[sdiy] temperature testing

[Neil Johnson]

Hi Brian,

> The key is that if any part your circuit does overheat, then you'll need to fix the design. You need multiple thermocouples in order to find out which chip is the one that's overheating, and that in turn will tell you which part of the circuit needs adjustment.

You don't need a thermal oven to find out which chips are overheating.
A finger tip will usually suffice, or if you don't like pain then a
thermal imaging camera (quite cheap nowadays from China).

> There's also the fact that different chips are rated for different maximum temperatures, so you'll need to know the precise temperature of each chip to know whether your product has "failed." You can't wait for total failure of the product itself, because not every chip is going to instantly stop working just because you've gone above its temperature limit. However, you want to make sure that your circuit design prevents each individual chip from exceeding its ratings, otherwise you have no guarantee that your product will last.

Commercial grade components are specified to 70 degrees C, die
temperatures maybe to 125 C. The main purpose of a thermal oven is to
see how a circuit behaves at temperatures other than ambient.  As the
designer you should already know which components are going to be
dissipating heat and design accordingly.  A chamber can then be used
to verify your design.

For example, one time I had a product that would get hot.  We knew it
would get hot, and designed in heat spreaders.  What we wanted to find
out was how hot the external case would get at elevated ambient
temperatures, say 45 C.  Some 24 hour tests in a chamber provided the
answers.

Another use I've put my "chamber" to was to run extended memory tests
on DRAM at high ambient temperatures.  Again, this was nothing to do
with finding out what got hot - we knew that - but wanted to find out
what effect temperature had on DRAM reliability.

> I've seen the process and the mess of thermocouples on prototypes, but I've never actually run these tests myself. I'm guessing that if the ambient temperature exceeds the chip ratings, then there's no circuit that could pass the test without refrigeration.

Ambient rarely exceeds chip ratings, unless you work in a chippy.
What you can do is following thermal calcs determine whether the
system can work reliably at the design's stated maximum working
ambient temperature.  For example, a synthesizer module might be
expected to work reliably in a case with a large power supply, where
the internal ambient temperature has risen to, say, 50-60 C.  A
thermal chamber allows you to run your module at that temperature
repeatably for extended testing.  For example, you can test your VCO's
stability with temperature in a more reliable and repeatable way than
you can with a hair dryer.

> Standard fans can only help when ambient air is cooler than the parts need to be. I've seen the reports, and they mention which chip exceeded its temperature limits, and under what conditions. This usually allows the circuit designer enough clues to reduce the heat in the problem areas.

A thermal imaging camera gives a very good picture of where the
hotspots are.  Definitely easier way to find hotspots than poking
around with thermocouples!

Neil
-- 
http://www.njohnson.co.uk