79L05 stability, output capacitance

I use the 78L/79L in an old product.

The 79L brings back bad memories of a "bad batch" from Jameco with intermittent power-cycling start-up problems. Intermittent problems are the worse to trouble shoot.

We've bought ~1k since then and no problems. George H.

Sure, to drop a few volts so half of the dissipation is distributed into this zener. SOT89 on the negative regulator is north of 100C/W.

I've done worse. Like sliding a Skil saw a bit sideways to cut a wide groove. Did it for a neighbor and she exclaimed sort of in disgust "I saw what you just did!". But the piece of wood was ready to go without using a router which we didn't have.

Note to everyone: Don't do this!

But it would feel like when Gloria Stivic (All in the Family) lost a bet and was thereby forced to eat a Lima bean.

-Lasse

It's the shunt version of a lead-lag feedback network. The big cap with the ESR hauls the gain down with a low corner frequency, then goes away and lets the ceramics and plane capacitance look after the high frequency stuff. For that to work, the lossy cap has to be much bigger than the total amount of fast capacitance.

A really big aluminum electro works well with biggish ceramics, but given what high value tantalums cost, the temptation would be to use too low a value. An Al-poly with a series resistor might be a nice solution.

Cheers

Phil Hobbs

Look at all the band-aids on his fingers! And is one pinky kind of short?

I meant using a zener *instead* of the regulator. No stability problems!

I decided that a SOT89 transistor can dissipate around 3 watts if it's soldered to a big copper pour, but it sounds like you don't have room for a big pour.

I'm playing around with a circuit that will push some SOT89 phemts pretty hard, 2 watts maybe. The possibilities are

Use one of those winged surface-mount heat sinks, over the phemt.

Cut slots on the board under the wings, and the exhaust fan (neg pressure on top of the board, fresh air entering box from below) will blow air directly into the wings.

or

Via the heat to a copper pour on the bottom of the board. Stick a pin-fin heat sink there, on the bottom. Arrange air inlets in the bottom of the box to blow directly onto the pinfins.

In high speed stuff, sometimes you just have to push small parts.

The magic number is around 2.5 x.

Sum up all the low-ESR bypasses on the rail. (You don't ever need to use more than 0.1's scattered around. Even that's probably not worthwhile, it's just common.)

Use a bulk cap >= 2.5x that total value.

Use ESR equal to the total bypass reactance at the crossover frequency.

Frequency is the hard part to find out, for an arbitrary regulator, but you can probably estimate the loop GBW from the impedance or phase response of the regulator. Or, just use whatever ESR the datasheet recommends.

Tim

The '723, qua regulator, was unlikeable (flexible, but too many parts). It had a temperature compensated zener, though, and was a dandy low tempco standard for voltage, and THAT was/is quite useful. Pdip '723 is still a 150ppm/C reference for under a quarter in thousand qty.

What do you mean, too many parts? The Motorola thinggie was a POS and used over twice the parts, umpteen extra diodes, resistors, etc in a tangle ready to choke the IC.

So is a 78M05 and (probably unlike the 723) you'll still be able to buy it in 10 years cheaply from regular distributors.

--sp

But that would make global warming :-)

Nice. The ball is especially cool. In my case it was a hand saw where there is the additional risk that it bucks and then wants to fly into your face.

Nearly all people I know who lost fingers did so while using a circular saw. One while lifting a running lawn mower off the ground.

I will have a large ground plane but the net where the tab connects to isn't going to be great. Also, the negative ones have a worse heat sinking resistance versus the positive ones. Probably the tab isn't the main path.

When I need this much I use TO220 and stand them up.

This unit won't have air flow. Usually I even try to design without because evetually all that clogs with dust or a fan bearing starts to go. And it's noisy.

That is a good method if you can get the heat into the ground plane.

Luckily most of my apps are pulsed. Just not this one.

Why not? It's just a dado (or plough the other direction). It's perfectly safe.

Jameco is basically floor sweepings, no?

I don't know, I forbid any more Jameco parts.

(so far that's worked.) George H.

Not exactly; 78M05 is 200ppm/C typical, '723 is 150ppm/C maximum. Using a LM317 (or any of the 78xx) as output transistor, for a '723, is a bit of a win; your error amp is not next to the power dissipation, and current in the LM317 output terminal generates an IR drop that's INSIDE the feedback loop instead of outside.