Needing more than a spark test?

[graham-xrf]

At this point, I am inclined to agree with you.

The current circuit is awkward to say the least. It's not that easy to simply insert the bias with a simple battery, at least not with further surgery. The existing circuit has one side of the PIN diode connected to the op amp which is sitting at 2.9V not 0V. I would need to remove a resistor or diode to properly add in the battery.

Considering that the circuit isn't stable in the open air, it's hard for me to think that I want to spend much more time with it. Wrapping the unit with a grounded foil, and some insulating plastic is just adding capacitance to the circuit. Seems its a miracle that it does anything useful at all.

It would be nice to remove the diode from the PCB and start afresh. I can see why you want to saw it off. I thought, maybe there is some redeeming value in the circuit, let's try it out. Well I did try it.

I have concluded, it's not a very good basis to work with. And yeah touching the diode was not ever going to be good. I was expecting railing from that. There's no way that op amp can deliver the gain of 66Meg. The op amp has to be hitting some over riding limit, like slew rate or something like that. Back to the books, and simulations. Guess I will have to attack why I had issues with LTspice...

You can go some way to doing this, enough to perhaps display a real Am241 signal. Wiring a twisted set of hookup wire from a 9V battery onto one end of D3, or R20, or maybe even C20 gets the bias on. then take that 9V also to some place 9V is supposed to be, like one end of C15, or Ps8, or C1.

Next step is to forcibly persuade U1A to behave, by shorting out two of the three R9, R10, R11, and putting (say) 220K across the last of the three..

At this stage, if you stick a source onto the diode with a little dob of some sort of goop, or tape over it with a tiny offcut of gaffer (I mean "Duck") tape, then you might be able to probe one end of R4, and show us a real pulse.
I will admit, I lost heart before I got there, and decided to salvage only the diode. About that same time, I was checking out PyMCA, and testing how fast I could go bit-banging a Raspbery's GPIO pin. Thus yes - I got distracted!

 

[graham-xrf]

Has anyone got any thoughts on cheap, effective screening boxes or covers? The Harwin product is about £16 for one. It may be enough just to cover over the circuit with a piece of 1mm aluminium sheet, secured at the corners on metal stand-offs. I don't much fancy the wrapped in metal tape. I want to have test points access

Maybe some sort of separation between amplifier/ADC and the rest. Perhaps a separate screen over the computer parts. I have to take care here, I want the WiFi and BlueTooth to keep working.

 

[WobblyHand]

You can go some way to doing this, enough to perhaps display a real Am241 signal. Wiring a twisted set of hookup wire from a 9V battery onto one end of D3, or R20, or maybe even C20 gets the bias on. then take that 9V also to some place 9V is supposed to be, like one end of C15, or Ps8, or C1.

Next step is to forcibly persuade U1A to behave, by shorting out two of the three R9, R10, R11, and putting (say) 220K across the last of the three..

At this stage, if you stick a source onto the diode with a little dob of some sort of goop, or tape over it with a tiny offcut of gaffer (I mean "Duck") tape, then you might be able to probe one end of R4, and show us a real pulse.
I will admit, I lost heart before I got there, and decided to salvage only the diode. About that same time, I was checking out PyMCA, and testing how fast I could go bit-banging a Raspbery's GPIO pin. Thus yes - I got distracted!

More surgery required. Worth a try. As is, exposing to Am241 shows nothing. Or rather there is no obvious effect. Pulled R9 & R10. Put a jumper around them. Left one of the 22Meg in place. Removed R20, so no more leaking through D3. Seemed slightly quieter for a few moments, then got back to noisy. Now that the PIN is isolated, I can add a 9V battery between VPD and ground. That should give about 6V bias. (9V - 2.88V). I will use the 9V battery clip that I yanked from the old smoke detector. If I need to lower the gain on U1, I can use R20, which is about 100K. Hope I don't lose it! Found the only "live" 9V battery in the house. It measures 8.9V, which is good enough. Hooked up the 12V main power. It's behaving the same. Noise, centered about 2.94V, with a significant shape to it. Basically it's not working. I will reduce the gain again, eliminating the 22Meg resistor and adding the 100K. But basically there's no functionality. There's no response to the source, except perhaps for the increase capacitance. Over 1cm away - no effect from the source. This circuit is currently inop. Next step replace R11 with a 100K resistor. No difference, the noise is reduced but it looks quite similar. It's not apparently working. Oh well, not unexpected. That's enough for now. U1 isn't apparently behaving. No pulses coming out of R4, just some synchronous garbage, some 60 Hz and some 20 mV noise with spikes.

Ahah, a random discovery! I had a cheap wall wart plugged in but not used. It is to charge LiOn batteries. It was making a lot of the synchronous trash showing up on the scope, now it is a lot quieter. Now to unplug more of that kind of stuff... Now the RMS noise is at 6mV. There's stuff running around, but it's hard to tell if its real or not. Looks like 100 MHz. I am getting no evidence of pulses using 100K. I am AC coupled on R4 at 20mV/div with the scope set to 200us/div. The source is resting on the diode. Single sweep trig set to 33mV, so it won't auto trigger. In about a minute I get a trigger. Sadly it looks like nothing that came from the front end. Guess there's tomorrow.

 

[graham-xrf]

More surgery required. Worth a try. As is, exposing to Am241 shows nothing. Or rather there is no obvious effect. Pulled R9 & R10. Put a jumper around them. Left one of the 22Meg in place. Removed R20, so no more leaking through D3. Seemed slightly quieter for a few moments, then got back to noisy. Now that the PIN is isolated, I can add a 9V battery between VPD and ground. That should give about 6V bias. (9V - 2.88V). I will use the 9V battery clip that I yanked from the old smoke detector. If I need to lower the gain on U1, I can use R20, which is about 100K. Hope I don't lose it! Found the only "live" 9V battery in the house. It measures 8.9V, which is good enough. Hooked up the 12V main power. It's behaving the same. Noise, centered about 2.94V, with a significant shape to it. Basically it's not working. I will reduce the gain again, eliminating the 22Meg resistor and adding the 100K. But basically there's no functionality. There's no response to the source, except perhaps for the increase capacitance. Over 1cm away - no effect from the source. This circuit is currently inop. Next step replace R11 with a 100K resistor. No difference, the noise is reduced but it looks quite similar. It's not apparently working. Oh well, not unexpected. That's enough for now. U1 isn't apparently behaving. No pulses coming out of R4, just some synchronous garbage, some 60 Hz and some 20 mV noise with spikes.

Ahah, a random discovery! I had a cheap wall wart plugged in but not used. It is to charge LiOn batteries. It was making a lot of the synchronous trash showing up on the scope, now it is a lot quieter. Now to unplug more of that kind of stuff... Now the RMS noise is at 6mV. There's stuff running around, but it's hard to tell if its real or not. Looks like 100 MHz. I am getting no evidence of pulses using 100K. I am AC coupled on R4 at 20mV/div with the scope set to 200us/div. The source is resting on the diode. Single sweep trig set to 33mV, so it won't auto trigger. In about a minute I get a trigger. Sadly it looks like nothing that came from the front end. Guess there's tomorrow.

Wow! You are tenacious!
Also, you are several hours behind me. I woke up after dozing, and now i have to clear up, and get to bed proper.
Before I do, let me say that what you are trying with an opamp - any opamp, is near impossible. 22MΩ is so high that without some other resistor along with the 22MΩ to help define the gain, you have nearly an open loop opamp. I suppose 22MΩ connected as a feedback resistor can tame it.

I get partly fooled by the junction between R16 and R16 in the divider chain. There are two connections sharing a single blob. I never do that! I always offset them so each shows its own connection separately. Anyway, that R2 and R16 provide the non-inverting connection to 0V+offset, and the offset voltage to allow the whole thing to operate between 0V and 9V.
BUT..
This is where we get to the difference between a TIA and a opamp circuit that will be reasonably well behaved.
Pin1 of U1A, the non-inverting input, is being used as a "shifted" zero signal reference, I think at 2.888V, heavily decoupled by C3.

Without some sort of resistor between the inverting input, and the reference return, to define the gain, it acts as a voltage follower with a very high feedback resistance. Now suppose that "resistor" was replaced by a photodiode, and the other end effectively grounded by a huge 1uF capacitor (C4). Any little "current" that comes out of the photodiode has no place to go except up the inverting input, which is an incredibly high input impedance with a consequence. The amplifier output will adjust the voltage so that that current can go up the feedback resistor. The voltage it makes is the input current x the Rf value.

So - if the output gets to as much as (say) 2mV, with Rf = 22MΩ, that was done with a input current of 90pA !
With another gain x100 in U1B, you are looking at 200mV. That is a decent size to see on a scope.

Getting something like that to actually happen, to actually measure, is likely to be difficult.
Mark has managed this. He has shown us scope pulses that look something like we expect, and now he is measuring and plotting them.
You can see why I feel I need to catch up, and verify what is going on with a TIA, and a Am241 source

 

[WobblyHand]

You are tenacious!

Perhaps bullheaded. Or stubborn.

Anyways, I honestly see no effect of the source. I do see some effect of local capacitance, even getting my hand 3-4 cm away. At first I thought that was due to the source, but I was mistaken, it's just my hand capacitance.

I will fiddle with this some more tomorrow. Will drag out the microscope, to make it easier on me to solder stuff.

 

[graham-xrf]

Perhaps bullheaded. Or stubborn.
Anyways, I honestly see no effect of the source. I do see some effect of local capacitance, even getting my hand 3-4 cm away. At first I thought that was due to the source, but I was mistaken, it's just my hand capacitance.

This is why we need to start at the source.
For all I know, the "cheapo" sources I bought might all be counterfeit.
Smoke detector innards without the rest of the gadget, a mass produced item, is an "unusual" item anyway.
[Dark thought.. Mark might be the only guy with a "genuine" one].
[Even darker thought .. the low count rate might be because he only has one or two "genuine ones" ]

I wonder if a source will do anything to a piece of photographic print paper, or film? I have some of that stuff in the loft, but no developer.

I have had the email from Aliexpress.com. It seems my stuff has "been shipped". It has to be the glittering vibrating nuclear detector
"Shipped" is probably literally true, on the ocean wave, and all that, so it will be "a while" (as defined by Penny in "The Big Bang Theory" comedy).
------------------------------------------------------
How we do this?
Everything we experience in trying to start up, and convincingly test a TIA amplifier like this, including approaching hand capacitance, etc. will be happening again, regardless one is just messing with a Pocket Geiger, or starting up one's own newly assembled pride and joy own design. Mark has a totally shielded aluminium box, and I guess something like that is needed.

I think I have to figure out some test arrangements. The trickiest is to find a way to introduce a tiny input that is not also just a huge resistor noise.
Since the Pocket Geiger is sensitive to hand capacitance, don't offer it the hand. Instead, through maybe a few pF of old-school thin wire ended capacitor, on the end of a tied down coax, and then use a signal generator, or some little anything making pulses, attenuated down, of course.

Very likely, just getting a physical probe onto any places of interest, is going to be awkward, even though I intend there to be available test points tracked out.

Perhaps some thin 2mm coax, or just using twisted pairs in a piece of network cable might do.
Probably, when trying it out, it may be necessary to put a piece of grounded oven cooking foil over the whole thing - damned inconvenient!

Worry about busting the input by static charges. Tera-ohms inputs with tiny FET gates won't take much of a transferred charge from a hand coming too close, to get to internal voltages that can cause damage.

 

[WobblyHand]

For testing I tacked on some 30awg wire to the A test pads. I twisted the signal and ground wires.

I'm wondering if I damaged the front end or not. It's still goes into square wave mode if I get too close, so it does work on a gross scale. Will drag out the stereo microscope and look over my handiwork. Maybe there's a solder sliver. I think it's a marginal design and I'm attempting to run the circuit outside of its margins!

To my dismay, I have discovered that I don't have many 0402 parts. To do much work on this PCB I'll need to get a kit. I suppose I can use them later. Haven't had to build up too many SMD boards at home. At work we regularly used 0402 and 0201s. Times are a changing. Can't get the big packages any more, just the tiny stuff.

Well, think I will try to get back to Spice to see if I can get something to work at all. I'll build up something similar to the @graham-xrf model and see if I can get it to work. Also hope to trawl this thread and see if I can figure out what Mark did. I need to see the schematic all on one sheet so to speak, so I see the whole picture in my mind. Might be missing something.

 

[graham-xrf]

For testing I tacked on some 30awg wire to the A test pads. I twisted the signal and ground wires.

I'm wondering if I damaged the front end or not. It's still goes into square wave mode if I get too close, so it does work on a gross scale. Will drag out the stereo microscope and look over my handiwork. Maybe there's a solder sliver. I think it's a marginal design and I'm attempting to run the circuit outside of its margins!

I accept you may not want to do this. For the Pocket Geiger, it's like 23:59 in the Last Chance Saloon!

Perhaps find a way to disconnect the X100-7 (D1).
That will leave you a voltage follower, albeit with 22M in it's feedback, but it is, in theory, a follower, so long as you stay away from it's non-inverting input. Therefore, short out R11. Put a link across it, and you have a real follower - with gain of only 1
That should drastically slash the high gain effects, forcibly going for better behavior, and allowing checking rails, and offsets.

The voltage at the non-inverting input should start (107.5/335) x your battery voltage.
If 9V battery, then expect 2.88V. That voltage had better also be on the link ex-R9, R10, R11, to high accuracy. It will let you know whether the first opamp is likely working.

Then check the output of U1B, seen on one end of C7, or R6, or R4, or even on JP2 connector Pin$4.
See the very same voltage 2.88V.

That single IC, both opamps, was supposedly delivering the entire linear part of the gain of the Pocket Geiger, to end up with what is shown in their blurb, (here attached for convenience). I can't tell whether the pulse shown was a slugged comparator output or not.

For us, we can now remove the short across the 22M, or better, temporarily let it become 100K or so be in one of those R9, R10, R11 places.
Also temporarily, we can now apply a signal through a series 10K or so to the inverting input, which is handily on where R9 is (or was).
Even the connection to the diode might be used. It may not matter if D1 is in circuit not. The point is, we can put in a test signal via the 10K, giving the stage a gain of about 10. Yes, it might oscillate, but I think not.

We can now increase, either the input signal, or the gain by altering the value of R11, exploring how far it gets before the output on the connector crashes the rails, and gives us fat square waves. If it's broke, we would know by now.

Well, think I will try to get back to Spice to see if I can get something to work at all. I'll build up something similar to the @graham-xrf model and see if I can get it to work. Also hope to trawl this thread and see if I can figure out what Mark did. I need to see the schematic all on one sheet so to speak, so I see the whole picture in my mind. Might be missing something.

I don't think you missed anything.
The Packet Geiger did take that signal to the JP2 Pin $PS4. Other than for testing, I don't think it was used.
In actual operation, I don't think the U1 outputs were driven to the rail.
Pocket Geiger used the bang-bang outputs from the comparators.

Of course, a comparator is like an open loop opamp. Its output will be on one rail, or the other.
In effect, it adds a huge gain. If the signal crosses the threshold at at all, either direction, the comparator flips!
This kind of "gain" may be OK for counting events, but is useless to us for gauging the original pulse energy.

The size of the pulse flipping the comparators might be quite small. For using it to measure energy, it might need to be bigger. That would be the reason more gain needed to be applied, as was done on Mark's signal conditioning board.

Incredibly, there is a 4700pF (C16) across the output of U3B. When I saw that, I thought the designers must know something I did not, or else it's a mistake in the drawing. Maybe the physical circuit has the capacitor on the other side of R18. If they wanted to shape the pulse, that would be the way.

Mark's signal conditioning board.
That one added the more linear gain, with options to change the gain in the path. We have the circuit he posted.

Yes indeed, simulation is always my first stop. It lets one get all sorts of stuff right, before tangling with the hardware.

 

[WobblyHand]

Digging through the TI website, the datasheets for the LM662, some app notes and some trans-impedance amplifier stuff, I am finding the recommended feedback capacitance for stability, relies on a particular bias. I redid the recommended values and they come out remarkably close to the values on the board. The Pocket Geiger is using a bias of ~ 33V which lowers the capacitance of the diode. I guessed it was about 50pF. Running through the numbers gives me about 0.8pF for the capacitance for the amplifier - which is close to what they used. (1pF) Using my microscope I also saw cuts in the top layer between the pads of the 22Meg resistors, in an effort to reduce the capacitance. These cuts were fabricated and made during the PCB manufacture. Kind of obvious under the microscope - at first I thought they were silkscreen, but they are slots in the FR4.

My bias of only 6V across the diode, yields a much high PIN diode capacitance, so I need a much higher compensating capacitance. I estimate I need about 20pF or so. Actual number came out to 18.9pF. Interesting. Think I will increase the bias further.

I was mistaken about the 0402's. The resistors seem to be 0603's. I found some a couple of 0402 parts, they are quite a bit smaller. No wonder I could solder them, they are the big ones! I don't have a stockpile of 0603's either, probably need to buy a kit or two, R's and C's.

 

[WobblyHand]

More limitations of this amplifier. The slew rate. The LMC622 has a slew rate of 0.8V/us for a 3.3Vpp output, this implies a full power bandwidth of about 67KHz, or 15us. I'd expect we don't see the full pulse if it is shorter than 15us due to the slew rate limit. Or its really distorted. Ok, back to figuring out a circuit model... For the moment, I am not going to play with physical circuits.

Back to LTspice. What ever you do don't open the model of the generic op amp. LTspice froze and was totally unresponsive when I clicked on level1. I had to kill the process as in $ kill -9 pid. Lost my schematic I was entering. This is going to be a tough slog it seems. Maybe it isn't an LTspice kind of a day...