Needing more than a spark test?

[graham-xrf]

Just to be a contrarian...

Going back to the NIST X-Ray mass attenuation data, lead's absorptivity at 60KV is 5.02, if thickness is in cm. The intensity of an exit beam passing through a sheet of lead can be calculated using: exit-intensity = incident-intensity * e^(-5.02*t). If you want to attentuate the beam by 99% -- i.e., exit-intensity = .01*incident-intensity this equation predicts you'd need .92cm to do it: ln(.01) = -4.6, so t = 4.6/5.02 (there are negative signs for both the numerator and denominator so they cancel).

In contrast, lead's mass attenuation for 6KV x-rays, which is close to the XRF spectra of interest, is 467 so you'd only need about 99 microns to drop the intensity by 99%. So an order of magnitude energy change results in about two orders of magnitude in the attenuation factor.

The highest energy photon of interest is about 59KeV, that being the most a Am241 can wring out of it's quantum probability electron re-arrangements.

Tungsten will shine at 59.318KeV, and 67.244KeV , so one inside of 59.541KeV, and one outside of Am241 excitation.
If Am241 shows up so clear in the Theremino plot on page #9, then so will tungsten it gets fired up.
The other two glows are 8.4KeV and 9.7KeV.

We set the lead thickness to sufficient to block up to this energy - or we end up looking at a whole lot of unwanted scatter.

 

[RJSakowski]

Just to be a contrarian...

Going back to the NIST X-Ray mass attenuation data, lead's absorptivity at 60KV is 5.02, if thickness is in cm. The intensity of an exit beam passing through a sheet of lead can be calculated using: exit-intensity = incident-intensity * e^(-5.02*t). If you want to attentuate the beam by 99% -- i.e., exit-intensity = .01*incident-intensity this equation predicts you'd need .92cm to do it: ln(.01) = -4.6, so t = 4.6/5.02 (there are negative signs for both the numerator and denominator so they cancel).

In contrast, lead's mass attenuation for 6KV x-rays, which is close to the XRF spectra of interest, is 467 so you'd only need about 99 microns to drop the intensity by 99%. So an order of magnitude energy change results in about two orders of magnitude in the attenuation factor.

From the NIST site:
******************************************************************************************************************************************************
A narrow beam of monoenergetic photons with an incident intensity Io, penetrating a layer of material with mass thickness x and density ρ, emerges with intensity I given by the exponential attenuation law

(eq 1)​

Equation (1) can be rewritten as

(eq 2)​

from which μ/ρ can be obtained from measured values of Io, I and x.

Note that the mass thickness is defined as the mass per unit area, and is obtained by multiplying the thickness t by the density ρ, i.e., x = ρt.


*********************************************************************************************************************************

The t in your equation is mass/unit area and has to be divided by the physical density to get the thickness. So .917/11.35 = 081cm or .81mm of lead will be required for 99% attenuation.

 

[graham-xrf]

So.. if 0.81mm of lead causes 99% attenuation, then the transmission is 1-0.99 = 0.01
Then 2 x 0.81mm = 1.62mm causes 0.01 x 0.01 = 0.0001 , or 99.99% attenuation

1.8mm of lead (the common code#4) is more than that 1.62mm. It will do.
(1.8/0.81 = 2.222 (2's forever). Code #4 lead amounts to 2.22 layers of 0.81

Using the simple observation..
Transmission = 0.01 ^ (no. of 0.81 thicknesses) = 0.01 ^ 2.22222222?
= 3.5938136638E-5
.. and attenuation (1 - 3.5938136638E-5) = 0.999964061863

Even code #3 lead (1.32mm) is worth 1.6296 layers of 0.81.
0.01^1.62 = 0.000575439937337
So attenuation is 1-transmission ≅ 0.999449.

So - it is 1.32mm lead if you like 99.945%
But if you want to push the decimals for a satisfying set of 9's, use the 1.8mm.
We are not trying to safety shield smoke detector radiation. We are trying to limit responses and noise from unwanted stuff, and get a clean plot.

OK - I get it. We don't need loads of thick lead!
( I can already imagine the lady of the house saying.. "It's bad enough you bring your radioactive junk into the house, and now you say your less than 1mm thick lead shielding is going to let as much as 1% leak out" ) ? !!!!!

 

[homebrewed]

The t in your equation is mass/unit area and has to be divided by the physical density to get the thickness. So .917/11.35 = 081cm or .81mm of lead will be required for 99% attenuation.

Aha! Thanks for the clarification! I thought that a cm of lead seemed like overkill. That also means my calculations on scintillator dimensions are off, too....they can be even shorter and still absorb most of the incident ~6Kev photons.

 

[graham-xrf]

It occurs to me that a simple cheap steel tube would stop the stuff, but we need to control, collimate, and collect without scatter from sundry materials. Lead looks the best, and if we do it right, given that photons still cannot go around corners, we can even pehaps arrange the geometry that XRF from lead in the test sample can be counted, because we know it did not come from the shielding.

 

[homebrewed]

I'm not planning on using steel for shielding because it could produce a substantial contribution to the "background" XRF signal. Lead does have an x-ray line at 10.549Kev, quite a ways away from vanadium through nickel. Nickel is the closest at 7.472Kev.

 

[graham-xrf]

I'm not planning on using steel for shielding because it could produce a substantial contribution to the "background" XRF signal. Lead does have an x-ray line at 10.549Kev, quite a ways away from vanadium through nickel. Nickel is the closest at 7.472Kev.

I agree - and it's not to hard to source some lead sheet.

Re: False positives from the shielding
This is a trick of geometry. If the shielding (a truncated inner ring cone) is so arranged to shadow the Am241 sources such that responses from the shielding can't get back to the scintillator, because they cannot do any more than diffract at an edge, and cannot go around corners, nor reverse direction, we have substantially reduced the shielding scatter.

Then go a little further. If the short lead collimator respond with X-Rays off the "hot" side, those X-Rays cannot re-stimulate any lead into fluorescence. I am still messing with the angles in a sketch, but I think one could eliminate response from shielding lead, and even get a genuine xrf response from a test lead sample, or lead in alloys, if placed in the central un-shadowed region.

The 60/40 mix of tin and lead in traditional solder will do as a check sample.

I don't see shadowing, collimating, nor controlling the "beam" reduces the output. All the photons that hit the shielding were headed in a useless direction anyway. Neither do those "response" photons provoked a from a gamma clipping the edge of the shielding get to make it up the scintillator (separate) shield.

When I have the cross-section sketch figured out, I will post it here.

The "slow pulse" loose end.
If you can recall, you once came across the rationale for "pulse stretching" with a low-pass filter.

I think the idea was that losing the higher frequency components from the photomultiplied raw scintillation waveshape (by filtering), to leave a delayed, attenuated, pulse that still contained enough information to identify the material. I think all that gets used is the peak amplitude.

This by simply storing the peak value on a analogue peak detector circuit, the accessed and captured at leisure with a low cost, and hence very slow A/D converter.

There was also, I think, a collection of spectra where the contributer was using some data modification technique to "make the bumps stand out better".

This is not the way I was thinking, but since the cost of a good A/D converter is anywhere from about $6 to more than $40, I wanted at least to understand it.

If you still have the link, or a stored explanation pdf, I would love to get at it. Many thanks if you can find it again.

 

[graham-xrf]

Should we call this thing "HM-XRF" or something?

 

[homebrewed]

If you still have the link, or a stored explanation pdf, I would love to get at it. Many thanks if you can find it again.

It might be this one: theremino data processing. You will need to tell your browser to translate it to English. This page has PDFs regarding their data processing as well but you will need to scroll about halfway down this very long page. Look for links to things like "ThereminoMCA_Deconvolution_Eng.pdf". You don't need to translate many of these to English (but the English versions have some peculiarities that need puzzling out). For instance, I think they are using "pictures" when they mean "pixels".

The relevant info on their web site is a little hard to track down, sometimes it is necessary to page through the entire page before you get to what you want.

The Theremino folks claim that pulse stretching improves the quality of their results, something I'm still trying to get my head around. I suspect the filter network behaves like a charge amplifier -- basically a passive (and fast/cheap) integrator, so the peak value is proportional to the total energy in the pulse. And it automagically resets so previous pulses don't affect the quality of the data. THIS makes sense to me: but I have been wrong before. Once or twice anyway

I need to find a decent Spice simulator that runs on Linux to check some of these ideas out. EasyEDA claims to include a simulator in their free web-hosted schematic capture and PCB layout tool so I will look into that to see how difficult it is to use. Since it's browser-based it runs independently of the OS.

 

[graham-xrf]

I need to find a decent Spice simulator that runs on Linux to check some of these ideas out. EasyEDA claims to include a simulator in their free web-hosted schematic capture and PCB layout tool so I will look into that to see how difficult it is to use. Since it's browser-based it runs independently of the OS.

Thanks for the Theremino link

Re: SPICE
ng-spice works with Kicad, gEDA, (forked lepton-eda) and a number of others.

The full blown tool is qucs. That is more than original Spice network matrix solving. It extends into harmonic balance and suchlike.

Getting up an instant powerful spice-only tool, use immediately, free, is to get LTSpiceXVII from Linear Technology (now part of Analog Devices). It loaded into my Linux Mint, and when I clicked on it, the system fetched and installed wine, and installed it.

LTSpice was aimed at letting prospective customers figure out designs for switched-mode power supplies, using quite elaborate models of LT products as sub-circuits and fast mathematical models, but it is a general purpose SPICE engine.

Now that Linear Tech is part of Analog Devices, I am about to check out if the A/D converters are modeled.



https://www.analog.com/en/design-center/design-tools-and-calculators/ltspice-simulator.html#
You can, of course, run it under Windows. Your mileage will vary. It runs perfectly OK on Linux under Wine.

Kicad can make any PCB you like, and produces files to have the made online, and posted.

If you want to keep your EDA programs in-house, then..
apt update
apt install kicad
or .. just use your package installer - Synaptic, whatever.

gEDA, or it's fork Lepton-EDA also has all the power, but it takes a significant effort to go up it's learning curve, make the models, get the connections between footprints and schematics to work seamlessly.

Right now, believe it or not, sketches in school-style 5mm squares feint ruled paper using colour crayons, pencils, and generally anything that makes a mark actually works.

Getting back to the point about A/D sampling. Everything I know says that one should grab as much information as possible, with as high a sample rate as possible, and then apply the smarts. If all we need is the value of a peak, and we have a way of "telling" if there are two peaks, or some other combination, then we can use a simpler analogue electronics and a slow sampler. I am willing to explore both ways, but looking at the titles in the PyMca software, clearly they want a good representation of the whole scintillation.

Re: Theremino. All the code related to the PIC at the heart of it is compiled from C sources available. Not so the display. The "app" that goes on a PC, apparently to handle various data protocols for external devices, communicated via USB, is a closed source proprietary .exe executable, and no app for a phone that I could see. I won't be using it. The Pi is low cost enough, and powerful enough to do it's own video, and if you really do need to have it show on some other PC's monitor, then use NoMachine and a network. A USB link to phone which provides a remote display - like a GoPro, can be a separate project.