Needing more than a spark test?

[WobblyHand]

The operating voltage for PMTs varies depending on the application and specific tube. One website that uses the same setup I got specifies 650V, which is on the low side -- but has the advantage of lower noise. The current each dynode requires depends on the dynode, due to the gain. It increases for each successive dynode, so the divider chain also has capacitors across the last few resistors in the divider chain to reduce the voltage drop.

The Theremino approach recommends 10 megohms for each resistor in the chain (except for a 20 Meg at the cathode end) for a total of 120M. This is to permit the use of a HV supply that only needs to provide a few micro-amps -- it reduces the cost, and also is a lot safer.

I have attached a drawing I found online that shows how the PMT is wired.
View attachment 491013
It's dirt simple compared to the TIA schemes we've talked about using for the silicon PIN diode detector, but that's mostly due to the high gain of the PMT.

So if the PMT device consumes no current, the string of resistors has about 5.4uA through them at 650V. The behavior you are seeing is like a cap is charging. Maybe the resistors with the caps are open or a wrong value? For a clue, you could plot current vs time or something like that? What are the resistors mounted to? Standoffs or a PCB? Is it possible that there are leakage paths that are interfering?

Kind of tough to measure 120Mohms. But can you measure the 10 and 20 Mohm resistors insitu? Remove the supply and short out the caps first! I know you know that, but this is in case someone reads this in the future.

HV is strange stuff, have to be careful, or you can get bitten badly. HV dielectrics can have some memory. Keep shorts across the terminals until use. A shorted cap can appear to spontaneously have a voltage on it after being discharged. I have some energy storage caps, they are always shorted because I've seen this behavior. There can be enough residual charge to fry you in some cases. If I recall correctly, 100mJ can stop your heart. I have a 3 200J caps waiting for a whacko project of mine. I always run the calculations on caps to see how much energy they store, and behave accordingly. Energy = 1/2 * C * V*V, where C is in Farads and V is in Volts. Units of Energy are Joules or watt-seconds. If you have lots of volts, it doesn't take much capacitance to become a problem.

That was a public service announcement. I get the feeling that you've dabbled in this stuff before, so my comments aren't directed to you. If I had my act more together, I'd be playing with HV again. I had made an HV supply and multiplier but it appears that it's lost to the ages. Pity, since it had a tube color TV flyback transformer in it. Those guys are practically impossible to find any more. I used a 1B3GT vacuum tube for the rectifier. The filament was run by 2 turns on the flyback transformer. Somehow it would self start. Everything was submerged in oil to prevent corona discharge. Got about 50kV out of it. It could go higher, but the rectifier started to produce X-rays. I could detect the onset by the fact that the inner surface of the glass was starting to fluoresce. At that point I stopped and waited for HV solid state rectifiers to come down in price. Eventually I got some, but they leaked too much, which makes your power supply bigger. Maybe someday I'll play with that some more.

 

[homebrewed]

So if the PMT device consumes no current, the string of resistors has about 5.4uA through them at 650V. The behavior you are seeing is like a cap is charging. Maybe the resistors with the caps are open or a wrong value? For a clue, you could plot current vs time or something like that? What are the resistors mounted to? Standoffs or a PCB? Is it possible that there are leakage paths that are interfering?

Kind of tough to measure 120Mohms. But can you measure the 10 and 20 Mohm resistors insitu? Remove the supply and short out the caps first! I know you know that, but this is in case someone reads this in the future.

HV is strange stuff, have to be careful, or you can get bitten badly. HV dielectrics can have some memory. Keep shorts across the terminals until use. A shorted cap can appear to spontaneously have a voltage on it after being discharged. I have some energy storage caps, they are always shorted because I've seen this behavior. There can be enough residual charge to fry you in some cases. If I recall correctly, 100mJ can stop your heart. I have a 3 200J caps waiting for a whacko project of mine. I always run the calculations on caps to see how much energy they store, and behave accordingly. Energy = 1/2 * C * V*V, where C is in Farads and V is in Volts. Units of Energy are Joules or watt-seconds. If you have lots of volts, it doesn't take much capacitance to become a problem.

That was a public service announcement. I get the feeling that you've dabbled in this stuff before, so my comments aren't directed to you. If I had my act more together, I'd be playing with HV again. I had made an HV supply and multiplier but it appears that it's lost to the ages. Pity, since it had a tube color TV flyback transformer in it. Those guys are practically impossible to find any more. I used a 1B3GT vacuum tube for the rectifier. The filament was run by 2 turns on the flyback transformer. Somehow it would self start. Everything was submerged in oil to prevent corona discharge. Got about 50kV out of it. It could go higher, but the rectifier started to produce X-rays. I could detect the onset by the fact that the inner surface of the glass was starting to fluoresce. At that point I stopped and waited for HV solid state rectifiers to come down in price. Eventually I got some, but they leaked too much, which makes your power supply bigger. Maybe someday I'll play with that some more.

I have experienced the recovery effect that HV capacitors can exhibit. By being bitten by it. Fortunately, the capacitor wasn't that big in terms of capacitance. The big caps used for things like crushing cans are a different story. If I saw one lying around without shorting wires on it I'd stay well clear of the thing, it could be a lurking death trap.

Long back I made a HVDC supply using a flyback and vacuum tube diode, much like you. I didn't do much with it other than make some ozone. I had plans for making my own electrostatic dust precipitator but the voltage was TOO high. Even back then I knew that ozone was bad news.

Back to my PMT/HV problem, I opened up my PMT and eyeballed it very carefully. No obvious problem. All the SMT resistors and caps are on a two-sided PCB, which is soldered to the PMT pins. I ohmed out the connections from the PMT pins to resistors and they were OK, too. I think my DVM can measure up to 10 meg so I can check the individual resistors. The thought occurred to me that perhaps I didn't properly wire up the RG-58 to the output of my HVPSU so I made a 1000:1 attenuator using some through-hole 10 Meg resistors. I haven't had an opportunity to test it yet.

Another thought I had was perhaps the BNC cable I bought for the connection between my detector and processor board was defective -- it was an inexpensive one I bought off ebay. But it ohmed out OK.

There's a relatively easy way to measure high-value resistors w/o using a high voltage. It's a simple form of SMU (source-measure unit) I made for measuring low-value leakage currents on the internal nodes of integrated circuits. It uses one op-amp.

It looks like this:

The feedback resistor R1 performs two functions -- it forces "V" volts on the inverting input, and the current flowing into the DUT appears as a voltage on the output. The voltage equals V + I(DUT)*R1. The DVM is connected to common-mode out the forcing voltage V, and the connections are reversed so + volts in equals a positive reading on the DVM. As it is, it would be easy to measure 10 megohms. To measure 100's of megohms the feedback resistor would need to be higher. But opamps with fA bias currents are relatively easy to come by these days so that's not a big deal.

Measuring capacitor leakage current might not work because there's a good chance the circuit would oscillate.

 

[homebrewed]

I have experienced the recovery effect that HV capacitors can exhibit. By being bitten by it. Fortunately, the capacitor wasn't that big in terms of capacitance. The big caps used for things like crushing cans are a different story. If I saw one lying around without shorting wires on it I'd stay well clear of the thing, it could be a lurking death trap.

Long back I made a HVDC supply using a flyback and vacuum tube diode, much like you. I didn't do much with it other than make some ozone. I had plans for making my own electrostatic dust precipitator but the voltage was TOO high. Even back then I knew that ozone was bad news.

Back to my PMT/HV problem, I opened up my PMT and eyeballed it very carefully. No obvious problem. All the SMT resistors and caps are on a two-sided PCB, which is soldered to the PMT pins. I ohmed out the connections from the PMT pins to resistors and they were OK, too. I think my DVM can measure up to 10 meg so I can check the individual resistors. The thought occurred to me that perhaps I didn't properly wire up the RG-58 to the output of my HVPSU so I made a 1000:1 attenuator using some through-hole 10 Meg resistors. I haven't had an opportunity to test it yet.

Another thought I had was perhaps the BNC cable I bought for the connection between my detector and processor board was defective -- it was an inexpensive one I bought off ebay. But it ohmed out OK.

There's a relatively easy way to measure high-value resistors w/o using a high voltage. It's a simple form of SMU (source-measure unit) I made for measuring low-value leakage currents on the internal nodes of integrated circuits. It uses one op-amp.

It looks like this:
View attachment 491165
The feedback resistor R1 performs two functions -- it forces "V" volts on the inverting input, and the current flowing into the DUT appears as a voltage on the output. The voltage equals V + I(DUT)*R1. The DVM is connected to common-mode out the forcing voltage V, and the connections are reversed so + volts in equals a positive reading on the DVM. As it is, it would be easy to measure 10 megohms. To measure 100's of megohms the feedback resistor would need to be higher. But opamps with fA bias currents are relatively easy to come by these days so that's not a big deal.

Measuring capacitor leakage current might not work because there's a good chance the circuit would oscillate.

I think I made an error in the polarity of the connections to the DVM.....

 

[homebrewed]

I found out why the PMT voltage divider isn't drawing any current. One of the 10M resistors is missing. It looks like there perhaps never _was_ a resistor there, which is why I didn't swap in a replacement. If that's the case my PMT/scintillator may be the equivalent of NOS.

I found the problem by measuring current, rather than ohms -- my DVM maxes out at 9.99 megohms. I used one of my bench supplies set to its maximum output -- 15V -- and worked my way through the PMT pins, dynode-to-dynode, until I found an open.

 

[WobblyHand]

I found out why the PMT voltage divider isn't drawing any current. One of the 10M resistors is missing. It looks like there perhaps never _was_ a resistor there, which is why I didn't swap in a replacement. If that's the case my PMT/scintillator may be the equivalent of NOS.

I found the problem by measuring current, rather than ohms -- my DVM maxes out at 9.99 megohms. I used one of my bench supplies set to its maximum output -- 15V -- and worked my way through the PMT pins, dynode-to-dynode, until I found an open.

Glad you found it! Yeah, an open circuit won't help. All the nodes would be at the same voltage effectively as well, which won't bias the dynodes correctly. Hopefully you should be getting some signal now.

 

[homebrewed]

Glad you found it! Yeah, an open circuit won't help. All the nodes would be at the same voltage effectively as well, which won't bias the dynodes correctly. Hopefully you should be getting some signal now.

I am! And it's FAR cleaner than anything I saw from the PIN detector:



The undershoot indicates that my pole-zero cancellation circuit element(s) need to be changed. I'm using it because the Theremino folks claim it improves the energy resolution.

BTW for the longest time I thought my HVPSU had some serious overshoot at turn-on but it was an artifact of my DVM's range-change as the voltage ramped up. The power supply is very well-behaved in that regard.

The pulse amplitude is a little low for getting the best results out of my outboard ADC, but once I address the pole-zero cancellation issue I will try using my MCA S/W to see what I get. Almost there.....!

 

[homebrewed]

Well it looks like simulations and real-life don't agree, in terms of the pole-zero cancellation stuff. At least, not with my design. Which is odd because it's very similar to the Theremino approach. My simulation results are pretty close to what the Theremino design is claimed to get.

I'm starting to think that it's related to differences between my HV power supply and theirs. They are quite different. I can't directly monitor the HV but I _can_ AC couple variations in the voltage and examine them.

 

[homebrewed]

I found the Pspice model for the Theremino-designed filter-amplifier board, so I could compare my design and simulations to theirs. Pretty much identical. Except for one very interesting bit. Their model includes the input network to a USB sound card. It basically is a high-pass filter, and I observed a real nasty undershoot on that node. So at this point I'm thinking that their pole-zero cancellation scheme isn't worth the space on the PCB!

If they really wanted to eliminate undershoot at the actual measurement point, they blew it right there. Any high-pass A.K.A. capacitor-coupled circuit will exhibit this effect, in the absence of a pole-zero cancelling network. As far as I know the only way to minimize it w/o such a network is to set the high pass filter's time constant to a very large value, compared to the time scale of the signal. And it won't eliminate the undershoot, just reduce it to a low value.

However, neither simulation agrees with what my actual setup is doing. That's troubling.

 

[WobblyHand]

I found the Pspice model for the Theremino-designed filter-amplifier board, so I could compare my design and simulations to theirs. Pretty much identical. Except for one very interesting bit. Their model includes the input network to a USB sound card. It basically is a high-pass filter, and I observed a real nasty undershoot on that node. So at this point I'm thinking that their pole-zero cancellation scheme isn't worth the space on the PCB!

If they really wanted to eliminate undershoot at the actual measurement point, they blew it right there. Any high-pass A.K.A. capacitor-coupled circuit will exhibit this effect, in the absence of a pole-zero cancelling network. As far as I know the only way to minimize it w/o such a network is to set the high pass filter's time constant to a very large value, compared to the time scale of the signal. And it won't eliminate the undershoot, just reduce it to a low value.

However, neither simulation agrees with what my actual setup is doing. That's troubling.

Parasitics, imperfections, etc. Components aren't pure R, L or C. Or some kind of coupling. For the most part, the pulse looks pretty good. Does the existing pulse shape prevent meaningful analysis?

 

[homebrewed]

Parasitics, imperfections, etc. Components aren't pure R, L or C. Or some kind of coupling. For the most part, the pulse looks pretty good. Does the existing pulse shape prevent meaningful analysis?

I think the pulse shape will do for identifying a lot of elements, but I'm not so certain w/regard to separating, say, the iron and cobalt (or manganese) peaks. They are very close together so everything has to be running at peak efficiency (pun intended). The Theremino approach uses deconvolution to improve the effective energy resolution, and I'm pretty sure it will be necessary if we want to identify different alloys.

In terms of device parasitics, the pulses are pretty slow. It's hard to believe that the parasitic L of SMT capacitors would play a role. Or parasitic C of resistors etc. I still think there's something unaccounted for in my setup. If nothing else, it would be good to figure it out so if someone else builds a similar system they will get consistent results. Ultimately the ability to identify different alloys will be a crowdsourced bunch of spectra that have to be consistent enough for all to use & get believable results, so a wonky system could derail that.

It IS true that the current pulses coming out of the PMT are pretty fast. But they are almost immediately slowed down by the relatively large capacitance of the coaxial cable. The Theremino spice model includes an asymmetric pulse with a 300nS risetime and 3uS fall time, probably due to the coax capacitance. The PMT pulls a wad of current out the coax, discharging it pretty fast....and then the 2Meg resistor in series with the HVPSU (slowly) charges it back up.