Removing DC offset from AD8220 output (Instrumentation Amplifier) in electromagnetic application

Hello,

are you sure that your problem is caused by a signal offset produced by your flowsensor? In your sketch I cannot recognize an input bias current return path (page 21 of the datasheet). Of course your sketch is not a full schematic, but if the input bias current return path is really missing, the common mode voltage of the amplifier will drift undefined beyond the allowed range, and this can lead to unwanted offset voltages at the output. In that case high impedant resistors from the amplifiers inputs to GND will help (similar to fig. 60 in the datasheet). Maybe your offset disappears (or is strongly reduced) once you provide the necessary current path.

If the offset is really produced by the sensor itself, an AC-coupling would be the standard solution. It will only work, if your sensor is AC driven (its polarity is switched inbetween individual measurements). Fig. 60 gives one possible example for AC-coupling, the standard approach would be a shown in fig. 64. For the correct dimensioning you would have to specify in numbers what frequency your AC-driven sensor is using and what you specifically mean with statements like "should be fast enough ... not having a big delay ... should be in real-time". 

best regards

Achim

Dear Achim,

Thank you very much for your nice and useful answer.

We have implemented all of these details you mentioned, but there are some issues, for example for adding a capacitor (uF) and 1M ohm before AD8220 to form a high pass circuit to block the DC content of the signal, the delay would be:

assuming:

R=1 MΩ=1×106Ω

C=10 μF=10×10−6 F

then:

τ=R×C

τ=10 S

Regarding the AC coupling circuit, we also implemented that, but there is also a delay, which is not acceptable for our design:

Our system is ultra low power and battery based, so delay is very important for us and should be as small as possible. This is our first schematic for your reference:

Would you please let me know if there is still a way to remove this DC offset in a very fast way? (with minimum delay)

Thank you in advance for your help

Hello,

thanks for the explanation and the schematic. It clarifies some aspects. So it seems, that you already have resistors providing an input bias current return path (R5 and R6, though the numbers are somehow hard to read. I can also not clearly read the value of the gain setting resistor.)

Other aspects are still unclear to me. The most important one is: do you drive your MID-sensor with AC or with DC? AC typically means, that you periodically switch the polarity of the flowsensor. If this is the case, the next question is: what frequency can you use to switch the sensor? This is decisive for choosing the correct edge frequency of the AC-coupling. The higher the frequency you can use, the easier the AC-coupling. Maybe you have choosen the edge frequency of your highpass filter much lower than necessary.

If you sensor is DC-driven and not periodically switching its polarity, then forget about AC-coupling. AC-coupling only works for AC signals. Then I have no solution idea, because for a purely DC driven sensor I don't know how to differentiate between an offset and a real (DC) flow signal.

Finally it's worth to discuss, what exactly is meant by "delay". It is true, that an RC-highpass with a time constant of 1s takes several seconds to settle. But this settling time does not mean the same as a delay of the input signal. If you switch your circuit on, it takes indeed several seconds until the the AC-coupling is setteled and the offset is completely suppressed. But once it is suppressed, any change of the input signal (e.g. a polariy switch of the flowsensor or a sudden change in flow speed) appears immediately at the output of the amplifier. So the AC-coupling does not cause a signal delay, it causes a neccesary settling time after power of the circuit is switched on.

If this initial settling time after power on is your real concern, you might consider something like the following process:

1) use an DAC ouput to drive the Ref-pin of the AD8220. At power on, start with a voltage of e.g. 2.5V at VRef

2)  after power on give enough time for your MID-sensor signal to settle. Then measure the AD8220 output with your ADC and change the DAC-voltage driving Vref to bring the AD8220 into its linear operating range. It's not necessary to bring the AD8220 output to exactly 2,5V, it's sufficient to have it in it's linear region.

3) take this output voltage of AD8220 and store it as ADC-value A.

4) switch the polarity of the MID-sensor and again give enough time for the sensor to settle.

5) read the new output voltage of the AD8220 as ADC-value B. Calculate if the signal chain is still in its linear region.

If both values A and B were measurend with the signal chain in its linear region, the difference ADC_A - ADC_B is the sensor signal you look for.

If value B is not in the linear region, you have to adjust Vref in a way, that the measurements with both polarities are in the linear region. For polarity A you have all necessary information. For polarity B you may still have to search for an appropriate value at Vref. If you found it, you can calculate a Vref value, for which both polarities of the sensor keep the signal chain in its linear operating range. If you don't find a solution for that, then the gain you have choosen is too high.

This process could be seen as something like a "programmtically driven AC-coupling" instead of the analog AC-coupling with RC-filters at the input or integrators at Vref. It can run faster, as you don't have the exponential approximation to the final value (which gets slower the closer you get to the final value). And it can run faster because you don't have to reach the exact value of the perfect offset suppression. It is enough to suppress the offset well enough to bring the signal chain into the linear working range for both polarities of the sensor.

In principle you might also use this as an "indirect measurement": what Vref-value does it take to bring the AD8220 output to exactly 2,5V. If you have this information for both polarities, you can calculate the sensor signal from that.

best regards

Achim

Thank you Achim,

You explained  very nice al the possible details, I appreciate it. 

Here are my answers:

The values are R6=1M, R5=1M, R gain=4.9 K

We send pulses that its frequency is 25HZ for 4inch meter.

The main problem is that our flow meter is battery-based so after excitation and measuring we turn off our analog circuits and after few seconds we turn on the analog circuit. In this situation waiting during the settling time will consume the battery. 

According to the schematic we have used a DAC and connected it to Ref-pin. 15ms after truing on the analog circuit before excitation we measure the DC signal in the output of AD8220  then we adjust the DAC value the we start to excitation and measurement. I also attached a PDF of the schematic to have more visibility. 

1- In this circuit, in practical mode, Would you please let me know, what is the maximum value of the offset voltage at the input of the AD8220 (Vin)?

2- In the case that the direction of the electrodes does not change, is it possible to change the polarity of the input offset voltage?

PDF

thanks for the additional details and the well readable pdf-schematic.

Andromeda8 said:

We send pulses that its frequency is 25HZ for 4inch meter.

Then it should be possible to improve the settling time of the AC coupling by ~2 orders of magnitude. You had choosen the edge frequency of the AC-coupling at 0,016Hz  (1/(2*pi*10µF*10MOhm)). Speed it up by choosing 100nF and 1MOhm. The edge frequency of the AC-coupling will still be one order of magnitude below the operating frequency of your flow meter. And the settling time will be a factor of 100 faster than before.

Andromeda8 said:

1- In this circuit, in practical mode, Would you please let me know, what is the maximum value of the offset voltage at the input of the AD8220 (Vin)?

That is in general a question of supply voltage, input voltage (common mode and differential) and Vref. It is usually examined using the diamond plot. Under the following link

tools.analog.com/.../

you can find an online-tool, where you can examine the allowed voltage ranges for your setup.

Andromeda8 said:

2- In the case that the direction of the electrodes does not change, is it possible to change the polarity of the input offset voltage?

In this case, we are not discussing the offset of the amplifier circuit but the offset of the sensor. Unfortunately, I do not know how to influence the offset of your sensor.

best regards

Achim