AD7387 describes a register write sequence where bit 15 is set, D[14:12] is the register address and the remaining D[11:0] bits is the register value to write. The current sequence I've constructed is a hard reset, wait 10 ms, read the CONFIGURATION2 register asserting the default value of 0x000 is returned, then a register write is performed on CONFIGURATION2 with SDO set to 1-wire mode, and finally read back the register confirming the desired value has been set. This sequence was proven to work with the EVAL-AD7386FMCZ connected to a SAMRH71. However, utilizing the same exact sequence and a custom PCB with a AD7387 all register writes appear to fail as confirmed during the read-back step. I've attached a waveform of the failure to write. The only major difference I can see between the ADI demo board and the custom schematic is missing 100 ohm resistors on the SDO line, but the signal is not marginal so I do not think this is a problem and I've clocked the bus as slow as 250 kHz. I can consistently read from the chip as confirmed by looping reads of the ALERT_HIGH_THRESHOLD which returns the expected 0xFFF. 
