Needing more than a spark test?

[graham-xrf]

I think I found the way it actually works!
The X-ray photon hits the silicon PIN diode.
For silicon, every little 3.66 ± 0.03eV pulls an electron away from a silicon atom, creating a free electron that can move under the influence of electric fields. It leaves a "vacancy" in the silicon atom - a classic "hole" acting like a positive charge. It's an electron-hole pair. Left to itself, in something even slightly conducting, it will very rapidly re-combine, though possibly promiscuously. (It finds a different -oh whatever..)

In regular diodes, just having P+ type holes material up against N- type electrons-to-spare material results in a brief current, and a depletion region where the electrons and holes found each other. Reverse bias can widen the region, and forward bias can narrow it, until it disappears, and diode conducts. In a PIN diode, there is a built-in intrinsic pure silicon layer between the P and the N. Any electron-hole pairs generated in this layer will extremely rapidly cross to the P and N layers. Except for some electron-hole pairs made by being shook up by heat, or encouraged to get pulled free by bias current, (ie. noise), every 3.66eV gets you an electron's worth.

Current is not the flow of electrons
Oh yes it is! Oh no it isn't!
Actually it's the latter. Electrons can take weeks to get around a circuit, and many may never do if the current is a reversing AC. Think if the power generator is pushing electrons from a windmill in the North Sea, and your light bulb is on the grid 100 miles inland. It is the effect of their fields is what transfers the energy. The analogy is a bunch of snooker balls in a line. Hit the the one on the end, and the one on the other end goes on it's way. So it is with our PIN diode.

So - let's say we whack our alloy with some 59.5keV from the smoke detector thingy, and it hits a molybdenum atom.


The L-shell is outermost, with electrons that are the easiest to get excited. It lets go some 2.29316keV.
That one is going to deliver a pulse of 2293.16 ÷ 3.66 = 626.54 electron-hole pairs. Others too - but lets do this one.
Lets gloss over what is 0.54 of an electron If that 626 electons moves in the circuit, it's a pathetic tiny current, and I would think somewhat too small to ever see against the other racket. Nor will all of them make it. There is an efficiency involved. Yet our X-100 specification sheet shows it has a 20% probability of making it.

OK then, just because it got lucky, and was among the 20%, that does not mean our circuit amplifier can see enough of it to give it gain, and offer at an ADC to be reliably have it's count put in a "2.29keV" bucket - does it?

So how much actual amps is 626 electrons? Each one is 1.602176634E-19 coulombs.
626 of them is 1.602176634E-19 x 626 = 1.0029625728E-16 coulombs.
That load of charge makes a funny, roughly triangular pulse about 13uS long, the integrated area under being proportional to the electron-volts.
If current is Coulombs/Sec, and we use 1/2 base x height approximation, then it's 1/2 x 1.00296257288e-16 ÷ 13E-6
The answer is 3.86 pico-amps !

That's disheartening! I know it was a rough-and-ready calculation, but now, hopefully, someone can tell me where I messed it up.

 

[graham-xrf]

Now trying for something like 15fA/√Hz JFET augmented transimpedance amplifier!

 

[homebrewed]

Playing around with my LTSpice simulation, your numbers don't look too far out of line: with 1e-16 coulombs delivered as a 10us triangular pulse I get a peak voltage of around 10mV. Not inconsistent with what I see coming out of my PocketGeiger, with the exception of some giant pulses that may be due to something else like cosmic rays.

 

[homebrewed]

This kind of detection efficiency is not confined to solid state detectors. A look at "high performance" scintillator crystals that have a good proportional response to x-rays showed that they generate something like 30-60 photons per Kev, so low-noise electronics are needed for them as well.

All this might seem discouraging but we already know that the Theremino and Open Physics Lab folks have done this kind of stuff.

 

[graham-xrf]

This kind of detection efficiency is not confined to solid state detectors. A look at "high performance" scintillator crystals that have a good proportional response to x-rays showed that they generate something like 30-60 photons per Kev, so low-noise electronics are needed for them as well.

All this might seem discouraging but we already know that the Theremino and Open Physics Lab folks have done this kind of stuff.

I can get in a flap sometimes
One has to look carefully at datasheet, and appreciate the difference between a noise voltage specification and a noise current specification, and why both are important.
The LTC6269 dual opamp I was after using has a voltage noise (1MHz) of only 4.3nV/√Hz
It has a current noise (100kHz) of only 5.5fA/√Hz
It has input capacitance of 450fF
It has a bias current of 3fA

Just as it is, without the front-end FET, that matches ultralow noise JFET IFN147 + opamps design from Linear Tech.
We should be able to see 3pA.

I got the LTC6269, not only because of the noise performance and gain-bandwidth product, but also because of the package being able to go straight onto the Pocket-Geiger board for experimentation. I forgave the high price (£10.42). Now, I will be going after my own amplifier PCB direct. I just can't do the stuff I want to on the Pocket-Geiger. Also, putting the diode on it,s own little board at right angles off one end edge of the amplifier makes it easier to swap out - like when we discover a different photodiode.

There are other low noise opamps, including one which has actual through-board pins!
I won't use through holes, but I am thinking to try one of the Linear Technology types that that come in small outline SO8, and SO16.
The layout I intend will have a guard ring, and a ground plane under the feedback resistor.
I think there are opamps costing in the range £3 to £6 that might well work just fine.
I have brought out my (somewhat older) digital scope. From "Hewlett Packard" before they became "Keysight".
It's a 54520A, dual channel 500MSa/sec 500 MHz thing.
We shall see.

 

[graham-xrf]

@homebrewed :
Re: The Size of Components
I am making some choices now, looking ahead so they don't come back and give me a hard time later.

So far, I have a main amplifier that has a tiny MSOP package. It happens to fit on one main place on the old Pocket Geiger board, and also the supplier, (Mouser) only keeps stock of MSOP and DFN-10 style parts of the LTC6869 Linear Technology amplifier.

The data sheet shows that is also manufactured with a SO8 package, which would be a whole lot easier to handle.
Recognizing that SO8 may be harder to come by, then regardless of the outlets for a SO8 package version, we have the option of just choosing an amplifier that is more available in SO8. It does not have to be the first I found, and if it was cheaper than the £10.40, so much the better.

This stuff would not matter if one was designing with the intention of commissioning the manufacture of lots of PCBs, pre-assembled by some robotic thing, but for us, we need to experiment with them with our big shaky fingers, and maybe help some folk get a few of them together. Thus I am trying for a build that uses SO size stuff, where I can. It's an intention really. The 16-bit ADC was inevitably in a 0.5mm pitch quad flatpack!

OK, OK, so you can guess that I spent a while closely searching the floor - again!

 

[WobblyHand]

@homebrewed :
Re: The Size of Components
I am making some choices now, looking ahead so they don't come back and give me a hard time later.

So far, I have a main amplifier that has a tiny MSOP package. It happens to fit on one main place on the old Pocket Geiger board, and also the supplier, (Mouser) only keeps stock of MSOP and DFN-10 style parts of the LTC6869 Linear Technology amplifier.

The data sheet shows that is also manufactured with a SO8 package, which would be a whole lot easier to handle.
Recognizing that SO8 may be harder to come by, then regardless of the outlets for a SO8 package version, we have the option of just choosing an amplifier that is more available in SO8. It does not have to be the first I found, and if it was cheaper than the £10.40, so much the better.

This stuff would not matter if one was designing with the intention of commissioning the manufacture of lots of PCBs, pre-assembled by some robotic thing, but for us, we need to experiment with them with our big shaky fingers, and maybe help some folk get a few of them together. Thus I am trying for a build that uses SO size stuff, where I can. It's an intention really. The 16-bit ADC was inevitably in a 0.5mm pitch quad flatpack!

OK, OK, so you can guess that I spent a while closely searching the floor - again!

@graham-xrf think you are making a fuss out of these component sizes. I have used DFN-8 and MSOP-8 parts on adapter boards. Using a stereo microscope so you can see what you are doing, put some tiny dabs of solder paste on the pads. Position the part on top of the paste. It doesn't have to be perfect to work, since the solder surface tension will align the part. A little hot air, and you can breadboard the assembly as if it were a DIP. It really isn't that hard, even with my wobbly hands. The hot air need not be very forceful, in fact it is better if it is gentle. It is surprisingly easy to do. The hot air guns are rather inexpensive and great for this or repair work.

I, like you, would rather work with slightly bigger parts, but alas, they are hard to find. Attempted to source some parts from all over the world, but they were unavailable. Or, they said they had them and later cancelled my order after a month. Adapter boards were my solution. Not elegant, but functional. Like you, I've spent a bit of time on the floor looking for parts that have flown off to who knows where. Maybe 1/2 the time I find them. I continue to look if they are one of a kind.

 

[WobblyHand]

Is this the basic circuit that you (two) have started from?

I presume you are using a different amplifiers? (LTC6869? and ?) Is anyone here sharing schematics, or are you keeping them private? I have used LTSpice before, simulated my Chebychev active filter on it. Did find it didn't do that well for dc offset issues. Had to make changes in situ to get my filter to work. (Saturation from multiple DC coupled stages.) Made two different boards, each one needed tweaks to get them to work.

Have to ask a pointed question, will this basic x-rf machine be open source? If it is, I'm very interested. But it's hard to keep up with you since a lot of details are seemingly hard to find in this voluminous thread. If you could point me to a schematic that is "a rough snapshot in time but mostly works", I'd appreciate it. It would be fun to build up and try it out.

 

[homebrewed]

I've soldered QFN and DFN parts down using a hot plate. As @WobblyHand has mentioned, surface tension does a pretty good job of aligning parts -- it's pretty cool to see the part scoot over and center itself once the solder melts. I now have a toaster oven for the purpose, although I haven't used it for that yet. It was much cheaper than a lab-grade hot plate, and at that point it was my dime, not my company's.

For hand-applying parts to boards, especially ones with leadless packages, I found it handy to add silk screen registration marks around their footprints -- it makes it a lot easier to place them near their final location. That was back when I used expressPCB a lot. They didn't have much in the way of footprints for QFN/DFN packages so I had to create them. After that much effort it was a relatively minor thing to add registration marks, usually at the corners.

Nowadays I use EasyEDA for the PCB design and their associated PCB fab arm, JCPCB, for the boards. Web-hosted and a pretty decent parts library. They have an autorouter but I found it more annoying than useful.

 

[homebrewed]

Is this the basic circuit that you (two) have started from?
View attachment 392587
I presume you are using a different amplifiers? (LTC6869? and ?) Is anyone here sharing schematics, or are you keeping them private? I have used LTSpice before, simulated my Chebychev active filter on it. Did find it didn't do that well for dc offset issues. Had to make changes in situ to get my filter to work. (Saturation from multiple DC coupled stages.) Made two different boards, each one needed tweaks to get them to work.

Have to ask a pointed question, will this basic x-rf machine be open source? If it is, I'm very interested. But it's hard to keep up with you since a lot of details are seemingly hard to find in this voluminous thread. If you could point me to a schematic that is "a rough snapshot in time but mostly works", I'd appreciate it. It would be fun to build up and try it out.

Similar, but not exactly. The starting point is the PocketGeiger, sold by Sparkfun. It uses a 10x10 silicon PIN diode for the detector. The bias voltage is higher, about 25V, generated by an on-board switcher. For about $70 you get the detector and electronics. If bought on its own from Digikey you will pay more than that for just the detector!