Needing more than a spark test?

[WobblyHand]

Even a cable in the wall nearby running power to something, makes a magnetic 60Hz that goes right through metal shielding. It's true that as the frequency goes up beyond into RF, with shorter wavelengths, magnetic fields lose the ability to get inside a Faraday cage, but 60Hz is so low, it may as well be DC.

When I get to this, I may well be putting the amp inside a magnetic shield. If not a mu-metal cover, then some toroid cores.

It's likely there is conducted 50/60Hz in the circuitry. Perhaps radiated. Can the whole setup be run on batteries right now? Would eliminate conducted 60Hz and all the power supply noise? I've rarely run into the case where radiated 60Hz is enough to mess things up on a proper PCB. It happens, but not that often. I've seen a lot of people try mu metal and it had little effect. Perhaps your experience is different. Edit: Just reread part of the thread. Yeah, I see where you are coming from...

Still, if I were to bet, it's not likely that magnetic 50/60Hz coupling, is the root issue.

 

[homebrewed]

Even a cable in the wall nearby running power to something, makes a magnetic 60Hz that goes right through metal shielding. It's true that as the frequency goes up beyond into RF, with shorter wavelengths, magnetic fields lose the ability to get inside a Faraday cage, but 60Hz is so low, it may as well be DC.

When I get to this, I may well be putting the amp inside a magnetic shield. If not a mu-metal cover, then some toroid cores.

Am I recalling correctly that you got yourself some mu-metal around the time that a scintillator+PMT was on the table? I know you have a PMT and an NaI(Tl) scintillator so a mu-metal shield is the logical next step.

I bought a sheet of the stuff but was a bit discouraged when I read that forming it to any great extent caused its permeability to drop. The real good stuff is annealed in a hydrogen atmosphere, something a bit dicey for a DIYer. "Machinist blows house up while experimenting with hydrogen furnace"......hmm.

Now, there is a nitrogen/hydrogen mix called Forming Gas. The hydrogen concentration is low enough that it's below the explosion threshold so it's the preferred delivery vehicle for hydrogen in a lab setting. I used it in grad school for some experiments (all actually vetted by the professor who oversaw the lab). No idea how available it is, nor how much it might cost for a tankful. I did the work in 1973 so things may have changed since

 

[graham-xrf]

Am I recalling correctly that you got yourself some mu-metal around the time that a scintillator+PMT was on the table? I know you have a PMT and an NaI(Tl) scintillator so a mu-metal shield is the logical next step.

I bought a sheet of the stuff but was a bit discouraged when I read that forming it to any great extent caused its permeability to drop. The real good stuff is annealed in a hydrogen atmosphere, something a bit dicey for a DIYer. "Machinist blows house up while experimenting with hydrogen furnace"......hmm.

Now, there is a nitrogen/hydrogen mix called Forming Gas. The hydrogen concentration is low enough that it's below the explosion threshold so it's the preferred delivery vehicle for hydrogen in a lab setting. I used it in grad school for some experiments (all actually vetted by the professor who oversaw the lab). No idea how available it is, nor how much it might cost for a tankful. I did the work in 1973 so things may have changed since

Yes - I do have some. Yes, bending, drilling, and doing all the stuff you might dramatically drops the mu value. Even knocking, dropping etc.
The mu-metal shields used to place around CRT oscilloscope tubes, in production, were all made up, then put through heat treatment annealing, before painting. Mu-metal as tape in a roll was the standard stuff for making specialist current transformer cores.

I also have a length of magnetic shielding in the form of about 1 metre long of what looks like tension spring about 5mm diameter, all soft and wobbly. One puts double screened twisted pair instrumentation wire up it. The outer screen is connected at only one end, to prevent induced currents on it's outer letting it act like an antenna. Complicated and expensive!

Then I discovered high-mu ferrite! Salvaged from old PC switcher power supplies, It can be had in very high mu values, and has none of the shifts in material properties like mu-metal. A ferrite toroid over a cable of twisted pair, screened, will stop all common-mode currents, and is called a "braid-breaker" for that reason. I thought to stay with stuff that gives least problems, and is cheap and easy to get.

 

[WobblyHand]

The ferrites for cables are great problem solvers.

 

[graham-xrf]

It's likely there is conducted 50/60Hz in the circuitry. Perhaps radiated. Can the whole setup be run on batteries right now? Would eliminate conducted 60Hz and all the power supply noise? I've rarely run into the case where radiated 60Hz is enough to mess things up on a proper PCB. It happens, but not that often. I've seen a lot of people try mu metal and it had little effect. Perhaps your experience is different. Edit: Just reread part of the thread. Yeah, I see where you are coming from...

Still, if I were to bet, it's not likely that magnetic 50/60Hz coupling, is the root issue.

Sure, if you don't connect it to power involving anything that ends up plugged into a wall, you eliminate a major interference route to noise. Know that every VFD on that mains is happily sending wideband high frequency harmonic currents all over the place. Unfortunately, a sensitive gadget, even if run off batteries, can still get messed by the magnetic component of low frequency fields.

I had a Metrix multimeter, sadly now lost, which seemed to to think it was measuring a strong 50Hz frequency, even though I was parked at the side of a field while watching a wood chipper machine - until I spotted the power lines above the field. I am told they cause a "small voltage" on the damp field between hoofs such that cows will not go under them.

All this stuff is OK! I am certain that Mark will chase this down, and right now, he is the one who has got actual electrons moving in real kit. Now that we have thought of it all, the design will be proof against it.

You are right that the way is to get the counts as low as possible before any signals are allowed, and then keep eliminating unwanteds as more of it is enabled. I am pretty sure this is happening anyway.

 

[WobblyHand]

High power lines are a different matter. Remember being near by a high tension line and hearing corona and sensing the electric field. It was an odd and uncomfortable sensation. Made me want to leave the area. Wouldn't surprise me if the livestock could hear the corona or sense the increase ozone production in the area - along with sensing the fields.

Similarly with high current devices, or devices switching high currents, like a VFD, there can be all sorts of EMI trash generated including strong magnetic fields. I wasn't envisioning running an X-RF off the same mains as a VFD - at least not on purpose!

The battery idea was to greatly reduce the power line coupling into the circuitry. If it was conduction through the lines, then that path would be eliminated. Just part of the elimination process. Got to be methodical about it, and just try stuff. Mark is doing a great job at it, just offering up thoughts to stimulate the thought process.

 

[graham-xrf]

The battery idea was to greatly reduce the power line coupling into the circuitry. If it was conduction through the lines, then that path would be eliminated. Just part of the elimination process. Got to be methodical about it, and just try stuff. Mark is doing a great job at it, just offering up thoughts to stimulate the thought process.

Exactly this same idea is why, in my design, the PIN diode bias is made using 3 lithium DC battery cells. All the ways I could derive bias from the oscillator type source in the Pocket Geiger, no matter how I "filtered" it, had unacceptable noise.

We do have to interface with a computer thingy which, even if we use a battery set, has computer clocked stuff going on in it. I do use the 3.3V output, filtered, and then cleaned up with a low noise linear vregulator to make the 2.5V for the ADC, and that regulator is a bit special. Yes, it does have the low noise, but it has a bandwidth so high that it can "regulate out" even the remnants of noise racket from power supply and computer currents on the 3.3V.

Then, I want use one of the low noise wideband op-amps to active divide the clean 3.3V, using two resistors, shunted by capacitors, to make the 0V for for itself, and the rest of the TIA small signal op-amps. The 0V is actively driven with the opamp low noise output, and in that way, we make available the positive and negative supplies for itself, and all the others. We set positive to +2.50V, so the ADC cannot get hurt, and I may include some extra way of ensuring that. The negative end is at -0.8V, which is enough to ensure the inverted signals never hit the rail. This is one of the few times one ever sees an op-amp output directly connected to a 0V, and we keep in mind that it is not the same "0V" that the original 3.3V came with. We have to call it "0VA" or "0VB", or something

 

[WobblyHand]

I don't understand the nature of the input signal. There's no rail to rail TIA, so the V- is created to avoid that issue? The signal is primarily positive going? The negative supply is just to handle the undershoot?

 

[homebrewed]

I don't understand the nature of the input signal. There's no rail to rail TIA, so the V- is created to avoid that issue? The signal is primarily positive going? The negative supply is just to handle the undershoot?

I think the purpose is to avoid overdriving the ADC. If you look at my signal conditioning board you will see I have a resistor + diode hung on its output. The diode is connected to the Teensy's +3.3 so it's a clamp, designed to accomplish something similar.

 

[graham-xrf]

I don't understand the nature of the input signal. There's no rail to rail TIA, so the V- is created to avoid that issue? The signal is primarily positive going? The negative supply is just to handle the undershoot?

No "undershoot" is allowed. In my circuit, a 45pA pulse, with a 820K TIA gain and the FET in place makes a 39uV signal, negative going.
The following stage lays on another 41dB of gain, bringing it up to about 5.2mV positive going.
If we decided that is as big as the pulse gets, we put in another x 120 gain, to finally get a 650mV pulse copy of the original current, still positive going.



**************************************
In another scenario, where I assume a "big" diode pulse is 2.85nA, The first stage gain with the FET + Rf =330K delivers a negative going 840uV.


Then, a +20dB inverting stage delivers a positive going about 86mV
A final +27dB non-inverting gain stage brings it up to +2.0V.
The internal reference is 2.048 volts
What final gain we use all depends on how big a current pulse happens from the most energetic photon we ever care to display
This is why I am after being able to have switched gain ranges.
In theory, a Am241 source could provoke a pile of 59.3KeV pulses from tungsten, along with some 8.4KeV and 9.7KeV. We only find out when we show it a carbide end mill



******************************
The op-amp active driven 0V supply circuit

It is a circuit I have used before, and I completely understand how it can confuse. Imagine temporarily that the output is not connected to 0V, but instead is called Vout. Then imagine the negative end is indeed connected to the "0V". We then have a simple circuit where "0V" is the negative supply for the opamp, 3.3V is the positive supply. and the voltage divider delivers +0.8V, which is buffered by the opamp unity gain follower.

Now if instead we "disconnect" the 0V from the bottom of the 3.3V supply, and connect the output of the opamp to that 0V. The circuit still operates in exactly the same way, except instead of asserting that the opamp output is +0.8V, it has to assert that V+ is 2.50 volts, and that V- is -0.8V, up to the limit of the current the opamp output is capable of giving. It becomes a dual regulator for the TIA low current circuits, limited to +2.5V, and offering the possibility of -0.8V for the op-amps, so never requiring them to go all the way to the rail.

The key thing is that the 3.3V supply negative end, if it had a 0V from anywhere else, is not the same 0VA as we are using for the TIA.
The scheme works well, but has it's inconveniences. You get the same result using two separate regulators, but the need is to contrive a no-noise negative supply. The really important thing is that no signal drives or powers the ADC above +2.5V. It's absolute maximum is 2.7V, and not recommended ever.

You may be right that the good way is just to use batteries.