I don't know which exact amplifier you have; just the manufacturer. They have over 40 instrumentation amplifiers in production.

 

http://www.analog.com/static/imported-files/data_sheets/AD620.pdf

Guess it's not rail to rail.

It says:

The reference terminal potential defines the zero output voltage and is especially useful when the load does not share a precise ground with the rest of the system. It provides a direct means of injecting a precise offset to the output, with an allowable range of 2 V within the supply voltages. Parasitic resistance should be kept to a minimum for optimum CMR.

So I just put my offset voltage here I guess right?

 

Pardon me for butting in like this, but I've been there, done that with op-amps and microcontrollers once or twice...

Wouldn't it be simpler to amplify the signal-of-interest by 40 (say) to give you the upper limit of 5V, and code the lower limit in software?

That way, you don't have to have the amp doing all the work in terms of limiting the upper and lower voltage ranges. You can just ignore ADC counts less than (say) 820. That will still give you a range of roughly 200 counts (assuming you're using a 10-bit ADC), where a count of 820 = 100mV and 1024 counts = 130mV.

Don't forget, you can use simple software filtering instead of physical components too - either a running average or Kalman filtering might be an alternative to the "jitter" you're experiencing. This assumes of course that you have code space and horsepower to spare!

I apologise if this is too silly and simple for your particular problem. I'm taking the hobbyist approach to solving a measurement problem in software and hardware, and it's neither elegant nor clever!

-Pete

 

Yes PC_Pete, this could work, never thought about ignoring the lower limit, I'll look into that filter too. Thanks! I'll let you know how it goes. Thanks guys for all the many suggestions, especially SgtWookie for all the advice!

 

Well, 110mV x 40 =4.4, and 130mV x 40 = 5.2v; so that's over the top.

One of the big reasons for using an instrumentation amplifier with an actively driven shield is to help eliminate the common-mode noise. Your CMRR should improve dramatically. You can use a gain of, say, 20 for the instrumentation amplifier, giving you an output range of 2.2v to 2.6v - of course, you'll need a negative supply to use the amp you have - and from there, you can use the original circuit to subtract 2.2v and with a gain of 5, you'd have an output range of 2v.

You could also switch to an RRIO instrumentation amplifier.

I don't know how far away your FET/49.9 Ohm resistor is from your ADC, and I suppose you can't tell me either. The further away it is, the more likely you are to pick up noise.

 

Oh ok I see. Yes I was going to use a gain resistor of 3.33k giving me a gain of about 38.14, so .09V --> 3.43 and .130V --> 4.96V. Unfortunately I see a problem here. The minimum input is -Vs + 1.9, and the maximum +Vs - 1.2 (I was wondering why it wasn't working and then I read the datasheet carefully) and the output is -Vs + 1.1 minimum and +Vs - 1.2 maximum. So basically inputs or outputs can't go near the rails. Since I am using a single supply with a low input, this presents a problem. Any ideas, other than using a dual supply?

I don't know how far away your FET/49.9 Ohm resistor is from your ADC, and I suppose you can't tell me either. The further away it is, the more likely you are to pick up noise.

No I can tell you, I just don't want to have a full schematic floating around out there just in case it became a issue in the future. it's not very far, on a PCB it would be close, but since this is a solder protoboard yes there are lots of wires so that is probably where I am picking up the noise, it's maybe about 5 inches away.

 

Yes, wiring acts like a bunch of antennas. You might try the "dead bug" style, or "manhattan" style of prototyping. Use a copper clad board, glue the top of the IC's to the board, and use the board as a ground plane, and point-to-point wiring for all of the connections. Put everything as close together as you can.

If you went to SMT/SMD's, you could reduce your length of wiring a good bit.

As far as the instrumentation amp being not rail to rail; you might look at single supply amps, or RRIO amps. Even a single supply amp would be more suitable than what you have now.

 

So I am ordering some RRIO op amps, but in the mean time I am just using the AD ones I have now with a dual power supply, and what do you know it works! I am able to read it with decent precision, and it is much more stable (I also average the values out as suggested and that seems to work just fine), doesn't move around much at all! So for now it seems to work. Thanks! I'll let you know if I have any more questions.

 

Glad you got it working OK. I'm actually a bit surprised it got resolved so quickly, with so many unknowns at this end.

 

Yeah me too, but yeah it works, I am able to measure the "resistance" of my JFET quite accurately (accuracy of .3 ohms maybe, which is all I need).

 

Everything you do in a SPICE simulation is an approximation. Lots of "shortcuts" are taken along the way, as trying to model ALL of the parameters of a device would not only be very labor-intensive, you would wind up with a very complex mathematical model that would take just about forever to run. You might as well just build the thing to begin with.

I remember starting a SPICE simulation in college on a 386sx and coming back an hour later to see if it was finished. Now you can tweak values and see the results in realtime.

I couldn't imagine trying to do the Monte Carlo simulations manually that were run through slide rules on the moon launches. One forgotten decimal place and *boom*.

 

I don't know which exact amplifier you have; just the manufacturer. They have over 40 instrumentation amplifiers in production.

I know this is an old post but just for those who follow down this path ... AD820 is RRO, and just about RRI, it works from +V and -V (18 V) and is available in a DIP ... which makes it a rare bird. Dual is AD822. AD820 offset pot is a 20K returned to -V, like LM741 style.

 

See

 

Thank you for your contributions.

Since this thread is very old, it is locked.