Although the units work at the voltages you describe, we have concluded that we
cannot guarantee such performance over all temperature, supply voltage, and
Now, to clarify the technical point you raise. The reason why our restriction
is important has to do with the way our part works. Internally, we have two
dice. The input die (the "Driver"), send a short pulse to the transformer on
the second die (the "Receiver") whenever it sees an input logic transition.
The magnitude of this pulse is scaled (approximately) by the magnitude of
VDD1-GND1. This pulse then couples inductively to bottom coil of our
transformer. At the output of the transformer, we have threshold circuitry
looking for these induced pulses. The threshold level is roughly scaled up or
down by the magnitude of VDD2-GND2. So, under certain temperature and process
variation conditions, we have difficulty operating when VDD1-GND1 is at 3.0V
and VDD2-GND2 is at 5.5V. The reason is that the driver pulses are at their
minimum in amplitude while the receiver threshold is at its maximum level. We
cannot guarantee that the induced pulses will also get detected by our receiver
because of this.