Assistance in designing measurement circuit for Tolman-Stewart experiment: difference amplifier

Something like this maybe:

You could analog subtract the diffamp outputs, or digitize both and do the subtraction (and some filtering) in a PC.

Those big coils will be great antennas, so there needs to be some emi/lowpass filtering ahead of the amps.

On a sunny day (Sun, 14 Jul 2019 19:42:28 -0700) it happened John Larkin wrote in :

Did you forget a resistor in the - of the top amp input filter?

Yes. I'll knock 10% off my design fee.

Thanks!

A question: the capacitor before the last amplifier -- isn't that usually u sed to remove DC? The signal we have is so slow (a pulse over 0.5s) I feel like I should be worried that the capacitor will remove parts of the pulse . Should this not be a concern?

Of course it's a concern. Use the right cap.

I still think that the counter-rotating torsional oscillator and lock-in detector is a far better way to do this. It solves a heap of electrical and mechanical problems.

For sure the oscillating coil or oscillating cylinder of metal (that's how they did it in 1924) is better. There is a version from the 80s that did that method and got very nice results.

With the braking coil, the math behind the physics is quite simple and can be understood after the first year in college. The oscillating one has much more complex math, that's why we hesitate.

The oscillating thing with lock-in detection does involve some slightly more complex issues, but not something that undergrad-physics students should have difficulty with. The average acceleration of an oscillating-rotation disk sounds like first-year physics to me.

If a torsion bar connects the coils, and they counter-rotate, the midpoint of the torsion bar doesn't move. So put the electronics there and bring out as many wires as needed. No flexing, no slip rings, big signals.

They could do that at Cornell!

The derivation of the charge to mass ratio is more complex. But maybe we will try after we get this version working!

I have two questions about your circuit.

First, why do you suggest using the two AD8229 differential amps? Why are those better than just taking the two outputs of the two coils and putting them into a TIA or just a simple inverting amp, and from there into the summing amplifier?

Second, on your low pass filter and emi filtering, what is the resistor closest to the coil to ground for? The capacitors surrounding the resistor look like a pi-filter to me, but the first resistor I don't understand what it is for.

Thanks!

To keep magnetism at bay, I suggest this nonmagnetic motor mechanism

The common-mode input range of the amplifier includes ground, but not necessarily a floating input. Nominally, grounding one end of the coil (or a center tap) would be just as good. At the nanovolt scale, common-mode rejection is going to be good enough for any decent amplifier.

Any possible positive feedback (even just adjacent-pin leakage) might be worth worrying about; guard rings or lifted-pin wiring recommended if an input pin is adjacent to power or output pins.

That somewhat optimizes s/n ratio and makes it easier to trim the relative gains a few per cent, to null ambient field noise.

You need an overall gain of a million or so. 1000 in each diffamp, and another 1000 in the opamp.

The diffamp needs a defined DC common-mode voltage, which is ground in this case. We can't float the coil into the diffamp. I'm trying to keep everything symmetric so's to let the diffamp cancel common-mode noise that the coil picks up electrostatically from the world.

The common-mode resistors can be large, many times the coil resistance.

As you may have noticed, I'm not a big fan of using a TIA here.

Squirrel cage induction rodent.

??? simple harmonic motion is highschool stuff. The Fourier transform is introduced at first year, but that's not really needed to analyze this.

Ok, I think I get it. But why put the resistor on both sides of the coil? Doesn't that set the reference voltage on both sides? I thought we would want to set the reference only at one point along the coil.

JL can jump on me if this is wrong. But I look at the resistors to ground and 'DC bias' resistors for the amp input. You could leave one out and that input of the amp would know about ground through the coil. But I think he's trying to keep everything symmetrical.

One potential problem with this type of scheme (filtering two signals and then subtracting) is that if the two filters only match to 5% (or something) then the subtraction won't be perfect and you still have ~5% of the common signal remaining. (Still 5% is better than

George H.

Symmetry (might) keep common-mode electrostatic noise from becoming differential mode signal. That theory sort of assumes that the coil is symmetrically exposed to any e-field noise, which it likely isn't.

The input bias current of the diffamp is small, so a single largish resistor to ground would keep the DC common mode voltage from drifting away.

Symmetry looks nice.

Yes. Asymmetry in individual filters (cap tolerances, mostly) can also convert common-mode noise to diff mode.

We're contemplating a voltage gain of multiples of 1E6 here! More than the open-loop gain of most opamps.

Right! (I don't know the right words.) Our (my) lockin has an int amp input. With AC on both channels... well stuff only cancels down to wherever the LF DC/AC edge is. There must be a similar mis-matching on the HF end, but I've not seen it. Re gain: Well one could always add another stage, at the high end more gain is another inverting opamp.

I have no idea where the noise is... EM-wise in the few Hertz region. In the audio range.. say 100 Hz to 100 kHz, an electro- static shield (metal box.) helps a lot. AC 60 Hz magnetic fields and harmonics are hard to shield, mostly I try to reduce area and move away from the source. George H.