There are two issues I’ve observed with v4.4 that I’m going to have to fix with another update. One is that the charge pump has some difficulty keeping up with the choppy current demands of the pilot generator. The issue is the 10 kHz clock rate of the chip. Fortunately by just strapping pin one to Vin the clock rate can be increased to 45 kHz. Early experiments seem to show that this is good enough. The bottom edge of the pilot signal seems to take a little extra time to go all the way down, but in principle the car never sees the negative portion anyway, so unless it winds up being a problem I’m going to declare it good enough. I’ll try adding more reservoir capacitance to the prototype to see if it helps and if it does, that’s just a component upgrade.
The other issue is that our Chevy vehicles seem to cause false GFI trips when power is first applied. I’ve spent some time simulating the circuit and now believe that the 10 µF cap I removed from the circuit some time ago may have served as an integral function. I’m going to try hacking it back into the circuit to see if the falsing goes away.
If both of those issues are taken care of, then v4.4.1 will be the one for the foreseeable future.
On the AC board side, v4.3.3 is likely to be good enough (with parts substitutions to decrease the GCM sensitivity), but I am going to try an experimental version substituting an LM393 comparator for the LM358 op amp. On that board, a comparator is actually what the circuit wants, and by leveraging the open drain output, we can reduce the parts count slightly, at the cost of inverting the sense of the signal (so a corresponding #define in the firmware will be needed).
UPDATE: The GFI fix seems to have worked, but longer term testing will be required to absolutely confirm it. The added capacitance on the charge pump, however, didn't do any good. Really, though, the only alternative would be going back to the MC34063 based inverter, which I might do, but I still think this is good enough for now.
UPDATE 2: It's a Christmas miracle! By changing the MOSFET turn-on resistors I've reduced the switching current enough to make the charge pump happy. The rise and fall times increased just a bit, but I believe it's still close enough to the spec requirements to score.