Nuclear Instruments and Methods in Physics Research A 730 (2013) 124–128

Contents lists available at ScienceDirect

Nuclear Instruments and Methods in

Physics Research A
journal homepage: www.elsevier.com/locate/nima

Real time digital pulse processing for X-ray and gamma ray
semiconductor detectors
L. Abbene n, G. Gerardi, F. Principato
Dipartimento di Fisica e Chimica, Università di Palermo, Viale delle Scienze, Edificio 18, Palermo 90128, Italy

art ic l e i nf o a b s t r a c t

Available online 25 April 2013 Digital pulse processing (DPP) systems, based on direct digitizing and processing of detector signals, have
Keywords: recently been favoured over analog electronics, ensuring higher flexibility, stability, lower dead time and
X-ray and gamma ray spectroscopy better spectroscopic performance. In this work, we present the performance of a new real time DPP
Digital pulse processing system for X-ray and gamma ray semiconductor detectors. The system is based on a commercial digitizer
Pulse height analysis equipped with a custom DPP firmware, developed by our group, for on-line pulse height and shape
Pulse shape analysis analysis. X-ray and gamma ray spectra measurements with cadmium telluride (CdTe) and germanium
Real time processing (Ge) detectors highlight the excellent performance of the system both at low and high rate environments
High photon counting rate (up to 800 kcps). These results make the proposed DPP system a very attractive tool for both laboratory
research and the development of advanced detection systems for high-rate-resolution spectroscopic
imaging, recently proposed in diagnostic medicine, industrial imaging and security screening.
& 2013 Elsevier B.V. All rights reserved.

1. Introduction Due to the high performance of the new analog-to-digital

converters (ADCs), digital pulse processing (DPP) electronics,
There is an increasing variety of X-ray and gamma ray applica- based on the direct digitization of the preamplifier output pulses,
tions requiring robust detection systems able to count each imping- has been recently proposed [1–5]. In digital systems ADC directly
ing photon and perform accurate measurements of photon energy converts the preamplified signals, while in analog ones the
even at high photon counting rates. Optimum detection systems digitization is performed after the analog shaping amplifier. This
should provide high throughput (i.e. the photon counting rate in the new approach leads to better results than the analog ones, i.e.
measured spectra related to the input photon counting rate), very better (i) stability, (ii) flexibility, (iii) reproducibility, (iv) energy
good energy resolution and good stability of the energy peak position resolution, (v) throughput, (vi) easy dead time modeling and (vii)
in the measured spectra over the widest range of input counting shape preservation of the pulses for further analysis. It is possible
rates. Besides the properties of the detectors and preamplifiers, the to use complex algorithms, that are not easily implementable
characteristics of the pulse processing electronics are crucial in the through a traditional analog approach, for adaptive processing and
development of high-rate-resolution spectrometers. Traditionally, optimum filtering. The digital analysis also requires considerably
pulse processing electronics consists of analog shaping amplifiers less overall processing time than the analog ones ensuring lower
coupled to multichannel analyzers (MCAs): the detector signals dead time and higher throughput, both very important under high
(preamplifier output signals) are shaped and filtered by the shaping rate conditions. There is no additional dead time associated with
amplifier and finally processed by the MCA. The MCA is capable of digitizing the pulses and so there is no MCA dead time. Several
evaluating and digitizing (analog-to-digital converters) the height of DPP systems are composed of a digitizer and a PC wherein the data
the output pulses from the shaping amplifier and generating the are recorded and analyzed (off-line analysis) [4,6–12]. Real time
energy spectrum. Pulse pile-up (tail and peak pile-up) and dead time data processing [1,3,5], in which the signals are acquired, recorded
losses are the major drawbacks at high photon counting rates and, and processed on-line, is obtained by using digitizers with local
despite the implementation of sophisticated baseline restoration and memory and field programmable gate arrays (FPGA) where pulse
pile-up rejection techniques, analog electronics fails in these hostile processing algorithms can be implemented (DPP firmware).
environments. In this work, we present the performance of a new real time DPP
system for X-ray and gamma ray semiconductor detectors. The
system is based on a modified version of a commercial digitizer
equipped with a custom DPP firmware, developed by our group, for
on-line pulse height and shape analysis. Experimental results on
n
Corresponding author. Tel.: +39 091 23899151; fax: +39 091 23860815. cadmium telluride (CdTe) and germanium (Ge) detectors, at both
E-mail addresses: [email protected], [email protected] (L. Abbene). low and high photon counting rates, were reported.

0168-9002/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.nima.2013.04.053
 L. Abbene et al. / Nuclear Instruments and Methods in Physics Research A 730 (2013) 124–128 125

2. Detectors, sources and analog electronics

To investigate the performance of the DPP system, two differ-

ent semiconductor detectors (CdTe and Ge) were used. As well
known, CdTe and CdZnTe detectors are very appealing for X-ray
spectroscopy in the 1–100 keV energy range [13–19], while ger-
manium detectors are excellent spectrometers in the gamma
energy range (4 200 keV) [14,20]. We used a planar CdTe detector
(XR100T-CdTe, S/N 6012, Amptek, USA) and a coaxial germanium
detector (GEM40P4-76, Ortec, USA). Both detectors are equipped
with resistive-feedback preamplifiers.
For comparison, a fast analog pulse processing chain was also
used. The preamplifier output pulses were shaped by an analog
shaping amplifier (Amptek PX2T). The semi-Gaussian output
pulses, characterized by a peaking time of 630 ns, were acquired
Fig. 2. The digitized preamplifier output waveform (white line) and the shaped
by a standard multichannel analyzer (MCA-8000A, Amptek, USA).
pulses (red line) through the fast shaping operation. A zoom of the picture clearly
The linear amplifier is also equipped with a baseline restorer shows the good detection of the pulses from the waveform. The pulses represent
circuit for high rate measurements. the 109Cd photons impinging on a CdTe detector with a photon counting rate of
We measured the system response to four X-ray and gamma ray 830 kcps. (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)
calibration sources (109Cd: 22.1, 24.9 and 88.1 keV; 57Co: 122.1 and
136.5 keV; 137Cs: 661.7 keV and 60Co: 1173.2 keV and 1332.5 keV).

therefore, working as the classical pole-zero cancellation techni-

3. Overview of the DPP system que. The width of each shaped pulse is equal to Td+Tp, where Td is
the delay time and Tp is the peaking time of the preamplifier
The digital system is composed of a digitizer and a PC wherein the output pulse. To optimize the detection of the events and their
user can set the DPP parameters, choose the working mode and energy measurement, the DPP method is characterized by two
display the results by a Labview program, developed by our group parallel shaping modes: a “fast” SDL shaping mode and a “slow” SDL
(Control and Display software). The pulse processing analysis is shaping mode. The fast SDL shaping, characterized by a short delay
performed by using a custom DPP firmware, developed by our group time, is used to detect the pulses from the preamplifier output
and implemented on the digitizer. Fig. 1 shows a block diagram of the waveform with high time resolution, as shown in Fig. 2. To
system coupled to a detector. To digitize the output pulses from perform the pulse height analysis, the DPP method used the
preamplifiers, we used a commercial digitizer (DT5724, CAEN S.p.A., combination of a stage of differentiation (slow SDL shaping mode)
Italy) [21], housing high speed ADCs (14 bit and 100 MS/s), a buffer followed by a stage of integration (low-pass filter). Each pream-
memory and ALTERA Cyclone EP1C20 FPGAs. The digitizer is plifier output pulse, selected by the fast SDL shaping (selection
equipped with 4 channels. Each channel is characterized by one described in Section 4.2), is analyzed for height estimation within
ADC with three full scale ranges (71.125 V, 70.5625 V and a time window ST, termed as snapshot time. The selected pulses are
70.2813 V). The data stream from each ADC is written in four shaped by using a slow SDL operation, characterized by a longer
circular memories without dead time. The digital pulse processing is delay time than those of the fast ones. In order to obtain pulses
carried out by the dedicated FPGA, in which the DPP method is characterized by a width equal to ST/2+Tp, this operation used a delay
implemented (DPP firmware). The produced data are packed and time of ST/2. The maximum amplitude of each shaped pulse is at the
another FPGA collects and sends them, via USB channel, to the PC. center position of the snapshot time window ST. The slow SDL shaping
is used to perform (i) a baseline restoration, by evaluating the mean
value of the samples preceding the leading edge and (ii) a pulse shape
4. DPP method analysis, by evaluating the maximum amplitude and the rise time of
the shaped pulses. Finally, the pulse height estimation is performed by
The DPP firmware is based on a revised version of a DPP method, applying an optimized low-pass filter to all the samples of the shaped
developed by our group and successfully used for off-line analysis pulses within the ST window.
[9–12]. In the following subsections, we summarize the main To perform the shape analysis of the pulses, the method is able to
operations performed by the method. measure the peaking time of the analyzed pulses (after slow SDL
shaping). The rise time of the pulses, i.e. the interval between the
4.1. Pulse height and shape analysis times at which the shaped pulse reaches 10% and 90% of its height
(after baseline restoration), is first evaluated. The times, corresponding
Once digitized, the preamplifier output pulses were shaped by to the exact fractions (10% and 90%) of the pulse height, are obtained
using the classical single delay line (SDL) shaping technique through a linear interpolation. Finally, the method estimates the
[4,10,20]. Each shaped pulse is obtained by subtracting from the peaking time equal to 2.27 times the rise time (i.e. about five times
original pulse its delayed and attenuated fraction. The attenuation the time constant) with a precision of 2 ns.
is used to avoid undesirable undershoots in the shaped pulses, To compensate incomplete charge collection, pulse shape
discrimination (PSD) and linear pulse shape correction (PSC)
techniques [11] were implemented.

4.2. Pile-up rejection

Because the width of each shaped pulse is well known, it is

possible to perform a pile-up inspection within the fast shaping
Fig. 1. Block diagram of the DPP system. operation (fast pile-up rejection). If the width of a shaped pulse
126 L. Abbene et al. / Nuclear Instruments and Methods in Physics Research A 730 (2013) 124–128

exceeds a maximum width threshold (ΔTfast), the pulse is rejected

(i.e. classified as representative of pile-up events) and it is not
analyzed by the slow shaping procedure.
Moreover, to avoid interference between adjacent pulses, the
method performs a further pulse selection during the fast shaping
operation (slow pile-up rejection): a pulse will be rejected for the
pulse height analysis if the time spacings between the pulse and
both the previous and the following pulse are lower than ST/2.
As shown in our previous works [9,11,12], the pulse shape
analysis allows to minimization of peak pile-up events, i.e. over-
lapped preamplifier pulses within the peaking time that are not
detectable through the fast shaping operation. Therefore, we used
a pulse shape discrimination (PSD) technique to minimize pile-up Fig. 4. The snapshot waveform. Each preamplified pulse is presented within a time
events, based on the selection of a range of peaking time values of window equal to the selected snapshot time ST.
the pulses that are not piled up.

4.3. Dead time correction

The well defined dead time modeling is one of the most

interesting characteristics of a DPP system. Knowledge of the dead
time and of its analytical model is very important to estimate the
true input counting rate. Both the maximum width threshold ΔTfast
and the snapshot time ST are dead times for the DPP system and
they can be modeled as paralyzable dead times. Concerning the
fast shaping operation, the true photon counting rate ntrue is
related to the measured photon counting rate mfast through the
following equation:
mf ast ¼ ntrue expð−ntrue ΔT f ast Þ ð1Þ

It is possible to evaluate the true rate ntrue from the measured

rate mfast by solving Eq. (1) iteratively. Therefore, the DPP system is
always able to estimate the true rate through Eq. (1), the measured
rate and the knowledge of the dead time. Fig. 3 shows the rate of
the events analyzed by the fast (ΔTfast ¼ 200 ns ) and the slow
pulse shaping (ST ¼6 ms) channels versus the true input counting
rate (109Cd events measured by using the CdTe detector). The true
input counting rate was estimated from mfast and by using Eq. (1).
As shown in Fig. 3, the measured rates from both the fast and the
slow shapings are in agreement with the paralyzable dead time

Fig. 5. (a) Measured 57Co spectra using PSD and no PSD techniques. After PSD, we
obtained an energy resolution of 0.60% FWHM at 122.1 keV. (b) An energy
resolution of 0.80% FWHM at 122.1 keV was obtained after non linear PSC, with
no reduction of photon counts.

model. This result highlights that our DPP system is always able to
Fig. 3. Throughput of the DPP system: the measured photon counting rates from estimate the true rate of the impinging photons through the fast
fast and slow shaping (red and blue points, respectively) versus the true input channel, even for low rates in the spectrum (slow channel).
counting rate. The red and blue lines are the paralyzable dead time model functions Therefore, it is possible to use long ST values (low throughput)
for the fast (dead time of 200 ns) and slow (dead time of 6 ms) shaping, respectively.
The system analyzed the 109Cd events measured with a CdTe detector. (For
for optimum pulse height analysis (optimum energy resolution)
interpretation of the references to color in this figure legend, the reader is referred without perturbing the correct estimation of the input photon
to the web version of this article.) counting rate.
 L. Abbene et al. / Nuclear Instruments and Methods in Physics Research A 730 (2013) 124–128 127

4.4. Working modes shape correction (PSC). We used a snapshot time ST¼ 40 ms.
Despite the strong reduction of the photon counts (93%), PSD
To obtain the overall characteristics of the impinging photons produced good spectral improvements in the spectrum: energy
and to perform any physics application, the DPP can transmit resolution of 1.6% FWHM at 122 keV with no correction and of 0.6%
several results to PC: (i) the input waveform, (ii) fast and slow FWHM after PSD. Good results were also obtained (0.8% FWHM at
shaping waveform, (iii) energy and peaking time list, (iv) energy, 122 keV) by using the non-linear PSC, with no reduction of photon
peaking time and the time of occurrence list, (v) time of occur- counts.
rence and the fast pulse width list and (vi) the snapshot waveform, We also measured 109Cd spectra with the CdTe detector at
i.e. a sequence, with length chosen by the user, of preamplified various rates (up to 830 kcps) by using the DPP system and the
pulses with their time of occurrence. In the last working mode, the analog system. We first compared the performance of both
system transmits the preamplified pulses selected for the pulse systems at the same throughput (Fig. 6). The results, shown in
height analysis, i.e. isolated from the preceding and following Fig. 6(b), point out the better performance of the DPP system at
pulses in the input waveform. Each pulse is presented within a high photon counting rates. Moreover, Fig. 6(b) shows further
time window equal to the selected snapshot time. Contrary to performance improvements at a lower throughput. In this last
what happens in an analog pulse processing system, the pream- case, we used a longer snapshot time (ST¼ 7.5 ms) looking for the
plifier pulse shape is preserved. Fig. 4 shows an example of a short best energy resolution. The 109Cd spectra at 830 kcps, measured
snapshot waveform. Each working mode is set through the control with both systems, are shown in Fig. 7. The better performance of
and display software. the DPP is clearly visible: energy resolution of 4.6% FWHM at
22.1 keV. Peak pile-up events were minimized by using the PSD
technique. The strong shift of the energy peak position for the
5. Measurements and results analog system and the absence of shift for the digital ones are also
well evident (the energy calibration of both systems is performed
Fig. 5 shows 57Co spectra, measured with the CdTe detector, at low rate).
after digital pulse shape discrimination (PSD) and non-linear pulse We also tested the DPP system with a germanium detector in
the gamma energy range. Figs. 8 and 9 show the measured 137Cs

Fig. 6. (a)Throughput of the DPP system: high and low conditions. (b) Energy Fig. 7. 109Cd spectra measured with the CdTe detector by using both (a) analog and
resolution (FWHM) at 22.1 keV at various rates measured with both the DPP system (b) digital systems. The counts were normalized to the total number of the detected
(high and low throughput) and the analog electronics (high throughput). events.
128 L. Abbene et al. / Nuclear Instruments and Methods in Physics Research A 730 (2013) 124–128

presented. The system is based on a modified version of a

commercial digitizer equipped with a custom DPP firmware,
developed by our group, for on-line pulse height and shape
analysis. X-ray and gamma ray spectra measurements with CdTe
and Ge detectors showed the excellent performance of the DPP
system, especially at high rates: (i) low photopeak centroid shift,
(ii) low worsening of energy resolution and (iii) the minimization
of peak pile-up through pulse shape analysis. Moreover, a com-
parison with a fast analog electronics highlighted the better high-
rate-resolution capabilities of the digital approach. The high
flexibility of the system was also pointed out. Due to the several
working modes, the system allows investigations on the overall
characteristics of the impinging photons and so the possibility to
perform any physics application.
The DPP system is a very attractive tool for both laboratory
research and for the development of advanced detection systems
for high-rate-resolution spectroscopic imaging, recently proposed
in diagnostic medicine (computed tomography and mammogra-
Fig. 8. 137Cs spectrum measured with the Ge detector by using the DPP system.
Energy resolution FWHM of 0.29% at 662 keV. phy), industrial imaging and security screening.

References

[1] M. Bolic, V. Drndarevic, Nuclear Instruments and Methods A 482 (2002) 761.
[2] J.M. Cardoso, et al., Nuclear Instruments and Methods A 522 (2004) 487.
[3] M. Arnold, et al., IEEE Transactions on Nuclear Science NS-53 (2006) 723.
[4] G. Gerardi, et al., Nuclear Instruments and Methods A 571 (2007) 378.
[5] T. Papp, J.A. Maxwell, Nuclear Instruments and Methods A 619 (2010) 89.
[6] U. Bottigli, B. Golosio, G.L. Masala, P. Oliva, S. Stumbo, P. Delogu, M.E. Fantacci,
L. Abbene, F. Fauci, G. Raso, Medical Physics 33 (2006) 3469.
[7] A. La Manna, L. Abbene, F. Fauci, G. Gerardi, G. Raso, S. Del Sordo, L. Barone
Tonghi, S. Stumbo, Nuovo Cimento C 29 (2006) 361.
[8] L. Abbene, A. La Manna, F. Fauci, G. Gerardi, S. Stumbo, G. Raso, Nuclear
Instruments and Methods A 571 (2007) 373.
[9] L. Abbene, G. Gerardi, F. Principato, S. Del Sordo, R. Ienzi, G. Raso, Medical
Physics 37 (2010) 6147.
[10] L. Abbene, G. Gerardi, S. Del Sordo, G. Raso, Nuclear Instruments and Methods
A 621 (2010) 447.
[11] L. Abbene, G. Gerardi, Nuclear Instruments and Methods A 654 (2011) 340.
[12] L. Abbene, G. Gerardi, F. Principato, S. Del Sordo, G. Raso, Sensors 12 (2012)
8390.
[13] S. Del Sordo, L. Abbene, M. Zora, G. Agnetta, B. Biondo, A. Mangano, F. Russo,
E. Caroli, N. Auricchio, A. Donati, F. Schiavone, J.B. Stephen, G. Ventura,
Fig. 9. 60Co spectrum measured with the Ge detector by using the DPP system. G. Bertuccio, S. Caccia, M. Sampietro, IEEE Transactions on Nuclear Science
Energy resolution FWHM of 0.17% at 1332 keV. NS-52 (2005) 3091.
[14] S. Del Sordo, L. Abbene, E. Caroli, A.M. Mancini, A. Zappettini, P. Ubertini,
Sensors 9 (2009) 3491.
[15] L. Abbene, S. Del Sordo, F. Fauci, G. Gerardi, A. La Manna, G. Raso, A. Cola,
and 60Co spectra. We used an ST of 25 μs and 30 μs for 137Cs and
60
E. Perillo, A. Raulo, V. Gostilo, S. Stumbo, Nuclear Instruments and Methods A
Co spectra, respectively. The energy resolution obtained with the 583 (2007) 324.
DPP system is in agreement with the performance specifications of [16] L. Abbene, S. Del Sordo, E. Caroli, G. Gerardi, G. Raso, S. Caccia, G. Bertuccio,
the detector: energy resolution FWHM of 0.29% and 0.17% at Journal of Applied Physics 105 (2009) 124508.
[17] N. Auricchio, L. Marchini, E. Caroli, A. Zappettini, L. Abbene, V. Honkimaki,
662 keV and 1332 keV, respectively. Journal of Applied Physics 110 (2011) 124502.
[18] F. Principato, G. Gerardi, A.A. Turturici, L. Abbene, Journal of Applied Physics
112 (2012) 094506.
6. Conclusions [19] L. Abbene, et al., Nuclear Instruments and Methods A, http://dx.doi.org/10.
1016/j.nima.2013.03.016, in press.
[20] G.F. Knoll, Radiation Detection and Measurement, third ed., Wiley, New York,
The performances of a new digital pulse processing (DPP) 2000.
system for X-ray and gamma ray semiconductor detectors were [21] 〈http://www.caentechnologies.com〉.