I have followed the input connections as outlined in the schematic for the ADE7953 evaluation board, but have chosen to not include the connection from the AC neutral directly to ground, believing that connecting it via a 1K resistor to the VN input would work properly for a differential measurement. The AC line side goes through a resistor divider to the VP pin as per the schematic. I was able to calibrate it and it appeared to work correctly until a few units were deployed and we started seeing odd voltages being reported by these units.
Our power supply is very quiet on the four-layer PCB and I don't see any strange noise sources. We average the VRMS readings 60 times over one second and I have even tried syncing the readings with the zero crossing signals, but they still fluctuate plus and minus several volts while the input signal stays steady. Worse yet, when a motor load is energized almost anywhere on the same AC line (not even the measured line, any line nearby), the voltage readings go 20 - 30 volts higher until the load is shut off again. I have also noted that the zero crossing output pin is rather unsteady and when a motor load is connected, the zero crossing output often goes low and does not toggle at all.
We originally had the neutral connected to the ground of the board, but ran into serious issues with AC currents on the system ground. Even now, if I place an AC ammeter from the neutral to the system ground I get several hundred milliamps of AC current flowing, which is unacceptable. Plus, there is the very real safety issue of potentially having the system ground at line levels if the board is miswired.
This is an embedded application. As there will be retrofit uses for this product, I am not willing to make this connection to the system ground, but I think that this is why I can't measure accurate voltages. I need support in correcting this situation as we have units deployed and an order in-house for more. I can supply my schematic if needed. Please advise as soon as possible.