NetNews Usenet Archive 1992 #19/NN_1992_19.iso/spool/sci/electron/14956

Internet Message Format | 1992-08-29 | 3.8 KB

Path: sparky!uunet!cs.utexas.edu!uwm.edu!ogicse!verdix!bit!thomas
From: [e-mail] (Thomas Walter)
Newsgroups: sci.electronics
Subject: Re: How hot can the 486-50 cpu run ?
Summary: Silicon - max temperature and how to figure JUNCTION temp.
Message-ID: <[e-mail]>
Date: 28 Aug 92 17:20:22 GMT
Article-I.D.: bit.414
References: <[e-mail]> <[e-mail]> <[e-mail]>
Organization: BIT, Portland, OR
Lines: 80



I can't take it any more, so here are my thoughts on "how hot will it
run".

Q: How hot will that silicon run?
A: Good question, how long do you want it to last?

Seriously, some silicon will function up to the 200C mark, but you
may find your socket melted ! There are factors such as how robust
the design is, and what sort of process is used. One of the oil drilling
companies poured a ton of money into developing a IC sensor that could with
stand 300C but that was not your run of the mill silicon sensor! [To my
knowledge it never was developed, great opportunity for a niche market!]

For reliability, it would be best to keep the chip cool as possible. At the
higher regions it may start to act "flaky". If you were to test the IC at
various temperatures, versus supply voltages, and different temperatures
you would get a series of "schmoo plots" showing the typical operating ranges.

I've worked with automotive IC's that were "hot chucked" at wafer sort to
140C. The idea was to screen any parts that did not perform at temperature,
before they were mounted in a package, or on a hybrid.

Motorola specified some of its ECL parts to have maximum temperature junction
of 165C for ceramic packages, 140C for plastic packages. (MECL 10K/10KH family)

PACKAGE:

The junction temperature, Tj, for a given junction-to-ambient thermal
resistance O-ja (That "O" is supposed to a be a "theta" character, which
isn't on my terminal!), power dissipation Pd, and ambient Ta, is given
by:

    Tj = Pd * Ota + Ta

 example:
 To figure out how hot the IC is, you need to know the thermal characteristics
 of the package. I've worked with some 30W ECL devices, that have a total
 theta of 2.5C/W [heat sinked; with 500 cfm air flow]. That would allow the
 die to keep well below the design limit of 125C (mainly set for long term
 reliability). Given an ambient temp of 25C + 30W*2.5C/W = 100C.

If a heat sink with thermal resistance Osa (sink to ambient) is used and the
thermal resistance from junction to case, Ojc, is given, the:

     Tj = Pd (Ojc + Ocs + Osa) + Ta

 example:
 Given a 5 Volt linear regulator, with a 12v input, delivering 1 Amp of
 current; you would be dissipating 7 watts of power, 1Amp*(12-5). If the
 regulator had a 0jc of 15C/W [National Semi "H" package]; 0sc of 1W/C
 (heat sink compound); and a Osa of 8W/C (heat sink);

    Tj = 7W (15C/W + 1C/W + 8C/W) + 25
           = 7 W (24C/W) + 25C
           = 193C  !!!! OUCH!

    (hint; the regulator will go into thermal shut down first!)

Repeating the above with a "K" case with a 3C/W;

    Tj = 7W (3C/W + 1C/W + 8C/W) + 25
       = 7W (12C/W) + 25
           = 109C

Hope that helps a little bit in explaining "how hot is it". I have picked
some typical values, as an explanation. This should cover 95% of the
Sci.electronics readers. My examples are towards worst case, actually a
T220 package would have worked fine. Those in the semiconductor industry
will hopefully accept my brevity. I am open to further questions.

Here at BIT (Bipolar Integrated Technology) we work with a lot of high power
ECL devices, and have a "computer controlled wind tunnel" for thermal
characterization of our devices. This machine gives Theta j-a numbers from
still air to a 1000 lfm of air flow.

Hope this helps clarify "how hot",

Thomas Walter                     bit![e-mail]