INA823: Output offset (~9 mV with 0V input) using +5V supply

Part Number: INA823
Other Parts Discussed in Thread: ADS1115, LM2776, LM7705

Hello IT experts,

I am using the INA823 in a simple power configuration (+Vs = +5 V, –Vs = GND) to measure 0–10 V and 0–20 mA analog signals.

When the input is 10V or 20mA, the output is about 2V, which is correct.
However, when the input is 0 V (or 0 mA), the output does not drop completely to 0 V, but instead remains around 9 mV, which affects the reading on the ADS1115 ADC.

By applying a small negative supply (–5 V) to the Vs– pin, the output offset decreases considerably to about 168 µV.

  1. Is this residual voltage (~9 mV) normal behavior when operating with +5 V supply only?
  2. Is it necessary to use a negative rail (e.g. with the LM2776) to get accurate readings close to 0V at the output, or is there some recommended configuration to achieve true 0V in single supply mode?

Thank you in advance for your support.

Best regards,

image.png

  • Hey Ivan,

    You have correctly identified the problem and offered a perfectly viable solution!

    What appears as the Vos is actually being generated by the swing illimitations of the INA. When we consult the datasheet, we see that the device will always drive within 150mV of the rails with a 10kOhm load. 

    Realistically, it is not unexpected to see the device actually swing much closer than 150mV, though we should sufficiently guard band our output to ensure we do not run into non-linearities here. 

    To solve our swing to rail problem, we either need to swing less, or push the negative rail away from GND. As you have mentioned, a device like LM2776 may work well to accomplish this function. This switched capacitor negative supply switches at 2MHz, so there may be slightly additional injection of HF noise from the V- rail. LM7705 is another suitable option for this function.

    The other solution is to rescale the output and add a non-zero reference voltage. Since your voltage/current sense goes down to zero volts, the easiest method is often to add a slightly positive reference voltage, ensuring the output will have extra room from the V- rail. Consequently, you would need to compensate for this on the digital side, though this is still preferred in many scenarios where digital processing is possible. 

    Both solution have advantages and disadvantages:

    Option 1 (add a negative rail).

    Good  Intuitive and simple to implement. Slightly more noise from the negative LDO, but still very reasonable in most applications. 

    Bad  Now the INA could swing to -5V in a fault condition. This would potentially damage the ADC and create a high current conduction path. The ADC will still not be able to resolve 0V, and there may be opportunity for offset errors to produce slightly negative voltage on the ADC when the Vdiff on the INA is 0V. 

    Option 2 (add a positive Vref and digitally rescale the output).

    Good  The INA can never overdrive the ADC if they share a voltage rail. No need for switched capacitor negative rail generation. 

    Bad  We must create a way to generate a slightly positive Vref (maybe use a voltage divider with buffer amplifier/ REF device). We must also digitally rescale our output to translate the new output to the system. 

    Both solutions are valid, and can be implemented for your use. 

    Please let me know if you have any questions here. 

    Thanks

    Jacob

  • Hi Jacob,
    Thank you very much for the detailed explanation and possible solutions.
    Best regards!