Cooling a TO-220 7812 regulator

I suspect that under those load conditions you will get away with it #1. I'm pretty sure the package is well able to handle that dissipation and if it can't it will fold back to thermally protect itself (in which case you will have to add a bit more heatsinking).

I once built a 78xx based PSU that was in a bright bare metal aluminium extruded box expecting it to be more than adequate heatsinking. It wasn't! The bright metal was a terrible radiator and it got hot enough to go into thermal foldback. A bit of matt black paint fixed that.

Thanks. I'm inclined to think that way too. Calculating temerature rise and heatsink requirement for steady-state dissipation is simple enough and it's easy to make a good guesstimate from experience. But transient cases like this are different.

Option 1 sounds fine to me. It will barely get warm.

You don't need it, but internal plane layers help spread out heat too.

The 'up to 4W' burst is timed to 1 sec. I think there's enough thermal mass to keep Tj from reaching a critical temp in that time. But then that's where I'm asking for the group's collective wisdom.

I do that when there's a significant amount of continuous dissipation.

No sensitive parts anywhere close by.

I've done that many times. In this particular case though, the PCB has to be more than an inch away from the wall.

Oh yeah, I've used some huge heatsinks too, when I was into audio and output transistors could easily dissipate several tens of watts each - and cooling fans were not easily available here. And then there was the time I made a limited production using 40A SSRs some 30 years ago.

Thanks for the input. Calculating heatsink requirement for continuous dissipation is one thing. Short bursts are another.

You can estimate the mass and specific heat of a fet and do the math. It's even easier to test one. You could do that quickly.

Given a TO-220 gadget, it would be nice to know the temp rise per joule dumped into it. Maybe I'll measure that.

First google hit:

Based on that, 4 watts for 1 second works out to about 8C rise. The PCB will help a little.

That's good to know. Even if those data are a bit off, it still leaves plenty of headroom for a safe operation including hotspots on the chip.

BTW I think it was you who mentioned Saturn PCB Toolkit in a thread several years ago. I've been using it since then. It's a useful tool.

There was a time when they seemed to have blocked downloads from outside the US. They didn't actually say so and it took me some time to figure it out. I managed to get around that with a VPN. I'm using v8.39 now.

Saturn is great, but the current version defaults to a hideous color scheme, which fortunately can be fixed.

It's not that difficult to weigh a chunk of material and calculate it's heat capacity. We did it back in 1993.

Sloman A.W., Buggs P., Molloy J., and Stewart D. “A microcontroller-based driver to stabilise the temperature of an optical stage to 1mK in the range 4C to 38C, using a Peltier heat pump and a thermistor sensor” Measurement Science and Technology, 7 1653-64 (1996)

and measured the step response of the system as well, with three different Peltier junctions between the point we were stabilising and the heat sink.

It's discussed in the third page of the paper. It does get a bit messy, but it is do-able.

You mentioned single sided board so I would be wary of just having the regulator standing straight up anchored only by the legs. Without plated through holes a s/s board joint cannot reliably support the weight of a TO220 package if the unit is dropped or shaken. My preference would be to screw it down to board if only for mechanical reasons-then the bolt hardware and area of copper underneath provide a little bit of extra heat-sinking it also gives more allowance against someone installing the unit next to a furnace or somewhere hotter than 35degC

I think he has a 10:1 or so thermal safety margin. Most fets are rated for 150 or 175 degC max Tj, and can stand a lot more in real life.

I agree about the standup vibration issue. Why not use a surface-mount dpak?

But yeah, still losts of headroom in this application.

Good point. With TO-220s, particularly those without a heatsink, I usually pull the middle leg out to form a triangle with the other two, and bend them on the copper side for a netter grip. If it's mounted on a heatsink, even a small one, the HS is almost always bolted onto the board.

I considered using the D2pak version but decided against it. One reason is that I live in one of the most remote regions of a third world and we have nothing like Digikey here. So I've made it a habit to use only the most easily available parts. The D2PAK version of 78xx isn't in that category.

Forgot to mention that I've decided to place the 7812 on the copper side and bolt it directly to the PCB. Changed the layout to accommodate the reversed pin orientation.

Cheap people make cheap shots.

Can you receive shipments from the USA?

No universal health care makes the USA distinctly third world.

It's not a cheap defect - health care in the US (when you can get it) costs half as much again more than same health care in expensive places like German, Switzerland and the Netherlands.

It showed up in the Covid-19 death rate. 3642 deaths per million isn't the worst performance around, but it's surprisingly third world.

Once you have bolted it down you might as well flow solder between the heat sink tab and the pcb. This will give you the best possible thermal contact.

John