Needing more than a spark test?

[graham-xrf]

Acronyms, buzzords, jargon!
Re: Theremino "Minimize the FWHM"
Duh! Should I have guessed that means "Full Width Half Maximum"?

[Edit: And while we are on this, "MCA" as seen in PyMca and ThereminoMCA apparently means "Multi-Channel Analyzer"]

 

[graham-xrf]

The Theremino way
OK, I have read it. I don't agree with a whole lot of it, but I take nothing away from the guys who that the guys who made real kit originally, and did real stuff. I can skip over an actual critique unless someone really wants to know all the points where I thought otherwise. This is boring technical stuff, and I want to get to the point where HM readers can see pictures of real stuff in action.

I must mention his points (8) and (9).
Apparently advocating 16-bits rather than 14-bits, and mentioning at least 20 samples per pulse. His example pulse is 5uS long. the 5uS/20 = 2.5E-7 or 250nS intervals.
That would be 4MS/sec, which looks at odds against his final solution of using a sound-card A/D at 192KS/sec.

We must carefully make clear the difference between a pulse from a scintillator and the peak on a histogram made up of counts per second in "buckets" of similar amplitude. horizontally sorted by amplitude. You can get a high count of a low amplitude pile of pulses, which would make a higher peak of counts, but on the low energy side of the spectrum

I entered the Theremino signal conditioning circuit into LTSpice simulation.
It uses transistors from 40 years ago, and honestly, this is not the way to do instrumentation when there is so much high bandwidth low cost integrated circuits available.

How to do peak detection --> https://www.analog.com/en/technical-articles/ltc6244-high-speed-peak-detector.html
Many A/D converters have built-in track/hold.

The simulation speaks for itself.
The green trace is a fast getup to generate a pulse similar to the one in the document.
I could have used a carefully defined piecewise waveform, but this was faster.
Notice that the Y-axis zero is suppressed in some of the traces.
The green trace starting has about 115mV amplitude. What survived the C8/C9 "high-pass filter" to the testpoint fits between 591mV and 593 mV. Massive amounts of signal thrown away! No buffering.




What you have at the end is a low amplitude smeared lomp, and I would not agree that it's tiny amplitude is more easily separated from other amplitudes with this circuit. Of course, it could be that the whole pulse was much larger to start with, and all the mumbers are bigger. Even so, the degradation of the pulse information is obvious!

Using BC237 transistors in a circuit that I would have binned 40 years ago! Sorry - but I go my own way with this.
I ws looking for a way to make it extremely low cost, and if capturing the peak and counting them is all it takes, this can be done better.

BUT, I an not sure about capturing scintillation peaks, and interpreting "pileups", etc.

Just starting out on this. Expect me to pretty much get right down to the bones of it. If it REALLY works, then we shall have some. Otherwise, it get ignored.

I include the save LTSpice files if anyone wants to play with it.

 

[rwm]

By "pile ups" I assume you mean simultaneous events? Can we just ignore these at first and only count the isolated events within a relatively short time interval? You might need to adjust the Am source activity if you are getting too much flux and not having a lot of discrete peaks.
Robert

 

[graham-xrf]

By "pile ups" I assume you mean simultaneous events? Can we just ignore these at first and only count the isolated events within a relatively short time interval? You might need to adjust the Am source activity if you are getting too much flux and not having a lot of discrete peaks.
Robert

Yes indeed, but you need a way of "recognizing" a pulse as being isolated, and of a set of similar height pulses, to have it be counted into the right bucket. I don't think we need limit activity. The article talks of 50,000 pulses/sec, and I have read of some sample runs taking several minutes. If the scintillations arrive, glow, and die at a pace determined by the scintillator response, that would be the constraint. I have read that Csi(Tl) scintillations are about 1uS duration.

Either way, I am going to first get the PMT and scintillator up, and measure with a storage scope. I want to get used to what happens, the magnitude and speed of the signal. Then I will figure out what it takes to measure it, sort the measures, and count into the right buckets. I am thinking a 5Msamples/sec 16-bit A/D converter would do it, but I am still intrigued by using a video bandwidth op-amp peak detector , with a trick trigger comparater to "detect" when the slope of the pulse has turned around, and so capture the peak. In such a scheme, it may be possible to use the audio input channels of the little computer to get the 16-bit captures, and there being two channels, one can double the capture to 2 x 192kHz.

It should be possible to discard signals from pileups. It would be great if the gamma from only one smoke detector source was enough.

 

[homebrewed]

I'm not disputing your simulation results, but something to consider is the effective CPS for your simulation is 2E5, while the Theremino expectations are in the 1000CPS range. The pulses should be spaced at least by 10 time constants to avoid that slow downward drift of V(output). Theremino also recommends using USB sound cards that can acquire at 192KSPS (still far short of what would be needed to capture a full-bandwidth pulse out of a PMT or SiPM).

The Theremino approach appears to employ a lot of data processing -- both analog and digital, including throwing away a lot of pulses that don't fit their criteria. To me this suggests that their main intent is to manage scintillator+PMT noise -- the heavy filtering plus (least squares??) curve fitting would do that. A more modern approach using low-noise electronics and a faster ADC might produce even better results.

I believe that SiPM's are inherently quieter so that's something to think about. And cooling one with a TE cooler wouldn't be too difficult. If taking that approach I would stay away from scintillators that are even slightly hygroscopic. Unfortunately, it seems that the best scintillator materials are extremely hygroscopic.

 

[graham-xrf]

I'm not disputing your simulation results, but something to consider is the effective CPS for your simulation is 2E5, while the Theremino expectations are in the 1000CPS range. The pulses should be spaced at least by 10 time constants to avoid that slow downward drift of V(output).

Ahh yes - please don't misunderstand. I get it that the pulses might come very slow, every now and then, but once the photon hits the scintillator, the glow response duration is microseconds. I set the simulation time to count two pulses, to nearly mimic a coincidence pile-up. There is no implication that the simulation pulses are continuously coming at 200kHz. In practice, there might be smaller, and larger scintillations, maybe separated, or sometimes on top of each other, randomly timed.

A PMT or Si(PM) does not know when there were two or more glows simultaneous.What it sees is additional energy, so making a "higher" pulse, that may have some other duration characteristic. Recognizing these, and coming up with ways of excluding them, or dealing with them, is part of the game.

Consider an analogue peak detector with a trigger and a run-out timer to flag any pulse durations longer than (say) 1.6uS. That would deliver only genuine single scintillations. Others so exactly simultaneous as to "look" like a single can be rejected for being over an energy threshold. Exactly simultaneous smaller scintillations could masquerade as a larger single, but not frequently enough to much adjust the statistics bucket.

Of course, much more "recognition" could be done with software, but that requires a very high resolution sampling.

I believe that SiPM's are inherently quieter so that's something to think about. And cooling one with a TE cooler wouldn't be too difficult. If taking that approach I would stay away from scintillators that are even slightly hygroscopic. Unfortunately, it seems that the best scintillator materials are extremely hygroscopic.

You may be right, but in my experience, reverse biased avalanche diodes are so spectacularly good at generation of wideband white noise that they are used loud noise generators to make standard noise calibration signals. There is a constant noise coming out of them.

Nor am I convinced about "noisiness" in the PMT tube circuits. They may be, but not in the way as in the article. A capacitor across a tube divider is to supply electrons, not a crude "smoother" of waveform noise. We can limit noise with inductance along with the capacitance. Noise from a resistor (from molecules banging about therein) is √4*Kb*T*R. The higher the resistance, the more the noise to be short-circuited by the capacitor, and it is not good at it!

I am more used to trying for the extremes of low noise amplifiers in another context, where the gain is more like 50dB or 60dB, the signal input around -110dBm to -120dBm, and the galactic and atmospheric is 10K to 100K noise temperature equivalent. If one is truly after low noise signal processing, the wrong way is to allow a 1Meg resistor to contribute, and then offer it at a BC237.

 

[graham-xrf]

Yes - still looking for a cheap & cheerful circuit dodge. If anything can be hung together that uses one smoke detector, one cheaper diode from Mouser or Digi-Key, and some bit of recycled transparent plastic, and some small electronics almost anyone can hang together, then I will try for it.

PMT is to be tried, including with non-hygroscopic scintillators, and also Si(PM) avalanche diode. My low cost Ukrainian PMT is still "in transit".

 

[Tim9]

I don’t overthink this stuff. I used to. If it looks like it’s got a lot of chrome/ nickel in it.... I touch with file first and then just chuck it up. I’ve had some stuff that looks like high stainless actually turn nicely. Other stuff looks similar but it turns with a stringy nightmare bird nest.
PS....you guy are impressive. I’m just lazy....my bad

 

[graham-xrf]

I don’t overthink this stuff. I used to. If it looks like it’s got a lot of chrome/ nickel in it.... I touch with file first and then just chuck it up. I’ve had some stuff that looks like high stainless actually turn nicely. Other stuff looks similar but it turns with a stringy nightmare bird nest.
PS....you guy are impressive. I’m just lazy....my bad

Let's go there a bit. Can one tell what is in it from the way it turns, looks, smells, rings, chips, weighs?
I think there are too many variables, including the wear and condition of the lathe, but I have seen there are some "difficult" materials that turn a good finish only when quite a deeper cut is used. I see on YT a video of a guy even turning tungsten carbide, and having a hell job with the fine chips and dust, keeping them out of the machine - and his lungs!

There are probably lots of HM guys with a fine-tuned sense of sight and feel and smell who can tell in the first seconds what the steel likely is. They may not be able to tell the difference between 4140 and EN24T, but they can tell which stainless is messing up the tool. Handling those that work-harden the pass before the final cut must be maddening!

I had thought a carbon arc against the material would yield light spectrum that contained all the materials signature in the spectrum. Maybe it does, but optical spectrum analyzers are many $$$. I am intrigued by the notion of capturing the flare on a smartphone through one of those toy hologram filters that make rainbow colors, and then analyzing the stripe of spatially dispersed pixels. (OK - maybe been done before, but I don't know the trick yet)

For now, I am going to have a little play with XRF. After all pushes some good buttons. Consider..
1) I am messing with stuff that is RADIOACTIVE!
2) I get to play with CHEMISTRY junk like they wouldn't let me do at school!
3) Dang - but that thing uses HIGH VOLTAGE - like KILOVOLTS even!
4) It's got glowing crystals and X-RAYS!

It does everything except come with explosions and rocketry. What's not to like?

 

[homebrewed]

It does everything except come with explosions and rocketry. What's not to like?

Surely there's a way to fit them into this .