Driving a dumb chopper: 24V supply, NMOS H-bridge, 12V gate drive, 50% duty cycle at 200kHz. All I need is complementary drive to the left and right-hand side gate drivers. Trouble is, my gate drive circuitry needs 12V. Old skool metal gate logic would suffice, but fancy that, no one makes special circuits out of it. :)
Options:
- Use, e.g., LT's TimerBlox, or logic (e.g., Schmitt trigger oscillator at 400kHz, followed by a T flip-flop to generate complementary outputs and 50% duty) and level shifter. Ugly, the level shifter wastes all the components the TimerBlox was supposed to save.
- I *could* use a TimerBlox, *if* I use logic level MOSFETs. But my simulations show this will at least double the losses, based on typical MOSFETs. I might go shopping to find others more suitable but Rds(on) will always be higher than with 10-12V drive; the expense is an overkill transistor with disproportionately higher Qg, which means more drive, more supply consumption...
- Skip the gate drive circuit I have, and use one of those VCOs with gate drives included (e.g., IR27xxx), downside being I need four gate outputs, not two. I don't know of any offhand that do that in one chip; the 8-pin models don't support synchronization so this would be a challenge (even more level shifting? fuggedaboudit).
- Do the Schmitt trigger + T flip-flop, but implement it in CD4000 MOS. Downside: I'm pretty sure no one makes CDTinyLogic4000's, so that's two SO-14's almost entirely wasted.
- Use an SMPS controller. I'm intently eyeballing UC3825, the "improved TL494". It's fast enough and has true complementary outputs, but it has limited duty cycle (dead time), as do most PP/ bridge controllers. I can tolerate plenty of deadtime, but it's not as elegant as having simple complementary outputs. It's also an SO-16 or whatever, when all I really need is an SOT-23-6, but still, it's better than two SO-14s.
- Just make a f****ng astable multivibrator*. We live in a world of everything integrated, but hey, what works, works... Plus, it saves one of those damn $5 chips. Not that the placement of all those resistors is any cheaper.
Optimal solution: a TimerBlox (or CD4000 logic oscillator, or..) that runs from 12V. CMOS outputs drive a nominal amount of current (>5mA) at roughly supply voltages, give or take drop. Heck, crude bipolar outputs would be acceptable too (e.g., a 555 with complementary outputs).
*Last time I had this same problem, I built the multivibrator out of the very power stage itself (2N4401 for the win). And the time before that, I had some FPGA outputs to spare, so I just clocked some square waves out of that. Which still required a level shift, but a transistor per phase happened to be enough in that case.Tim