DP5

1. INTRODUCTION

The DP5 is a high-performance digital pulse processor. The DP5 is one component of a complete nuclear spectroscopy system that also includes:

1. detector and preamplifier;
2. power supply.

A complete system can be assembled by combining the DP5 with one of the Amptek detectors, a preamplifier (several options and configurations can be used) and an Amptek PC5 power supply. The user can also supply their own detector, preamplifier and/or power supply. The DP5 is designed for use with high resolution solid state detectors, but can also be used with scintillator/PMT systems, proportional counters and other detectors. The DP5 is a printed circuit board with electronics that is primarily suitable for OEM applications as part of a complete system.

DP5 is the second generation of digital pulse processors (DPP) that replace the amplifier-shaper and multichannel analyzer used in analog systems. Digital technologies improve several key parameters: 1) higher performance, in particular, higher resolution and higher count rates; 2) significant system flexibility due to a large number of possible settings selected using software; 3) improved stability and reproducibility. DPP digitizes the output signal of the preamplifier, applies digital signal processing in real time, detects the amplitude peaks and places them in the histogram memory. The spectrum is then transferred to the user's computer.

In the standard configuration, only three connections are required: power (+5 VDC), communications (USB, RS232 or Ethernet), and an analog input from the preamplifier. An auxiliary connector provides several additional inputs and outputs used when integrating the DP5 with other equipment. This includes an MCA gateway, timing outputs, and eight SCA outputs. The DP5 also includes an "interconnect" designed primarily for interfacing with Amptek power supply boards, but is available to OEMs. The DP5 comes with ADMCA software for data acquisition and detector settings control, as well as DLL libraries for integrating the hardware with customer software. Optional additional hardware includes X-ray spectrum analysis software, several collimators and mounting hardware, and X-ray tubes to create a complete compact X-ray fluorescence analysis system.

Figure 1-1. Photograph of DP5 (left) and the characteristic spectrum of 55Fe obtained with the XR-100SDD detector.

2. DESCRIPTION OF DP5

A complete, standard nuclear spectroscopy system includes several key components:

1. Detector
2. Preamplifier
3. Pulse conversion board (including pulse shaper, pulse selection circuit, pulse counter, multichannel analyzer, data acquisition and control interface)
4. Power supply
5. Packaging or case
6. Software for setting up the detector, collecting and analyzing the received data.

The DP5 is a digital pulse processor that implements the functions described in (3) and is one component of a complete spectrometric system. The DP5 has been designed to provide maximum versatility and can be adapted for use in a wide variety of systems. Designed as a single small board, the DP5 is the most suitable solution for integration with OEM solutions. This article provides a detailed specification and application examples for the DP5 board.

2.1 Functions of the main blocks

Figure 2.1 shows how the digital pulse processor (DPP) is used to process signals in the complete chain of the nuclear instrument system and its main functional blocks. The DPP digitizes the output signal of the preamplifier, applies digital processing to the signal in real time, determines the maximum amplitude (in digital form) and places them in a memory buffer, creating an energy spectrum. The pulse selection circuit can exclude pulses from the spectrum using various criteria. The spectrum is then transmitted through the DPP interface to the user's computer.

Dpp digitizes the preamplifier output, performs digital signal processing in real time, detects the peak amplitude and stores it in a memory buffer, this amplitude can be rejected by electronics depending on the criterion used.

Analog Preamplifier (Pre-filter) : The Dpp input is the output of the analog charge-sensitive preamplifier. The analog pre-filter chip prepares the signal for digital processing. The main functions of this circuit are (1) applying appropriate gain and mixing to "hit" the signal in the appropriate ADC range (2) filtering the signal and shaping it to optimize digitization.

ADC : The 12-bit ADC digitizes the output of the analog preamplifier at a frequency in the range of 20 – 80 MHz. The stream of digitized values is transmitted to the digital pulse shaper (Digital Pulse Shaping) in real time.

Digital Pulse Shaping : The ADC output is continuously processed using a pipeline architecture to generate a pulse in a form convenient for subsequent processing in real time. Pulse shaping is standard, similar to any other amplifier-shaper. The shaped pulse is a pure digital unit. The output can be redirected to the DAC, for diagnostic purposes, but this is not a mandatory requirement.

Inside the pulse shaper there are two components for signal processing - these are fast and slow channels, which are optimized for processing various information of the current pulse chain.

The slow channel has a long pulse formation time, which is necessary to obtain accurate pulse amplitudes. The peak height value for each pulse in the slow channel is the output signal value of the pulse former.
The fast channel is optimized for obtaining time information, namely, for detecting pulses that overlap in the slow channel, measuring the count rate, pulse rise time, etc.

Pulse Selection Logic: Eliminates pulses that cannot be accurately measured. Includes pile-up rejection logic, time discrimination, etc.

2.2 Analog preamplifier

The Dp5 is designed to process signals coming from a charge sensitive preamplifier used with solid state radiation detectors. These signals have (1) a small amplitude in the range of a few mV (2) a fast rise time (10 ns (or µs)) (3) and a small amplitude. These signals (steps) can be seen in the upper parts of Figure 2.2 . These signals are not suitable for digitization because of their small amplitude. The analog preamplifier prepares these signals for further digitization (blue curve).

The analog amplifier performs the following functions: (1) A high-pass filter with a time constant of 3.2 µs so that the pulses no longer overlap, (2) Amplified signal so that the largest pulses have an amplitude of approximately 1 V, (3) Shifted signal to fall within the range of the ADC. The output of the analog amplifier is shown in the figure by the blue line.

By default, the analog amplifier is configured for use with Amptek's XR100CR family of detectors (solid state detectors with a resettable preamplifier).

Systemgain

The system gain is measured in units of channels/keV: this gives the channel number in which a particular energy peak will appear. It is the product of three terms: (1) the gain of the charge-sensing amplifier (in units of mV/keV), (2) the total gain of the voltage amplifier (this is the product of the coarse gain and fine gain), (3) the gain of the MCA analyzer (channels per mV).

For Amptek's XR100CR detectors, the gain is typically 1 mV/keV. The MCA gain of the analyzer is given by the value of the selected number of channels (e.g. 1024) divided by the voltage corresponding to the channel in which the peak is located. In Amptek digital processors, this value is typically 950 mV. The DP5 gain is the product of the coarse gain and the fine gain. For example, if the fine gain is 1.00 and the coarse gain is 66.3, then the system gain is (1 mV/keV)(66.3)(1.00)(1024 channels / 950 mV) = 71.5 channels/keV. 1/71.5 channels/keV = 14 eV/channel is the MCA calibration factor. The full scale energy would then be 1024 channels / 71.5 channels per keV = 14.3 keV. However, these values are approximate due to manufacturing tolerances of feedback capacitors, resistors, etc. (errors amount to a few percent).

Reset and Continuous Preamplifiers

The charge-sensing amplifier produces a voltage proportional to the time integral of the current. The integrator eventually saturates because the current through the diode continually increases. There are two ways to maintain the preamplifier output in the desired range: reset and continuous feedback. Figure 2-4 (left) shows the output of the reset preamplifier over a long period of time: many small steps (a few mV) force the output signal to approach the negative limit (- 5 V) linearly over a period of several seconds. Then a reset pulse is triggered so that the output signal is set to + 5 V over a period of several µs. The reset amplifier provides a minimum of electronic noise and is therefore used in detectors. A very large transition during reset can affect signal processing, so the DPP includes "lock out" logic designed to eliminate unwanted effects.

Another traditional solution is to create a small feedback loop that restores the input signal to a value close to ground. In the simplest case, the feedback resistor Rf is placed in parallel with the feedback capacitor Cf, across which the current is integrated. After a voltage step ΔV due to the interacting signals, the output signal gradually drifts to the initial value, with a feedback time constant, as shown in Fig. 2-4 right. In the figure, this time constant is equivalent to 500 µs, which allows accurate calculation (integration) of the total charge, but causes pulse pile-up. The feedback resistor increases the electronic noise, so this circuit is not used in amptek detectors.

2.3 Pulse Shaping.

Slow channel.

The slow DPP channel is optimized for accurate peak height counting. It uses trapezoidal pulse shaping, an example of which is shown in Figure 2-5. This pulse shape provides the optimal signal-to-noise ratio for many detectors.

The user can adjust the rise or fall time (these times must be the same) and the duration of the flat top over many steps. A semi-Gaussian amplifier with a pulse shaping time of τ has a peak rise time of 2.2 τ and is comparable in performance to a trapezoidal pulse with the same peak rise time. A DPP with a peak rise time of 2.4 μs is equivalent to a semi-Gaussian shaper with a time constant of 1 μs.

Adjusting the peak rise time is a very important component in optimizing the system configuration. Usually, there is a trade-off: the shorter the peak rise time, the smaller the dead time, which increases the throughput and count rate, but with an increase in the peak rise time, the electronic noise of the system also increases. Optimal settings strictly depend on the type of detector and amplifier, as well as on the goals set. Electrical noise has a minimum at a certain value of the peak rise time. At peak rise times greater or less than this value, the noise value will increase, which will degrade the resolution.

If the peak rise time is very long compared to the incoming sample rate, pile-up will occur.

Fast channel

The fast channel is designed to detect pulses that overlap each other in the slow channel. The fast channel is used to reject pulses that are too close to be distinguished in the slow channel, and to determine the true count rate (corrected for events that were lost in the dead time of the slow channel). The fast channel also uses trapezoidal pulse shaping, however, the rise time of the peak in this case is in the range of 100-400 ns. Figure 2-6 shows the basic operation of the fast channel, pulses are measured with a peak rise time of 100 ns. As can be seen on the right, pulses that lag each other in time by only 120 ns are separately counted in the fast channel.

BASELINE RESTORATION (Baseline Pulse Reconstruction)

The amplitude of the pulses is implicitly calculated relative to the baseline. Any random fluctuation in the baseline, or any systematic change in it, will distort the amplitude measurement. The baseline is commonly called the "ground", but this is somewhat ambiguous since the ground is just some reference for the voltage measurement. If this baseline changes with time, count rate, or anything else, these distortions will show up in the measurements.

The baseline peak of a digital processor has significant differences from a traditional analog shaper-amplifier. This is due to the fact that the pulse after passing the chain has no effect on other pulses going along the chain (this is how I understood it!!!). This is a fundamental difference from analog differentiators and leads to a significant increase in the stability of baselines at high count rates.

Dpp has an asymmetric baseline with several different settings. DPP BLR uses negative peaks from random noise to determine the baseline. The negative peaks only appear when there is no signal, so if they are stable, then the baseline is stable, regardless of the count rate. BLR typically produces a shift comparable to the value of the RMS noise. There are two independent parameters, UP and DOWN, each of which can be set to four positions: Very Slow, Slow, Medium, and Fast. These are essentially slew rates in the baseline response. Setting both UP and DOWN to Very Fast will cause BLR to respond very quickly to any change in the baseline. It should be emphasized that the optimal setting strictly depends on the details of the practical application: the nature of the fluctuations, etc. If the peaks are found to shift to lower channels at high count rates, then increase the UP slew rate or decrease the DOWN slew rate. If one observes occasional “bursts” in the system which cause the spectrum to shift to higher channels (often manifesting as bursts of noise above the threshold), then decrease the UP slow rate or increase the DOWN slew rate.

2.3.2 Selecting pulses.

Dpp uses thresholds to detect pulses. Both channels (fast and slow) have their own thresholds. The noise is usually higher in the fast channel, and the best option for the fast channel is to set the threshold slightly higher than the noise. The threshold of the slow channel is used to determine which events will be added to the spectrum. Events with an amplitude smaller than the slow threshold are ignored. The threshold of the slow channel is equivalent to the lower-level discriminator (LLD).

The fast channel threshold also functions as a low-level discriminator and is used to achieve the following effects: (1) The event rate measured in the fast channel is the incoming flow measured by the incoming count rate (ICR) detector. (2) Pile-Up Rejection (PUR) is the logic that distinguishes between events that overlap in the slow channel but differ in the fast channel. (3) Rise Time Discrimination (RTD) uses the amplitude of the signal received in the fast channel to measure the current at the beginning of the pulse. PUR and RTD will be discussed in more detail below.

The correctness of setting these thresholds is very important in terms of obtaining the correct and most accurate information. Incorrect threshold settings entail many problems that arise for users. For example, if the threshold is too small in a fast channel and the PUR function is enabled, then each event will be rejected and, accordingly, no signal will be received. Similarly, if the threshold of a slow channel is very large, all events will be rejected.

Pulse-Up Rejection

This logic is used to detect two interactions that occur so close in time that the output merges into a single event with distorted amplitude. PUR uses a "Fast-Slow" system. Figure 2.7 shows the operation of DP4 for pulses (events) that are close in time.

In figure (a) two events are separated by less than the peak rise time, while figure (b) shows a completely different picture, where the pulses are well resolved, hence the time between their occurrences is much longer. In case (a), the output signal is the sum of the two pulses. However, the signals at the analog output (a) are resolved. For a near-triangular pulse shape, overlap occurs only when the two events are resolved by less than the peak time. The interval used in Dpp to reject events by criteria 1 – dead time and 2 – PUR is the sum of the rise time and the flat-top duration. If PUR is enabled and the two events are separated by more than the fast channel two-pulse resolution time (120 ns) and less than this interval, both will be rejected. If pile-up rejection is enabled and two events are separated by more than the fast channel pulse pair resolution (120 nsec) and less than this interval, both are rejected. Events that exceed a threshold in the fast channel trigger the pile-up reject logic.

Reset Lockout. (lock after reset)

As noted earlier, many preamplifiers use a pulse reset to prevent the preamplifier output from saturating. Reset generates a very long signal in Dpp, which leads to amplifier saturation, register overflow, etc. Therefore, Dpp has a reset detection circuit (which detects very large, negative pulses) and logic to block signal processing for some time after reset, this time serves to restore normal operation. Dpp offers the user to enable or disable the reset function (reset should be disabled for preamplifiers with continuous feedback). The user can also specify a time during which signal processing will be disabled. If the interval is chosen very small, then the signal shape (and therefore the spectrum) will be distorted. At high count rates, reset pulses appear much more often, so if this interval is chosen very large, therefore, the dead time of the detector will be very large.

RiseTimeDiscrimination

In some applications, it is important to discriminate the pulses based on the duration of the transient current through the detector to the preamplifier. For example, in some Si diodes there is a non-depleted region with a weak electric field. Radiative interactions in this region will generate a signal current, but the charge moves through this region slowly. Such interactions in this region can lead to various spectral distortions: background values, shadow peaks, peak asymmetry. In CdTe diodes, the carrier lifetime is so short that slow pulses exhibit charge deficit. These low-amplitude pulses distort the spectrum. In scintillators, pulse shape discrimination allows gamma rays to be distinguished from neutrons. Such pulse shape discrimination can be used in Dp5 using the RTD command.

Risetime discrimination rejects events with a long detector current that result in a slow rising edge in fast and slow pulse shapes. DP5 uses as a selection criterion a comparison of the peak amplitude in the fast channel to the peak amplitude in the slow channel. If this ratio is significantly large, the rise time is fast and therefore the pulses are considered valid. If the ratio is small, the pulses are rejected. Because the fast channel is inherently much noisier than the slower shaped channel, an RTD threshold is also implemented on the shaped channel. Events that are below this threshold (the “RTD Slow Threshold”) are not processed by the RTD and are therefore accepted (however, they may be rejected by PUR or other logics). RTD is often used to describe interactions that occur deep in the detector, at high energies; low energy events are unlikely to benefit from RTD on because they are below the RTD threshold and are therefore accepted.

Gate (Input Control Signal)

The control input signal is used with external circuitry to determine which events are included and which are excluded from the spectrum. The signal can be high active or low active (or off). If the signal is not present (off) then all events (that satisfy the selection criteria listed above) are counted. If the detector count activity is high (low) and the control input signal activity is high (low) then the event is recorded as valid. When the count rate is zero (gated off), the spectrum acquisition timer is also disabled, so the exact count rate can be determined. Adjusting the duration of this signal is very important. If the signal is active during the fast threshold triggering, then the event is interpreted as a fast count rate (count rate in the fast channel). If the signal is active when the peak detect is triggered (the peak detect is triggered), then the event is recorded at the slow count rate, and therefore displayed in the spectrum. Reminder: The fast and slow channels have different trigger times, and therefore have different pulse generation times.