Needing more than a spark test?

[rwm]

Looks like you could drive that out of the carrier with the correct shop made punch and die?

 

[homebrewed]

Looks like you could drive that out of the carrier with the correct shop made punch and die?

Yes, you probably could. The steel used to make them appears to be pretty soft. The actual bit of foil containing the Am241 is in a recess so it wouldn't be at risk.

 

[homebrewed]

Regarding my comment about the choices of material suitable for making the focus ring. In retrospect I don't think acetal or Derlin would be good choices. In my experience, glues don't bond well to them. Since my aperture plate is made of lead, it really doesn't matter what material is used (in terms of the XRF performance). It could be just about anything a hobbyist has the ability to accurately shape or print.

I have some 6061 aluminum rod that has a large enough OD so I will just use that. To prevent any unsupported overhang of the sources, the focus ring's OD should be a minimum of 1.035 inches. With that, its height will just be (.707)*.250 = .177 inch/4.5mm. 1.035 inches is the minimum it should be -- it can be as large as will fit in the enclosure.

Having the sources a bit below the top surface of the focus ring would provide a bit of protection from samples inadvertently being shoved up against the ring. A further refinement might be to make the focus ring deep enough so the focal point of the sources is coincident with the mouth of the ring. So no guessing on how to place the sample for the highest count rate. That is, if the count rate turns out to be an issue.

With this design, assuming that the sources are .250" in diameter, the "focal point" will be .455 inches above the aperture plate. If the aperture plate is 1/16" thick and the detector is placed .080"/2mm behind the plate, the detector will be about .6 inches from the sample. These numbers are consistent with my estimates obtained with my OpenSCAD model.

Once the focal point is known, we also know how tall the focus ring needs to be, .455 inches. I'd bore one end of my rod to an ID of 1.035" and a depth of (.455-.177 = ,.278") then flip it around, bore the remainder out to .681", and finally cut the 45 degree taper. The ID of the focus ring is larger than the aperture plate hole because the shield ring has to fit in there. This design is the one where the sample is placed in contact with the focus ring -- assuming the sample is larger than the mouth of the ring.

It might be tempting to decrease the focal point more to get everything even closer together, but this will increase the angle between the sample surface and incident x-rays, effectively reducing the flux density. My gut feeling is that a source tilt of 45 degrees is about optimum, but at present I don't have any calculations to prove it.

 

[graham-xrf]

@homebrewed Hi Mark.
I know it looks like we have a "focal point", and indeed the original diagram where I worked out the geometry makes it look like there is one, but in this case, I don't think there really is one in the way we would like. I imagine only the extremes of possible paths and the stuff that might get hit. Those X-rays that escape the shielding shadows will generally hit material everywhere outside of the shadows, and by tilting the sources to best get gamma irradiation at places directly under the diode should increase the count.

Then, when the material does deliver X-rays, again, they will start from everywhere there was material that could get a dose of gamma. X-rays will scatter in all directions everywhere, but, if lots are directly under the diode, such that the diode size occupies lots of the available exit directions, they can't escape, and we increase the count. Wow! It would be so great if we could use some kind of "lens" to concentrate and gather this stuff, but I cant think of there being a "focus" in the sense I wish for. Still, I do know what you mean.

 

[graham-xrf]

Secondary excitations?
Yeah - jumping the gun a bit, but I was thinking ahead, playing "what if"?
Suppose we put a sample of pure stuff in, one of your expensive cubes (say), and get a single bin of it's count. Looks OK.

That situation would obviously change with some alloy mixture. We should get some other bins, but we hope that the counts would show the proportions of the elements in the alloy, taken as fraction of the total. One expects that regardless the proportions, the mere presence of certain bins would usually be enough to identify the alloy.
BUT..
The gamma doing the radiating might not be equal-opportunity in hitting on atoms. Some might be "bigger", or something like that.
Then also, what if some of the the X-ray photons we would love to count happen to be of high enough energy to be more than some (low) L-shell energies or the other stuff in the sample. Our "count" photon gets diverted into falsely increasing the count of one of the other elements in the mix, instead of making it to our diode. (Hell, I can be such a pessimist sometimes!)

Yeah, we likely have to build a set of "signature" responses for various common materials. This begs the question. Does one get the plot, and then look through a set of plots to then say "Ahh - that must be it"? Alternatively, do we aspire to give it such smarts that it comes up with the most probable plot by itself, overlaid onto the plot in a different colour, and a text box with the name of the alloy? What did that Theremino thing do?

 

[WobblyHand]

Don't know about the Theremino. How about Principal Component Analysis? There's some python libraries that would run on your Pi. Unfortunately, the principal components aren't guaranteed to line up with physical quantities, but they might. Be fun to try. Run Multiple hypotheses and pick the most probable?

This is a problem that has been solved in commercial products, so we just need to dig for some papers, or ugh, glean stuff from marketing blather.

 

[graham-xrf]

Don't know about the Theremino. How about Principal Component Analysis? There's some python libraries that would run on your Pi. Unfortunately, the principal components aren't guaranteed to line up with physical quantities, but they might. Be fun to try. Run Multiple hypotheses and pick the most probable?

This is a problem that has been solved in commercial products, so we just need to dig for some papers, or ugh, glean stuff from marketing blather.

Don't let's put too much into this aspect right now. I have yet to reap all the goodies from the CERN open source software. I am trying for my circuit board right now, and getting the ADC working in a trivial way, just doing the basics.

 

[WobblyHand]

Don't let's put too much into this aspect right now. I have yet to reap all the goodies from the CERN open source software. I am trying for my circuit board right now, and getting the ADC working in a trivial way, just doing the basics.

No worries! Just throwing something up in the air. Believe me, you are faaaaaaaaaaar ahead of me. I haven't designed any circuit boards for this. Waiting for the dust to settle, as they say. Haven't bought anything, just following along for the moment. Really quietly cheering this along.

I'm sure you can get the ADC to sample. I'd recommend it be either externally triggered or via DMA and an internal timer. I had a devil of a time getting my homemade doppler chronograph running correctly until I hooked up the sampling trigger to a timed DMA process. Otherwise I just lost sensitivity due to timing jitter. Had to design and build a front end for it as well, with a 7 pole Chebychev analog active low pass filter, with low noise, since all the published schematics were hopelessly bad for aliasing. It was tricky designing and testing it at home, since I no longer had access to expensive test equipment. Had to make an Arduino swept tone generator to test out my filter. The sweeper wasn't very good, but it was good enough for the characterization.

I haven't tried DMA sampling in an RPI, but in an M4 Arduino, it is certainly doable. The 12 bit ADC is running continuously at 60 KHz sample rate with concurrent signal processing (FFT's and automatic target detection) and driving a small TFT display. Took a while to get it to work to my design specs, but was a fun project. You know, I ought to make a housing for this...

I'm sure this X-RF project will turn out well. If there's anything I could help out with, let me know.

 

[homebrewed]

Don't know about the Theremino. How about Principal Component Analysis? There's some python libraries that would run on your Pi. Unfortunately, the principal components aren't guaranteed to line up with physical quantities, but they might. Be fun to try. Run Multiple hypotheses and pick the most probable?

This is a problem that has been solved in commercial products, so we just need to dig for some papers, or ugh, glean stuff from marketing blather.

IIRC, PCA is one of the commonly-used methods to extract elemental information from spectra that exhibit a lot of overlap. This is not a new or unexpected problem, all "proportional" x-ray detectors have some spread in the detected photon energy. It's statistics.

Another approach is deconvolution, but it requires a consistent energy distribution (usually assumed to be a Gaussian distribution).

My scheme of using energy filters may also be helpful but we need to iron out the basic geometry, detector, electronics and MCA S/W first.

 

[homebrewed]

@homebrewed Hi Mark.
I know it looks like we have a "focal point", and indeed the original diagram where I worked out the geometry makes it look like there is one, but in this case, I don't think there really is one in the way we would like. I imagine only the extremes of possible paths and the stuff that might get hit. Those X-rays that escape the shielding shadows will generally hit material everywhere outside of the shadows, and by tilting the sources to best get gamma irradiation at places directly under the diode should increase the count.

Then, when the material does deliver X-rays, again, they will start from everywhere there was material that could get a dose of gamma. X-rays will scatter in all directions everywhere, but, if lots are directly under the diode, such that the diode size occupies lots of the available exit directions, they can't escape, and we increase the count. Wow! It would be so great if we could use some kind of "lens" to concentrate and gather this stuff, but I cant think of there being a "focus" in the sense I wish for. Still, I do know what you mean.

Yes, the name "focus ring" is misleading. More correct might be something like "director" ring, since it just positions the sources to obtain the highest possible fluence of the 60Kev primaries, consistent with the added requirements of shielding the detector from the same primaries and maximizing the number of XRF photons captured by the detector. It doesn't increase the apparent brightness of the primary x-rays.

There _are_ devices that can focus x-rays. One type is a grazing-incidence mirror, typically made with alternating layers of material to behave like a dielectric mirror, but at x-ray wavelengths. High-z and low-z metal films are used in this case. The other is a diffractive lens. Both are far beyond the capabilities of hobbyist types to fabricate; and in the case of the diffractive lens, at least 50% of the x-rays never make it past the lens. I also doubt we will ever find something like them on ebay