by less than 1:4 ms (the value of parameter D) are not distinguished by the present algorithm and are treated as a single pulse. This is a large improvement in dead time in comparison with the conventional acquisition where 426 ms are used as shaping time. For the online data acquisition, the N 1 and N 2 were determined dynamically to adapt to the time interval between two consecutive pulses. If N 1 or N 2 are smaller than 1 ms then a flag is set for that event. For the counting rate of the present experimental setup the number of these pulses is negligible. For investigating the effect of the digitizer’s number of bits on the amplitude resolution of the detector, three different full scale ranges of the digitizer were used: 250, 500 mV and 1 V for the same gain of the detector’s preamplifier. A slight deterioration of the resolution was observed when the full scale range was increased, but still 2.3 keV amplitude resolution was obtained for the 1 V full scale range. It can be concluded that a digitizer with 12 bits amplitude range is sufficient for used with HPGe detectors at g-ray energies below 10 MeV.
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
304
5.2. Results for neutron inelastic cross-sections The fast digitizer was used for neutron inelastic scattering measurements in 206 Pb and 208 Pb with the setup described in Section 3.2 and with online signal processing. The two acquisition systems, based on the DC440 digitizer and on the conventional electronics, were used simultaneously only for comparison. The data were analyzed independently. Independent absolute efficiency calibrations were done. The ratio between the detection efficiency of the acquisition system with the fast digitizer and with the conventional electronics is given in Fig. 8. The efficiency losses due to the SRTR function in the conventional electronics are recovered in the fast digitizer. Three hundred percent is gained in the efficiency at energies around 400 keV and 15% at 1.4 MeV. Two examples of differential gamma production crosssection are shown in Fig. 9 for the 2614 and 583 keV transitions from 208 Pb nucleus. The measurement angles were 1101 and 1501. The results agree well within
the stated uncertainties. The magnitudes of the uncertainties are given as generic error bars at 6 MeV and at 15 MeV. The time resolution of the two acquisition systems can be compared when resonance structures are visible. This is the case for the main transition of 206 Pb (803 keV) at neutron energies immediately above the inelastic threshold. The resonance structures in the differential gamma production cross-section of the 803 keV transition are shown in Fig. 10 for the two acquisition systems. The data were recorded in parallel with the same detector (104% relative efficiency) at 1501. The two systems provide almost the same result. The small difference in the resonance amplitude may be the effect of the tail in the time resolution spectrum of the digitizer (see Fig. 6). This small difference does not outweigh the higher efficiency of the digitizer.
0.12 digitizer conventional
0.1
dσ/dΩ (barn/sr)
4.0
εd / εc
3.0
2.0
0.08 0.06 0.04 0.02 0
1.0 200
400
600
800
1000
1200
800
1400
900
Eγ (keV)
Fig. 8. The ratio between the detection efficiency of the digitizer and the detection efficiency of the conventional acquisition system with the SRTR function.
Fig. 10. High resolution differential gamma production cross-sections obtained for the 803 keV transition from 206 Pb with the fast digitizer and with the conventional acquisition system. 0.15
digitizer conventional
2614 keV - 208Pb 0.1
110°
0.05
digitizer conventional
dσ/dΩ (barn/sr)
0.15 dσ/dΩ (barn/sr)
1000
En (keV)
583 keV - 208Pb
0.1
150° 0.05
0
0 5000
10000 En (keV)
15000
5000
10000
15000
En (keV)
Fig. 9. Comparison of the differential gamma production cross-sections obtained with the fast digitizer and with the conventional electronics for 208 Pb. Both excitation functions were smoothed with a running average filter only for an easier comparison. Left: differential gamma production cross-section at 1101 for the main transition of 208 Pb nucleus (2614 keV). Right: differential gamma production cross-section at 1501 for the second transition of 208 Pb nucleus (583 keV).
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
6. Conclusions A new data acquisition system based on a 12-bits and 420 MSamples/s fast digitizer was created to increase the detection efficiency in the ðn; xngÞ ðx ¼ 1; 2; 3Þ cross-section measurements with large volume HPGe detectors at GELINA. The time and amplitude resolutions obtained for the same detectors with the fast digitizer and with the conventional electronics were compared. Different signal processing algorithms were tested off-line on the same sets of raw signals (as recorded from the output of the preamplifier) to establish the most suitable one for the online acquisition. The trapezoid algorithm for the amplitude and the CFD algorithm for the timing were preferred for the online signal processing. With the fast digitizer the amplitude and the time resolutions were identical with the results obtained with the conventional electronics. The advantages of the fast digitizer are firstly the fact that the time spectrum of the digitizer has a much smaller tail than the CF function of the CFD module and secondly that the minimum resolving time of trapezoid algorithm is only 2:5 ms, much smaller than the one of the spectroscopic amplifier. Therefore, the total dead time of the digitizer is 2:5 ms which allows the separation of neutron-induced events from the gamma-flash induced events for neutrons up to 20 MeV at the 200 m flight path length. The pile-ups between neutron-induced events were neglected, due to the very low counting rate. The differential gamma production cross-sections obtained with the data acquisition system based on the fast digitizer and with the data acquisition system based on the conventional electronics were compared for two sets of measurements, with 206 Pb and 208 Pb samples. The two acquisition systems give the same result for the differential gamma production cross-sections and the advantage of the fast digitizer is a higher detection efficiency, that increases from 15% at 1.4 MeV up to 300% at 400 keV. The increase in the efficiency of the fast digitizer has two sources:
The digitizer resolves the pile-up of the neutron-induced events with the gamma-flash induced events. Efficiency of 10–20% is gained in this way for every detector. The digitizer does not reject the pulses with a slow rise time and has the same efficiency as the CF function of the conventional CFD module. The tail of the time resolution spectrum is significantly reduced for the digitizer compared with the analog CF module.
It was observed that the present data acquisition with the fast digitizer is more sensitive to irregularities in the output signal of different detectors compared with the conven-
305
tional electronics. The performances of the fast digitizer are lower for detectors with lower signal-to-noise ratio, with two components of the decay signal or with changes in the signal offset (v0 ) during the data recording. Such effects were observed especially for old detectors. From the results of the tests it was concluded that the 12-bits and 420 MSPS fast digitizer can be used successfully with large volume HPGe detectors for low counting rates as it is the case for the ðn; xngÞ setup at the 200 m flight path station at GELINA. Mainly because of the transfer speed between the digitizer and computer the present acquisition system is limited to a counting rate of about 100 Hz. Substantially, higher counting rates can be accommodated by a fast digitizer if an on-board data processing would be possible, e.g., if the digitizer has on-board Field Programmable Gate Array (FPGA) that can be used for a complete signal processing. In this manner, the limitation imposed by the transfer time between the digitizer and the PC can be completely eliminated. The on-board signal processing would simplify even further the electronic scheme given in Section 3.2, avoiding the use of any conventional electronic module for the external validation. Acknowledgements L.C.M. and C.B. are grateful to the EC/JRC for financial support. References [1] L.C. Mihailescu, L. Ola´h, C. Borcea, A.J.M. Plompen, Nucl. Instr. and Meth. A 531 (2004) 375. [2] L.C. Mihailescu, C. Borcea, A.J.M. Plompen, Tests for the use of a fast digitizer in a time-of-flight measurements with large volume HPGe detectors, in: Proceedings of the Enlargement Workshop on Neutron Measurements and Evaluations for Applications, NEMEA2, Bucharest, 2005, p. 109. [3] L.C. Mihailescu, Neutron ðn; xngÞ cross-section measurements for 52 Cr, 209 Bi, 206;207;208 Pb from threshold up to 20 MeV, Report EUR 22343 EN, ISBN 92-79-02885-5, ISSN 1018-5593, European Communities, 2006. [4] L.C. Mihailescu, C. Borcea, A. Koning, A.J.M. Plompen, Nucl. Phys. A 786 (2007) 1. [5] S. de Jonge, Fast time digitizer, Internal Report GE/DE/R/24/87, JRC-CBNM, Geel, 1987. [6] J. Gonzalez, C. Bastian, S. de Jonge, K. Hofmans, Modular multiparameter multiplexer, MMPM, hardware description and user manual, Geel, 1997. [7] G.F. Knoll, Radiation Detection and Measurement, third ed. Wiley, New York, USA, 2000. [8] V. Radeka, IEEE Trans. Nucl. Sci NS-19 (1) (1972) 412. [9] V.T. Jordanov, G.F. Knoll, A.C. Huber, J.A. Pantazis, Nucl. Instr. and Meth A 353 (1994) 261. [10] A. Georgiev, W. Gast, IEEE Trans. Nucl. Sci. NS-40 (1993) 770.
Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305 www.elsevier.com/locate/nima
Data acquisition with a fast digitizer for large volume HPGe detectors L.C. Mihailescua,b, C. Borceaa,b, A.J.M. Plompena, a
European Commission, Joint Research Centre, Institute for Reference Materials and Measurements, B-2440 Geel, Belgium b ‘‘Horia Hulubei’’ National Institute for Physics and Nuclear Engineering, P.O. Box MG-6, 76900 Bucharest, Romania Received 21 March 2007; received in revised form 8 May 2007; accepted 8 May 2007 Available online 25 May 2007
Abstract A 12-bits and 420 MSample/s waveform digitizer was used with large volume High Purity Germanium (HPGe) detectors for measurements of gamma-ray production cross-section from inelastic neutron scattering. For these time-of-flight measurements at the GELINA white neutron source, the use of a fast digitizer significantly increases the efficiency compared with a conventional data acquisition system. First, the pulse processing time required to obtain the amplitude with high resolution is significantly reduced so that pulse pile-up with the prompt gamma-burst is eliminated for neutron-induced events. Second, an improved time response is obtained for which the amplitude and rise time dependence is strongly reduced compared to that of a conventional constant fraction discriminator. Excellent energy and time resolution is obtained with algorithms suitable for on-line signal processing, so that data storage is under control. Bench tests are presented that compare methods of signal processing. For the best method, the data acquisition system based on the fast digitizer was tested during measurements of gamma production cross-sections for 206 Pb and 208 Pb. A direct comparison was made with results obtained with conventional electronics operated in parallel. r 2007 Elsevier B.V. All rights reserved. PACS: 29.30.Kv; 28.20.v; 25.40.h; 24.10.1 Keywords: Fast digitizer; Large volume HPGe detectors; Time-of-flight; Neutron inelastic scattering
1. Introduction Precise neutron-induced cross-sections are needed for nuclear applications under study like the new generation of reactors (the Generation IV initiative) or accelerator driven systems. At GELINA there is a continuous interest for new measurements of neutron-induced cross-sections with improved energy resolution and total uncertainty. New measurements of neutron inelastic, ðn; 2ngÞ and ðn; 3ngÞ cross-sections provide one example. For this purpose, a new experimental setup was created at the 200 m flight path station [1,2] of the GELINA white neutron source for highresolution time-of-flight (t.o.f.) measurements. The measurements are based on the ðn; xngÞ-technique which involves detecting the g-rays emitted following inelastic scattering, (n,2n) and (n,3n) reactions. The gamma production cross-section is the most important quantity Corresponding author. Tel.: +32 14 571 489; fax: +32 14 571 381.
E-mail address: [email protected] (A.J.M. Plompen). 0168-9002/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.nima.2007.05.231
determined in such an experiment. It is obtained for every observed transition in the full neutron energy range, from threshold to about 20 MeV. Large volume HPGe detectors of 75–104% relative efficiency are used for the detection of the g-rays. The incident neutron energy resolution for the main transitions is determined by a time resolution of 4–8 ns over a time-span of 20 ms. The desired total uncertainty of the gamma production cross-sections of about 5% for the most intense transitions requires long acquisition times (e.g. 500 h). Therefore, improvements of the detection efficiency are important to increase the flexibility and use of this experimental setup. Despite the very good results obtained with the setup shortly described above and in more detail in Refs. [1,2], two evident limitations emerged: (i) the acquisition system based on conventional electronics has a dead time of about 10 ms which is significant as a result of the strong prompt gamma-flash of the accelerator and (ii) the slow-rise-timerejection (SRTR) function of the constant fraction discriminator (CFD) was used to achieve good time
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
resolution for the large volume HPGe detectors resulting in a strong loss of efficiency for gamma-ray energies below 500 keV. As it will be shown later, the time spectrum obtained with the Constant Fraction (CF) function has one peak with a long tail that lasts for about 700 ns for an HPGe detector with a 104% relative efficiency. This tail is due to the pulses with a slow rise time and they occur more frequently at low energies of the g-ray. The SRTR function rejects all the events in the tail. The rejected events are about 55% for the detector mentioned above in the case of a 60 Co source. At GELINA, the neutron production is based on a linear accelerator and on a natural uranium target. Electrons are accelerated up to 150 MeV by the linear accelerator and then stopped in the U target producing bremsstrahlung. The neutrons are produced through ðg; xnÞ and (g, fission) reactions on the U target. Due to this production mechanism, every neutron burst is preceded by a flash of g-rays (bremsstrahlung radiation). This gamma-flash is scattered on the sample placed at the 200 m flight path station and then detected by the HPGe detectors. The gamma-flash induced events in every HPGe detectors amount to about 10–20% of the total number of bursts. This percentage depends on the properties of the sample, the detection efficiency of the HPGe detector and the angle of observation. To avoid complicated dead time corrections due to the pile-up of the neutron-induced events and the gamma-flash induced events, all the neutron bursts in which a gamma-flash induced event was detected had to be rejected. This has been done in hardware, constructing an inhibit signal [1,3,4]. Preserving the same number of detectors, the detection efficiency can be increased if the dead time is reduced and if the use of the SRTR rejection is avoided. The solution for these two limitations was a new data acquisition system based on a fast digitizer. The difficulties in using large volume HPGe detectors arise from the large variations in the rise time and in the shape of the pulse on the rising edge. A digital signal processing algorithm can deal better with such pulses. Here, the use of a particular fast digitizer is investigated to improve the dead time and time-response of the employed HPGe detectors so that high incident neutron energy resolution together with good gamma-ray energy resolution is preserved while significantly improving the detection efficiency. This paper will present the tests for t.o.f. measurements with a commercial fast digitizer, the digital algorithms used for the signal processing and the results for neutron inelastic scattering measurements. Throughout this paper the fast digitizer will be compared with the data acquisition system based on conventional electronics. The final comparison was done during measurements of neutron inelastic scattering cross-sections for two different enriched samples of 206 Pb and 208 Pb, where the output of preamplifier from the HPGe detectors was read simultaneously by the fast digitizer and by the data acquisition based on the conventional electronics.
299
2. The fast digitizer A fast digitizer is a flash analog-to-digital converter that records samples of the input signal with a high repetition rate and transfers them for signal processing. Two properties define the performance of a fast digitizer: the number of bits for the amplitude and the sampling rate. The number of bits defines the amplitude resolution and the sampling rate defines the time resolution. For the project described here, a commercially available DC440 digitizer from Acqirist was chosen. This has 420 MSamples/s maximum sampling rate which corresponds to a minimum sampling interval of about 2.38 ns and has 12 bits for the amplitude. The full scale range of the inputs can have the following values: 0.25, 0.5, 1, 2, 5 and 10 V. The input impedance is 50 O. Every DC440 card has two input channels and a common external trigger input. This card has the trigger time interpolation function that allows a very precise determination of the trigger moment. In this way, the arrival time of the external trigger is determined within one sampling interval. The clock accuracy of the card is less than 0.2 ppm. The preamplifier output of the detector is given directly to one of the two independent input channels. The DC440 module has no possibility of onboard signal processing and in consequence the full set of samples has to be transferred to the computer. There the signal is processed online or recorded on the disk. The digitizer system used here has two DC440 cards plugged in a CompactPCIt create. The create is connected to the computer through a PCI bus. An important limitation of the DC440 module is the transfer speed through the PCI bus (up to 100 MB/s). For the ðn; xngÞ cross-section measurements at the 200 m flight path station, about 10 000 samples ð24 msÞ are needed to cover the neutron energy range of interest (500 keV up to 20 MeV). In the worst case when both cards have a trigger simultaneously, 40 000 samples have to be transferred. The transfer of these data can be done in slightly less than 1.25 ms which is the time interval between two consecutive neutron bursts (800 Hz repetition rate of GELINA). Moreover, for the typical conditions of these experiments, the average counting rate is less than 10 neutron-induced events per second so that no dead time results from the data transfer.
3. The experimental setup Prior to implementing the online data acquisition with the DC440 digitizer we tested off-line the pulse processing algorithms on data recorded with a start–stop setup and a 60 Co g-ray source. Once the best pulse processing algorithms were chosen, the electronic scheme for the online data acquisition was created. The electronic schemes used for off-line tests and for t.o.f. measurements are presented in this section.
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
300
3.1. Start–stop setup with
60
Co source
simultaneously the time and the amplitude spectra and their coincidences.
To compare precisely the time and amplitude resolutions obtained with the DC440 digitizer and with the acquisition system based on the conventional electronics, start–stop experimental setups were created: one with the fast digitizer and another with the conventional electronics. The scheme of both setups is given in Fig. 1. The same detectors and the same 60 Co source were used in both cases. In the setup with the DC440 fast digitizer (Fig. 1, left) the HPGe preamplifier output was given to one input channel of the digitizer and the split output of the plastic scintillator (Pilot-U) to the second channel from the same card. The common external trigger of the digitizer card was the output of the scintillator CFD. The plastic scintillator output was given to the second input channel only to have a check for the precision of the external trigger. For every external trigger, the digitizer recorded 10 000 samples that were saved in binary files on the computer hard-disk. In the start–stop setup with conventional electronics (Fig. 1, right) spectra were recorded with both functions of the CFD module, SRTR and CF. The delay used for the output of the scintillator CFD had a precision of a few picoseconds. This delay and the reset signal were needed for a proper functioning of the multi-hit time-to-digital converter (FTD) [5,6]. This start–stop setup allowed saving
3.2. The setup for t.o.f. measurements For the measurement of the neutron inelastic scattering cross-section with the digitizer, the electronic setup of Fig. 2 was used. The purpose of the conventional electronic modules was only to reduce the counting rate to the rate of the neutron-induced events, i.e. avoid triggering by the g-flash. For this, an external trigger was constructed to select only the events that are generated within about 19 ms after the arrival of the gamma flash. This time range allows the detection of the neutrons from about 500 keV up to 20 MeV. The trigger is the t0 signal from the accelerator which signals the arrival of the electron burst at the uranium target. This trigger is validated by the presence of a neutron-induced signal in one of the two HPGe detectors connected to one card. For every external trigger, the digitizer card records 10 000 samples and transfers them to the PC where the signals are processed online. On the disk were recorded the amplitude and the time of the pulse. An additional flag that gives the number of pulses in every signal was recorded to check the consistency of the recorded data. During the measurements with the samples of 206 Pb and 208 Pb the output signals of all the four HPGe ADC
SA
PCI bus
PC
TFA
CFD
cable delay
CFD CFD
stop reset
Pilot-U
Pilot-U
external trigger
digitizer DC440
HPGe
HPGe
input 1 input 2
precise delay
FTD start
MMPM
PC
Fig. 1. Two independent start–stop experimental setups for testing different signal processing algorithms: one with the DC440 digitizer (left) and the other with conventional electronics (right).
Fig. 2. Left: the t.o.f. setup based on the fast digitizer Acqiris DC440. One card of the digitizer accommodated two HPGe detectors. Conventional electronic modules were still used only to reduce the counting rate to the neutron induced counting rate (about 6 counts/s in average). Right: the corresponding timing of the signals.
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
detectors of the setup were split to be connected to the digitizer and to the conventional acquisition system.
4. Digital algorithms for the fast digitizer The signals recorded from the HPGe detectors (see the upper panel from Fig. 3) were transferred to the PC and then analyzed to extract the time and the amplitude of every pulse. The difficulties in using the large volume HPGe detectors are due to the large variations in the rise time and in the shape of the pulse on the rising edge. Examples of the experimental pulse shapes are shown in -12 uk N2
vk
amplitude (arb. units)
-22 D
N1 -32 0
10000
5000
10
5
0 0
5000
10000
channels (1ch=2.38ns)
Fig. 3. Up: the preamplifier output as recorded with the digitizer (initial signal vk ) compared with the resulting signal (uk ) of the recursive algorithm from Eq. (6). The uk signal has a higher amplitude compared with the initial signal vk due to the correction for the ballistic deficit. Down: the result of the complete trapezoid algorithm, resulted from the differentiation of the signal uk .
Fig. 4 for a large volume HPGe detector with 104% relative efficiency. This detector has a crystal with the closed-end coaxial shape. The edges of the crystal were rounded. The crystal length was 6.3 cm and the diameter was 8.7 cm. The pulses from the output of the detector preamplifier were recorded with the DC440 fast digitizer. The rise time of the pulses varies from less than 100 ns to about 700 ns and these variations are larger with the increase of detector volume. The time and the amplitude algorithms that were tested in the present work are described in this section. 4.1. Timing On the initial signal, vk (k is the sample index), a pulse finding algorithm was applied which identified the number of pulses. This algorithm consists of a differentiation of the signal followed by the verification that peaks in the differentiated signal exceed a certain threshold. If no pulse was found, the signal was discarded. In case pulses were found, a digital Timing Filter Amplifier (TFA) algorithm was applied to reduce the noise and to bring to zero the offset of the signal. This algorithm consists of a two step recursive expression with discrete output signal ok : t2 D D=ð2t1 Þ ð1Þ D D=ð2t2 Þ ð2Þ ok ¼ A e ok e ok (1) t2 t2 t1 t1 where t1 is the integration time constant, t2 the differentiation time constant and A the amplification (here, A ¼ 1). D is the sampling interval (2.38 ns): oð1;2Þ ¼ k
1 X ðvkl elD=t1;2 Þ.
(2)
l¼0
Each factor o1;2 k can be written recursively: D=tð1;2Þ ð1;2Þ ok . oð1;2Þ kþ1 ¼ vk þ e
(3)
With the assumption that the input signal is constant before the first sample k ¼ 1 is recorded, from Eq. (3) it results
-20
Voltage (arb. units)
301
oð1;2Þ ¼ v1 1
-25
-30 0
200
400 600 Time (ns)
800
1000
Fig. 4. Different shapes of the signal from the preamplifier output of a 104% efficiency detector as recorded with the DC440 digitizer. Rise time values up to 700 ns are observed. The corresponding energy of these pulses is about 600 keV.
1 X eD=ð2t1;2 Þ ðeD=t1;2 Þl v1 D=t1;2 l¼0
(4)
and therefore o1 ¼ 0. On the output of the TFA algorithm, we tested three different algorithms for the determination of the time: Leading Edge Threshold (LET), Extrapolated Leading Edge Threshold (ELET) and Constant Fraction Discriminator (CFD). All of them have a well-known equivalent in the analog electronics (Ref. [7]). The LET algorithm is the simplest solution for the time determination. The time of the pulse is the moment when the signal crosses a given threshold, l. The LET algorithm is sensible both to time jitter and walk effect. The ELET algorithm [7] consist of two thresholds l 1 and l 2 , with l 1 ¼ k l 2 and k usually a small integer number. The ELET algorithm is insensitive to the walk effect as long the signal rises linearly.
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
The CFD algorithm used here is the digital version of the method described in Ref. [7]. The recursive equation used to construct the CFD signal ck is ck ¼ fokd ok
with kXd
(5)
where d is the delay and f the attenuation constant. To reduce the computing time, the CFD signal ck was constructed only on a short region around the pulse of interest. The digital pulse processing allowed also adjusting the baseline before the determination of the zero-crossing point. 4.2. Amplitude determination In conventional electronics the key element for pulse amplitude determination is the spectroscopic amplifier. This provides simultaneously the proper pulse amplification to match the requirements of the ADC that follows and the pulse shaping, reducing the long tail from the preamplifier. In choosing the most appropriate shaping method, a compromise between three general concepts has to be found: signal-to-noise ratio, pile-up of the pulses and ballistic deficit. The best signal-to-noise ratio is given by the cusp-like shaping followed by the triangular shaping. To overcome the ballistic deficit a shaping with the flat-top comparable with the maximum rise time is needed. The trapezoidal shaping meets simultaneously good signal-tonoise ratio and good compensation for the ballistic deficit [8]. The basic function of the trapezoid algorithm is to convert the exponentially decaying signal from the preamplifier into a trapezoidal signal. This implies the convolution of the preamplifier signal (input signal vk for the digitizer) with different time dependent functions. Several proposals for such methods can be found in the literature (Refs. [9,10]), but the differences between them is only the order in which the convolutions are done. The algorithm described here was optimized also for a high speed calculation. In the algorithm used here, two steps can be distinguished in the construction of the trapezoidal output signal. First the preamplifier signal is corrected for the finite decay time t and converted into a step function. The correction for the finite decay time t was done using the following relation: ukþ1 ¼ uk þ ðvkþ1 v0 ÞeDt=2t ðvk v0 ÞeDt=2t
(6)
where Dt is the sampling interval. The resulted step signal uk is shown in the upper panel of Fig. 3. The second step is to differentiate the signal uk obtaining the trapezoid shape. To maximize the computing speed and using the timing information obtained already, only two averages on the uk signal are done, one to the left of the starting time (on the interval N 1 ) and another to the right (on the interval N 2 ), after a maximum rise time D (Fig. 3, up). This average gives the maximum height of the trapezoid (Fig.3, down). Moreover N 1 and N 2 can be
dynamically determined as the maximum time interval between two consecutive pulses. Prior to the application of the trapezoid algorithm it is necessary to determine the decay time, t, of the preamplifier and the signal offset, v0 . The decay time t was determined from a least square fit of the decaying part of the pulse. The fit was done once for every detector, for a large number of pulses to assure a small statistical error. A Gaussian distribution was obtained for the decay time t. The centroid of the distribution was taken. For the 104% relative efficiency detector t ¼ 47:8 ms with the full width at half maximum (FWHM) of 0:7 ms. This distribution comes mainly from the noise of the signal and from the finite time interval (number of channels) used for the fit. 5. Results and discussion 5.1. Results with the
60
Co source
The timing algorithms mentioned in Section 4.1 were tested off-line on the same sets of signals. The best results for each algorithm are shown in Fig. 5. For the TFA algorithm, an integration time t1 ¼ 10 ns and a differentiation time t2 ¼ 200 ns were used. The parameters used for the CFD algorithm were d ¼ 31 ns and f ¼ 0:25. A short delay gives good results because it determines the zerocrossing very close to the starting time of the pulse where the rise often appears nearly linear (Fig. 4). The FWHM is 6.2 ns for the CFD algorithm and 18 ns for the worst case, the LET algorithm. A threshold l ¼ 15 keV was used for the LET algorithm while for the ELET algorithm l ¼ 15 keV and k ¼ 2. The parameters of the algorithms were obtained from trials with different values. Following this comparison, the CFD algorithm was chosen for the online data acquisition. Fig. 6 shows the comparison between the time resolution obtained with the digitizer and with the conventional electronics for two detectors with different relative 600 LET ELET CFD
400 Counts
302
200
0 0
100
Fig. 5. Time resolution with the algorithms.
200 Time (ns) 60
300
400
Co source and different digital
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
303
1000 75.9 %
Digitizer SRTR CF
Digitizer SRTR CF
104 %
1000
100 Counts
Counts
100
10
10 1
0.1
1 0
100
200
300
400
500
600
Time (ns)
0
200
400
600
800
1000
Time (ns)
Fig. 6. The comparison of the time resolution with the digitizer and with the conventional electronics for detectors with different relative efficiencies: left: 75.9% and right: 104%. The two functions of the analog CFD module were used: Constant Fraction (CF) and slow-rise-time-rejection (SRTR).
400 digitizer-trapezoid conventional system
Counts
300
200
100
0 1326
1328
1330 1332 Eγ (keV)
1334
1336
Fig. 7. The amplitude spectrum of the 60 Co source around the 1.33 MeV peak with the digitizer and with the conventional electronics for a 104% relative efficiency detector.
efficiencies: 75.9% and 104%. Energies E g X150 keV were selected in all these cases, both for the digitizer and for the conventional module. The fast digitizer and the conventional electronics give the same FWHM(5.8 ns for 75.9% detector and 6.2 ns for the 104% detector). Moreover the full width at 10th maximum (FWTM) is almost the same for the digitizer and for the conventional electronics. The difference is in the tail, defined here as all the counts outside one FWTM. The CF spectrum has a large tail that extends to 400 ns for the detector with the smaller volume and to 700 ns for the larger detector This tail represents 37% and 55% of the full spectrum, respectively. The fast digitizer preserves the detection efficiency of the CF function and the tail is significantly reduced to about 21% of the number of counts. The SRTR spectrum has no tail because all events with zero-crossing times later than leading edge times are rejected. This rejection is the reason for the lower detection efficiency and is more severe for lower gamma-ray energies.
The amplitude spectrum obtained both with the digitizer and with the conventional acquisition is shown in Fig. 7 for energies around the 1.33 MeV peak of 60 Co. FWHM ¼ 2:1 keV at 1.33 MeV for both the digital and the conventional system in the case of the 104% relative efficiency detector. This resolution was obtained with the following values for the trapezoid algorithm (Section 4.2): N 1 ¼ N 2 ¼ 3:3 ms and D ¼ 1:4 ms. The value of D was chosen to be larger than the maximum observed rise time and to avoid any shape fluctuations in the pulse at the end of the rise. Different values of N 1 and N 2 were tried between 0.5 and 5 ms. With the increase of the averaging intervals N 1 and N 2 the energy resolution improves and saturates around 3 ms. A minimum of about 1 ms was needed to still have a reasonable resolution (better than 2.5 keV). From these tests it was concluded that a minimum of 2:5 ms spacing between two consecutive pulses (pile-ups) is needed in the present trapezoid algorithm to have an amplitude resolution better than 2.5 keV at 1.33 MeV. Pulses separated by less than 1:4 ms (the value of parameter D) are not distinguished by the present algorithm and are treated as a single pulse. This is a large improvement in dead time in comparison with the conventional acquisition where 426 ms are used as shaping time. For the online data acquisition, the N 1 and N 2 were determined dynamically to adapt to the time interval between two consecutive pulses. If N 1 or N 2 are smaller than 1 ms then a flag is set for that event. For the counting rate of the present experimental setup the number of these pulses is negligible. For investigating the effect of the digitizer’s number of bits on the amplitude resolution of the detector, three different full scale ranges of the digitizer were used: 250, 500 mV and 1 V for the same gain of the detector’s preamplifier. A slight deterioration of the resolution was observed when the full scale range was increased, but still 2.3 keV amplitude resolution was obtained for the 1 V full scale range. It can be concluded that a digitizer with 12 bits amplitude range is sufficient for used with HPGe detectors at g-ray energies below 10 MeV.
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
304
5.2. Results for neutron inelastic cross-sections The fast digitizer was used for neutron inelastic scattering measurements in 206 Pb and 208 Pb with the setup described in Section 3.2 and with online signal processing. The two acquisition systems, based on the DC440 digitizer and on the conventional electronics, were used simultaneously only for comparison. The data were analyzed independently. Independent absolute efficiency calibrations were done. The ratio between the detection efficiency of the acquisition system with the fast digitizer and with the conventional electronics is given in Fig. 8. The efficiency losses due to the SRTR function in the conventional electronics are recovered in the fast digitizer. Three hundred percent is gained in the efficiency at energies around 400 keV and 15% at 1.4 MeV. Two examples of differential gamma production crosssection are shown in Fig. 9 for the 2614 and 583 keV transitions from 208 Pb nucleus. The measurement angles were 1101 and 1501. The results agree well within
the stated uncertainties. The magnitudes of the uncertainties are given as generic error bars at 6 MeV and at 15 MeV. The time resolution of the two acquisition systems can be compared when resonance structures are visible. This is the case for the main transition of 206 Pb (803 keV) at neutron energies immediately above the inelastic threshold. The resonance structures in the differential gamma production cross-section of the 803 keV transition are shown in Fig. 10 for the two acquisition systems. The data were recorded in parallel with the same detector (104% relative efficiency) at 1501. The two systems provide almost the same result. The small difference in the resonance amplitude may be the effect of the tail in the time resolution spectrum of the digitizer (see Fig. 6). This small difference does not outweigh the higher efficiency of the digitizer.
0.12 digitizer conventional
0.1
dσ/dΩ (barn/sr)
4.0
εd / εc
3.0
2.0
0.08 0.06 0.04 0.02 0
1.0 200
400
600
800
1000
1200
800
1400
900
Eγ (keV)
Fig. 8. The ratio between the detection efficiency of the digitizer and the detection efficiency of the conventional acquisition system with the SRTR function.
Fig. 10. High resolution differential gamma production cross-sections obtained for the 803 keV transition from 206 Pb with the fast digitizer and with the conventional acquisition system. 0.15
digitizer conventional
2614 keV - 208Pb 0.1
110°
0.05
digitizer conventional
dσ/dΩ (barn/sr)
0.15 dσ/dΩ (barn/sr)
1000
En (keV)
583 keV - 208Pb
0.1
150° 0.05
0
0 5000
10000 En (keV)
15000
5000
10000
15000
En (keV)
Fig. 9. Comparison of the differential gamma production cross-sections obtained with the fast digitizer and with the conventional electronics for 208 Pb. Both excitation functions were smoothed with a running average filter only for an easier comparison. Left: differential gamma production cross-section at 1101 for the main transition of 208 Pb nucleus (2614 keV). Right: differential gamma production cross-section at 1501 for the second transition of 208 Pb nucleus (583 keV).
ARTICLE IN PRESS L.C. Mihailescu et al. / Nuclear Instruments and Methods in Physics Research A 578 (2007) 298–305
6. Conclusions A new data acquisition system based on a 12-bits and 420 MSamples/s fast digitizer was created to increase the detection efficiency in the ðn; xngÞ ðx ¼ 1; 2; 3Þ cross-section measurements with large volume HPGe detectors at GELINA. The time and amplitude resolutions obtained for the same detectors with the fast digitizer and with the conventional electronics were compared. Different signal processing algorithms were tested off-line on the same sets of raw signals (as recorded from the output of the preamplifier) to establish the most suitable one for the online acquisition. The trapezoid algorithm for the amplitude and the CFD algorithm for the timing were preferred for the online signal processing. With the fast digitizer the amplitude and the time resolutions were identical with the results obtained with the conventional electronics. The advantages of the fast digitizer are firstly the fact that the time spectrum of the digitizer has a much smaller tail than the CF function of the CFD module and secondly that the minimum resolving time of trapezoid algorithm is only 2:5 ms, much smaller than the one of the spectroscopic amplifier. Therefore, the total dead time of the digitizer is 2:5 ms which allows the separation of neutron-induced events from the gamma-flash induced events for neutrons up to 20 MeV at the 200 m flight path length. The pile-ups between neutron-induced events were neglected, due to the very low counting rate. The differential gamma production cross-sections obtained with the data acquisition system based on the fast digitizer and with the data acquisition system based on the conventional electronics were compared for two sets of measurements, with 206 Pb and 208 Pb samples. The two acquisition systems give the same result for the differential gamma production cross-sections and the advantage of the fast digitizer is a higher detection efficiency, that increases from 15% at 1.4 MeV up to 300% at 400 keV. The increase in the efficiency of the fast digitizer has two sources:
The digitizer resolves the pile-up of the neutron-induced events with the gamma-flash induced events. Efficiency of 10–20% is gained in this way for every detector. The digitizer does not reject the pulses with a slow rise time and has the same efficiency as the CF function of the conventional CFD module. The tail of the time resolution spectrum is significantly reduced for the digitizer compared with the analog CF module.
It was observed that the present data acquisition with the fast digitizer is more sensitive to irregularities in the output signal of different detectors compared with the conven-
305
tional electronics. The performances of the fast digitizer are lower for detectors with lower signal-to-noise ratio, with two components of the decay signal or with changes in the signal offset (v0 ) during the data recording. Such effects were observed especially for old detectors. From the results of the tests it was concluded that the 12-bits and 420 MSPS fast digitizer can be used successfully with large volume HPGe detectors for low counting rates as it is the case for the ðn; xngÞ setup at the 200 m flight path station at GELINA. Mainly because of the transfer speed between the digitizer and computer the present acquisition system is limited to a counting rate of about 100 Hz. Substantially, higher counting rates can be accommodated by a fast digitizer if an on-board data processing would be possible, e.g., if the digitizer has on-board Field Programmable Gate Array (FPGA) that can be used for a complete signal processing. In this manner, the limitation imposed by the transfer time between the digitizer and the PC can be completely eliminated. The on-board signal processing would simplify even further the electronic scheme given in Section 3.2, avoiding the use of any conventional electronic module for the external validation. Acknowledgements L.C.M. and C.B. are grateful to the EC/JRC for financial support. References [1] L.C. Mihailescu, L. Ola´h, C. Borcea, A.J.M. Plompen, Nucl. Instr. and Meth. A 531 (2004) 375. [2] L.C. Mihailescu, C. Borcea, A.J.M. Plompen, Tests for the use of a fast digitizer in a time-of-flight measurements with large volume HPGe detectors, in: Proceedings of the Enlargement Workshop on Neutron Measurements and Evaluations for Applications, NEMEA2, Bucharest, 2005, p. 109. [3] L.C. Mihailescu, Neutron ðn; xngÞ cross-section measurements for 52 Cr, 209 Bi, 206;207;208 Pb from threshold up to 20 MeV, Report EUR 22343 EN, ISBN 92-79-02885-5, ISSN 1018-5593, European Communities, 2006. [4] L.C. Mihailescu, C. Borcea, A. Koning, A.J.M. Plompen, Nucl. Phys. A 786 (2007) 1. [5] S. de Jonge, Fast time digitizer, Internal Report GE/DE/R/24/87, JRC-CBNM, Geel, 1987. [6] J. Gonzalez, C. Bastian, S. de Jonge, K. Hofmans, Modular multiparameter multiplexer, MMPM, hardware description and user manual, Geel, 1997. [7] G.F. Knoll, Radiation Detection and Measurement, third ed. Wiley, New York, USA, 2000. [8] V. Radeka, IEEE Trans. Nucl. Sci NS-19 (1) (1972) 412. [9] V.T. Jordanov, G.F. Knoll, A.C. Huber, J.A. Pantazis, Nucl. Instr. and Meth A 353 (1994) 261. [10] A. Georgiev, W. Gast, IEEE Trans. Nucl. Sci. NS-40 (1993) 770.