# Instrumentation amplifier cross-connection circuit

I am trying to implement a circuit from an ADI app note that I have not seen anywhere else : Here it is in PDF format which is a copy of this article. I strongly encourage reading it as it is quite short and interesting, providing much more insight into the workings of the circuit than I could myself. Here it is anyway : Note that the INAs are horizontally mirrored which can further add to the confusion (look at the input pins).

Without going into too much detail I'm using this circuit for its high input impedance and common mode rejection ratio, and the differential output to drive an ADC (which will provide further CMRR). I am also using a single supply which is more convenient for my use case, the AD8224 should tolerate GND centered signals within reason and simulation confirm this, although I am unsure if this has any influence on the issues I am having.

Unfortunately my circuit doesn't behave quite as expected, from my own tweaking of the resistor values I have the following observations :

• Removing any one of the two resistors (leaving a short) make the circuit completely unstable. From the article I understood that R2 is for setting the gain and R3 for attenuation and using both at once was unnecessary.
• The gain of the circuit seems to be dictated by the lowest of the two resistors, irrelevant of which, and in accordance with expected values based on AD8224 datasheet using a single gain resistor.

Here is an example with a measured / simulated gain of 10, swapping resistor values makes no difference on the output : Here are the few measurements from trying different resistor values, hopefully you can make sense of it because I can't. Of course will try any suggested configurations and report with the results.

• Both resistors @ 50Kohm gives a gain of -100
• Both resistors @ 500ohm gives a gain of -10
• R2 @ 5000ohm and R3 @ 500ohm gives a gain of 100 (same as 50K and 500)
• One resistor removed (open) and other one @ 500ohm also produces a gain of 100
• In each case swapping the resistors has no effect on the outcome
• Have you actually put it together on a PCB and tested the real thing, or are we talking about spice results only? It's not clear from the question. Now, SPICE is telling you the right thing, but a physical circuit would be more assertive: I expect the chips would run much warmer than normal. Aug 11, 2022 at 14:59
• Tested with R2 OR R3 ... Seems a very interesting thing. Gain change, +/ - when one resistor swapping position. And adding Vcm is easy. Aug 11, 2022 at 15:23
• My bad, it seems after a more thorough read of the article you need to place a HIGHER value resistor to achieve attenuation in R3.G = 1 w/ 50Kohm in R2 and no R3, G = -1 w/50Kohm in R3 and no R2 (outputs inverted) and G = -2 w/ 100Kohm in R3 and no R2.
– Ryan
Aug 11, 2022 at 15:31
• Here the link : TRAN Analysis : dropbox.com/s/u4mynbccuclzeki/… Aug 11, 2022 at 15:37
• Sorry ... I forget. microcap v12 file as usual ... Aug 11, 2022 at 15:48

This equation is part of the document you referenced. $$G = 2 \cdot \left( \frac{R_1}{R_2} - \frac{R_1}{R_3} \right)$$

In this case, R2 corrersponds to your 50k resistor, and R3 to your 500 ohm resistor. R1 is internal to the AD8224, and is effectively 49.7kOhm.

• Removing any one of the two resistors (leaving a short) make the circuit completely unstable. From the article I understood that R2 is for setting the gain and R3 for attenuation and using both at once was unnecessary.

Yes, shorting either of your gain resistors will cause the R2 or R3 term to become undefined (and go toward infinity). Unstable output is the expected result.

• The gain of the circuit seems to be dictated by the lowest of the two resistors, irrelevant of which, and in accordance with expected values based on AD8224 datasheet using a single gain resistor.

Again, this is expected: you're setting the overall gain of the circuit by dividing R1 by two different values. If one instance is dividing by a comparatively large number (50k), and the other by a small number (500), then you can expect the "small" division to dominate the large.

Here is an example with a measured / simulated gain of 10, swapping resistor values makes no difference on the output.

This is also predicted by the equation, except that you should see the phase change by 180 deg.

• I see, this makes sense. The 180 deg phase shift is not something I looked at but outputs are differential either way and I'm not too concerned which leads which, although it is my bad for saying they was no difference. That being said, how come a 500ohm resistor at either R2 or R3 results in a gain of 10 ? I was under the impression from the formula that in position R3 this would result in attenuation instead.
– Ryan
Aug 11, 2022 at 15:09