No, you can always leave out anything you later decide is redundant; it's a lot easier than adding something in later on.
I'd have stayed with the single transistor cut-out; having two transistors is only needed for bidirectional control and you never want any current from battery to D+. Two transistors will double the power loss and heat to dissipate.
Thanks. OK, I went back to one transistor. I also made some other simplifications.
herbie1200 suggested that I don't really need the whole ignition sense and idiot light control thing. The original warning light will still work the same as it always did if left connected to D+ as it is in the old regulator. And, the voltage regulator can go to sleep when D+ goes to 0 (for some time) and wake up on rising edge. (I just went to the garage and checked and D+ is at about 1v at low idle, so it's not 0 when the engine is running.) I can always add bells and whistles later.
I have just started playing around with PCB layout. The original enclosure is about 60 x 50 x 40 mm. Getting all this on a single 60 x 50 board looks tough! So, now I am suddenly enthusiastic about simplification. I think the prototype with through hole parts might have to go on a double-sized board. Later, I can switch as much as possible to surface mount, and maybe populate both sides? Or use long headers to stack a second board (like feathers and hats do)? Maybe a second board that is a metal plate that I can use for heat sinking the big components? Also, the MCU evaluation board takes up most of the space, although I guess all I really need is two 30-pin headers and I could place (short) things between those. But, just placing the components I need from that directly on my board should take up a lot less room. The MCU itself is tiny.
The mechanical regs do not have any other way of cutting the connection than by reverse current, it's a weakness not a requirement.
The cutout worked as a "latching relay", turning on when D+ reached a certain voltage and connecting it to the battery. Based on voltage alone, that would then have stayed connected until the battery went flat, so the cut-out part is a second current-operated coil around the cut-in relay that counters the field from the voltage coil when current flows from battery to D+, allowing the relay armature to release.
That's one area where I'm hoping to gain some efficiency. (Another big one being the regulator curve.)
You could also keep the B+ sense circuit at the battery side of the power switch; as you have the 0.1uF cap on the divider, the resistors could be eg. ten times larger and still get a good reading, avoiding any voltage loss in the power switch.
Sounds good. The 0.1uF value was kinda picked out of the air (and the resistor values certainly were). At one point I was considering using another analog input to measure that voltage loss, but then I decided to instead add a header pin for an external input that I can use for calibration. (E.g., voltage at the battery terminals?) I don't know how stable this will be with temperature. The MCU has an internal temperature sensor, which might be useful.
Anyway, here's the latest:
https://oshwlab.com/carlk3/volt-reg-1-1_copy