Gamma Spectrometry Forum

I recently made a probe for gamma spectroscopy with a LYSO crystal matrix (5x5 crystals, each 4x4x20mm, for a total of 20x20x20mm cube) and a Hamamatsu R9420 tube (FWIW both the LYSO and the tube were bought from Tom at Irad on ebay). The problem I have is that the LYSO has an intrinsic background due to the presence of 176Lu which does emit a number of low-energy gammas in its decay cascade.
However, I did not expect the background to be that high. I'm getting approx. 140000cpm with no gamma source nearby! (and some lead shielding). Apparently it's normal for this type of crystal. And since the source of the gammas is inside the crystal itself, the counts are very high even though the crystal when counted with an external detector (a 3inch BC412-based one) only gives around 5-6000cpm.

I have a 137Cs source (10uCi) which I'm trying to use to calibrate this new detector, but have some trouble with background subtraction. Here are below some of my most recent (and best so far) attempts.
The parameters I'm using are: 1000V tube bias, 100mV rejection (there are a LOT of pulses below 100mV, X-rays mostly which I'm trying to eliminate from the equation), 2mV/channel, range 2V (=1000 channels). This is not with a Theremino-like adapter but with a special software-controlled voltage source and acquisition is with a Picoscope2000 USB oscilloscope after a signal conditioning amplifier. Both are driven by a custom software interface.

I get (among others) an output as text files containing a listing of channel numbers and number of pulses counted in the channel during the experiment. I did some text processing with bash and awk to obtain test minus background (with/without the 137Cs source) then render graphs in LibreOffice. I noticed that the counts appeared to be a little shifted between channels between experiments so I also did a manual shifting of one of the file counts (the background) by 1 row to better fit the data. I didn't correct negative values but rather left them in place as-is. Shifting the background file values by 2 places down did not improve things, quite the contrary.
It's still not very neat. All counts were over 10 minutes.
I am concerned mainly with the maximum peaks of the lutetium and cesium mostly overlapping... And they shouldn't.

Silviu,

The most obvious problem I can see is the high count rate saturating your sound card (assuming you are using a sound card with Theremino).

Let's do the numbers, your sound card is sampling at say 96,000 kHz. so the time between samples is 1 s/96 ks = 10.4 µs, so in order to get a pulse height Theremino probably reads 16 samples, so 10.4 µs * 16 = 166 µs. Now there is a specific calculation for dead time which I can't recall now, but I think you need a pause in the order of twice this much so now you are up to 500 µs, which means the maximum count rate you can read is 2000 cps.

Your detector is already exceeding this before you start counting, so not surprising the results are poor.

If you want to get this working with a sound card I would suggest the following;

1) Reduce the size of the crystal array
2) Find a sound card with 192 khz sampling
3) Try using PRA and reduce the number of samples to integrate from 16 to 8 (un-tick the boxes in settings pulse shape)

Simply doubling the sound card sate and halving the integration will quadruple the count rate.

Steven

PS: Silviu, can you please change your pseudonym to your real name, or add your real name in the signature, so we can address you without looking up your introduction post.

I don't use a sound card, or a Theremino-like driver. It's in my post above: for signal acquisition I use a USB oscilloscope. It samples at more than 20Msps, so that is not the problem. I can post signal waveforms if you want, they are very clean. It's not a sampling issue.

(edit) here are typical pulse shapes that I get

Here's another run. This time 5 minutes only, and I turned on signal integration in the software. Not much better.
The last picture has a typo: channel width is still 2mV, not 1mV.
First picture is a superposition of the background (green) and cesium (blue) counts, with the red being an average of the cesium counts,as generated by the software.

Silviu,

I read your post again more carefully, this time I see you are using a scope to sample the pulses (sorry)

With these high count rates I imagine there is going to be some pile up, is there a way to filter out the corrupted pulses?

Steven

Not unless I do it manually - not a nice prospect for 140k pulses/min. I have been watching pulses fly by and while I do see occasionally pulses riding on top of each other, I estimate they can't be more than a few percent of the total.

Silviu,

Here is an extract from the PRA manual explaining how the pulse shape filter is done. Possibly you can run a script through the data to generate a tolerance value and then reject counts with unacceptable tolerance.

Maybe it can be of some help.

Steven

That would be nice. I don't have access to the raw pulse data however, the scope's custom software only lets me access processed data such as the images I've attached. And some statistics in text format. I can record the shape of every pulse counted but only as .png images.

(edit) The software does have a filtering option, but it's non-intuitive and poorly documented. It takes a large table of float values as a parameter but I'm not sure what the parameters represent. I'm trying to get in touch with the maker of the software to get better documentation.

After consulting with experts on LYSO and my detector setup, this is the best I can do. Counted background for 10 minutes, then my 10uCi cesium source for 10 minutes, subtracted the first from the second, cut off the first 20 channels which were noise, graphed. This was done with just one small LYSO crystal, 4x4x20mm as scintillator. 2 crystals or more on the PMT give worse results. It's very, very poor resolution. I think I'll give up on trying to make this work, it's using up too much of my time. I have a large NaI(Tl) crystal coming in the mail and I already have the PMT prepared for it, I'll just use that.

You can see the cesium (well, barium) peak, but it looks disgusting. And that's with a strong source located very close to the detector. The CPS for background were around 95, with cesium around 142. Maybe in a professional setup with finely tuned detecting systems LYSO can be a viable option, but in my amateur conditions it's just not working out.

Silviu,

Are you still using your scope to make these spectra?

If so, I think you will see an improvement when you use the proper PRA software with pulse shape filtering.

Steven