Optimizing Amptek DP5 for High Count Rates
Uploaded by
Filipe LaínsOptimizing Amptek DP5 for High Count Rates
Uploaded by
Filipe LaínsFAQ High Rate Performance in Amptek FASTSDD® Detectors
Amptek’s FASTSDD® detectors can be used at count rates exceeding 1 Mcps. Operating at such high rates will
lead to improved statistical precision or to reduced measurement times, both of which are extremely valuable. As a
general rule, operating at high count rates is advantageous1. But this is only true if the spectrometer is properly
configured, and even when properly configured, there are trade-offs. The primary goal of this application note is to
explain (1) what can be expected when operating Amptek’s FASTSDD® detectors at high count rates and (2) how to
optimize the spectrometer settings for best operation at high count rates.
What is typically observed for the performance of the FASTSDD® at high count rates?
The charts below show results measured at Amptek, using a 25 mm2 FASTSDD®, an Amptek DP5-X digital pulse
processor, with a 50 kV X-ray tube (Ag anode) exciting a 99.9% pure Zn target. Data were taken at T pk=0.1 s (for the
highest rates) and at Tpk =1.0 s (for lower rates but better resolution). Various signal processing parameters in the
DP5-X were tuned to optimize the system for high count rates (how to optimize is discussed later in this application
note). The tube current was adjust to give input count rates from 10 kcps to over 1 Mcps; spectra were measured to
obtain a fixed number of counts in the photopeak channel.
The plots below show spectra measured at Tpk=0.1 s (top two) and at Tpk=1.0 s (bottom two), at input count
rates of 10 kcps, 100 kcps, 500 kcps, and 1 Mcps . The plot on the left shows that the Zn photopeaks are unchanged:
the resolution of the Zn K peak is 167 eV FWHM at Tpk=0.1 s and 152 eV FWHM at Tpk=1.0 s. Notice that the peak
to background and secondary peaks are unchanged even at the highest rates. There are sum peaks; the sum peak
rates are as expected, they are Gaussian, and their FWHM is as expected and does not change with rate. The plots on
the right shows that the spectra are as expected: at the highest rate, the higher order sum peaks and the sum
continuum are seen.
1
Redus, R.H., Huber, A., “Figure of merit for spectrometers for EDXRF,” X-Ray Spectrom (2012)
AMPTEK, Inc. 14 DeAngelo Drive, Bedford, MA 01730-2204 USA
+1 781 275-2242 Fax: +1 781 275-3470 [Link] sales@[Link]
Products for Your Imagination
The plots below show the throughput (left) and change in photopeak FWHM and centroid (right) in the spectra
shown above.
What are the intrinsic limitations to operation at high count rate?
Peaking Time, Count Rate, and Resolution
The most important factor determining both count rate and energy resolution is the peaking time. As the two
charts below show, to operate at high count rates, the system must be configured for a short peaking time, and this
necessarily increases the resolution. The user needs to understand what is best for each particular application and
adjust the peaking time appropriately.
Dead Time, Throughput, and Sum Peaks
Because it takes some time to process each X-ray, (a.k.a. the dead time per pulse, roughly equal to Tpeak+Tflat) and
the X-rays interact at random time, there is always some probability that pulses will overlap in time. This means that
some pulses will not be counted, reducing the throughput.
When pulses overlap, the result (called “pile-up”) is a pulse with distorted pulse height. Amptek’s signal
processors include pileup reject (PUR) logic to eliminate the vast majority of these pulses. The PUR logic does require
a minimum time between the pulses (the “pulse pair resolving time”); two pulses occurring within this interval will (a)
yield a single pulse with height equal to the sum of the two pulses and (b) not be rejected. The result is a “sum peak”.
The fact that there is a dead time per pulse, that it limits throughput, and the existence of a sum peak are
intrinsic to any pulse counting radiation detection system. Please refer to Amptek’s application notes to understand
how to accurate determine the true, input count rate with Amptek’s digital pulse processor 2. For background
information on dead time, sum peaks, and so on, there are excellent textbooks available 3.
Are there any anomalies at high count rates?
If the system is configured correctly, there should be no “anomalies.” We expect sum peaks (and sum
continuum); we expect dead time, as shown above. We expect small changes in FWHM and centroid, as shown above.
2
Products for Your Imagination
If the system is not configured correctly, then many types of anomalies are indeed possible. Larger dead times,
larger shifts in centroid or FWHM, are possible. Various distortions of the photopeak shape, e.g. tails on the upper or
lower end, excess low amplitude counts, these are all possible.
If a customer sees “anomalies” at high count rates, please contact Amptek technical support. Please save a .MCA
file at low count rate and another .MCA taken at high count rates, and send these to Amptek. Seeing the complete file
is quite important, so Amptek support has the raw data and can view all the configuration settings.
What is the maximum count rate for the Amptek FASTSDD®?
It depends on the requirements for spectral quality and purity. With the shortest peaking time (50 ns) and flat
top (12.5 ns), the system can operate at over 4 Mcps input rate. But the resolution will be quite poor (due to both
noise and ballistic deficit) and the resolution and centroids will shift. For some applications, which involve counting
strong peaks in a wide ROI, this may be adequate.
For applications requiring good quality spectra, input count rates of 1 Mcps are quite achievable, as shown
above. With careful tuning and certain applications, input rates can go higher, even to 2 Mcps.
Is Amptek’s default configuration the best for high count rate?
No! By default, when Amptek tests a system at the factory, it is configured for the best energy resolution at the
55
Fe photopeak. It uses a relatively long peaking time, so clearly T peak needs to be shortened. As shown below, other
parameters will likely need to be changed as well.
What configuration do you recommend for high rates?
The table below shows the key parameters which should be adjusted, for Amptek’s digital pulse processors. The
first column shows settings from Amptek’s default configuration, used in the factory to measure 55Fe resolution. The
second and third columns show the configuration settings used for the spectra shown above. The second column
used Tpeak=1.0 s and is appropriate for count rates up to of 50 to 300 kcps. The third column used T peak=0.1 s and is
recommended for count rates of 300 kcps to 1 Mcps or above.
3
Products for Your Imagination
What factors are important, if I want to optimize the configuration for myself?
Tpeak, Tflat, Tfast
As noted above, Tpeak is the most important parameter in determining the maximum count rate. A short peaking
time allows operation at higher count rates but leads to degraded resolution (the noise bandwidth of the filtering
electronics is greater). The maximum output count rate is found at a dead time of 68%, but in general, we
recommend limiting the dead time to 50% or lower. A higher dead time implies a larger correction and this can
degrade accuracy.
For SDDs and FASTSDDs, we recommend a Tfast of 0.1 s. You can make it shorter, but since this gets convolved
with the charge collection time in the SDD, you don’t gain much in the ability to resolve closely space pulses which the
noise in the fast channel increases significantly. For the lowest energy applications, e.g. EDS, where Tpeak is long for
the best resolution, you can increase Tfast to reduce the noise threshold in the fast channel.
For most peaking times, a Tflat of 0.2 s is recommended. For the shortest peaking times, this can be reduced,
but if it is made too small, an effect called “ballistic deficit” will make the FWHM worse and the photopeak shape will
be non-Gaussian, with a tail towards lower amplitudes. A Tflat shorter than 0.15 s may give ballistic deficit. Note that
Tflat should always be longer than Tfast; if it is shorter, the shape of the sum peak is distorted.
Gain
The gain of the digital processor should always be set so that the full spectrum is observed on the screen: off
range pulse lead to distortions. At high rates, the gain should be reduced further: there can be tail pileup at the ADC
input, and when this occurs, there will be anomalies in the spectrum. There is no “hard and fast” rule since it depends
on both the rate and the spectrum, but as a general rule, lower gain is important to support high rates.
BLR and PURS
In early 2019, Amptek introduced a new version of firmware for the DP5 and DP5X processors, which included a
new and improved baseline restoration algorithm (mode 3) and a “secondary” pileup rejection algorithm. This
secondary pileup reject can detect and reject many events which were not rejected in the older, primary algorithm, so
it reduces the sum peak intensities by typically half. These are described in a separate application note; please refer
to that note for the
Reset lockout
For high rates, we recommend a reset lockout interval of 15 s for the FASTSDD. Every time the preamp resets,
the pulse shaping is distorted; data acquisition is stopped for the lockout interval, to avoid distortions. If the interval is
shorter than 15 s, then some distortions can be seen in the spectrum (typically a tail on the photopeaks). If the
interval is much longer, then the difference between acquisition time and real time becomes quite long and the
effective measurement time increases.
Thresholds
After the other key parameters are adjusted (Tpeak, Tflat, Tfast, gain) then it is important to adjust the fast and slow
thresholds. This is discussed in the DPPMCA Help file.
Common questions
Powered by AIOperating at count rates exceeding 1 Mcps is beneficial where high statistical precision or reduced measurement time is required. At such rates, the resolution degrades due to increased noise and ballistic deficit. It's suitable for applications involving strong peak counting in wide regions of interest (ROI) where precision in data is more critical than resolution quality. Trade-offs include poorer spectral resolution and potential centroid shifts, which may not be tolerable in applications requiring high spectral fidelity .
The peaking time is essential in determining both count rate and energy resolution for the detectors. A shorter peaking time allows operation at higher count rates, which is crucial for applications requiring improved statistical precision or reduced measurement times. However, this also increases noise and degrades resolution (ballistic deficit), affecting the spectral quality. Thus, the user must balance count rate performance with resolution considerations according to their specific application requirements .
Dead time affects throughput significantly as it represents the processing time during which incoming pulses cannot be counted. The overlap of pulses (pile-up) further reduces throughput. Efficient processing methods, including pileup reject logic, help minimize these losses. However, there is always some probability of occurrence of sum peaks due to pulse overlap. Controlling dead time, ideally around 50% or below, and optimizing processing with features like secondary pileup reject algorithms enhance throughput .
Amptek’s advanced firmware includes improved baseline restoration algorithms and a secondary pileup rejection algorithm. The latest baseline restoration (mode 3) ensures more accurate signal processing at high rates, while the secondary pileup reject algorithm reduces the intensity of sum peaks by about half compared to older algorithms. These enhancements help maintain spectral fidelity at high count rates, reducing distortion and enhancing accuracy .
The pulse pair resolving time determines the minimum interval required between two pulses to avoid pile-up distortions. If two pulses occur within this time, a single pulse with a height equal to the sum of the original pulses is produced, contributing to sum peaks. These peaks affect the quality of the spectral data and can introduce inaccuracies if not managed properly through features like pileup reject logic .
Amptek's default configuration is optimized for energy resolution, using a relatively long peaking time which is unsuitable for high count applications. For high count rates, it is crucial to shorten the peaking time and adjust parameters such as Tflat, Tfast, and gain. These changes help manage dead time and resolve sum peaks. Other critical adjustments include reducing the gain (to avoid pulse distortion) and setting the reset lockout interval appropriately to prevent distortions from preamp resets .
Intrinsic challenges include dead time, throughput limitations, and sum peaks. At high rates, the dead time increases, leading to a lower throughput. Sum peaks arise from pulse pile-up when pulses overlap during processing. Amptek detectors utilize pileup reject (PUR) logic to mitigate these effects, though it does not eliminate them entirely. The optimal configuration—adjusting parameters like Tpeak, Tflat, and gain—can significantly reduce anomalies like centroid shifts and distortion in spectra .
Improper configuration can lead to larger dead times, centroid shifts, and changes in FWHM. Photopeak distortion might appear as tails or excess low amplitude counts. These issues manifest as spectral anomalies, such as inaccurate peak identification or quantification. Ensuring the correct setup of parameters like peaking time, gain, and thresholds is critical to reducing these risks. Anomalies can diminish data quality and reliability, necessitating detailed configuration review .
A reset lockout interval prevents data acquisition distortions caused by rapid preamp resets. A recommended interval of 15 μs helps avoid distortions such as photopeak tails. If too short, reset-associated noise degrades spectral data; if too long, it increases the time discrepancy between acquisition and real time, affecting measurement efficiency. Thus, optimal settings are pivotal to balance data integrity and throughput .
Adjusting gain is crucial at high count rates to prevent spectral distortion due to tail pile-up at the ADC input. A lower gain helps ensure the full spectrum is visible across the screen, minimizing bias and anomalies. Without proper gain management, high count rate operation can result in significant distortions, especially in pulse height, leading to erroneous spectral interpretations .