From: Keith R. Williams
Subject: Re: TO-220 on a big aluminum sheet
Date: Fri, 3 Jan 2003 23:15:06 -0500
X-Newsreader: MicroPlanet Gravity v2.60
In article ,
> On Fri, 3 Jan 2003 14:30:45 -0500, Keith R. Williams
> >In article ,
> >jjlarkin@highSNIPlandTHIStechPLEASEnology.etc says...
> >> On Fri, 03 Jan 2003 04:39:57 GMT, Spehro Pefhany
> >> wrote:
> >> >On Thu, 02 Jan 2003 15:57:10 -0800, the renowned John Larkin
> >> > wrote:
> >> >
> >> >>
> >> >>see a.b.s.e.
> >> >
> >> >The nomograms in the GE SCR manual for doing this sort of thing are
> >> >pretty messy looking. Thanks for the info.
> >> >
> >> >Best regards,
> >> >Spehro Pefhany
> >> The other cool thing to try would be to plot theta versus plate area;
> >> but too much work. It's a shame that thermal modelling software is so
> >> expensive.
> >Why not use Spice or some other circuit simulator? The thermal
> >gurus I worked with 25 years ago at IBM used IBM's circuit
> >simulator program (ASTAP) for some pretty sophisticated thermal
> >modeling of multi-chip packages, even entire mainframes.
> > Keith
> Spice is hard to use for distributed (diffusion model) situations with
> geometry issues. A real thermal modeling program would chop the sheet
> up into hundreds (at least) of finite-element chunks, and do the
> conduction+convection+radiation analysis of each until the thing
> settles to equilibrium. If there is enough heat to cause significant
> air motion, we add a fluid dynamics analysis on top of that.
It was all done on circuit sims. I thought it rather amazing
myself, but libraries make it all come together.
> I have used Spice to model simple thermal situations, especially for
> transient (pulsed) heating. But you've got to chop up the metal into a
> small number of idealized (isothermal, ie capacitor) blocks, and
> that's not often very realistic.
Even before the TCMs (Thermal Conduction Modules) they were doing
LEM (Liquid Encapsulated Modules) dynamic simulations. Each had
thousands of pins (all modeled) and LEMs employed a phase change
so the computations were nuts. The problems were on the same
scale as the circuit sims I was doing at the time; perhaps an
hour or six of mainframe (360/85) time.
> Oh, incidentally, this is from one of my engineering notes files...
> TRANSIENT THERMAL MODELLING
> Analog circuit analysis techniques can be applied to thermal systems.
> The following equivalences are accurate to about 10%...
> ELECTRICAL THERMAL
> 1 amp 1 watt
> 1 farad 1 gram aluminum
Interesting that aluminum comes out this close.
> 1 volt 1 degree C
> 1 second 1 second
> 1 ohm 1 degC/watt
Theremal resistance is, well...
> So, with appropriate unit scaling, Spice can be used to model simple
> thermal systems.
and indeed some very complex ones. The crank-handle may need to
be a bit longer for some, but thermal analysis is a parallel to