Signal digitizers convert detector pulses into sampled waveforms that software can store or analyze in real time. For engineers designing measurement and data acquisition systems, a key decision is whether to use specialized digitizers or adapt lower-cost, general-purpose digital oscilloscopes. A controlled comparison using the same acquisition and analysis software shows how the options differ in energy resolution, pulse shape discrimination, timing, and event rate.
The results come from tests with radiation detectors, so they do not establish performance for every sensor or application. They do, however, illustrate practical trade-offs that matter when selecting waveform acquisition hardware: sampling rate and input range are only part of the decision. Trigger behavior, buffering, channel count, and the ability to process events quickly also affect system suitability.
How signal digitizers fit into data acquisition
Traditional detector readouts may use separate modules for tasks such as shaping, timing, and coincidence detection. A digitizer samples the electrical signal with an analog-to-digital converter (ADC), creating a numerical waveform that can be analyzed by software. This approach can consolidate functions, although some detectors still require a preamplifier.
The comparison evaluated two dedicated CAEN digitizers—the DT5725 and DT5751—and two lower-cost instruments: the Digilent Analog Discovery 2 (AD2) and Red Pitaya STEMLab 125-14. To make the results comparable, all four were connected to a distributed data acquisition system. The hardware interface changed for each instrument, but the waveform-analysis algorithms remained the same.
That consistency matters: comparing devices with different software methods could confuse hardware differences with analysis effects. The shared processing calculated pulse integrals and timestamps, then used those values to generate energy spectra, pulse shape discrimination (PSD) results, and time-of-flight distributions.
Acquisition rate: waveform handling matters
Sampling rate describes how many samples an instrument records per second while capturing a waveform. Acquisition rate, by contrast, is the number of events processed per second. A high sampling rate does not guarantee a high event rate: data transfer, buffering, and processing can become bottlenecks.
In the reported setup, the CAEN digitizers reached acquisition rates up to 600,000 events per second per channel with approximately 1% dead time. A controlled pulser test reached up to one million events per second per channel. When full waveforms were transferred over an optical connection, the reported rate was lower—up to 30,000 events per second per channel.
The STEMLab 125-14 reached 4,000 events per second per channel, while the AD2 reached 100 events per second in this configuration. The AD2 did not buffer multiple events, so the acquisition system had to request each event individually over USB. Both low-cost instruments acquired waveforms rather than producing the processed PSD events available from dedicated firmware.
These figures describe the tested system, not a universal maximum for every configuration. They nevertheless show why event throughput should be tested end to end rather than inferred from ADC sampling specifications alone.
Spectroscopy and pulse shape discrimination
For spectroscopy, the system used sodium iodide and lanthanum bromide scintillation detectors. Energy resolution was assessed by measuring the width of known gamma-ray peaks relative to their peak energy. At the 662 keV line, the DT5725 recorded resolutions of 7.578% with NaI(Tl) and 3.288% with LaBr3(Ce). The STEMLab results were 7.726% and 3.376%, respectively. The AD2 recorded 9.81% and 3.916%.
The DT5725 performed best overall in these spectroscopy measurements. The STEMLab was close to the dedicated units, while the AD2 had the weakest results among the four. The tested devices differ in sampling rate, ADC resolution, and input dynamic range, so the results should be interpreted as system-level comparisons rather than as a ranking based on any single specification.
For PSD, the test used an EJ301 liquid scintillator and a californium-252 source to distinguish neutron and gamma events. The analysis compared short- and long-gate pulse integrals: differences in pulse decay shape produce different PSD values. Performance was summarized using a figure of merit (FOM), with higher values indicating better separation between the measured distributions in the selected energy range.
The DT5751 achieved the highest FOM at 2.50, followed by the DT5725 at 2.17, the STEMLab at 1.91, and the AD2 at 1.61. However, a single FOM does not describe every feature of the distributions. For example, the DT5751 distribution curved at higher energies, meaning a simple fixed PSD threshold would not provide uniform discrimination across the full range.
Timing performance and system constraints
Timing was evaluated using the width of the time-difference distribution between two detector channels. The DT5751 produced the narrowest distribution, with a full width at half maximum (FWHM) of 0.748 ns. The DT5725 measured 0.898 ns, the AD2 1.19 ns, and the STEMLab 1.34 ns.
These measurements concern relative timing within a short coincidence window, not unrestricted absolute timing. The CAEN digitizers provide an absolute timestamp from the start of acquisition. The two oscilloscopes did not provide equivalent absolute timestamps in this setup, so their timing use was limited by the channel buffers; the acquisition software simulated absolute timestamps using the computer clock.
Channel behavior is another design consideration. The CAEN digitizers can acquire channels independently with the specified firmware configuration. The AD2 and STEMLab channels share a trigger and cannot operate independently. The STEMLab offers a different practical advantage: as an embedded computer, it can control acquisition and store data without a separate acquisition computer.
Choosing hardware for the application
Low-cost oscilloscopes can be viable for prototypes or applications with modest event rates, provided their trigger, channel, timing, and data-transfer limitations fit the task. In this comparison, the STEMLab delivered spectroscopy and PSD results closer to the dedicated digitizers than the AD2, but its event rate was much lower than the reported CAEN performance. The AD2 offered the lowest acquisition rate in the tested configuration.
Dedicated digitizers are the stronger fit when high event throughput, independent channels, or wider-window absolute timing is required. Their embedded processing and triggering options may also be relevant, although advanced firmware capabilities were not evaluated in this comparison. Channel count and total system cost should be considered alongside performance: the tested specialized units offered more channels than either oscilloscope.
The practical takeaway is to select waveform acquisition hardware against the complete measurement requirement—not price or sampling rate alone. Validate event throughput with the intended data path, check whether channel triggers and timestamps support the measurement, and assess signal quality using the analysis method your application will actually use.