transistor with a lot of beta

The 2N5087 (NPN) and 2N5089 (PNP) are hard to beat, I think. Have high beta from Ic from 1uA to 50mA. But Vceo is just 25V. Probably OK for most things anyway.

The BC547C (NPN) and BC557C (PNP) are even higher gain for Ic >1mA to 100mA, and Vceo of 45V.

-RS

I don't think either of those have high enough Ic absolute maximum ratings for this application. Passing ~ 300mA.

I think we're looking at something in a TO-126 package at least. I'd like to avoid a TO-220 if possible.

I would suggest a jellybean switching buck regulator to save you a lot of dissipation. All modern digital boards have those.

and if you can afford a little more money (like +0.02USD re 2N5087), the MPSA18 (NPN) is a real hot cat. High betas from 10uA to 10mA. Vceo of 45V, Ic-max 0.2A.

That's definitely a possibility.

Given the power requirements for the tubes, however, it's likely going to be powered via a wall-wart, so efficiency doesn't matter as much.

Oh, OK, I read your post to mean each tube was being separately controlled (& so Ic at 40mA).

For Ic around 0.3A then the 2N4401 is fairly good, beta ~ 150. But with Vce of 3V, you'll be asking it to the dissipate 0.9W (too much for 2N4401 alone - 0.625W). Either add a series voltage-dropping Resistor, or move up from TO-92/SOT-23 type package.

But isn't that 300mA made up from 7 x 40mA? - So why not just spread the loading out with penny-cost cool-running jelly bean per filament rather than a boutique dollar-cost hot-running power device?

piglet

absolute maximum

Oh, OK, I read your post to mean each tube was being separately controlled (& so Ic at 40mA).

For Ic around 0.3A then the 2N4401 is fairly good, beta ~ 150. But with Vce of 3V, you'll be asking it to the dissipate 0.9W (too much for 2N4401 alone - 0.625W). Either add a series voltage-dropping Resistor, move up in price range, or move up from TO-92/SOT-23 type package.

The Zetex (now Diodes Inc) ZTX618 is beefier, handling 1W in a TO-92. But price is significantly higher.

That's certainly a possibility. Unfortunately the data sheet has little information (none) about how much current the reference pin can source/sink.

Ideally, I guess if it's going to be an offline thing, the best plan would probably be to use a transformer/wall wart with a low voltage output and voltage double part of that that for the main supply, and then run the filaments off the filtered but unregulated main voltage. If that makes any sense.

er, run the filaments off a filtered but not-voltage-doubled part.

Okay. It's often said on this newsgroup to state "what you want to do", because you can lose the forest for the trees in looking for solutions.

Required voltages and currents: +5 at a few mA for display IC, +5/-5 at a few mA for analog section, +2 at around 300mA total for filaments, around +10-12 at about 50mA for tube anodes.

Offline supply - some wall wart. If it's battery powered, then it's going to have to be a switcher and a big battery.

Are the filaments isolated from the cathodes? Because if so, you can put filaments in series for convenience.

Why not use a +12V supply and generate ground at 5V below the +12V rail? Then you'll have +5V, -7V, 12V total for the anodes, and you're only left with the task of finding 2V for the filaments. I'd be tempted to see if I could series-connect them across the 12V.

Oh, and level-shifters for the anodes, but you knew that.

As you have seen in the other responses, the dissipation *is* a potential problem. 1 Watt makes a small component run quite hot.

The tubes are direct-heated: a filament, not a heater. In this use, there is little to be gained from indirect heating, but plenty to lose (size, power).

To the OP: Is a LM317 or similar out of question?

Unfortunately, it looks like the filaments are the cathodes as well, not a cathode + heater type deal.

I could always current regulate each cathode individually off the +12 instead of voltage regulate. LM317LZs in TO-92 are about ten cents each...