We have designed an energy metering device based on the ADE7978/ADE7933/ADE7932 family, following the datasheet guidelines. We have also managed to calibrate the measurement with the precision required. However, we have found that sometimes when we disconnect the device and connect it again with the same setup configuration, the measurement prior to the calibration varies, which causes the calibration to be incorrect. In this sense, we have the feeling that the problem is related to the configuration of the ADE7932 input circuitry. Even though we have followed the schematics provided in the datasheet:
We have separated the input circuitry between two boards (PCB A and B) as you can see in the following picture:
Signal N is the neutral and signals L1_P and L1 are both ends of the shunt resistor placed on L1 phase, the three of them marked in yellow (PCB A).
Signals IP, IN, V1P and AGND are marked in green (PCB A) and blue (PCB B), and they are connected between the two boards using through-hole 1.27mm-pitch pins. This means that the GND planes layout guidelines (datasheet page 88) are not followed on PCB A, but we believe that, in principle, this would only affect EMI issues.
Are the IP, IN, V1P signals too sensible to be routed through these pins? Is it critical to not have the input circuitry in the same board?
Thanks for your help and let me know if you need more information.