The secret of PRA Tolerance Correction revealed

PRA, BeqMoni, Theremino, Fitzpeaks and beyond!
Post Reply
User avatar
Steven Sesselmann
Posts: 626
Joined: 27 Apr 2015, 11:40
Location: Sydney
Contact:

The secret of PRA Tolerance Correction revealed

Post by Steven Sesselmann » 21 Sep 2015, 22:21

I guess many of you have wondered what the mystical Tolerance Correction in PRA does, so I will try to explain it here.

Sound card sampling has one weakness, it occasionally under-samples a pulse, this happens because a pulse may have a peak which is so short that it falls between two samples, so when the software integrates across the 16 samples the pulse height comes in slightly low. Understandably for the same reason a pulse can never be over-sampled, so if we had a scenario where the sound card was receiving pulses of exactly the same pulse height, we would expect to see a non gaussian peak, with the bins slowly leading up to the peak, but no counts in the bins to the right of the peak.

Over sampling may happen when two random pulses overlap, but this is a different problem often referred to as PPU (Pulse Pile Up)

It is not easy to see this with NaI(Tl) spectrometry because the resolution of the detector is not good enough, but we can see it if we use an electronic pulse generator.

I actually bought a pulse generator kit on eBay for $3.50 and built myself a simple battery operated circuit with an adjustable pulse.

http://www.ebay.com/itm/231475560234?_t ... EBIDX%3AIT
(no association with the seller)
pulsegenerator.jpg
Pulser
pulsegenerator.jpg (40.05 KiB) Viewed 2453 times
I intend to use it to calibrate the preamp on the GS-PRO models, so I don't have to spend hours every day handling radioactive sources (ALARA).

This pulser generates uniform pulses and with the GS and PRA will produce a peak with a resolution of around 1.5% (without any filtering), the pulses are far enough apart so there is no PPU.

Tightening up the shape tolerance filter will improve the resolution by removing some counts in the bins to the left of the peak, but for some experiments it is not satisfactory to discard counts, so this is where the tolerance correction can come in handy.

Marek (PRA author) realised that there was a linear relationship between the amount of pulse distortion and the under-sampling, so he coded a field into the settings called Tolerance correction. This function essentially looks at the level of distortion and applies a correction to the sampled pulse height.

The best way to use this method is to leave the shape tolerance very loose (I set it to 10) which means virtually no pulses are lost, and then experiment with the Tolerance correction until you find the optimal resolution.

Below is an animated gif where I keep increasing the tolerance correction by 10 at the time.

You might be able to see that the optimum value is somewhere between 40 and 50, because once I go above 50 the peak moves to the right and the resolution becomes worse again.
tollcorr.gif
Tolerance Correction
tollcorr.gif (454.48 KiB) Viewed 2453 times
Anyone who wants to investigate this further should look it up in the manual that comes with PRA 10

Have fun..
Steven Sesselmann | Sydney | Australia | gammaspectacular.com | groundpotential.org | rephopper.com | beejewel.com.au |

User avatar
Geoff
Posts: 96
Joined: 01 May 2015, 12:21
Location: Marathon County, Wisconsin
Contact:

Re: The secret of PRA Tolerance Correction revealed

Post by Geoff » 22 Sep 2015, 07:41

Very interesting! I had been fumbling my way through setups for the most part, a lot of trial and error. Seeing the tolerance correction illustrated like this really helps me understand what exactly is going on as new values are used.
Geoff Van Horn

Anchorage and Deadhorse Alaska

User avatar
Steven Sesselmann
Posts: 626
Joined: 27 Apr 2015, 11:40
Location: Sydney
Contact:

Re: The secret of PRA Tolerance Correction revealed

Post by Steven Sesselmann » 22 Sep 2015, 20:51

Had another play with the pulse generator and PRA today, my setup is as follows;

The pulse generator outputs the signal through a BNC, it has an internal 3kV coupling capacitor, so it can be connected directly to the HV on the GS-1100A or the GS-PRO, the pulse goes through the GS preamp and is output through the audio jack in the normal way.

With the PRA shape threshold set to 50, I was able to record a spectrum with a peak at 30 arb.u having 2.5% resolution. After experimenting with the tolerance correction I arrived at the number 67 which gave me the optimum resolution, and the PHR came down down from 2.5% to 1.5% without loosing any counts.

I guess this shows us how the weakest link is not the sound card, the software or the GS, but as we all knew it is the NaI(Tl) detector and PMT.

With this kind of resolution I don't see any reason why sound card spectrometry could not be used with an HPGE detector, especially since these systems are usually of a very low count rate.

CZT is another high resolution detector which could work well, in fact I have tested this myself, but only with a rather old and maybe not so good CZT detector.

Always pushing the boundary ;)
Pulseresolution.png
PHR
Pulseresolution.png (73.51 KiB) Viewed 2426 times
Steven Sesselmann | Sydney | Australia | gammaspectacular.com | groundpotential.org | rephopper.com | beejewel.com.au |

Post Reply

Who is online

Users browsing this forum: No registered users and 2 guests