> ...you also have to know at what temperature (over ambient) that the board
> will be running at...

Remember, what the chart is saying is that if you apply a specified current
to a specified trace size, the temperature of the trace will rise by a
specified amount.

> ...As an example, say I need 5amps and 20 degrees over ambient is a safe
> assumption. The chart gives ABOUT 100 square mils, which corresponds to
> ABOUT a 75mil trace width on the 1oz copper curve.

Are you sure a 20 degree rise in that trace will be acceptable?  If your
equipment is designed to operate at 50 degrees C, that trace will be 70.
You might also have to take into consideration what else might be thermally
coupled and where hot-spots might occur (just because the average
operating temperature inside an enclosure is 50 degrees does not mean that
it is uniformly distributed).  Also, kinks and bends will cause local hot
spots within the line.

Anyway, I tend to lean towards the conservative anywhere I can.  If I have
to neck down to the minimum I will but for as small a length as possible.
In this particular case, I would use a 100 mil trace width if it is for a
limited length, (say less than 6 inches) had good air flow, and no
significant thermal coupling.

I assume this is a power connection and not a signal so the electrical
characteristics of the line (resistance, etc...) that may change with
temperature are not of real concern.  If it were, you would have to check
with your engineer about what would be an allowable rise (especially for
higher freqs).


     ^^^^^^     [log in to unmask]       "It's a damn poor mind that can
     |o  o|     voice: (512)251-2341          think of only one way to spell
---000-()-000---fax:   (512)251-0498          a word."   --Andrew Jackson