I'm confident that the Spice simulation shows what is shows. Moreover, everything you've stated about the propriety of using a P-fet vs. an N-fet is quite true, though, as I pointed out, P-fets are not so ubiquitous. You can order 'em here in the U.S, and have them in a week, but where the O/P is, he may have different results.

I had the N-fet placed in the circuit in question because we had 'em on hand and they were in an convenient-to-apply package. Since the current was small, I concluded power dissipation wouldn't be an issue. Having "gotten by" with this before, I didn't see any reason not to give it a try. This was also not for a production application. It certainly would have gotten more effort had it been so.

While there's normally a considerable Vds at substantial currents, this isn't consistent with what's been observed in the circuit that's been used here, probably because the load current is much smaller.

I'd be surprised if the current required to run an MCU exceeds 200 mA, and we don't know what the O/P's load current will be. In the application in which we did this last week, 5 volts was not required, though desired, and the current was, small. However, the N-fet that was used not only produced adequate output voltage, but had Vds clearly below what the simulation suggests for the device you chose, which has "better" spec's all around. I suspect the load current is a factor.

Again, I don't disagree with your suggestion that a P-fet would be a more effective device in this application, but, having seen that the O/P isn't here in the U.S. where "everything is available," I suspected he'd have less difficulty with finding a logic-level N-channel device than a P-FET. My own recollection of browsing surplus electronics shops, not that those are the only option, was that I never saw a P-Fet, though I routinely saw different types of N-Fets. I certainly don't recall seeing one when I actually thought one would be of use. Right now, I believe I have exactly one type of logic-level P-Fet on hand, and it's on the same die with its N-channel complement.

It would be interesting to see what LTSpice thinks would happen if the load were 51 ohms rather than 1 ohm, since you already have the simulation set up. After all, for this sort of application, you don't want a device that loses one volt of Vds irrespective of Ids, since that has to go somewhere, probably as heat.

RE