Needing more than a spark test?

[homebrewed]

I am not sure that it is so. I am OK to get convinced.

I understood it differently.
The incoming the x-ray photon has a fixed energy. It causes the release of carriers. They can only be electrons released from the photoelectric effect, and quite exactly, the energy used in that action corresponds to the energy one might derive from dissipating those electrons.

That number of electrons becomes a charge, on a capacitance, all very suddenly, which will run away and dissipate as currents in whatever it discovers itself connected to. Part goes into the reverse bias resistance. The rest goes into whatever is outside, including a TIA input.

I never saw the current one gets out of a photodiode as a straightforward analogue of the photon energy in Joules of what arrived.
It's not a big problem, because if the content of the current pulse is indeed already a measure of incoming energy, then counting up the waveform of current conveniently does not need squaring.
That said - I will find out, one way or another.

It's always good to review the premises of a complex system like the one we're working on. Sometimes incorrect assumptions become embedded in the "lore" of the art/science, similar to the creation of an Urban Legend.

I personally encountered that in the case of a problem affecting a device that was meant to be placed in front of a CRT to create a 3D display. The combination created annoying display artifacts that were commonly believed to be unavoidable due to the physics involved but that assumption turned out to be incorrect. A simple change to the wiring harness fixed the problem, but it wouldn't have happened if I hadn't questioned the assumption regarding the nature of the problem.

 

[WobblyHand]

Our assumptions can certainly lead us astray. It's good to take stock and review what we think is happening.

Got one bathroom with full running water, so at least the house is habitable. Only 1-1/2 more bathrooms to go. What a day this has been.

 

[homebrewed]

I have had a chance to test my ADC board and driver S/W. It took a bit of debugging to get the driver working, basically because I didn't pay enough attention to some important definitions in core_pins.h so I actually wasn't setting the ADC's control pins.

After resolving that, I got some results but they were disappointing My initial test code sets up a circular buffer that contains 1024 16-bit values, then analyzes the data and calculates the average number of bits in each sample. This, assuming the input is grounded. I was getting about 1,000 counts on average. Not so great. But I had some debug code in the ADC's ISR that toggled the LED output to provide a kind of "heartbeat" to verify that the ISR was working (it was). I disabled that -- and that alone -- and the average counts dropped to 3 per sample. Now, I'm still not convinced that is real, although I can easily believe that a fast ~50mA current pulse right around the time the ADC is triggered could result in some system noise! 1,000 counts out of 65536, with a Vref = 2.4V, is about 37mV. 3 counts is about 110 microvolts. This calculation is an average so not necessarily good when it comes to processing a digitized single pulse -- but I'd say it is encouraging.

The next step is to connect a signal generator to the input and do some things -- first plot the waveform to make sure I'm correctly assembling all 16 bits and if that looks OK calculate the FFT.

But so far the results are encouraging. At the least, I believe that the 8 LSB's are correct -- so that's something.

The other caveat is that I didn't aim too high in the sample rate for my initial evaluations. 100KSPS. No idea at this point if the baseline noise level will go up when the sample rate goes up or not.

 

[graham-xrf]

I have had a chance to test my ADC board and driver S/W. It took a bit of debugging to get the driver working, basically because I didn't pay enough attention to some important definitions in core_pins.h so I actually wasn't setting the ADC's control pins.

After resolving that, I got some results but they were disappointing My initial test code sets up a circular buffer that contains 1024 16-bit values, then analyzes the data and calculates the average number of bits in each sample. This, assuming the input is grounded. I was getting about 1,000 counts on average. Not so great. But I had some debug code in the ADC's ISR that toggled the LED output to provide a kind of "heartbeat" to verify that the ISR was working (it was). I disabled that -- and that alone -- and the average counts dropped to 3 per sample. Now, I'm still not convinced that is real, although I can easily believe that a fast ~50mA current pulse right around the time the ADC is triggered could result in some system noise! 1,000 counts out of 65536, with a Vref = 2.4V, is about 37mV. 3 counts is about 110 microvolts. This calculation is an average so not necessarily good when it comes to processing a digitized single pulse -- but I'd say it is encouraging.

The next step is to connect a signal generator to the input and do some things -- first plot the waveform to make sure I'm correctly assembling all 16 bits and if that looks OK calculate the FFT.

But so far the results are encouraging. At the least, I believe that the 8 LSB's are correct -- so that's something.

The other caveat is that I didn't aim too high in the sample rate for my initial evaluations. 100KSPS. No idea at this point if the baseline noise level will go up when the sample rate goes up or not.

To isolate an ADC from the computer that is trying to get data from it is not easy when the ADC is able to read down in the microvolts. The very act of "clocking" on it can bounce things around. With some care, it can be done without resorting to opto-isolator tricks.

The common 0V return for digital stuff has to meet the analog stuff somewhere, or it would not be possible to transfer any information. That "star" point is in the ADC chip, which has careful separate circuit paths for the separated power rails. Aside from ensuring the OV ground plane has a slot in it to force the return currents to keep to their separate paths, it can help to use higher impedance in the digital lines to limit the currents in the return. I once used 100Ω resistors in the digital data lines, and hoped it would not limit the speed. In fact, they did damp ringing, and actually helped.

Going straight for waveforms would not be my first act. Shorting the input, and trying to get all counts as low as possible is what I found convincing. Then offer it a DC derived from the reference by two resistors, with 10nF or so capacitors across them, and see what the count looks like. Counts out of 16 bits as low as 250 and up to 1800 can be "usual" before going for cleanup. I would settle for anything less than 80. I once got an ADC to count around 39, +/- about 5 counts. This was OK, because the signal it had to look at had more than that in noise anyway.

If you are already looking at counts as low as 3, and you trust the 8 LSB, then I am envious!

 

[graham-xrf]

Our assumptions can certainly lead us astray. It's good to take stock and review what we think is happening.
Got one bathroom with full running water, so at least the house is habitable. Only 1-1/2 more bathrooms to go. What a day this has been.

Bruce - I sympathize, and I know exactly what it is like to have to keep critical house infrastructure working.
Well done on the progress!

I admit I am having some trouble visualizing what a completed place is like with 2+1/2 bathrooms!
(The dash was a minus sign ?)

 

[graham-xrf]

Looking at the lighter ones..
I mean aluminium, and magnesium. The X-rays they emit are low energy, and may not have the power to exit their own sample from the depth they were excited at. It means the efficiency of the return is low. Then again, they have to get into the detector, That too, is at low efficiency.

It seems that if one wants to to XRF in that region, one needs a SDD (Silicon Drift Detector), and a big bank balance!

 

[WobblyHand]

Bruce - I sympathize, and I know exactly what it is like to have to keep critical house infrastructure working.
Well done on the progress!

I admit I am having some trouble visualizing what a completed place is like with 2+1/2 bathrooms!
(The dash was a minus sign ?)

In US real estate parlance, a full bath has a tub, shower sink and toilet. The laundry room has just 1/2 of those items, a sink and a toilet. Strange language, isn't it! I suppose I added to the confusion by stating 2-1/2, meaning 2.5, or I suppose 2,5 depending on your locale.

Due to a significant heat rise from energy hungry space heaters, all of the bathrooms are now functional. Nothing burst, for which I am exceedingly grateful. Was a long day. Some of our modern appliances took umbrage with the low temperatures, still haven't cleared their codes. Our dishwasher thinks it has a blockage, and the washing machine is throwing an unbalanced load error, even last night when I tested it with no load inside it. Life with smart but dumb electronics. Besides that, things are mostly sorted.

This morning was more seasonal, being 37F higher than yesterday morning. 20F is not warm, but lots warmer than -17F. Today it should get to 45F, which compared to last morning is tropical!

Eager to get back to pondering XRF. Did you ever find pricing information on the SDD's? Just curious what they cost, considering their alleged advantages. Even an order of magnitude cost estimate could be helpful. Considering that the current diode is let's say 70 USD, what might an silicon drift diode cost. If it's less than 150, might be worth looking into. If 700, well, don't think it's viable for the target group. At least for me, 700 USD just for the SDD is too high for my casual interest. Hopefully, the PIN diode is adequate for our application. It does concern me that it's going out of production. Would make it tough for someone reading this thread five years from now to find out it's impossible to make any more. Most people just can't diffuse their own diode!

 

[graham-xrf]

This morning was more seasonal, being 37F higher than yesterday morning. 20F is not warm, but lots warmer than -17F. Today it should get to 45F, which compared to last morning is tropical!

We have been seeing reports on BBC News about the Arctic blast you folk have been having, and record low temperatures at Mt Washington.

Eager to get back to pondering XRF. Did you ever find pricing information on the SDD's? Just curious what they cost, considering their alleged advantages. Even an order of magnitude cost estimate could be helpful. Considering that the current diode is let's say 70 USD, what might an silicon drift diode cost. If it's less than 150, might be worth looking into. If 700, well, don't think it's viable for the target group. At least for me, 700 USD just for the SDD is too high for my casual interest.

I had been looking around for alternative PIN diode all the time. Also now getting deeply into getting at the expression for current from a single photon hit. This is tricky, because all expressions I find are about the rate of input (Watts) of optical power, which can be had in photons/second.

When we get into bandgap energies, and what happens in the production of electron-hole carrier pairs in a depletion zone, and how what happens includes the carrier acceleration effects that take some energy from the bias. The extreme form of this is SPADs (Single Photon Avalanche Diodes). They are extremely sensitive, with gain, kind of like a solid state version of a PMT, but by their nature, are just event-counting Geiger mode devices.

Look on Wikipedia, and discover how unusual and difficult it is to even contrive single photon sources.

The XRF industry is the big gear here. They want resolutions of one or two hundred EV. They want to see energies down to Magnesium, Aluminium, Sodium, and Oxygen, using only Peltier coolers to -30C instead of having to deal with liquid Nitrogen.
Just look at the Ketek site --> https://www.ketek.net/sdd/
Check out the spectra here --> https://www.ketek.net/sdd/vitus-sdd-modules/
When it's hard to find a price, even just a ball-park value, without getting a quote, you know it's expensive!
If an SDD device cad be had for under $250, then maybe, but I am happy to play with what I have.

In 2021, Ketek sold it's silicon photomultiplier assets to Broadcom.
I am not sure what all that means because SDD's are not photomultipliers like SPADs, and are really for materials detection rather than high speed optical communications.

Hopefully, the PIN diode is adequate for our application. It does concern me that it's going out of production. Would make it tough for someone reading this thread five years from now to find out it's impossible to make any more. Most people just can't diffuse their own diode!

Do not worry about it too much. Despite that we are exploring the hell out of PIN photodiode XRF technology devoted to a device that costs a very low fraction of what is sold in industry, we are flexible/adaptable enough to manage with what we can find. With the diode you have, there is a reasonable chance we can get spectra as good as the PMT.

My scintillator PMT combination, all "new old stock", cost less than getting a SparkFun Pocket Geiger. Of course, a PMT does not need anything like the huge gain ultra-low noise kit, because of the stages of photo-multiplication, but instead, needs a high voltage generator. Not difficult, but a different project. I think Mark already has one.

 

[homebrewed]

It does concern me that it's going out of production. Would make it tough for someone reading this thread five years from now to find out it's impossible to make any more. Most people just can't diffuse their own diode!

An alternative would be to combine a piece cut down from this stuck down onto one of these. CdWO6 is reported to be good for low-energy xrays, non hygroscopic and offers about 8% resolution. The 20 x 20mm scintillator could be cut into a bunch of smaller pieces so the cost wouldn't be that bad on a per-detector basis.

If our PIN diode doesn't provide good-enough energy resolution to determine alloy composition I think that's the direction I will take.

At present I don't have a PMT but I do have a couple pieces of LYSO scintillator I can play with.

 

[graham-xrf]

An alternative would be to combine a piece cut down from this stuck down onto one of these. CdWO6 is reported to be good for low-energy xrays, non hygroscopic and offers about 8% resolution. The 20 x 20mm scintillator could be cut into a bunch of smaller pieces so the cost wouldn't be that bad on a per-detector basis.

If our PIN diode doesn't provide good-enough energy resolution to determine alloy composition I think that's the direction I will take.

At present I don't have a PMT but I do have a couple pieces of LYSO scintillator I can play with.

This one is actually more attractive in concept in some ways than the PIN photodiode.
What we need is some sort of graph of cadmium tungstate response vs KeV.
From the numbers given, it says 12 to 15 photons/KeV. So even looking at Am241 source, that comes to 714 to 893 photons.
I am thinking that is kinda not much, and it's of a sort of sky blue colour, which is OK, if it were bright enough to see some.