AD8420 4-20mA Receiver grounding

I have a simple AD8420 configured as a 4-20mA receiver powered by +12v/GND .  The issue I have is that the output is only valid if I tie the return of the 4-20mA source (AI420_4_N) to the same ground as my board (DGND).  

I'm within the common mode requirements ie with 10mA input, I have 1.10V on "+" input and 0.78V on "-" input reference to my board ground, which is expected 0.43mV input voltage (across the 43 ohm resistor).  without a common ground, the output is ~150mV and slowly decays.  With the common ground I get my expected 1.93V out.

(These numbers are with R202 removed as I thought that could be causing some issues)

The reference design shows the common ground but I thought that was just for that externally powered loop design.  I've done a similar design with the AD8429 and didn't have this problem...

thanks

  • Robb,

    The datasheet difference between AD8429 and AD8420 as far as I can see seems to be in the input bias current magnitude and direction. AD8429 is positive (flowing into the device) and large (150 nA.) while the AD8420 is negative (flowing out of the device) and much smaller (27 nA.). Current flowing out of the device tends to make the inputs rise without limit if there is no DC return to DGND. Any inamp always needs a DC path for its input currents to flow through and eventually back to the inamp power supply. Are you able to identify that path? I don't see it from your diagram. (Hopefully the data sheet is accurate in reporting the sign of the bias current correctly).

    Maybe if you add a 10 Meg resistor from each input (R1 and R2 in Harry's figure 5-4) to ground it will prevent the common mode from rising to where the DUT fails to function.

  • thanks for the help.  The 10Meg resistor did the trick.  The design with the AD8429 might of worked as we had some protection diodes on the inputs referenced to ground.  Maybe enough leakage to give a reference for the input?..

    Harry - the voltages were measured referenced to the board ground.  Probably those were stable as the meter provided a return path through it's input impedance (>10Meg).