0
\$\begingroup\$

I have several separate identical PCBs with identical ICs on them. It's fairly critical for this application that all the IC's are synchronized in regards to temperature. E.g. if one increases by 1 deg, they should all increase by 1 deg.

There's two solutions i'm thinking of. One is to simply enclose everything in an enclosure, and create an oven like effect, but this may also create cooling issues.

The other is to use a heat pipe that is snaked across the PCBs and thermal pasted to each of the IC's. I assume that this will help synchronize the temperatures as the heat pipe should stabilize to the mean temperature?

Is there anything else I should be looking at doing?

\$\endgroup\$
  • 1
    \$\begingroup\$ Can you equalize the power dissipation in said ICs? This is an unusual requirement, maybe there is a better solution. \$\endgroup\$ – Spehro Pefhany Aug 7 '18 at 3:41
  • \$\begingroup\$ What's the nature of the ICs in question? If you could post links to the datasheets, that would be the best. \$\endgroup\$ – Nick Alexeev Aug 7 '18 at 4:32
  • 1
    \$\begingroup\$ synchronize verb "to occur at the same time," from Greek synchronizein "be of the same time". The word you are looking for is "equalize". \$\endgroup\$ – Transistor Aug 7 '18 at 7:42
  • 1
    \$\begingroup\$ Is this to reduce part-to-part variation due to temperature? If you ask about the problem you are trying to solve (e.g., which parameter is varying too much with temperature), someone may be able to suggest a simpler solution. \$\endgroup\$ – Justin Aug 7 '18 at 14:29
  • \$\begingroup\$ It's basically a hack we're using to keep PLLs on our chips coherent over a long time period. It's a hack as the chips were never designed for this, but regardless it works well and is very cheap to implement. Due to small variances in the chips, the heat output varies very slightly between them over time causing drift. By using a heat pipe/oven effect i'm hoping to force them to stay all at a mean temperature. I've already tested the oven like effect, and it definitely works. \$\endgroup\$ – Projectile Fish Aug 8 '18 at 4:26
1
\$\begingroup\$

You don't want a heat pipe. It is meant to transfer heat from one place to another, not to keep everything attached to it at the same temperature.

If you want to keep all of the chips at the same temperature, you want a high thermal conductivity material joining the chips together. With unlimited resources, the best that you could do would be to thermal paste them to a solid chunk of diamond, which has the highest thermal conductivity of currently available materials.

Since that is probably not cost effective, the next best possibility would be to use Pyrolytic Graphite Sheet (PGS). It is a layered material that conducts very well in plane, and not so well vertically. The graph below shows a comparison of the thermal conductivity versus copper and aluminum.

enter image description here

You can google around for "pyrolytic graphite heat spreader" and find places that sell it laminated to a backing material for physical support, since it is relatively flexible.

Your most cost effective and easy to obtain solution would be to thermally connect all of the chips using high quality thermal paste and a copper bar as thick as you can afford.

\$\endgroup\$
  • \$\begingroup\$ Isn't a heat pipe basically a way to make a material with very high thermal conductivity? If I Google for heat pipe thermal conductivity I get measures on the order on 100kW/mk, much higher than any raw material. Having a look on YouTube for thermal camera images of heat pipes, I see that a heat pipe very quickly stabilizes to the source temperature at all points, whereas pure metals like copper take much longer. \$\endgroup\$ – Projectile Fish Aug 8 '18 at 4:17
  • \$\begingroup\$ A heat pipe is meant to transfer heat from a hot area to a colder area. For example in a laptop, it transfers it from the CPU to another heatsink which has a fan blowing on it to release the heat external to the laptop. It depends on boiling a liquid at the source which is then condensed at the cooler destination. In order to work, the temperature of the source needs to be high enough to boil the heat transfer liquid. If you want to just keep things at the same temperature, it is not the ideal tool. \$\endgroup\$ – crj11 Aug 8 '18 at 12:22
  • \$\begingroup\$ The heat pipe would work as well as any other thermal conduit. It's more of a cost vs. heat transfer. It does not matter if the heat is transferred to a heatsink or not. Heat always moves from hot to cold. Therefore if the component were cooler than the heat pipe (or other medium) it will absorb heat from the pipe and vice versa. This is essentially what the OP desires. In this case the assembly cost may be the highest cost if it is a production project. Without a photo of the PCB I cannot offer much more. \$\endgroup\$ – Misunderstood Aug 13 '18 at 12:26
0
\$\begingroup\$

Not sure if this is an XY problem, but upon reading your question I remembered this:

Some bench multimeters have a plastic chamber where the voltage reference has a heating element and a temperature sensing transistor:

enter image description here (From "Keysight 34470A 7.5 Digit Multimeter Teardown" by EEVBlog)

LTZ1000 pinout shows heating element between pins 1+2 enter image description here

Included on the chip is a subsurface zener reference, a heater resistor for temperature stabilization, and a temperature sensing transistor

\$\endgroup\$
  • \$\begingroup\$ TO-5 metal cans, they do still exist! \$\endgroup\$ – Janka Aug 7 '18 at 15:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.