Until the program is loaded into the DDS, it will latch the output of the PA on, causing an overcurrent failure of the transistor.
In this doc (
https://app.box.com/v/3ysyx-444reop) in Figure 17 on page 18,
I show the behavior of the gate (which means that it is coming straight from the DDS through the FET driver). It goes through some sort of power-on behavior which causes the gate to pulse and since the drain voltage is already applied, the transistor turns on, etc. Unfortunately, the gate (and thus the DDS) behavior was not at all consistent - each time you turned it on, it looked slightly different (different pulse length, different number of pulses, etc.), which lead me to believe that it is not the result of some sort of UVLO or POR, which I would think would be the same each time you turned it on.
My solution was to add a drive loss protection circuit, and I have not lost another device since.
What you describe is the elegant and preferred solution, IMO. The caveman solution in this case (in addition to protection passives all over the place) was to change the power supply sequencing so that the DDS supply goes on first to let the DDS and the gate go through its (unpredictable) turn on sequence then - after some period of time where it is safe - apply the drain voltage while the gate is on but still idle. Obviously if you wait too long to apply the drain voltage, the gate will already have a squarewave drive applied and you risk problems not unlike the problem you are trying to solve.
I did not do this in the end because I thought that the modifications would be a bit advanced for some and I was trying to keep things on the simple side.
The later versions of his transmitters I noted the elimination of the modulation transmitter, and instead feeding the PA directly from the modulator. This may well have solved the problem, as the modulator has overcurrent sensing and protection built in.
My understanding is that his 100W TX uses a PWM.
I have a PWM circuit in the works that is meant for a wide-variety of mid and low-power transmitters, and I suppose could be an add-on to replace the Class-D amp and modulation transformer in a Stretchy 40W TX. It generates its own internal and external DC voltages and has a simple power sequencer (TI LM3880) to apply those voltages in an ordered manner, creating a safe and predictable turn-on and turn-off sequence for the whole TX. The revision 2 PC board is back from the fab house but I am focused on other things at the moment and haven't populated it yet.