Populating DNP resistors R5 and R6 can solve communication issues

[mohamed-dek1]

While testing new AMDS boards, we observed an issue where the sensor data appeared to be valid at the communication level—no corrupted packets and no timeouts—but the actual numeric voltage/current readings were incorrect.

A strange behavior was found where connecting a Saleae logic analyzer to the current-sensor card’s SPI signals caused the readings to suddenly become correct. Importantly, the communication between the AMDS and AMDC was functioning reliably in the sense that data was being exchanged without errors, yet the values received by the AMDC did not match the true sensor outputs. Because of this, the issue was clearly occurring in the communication path between the AMDS and the sensor card.

After reviewing the AMDC and AMDS schematics and tracing the signal path between the boards, we found that the GND reference on the 3.3V side of the isolation transceiver (U11) was floating. This caused the differential transceiver to operate with an unstable or elevated ground reference, which resulted in incorrect (effectively nonsense) SPI data.

On earlier AMDS boards from previous orders, resistors R5 and R6 were populated. These tie the shield to the local GND, ensuring that the AMDS 3.3V-side ground is properly referenced to the AMDC common. The new boards were assembled without these components, leaving the ground reference floating.

Soldering R5 and R6 onto the new AMDS boards restores a proper ground reference and immediately resolves the incorrect-data issue. After confirming this fix on multiple units, we recommend ensuring that these resistors are populated on all newly manufactured AMDS boards.