Re: [N8VEM-S100:2049] Voltage Regulator

[David Riley <frave...@gmail.com>]

On Nov 13, 2013, at 5:10 PM, Vince Mulhollon <vincemu...@gmail.com> wrote:

> On Wednesday, November 13, 2013 11:07:47 AM UTC-6, David Riley wrote:
> You're not going to get significantly better thermal performance out of a TO-3 attached to a small heatsink than you will out of a TO-220 attached to a similarly sized one (though it is obviously better than zero benefit).
> 
> Check the datasheets and run a thermal simulation...  Its huge like a factor of ten sometimes. Max junction temp 150C and junction to package thermal resistance of 0.5 C/W will always beat max junction temp 125C and package thermal resistance of 5 C/W.  The point of my long explanatory post was it doesn't matter anyway for our project's power levels because the short ckt protection is kicking in long before the thermal protection ckt is kicking in, so thats why its OK to use the TO-220 3 amp regs.  As long as the heatsink can keep up, the junctions will be chilling.

Agreed on the short circuit protection.  My point is that the thermal impedance of a small TO-3 heatsink (to ambient) isn't usually that much better than that of a reasonably large TO-220 heatsink (one that takes up a similar amount of board real estate).  It's more a matter of radiating area; TO-3 packages obviously have a much smaller \theta JC than a TO-220 does.  TO-3 devices are also usually designed for high current, though, so as you mentioned, their protection is usually set to kick in somewhat higher.

> Now if we were turning 14 volt car power into 5 volts, then the numbers aren't so pretty.  That would be like 10 watts per amp, and 3 amps would mean 30 watts, and a cheap TO-220 linear reg at 5 C/W package design would mean the junction roasts at 150 degrees hotter than whatever the heatsink is at (whoops) so you'd have to use dry ice cooling on the heatsink for the junctions to survive, but a TO-3 mounted junction would be a mere 15 degrees warmer than the heatsink, which is nothing.

Fair enough.  But you'd have to be nuts to try to drop 14v down to 5v for a current that large; you might as well be using vacuum tubes at that rate (and they'll take less power than that for the filament).  9v down to 5v isn't pretty either; you're talking 55% efficiency, which is pretty horrifying unless you're making a combo S-100 board/coffee maker.

> Its all moot anyway; TO-3 regs aren't going to be made anymore, maybe ever again, the market is flooded with non-operative TO-3 counterfeits, switcher modules are cheaper than TO-3 + heatsink cost now, high end modern TO-220 packages work almost as well as ancient TO-3 designs and we aren't thermally limited anyway.

TO-3 transistors (not regs, so you'd need space for the reg circuitry as well) certainly are still made, though they're being supplanted by TO-263 (which is much easier to electrically isolate without sacrificing much thermal performance) and surface mount parts.  The counterfeits are a big problem; I've been lucky to get away without running into any, but I've heard lots of awful stories.  And yes, a switcher "brick" is a much better solution where possible, and you can usually get at least a 5-6A switching brick in the space of a TO-3 plus heatsink, without all the heat to dissipate (and some power savings as well).

Really, the bottom line is that a linear regulator is a poor solution unless you REALLY can't deal with the potential cost/complexity of a switching solution, or if you REALLY need a quiet analog line (and you should be running on a much lower Vin-Vout than 3v on modern electronics if you need that!).


- Dave