1 Introduction

Digital pulse processing is a signal processing technique in which detector signals (pre-amplifier or shaping amplifier output) are digitized and processed in order to extract information. Such techniques have undergone rapid development in recent decades^[1-3]. These new techniques have been applied in many fields^[4-6]. The main advantages of digital pulse processing (compared to analog-based processing) have been summarized in previous studies^[7]. In this study, we developed a custom-built digital pulse processing method comprised of a FPGA (field programmable gate array) based acquisition card with a 16-bit resolution at a sampling rate of 100 MSPS (million samples per second). Moreover, we developed a method to analyze pulse height in order to determine the surface emission rates measurements for α and β emitters.

The main objective of this study was to develop a digital pulse processing method to replace the current instrumentation-based 2πα and 2πβ emitter measurement system ^[8]. The instrumentation-based system serves as a national primary standard at the National Institute of Metrology (NIM), and ensures the traceability of α and β surface emission rates. A basic block diagram of the standard analog emitter measurement system is shown in Fig. 1. In the standard analog system the detected α or β signal is processed through the pre-amplifier, the pulse shaping amplifier, and the single-channel pulse-height analyzer (SCA), successively^[9-12]. The pulse counter records signals when there are pulses generated from the SCA. The amplitude of the analog pulse at the output of the pulse-shaping amplifier is typically proportional to the energy of the detected event of the particle. Selection of a range of signal levels at the output of the amplifier is equivalent to the selection of a range of energies for these events^[13]. The SCA accomplishes the selection by producing an output logic pulse for the pulse counter when the amplitude of an input signal falls within the pulse-height window established by 2 preset threshold levels. In our method, signal acquisition and analysis determine the functions of the SCA and pulse counter.

Figure 1.Full-screen View

Simplified block diagram of the standard analog emitter measurement system.

The advantages of utilizing a digital pulse processing system are as follows: First, signal capture and processing allows for offline analysis and corrections, including noise cancelation, dead time correction, background subtraction, etc. Second, control procedures provide flexible options for pulse width fitting and threshold level setting (as determined by the needs of the surveyor). Third, energy extrapolation can be conducted to correct for β counting loss due to the continuous β energy spectrum. By manipulating the selected pulse, one can change the threshold for β counting in the data, and thus avoid statistical uncertainty from repetitive measurements. Furthermore, digitized pulse processing allows for further operations, such as shape discrimination for different particles.

2 Structures of the System

Our digital pulse processing method consists of an acquisition board and a computer running the control and analysis program, as shown in Fig.2. The acquisition board captures and digitalizes the pulse, and then selects the height of the pulse before transferring it to the computer. We used USB (universal serial bus) protocols for transmission, specifically a USB protocol with a micro USB peripheral controller on board called CYPRESS CY7C68013A^[14].

Figure 2.Full-screen View

Simplified block diagram of the digital pulse processing system.

In order to meet the different requirements for the acquisition of α and β emitter signals, we utilized separate conditioning amplifiers, in the front of the input side, to ensure the range of signal amplitudes was suitable for the range of the ADC (analog to digital converter) inputs^[15]. According to the tests completed using the oscilloscope (Agilent Technologies MSO6034A), the α signal from the pulse-shaping amplifier (ORTEC 572A) had an amplitude range of 0 to 1 V, and the β signal had an amplitude range from 0 to 10 V. Both signals were single-ended from the pulse-shaping amplifier with a 50Ω characteristic impedance coaxial cable. To create a differential ADC with an input range from -1 to 1 V, we chose TI LMH6554 (a conditioning amplifier chip) because of its exceptional signal fidelity and wide large-signal bandwidth (which is necessary for 8 to 16-bit high-speed data acquisition systems)^[16,17].

In order to acquire both α and β signals, we used an ADC with 2 channels of 16-bits and 100MSPS. According to ISO 8769, for the emission rate measurement of the β source, the measurement threshold should be 590 eV, which is equal to 10 percent of the energy of the 55Fe X-ray. So the voltage should be 20 mV at 1600 V high-voltage supply on the basis of experimental measurements. As the signal pulse has 10 V margin (maximum), the minimum voltage was chosen to be 2.5 mV to increase the precision as much as possible, and to determine the emission rate measurement of the β source. Moreover, 100MSPS is sufficient for sampling because the width of the signal, which is able to trigger the amplitude discriminator needs to be less than 100 ns. As a result, we chose ADI AD9268 (a standard ADC chip) for the acquisition process. According to the completed tests, the SNR (Signal-to-Noise Ratio) of AD9268 at 100 MHz was 74.3 dB (Signal generator: Tektronix AWG5000 series: 600 MS/s maximum sampling rate, 2.5 G bandwidth, 14-bit D/A resolution, 16 M memory length). While it was lower than the 78.2 dB shown in the datasheet of the device^[18], it still meets the requirement of the minimum voltage precision.

In this study, we used a digital to analog convertor (DAC) and a controllable high-voltage power supply to achieve high voltage control. The DAC provided a control voltage of 0 to 5 V (as determined by the configuration set by the user on the computer). The high-voltage power supply transformed 15V DC input into 0 to 3 kV output voltages, varying linearly with the control voltage. We chose TI DAC8881 (a DAC chip) for our method because of its linearity and low-power operation^[19].

3 FPGA Logic Module

The schematic diagram of the FPGA is shown in Fig.3. As there is a 16-bit channel for both the α data and the β data in the AD9268, the interface for the ADC is a 32-bit complementary metal oxide semiconductor (CMOS) with a clock of 100 MHz. We utilized the ALTERA Cyclone V in order to guarantee that the FPGA would work well and have sufficient memory blocks^[20]. The data sampled by the ADC was screened in the amplitude discriminator. When the input signal exceeded a preset acquiring threshold level, the amplitude discriminator generated a trigger signal. This threshold level was set below that of the traditional SCA in order to acquire more pulse signals for analysis. To reduce the effects of the high-frequency noise, we established a hold time in order to ensure the over-threshold state was not instantaneous. After the over-threshold was detected, we introduced a fixed dead time in order to prevent any other over-threshold detections. This helped us discard the tails of pulses with long time widths and the small pulses after a main pulse (an occurrence known as the ringing effect).

Figure 3.Full-screen View

Schematic diagram of the FPGA.

Signals were entered into the data buffer, in the form of data packages, when all the requirements were met. The data package, as shown in Table 1, had 128-byte data for each selected signal pulse. Each data package had 60 sample points. At the sample rate of 100 MHz, we obtained a sample point every 10 ns. As such, in order to record a pulse with a time width of about 10 μs in just one package, a mid-value point was extracted from every 16-sample points. This process sacrificed some of the utilization rate of the sampling, but it was necessary for analysis. For pulses with different time widths, the extraction ratio can be adjusted through configuration.

All the data collected in the data buffer was transferred to the USB peripheral controller at a speed of 48 million bytes per second (MBPS). According to the tests we conducted, the speed of the transmission between the USB peripheral controller and the computer was 11.2 MBPS. When there were 20,000 pulses per second, 5.12 Mbytes of data were transferred (20,000 x 2 x 128 byte). Our tests show that our method works very well at this pulse rate (and meets the requirements for most emitter measurements).

The surveyor controlling the system can conveniently set the parameter configuration of the amplitude discriminator and high-voltage controller by transferring the command to the FPGA through the USB peripheral controller. As there is a time margin in data transmission, occupation of the USB data bus has no influence on our method’s operations (which we proved conclusively in our tests).

4 Control and Analysis Programs

The analysis program contains an online mode and an offline mode. The flow diagram of the analysis program is shown in Figure 4. In the online mode, before the acquisition, the surveyor must set the parameters of the amplitude discriminator in the FPGA (the parameters being the acquiring threshold level, hold-time, acquisition time, and the extraction ratio). Each configuration command is also sent as a command package, as shown in Table 2.

Figure 4.Full-screen View

Flow diagram of the control and analysis program.

The received data is stored in the computer as a hex file (in the form of a data package). For convenience sake, the data is stored in one file every minute in sequence. The pulse-shape analysis process automatically reads the file regardless of whether it is in online mode or offline mode. In the analysis, the pulse amplitude was extracted from the data then compared with threshold. The threshold level can be set according to the demands of the surveyors. Thus, the pulse counts in the range of the pre-set threshold are easily calculated out. Besides the amplitude analysis, the width of the pulse is also taken into account in order to exclude ambient noise. FWHM (Full Width at Half Maximum) is calculated for every pulse. The signal with a FWHM in a certain range is regarded as the signal pulse, and the other pulses are treated as ambient noise pulses^[21].

By utilizing the procedures outlined above, the counts of the α and β pulses (at specific threshold levels) provide conclusive results. According to the spectrum of the α source, the energy of the pulses is concentrated near its spectral line. Therefore, we only need to set an appropriate threshold level and make the necessary calculations in order to determine the event counts for the α emitter. However, for the β emitters, according to the spectrum of the Sr-90/Y-90, the energy of the pulses is distributed over the entire spectrum. As such, we need to extrapolate in order to obtain the zero-energy counting rates. The spectrum of the β sources can be obtained by setting threshold levels according to the results of each signal measurement. The counting rates of the different threshold levels can then be used to determine the linear fitting, which, in turn, yields the counting rate of the β source. Importantly, the dead time during the acquirement of the signal should be counted in order to correct the counting rate^[22]. The corrected counting rate is

where N is the counting rate (s-1) of measurement, and τ is the dead time (μs) of the acquisition (which can be set by the surveyors).

5 System Validation

In order to test our method’s ability to reconstruct pulses, we first used a signal generator (Tektronix AFG3251 series) to simulate the pulses generated by the α and β sources. Both of the α and β pulses from the pulse-shaping amplifier were Gaussian-like pulses. We used a 50 Ω resistance channel to simulate the pulses and fanned them out by the connector. We sampled the signals with an oscilloscope (Tektronix MSO5204) and our method simultaneously. The α pulse had a 2.4 μs FWHM with an amplitude of 200mV, as shown in Fig. 5 (a). The β pulse had a 2.4 μs FWHM with an amplitude of 5V, as shown in Fig. 5 (c). The pulses were trigged by a random external trigger source at a 20 K/s trigging rate. The results of the test demonstrate that our method works very well. The sampled pulses were reconstructed in order to compare them with the pulses sampled by the oscilloscope. The restored α pulse sampled by our method is shown in Fig. 5 (b). The reconstructed β pulse sampled by our method is shown in Fig. 5 (d). Compared to the measurements from the oscilloscope, we found that our method recreated the shape of the pulse with high fidelity. This level of accuracy is crucial for pulse shape analysis.

Figure 5.Full-screen View

a) Simulated α pulse sampled by the oscilloscope; b) Reconstructed α pulse sampled by our method; c) Simulated β pulse sampled by the oscilloscope; d) Reconstructed β pulse sampled by our method.

In order to verify the accuracy of the acquisition channels, we used a signal generator to generate sine waves for amplitude tests. The sine waves had a 20 μs period and various peak-to-peak values. After capturing and digitalizing the sine waves, the peak-to-peak values (Vpp) of the reconstructed waves were measured and compared to those it had generated. The results of the amplitude tests of the α channel are shown in Table 3. The results of the amplitude tests of the β channel are shown in Table 4. We found minor errors in both channels, which were all below 1% and sufficient for processing.

We conducted the spectrum measurement in the National Metrology Institute of China, using the newly designed multi-wire proportional counter. The sources used for measurement were the α emitters (Pu-239 and Am-241) and the β emitters (Sy-90/Y-90). The emitters were incorporated into the surface of an anodized aluminum foil approximately 0.3 mm thick. The anodized aluminum foil was then placed in the center position of the detecting area. The measurements were taken at 0.1 MPa with 100 mL/min gas flow speed with a mixture of 90% argon gas and 10% methane by volume. The pre-amplifier was a CAMBERRA MODEL 2006 and the pulse-shaping amplifier was an ORTEC 572A. The voltages were set as 700 V for the α emitters and 1600 V for the β emitters. The spectrum for the α emitter Pu-241 obtained by using our method is shown in Fig. 6. The spectrum for the β emitter Sy-90/Y-90 obtained by using our method is shown in Fig. 7.

Figure 6.Full-screen View

Spectrum of the α emitter Am-241 obtained by using our pulse processing method.

Figure 7.Full-screen View

Spectrum of the β emitter Sr-90/Y-90 obtained by using our pulse processing method.

We conducted a counting rate measurement under the conditions of the spectrum measurements outlined above. As there were two output channels on the pulse-shaping amplifier ORTEC 572A, we were able to simultaneously conduct measurements using both methods (one channel using our method and the other channel using SCA and the pulse counter). The acquiring threshold of our method was set to 40 mV for the α channel and 10 mV for the β channel. The threshold of SCA was set to 60 mV for the α source and 20 mV for the β source. Analysis processes of our method set the threshold value equal to that of SCA. The dead time of our method was set to 10 μs and the hold time was set to 100 ns.

According to the measurement results, the counts measured by our method were greater than that displayed by the pulse counter. This result reflects the more precise identifiable pulse width in our method than of that in SCA. In order to reduce the effects of ambient noise, we conducted measurements at a working voltage of 1600 V (without emitters in the detector). The mean value of the ambient counting rates was 34.911 for our method and 30.988 for the ORTEC SCA.

For the sake of correctness, we conducted additional measurements with different set dead times in order to verify the algorithm of our dead time correction. As shown in Table 5, we used the same source Sr-90/Y-90-2 with activity at magnitude of 10² for all measurements. In spite of the measurement errors, the results indicated that our dead time correction was accurate and effective.

Factoring in the dead time corrections and ambient noise subtraction, we conducted measurements for each of the 6 sources in order to compare our method with the ORTEC SCA and pulse counter. The uncertainty of the measurements was 0.397% and 0.885% for the α and β channels, respectively. The results are shown in Table 6. We found that for the α channel, the percent difference between our method and the current primary standard was below 0.3%. The percent difference for the β channel was below 1%. These results demonstrate that our method is as accurate and precise in the measurement of the 2πα and 2πβ emission rate measurement as the primary standard method.

6 Conclusion:

In this study, we presented a pulse processing method for 2πα and 2πβ emission rate measurement. Test results indicate that our method is precise in pulse reconstruction and acquisition accuracy. Moreover, our method works well in obtaining spectrums because it is in good agreement with the counting rate measurement of the primary standard method used at the NIM. Furthermore, our new method is more convenient than the present method because it is operated solely by computer, which allows the digitized data to be quickly and easily analyzed. Importantly, our method is applicable to most 2πα and 2πβ emitter measurements. We will conduct further research into automatically identifying the α and β sources by analyzing the shape of the acquired pulses. We will also conduct further research into the analysis of multiple particle resources.