Needing more than a spark test?

[homebrewed]

I bought some 1uf plastic SMT's for a future from-the-ground-up detector board. The plastic caps are substantially larger than 1206's.
X7R and X5R ceramics are highly microphonic, particularly if they have a voltage across them. They are real similar in composition to real-deal piezoelectric devices. They also have higher leakage than polymer caps. There are reasons that it's a bad idea to put high-K ceramic capacitors in the signal path of audio amplifiers.

On a different subject, I have a couple of XRF spectrum I want to share. The first is for 1018 HRS, and the second is for aluminum. I'm _finally_ seeing a peak I believe may actually represent iron! But there also is a pretty strong peak from something else. That peak looks pretty much like the aluminum-only sample.

I think that high aluminum background is secondary xray fluorescence coming from my aluminum "focusing ring", perhaps generated by 6Kev xrays from the steel sample itself (!). I may have shot myself in the foot there, so you folks going with a plastic 3D printed focusing ring will likely get much better results right out of the chute.

Iron + "X" (the iron is the peak on the far right):


Aluminum + a little bit of "Y":


As you can see, the "X" looks a lot like aluminum and the "Y" looks like it could be iron, possibly coming from my little source disks -- or perhaps fastener hardware in my enclosure. I will try strategically placing some lead shielding here and there to see if I can get rid of the extraneous iron peak.

Right now I'm reluctant to scavenge my sources, preferring to keep the current setup for comparison purposes. So I just ordered 8 more sources from Ali Express, too. They are selling them for a lot less than Ebay right now. The sources should arrive the first week of March. This isn't too bad because I want to spend some time on the driver code for my 1MSPS ADC. It will be interesting to see how many "noise bits" my board design has.

 

[graham-xrf]

Mark - can you post the Ali Express link?
Are they pure all the way through?
If the iron is unwanted "any old iron" it may have all sorts of stuff in it, like silicon.

Lead shielded collimation, and it's geometry should prevent even the 10.5KeV and 12.6KeV from lead itself getting in, but as you point out, these may be causing a secondary excitations. At the base, where we put the test sample, we need either the whole aperture to be made of the stuff we are testing, or the sample be put on something that will "just absorb".

Placing it on a lead sheet, or any other metal, may just be the way to get unwanted returns that we have to learn to recognize.

This stuff is getting complicated!

Speculating madly! ...OK, a sheet of plastic, say polycarbonate or acryllic, to let incoming photons pass and with enough space behind they will end up where they are unlikely to make it back. This is hardly "portable"? Hmm.. we need a trick way to ensure that what we put there, without the test sample, is subtracted.

Subtract?
So what happen if we subtract one plot from the other, and either deny, or change sign on all negative buckets? If we are left with random uncorrelated counts, then they were noise error. If we see definite peaks and bumps, there was something going on in one that was not in the other, amid a whole pile of (maybe valid) stuff that was in both?

 

[homebrewed]

Mark - can you post the Ali Express link?

It's the same web page that Bruce posted a link to earlier: here .

Once the background doesn't talk so loud, so to speak, doing some subtraction would be worth trying. Scaling the background spectrum will be necessary if its magnitude is substantially due to fluorescence from the sample. Before worrying about it a lot I really want to see how much that background can be suppressed by going with a plastic holder for the sources.

I agree that stuff like 1018 isn't necessarily just iron with carbon in it. But when counts from the "brand X" component outnumber counts from the sample there's got to be something else going on.

 

[graham-xrf]

It's the same web page that Bruce posted a link to earlier: here .

Oops! Sorry I misunderstood. I was thinking the set of pure test calibration elements you have.

That said, I signed up to AliExpress and bought 10 more of the sources.
I am amazed at the price difference to any from eBay. There was a shipping charge, and tax, but they were £1.30 each, compared to more than £8.00 from eBay. The total for 10 from Ali Express came to a few pennies over £20.

Once the background doesn't talk so loud, so to speak, doing some subtraction would be worth trying. Scaling the background spectrum will be necessary if its magnitude is substantially due to fluorescence from the sample. Before worrying about it a lot I really want to see how much that background can be suppressed by going with a plastic holder for the sources.

Unfortunately, this kind of racket is random. It's not like closing the lens on a camera, and saving the dark response, and then subtracting from the image. Sampling over time, and then removing those that that do not consistently reappear is a bit digital filter type brutal, but might be tried.

Way back, I did attempt to figure what count rate we might expect. I am going to re-visit that. I am concerned that we just do not have enough gamma making returns to count, which is why I bought more sources. I had thought that six would be plenty, but now I am considering crowding in as much as I can reasonably fit in. The first try will be with eight.

For background noise in your setup, if you use plastic source mounts, I take it you still have some lead collimation to stop the diode being hit from the side through the plastic.

Another thought occurs. What might be in the ceramic case material the PIN diode is mounted in? Might the case be hit by lots of Xrays coming from the test sample, and making some returns of its own?

Do we really have to have the diode see the outside through a little 10mm square aperture in lead sheet?

 

[WobblyHand]

Subtraction reduces correlated noise but not random noise. Subtraction increases the noise in general. Averaging reduces noise along with its "equivalent" a narrow band filters. Subtraction can be used to reduce DC offset with little penalty. The key with subtraction is to have the estimated value have very low variance, then the resultant variance is dominated by the input signal. However, the variance always increases. Entropy always increases.

I bought 8 sources, hope that's sufficient.

As for shielding, I was planning to form a lead aperture for the detector. The less stray stuff that gets in the better.

The PIN diode itself obviously has conductors in it or it's lead frame. There may be heavier metals in the form of plating. Whether they are a source of stray radiation due to them is up for discussion. I don't know if the ceramic is loaded with heavy stuff for favorable properties.

 

[WobblyHand]

Britannica states ceramic is 90-94% alumina, the rest is alkaline earth silicates.

 

[graham-xrf]

Britannica states ceramic is 90-94% alumina, the rest is alkaline earth silicates.

OK then - just speculating where the ton of other returns might be coming from..

Al2O3 would be the very white stuff I used to use for insulators in vacuum furnaces, and also for microwave feed parts, because of it's very high dielectric constant (~10). We are talking aluminium and oxygen. The casing is not pure white, so there is other junk in it, and I don't yet know what might be in the "alkaline earth silicates". Sodium hydroxide was much involved in manufacture from bauxite. Even so, let us assume it's these two.

My first thought is if one is showing the device some aluminium to test, we should expect the diode package, having a whole lot of aluminium in it, just might be making responses to primary excitation leakage, if we let it. Also, we have secondary showers of photons coming from stuff in the structure already hit, and also the non-expended energy coming out as new photons, in new directions, some happening to go straight into the diode silicon. I am thinking we are not safe from low energy returns from plastics.

Why we may have extra high energy excitations in directions we don't expect.
We may never be able to detect the tiny 277eV from carbon, nor 392eV from nitrogen. 54eV from hydrogen sounds impossible. We might be seeing their leftovers! I don't really know. All I can see is that logically, when we show it some aluminium, or iron, what comes back might be the responses we want, and also a whole shower of the remaining unused energies as noise photons. Is this notion totally crackpot?

There might be (some) 2.6KeV and 2.8KeV from chlorine in some plastics. These returns, from primary strikes of a 59.54KeV photons from the sources may be trivial, but we know from @RJSakowski that the "left over" energy that was not "used up" will go on to raise the level in other shells in the same atom, and also go on to raise the energy in new electron shells from entirely new atoms.

What is more, the direction the exit photon, hardly much depleted from it's encounter with plastics, can go anywhere. It finds aluminium, and oxygen. These are low energy adsorptions. Aluminium is 1.48KeV with 1.55KeV. Oxygen is a feeble 525eV. A 59.54KeV primary has lots left over after finding plastic, or the test material, or the alumina, and the new direction is uncontrolled. It can exit sideways!

The fully shielded PIN diode
I now think that the photodiode should be totally shielded from all unwanted photon paths, by having a 10mm x 10mm square aperture in lead sheet shielding as the only way in, except for the connections, and they can be covered from having to "turn the corner" past the shielding.

That leaves us to deal with with unwanted returns from the lead. The returns from Pb λ- shell are 10.55KeV and 12.61KeV, and their at worst. scatter leftovers at about 48.99KeV and 46.94KeV . Then, another generation down, possibly more leftovers racket at those energies reduced a bit more by aluminim, oxygen, carbon and hydrogen. All of these are stopped by having the lead thick enough, meaning 1/16" or more. The odd unlucky return might come from the very edge of the shield, but the atomic cross-section fat heavy lead atoms mean that 99.99% is expended in the shield.

Is this bonkers uncompromising? I agree, it's an extreme approach that may be well OTT. In the end, it may be shown to be unnecessary, in which case, one can dump the idea. That said, I would rather start out with it there, than drive myself nuts looking for the noise mechanism, and why there is a background racket.

 

[WobblyHand]

Strictly speaking, alkaline earth elements are not alkali. Alkaline earth elements are Be, Mg, Ca, Sr, Ba, and Ra. (Beryllium, magnesium, calcium, strontium, barium and radium.) Now there's hardly much chance of radium, but there could be calcium & magnesium.

 

[homebrewed]

The data inside an MCA array is a kind of average so the random nature of incoming counts becomes less and less apparent. Otherwise we wouldn't see a spectrum. This is why I think it may be beneficial to perform background subtraction -- it's not on a per-pulse basis. Far from it.

That said, it also is evident that the MCA array also contains a broad set of onesy-twosey counts that, statistically speaking, can never be eliminated. Gotta be careful when it comes to other approaches since we may be looking at alloys with just a few percent content of other elements. I'm inclined to take a hands-off approach as much as possible when it comes to making the spectra look pretty.

 

[graham-xrf]

Yes indeed. We don't care if we get messy spectrum plots, so long as we think the pulses that made them were real, and not too knocked about by noise.

My lead sheet might arrive Monday (ish).