Best heat sink compound?

[Bill Sloman]

Jim Weir <jim@rst-engr.com> wrote in message news:<8spmsv0m63763a7oq8ul7beanb7p8mt16e@4ax.com>...

You will get many answers to this, and as usual, "best" is a function of several
variables.

However, if you are out in East Undershirt hicksville some day and just HAVE to
get the job done, go down to the drugstore and get a small tube of that white
stuff you rub on your nose in the summertime (zinc oxide). It isn't the BEST
thermal conductor you can find, but it will do a fine job in most instances.

BTW, the "thin rubber fiber" that some other folks are talking about are from a
company called Bergquist if I'm not mistaken. Expensive, but damned good
thermal conductivity.

I've never been that impressed with Bergquist thermal pads. The
claimed conductivity isn't all that good, and we didn't seem to see
anything quite that high in practice.

Farnell seems to stock Warth International's graphite cloth pads, for
which they claim 0.05C/W thermal resistances for TO-3 devices, which
is pretty good.

The catch with graphite cloth is that it is electically conductive, so
you may need to use hard-anodised heatsinks (as recommended elsewhere
in this thread).

In 1993 we used graphite cloth to couple several biggish Marlow
Peltier-junctions to our heat-sinking arrangements, and were pretty
happy with the results. The paper we published on the project includes
some thermal time constant measurements for a sereies of Peltier
junctions, which we converted to a surprisingly consistent set of heat
capacities by a process that depended on estimating total thermal
resistances to ambient, which gave us some confidence in the thermal
resistances achieved.

------
Bil Sloman, Nijmegen

 

[DarkMatter]

On Tue, 02 Dec 2003 09:08:54 GMT, "Daniel L. Belton" <abuse@spam.gov>
Gave us:

Ok... Let me put it into numbers... when overclocking my computer, I
get approx. 400 mhz more out of it when using arctic silver over the
white goop...

Being an overclocker, you should know better than to give the
multiplied number as in 2.4 GHz OC'd to 2.8, a 400MHz gain.
Bullshit.

That is bullshit. What is the FSB clock change? That is the only
one that REALLY matters.

BTW, if that is not really your valid e-mail, it is illegal to use a
dot gov addy if it isn't legitimate. You could get nailed.

 

[DarkMatter]

On Tue, 2 Dec 2003 10:45:21 -0000, "Genome" <genome@nothere.com> Gave
us:

"John Larkin" <jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote in
message news:445osvk824g54mduli758nij2uj5iq3sa6@4ax.com...
On Mon, 01 Dec 2003 20:19:05 -0800, DarkMatter

But if you want to run a TO-220 or a TO-247 anywhere near its rated >max
power, a sil-pad will be lethal. }~~~

John

Missing Dribbly Devil Smiley added.\

Retard.

AND

Oh boy...

Ooooooo super cross post

Happy you could join?

 

[Genome]

"DarkMatter" <DarkMatter@thebarattheendoftheuniverse.org> wrote in message
news:f3tosv06na8u2vqpure3ur9cm82dn19cgr@4ax.com...

Gosh, I think I've just been DarkMattered.

Thank you.

AND

 

[Frithiof Andreas Jensen]

"John Larkin" <jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote in
message news:f1rmsv835did6o3251d5d9vkauglbop0b2@4ax.com...

The white thermal compound stuff is better. For serious heat sinking,

It sucks in production and it sucks during servicing - the Warth et. al.
thermal pads handles much, much better (and if you cannot get the heat out
with that, you are cutting it too close anyway).

don't use an insulator. If you must, use 0.5 mil
hard anodize on the heat sink.

hehe - above 60 V you *must* use an insulator if the heat-sink is accessible
to the user - besides I would not trust the anodising at all: One tiny piece
of burr under the device; Boom! This *will* happen in a production
environment..

*don't* use a silicone sil-pad or phase-change stuff.
They are both awful thermally.

No, Not Really - especially if the device is fixed with a clip instead of
that silly screw/washer combo.

 

[Roy McCammon]

DaveC wrote:

White "tooth" paste kind, or clear greasy "hair stuff" kind?

The conventional wisdom was that the tooth paste kind is best
if properly applied which means no more than 1 mil thick.

1 mil thick is difficult to achieve.

The greasy kind is probably silicon grease with is banned
in some environments.

I'd certainly try to get by with a sil pad in production.

--
local optimization seldom leads to global optimization

my e-mail address is: rb <my last name> AT ieee DOT org

 

[John Larkin]

On Tue, 02 Dec 2003 03:06:42 -0800, DarkMatter
<DarkMatter@thebarattheendoftheuniverse.org> wrote:

On Mon, 01 Dec 2003 21:05:19 -0800, John Larkin
jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> Gave us:


So, if the heatsink is flat, the grease has roughly a 100:1 advantage
over a sil-pad.


Hey retard. "The math" you jack off at the mouth about doesn't pan
out in reality.

The same transistor on the same sink varies by far less than a 100:1
difference. All the way up until the friggin solder is reflowing.

If one uses the "math" you refer to one can easily see that 7 mils
of sil-pad performs quite similarly to your 0.1 mil interface.
Actually better, since the temperature difference is only a few
degrees on the device. Why for a 70 times thickness, it conducts
across its thickness very nearly the same amount of heat. That
suggests, by the math, that the pad media is BETTER at the job. It is
certainly better when electrical isolation is involved, which often
times it is. Multiply your goop at 70 times the gap and watch it NOT
pass the heat the same, if at all. It is made ONLY for filling
micro-crevaces in the mating interface. If the interface has ANY gap,
the stuff is virtually worthless.

The thermal conductivity of filled silicone grease is close to that of
filled silicone rubber. That makes sense, as they're chemically almost
the same. The point is that the sim-pad *forces* you to have a big gap
between the device and the sink, and that's a lot of theta.

The conductivity, and efficiency of the sink to carry off the heat
is far more important.

What's important is the sum of all the thetas in series.

We run large high power devices on sinks all the time. We operate
them at proper design maximum levels, however. Not the scenarios you
suggest.

If one designs or uses such devices AT their rated power, one is
asking for a failure mode. DOH!

Not if you understand what you're doing.

You need to get a clue. Pads are much more stable over a much
longer period, and even withstand service operations.

Your brain obviously doesn't deal with FACT operations though, so I
guess you won't be learning anything anytime this decade.

Last year we designed a 20 KW peak output gradient driver amp for MRI
imaging. It uses 32 power fets in the output stage, each rated 300
watts continuous dissipation, and we run them at close to 300 watt
peak during a gradient pulse. If you put a sil-pad under a TO-247 it
will increase the thermal resistance by about 1 k/w, and thus
*increase* junction temperature by 300C. Try that on for a FACT.

The fets are clamp-mounted to hefty machined-flat copper heat
spreaders, on top of a huge aluminum sink, with nice gloppy white
silicone grease. I posted some pics to a.b.s.e. a while back.
Protection is by a microprocessor measuring everything and running a
realtime simulation of fet junction temp; I wrote that code, using
that dreaded 'math' again. They have been stone cold reliable in the
field, running right at 300 watts/fet. It's reliable beacuse it was
*designed*, using this 'math' stuff you don't like.

The nice thing about MRI is that the customer always buys 3 units (X,
Y, Z gradient fields) at a time, at about $9K per axis.

John

 

Robert Baer <robertbaer@earthlink.net> says...

The "arctic silver compound" is *bad* stuff.
It migrates all over hell and back, not only making a crappy looking
mess, but also has been reported to do damage to electrical contacts.

1) Did you really have to post this three times, wasting our time?

2) It didn't migrate when I used it.

3) Nobody but you is reporting that it migrates.
http://www.google.com/groups?as_q=%22arctic%20silver%22%20migrates
http://www.google.com/groups?as_q=%22arctic%20silver%22%20migrate
http://www.google.com/groups?as_q=%22arctic%20silver%22%20migration

4) One person reported that other compounds migrate but Arctic Silver
doesn't (search on Arctic Silver in this thread).
http://www.google.com/groups?&threadm=cYnZ8.92811%24%25%252.3989157%40news2.east.cox.net

5) If you get it on electrical contacts, you aren't following
the instructions - you put way too much on.

6) If you can see a "crappy looking mess", you aren't following
the instructions - you put way too much on.

7) "Reported" where? By whom?

 

[John Larkin]

On Tue, 2 Dec 2003 13:30:40 +0100, "Frithiof Andreas Jensen"
<frithiof.jensen@removethis.ted.ericsson.dk> wrote:

"John Larkin" <jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote in
message news:f1rmsv835did6o3251d5d9vkauglbop0b2@4ax.com...


The white thermal compound stuff is better. For serious heat sinking,

It sucks in production and it sucks during servicing - the Warth et. al.
thermal pads handles much, much better (and if you cannot get the heat out
with that, you are cutting it too close anyway).

don't use an insulator. If you must, use 0.5 mil
hard anodize on the heat sink.

hehe - above 60 V you *must* use an insulator if the heat-sink is accessible
to the user -

The hard anodize *is* the insulator, with a thermal resistance a tiny
fraction of a sil-pad's. If the heatsink is properly grounded to the
chassis, it's safe even if an insulator - any insulator - fails.

besides I would not trust the anodising at all: One tiny piece
of burr under the device; Boom! This *will* happen in a production
environment..

A burr will punch through a sil-pad as well. Don't have them.
Interestingly, the hard anodize process etches the aluminum nicely...
it's a soft-looking, very smooth and pretty finish, and hard as glass.
If you take your sharpest meter probes and push as hard as you can,
you'll just measure infinity ohms on a properly hard-anodized heat
sink.

*don't* use a silicone sil-pad or phase-change stuff.
They are both awful thermally.

No, Not Really - especially if the device is fixed with a clip instead of
that silly screw/washer combo.

Do the math. Numbers beat opinions every time.

John

 

[Spehro Pefhany]

On Tue, 02 Dec 2003 08:35:34 -0800, the renowned John Larkin
<jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote:

On Tue, 2 Dec 2003 13:30:40 +0100, "Frithiof Andreas Jensen"
frithiof.jensen@removethis.ted.ericsson.dk> wrote:


"John Larkin" <jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote in
message news:f1rmsv835did6o3251d5d9vkauglbop0b2@4ax.com...


The white thermal compound stuff is better. For serious heat sinking,

It sucks in production and it sucks during servicing - the Warth et. al.
thermal pads handles much, much better (and if you cannot get the heat out
with that, you are cutting it too close anyway).

don't use an insulator. If you must, use 0.5 mil
hard anodize on the heat sink.

hehe - above 60 V you *must* use an insulator if the heat-sink is accessible
to the user -

The hard anodize *is* the insulator, with a thermal resistance a tiny
fraction of a sil-pad's. If the heatsink is properly grounded to the
chassis, it's safe even if an insulator - any insulator - fails.

besides I would not trust the anodising at all: One tiny piece
of burr under the device; Boom! This *will* happen in a production
environment..


A burr will punch through a sil-pad as well. Don't have them.
Interestingly, the hard anodize process etches the aluminum nicely...
it's a soft-looking, very smooth and pretty finish, and hard as glass.
If you take your sharpest meter probes and push as hard as you can,
you'll just measure infinity ohms on a properly hard-anodized heat
sink.

Quoting another source:
"The Electroplating and Engineering Handbook by Lawrence Durney gives
an average breakdown voltage of 40V/uM or 1000V/.001", but because of
imperfections in the coating values can range from 250-1750V/.001"

Safety agencies require something like double the operating voltage
plus 1kV hipot test for line-to-earth breakdown, so around
1250-1500VAC (1 minute) for 120/240VAC operation. This is doable with
a thick enough anodize (maybe 5 thou). For applications like audio
amplifiers, it's not required because the isolation is provided
elsewhere.

*don't* use a silicone sil-pad or phase-change stuff.
They are both awful thermally.

No, Not Really - especially if the device is fixed with a clip instead of
that silly screw/washer combo.

Do the math. Numbers beat opinions every time.

Apples and oranges. 300W/device in a high-end product built and
maintained by careful responsible people vs. 5 or 10W/device in a
volume product slapped together in seconds by people who were working
in fields a year ago. You're both right, in different domains and with
vastly different definitions of "serious".

Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com

 

[John Larkin]

On Tue, 2 Dec 2003 10:45:21 -0000, "Genome" <genome@nothere.com>
wrote:

"John Larkin" <jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote in
message news:445osvk824g54mduli758nij2uj5iq3sa6@4ax.com...
On Mon, 01 Dec 2003 20:19:05 -0800, DarkMatter

But if you want to run a TO-220 or a TO-247 anywhere near its rated >max
power, a sil-pad will be lethal. }~~~

John

Missing Dribbly Devil Smiley added.

AND

Ooooooo super cross post

Genome,

as always, I appreciate your support.

John

 

[John Larkin]

On Tue, 02 Dec 2003 17:17:53 GMT, Spehro Pefhany
<speffSNIP@interlogDOTyou.knowwhat> wrote:

On Tue, 02 Dec 2003 08:35:34 -0800, the renowned John Larkin
jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote:

On Tue, 2 Dec 2003 13:30:40 +0100, "Frithiof Andreas Jensen"
frithiof.jensen@removethis.ted.ericsson.dk> wrote:


"John Larkin" <jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote in
message news:f1rmsv835did6o3251d5d9vkauglbop0b2@4ax.com...


The white thermal compound stuff is better. For serious heat sinking,

It sucks in production and it sucks during servicing - the Warth et. al.
thermal pads handles much, much better (and if you cannot get the heat out
with that, you are cutting it too close anyway).

don't use an insulator. If you must, use 0.5 mil
hard anodize on the heat sink.

hehe - above 60 V you *must* use an insulator if the heat-sink is accessible
to the user -

The hard anodize *is* the insulator, with a thermal resistance a tiny
fraction of a sil-pad's. If the heatsink is properly grounded to the
chassis, it's safe even if an insulator - any insulator - fails.

besides I would not trust the anodising at all: One tiny piece
of burr under the device; Boom! This *will* happen in a production
environment..


A burr will punch through a sil-pad as well. Don't have them.
Interestingly, the hard anodize process etches the aluminum nicely...
it's a soft-looking, very smooth and pretty finish, and hard as glass.
If you take your sharpest meter probes and push as hard as you can,
you'll just measure infinity ohms on a properly hard-anodized heat
sink.

Quoting another source:
"The Electroplating and Engineering Handbook by Lawrence Durney gives
an average breakdown voltage of 40V/uM or 1000V/.001", but because of
imperfections in the coating values can range from 250-1750V/.001"

Safety agencies require something like double the operating voltage
plus 1kV hipot test for line-to-earth breakdown, so around
1250-1500VAC (1 minute) for 120/240VAC operation. This is doable with
a thick enough anodize (maybe 5 thou). For applications like audio
amplifiers, it's not required because the isolation is provided
elsewhere.

I wouldn't use anodize alone to insulate line voltage. Pinholes and
whatever are possible, and AC lines spike to kilovolts now and then.
We have used 1-mil hard anodize on heatsinks standing off up to 200
volts, and it's been very reliable.

I wouldn't trust personal safety to any one single component or
insulator. I don't think UL allows that anyhow.

*don't* use a silicone sil-pad or phase-change stuff.
They are both awful thermally.

No, Not Really - especially if the device is fixed with a clip instead of
that silly screw/washer combo.

Do the math. Numbers beat opinions every time.

Apples and oranges. 300W/device in a high-end product built and
maintained by careful responsible people vs. 5 or 10W/device in a
volume product slapped together in seconds by people who were working
in fields a year ago. You're both right, in different domains and with
vastly different definitions of "serious".

My rant is, I suppose, that the thermal behavior of any product should
be calculated, and not guessed. It's not terribly hard to calculate
most of this stuff, but it's easy to mis-guess a Tj rise by a factor
of five or so. Even a cheap consumer product can benefit from a
thoughtful thermal design... might save a few cents!

John

 

[John Larkin]

On Mon, 01 Dec 2003 08:11:23 -0800, Jim Weir <jim@rst-engr.com> wrote:

You will get many answers to this, and as usual, "best" is a function of several
variables.

However, if you are out in East Undershirt hicksville some day and just HAVE to
get the job done, go down to the drugstore and get a small tube of that white
stuff you rub on your nose in the summertime (zinc oxide). It isn't the BEST
thermal conductor you can find, but it will do a fine job in most instances.

BTW, the "thin rubber fiber" that some other folks are talking about are from a
company called Bergquist if I'm not mistaken. Expensive, but damned good
thermal conductivity.

Jim

0.9 w/m-k is typically spec'd, and that's not 'damned good'; that's
terrible. In real life, they're not even that good.

John

 

[Spehro Pefhany]

On Tue, 02 Dec 2003 10:09:46 -0800, the renowned John Larkin
<jjlarkin@highSNIPlandTHIStechPLEASEnology.com> wrote:

My rant is, I suppose, that the thermal behavior of any product should
be calculated, and not guessed. It's not terribly hard to calculate
most of this stuff, but it's easy to mis-guess a Tj rise by a factor
of five or so.

Reliably getting 300W out of something the size of a pea* is probably
not intuitive the first time you do it. I've been looking at the
aluminum (or copper) base boards again- they can do 10oz conductors on
a 1/8" copper base. The dielectric is a "ceramic" filled polymer
layer, which I'd guess is that nasty BeO stuff.

Even a cheap consumer product can benefit from a
thoughtful thermal design... might save a few cents!

Which can add up. Sometimes they do seemingly stupid things- I used to
have a Japanese receiver that used a *heat pipe* to move the heat 6"
inside the case from the devices to the heat sink. Maybe the heat sink
was cheaper (a bunch of thin stamped fins) or maybe it was a marketing
"feature" or maybe the engineers just wanted to play around. <shrug>

* Boiled salted soybeans- they look like peas in the pod, with an
especially inedible pod, and the beans themselves are a delicious
drink accompaniment. Hard to believe that they can make something as
repulsive as Tofu/Dofu from them.

Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com

 

[Spehro Pefhany]

On Tue, 02 Dec 2003 10:09:46 -0800, the renowned John Larkin
<jjlarkin@highSNIPlandTHIStechPLEASEnology.com> wrote:

I wouldn't trust personal safety to any one single component or
insulator. I don't think UL allows that anyhow.

They demand an extraordinary level of reliability from anything
between line and "earth" in a product, presumably because they *know*
that 0.X% of products will be plugged into a 2-wire extension cord
with the knobby cut off or otherwise not properly earthed and if you
allow 0.Y% of caps to fail, then you'll get a certain non-zero number
of deaths by electrocution with a lot of such products out there.

We made an automated capacitor test rig for CSA to do this some years
ago. It tests a LARGE number of parts (hundreds) simultaneously, at
pulsed high voltage (kV), and in an oven at elevated temperature.

The single point failure thing is a good rule of thumb, but it is not
always required on simple mechanical devices- for example if the
housing is smashed (one failure) or the line cord is split you can
touch line voltage. I also wonder about cases like a safety-critical
device I have in front of me now- If this device fails, there will
likely be an explosion. The designer (not me) used two resistors in
series in one part of the circuit and two in parallel in another (it's
approved by the appropriate agencies, BTW). The thing is, they are
identical resistors, no doubt from the same production lot. If the mfr
is having a problem, say with their end-cap crimping machine, chances
are that *both* resistors could have the exact same problem. Automated
machinery tends to put them beside each other on the T&R. 8-(

Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com

 

[John Fields]

On Tue, 02 Dec 2003 03:06:42 -0800, DarkMatter
<DarkMatter@thebarattheendoftheuniverse.org> wrote:

On Mon, 01 Dec 2003 21:05:19 -0800, John Larkin
jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> Gave us:


So, if the heatsink is flat, the grease has roughly a 100:1 advantage
over a sil-pad.


Hey retard. "The math" you jack off at the mouth about doesn't pan
out in reality.

The same transistor on the same sink varies by far less than a 100:1
difference. All the way up until the friggin solder is reflowing.

If one uses the "math" you refer to one can easily see that 7 mils
of sil-pad performs quite similarly to your 0.1 mil interface.
Actually better, since the temperature difference is only a few
degrees on the device. Why for a 70 times thickness, it conducts
across its thickness very nearly the same amount of heat. That
suggests, by the math, that the pad media is BETTER at the job. It is
certainly better when electrical isolation is involved, which often
times it is. Multiply your goop at 70 times the gap and watch it NOT
pass the heat the same, if at all. It is made ONLY for filling
micro-crevaces in the mating interface. If the interface has ANY gap,
the stuff is virtually worthless.

---
Larkin is right and you're wrong. Go to

http://www.bergquistcompany.com/objects/SilPad_SelGuide_PDFs/SilPadSelectionGuide2002.PDF

and learn all about what you _think_ you are knowledgeable about but
obviously need some help to understand.

After you get through the theory part, take a look at the TO-220 thermal
performance of some of the various pads and I'm sure you'll be in for a
surprise considering that 2.0°C/watt means that the mounting surface of
the TO-220 package will be 200°C higher than the temperature of the
heatsink directly under the package if the transistor is dissipating
100W. That means that even with an infinite sink capable of staying at,
say, 25°C, no matter what, the mounting surface of the TO-220 will be at
225°C when it's dissipating 100W. Now look at theta jc and theta
sink-to-ambient to find out what the junction temp is going to be in the
real world when it's dissipating 100W and you'll probably be in for
another nasty surprise.

Next, to do a fair comparison you'll want to look at the thermal
characteristics of an interface comprising a 0.005" thick oxide film on
an aluminum substrate with a film of goop sandwiched by the TO-220
package and the anodized surface of the heatsink. Another surprise I'm
sure.
---

The conductivity, and efficiency of the sink to carry off the heat
is far more important.

---
No. Even with a perfectly efficient heatsink, if you cant get the heat
into it you can't get rid of the heat that way.
---

We run large high power devices on sinks all the time. We operate
them at proper design maximum levels, however. Not the scenarios you
suggest.

---
From your exposition so far, I doubt whether you have the capability of
determining what proper design levels are, other than something like
"Hey, Joe, we just blew out another one. Let's try a fan and see what
happens."
---

If one designs or uses such devices AT their rated power, one is
asking for a failure mode. DOH!

---
I'm sure it seems like that to you...
---

You need to get a clue. Pads are much more stable over a much
longer period, and even withstand service operations.

---
How about some numbers to substantiate your claim(s)?
---

Your brain obviously doesn't deal with FACT operations though, so I
guess you won't be learning anything anytime this decade.

---
Larkin seems to be pretty much caught up with what he needs to know in
order to succesfully do what he wants to do, so anything new he needs to
learn is going to be miniscule compared to the gargantuan effort you'd
need to make to even _start_ to catch up.

--
John Fields
Professional circuit designer
http://austininstruments.com

 

[John Fields]

On Tue, 02 Dec 2003 13:09:04 -0600, John Fields
<jfields@austininstruments.com> wrote:

Correction...

Next, to do a fair comparison you'll want to look at the thermal
characteristics of an interface comprising a 0.005" thick oxide film on
^^^^^^

0.0001"

an aluminum substrate with a film of goop sandwiched between the TO-220
package and the anodized surface of the heatsink. Another surprise I'm
sure.

--
John Fields

 

[John Larkin]

On Tue, 02 Dec 2003 18:18:48 GMT, Spehro Pefhany
<speffSNIP@interlogDOTyou.knowwhat> wrote:

On Tue, 02 Dec 2003 10:09:46 -0800, the renowned John Larkin
jjlarkin@highSNIPlandTHIStechPLEASEnology.com> wrote:


My rant is, I suppose, that the thermal behavior of any product should
be calculated, and not guessed. It's not terribly hard to calculate
most of this stuff, but it's easy to mis-guess a Tj rise by a factor
of five or so.

Reliably getting 300W out of something the size of a pea* is probably
not intuitive the first time you do it. I've been looking at the
aluminum (or copper) base boards again- they can do 10oz conductors on
a 1/8" copper base. The dielectric is a "ceramic" filled polymer
layer, which I'd guess is that nasty BeO stuff.

The bitch here, with thin copper, is the spreading thermal resistance,
and to calculate that properly you need nasty expensive finite-element
modeling software. Or just test a prototype!

BeO is falling out of favor for the obvious reasons. AlN is cool
stuff. For a filled epoxy or whatever, it won't make much difference,
as the epoxy dominates. Diamond filled epoxy is cool, though.

Even a cheap consumer product can benefit from a
thoughtful thermal design... might save a few cents!

Which can add up. Sometimes they do seemingly stupid things- I used to
have a Japanese receiver that used a *heat pipe* to move the heat 6"
inside the case from the devices to the heat sink. Maybe the heat sink
was cheaper (a bunch of thin stamped fins) or maybe it was a marketing
"feature" or maybe the engineers just wanted to play around. <shrug

* Boiled salted soybeans- they look like peas in the pod, with an
especially inedible pod, and the beans themselves are a delicious
drink accompaniment. Hard to believe that they can make something as
repulsive as Tofu/Dofu from them.

I'll stick with cranberry beans. Did you see my pics in a.b.s.e. a
while back?

John

 

[John Larkin]

On Tue, 02 Dec 2003 13:29:16 -0600, John Fields
<jfields@austininstruments.com> wrote:

On Tue, 02 Dec 2003 13:09:04 -0600, John Fields
jfields@austininstruments.com> wrote:

Correction...

Next, to do a fair comparison you'll want to look at the thermal
characteristics of an interface comprising a 0.005" thick oxide film on
^^^^^^
0.0001"

an aluminum substrate with a film of goop sandwiched between the TO-220
package and the anodized surface of the heatsink. Another surprise I'm
sure.

We like to hard anodize 0.5 mils (0.0005 inches), or maybe 1 mil for
higher voltages, like maybe above 100. Hard anodize is aluminum oxide,
basicly dirty sapphire. Its thermal conductivity is about 30 times
better than filled silicone, so 1 mil anodize has about 1/300 the
thermal resistance of a 10 mil sil-pad!

"Regular" anodize is pretty fragile to use to insulate transistors.

John

 

[Spehro Pefhany]

On Tue, 02 Dec 2003 11:36:06 -0800, the renowned John Larkin
<jjlarkin@highSNIPlandTHIStechPLEASEnology.com> wrote:

The bitch here, with thin copper, is the spreading thermal resistance,
and to calculate that properly you need nasty expensive finite-element
modeling software. Or just test a prototype!

Yes. It's not just Ohm's law anymore. Some CAD programs include an
introductory FEA program, but for thermal analysis (and transient
thermal analysis) you need deep pockets. For something very simple
(simple shape, ignore convection), it might not be that hard to hack
something up because it is then just Ohm's law for each snippet of the
part.

BeO is falling out of favor for the obvious reasons. AlN is cool
stuff. For a filled epoxy or whatever, it won't make much difference,
as the epoxy dominates. Diamond filled epoxy is cool, though.

I hear CVD diamond film is really nice.

Even a cheap consumer product can benefit from a
thoughtful thermal design... might save a few cents!

Which can add up. Sometimes they do seemingly stupid things- I used to
have a Japanese receiver that used a *heat pipe* to move the heat 6"
inside the case from the devices to the heat sink. Maybe the heat sink
was cheaper (a bunch of thin stamped fins) or maybe it was a marketing
"feature" or maybe the engineers just wanted to play around. <shrug


* Boiled salted soybeans- they look like peas in the pod, with an
especially inedible pod, and the beans themselves are a delicious
drink accompaniment. Hard to believe that they can make something as
repulsive as Tofu/Dofu from them.

I'll stick with cranberry beans. Did you see my pics in a.b.s.e. a
while back?

No, missed the photos, but caught the discussion here. I have not
knowningly run across them yet in the fresh form, though I've probably
eaten them in pasta e fagioli. They also seem to be called borlotti or
corn beans.
Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com