Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Potential heating problem




Over-etched pcb's can be a significant problem. We have rejected boards with
traces etched 40% below the nominal width. If someone starts out with traces
narrowed to fit between pads, etc. it is possible in production to see traces in
the .004-.005 range unless proper manufacturing practices are maintained. It would
be useful to inspect actual "failed" parts, if any exist. It is, however, not our
mandate to make something work over every possible combination of "junk". For
instance, the bulk resistance spec. should adequately filter out extremely high
resistance traces, if anyone actually performs the test.
Also, the trace heating calculator seems to yield results that do not agree with
my actual measurements, and the trace resistance varies depending on what current
and calculations are used. I think any of the methods we have discussed can only
be used to an order of magnitude range, and not to be relied upon too strictly.

Sterling Vaden

Jack Andresen wrote:

> With Stirlings high current tests and all the analysis by others, I
> believe that trace size is not the culprit we seek.
>
> The fact that several people have reported failures is very important.
> What other mechanisms could explain the failures? The failure of same
> boards may be due to a variety of other causes. A couple,
> 1) Poor interconnection. i.e. a high resistance connection, RJ plug to
> RJ jack.
> 2) Poor PCB plating. A high resistance through hole connection.
>
> Anybody have other ideas?
>
> Jack Andresen
> *******************************
>
> Comments
>
> "Schwartz, Peter" wrote:
> >
> > Jack:
> >
> > Perhaps the rather handy little calculator available for free download form
> > this website will help you out.  The specific file of interest is
> > PCBTEMP.ZIP.
> >
> > http://www.contrive.it/Download.html <http://www.contrive.it/Download.html>
> >
> > Peter Schwartz
> > Applications Engineer
> > Micrel Semiconductor
> > Phone:  408.435.2460
> > FAX:    408.456.0490
> > peter.schwartz@xxxxxxxxxx
> >
> >         -----Original Message-----
> >         From:   Jack Andresen [SMTP:jandresen@xxxxxxxxxx]
> >         Sent:   Monday, May 07, 2001 13:22
> >         To:     Sterling Vaden
> >         Cc:     Dave Dwelley; stds-802-3-pwrviamdi@xxxxxxxx
> >         Subject:        Re: Potential heating problem
> >
> >         Stirling,
> >
> >         To understand your figures. Your traces are 1.4 mils thick because
> > they
> >         are internaland have no overcoat, correct?. The traces are 10 mils
> > wide
> >         but how long and how much space between traces.
> >
> >         By my calculations, your traces have about .75 ohms per foot. I get
> > .43
> >         ohms from your stated voltage drop and the current.So the total
> > length
> >         of your traces is about 6.9 inches. Am I in the ball park?
> >
> >         With your total of 2.95 Watts, the equivalent surface are is about
> > 5.4
> >         sq in. with your 60 degree C over ambient, However, you say the
> > surface
> >         temperature is 90 to 107. How can the surface temperature be higher
> > than
> >         the temperature source, namely the PC traces. Please explain
> >
> >         Jack Andresen
> >         ****************************
> >         Sterling Vaden wrote:
> >         >
> >         > I have been running a little informal test of current carrying
> > capacity. I
> >         > am currently (get it) running 2.6 Amperes through a pair of traces
> > (.010 X
> >         > .0014) wired in series, embedded in the inner layers, one directly
> > above the
> >         > other, of the test board shown. The temperature rise is
> > approximately 60
> >         > degrees C over ambient yielding 98 to 107 C as measured on the
> > surface of
> >         > the pcb. No discoloration of the board material has been noted.
> > The traces
> >         > are dissipating 1.138 X 2.60 = 2.96 W. No catastrophic failure has
> > been
> >         > noted.
> >         >
> >         > Now. I predict that if traces are failing at .250 A, those must be
> > small
> >         > traces indeed!
> >         >
> >         > Sterling Vaden
> >         >
> >         > Dave Dwelley wrote:
> >         >
> >         > > Do these miniature traces affect our 20 ohm wiring budget?
> >         > >
> >         > > What's the typical DC resistance of a patch panel?
> >         > >
> >         > > Dave Dwelley
> >         > >
> >         > > Jack Andresen wrote:
> >         > >
> >         > > >But let me go back to original issue. I have looked at many
> > patch panel
> >         > > >PC boards and for various reasons, the traces run from 10 to 18
> > mils.
> >         > > >Possible reasons:
> >         > > >1) Making pairs of traces 100 ohm. 2) Trying to get between the
> > RJ45
> >         > > >pins. 3) Running a pair between the 110 punch down pins. 4) In
> > general,
> >         > > >there is very little space for both compensation and wiring
> > while
> >         > > >leaving margins between pairs of traces. 5) People are afraid
> > of too
> >         > > >fine traces
> >         > > >But it is important to recognize here is no general standard
> > for patch
> >         > > >panel traces (as pointed out in one of the responses).
> >         >
> >         >
> > ------------------------------------------------------------------------
> >         >                              Name: pcbCurrentCapacity.doc
> >         >    pcbCurrentCapacity.doc    Type: Winword File
> > (application/msword)
> >         >                          Encoding: base64