Needing more than a spark test?

[homebrewed]

In making a signal for the ADC, would not clipping at 2.5V be considered a "desirable" feature? My whole circuit deliberately uses 2.5V just so that is guaranteed to happen.

The problem is that the amplifier output was going nonlinear well before it hit 2.5V. I have a new spectrum for aluminum that looks much better, currently getting one for iron to compare. I'll post the results as soon as the iron spectrum is done. They both will be plots of the fitted peak voltage, not the pulse area.

 

[WobblyHand]

The problem is that the amplifier output was going nonlinear well before it hit 2.5V. I have a new spectrum for aluminum that looks much better, currently getting one for iron to compare. I'll post the results as soon as the iron spectrum is done. They both will be plots of the fitted peak voltage, not the pulse area.

I agree that clipping is bad, bad, bad. Generates all sorts of problems. Better to back down the gain. Clipping will result in the pulses not being properly integrated and sorted into the right bins. We want the highest gain that stays linear and is scaled to the ADC input.

 

[graham-xrf]

I agree that clipping is bad, bad, bad. Generates all sorts of problems. Better to back down the gain. Clipping will result in the pulses not being properly integrated and sorted into the right bins. We want the highest gain that stays linear and is scaled to the ADC input.

Yes - linearity to Hi-Fi standards and more!
Of course we need pulses that are not clipped. They cannot retain the information to allow discrimination of one to the next.
The energy analogue is the area under the waveform curve. Take enough samples in vertical stripes, and add them up.

More practically, add them up, and apply a scaling factor to normalize the value when one knows what a maximum for that gain range is, meaning calibration for a known element. Thus, if the gain gets crushed from a non-linearity, the area count is telling lies.

Related to being able discriminate between elements are present, these must include scaling those with energies outside of the 6KeV to 10KeV range, essentially to adjust their counts according to the inverse probability of their diode detection. I mean the inverse of the X-100-7 diode absorption probability curve.

We should not be facing even very small non-linearities. We can do very accurate analogue computing at high speed even to missile navigation accuracy, and it can be done with low cost components.

I am only OK with adding up the low-pass filtered, delayed, "stretched" pulse waveform if we know for sure that the entire waveshape analogue-of-area waveshape is preserved. I have not (yet) done the integration exercise to mathematically prove it. Right now, I am not confident that the output of U1 does that even once, even if we change the opamp for something else. I would be happier if it was a linear TIA amplifier, instead of an strange integrator "pulse stretcher". We then AC couple, and go again with U2, but that one has a response roll-off much higher, near 16KHz, so it only causes a little delay, and hopefully, no distortion.

Then we get to whatever further gain we apply. I suppose the passive high-pass AC coupler can't really hurt an already slow pulse. Mark has already caught up with and fixed the clipping, but going non-linear before getting to 2.5V is a problem. We should be able to compute exactly to within (say) 100mV or less of the rail, with no distortion.

I know I must be sounding like an electronics Devil's advocate here, and please forgive that I express misgivings. It's only because the total design case for the TIA amplifier à la mode de Pocket Geiger has not been made in a way that I can understand, and that may be a problem only for me.

 

[WobblyHand]

It's only because the total design case for the TIA amplifier à la mode de Pocket Geiger has not been made in a way that I can understand, and that may be a problem only for me.

You are not the only one. This design doesn't strike me as very sophisticated. In many ways, I'm surprised it works. It's certainly not a classical TIA circuit. However, as most consumer goods, it is designed to a price point.

Pretty sure it kind of works. Think changes are required for our application. Can't say I'm surprised, since our application was not part of the original Pocket Geiger design.

 

[WobblyHand]

Well 2 square feet of 1/16" lead showed up. Was surprised because it was ordered from eBay and delivered by Amazon. Even more surprised when I got two packages from Amazon, when I thought I'd get one. When I picked up the box, I thought, gosh this is heavy, what the heck did I order from Amazon that would be this heavy? Pretty small box and quite heavy - yeah lead. About 8lbs, the inner box is 2"x2"x12". Other box? 3d printing stuff, assorted nozzles, and 1kg of black PETG.

 

[graham-xrf]

You are not the only one. This design doesn't strike me as very sophisticated. In many ways, I'm surprised it works. It's certainly not a classical TIA circuit. However, as most consumer goods, it is designed to a price point.

Pretty sure it kind of works. Think changes are required for our application. Can't say I'm surprised, since our application was not part of the original Pocket Geiger design.

It works as a Geiger event detector, banging the LM393 comparator outputs up and down if a "pulse" traverses the levels set by the divider chain.

The comparator levels are 2.888V and 3.089V. The positive going waveform out of U1B, if started by a 2nA pulse from the diode, passes the U3B comparator 3.089V at about 14.5uSand makes it's output switch to negative (to 0V). That's a "the pulse has started" signal.

On the long slope down, the signal passes the 2.888V level, and makes U3A fire up a pulse up to positive (rail?) I think that would be" signal, but Pocket Geiger a "pulse is over, but the literature calls the negative pulse (SIG) for detecting radiation, and the other is (NS) for noise detection. The waveforms shown in the write-up do not happen at the levels shown. Regardless all about the LM393, you don't need it.

I know you did not get the simulation to work, and I am not sure why, but what you might do is keep the power supplies, and the artificial photon pulse maker, and choose some op-amp from what is already in their library. Pick an opamp from Analog Devices, or Linear Tech, that has a GBW of about 3.5MHz, and similar properties to LMC662. Alternatively, the plots I posted were for the LMC662 may do.

I truly don't want to be awkward, and right now, I am trying to catch up with the real hardware. This morning I payed near £25 for a ebay roll of Code 4 (1.8mm) lead flashing, including P&P. It's 203mm wide (8") and 1 metre long. That's more than I need, but some is to get melted, and some tries will have mistaikes. Tomorrow, I get done with my Mouser order for the charge pump voltage regulator, the notch filter option, and a bunch of other parts.

 

[graham-xrf]

Microphonics?
In the Pocket Geiger blurb, it says the NS pulse on the Pocket Geiger is a positive pulse signalling vibration noise.

I recall reading about how much piezo-type racket can be made by X7R capacitors that feel vibration on a PCB. I will be trying to find out about "microphonics" messing with the signal. You can imagine, with the very high gain, sound and vibration wiggling on a capacitor, especially like a coupling capacitor, might be an annoyance!

 

[WobblyHand]

No worries on the simulation. I will simply print out the schematic and enter it in by hand. It won't take that long. Must have done something to mess up this file. My other LTspice sims are unaffected, so that's good.

Been busy with a few things which are peripherally related. I umm upgraded to the Arduino IDE 2. It has caused me a lot of grief, because it seems the compiler has new settings and it is a lot less permissive than it used to be. Kind of enforcing good coding, I'd guess. My existing 2300 LOC ELS project doesn't compile error free under IDE 2, and it did just a few days ago under IDE 1. So trying to slowly fix that. I have everything backed up and a full IDE 1 running on an RPI4, so it isn't the end of the earth, but I sure would like to get it fixed. IDE2 will gladly update all your libraries for you. If you have a lot of code in IDE1, umm, don't do that. Do that later, manually, one at a time, and fix what got broken... I have had to make dozens of changes, and it still doesn't compile error free. And then I have to test it all on my lathe. Been a real pain in the neck.

I will try forming this lead sheet using my plastic die set. It might work, it may not. I expect to learn from it though. If I get lucky maybe I can even form a pocket. Might have to reprint the die set to make it easier, with shallower curves. If this doesn't look promising then I will design an aluminum die set for casting. That will be machined conventionally. I don't have a casting set up right now, but will probably get one if I make the die.

 

[WobblyHand]

Microphonics?
In the Pocket Geiger blurb, it says the NS pulse on the Pocket Geiger is a positive pulse signalling vibration noise.

I recall reading about how much piezo-type racket can be made by X7R capacitors that feel vibration on a PCB. I will be trying to find out about "microphonics" messing with the signal. You can imagine, with the very high gain, sound and vibration wiggling on a capacitor, especially like a coupling capacitor, might be an annoyance!

I don't know what that is about, but that is why I want to measure the existing circuit to see what's going on. I have to figure out my power supplies first though, as I just removed the switcher. If push came to shove I could reinstall the switcher, but I don't want to.

Capacitors can be replaced with something less microphonic, right? They might be bigger though.

 

[graham-xrf]

I don't know what that is about, but that is why I want to measure the existing circuit to see what's going on. I have to figure out my power supplies first though, as I just removed the switcher. If push came to shove I could reinstall the switcher, but I don't want to.

Capacitors can be replaced with something less microphonic, right? They might be bigger though.

As soon as you get the signal chain powered, and you are sure stuff happening at the beginning is making a racket at the end, you can test this aspect out for us. I suppose showing the Pocket Geiger a smoke detector might work to let you know it is alive, but then try for as quiet as possible.

Monitor the analogue output.
You may have a pile of unshielded pickup signal to contend with, but one might try short-circuit the photodiode, and take away its bias.
Then get something reasonably hard, and start tapping on the board, and on the coupling capacitors, and maybe on the chips.

If tapping on them, or on the board, or flexing the board slightly starts making outputs, then we have dielectric microphonic effects. We may have to "soft-mount" our final PCB to isolate it from direct hard connection to metal outer casing, choose capacitor types carefully, and such-like, but I am hoping these effects will be minimal. Long ago I discovered the (high) voltages that can come from a piece of coaxial cable getting moved about.