Hi folks,

Ive been reading up on everything google had to offer, including the related threads on this forum. However im drawing a blank and im about to lose my sanity over this. So here goes:

I have a simple set up to verify that the AD623 does its job at amplifying the signal _linearly_. The V- is grounded, V+ is connected to a voltage divider with a potensiometer (to adjust input voltage for testing), Vs+ is 5V, Vs- is grounded, Ref is grounded and i am of course measuring on the output pin.

The Rg is 5k Ohms, which makes the gain about 20.

The problem is: What i put on the V+ pin seems to have NO linearity with what comes out at the output pin. Im trying to check the gain factor by dividing the voltage at the output pin with the voltage at the V+(input) pin and the Gain factor is a LOT at very small voltage and almost nothing at higher gains. (Ofcourse it will stop amplification of the signal when Output reaches 5V, but i mean even before this happens).

I am very lost and totally out of ideas so PLEASE anybody help me!

Thank you so much for any help!

 

I have used this chip before, and I had trouble with it too.

I switched to making my own instrumentation amp using an op amp with a much lower offset error, since the voltages I was amping were very very low and the offset of the better version of AD623 is a very big 100 microVolts.

Anyway, see the attached images. I have two suggestions.

1. Did you include the capacitors and do any low-pass filtering as they recommend in the data sheet? See my first image. Note I was doing larger gain than you. Did you look at the output signal with a scope not just a voltmeter? Maybe you're getting oscillation.

2. Don't use AD623, rather use another instr amp or else build your own as I did in the other attached image. You can use the dual op amp as I have shown, or the more typical 3 op amp version. You can select an op amp with a low offset error, like 1 uV or 2 uV, and get better precision and make your own feedback network for the gain. Use precise resistors and you should get good results.

 

The problem is: What i put on the V+ pin seems to have NO linearity with what comes out at the output pin. Im trying to check the gain factor by dividing the voltage at the output pin with the voltage at the V+(input) pin and the Gain factor is a LOT at very small voltage and almost nothing at higher gains. (Ofcourse it will stop amplification of the signal when Output reaches 5V, but i mean even before this happens).

Can you give more info on this? What do you mean by "very small voltage"? When you wrote "at higher gains" did you mean "at higher input voltages"?

I wonder if it is related to the offset error of the chip, up to 200 uV.

 

I have used this chip before, and I had trouble with it too.

I switched to making my own instrumentation amp using an op amp with a much lower offset error, since the voltages I was amping were very very low and the offset of the better version of AD623 is a very big 100 microVolts.

Anyway, see the attached images. I have two suggestions.

1. Did you include the capacitors and do any low-pass filtering as they recommend in the data sheet? See my first image. Note I was doing larger gain than you. Did you look at the output signal with a scope not just a voltmeter? Maybe you're getting oscillation.

2. Don't use AD623, rather use another instr amp or else build your own as I did in the other attached image. You can use the dual op amp as I have shown, or the more typical 3 op amp version. You can select an op amp with a low offset error, like 1 uV or 2 uV, and get better precision and make your own feedback network for the gain. Use precise resistors and you should get good results.

Thank you for you reply! I have only decoupled with capacitors at the power rail. The source for my measuring voltages is a thermistor so since the value will change so slowly i thought there was no need for filters. I will try to get a hold of the capacitor values in your drawing and try to connect it. I have precise resistors and had never thought that the AD623 would prove itself so useless after what i read after googling last week.

I have voltages from 0-125mV where the part i care about is 95-125mV - what is the easiest way to accurately(relatively) amplify these voltages so that i can measure them at my ADC?

What do you suggest? Your alternative 1 or 2?

Thanks again!

 

I honestly don't think that #1 will help you. My voltages were smaller and I was doing low-pass filtering to eliminate noise.

I don't know what is the deal with the chip, except that the offset error is indeed very high for an instr amp. That should not be the problem with you, since you're in the 100 mV range. I was in the 3 mV range.

It sounds like one of your inputs is ground, and you're really amplifying a single-ended signal.

You're interested in the range of 95 to 125 mV, so it's not like you're scraping the ground.

Therefore, I would recommend this simple circuit:
http://en.wikipedia.org/wiki/Operational_amplifier_applications#Non-inverting_amplifier

A simple non-inverting amplifier with one op amp. Get an op amp with a true rail-to-rail and you won't even need a negative rail. As long as the op amp is true and has low offset error on the whole range, then the resistor accuracy will not affect linearity. Any resistor error will only affect gain.

Does that make sense?

 

Ok I will try with an Opamp now, an OPA337PA.

By the way, i am measuring on a 2 channel oscilloscope - both on the input and output!

 

Ok, I just hooked up (on a _breadboard_) a OPA337 opamp and it works perfectly. I feel like someone has been toying with me haha... Why is this, the simplest of solutions, so accurate - while the expensive AD623 does such a terrible job?

My setup now has a gain of 1 + (33k / 1k) = 34. I am measuring 108,6mV on the input and 3,69V at the output. This gives a real gain of 3,69/108,6mV=33,9779 which is five oceans closer to expected output than I could ever reach with the AD623.

Im thinking of connecting the ADC Ref to 4V then i have a pretty good resolution for measuring the output, that should work fine dont you think?

 

Good! I am happy when a simple solution makes a complicated mess disappear!

Use whatever reference you want. You have the control over the voltage range by changing the resistor ratio, so you could use whatever range you like. The values like 2.048V and 4.096V make the conversion to a voltage very easy, if that's what you're doing.

If you want good absolute accuracy, then a good reference voltage is critical. I always like the ISL6000x series. They are very accurate and low power, though a bit expensive around $4 each.

But if you have the luxury of calibrating the values, or if absolute accuracy is not critical, then you could even use the internal reference of a microcontroller. Those are typically not so accurate but they are a reference and no additional cost.