Gamma Spectrometry Forum

Hello all,

I want to share with you my resent results while attempting to build a better (than sound card) data acquisition system that is still cheap and suitable for hobbyists.

My focus (also with sound card data acquisition) is to have more than a single channel to make coincidence measurements. While it works with the sound card, the time resolution is rather limited: with a 192 kHz sampling rate one cannot easily get below a few micro seconds. Time resolution is crucial for coincidence measurements. Few nanoseconds resolution is desirable. I also would like to have more than two channels... I would be happy with four. There are four channel sound cards, but if I had to buy special hardware for my measurements, I could rather try to find something that fits my needs better.

I started playing around with the time-over-threshold method, where the detector signal is compared with a constant threshold voltage level. It works by measuring the time t0 when the rising edge of the detector signal crosses the threshold voltage, and time t1 when the falling edge crosses the signal again. The time difference t1-t0 has information about the pulse length, but also about the pulse height. The relation between pulse length and amplitude is approximately logarithmic if the rising edge is fast and the falling edge is an exponential decay. One problem I encountered was that I couldn't make the voltage threshold as low as I wanted without sacrificing energy resolution. I found some publications about "dynamic time-over-threshold" methods: after the leading edge of the signal crosses the threshold, the threshold is increased in order to intercept the falling edge at higher voltages, where the impact of noise is lower. EDIT: This is illustrated in the picture below (smaller error due to the intersection of the signal at a higher amplitude) For time measurements, I used an FPGA development board https://embeddedmicro.com/products/mojo-v3.html (for 75$) with some custom VHDL code. It records the time of the threshold crossing with 5ns resolution and sends the time value over USB to a PC. The analog voltage comparison is done on an add-on board for the FPGA board. More information about the device can be found here https://emsyfs.blogspot.com/2017/09/fpg ... ition.html and here https://emsyfs.blogspot.com/2017/10/gam ... -over.html.

I am quite happy with the result so far. The following spectrum was taken with a 3x3 inch NaI(Tl) detector from BICRON at -700 Volts and a LYSO crystal as a source. One more special thing about that measurement: there was no shielding or background subtraction applied. The clean spectrum is rather resulting from the fact that I used the LYSO (which is also a scintillator) in conjunction with a Silicon photomultiplier to give me another signal to "gate" my NaI detector. In other words: the NaI signals were only filled into the spectrum when within a 50ns time window there was also a singal from the LYSO scintillator. Here is a picture of this "active source" A few final comments to this method:
1) Number of channels can be easily increased. I synthesized the VHDL code successfully with 16 channels, but my front-end card only supports 4 channels.
2) Due to the logarithmic behavior, there is practically no saturation in the data acquisition. It can digitize signals with very high amplitudes like cosmic muons. (maybe the front-end saturates at some point).
3) The calibration is a bit more difficult, because a linear function will not work. But if the signal shape is known, this can be handled.
4) The time resolution of 5ns can most likely be improved (down to 1ns should be possible with a bit of effort).

Best regards,
Michael

Interesting !

Michael,

Very interesting and novel work. I can see the problem with time over threshold as you describe it, ho low thresholds the resolution will be affected by noise and baseline drift. I look forward to follow your progress. Would be interesting to see a comparison with a good NaI(Tl) detector and some common isotopes.

Steven

Hello Steven

Thank you! You are right in that I should do some comparison to conventional spectra. Yesterday I did a longer measurement that has no coincidences or gated spectra. Just a plain ~20 hours measurement of a Th lantern mantle under my 3x3 inch BICRON NaI(Tl) detector and a single channel of my ToT-DAQ box. The detector has a lead shield at the sides (top and bottom are open). I used -800 Volts. My HV comes from a hamamatsu C4900 module. The count rate during the measurement was approx. 800 per second.

I made a picture of the setup (I know, the HV part of my setup is still too open... but I know what I'm doing ;) ) The time over threshold histogram looks like this: and after a quick and dirty energy calibration, which goes like E=threshold*exp(ToT/tau) (EDIT: tau is the decay time of the detector signal) I get the following energy spectrum: I've compared the picture to the one here (http://www.gammaspectacular.com/gamma_s ... 2-spectrum) and it looks similar. I would say my resolution is even a bit better. Due to the rough energy calibration my peaks are all a bit too low. And I wouldn't trust the calibration at all in the x-ray region of the spectrum, because the deviation from the exponential relation is biggest here. I've written the event data to disk (543MB of rising and falling edge time stamps), so I can do improvements on the calibration later, replay the data and get better calibrated spectra.

Michael

Michael,

I agree, it's a very nice smooth spectrum, it must also be a good NaI(Ti) detector.

How long was your data acquisition?

Are you doing any pulse filtering other than time over threshold?

What pulse length are you working with?

How fast is the switching and the clock rate?

I guess if the whole thing is fast enough, then pulse pile up (PPU) becomes less of an issue.

Steven

Steven,

I got the detector from ebay. I guess I was lucky with this one :)

In this measurement I took data for 19 hours and 49 minutes, at the moment I'm measuring the background spectrum inside the shielding for a similar amount of time to make a background subtraction.

I use a similar front end parts as in my sound card setup: A resistor "Rd" discharges the anode of the PMT. This results in a sharp rising edge of the pulse and an exponential decay. In the sound card setup I used a MOSFET based voltage follower. In the time-over-threshold setup I use a JFET based amplifer (x5).

In both cases, the transistors have a huge input impedance and do not influence the pulse shape. The decay time (tau) of the signal is determined only by the capacitance of the PMT and the resistor "Rd". In the time-over-threshold setup I ended up using Rd=33 kohms. The resulting signal decay time is in the oder of 2.5 us. In the time-over-threshold histogram (see above) you can see how long the pulses are above the theshold: the majority of signals is shorter than 12 us.

The time to digital conversion is a counter that ticks with 200MHz. I send the counter value to the PC whenever a rising or falling edge is detected. This results in 5ns resolution. I want to improve that to 1ns in the future, then I can make the pulse decay time even shorter (perhaps 500 ns). Ideally I want to use a 50 ohm resistor.
Count rates can go rather high inside the hardware, but there is a bottleneck in the USB connection. It is too slow in the current version. Because of the limited data transfer rate I cannot get more than ~1500 counts per second over to the PC at the moment. With an improved data transfer rate to the PC, the count rate limit might be several 10000 counts per second per channel (nothing I ever want to have in my home lab).

Michael

Michael,

I did another Thorium spectrum last night with my 3" detector for comparison.

This is 5,000,000 pulses with a GS-1100-PRO and sound card sampling at 96 kHz.

Typical volume adjustment with K40 in the centre of spectrum.

Steven

Steven,

that spectrum looks very nice, even better resolution than what I got. Particularly for the low energies, I'm not really sure of how linear I can get my calibration. Your spectrum is very useful to investigate this. Thank you for posting this!

I have some questions about your setup:

Did you use shielding all around your detector?
Did you have some distance between the sample and the detector?
Did you use a lantern mantle or do you have a different Thorium source?
From your data (over night (8hours?) and 5000000 pulses) I get an average rate of approx 170 counts per second, is that correct?

Michael

Michael,

I was using the 3" detector which was refurbished by Luuk, before refurbishing it had a resolution of around 14% and after Luuk's secret treatment it came back 6.5%.

Did you use shielding all around your detector?

Only very light lead shielding, a cylinder with approximately 10 mm lead around the detector

Did you have some distance between the sample and the detector?

Yes approximately 1" gap between detector and source.

Did you use a lantern mantle or do you have a different Thorium source?

Yes it was an old lantern mantle in a plastic bag.

From your data (over night (8hours?) and 5000000 pulses) I get an average rate of approx 170 counts per second, is that correct?

I limited the recording to 5 million pulses in PRA, and went to bed, not sure what time it finished.

Steven