Amplification from the additional Rref Is Very Annoying

I'm using AD8222 in my sensor design where it is used to interface with Wheatstone bridge. On the prototype, +IN1 and -IN1 are tied together to fix 2.50000V.  +Vs is set to 12.0v and -Vs is set to 0v. REF1 is tied to a unity buffer op-amp (REF495) with a voltage of 2.5V. The gain setting is 351 with RG1 set at 141 ohm.


Theoretically, i would expect the OUT1 of 2.50000V. However, there is always an additional amplification of 0.02V. I suspect this comes from the amplification from the additional Rref. How can i get rid of the additional amplification of 0.02V. i attach the circuitry of the design.

  • Hi Tik Liang,

    What is your input range? When you apply the 2.5V at the inputs of the AD8222, did you short the +In and -In to make sure that there is really no differential voltage? Do you see the additional 0.02V as offset error or gain error?

    It seems that you are operating close to the diamond plot limit of the instrumentation amplifier as shown on this very useful tool. You can also learn more about input common-mode voltage vs output swing on this FAQ.

    You'll see that the output swing will be limited by the input common-mode voltage and depending on your target output range, you have to apply sufficient amount of input common-mode voltage which is ideally was set to Vcc/2 or 6V in your case.


    Any voltage at the reference pin will be amplified at gain of 1. Can you double check if you have accurate 2.5V going out from U4? I could see some resistor divider at the input pin of the buffer and its tolerances can contribute to 0.02V error that you are seeing.

    Also what is the purpose of TP1, TP2, TP3 and TP4? RG pins of the instrumentation amplifier are very sensitive to parasitics which if not properly taking care of, can results to ringing or worse oscillations at the output on the inamp.


    Best regards,

    Emman

  • Hi Emman,

    Just to clarify, I would like to made this design to have differential output. I assume AD8222 dual channel are symmetrical internally. The gain resistors for both channels are 141 ohm at the prototype testing. Is my design methodology correct in order to obtain the differential output?

    The input range is 2.5V (+-10mV), it is actually just a strain gauge transducer output. Yes, when I apply 2.5V at the inputs of the AD8222, I short the +In and -In and there is not voltage difference at the input.

    And thanks for the diamond plot tool, I had also reduce the VCC from 12V to 5V. So, input common mode voltage of 2.5V is actually sufficient now.

    I have also make sure the output from U4 to be very accurate at 2.500V.

    I did further investigation today and noticed that there is always some slight difference between Out1 (U2.15) and Out2 (U2.14). The difference could raise from 0.01V to 0.02V although I had shorted the both + and - inputs to 2.5V. As a result, the noise performance of this circuitry is bad. However, when I short both of them to Ground, the difference will go away.

    This clearly shows the gain resistor is the problem. The data pointed towards the gain error. i have actually want to made TP1, TP2, TP3 and TP4 as test points to solder an external potentiometer for gain tuning. But it looks like it backfires. The extra stub or traces could contribute to this imbalance and parasitic. I never expect the in-amp is so sensitive to the gain resistor. In order to made channel1 and channel 2 symmetrical, looks like I also need to make sure the Rg path symmetrical and as short as possible?

  • The problem go away if i use a very precise gaining resistor, 0.01%. Looks like it is caused by the channel to channel imbalance which is caused by the precision of the gaining resistors.

  • Hi Tik Liang,

    Glad that you've been able to solve the problem. Yes, generally, you have to make everything symmetrical and balance especially if you are dealing with differential signal. Have you consider doing the differential configuration suggested on Figure 50 of its datasheet? This circuit might be simpler and only uses single gain resistor.

    Best regards,

    Emman