gamma spectroscopy #3: Dynamic Time-over-Threshold (DTOT)

Introduction

In a previous post, I build a 4 channel data acquisition system for scintillation gamma spectroscopy using the Time-over-Threshold (TOT) technique. The results were promising, but one issue was that the resolution was getting bad if the threshold was too low. The reason for this is shown in the following picture.

There is always some noise on the signal (pink). The region where the trailing edge intersects the noise envelop is larger for thresholds that are closer to zero Volts. And this region directly creates the uncertainty in the Time-over-Threshold measurement. Note that this is not a problem for the leading edge because of its fast rising slope.

One could just put a higher threshold and the uncertainty will be less, but the disadvantage is that small signals below the threshold are not measured at all. 

Dynamic Time-over-Threshold

One possible solution is to start with a small threshold (just above the noise) and as soon as the leading edge is detected the threshold is increased dynamically.

The following picture shows a spice simulation including a small modification of the circuit. The dynamic threshold is created from the comparator output and is fed back into the previously unused second input of the differential amplifier.

Both lines in the graph show the inputs of the comparator. One can see that the threshold is dynamically increased and the trailing edge is intercepted at a higher slope. That effectively reduced the uncertainty of the trailing edge time measurement. I took the idea from this publication. My implementation is however different from what they mention in that publication. I've done the simplest possible extension to my existing circuit.

Measurements

I modified one of the channels of my PCB by adding the three components in the feedback path from the comparator output to the j-FET input.

After doing this I measured the comparator inputs with the oscilloscope and got the following picture (quite similar to the simulation).

Spectroscopy

Finally I made a setup with one single LYSO crystal and one 3x3 inch NaI detector. The detector signal was going into the modified channel of my acquisition box. I got an additional signal from a Silicon photomultiplier that was attached to the LYSO crystal. The LYSO setup can be seen in the following picture: the bare parts and the final assembly wrapped in Teflon tape, aluminum foil and black tape.

This signal was recorded with an unmodified channel (no dyamic ToT) of the acquisition box. The following is a pulse height histogram of the NaI detector for all signals that have a coincident signal in the Silicon photomultiplier (5 microseconds time window). 

The most important point here is the low energy limit of the spectrum as a result of the dynamic threshold behavior. The lowest visible peak is at about 28 keV and the spectrum extends down to half of this value (~20 keV). This was not possible without the dynamic threshold adjustment. The spectrum was calibrated using the known positions of the peaks. Labels on peaks show the decaying level and the FWHM resolution.

HPGe detector readout

At work, I got the chance to connect my device to the pre-amplifier output of a high-purity Germanium (HPGe) detector and a 60Co source. The count rate was around 400 Hz. I didn't have the time to do a lot of tuning, but the result looks promising

It is not so easy to estimate the actual resolution because of the nonlinear time-over-threshold values and the unknown threshold height. But I can see where the Compton edges are. Based on that I can guess that the threshold was quite high (maybe 400 keV). The resolution of the peaks in the histogram is 1% (FWHM/position). The conclusion is that the energy resolution of this setup was better than 1%.
It would be interesting to find how close I can get to the ~0.2% intrinsic HPGe resolution with a bit more tuning of the setup.
Another conclusion is that this method is definitely sufficient to read-out scintillation detectors.