Needing more than a spark test?

[WobblyHand]

@WobblyHand
Do tell about the printer. We can maybe benefit from some of your agonizing.
I mean, what model, and how hurtful to the wallet?

The long and short of it was I cried a lot and bought the Prusa i3 MK3S+. You can discover the latest price at: i3 MK3S+ The reasoning behind the final buy was I did not want to make the printer itself become a hobby. Read about a lot of other people, and they required modifications to print ok, and most if not all the modifications are already in the Prusa. It's not the most advanced printer, but it is solid enough, and gets prints done. And right now, that's all I want from it.

Even though I'm retired, I didn't want yet another time consuming hobby, I just wanted to be able to make some things that would be useful to me, like custom electronics housings. I made a housing for my doppler chronograph, and my ELS controller. Both units have touch screen displays. I basically printed directly from my FreeCAD designs. To have made those housings in metal would have been far beyond my machining capabilities. The downside of 3d printing is that it is not particularly speedy. One is building up an object one thin thread of material at a time.

Picture of my ELS box. Not shown are the GX aviation connectors, the cables, or the magnets. You can see the four pockets on the bottom side. I pressed in some 12mm magnets so the box stays upright and stuck to the headstock. I've made two of these boxes now. The second was an easy print - I didn't have to adjust anything - it just printed out, and everything fit. You can export an stl file directly in FreeCAD. The box was printed, then the cover. I used thermal M3 inserts to hold the cover on. Use a soldering iron set to 215C or so and slowly press the insert into the proper size hole. Works great.

I bought the 3d printer kit, hoping to save some money. I did, but it took me a while to build it. By the time I had finished, I thought the assembled unit was a bargain. Nothing was hard, but it took a while. There's a few places that you can go wrong, and you don't know it until sometime later in the process. Then you have to undo some stuff. Personally, there were some steps that are meant for very young nimble fingers, that took me a while to do, or they were mechanically awkward to perform. However, once it was assembled, it wasn't super hard to get going. There's some gotchas in 3d printing, that I stumbled over. But that's pretty much the same as in machining.

You can buy something a lot cheaper, but you will have to put in a bit of time. Classic tradeoff... Prusa's thing is it just works. My experience is they are right.

 

[homebrewed]

I think I found the program-hang problem I was having. After a pulse is processed I was waiting for the incoming signal level to go below the baseline, basically the DC level of the signal. This was to make sure that the pulse processor wasn't triggered by the tail end of another pulse that had come in during the processing. But if any upward drift in the baseline occurred the wait() function would never exit. To fix that, I changed the code so it would exit if the voltage was less than the trigger level, but a better approach might be to either wait until the signal had gone back down to the baseline OR if a long enough time had passed -- so it could never hang.

 

[homebrewed]

Here are two photos of spectra I got today. In both cases I had placed a stainless steel sample in front of the detector. The first was obtained using 15V bias on the detector, and the second with 35V on the detector.
Although the differences are subtle, it looks to me like the lower-bias result actually has more spread than the 35V result. None of the acquisition parameters were changed to acquire these. What do folks think?

15Vbias:


35V:

 

[homebrewed]

It also appears that the peak number of counts in the maximum MCA bin was a little higher for the 35V bias case. The voltage output of my bench supply for Vbias maxes out at 40V so I can try that, too. The DS indicates the maximum reverse voltage is 50V but that probably is one of those guaranteed-by-design specs. So plus/minus there. I wouldn't be surprised if it actually is much higher, but.... it wouldn't be a good idea to push the detector into breakdown. Even a short time in that situation could drastically alter its noise characteristics. In case anyone was thinking about using it as an avalanche detector. So-called "hot carriers" can be stuffed into the dielectric where the breakdown occurs and the resultant fixed charge and charge traps are known to cause big-time problems for things like bipolar transistors.

As in: significant beta degradation, particularly at low collector currents; and increased current noise. Elevated temperature baking after exposure to reverse current can help but in my experience transistors abused in that fashion never completely recover.

 

[graham-xrf]

It also appears that the peak number of counts in the maximum MCA bin was a little higher for the 35V bias case. The voltage output of my bench supply for Vbias maxes out at 40V so I can try that, too. The DS indicates the maximum reverse voltage is 50V but that probably is one of those guaranteed-by-design specs. So plus/minus there. I wouldn't be surprised if it actually is much higher, but.... it wouldn't be a good idea to push the detector into breakdown. Even a short time in that situation could drastically alter its noise characteristics. In case anyone was thinking about using it as an avalanche detector. So-called "hot carriers" can be stuffed into the dielectric where the breakdown occurs and the resultant fixed charge and charge traps are known to cause big-time problems for things like bipolar transistors.

As in: significant beta degradation, particularly at low collector currents; and increased current noise. Elevated temperature baking after exposure to reverse current can help but in my experience transistors abused in that fashion never completely recover.

Regarding persuading a reverse biased diode to get "more gain" before the transimpedance amplifier. Here we are not talking about straightforward diode avalanche breakdown. A true avalanche photodiode is the solid state equivalent of of a PMT tube. It can have gain 1e4 to 1e6 before encountering the amplifier. They have a couple of new types of noise generation.

The X-100-7, I think, is not one of that kind. It takes about 100V at least to get silicon there, and more usually, around 180V, but specially made ones might use over 1kV. I suppose you could explore the breakdown voltage in the usual way using a high value series resistor, but I would not do it to the only one I managed to bring into the country at way more cost than one might have thought!

Working with avalanche photodiodes is tricky, needing a voltage driven an active control loop to keep it stable. I have been learning most from Wikipedia. Although I know they are used in "fluorescence spectroscopy", it's hard to find actual examples and info. The majority use seems to be in optical communications. In Geiger mode, a single photon can cause an avalanche pulse of 1e8 carriers. The whole technology is about extreme speed and low transit times and use with pulsed laser diodes.

I read that in recent technology (2020), graphene layers can prevent degradation.
--> https://phys.org/news/2020-07-electron-sources-graphene.html

Way way back in this thread, I was looking about for one of these, until you showed us the Pocket Geiger robbery recycle plan.

 

[homebrewed]

SiPM's apparently consist of a whole bunch of parallel-connected diodes, each of which has an integrated current-limiting resistor. At least, that's the impression I get from looking at data sheets and manufacturer literature. It would be nice if they weren't so expensive, considering their rather simple construction; but there are only a few folks that make them so they get to charge whatever the market will bear.

 

[WobblyHand]

My Pocket Geiger is to arrive on Saturday. Is C16 (4700pF) really hanging off the end of U3B (LM393) voltage comparator?

Edit: What is the part U4? It claims it is an LT1651, but that doesn't seem to be an active part, anywhere. Ah, an LT1615 seems to have the same outline and function. Jeesh. I think according to the numbers, the switcher would supply about 13.5V to the NJM regulator. Not positive if that is a doubler, or tripler. Tripling the voltage would make it about 39V, give or take. That's a nice high bias voltage. What's interesting, or rather sad, is there's no information, at least on the Sparkfun site on how this works, or even how it is wired. What is the function of the JP2 block? What is the baseline configuration?

It would seem the top comparator threshold is 3.089V, this is "pulse". The bottom comparator threshold is 2.686V, this is noise. I understand the comparator outputs won't be used in our application, but it would seem they are using these "signals" to validate the pulse.

 

[WobblyHand]

Attempting to install LTSpice on my linux laptop. Used to have it before upgrading the OS. Installed via wine. It started the install and it offered to do an upgrade. This is as slow as Windows upgrades of yore... Downloading countless library files at 10KB/sec from the website... Feel like I have an old modem.

 

[graham-xrf]

My Pocket Geiger is to arrive on Saturday. Is C16 (4700pF) really hanging off the end of U3B (LM393) voltage comparator?

Edit: What is the part U4? It claims it is an LT1651, but that doesn't seem to be an active part, anywhere. Ah, an LT1615 seems to have the same outline and function. Jeesh. I think according to the numbers, the switcher would supply about 13.5V to the NJM regulator. Not positive if that is a doubler, or tripler. Tripling the voltage would make it about 39V, give or take. That's a nice high bias voltage. What's interesting, or rather sad, is there's no information, at least on the Sparkfun site on how this works, or even how it is wired. What is the function of the JP2 block? What is the baseline configuration?

It would seem the top comparator threshold is 3.089V, this is "pulse". The bottom comparator threshold is 2.686V, this is noise. I understand the comparator outputs won't be used in our application, but it would seem they are using these "signals" to validate the pulse.

Earlier today I worked up a (only in minor ways) ) modified LMC662 model imported from the TI PSpice model into LTSpice. There are various ways folk get imported models to work, including dropping the whole file into the drawing as a comment, but that gets very messy.

I did try offering my trusty photon current generator model at it (quick and dirty), just on the first opamp. I saw the input pulse, and the output seemingly kicked high, and slowly coming down.

This is preliminary stuff! The TI model is a bit simplistic. There is no internal noise model. One would have thought such a standard thing could have been included. It may be possible to just pinch a sub-circuit from some other opamp, and put it into the sub-circuit netlist, and then set it's values to correspond with the data sheet. I accept that a noise model is more complicated than that, because one wants to see the noise plot with frequency over a considerable range.

LTSpice also have an "alternative" file way of gathering a whole series of products into one big file of sub-circuits. I don't quite understand it all yet.There are also a whole bunch of "undocumented" features. There is a lot of great stuff in there, and I have tried some of them out. I discovered A-devices. The help file says they are undocumented because they are subject to frequent change. I think that is a load of bunkum!

The good stuff is in the file here ..

 

[graham-xrf]

I have had a (sort of) try at the LMC662C. I gave it a purist model of a photon current pulse about double the steady dark current.
The 6.0mV of input offset voltage is exactly what it says in the datasheet, and the model.
It lives with about -83mV on it's output from the input offset, and being the inverting input, expects to go negative which it sort of (eventually) tries to do.



So the right thing to do was make the simulation longer. I set it to 400uS. It lets us see the long, slowed down pulse, but I think it a reasonable looking shape starting from the 3nA height input. It's late here, so I won't go exploring my 45pA yet. The output finally gets back to something like ready for another go after about 360nS.


The next move was to give it a gain of 40,000. That's still high for a circuit hoping to perform to it's GBW limit, but I am not even sure the model is right in this respect. I have not had time to verify. The default was 10MHz, and this chip has it at 1.4MHz. So notice the oscillation!



When I altered the compensation capacitor to as much as 6pF, the oscillation stops. This does not need to be done when there is all that 66MEG in there, but also, the gain is not what one would expect just from raw numbers. It does seem to run out of steam!

So back to the amp, with gain 40,000. The oscillation may have stopped, but it retains a lump that just won't go away. It cant get it's output to faithfully follow the current input wave shape. I increase it to 10pF before I decided either the model was not so good, or the opamp was hitting it's slew rate limit, or something.

So how fast can we get it to go?
I changed Rf to 20K, to try for some speed. Cf still needed to be 8pF. We may not actually need to. The asymmetric shape is better represented by a somewhat slower response pulse, but not as far as what happens with Rf=66MEG



I have not yet tried to string up more opamps in a row. It seems to me that if the 400K gain (about) is adopted, with about 6.8pF across it, you get a pulse that is over in about 15uS, which is fine, but it's shape is like a symmetrical pulse, instead of the "lopsided" one modeled from the photon arrival.



This last one is about as fast as it can go. The height of the pulse is 931uV, so nearly a mV. It requires a further gain of about 2000, to make it a full 2V output, but keep in mind the original 2.85nA might be way more than the biggest we ever expect, or it might be less.

About the current input diode model.
One does not need to model the bias, because I simply added in C1 = 85pF, this being the PIN diode capacitance when the bias is -10V. The rest of what is in there came from the datasheet. I gave it a dark current. The voltage source V3 = 0 is my ammeter. I can measure the current pulse going in with it.

The 5.45mV offset is not as high as the 83mV offset that happens if we put 66MEG in there, but it is still enough to saturate anything following that has enough gain. There seems no choice but to use AC coupled stages. Ultimately, we verify a few things, and then try for the whole circuit chain. If LMC662C's are used in other places in this circuit, there is the possibility of continuous oscillation, which would get counted.

If we can get a modified Pocket Geiger gain set to do in practice what we see in simulation, then we may celebrate. Even so, LM662C is not my choice, and nor is the gain distribution. Sure, my psychology is saying "You already spent on the darn LTC6268-10, and a LTC6269 dual, so don't waste them"!

BUT - if the purchase of a Pocket Geiger, after some hacking with some resistors, etc. is all you need to get real deal XRF plots with some resolution in them, that's great!

OK - I burned the candle tooo far tonight. I gotta pack it in.