Problems calibrating between 1500-2600 keV

[ColoRad-o]

Steven--thanks for all of the information, based on experience and expertise! This is above the call of duty!

As a theoretical physicist, I tended to accept things as I measured them, rather than understanding how I could tweak things to make problems go away. (And my last exposure to [mostly] analog electronics was more than 50 years ago.) This is why I accepted 700 V PMT bias--on the plateau of count rates and the nominal operating voltage. I never twiddled with the SHP or VOL knobs, in my eagerness to measure things. Very non-linear calibration curve at high bias? I thought I'd just deal with it, but it does cause considerable pain.

Almost all of the data I show were taken 5 years ago, I was non-comprehending but methodical. Generally speaking all of my count rates are low--under 2000 cps, often a couple hundred cps.

For sound-card spectrometry (µs-scale NaI(Tl) pulses) the preamp rarely sets linearity if we avoid slew limiting and clipping. A simple, low-noise, unity-gain-stable op-amp with adequate GBW and headroom, followed by gentle anti-alias filtering and proper impedance matching to the audio CODEC, is usually best, over-shaping in hardware just adds dispersion that can't be un-done. My view, “less is more” clean gain, modest filtering, and good headroom beats elaborate analog shaping for PRA/Impulse.

"Less is more" is all I could cope with:) A fraction of order 15% of my pulses are clipped somewhat, and the manual reassured me that this was OK--probably cosmic rays. I assumed that PRA's shape method would ignore such pulses, but I don't know for sure what their impact is.

I really don't want to take the severe hit in count rate if I can avoid it--the samples I'm counting are feeble enough and life is short, and I can only fit so much into the sample region above the detector tube inside the shielding (a GS-STANDUP). But there's nothing like a big honking scintillator for sensitivity with respect to the fancy new hand-held gamma ray spectrometers, so I'll never stop using the GS system. (It IS annoying that their energy calibration is ridiculously good.)

Thanks again for your comments and suggestions!

[Sesselmann]

David,

I just ran some tests with GS-PRO and a GS-1515-NaI which is my least linear detector and the results are not crazy.

It is an old spreadsheet I made which normalises arb.u to KeV and plots.

These were relatively short runs so the peak locations were not 100% accurate, but good enough not to end up like a banana 🍌

If I took it to 900V we probably see more of a banana.

Steven

Linearity check

[jneilson]

Just addressing the initial question rather than any of the follow up discussion for now, but I see a few separate questions in there: both sources for that range, and piecewise calibration - and for both questions, it depends whether you're just doing energy calibration or whether you're trying to do efficiency calibration.

Piecewise energy calibration:
I have never heard of anyone doing that. The common approach is to use a polynomial fit rather than a straight line, but that's usually just a slight bend to the straight line, not something that would go piecewise at all. I see you're already using a 4th order curve - I'd perhaps play with that to see what happens - try going down to quadratic and up to maybe 6th order on the initial fit, before you try fitting to the errors.

Piecewise efficiency calibration:
There's two piecewise-sounding approaches I have heard of - purely interpolative (ie piecewise between each pair of calibration points); or a common one for HPGe calibrations would be a dual quadratic (piecewise with one curve at low energy and another curve at high energies - but generally the transition is in the couple-of-hundred keV range, not a 1000+ type thing)

Higher energy energy calibration peaks:
If you're only doing energy calibration, you can still use higher energy peaks from the U/Ra and Th chains, as that's what you're already using and NORM materials are relatively easy to get hold of - but you won't necessarily have a good calibration of how much you have if you want to look from an efficiency point of view. Depending on your amount of activity and how long you might be able to count for - might need to be a very long time, you might be able to see some of:
Bi-214 (U/Ra): 1509 (2%), 1583 (0.7%), 1729 (2%), 1764 (15%), 1847 (2%), 2118 (1%), 2204 (5%), 2447 (1.5%)
Ac-228 (Th): 1588 (3%), 1630 (1.5%),
Bi-212 (Th): 1620 (1.5%)

I've often used some of those peaks just from my natural background to help refine an energy calibration, especially for our portable HPGes.

There's also a couple of other spectrum features you can sometimes use:
Tl-208 2614 escape peaks: 2103, 1592
Co-60 sum peak: 2505

Higher energy efficiency calibration sources:
If you're looking for efficiency calibration, then you want something that's been accurately calibrated as a reference source, so it depends on what source manufacturers you have access to. We have E&Z mixed nuclide sources that contain Y-88 for the 1836 line to give us a point in that range. Spectrum Techniques seem to be one of the few suppliers that will ship to individuals rather than large facilities, so you might be limited to what they can sell, and I'm not sure they have anything in that range on their lists.

There's actually really only a small handful of nuclides that have lines in that range at any appreciable intensity and have a half-life long enough to be useful. Bi-207: 1770, Eu-154: 1596, Al-26 (1808) are the few that jump to mind, but you might need to go to alternative suppliers for those.

[ColoRad-o]

Joseph--thanks for your posting!

I have tried most of your suggestions (polynomial order) both recently and in the distant past without success. And the GADRAS-DRF software (more below) does a good job of producing a (typical-looking scintillator efficiency curve given the dimensions, material, and shielding for my detector.

As you can see from my posting curve on November 5, I have quite a large number of reproducible peaks ranging up to 2614 keV and, yes, I have used the Co60 sum peak too. So it's not finding peaks per se that is the problem, at this point. Rather it is the fact that my 2x2" NaI:Tl detector needs to be run at moderate bias voltages (I have used 700 V), for the particular PMT that came with my model, in order to get a stable (and conveniently high) count rate. [By stable: I mean on the 'plateau' of count rates that exists over a moderate bias range.] At that voltage there is quite a bit of curvature of the calibration curve. This is not a problem per se, but I find that the curvature gives rise to enough slop in the calibrated spectrum that, even when re-calibrated by InterSpec, the spectra do not perform well using GADRAS-DRFl. (This U.S. government-distributed software does whole-spectrum fits to identify proportionate contributions from, for example, natural radioisotopes in a sample.)

Annoyingly, data from a small, much less sensitive (and much less expensive) RadiaCode 103G system is much better behaved, almost definitely because it is very well calibrated 'out of the box'. As time permits, I may revisit the NaI issue, but meanwhile it is just much easier to run the RadiaCode for 1 week at a time and then combine spectra. The calibration of this model is very stable over periods of many months to years, so lends itself well to repeated runs. [It's battery powered but I don't really want to leave it plugged in for weeks at a time.]

Thanks again for your followup!