ADuM4471 secondary 5-volt regulators failing

Dear Sirs,

We're using a total of 18 of the ADuM4471 four-channel isolators to support 9 Integrated Power Modules in a magnet control system.  We were originally using the recommended Coilcraft CR798x transformers.

Everything works just fine.  However, during the open-loop startup sequence of the ADuM4471, the chip overvoltages its secondary side before the beginning of closed-loop operation and this causes the pass transistor of the internal 5-volt regulator to permanently short-circuit.  The external 5-volt load is a constant 20mA, well within specifications.

We can solve the problem by using fewer secondary turns on the transformer, but because the IPM will go into undervoltage lockout at 13.5V, we need *at least* 15.0V from the ADuM4471.  We are using the recommended circuit for +15V regulated output (figure 48 in the datasheet).

After a half dozen of these failures, we are now instead powering the ADuM4471 secondary side from a separate isolated regulated +15V supply.  After a few startups, another internal regulator failed.

These are production-quality circuit boards with careful layout and isolation.  Perhaps the isolator chip is too-lightly loaded at startup?

After reading through the ADuM4471 datasheet several times, it appears the breakdown voltage of the secondary VREG (Pin 20) may be very low, perhaps less than 20 volts which may be causing this problem.

We have decided to clamp VREG to 16.5 volts using a 16V zener and a 2-ohm resistor to protect the ADuM4471.  Our feedback network is calibrated to 15.1 volts.

The overall systems require extreme reliability for safety reasons.  Will the zener clamp solve our problem?  Perhaps an application engineer can suggest an alternative isolator which is better suited to the job, although we've browsed your product selection extensively.

Each IPM needs 6 forward channels and 2 reverse channels at no higher than 20kHz, and requires +15V at up to 50mA to run the IPM internals.  However because of our difficulties we are considering not to use the +15V output from the ADuM4471 to power the IPM.

Can you suggest a better part, or a better solution?  When it works, it works great.  The parts only fail on startup, they have never failed while in operation.

Thanks for your consideration