I usually have +5 volts available in VME modules, so I generally linear-regulate down from +5 to 3.3, 2.5, and 1.2 for Spartan3 fpga's. VME has lots of power and lots of air flow. My favorite trick is to use an LM1117 regulator with its ADJ pin grounded, to make 1.25 volts. A second LM1117 has its ADJ pin riding on the 1.25, so I get 2.5, all with no resistors.
Now this new gadget: it's an uncooled small box powered from a 12 volt wart. I don't need much 5 volts, so I switched directly to 3.3 and did the same LM1117 thing. Oops. The 1117 has about a 1.1 volt dropout, so I'm getting about +2.2 for Vccaux, sort of marginal.
So I'm thinking, why not yank the regulator and put a diode from 3.3 to make 2.5? So I pulled all the MELF diodes we have in stock. The current draw on +2.5 is about 40 mA, increasing to 50 mA after configuration (XC3S400, running mostly at 64 MHz). So I'm looking for a diode with 0.8 volts drop at 50 mA.
As expected, big 1 amp, low-voltage (100v) diodes have the least drop, around 0.64 volts. Higher voltage diodes, 600 and 1000 volts increase, to about 0.70. So I tried some 1-watt zeners in the forward direction. Bingo. A 5.1 volt Zetex part is 0.84, and an 8.2 volt zener is 0.805.
I wonder what might be the trend of zener forward voltage versus zener reverse voltage. Doping and stuff.
I'm going to spin the board after this batch is used up, for other reasons, so I suppose I'll do it right next pass. Probably go to switchers for most everything.
John