I'm not aware of major Maxim sourcing hurdles in the last couple of years -- above and beyond present circumstances, I mean. In the more distant past they've been rather notorious for this, and in the recent past I don't know that they've been much better at it (the only signal otherwise that I recall being a press release basically saying "we're better now, we swear", which seems free to disregard..). Any obvious changes since ADI acquisition?
For my part, I've done the mostly-discrete version:
The notes show the layout and schematic, more or less. I like to do a few quick sketches before carving up some copper clad. The circuit I breadboarded, to get the operation I wanted; it's free-running, not saturation-commutated (Royer), nor resonant (Baxandall / "ZVS"). The R + (R || C) feedback gives a fairly stable frequency with respect to supply voltage and load current.
This board is elevated above the metal chassis (double-stick tape and some blank FR4 strips) for isolation; the active circuit (MCU and stuff) is actually at unknown voltage (potentially mains)
Also, note the strips of foil at the bottom: can't really see them here, just the relief, as they're buried in solder, but they wrap around the edge, grounding the bottom side copper. Not a bad idea for improved EMI immunity on a carved-clad build.
(first tentative schematic)
Notes at the bottom show transformer small-signal parameters (that's turns on a P22 pot core), and excellent performance of the circuit: 8V OC, 4.3-5.7V into 22-1000 ohm load.
(final schematic, layout sketch.
Diodes became BAT54S on the build, BAT85 was breadboarded. Pretty much the same thing, anyways.
Note that this feedback network reduces quiescent current at DC, so if the transformer goes shorted (quenching oscillation), operation is hFE limited -- still an extremely crude limit, but some kind of current limit at least.
I rather like how simple the oscillator and rectifier layouts were. And the MAX232 and SFH6345s are no-brainers. Lots of wasted space because, hey, it's hand carved, easy is good.
Overall, this would be more expensive to produce (with these components) than with digital isolators (like the SI, AD or TI families), so I would recommend them in that context. This was just a prototype, and I happen to have all these things handy. The DC-DC converter is basically the worst part of it; there are drop-in solutions of course, but they're all over a couple bucks (in small quantities), and even then most aren't high voltage (2.5kV+) or reinforced. (Murata datasheets in particular you must read carefully: few of their parts are reinforced, and the voltage rating they give is usually Functional or Basic.) Plain transformers aren't much cheaper either, unless you get them in quantity and/or from China. Basically, if you want Reinforced, you're paying for the qualification moreso than the part.
(I suppose since I used a piece of TIW, this should be Reinforced, though it's salvaged so may be nicked. The enclosure is grounded anyway so I'm not worried.)
Regarding the MAX232, and also the MAX253 if you aren't using naked transistors -- these have been reliably produced for decades, and show no sign of stopping. 232 in particular is multiple sourced, along with similar devices from a half-dozen other manufacturers (with various capabilities including 3.3V operation, internal ESD protection -- or the lack thereof!, etc.). Of the MAX253, TI has a SN6501 equivalent, though they don't seem to be producing very many (or is it being phased out, I forget). Not sure who else makes similar.
Tim