Needing more than a spark test?

[homebrewed]

I have some refinements to the geometry that might allow us to eliminate the shield ring altogether. The idea is to move the detector a bit further away from the aperture plate, something on the order of .25 inch vs. .08 inch. The detector will see a slightly lower count rate: but that might be a good thing if we've got 8 Am241 sources blasting away about .5 inch away from the sample.

Apologies for the ongoing variations on this part of the system, but the changes are dictated by my observations and conclusions based on the construction process. Hopefully all this will result in a much easier to fabricate XRF system, without giving up much in the process!

 

[graham-xrf]

Indeed yes. Integrate the pulse, because it is the area under the pulse that quite accurately represents energy.
I am less sure about that if the pulse were stretched, and there are several kinds of stretching scenario.

One is pulse duration were stretched, but the peak amplitude maintained, or restored by gain. I think this amounts to increasing the area by a constant. This is OK

Another is the pulse is stretched, and the amplitude drops, such that the area is maintained. When I think it through, that is just a version of the first, where one neglected to restore by adding gain. This should also be OK

Another is where the stretching was accompanied by nonlinear distortion that falsifies the area energy. All filters bring delay. A set of Bessel low-pass filters, with roll-offs above the highest frequency component gives a constant group delay. A tapped delay line, with tappings summed in various way can allow one to make all sorts. We don't want to be doing high speed digital FIR, echo-equalizers, feedback fllters, etc.
Even so, even with "distortion", so long as the nature of the distortion is constant, being equal-opportunity for all signals, then at calibration time, if what you are using is the integrated pulse, your scheme should be just fine.

A very accurate, and immediate, value could be sampled off the end of an analog integrator. Whether one uses an opamp with capacitor, or goes instead for having the computer add up the set of samples through the pulse with constant, short enough, time periods, both situations require the recognition of the start, and a decided end to the pulse.

Getting an opamp integrator needs the trigger to reset it, instead of having it pump up and down on the waveform. A software integrator needs have some way of deciding when is OK to have another count-up. Either way, it feels good!

 

[homebrewed]

Getting an opamp integrator needs the trigger to reset it, instead of having it pump up and down on the waveform. A software integrator needs have some way of deciding when is OK to have another count-up. Either way, it feels good!

It occurs to me that if we've got a pulse width discriminator we already have some sort of trigger circuit going. The "start" signal could open an analog switch to allow the integrator to do its thing. Depending on the elapsed time, the "stop" signal would either trigger an acquisition or, in effect, say "naah I don't think so" and reset the integrator to wait for a good pulse to come along. This is some pretty minimal processing. The only potential issue is that we can't prevent noise from perturbing the trigger points so there will be some unavoidable jitter in the timing.

The same basic flow _could_ be done with a speedy enough A/D-equipped microprocessor and in fact wouldn't need a fancy circular buffer.

There are some issues like charge injection from the analog switches getting into the integrator but that should be a relatively constant offset. Anyway, well-designed analog switches have pretty low charge injection, on the order of pico-coulombs. This wouldn't be a concern with a microprocessor based design but there are other factors, like latency, that could have an impact on a mostly-digital approach.

 

[graham-xrf]

I am thinking your scheme should be OK once one has a pretty stiff signal with a good S/N ratio already captured in the first amplifier.

Already built-in on the Pocket-Geiger is the LM393 that was used to make an output slam up (down??) as the amplified signal passed a threshold, and the other output has it happen at a different threshold, all set by R1, R2, R15, R16. Unless you have already cut and re-purposed the tracks layout, there might be the basis for a trigger monitor, such that if the signal rises above something significantly above zero, is taken to mean a real pulse is there, and the software starts a capture.

The cute dodge here is that if the ADC is always kept running, sampling the signal all the time into a FIFO stack of (say) 20 long, or however many you deem enough to get the resolution number of samples through the pulse you want, you can have the pre-trigger values anyway. The software fetches it's sample set starting from a point a few samples before the trigger. You get the definite constant predictable solid trigger, and also get to fetch the pre-trigger values that were there a bit before it. They may not have been suitable for a reliable trigger, but they might be quite fine to improve the accuracy of the area count. This dodge may have to be applied post stretching. Of course, the whole scheme can be done entirely in the programming.

The returning "pulse is all over" signal might be a cheapskate way to tell you have a pulse overlap, so then know to discard the count. You could re-use the first trigger on it's way back. If the comparison is still up by the time the software has got to (stretched version) 13uS, or maybe 18uS worth, then that may be evidence of a smeared pulse that maybe overlapped and would be artificially extending the duration, and making a wrong integration count.

Getting an unrealistic high integration as evidence for rejection may be enough, but the uncompromising way is to limit the duration, and also know the pulse had subsided below a threshold at cut-off time.

 

[homebrewed]

I am thinking your scheme should be OK once one has a pretty stiff signal with a good S/N ratio already captured in the first amplifier.

Already built-in on the Pocket-Geiger is the LM393 that was used to make an output slam up (down??) as the amplified signal passed a threshold, and the other output has it happen at a different threshold, all set by R1, R2, R15, R16. Unless you have already cut and re-purposed the tracks layout, there might be the basis for a trigger monitor, such that if the signal rises above something significantly above zero, is taken to mean a real pulse is there, and the software starts a capture.

The cute dodge here is that if the ADC is always kept running, sampling the signal all the time into a FIFO stack of (say) 20 long, or however many you deem enough to get the resolution number of samples through the pulse you want, you can have the pre-trigger values anyway. The software fetches it's sample set starting from a point a few samples before the trigger. You get the definite constant predictable solid trigger, and also get to fetch the pre-trigger values that were there a bit before it. They may not have been suitable for a reliable trigger, but they might be quite fine to improve the accuracy of the area count. This dodge may have to be applied post stretching. Of course, the whole scheme can be done entirely in the programming.

The returning "pulse is all over" signal might be a cheapskate way to tell you have a pulse overlap, so then know to discard the count. You could re-use the first trigger on it's way back. If the comparison is still up by the time the software has got to (stretched version) 13uS, or maybe 18uS worth, then that may be evidence of a smeared pulse that maybe overlapped and would be artificially extending the duration, and making a wrong integration count.

Getting an unrealistic high integration as evidence for rejection may be enough, but the uncompromising way is to limit the duration, and also know the pulse had subsided below a threshold at cut-off time.

I like the fact that we are moving toward a general framework for the MCA part. Using the pocket geiger's on-board comparators for the trigger signal is a clever idea and certainly worth evaluating! I haven't disable that part so can easily test the idea out.

But as a cautionary note it's important to keep in mind that we have no fully-functional front end yet. I'm hoping to have a particular version of that going soon, but you never know what's going to come up. Case in point: I wasn't expecting to have to repair my lathe. Hopefully I've been able to get past my immediate need for the lathe. The bearings and transmission gears arrived yesterday....

 

[graham-xrf]

Oh boy! Bearings and busted gears!
Not unknown in this forum, but you will certainly have the sympathies of most folk here.

I still envy that your lathe is at least somewhat together. I had one lathe in pieces, and the other working OK until it too went into a mess. I won't be putting back the underneath drive in the way the previous owner bodge had it. Also, the whole angle-iron-based frame, with drawers is not really what I want to retain. When I say "angle iron", you would expect the sort a tad thicker than 5/32". In other places, it uses that 1/16" shelving stuff with perforations slots. If I can't weld up something somewhat improved, then even a stout wooden bench would work better. When I gave it a firm shove, I could see the movement relative to the wall! It now is to be re-located into my outhouse shop anyway, so I now want a better bench or stand under it.

Relentlessly, I keep having a stacked-up list of stuff I have to trawl through before I can get to play. My new-old-stock ex cold war surplus Geiger tube arrived from Ukraine on Friday, and all I can do with it is stash it for now.

The last thing I needed was was a yell for help from my wife at the end of the drive, arriving back from a shopping trip. "We have a problem with the car"! No power steering, and big red "battery" symbol on the dash had to mean the alternator drive belt, which also turns the hydraulics pump, was kaput. It's is a old(ish) C-Class Mercedes Estate. It must have been a tough work-out for her trying to get it home. The belt was OK, but the tensioner roller, and it's bearing, and retainer cap were a tipped over wreck. I opted to have my favourite non-dealership Merc car service guy bring his recovery truck with the clever ramp thing that just loads the car on. I could have have set about fixing that sort of stuff myself. I have been fixing car engines since I was about 10, but this time, I let someone else do it - and he had the parts in stock!

I am going up the learning curve with KiCad. It's open source Schematic + PCB, but now so capable that in industry, it is becoming preferred to stuff like Altium.

I actually think that our XRF front-end is the one part where we do know what to do. If it is going to work at all, then these bits must obey the physics. There is not much we can do to make those atoms respond any better than the way they would always have done. Agreed, we can get cute with the geometry, in trying to arrange that any responses towards the diode don't have much place left to escape.

For any in the future who would like to try some of this, we, and they, have ever fewer ways of getting hold of anything that reasonably be used to irradiate the sample. About all that is left is smoke detectors. Am241 can only be man-made, a by-product of stuff going on in nuclear power stations, and even these type of detectors are being "phased out" where possible. There is no serious health risk, but they may nevertheless be retired from use as a prudent avoidance measure. I am considering using up the thorium gas mantles I have, packed into a circular trough, held in with epoxy. A chemistry adventure with TIG rods and peroxide may also happen.

 

[graham-xrf]

Check out the little instrument this guy was using. I think it's a straight spectrum energy monitor, as opposed to a materials identification thing, but I don't know. What he does with the screen stylus to zoom expand on the peaks, and read off keV is fascinating.
Thorium Counter Top - Gamma Spectrum Analysis

 

[rwm]

This video shows more about how that works.

He says in the video that it will not measure below 40 KeV so it would not identify all the energies we need.

 

[graham-xrf]

This video shows more about how that works.

He says in the video that it will not measure below 40 KeV so it would not identify all the energies we need.

Hi Robert
Thanks for the extras
I was just looking at the "features" he was getting with the stylus.
As I understand it, most all the MCA programs all do have ability to expand the scale.

 

[homebrewed]

Just the transmission gears are on their way to going kaput. But to change them out I have to remove the spindle, which includes the front bearing. The stock bearings for the 7X lathes are just deep groove ball bearings so I'm taking advantage of the opportunity and will swap them out for angular contact bearings. Tapered roller bearings would be better yet but unlike the AC bearings they're thicker than the originals, which requires some additional changes to the lathe -- positions of other gears change, spacers need to be redone etc. So I decided to just go with the AC bearings. That will get the lathe up & running sooner, and should be much better than the originals.

I don't have a press so I'm going to use all-thread and an assortment of bushings and metal plates to pull/reinstall the bearings off the spindle and the rear end of the headstock. I may need to buy a longer gear/bearing puller to get the bearing off the spindle, but they're not all that expensive. Harbor Freight has a set of 3 for about $40.

There are two sets of gears in the transmission -- one pair is on the spindle and the other pair is on an adjacent intermediate shaft. The OEM replacements are plastic. There is a metal gearset available but they're currently out of stock and won't be back in stock until mid March. Since the first set of plastic gears lasted over 10 years I decided to just stick with the plastic gears. They also are one fifth the price of the metal gears and ARE in stock.