I got a test board today from OSHPark. This one is an LCD with a VCOCXO controlled by a 20 turn pot. The board has a pair of D-PAK 78xx regulators on it from a 2.1mm barrel connector.

Well, I didn't do the math properly on power dissipation. The 7833 almost instantly got rocket hot and shut down.

Let's review: 1A drop from 6V to 3.3V is 2.7 watts. Turns out, that's too much to ask for a D-PAK package. I'm going to have to go with a TO-220 through-hole regulator with a heat sink on it.

Some will ask why I don't just use a buck converter. That's a valid choice, and I've done them before. But at high currents like that, keeping the power clean is a challenge, and this application is going to be quite sensitive to that sort of thing.

For this test board, I'm going to just tack on a TO-220 package regulator and see how the rest of it works. The board around the OCXO was getting quite warm, even though the oscillator never actually started properly. That may wind up being an issue down the road as well. We'll hopefully know more later tonight.

EDIT: I wasn't able to get a TO-220 3.3v regulator this evening, but I've temporarily separately powered the 3.3v rail from my bench supply. It's not the most stable thing in the world, but it works. With it, I've been able to tweak the 20 turn pot to get it down to exactly 200,000,000 cycles every 10 GPS seconds - within 5 ppb of GPS, and it's stayed that way for a half an hour so far. Now, I have no idea how much phase noise is being introduced, but at least for this build, that's not really the point.