Please help me to figure out the source of offset

[miceuz]

Hi all,

I'm building a thermocouple interface based on AD8237 http://www.analog.com/static/imported-files/data_sheets/AD8237.pdf instrumentation amplifier - my schematics is below. Yesterday I had an opportunity to use a 7 digit multimeter to test the performance of my design. Could someone go thru the process I've used and comment on that was I doing wrong?

I can't interpret the results I get - I'm getting somewhat 65uV error over the range of my interest (0mv - 45mv). This is within the spec of instamp, but very on the edge really.

I've measured offset voltage by shorting the inputs of instamp. By subtracting reference voltage from the output and dividing it by gain of 101 I get an offset of 20uV which is totally ok - distribution graph has a big peak around that area.

Since I don't have an accurately controlled oven to test it with a thermocouple, I've made myself a test rig to emulate the thermocouple - I've hooked up a 1.2V reference and a resistor divider out of 100k resistor and a 10k pot, i'm powering it from the battery. I went and made 45 measurements - in average they show somewhat ~65uV error.

Max bias current for this instamp is 650pA -- this can't be causing additional 45uV of offset.

What could be causing this error? Is my setup wrong? What could I change to test it another way?

[Harvs]

If I'm reading your schematic correctly, you have the reference set at 49mV above ground which is also the negative supply rail, yes?

If this is the case it represents a fundamental difference to how the inamp was tested and spec'ed for the datasheet.  It has rails at 0 and 5V, and the ref in the dead center at 2.5V.  I'd say a 65uV offset is perfectly reasonable in this case, and I would typically expect worse.

Having said that, is it then going into an ADC?  If so, just cal it in software and be done with it.

BTW I have no idea what your graph is saying, you need to label what the traces are.

[miceuz]

If I'm reading your schematic correctly, you have the reference set at 49mV above ground which is also the negative supply rail, yes?

If this is the case it represents a fundamental difference to how the inamp was tested and spec'ed for the datasheet.  It has rails at 0 and 5V, and the ref in the dead center at 2.5V.  I'd say a 65uV offset is perfectly reasonable in this case, and I would typically expect worse.

I didn't know that reference voltage could have such a huge impact, it's not mentioned anywhere in the datasheet. Anyway, my concern is that 65uV is suspiciously close to 75uv max.
 

Having said that, is it then going into an ADC?  If so, just cal it in software and be done with it.

Sure, but I wanted to make sure I've done everything right on the analog side of things.

Regarding the graph - yeah, forgot to do that. The gray trace is the error and the red trace is error filtered by median filter. There is input voltage on x axis and error (input * gain - (actual output-reference)/gain) on y axis

[Harvs]

If I'm reading your schematic correctly, you have the reference set at 49mV above ground which is also the negative supply rail, yes?

If this is the case it represents a fundamental difference to how the inamp was tested and spec'ed for the datasheet.  It has rails at 0 and 5V, and the ref in the dead center at 2.5V.  I'd say a 65uV offset is perfectly reasonable in this case, and I would typically expect worse.

I didn't know that reference voltage could have such a huge impact, it's not mentioned anywhere in the datasheet. Anyway, my concern is that 65uV is suspiciously close to 75uv max.

Yes, the common mode voltages make a very big difference to any opamp's Vios. An inamp has both the Vcm that we see at the terminals, plus the reference voltage which forms part of the Vcm for the second stage.  You are operating both the Vref and Vcm miles away from where they have spec'd it.

The 75uV max spec is only valid for the test conditions they specify, you need to read those with any spec and understand that if you operate away from these conditions the device may not meet those specs (and potentially by a large margin).  Effectively the manufacturer wants to show their product in the best light, so a fair bit of marketing goes into what not to put in the datasheet.