opamp thermal errors

One method that I sometimes use in temperature controllers is a compound am plifier: first stage barefoot, mounted on the cold plate; second (much fast er) stage connected as an integrator, mounted on the board. That keeps the first stage's bias very nearly constant, both input and output, and so virt ually eliminates forcing due to thermal feedback via the amplifier.

Cheers

Phil Hobbs

The bias current is more sensitive than Vos to temperature. I recall computing that a '741 style op amp would, if input impedances were different by more than 4k ohms, have thermal errors dominated by the input bias current. I changed a misperforming design to reduce the effect, and it behaved better after that...

That's a zero drift type so it should cancel out the voltages it sees. But that doesn't account for thermal EMFs on the package.

YOu might get away with it if you pay attention to layout symmetry on the op-amp inputs, espcially if the power dissipation isn't changing too quickly or you can tolerate transient errors.

--sp

Luckily, my customers will usually measure zero offset with zero output, using a DVM. That's the ideal case for low offset!

Hmm, would you rather have smart or "less smart" customers?

George H.

Keep it cool with thermal bath of liquid?

Mercury would be good ...

Way too messy. Potting would help, but that's too messy for production too.

Venus is more "sexy"...

Why would you want to add insulation? If you can maintain a non-porous environment, then back-fill such an enclosure with Helium; failing that maybe Argon. With a little ingenuity, turn one surface (mounting one is ideal) into an internal part of a heat pipe; back-fill a suitable liquid (water and Ammonia are good except for the corrosion characteristics).

Epoxy conducts heat about 100x better than air.

Any sort of liquid cooling is a mess.

The easy way to make electronics is to solder parts onto circuit boards.

On Tue, 18 Aug 2015 20:55:37 -0700, John Larkin Gave us:

Alumina fiber filled epoxy even better..

We used to mix our own for potting HV power supplies on a couple of our designs. Since they are all pretty must custom, per application designs. It isn't something we do with all the designs. And I do not work there anymore anyway, so I do not know what they are doing now.

A drop of this epoxy, with a metal plate added on top would make a nice, flat interface surface for an added fin device as well.

I'd expect that epoxying a small plate of some sort to the top of the SOT23 would drop theta a whole bunch all by itself. But that's still messy in production. Looks like I'll have plenty of margin if I use two LM8261s as the power buffers, working in parallel, with the V- pin heat sunk to a PCB plane.

On Wed, 19 Aug 2015 07:10:48 -0700, John Larkin Gave us:

The metal plate is only there temporarily to make a flat interface surface. You remove it after the cure, and to facilitate that you can spray it with food grade dry lube first. The flat surface it provides levels the epoxy at the component profile height so what you end up with is a sot23 device surrounded by epoxy, but with a nearly bare top. Then any sink you add on top of that carries away heat as well.. So you get a thermal epoxy shroud around it and a flat surface to support the attachment of a small sinking device on top. It is a post assembly manual operation and very labor intensive. But as you already stated, it works better than air.

It is similar to what motherboard makers do when they glue a heat sink on top of a chipset chip in the motherboard.

I stuck a pin-fin heat sink to the top of an FPGA and measured the drop in chip temperature using an internal ring oscillator.

Then, just for fun, I replaced it with a piece of flat aluminum about the same size. Same temp drop. The pins are useless in free air, but the heatsink base spreads the chip hot-spot temp, as does the flat plate.

But with a fan, the pins do help a lot.

Pin heatsinks are a snare and a delusion, I expect. They have the same surface area as a normal finned heat sink, and many times the flow resistance due to all the turbulence.

Has anybody done measurements to support pins over fins?

Cheers

Phil Hobbs

only advantage I can see it that you don't have to consider orientation with regard to airflow

-Lasse

They do look cool.

I'm guessing that they work pretty well in my case above, with a fan blasting straight down onto the pins. Air can squirt out in all directions.

But no, I haven't measured it.

With a (flat) fin, there can be a lot of dead air space right at the surface. Don't you want turbulent air flow,to guarantee good heat transfer?