[Resolved] Unable to "tame" AD620 Instrumentation Amplifier.

Other Hardware General Electronics
1

I thought using an Instrumentation Amplifier would be as simple as it gets; "easy as pie":

  1. Two electrodes from the subject go straight to (+) and (-) inputs.
  2. Pin five is a third electrode on the subject to serve as a floating ground reference.
  3. A single resistor across pins 1 and 8 determine the gain.

    But I'm baffled by the waveforms I'm getting while attempting to fine tune it. My hope is that someone here really knows their waveforms (as well as Instrumentation Amplifiers) and can tell me where I took a wrong turn. Currently, I'm lost without a map.

The frequency in these pictures is 60Hz, so I know that part is coming from the power lines near by, although I'm running this project from battery power on a wooden table. The electrodes are on leads only 6 inches long, and when no resistor between pins 1 and 8 is supposed to mean unity gain; yet the 60 cps persists.

The below is what I get with the (+) and (-) electrodes connected to each other, and the ground reference electrode connected to the power (battery) ground, with pins 1 and 8 left open:

This seems outstandingly strange, because (+) and (-) shorted together should basically cancel any signal, and grounding the floating ground should -- if anything -- send the output off scale (making a flat line across the scope).

The next picture below is with no resistor and all three electrode-leads left open:

It seems the positive peak is totally cut off at 4 volts, while the negative peak folds over and points up. Very strange, because no resistor is supposed to mean "Unity Gain", so why such a large waveform? ...and from where?

The next picture is the same, only with a 1000 ohm resistor inserted between pins 1 and 8. In some ways, that seems to have added some gain, as it should; but why is the negative (folder over) peak gone?

This next picture is with the negative input and ground reference shorted to gether and connected to my body, with the positive lead left open. Strange in that it is so different from the other waveforms.

And if I touch the positive lead, this waveform goes off scale in the positive direction, making a flat line near the top of the scope as shown below:

The below appears when including the 1k gain resistor, (-) and floating ground connected together, (+) open, and nothing touching me.

What can I say?

This final image is the same as the one above it, but with the 1k resistor removed (supposedly reducing the signal to unity gain?)

I'm so lost. None of the above makes much sense to me. If no one has the answer, I'll re-design the circuit using op-amps. At least they behave themselves and follow logic.

2

Inputs have to be between -Vs +1.9volt (1.9volt above ground on a single 5volt supply), and +Vs-1.2 (3.8volt).
So having the reference electrode at ground potential is not going to work with a single supply.
You need a pos/neg supply, or a virtual ground (2.5volt).
The pictures show all sorts of clipping.
This opamp has 10Gohm inputs. Shielding/layout/decoupling is crucial.
And it doesn't have rail2rail outputs (-Vs+1.1 and +Vs-1.2).
Leo..

3

Lots of opamps phase-reserve if the inputs are either outside the normal range or different by more than
some voltage limit - such opamps cannot be used as comparators for instance. When in a stable
gain configuration with input in range and not overloading such saturation behaviour is immaterial,
don't worry about it. If you need a rail-to-rail instrumentation amp, choose a device that can do that,
or synthesize one from 3 opamps (its a thoroughly standard configuration).

4

Wawa:
You need a ... virtual ground (2.5volt).
....
This opamp has 10Gohm inputs. Shielding/layout/decoupling is crucial.

I started working on implementing your information last night, as soon as I found your replay. Everything you said was useful, but the above two lines were the most important, enabling me to get it working. Thanks! Now I can see the EMG signal when I flex my left thumb muscle (my goal for this sensor).

I still have a ways to go though. There is currently too much noise, resulting in not enough sensitivity to the desired signal. Searching Google, I found the below, and am about to build it, hoping the noise problem will be solved as well.

And I'm hoping that moves me toward the frequency spectrum I need. Charts show EMG as peaking at around 70 cps:

I looked up your term "10Gohm". Wow! That's one huge amount of input impedance. It should mean I only need small dry electrodes to do the job.

MarkT:
Lots of opamps phase-reserve....
....
If you need a rail-to-rail.... synthesize one from 3 opamps ... a thoroughly standard configuration.

All good stuff to know. I looked up rail-to-rail; ah, yeah. And, I've now seen the 3-opamp configuration on several sites. Thanks for helping me understand these things better.

5

If you're going to interface this opamp with an Arduino, and supply the opamp from a single 5volt rail (first diagram).
Then try connecting the reference electrode to the 3.3volt supply.
(Maybe via a 1k resistor, so accidently shorting it to ground won't kill the regulator)
Now your body is on a 3.3volt potential, right within the opamp's input voltage limits of 1.9volt and 3.8volt.
Leo..

6

Thanks, Wawa. It's on an Arduino Pro Mini, worn by the user and radioing back the data to the main Arduino. I tried the 3.3 volt pin as REF. But thought it too off-center, so am currently using two 20k resistors on an op-amp, getting very close to 2.5v.

7

Theoretically you need the middle of the input common mode range. ~2.85volt in this case.
But too much off center only matters if the opamp gain is 1x or 2x.
If you have a gain of e.g. 100x, input p/p voltage is so small that it doesn't matter.
The output will clip before the input does.
Leo..