Insulated and bare tabs of power devices

[Pimpom]

When active power devices are operated at dissipations
approaching their maximum ratings (after including appropriate
derating factors, heatsink properties and safety margins), the
case-heatsink thermal resistance sometimes makes up a significant
part of the total J-A thermal circuit. Some devices like those in
variants of TO-220 and TOP3 packages are offered with a choice of
bare metal or insulated mounting tabs. Datasheets often do not
make any distinction between the two versions regarding their
case-heatsink thermal resistances.

Would it be reasonable to assume that the c-hs Rth of an
insulated case is about the same as that of a bare metal case
mounted with a mica insulator (using a thermal paste in each
case)?

 

[Pimpom]

Examples? I don't think I've ever seen a power device data
sheet that
didn't list individual junction-to-case thermal resistances for
each
and every package option.

Actually, it's the case-to-heatsink Rth, not j-c, that I'm
concerned about.

I wouldn't make any assumptions -- other than if the data sheet
doesn't go into much detail, then that's an indication that the
part
is designed with insufficient care, and should be avoided if
possible.

One example is the popular LM3886 audio amplifier. It comes with
T and TF suffixes, the latter being the insulated version.

 

[Pimpom]

Did you have any specfic parts/data sheets in mind?

An example is the popular LM3886 audio amplifier. It comes with T
and TF suffixes, the latter being the insulated version. I've
come across several other devices although I can't recall a
specific type number off the top of my head.

The insulation they add has got to increase theta, usually a
lot.

The best way to heatsink a power device is to bolt it to a
bare, very
flat copper heat spreader block, with silicone grease, and then
bolt
the block to the main heat sink, with an insulator if
necessary. The
best practical insulator is hard anodize on very flat aluminum.

Of course, but it's not always practicable to use that mounting
method.

TO-220s are not the best package. Thermal footprint is small,
half of
that (the wing part with the hole in it) is not very effective,
and
the bolt hole is off center.

Here's 32 power fets, clamp mounted on copper heat spreaders,
then the
main aluminum heat sink. The only insulator is between the heat
sink
and the chassis.

ftp://jjlarkin.lmi.net/Amp.jpg

Unfortunately, I've never been able to connect to your ftp links,
either with IE8 or FF. Maybe it has something to do with my ISP.
But I'm familiar with the mounting method you described.

 

[Pimpom]

Oi. (Sound of forehead slapping). My reading comprehension is
down
today.

Case to heatsink is your responsibility. There are ap notes to
help,
but -- it's still your responsibility.

You can figure, though, that the case to heatsink thermal
resistance
with an insulated package will be less than the case to mica
tab to
heatsink thermal resistance of a non-insulated package, because
there's only one thermal interface, and no thermal insulator,
however
thin.

I'd go on an ap note hunt (and, alas, I can't help you with
guidance
on where to look, other than I know that the Motorola RF
transistor
ones tended to be very detailed and informative).

I agree, at least to some extent, that it's up to the end-user to
provide a good device-to-heatsink interface. And a mica or other
external insulator is certainly not an integral part of the
device and therefore outside the control of the device
manufacturer. However, an insulated tab _is_ a part of the device
and it would be nice if the manufacturer gave us some idea of
what to expect.

 

[Pimpom]

Once again, your confusion confuses me. If the manufacturer is
selling something in an insulated case and publishes a junction
to
case thermal resistance, they mean the junction to the
_outside_ of
the case, not the junction to some internal feature of the
case.

Then it's your job to figure out how to get the heat out of the
case.

That's assuming that the thermal resistance from junction to the
outside of the case is the same for both insulated and bare
cases. Is that a safe assumption? The mounting tab of an
insulated type is considerably thicker than that of a bare-metal
type, indicating that the metal is covered with a
not-insignificant layer of plastic. I would have thought that
this would make at least a small difference even if the datasheet
doesn't say so. With a device dissipating tens of watts, a
difference of 0.1 C/W could mean a difference of several degrees.

 

[Pimpom]

Let's review: You gave a long spiel about junction to heatsink
resistances, I assumed it was about junction to case, and said
I
didn't think that a data sheet would be inaccurate about this.

You said "no, no, it's about junction to _heatsink_". So I
went and
blattered on about that.

Now you're saying "no, no, it's about junction to case after
all".

Make Up Your Mind.

For a moment, you made me think that I'd screwed up and actually
started typing an admission. Then I went over my posts again and
saw that I didn't. You say "You gave a long spiel about junction
to heatsink resistances......". I didn't mention J-HS at all in
my opening post. It was about case-to-heatsink. Where I might
have been unclear was that I was thinking about the thermal
resistance from the metal tab (which I assumed to be inside the
plastic) to the heatsink. I assumed that the metal tab is the
same in both versions and that the plastic coating introduces an
additional theta.

I mentioned the junction-to-ambient resistance to point out that
the case-heatsink Rth can be a significant part of it. I repeat,
I never mentioned J-C at all in my initial post.

I realise now that assuming that there's a metal tab inside the
plastic type could be wrong. Maybe it's all plastic. Tomorrow,
I'll see if I can find an old insulated TO-220 and break it open
(it's well past midnight here now).

If I'm NOT wrong and both types have the same metal tab inside,
with an extra layer of plastic in the insulated type, then that
must introduce an additional Rth. My original question was
whether that additional Rth is about the same as that of a mica
insulator.

 

[Phil Allison]

"Pimpom"

One example is the popular LM3886 audio amplifier. It comes with T and TF
suffixes, the latter being the insulated version.

** Natsemi seem not to have published additional data for the thermal
resistance of the TF pack for the LM3886TF.

Using 2 thou mica and thermal grease on a smooth, flat heatsink results in
0.2 degrees C per watt (mounting base to heatsink) - according to Natsemi's
notes on the LM3886T. As this is a low voltage device, the plastic coating
on the rear tab of the TF version is likely to be very thin but possibly not
as thin as a 2 thou mica.

Comments on web forums suggest a figure of 1 degree C per watt.



.... Phil

 

[miso]

If it is not stated in the electricals of the datasheet, then you are on
your own.

Chips in plastic might have a slug in the package. Or they just use a
heavier bonding wire. [Heat transfer through the bonding wire is
proportional to the square of the radius, so a little helps a lot.]
Sometime they are put in the package upside down to get the die away
from the PCB.

You can measure the junction temperature by forward biasing a diode on
the chip. I can post how to do this if you are actually going to do the
work.

 

[Phil Allison]

"miso"

If it is not stated in the electricals of the datasheet, then you are on
your own.

Chips in plastic might have a slug in the package. Or they just use a
heavier bonding wire.

** Read the OP's question - fuckhead !!

 

[miso]

"miso"



** Read the OP's question - fuckhead !!

**** you very much.

 

[Pimpom]

Just copy and paste the address into the browser address bar to
see
them

I've done that before. Doesn't work for me, and I remember some
others saying that it doesn't work for them either. It works with
other links that don't open directly from my newsreader, but not
wih JL's. Tinkering with or turning off my firewall makes no
difference.

 

[Pimpom]

"Pimpom"



** Natsemi seem not to have published additional data for the
thermal
resistance of the TF pack for the LM3886TF.

Using 2 thou mica and thermal grease on a smooth, flat heatsink
results in 0.2 degrees C per watt (mounting base to heatsink) -
according to
Natsemi's notes on the LM3886T. As this is a low voltage
device, the
plastic coating on the rear tab of the TF version is likely to
be
very thin but possibly not as thin as a 2 thou mica.

Comments on web forums suggest a figure of 1 degree C per watt.

Thanks. That about agrees with the ballpark estimate I had in
mind. The difference between that and a bare metal + mica
mounting is quite significant at high dissipation levels.

 

[Mikko OH2HVJ]

Examples? I don't think I've ever seen a power device data sheet that
didn't list individual junction-to-case thermal resistances for each and
every package option.

MJE340 / MJE350 has been a pain in this sense - nice to have multiple sources,
but you never know if you get them with or without plastic insulation.
I think ST or OnSemi datasheet looked like metal visible on the cover sheet,
but the thickness in the last page indicated a plastic covered version.
Also, most of the datasheets do not list thermal resistance..

 

[josephkk]

When active power devices are operated at dissipations
approaching their maximum ratings (after including appropriate
derating factors, heatsink properties and safety margins), the
case-heatsink thermal resistance sometimes makes up a significant
part of the total J-A thermal circuit. Some devices like those in
variants of TO-220 and TOP3 packages are offered with a choice of
bare metal or insulated mounting tabs. Datasheets often do not
make any distinction between the two versions regarding their
case-heatsink thermal resistances.

Data sheets assume best possible metal to metal contact (short of welding)
if they discuss Theta(c-hs) at all.

Would it be reasonable to assume that the c-hs Rth of an
insulated case is about the same as that of a bare metal case
mounted with a mica insulator (using a thermal paste in each
case)?

Absolutely not. Electrical insulators are always thermal insulators as
well. The material properties go hand in hand.

?-)

 

[josephkk]

For a moment, you made me think that I'd screwed up and actually
started typing an admission. Then I went over my posts again and
saw that I didn't. You say "You gave a long spiel about junction
to heatsink resistances......". I didn't mention J-HS at all in
my opening post. It was about case-to-heatsink. Where I might
have been unclear was that I was thinking about the thermal
resistance from the metal tab (which I assumed to be inside the
plastic) to the heatsink. I assumed that the metal tab is the
same in both versions and that the plastic coating introduces an
additional theta.

I mentioned the junction-to-ambient resistance to point out that
the case-heatsink Rth can be a significant part of it. I repeat,
I never mentioned J-C at all in my initial post.

I realise now that assuming that there's a metal tab inside the
plastic type could be wrong. Maybe it's all plastic. Tomorrow,
I'll see if I can find an old insulated TO-220 and break it open
(it's well past midnight here now).

If I'm NOT wrong and both types have the same metal tab inside,
with an extra layer of plastic in the insulated type, then that
must introduce an additional Rth. My original question was
whether that additional Rth is about the same as that of a mica
insulator.

Thanks for the clear explanation of what you want to know. That isn't
quite what you expressed in your OP.

?-)

 

[josephkk]

If it is not stated in the electricals of the datasheet, then you are on
your own.

Chips in plastic might have a slug in the package. Or they just use a
heavier bonding wire. [Heat transfer through the bonding wire is
proportional to the square of the radius, so a little helps a lot.]
Sometime they are put in the package upside down to get the die away
from the PCB.

You can measure the junction temperature by forward biasing a diode on
the chip. I can post how to do this if you are actually going to do the
work.

Not necessarily available from a monolithic power amplifier.

?-)

 

[Ralph Barone]

Data sheets assume best possible metal to metal contact (short of welding)
if they discuss Theta(c-hs) at all.
Absolutely not. Electrical insulators are always thermal insulators as
well. The material properties go hand in hand.

?-)

Diamond?

 

[tm]

snip


Diamond?

Diamond is better than silver. The relationship between thermal and
electrical conductivity only holds for conductors.

 

[[email protected]]

Data sheets assume best possible metal to metal contact (short of welding)
if they discuss Theta(c-hs) at all.

Which is why I've been moving from the TO262 style to DFNs, and such, with
heat slugs. They're much better packages. Smaller, too.

Absolutely not. Electrical insulators are always thermal insulators as
well. The material properties go hand in hand.

Read it again. He specified mica between the metal. It still might not be
true, but your answer doesn't make sense.

 

[josephkk]

The datasheet of the LM3886 discusses it at some length. It just
doesn't mention any difference between the coated and uncoated
types. My concern is based on the assumption that the two types
have identical metal tabs and that the the insulated type simply
has a layer of epoxy coating on the tab. If this assumption is
true and both types are mounted _directly_ on a heatsink, then
the coated type will definitely have an extra thermal resistance
compared to the bare metal type.

When I asked about the case-heatsink thermal resistance, I meant
the metal tab-to-heatsink Rth in _both_ types. I probably created
some confusion by failing to make that clear. Sorry about that.

It may be that they seriously tried to equalize Theta(j-tab) between the
two for the first data sheet. No bets, no promises.

In general, yes. But there are a few partial exceptions like
mica, beryllium oxide, anodized Al, silicone-rubber, thermal
grease which are used for their relatively low thermal
resistances. I don't have any data about the epoxy encapsulation
used for semiconductors, but I'd expect them to have better
thermal conductivity than other common plastics, though worse
than metal.

BeO is the only partial exception, look up the materials properties, and
it has toxicity issues. The fundamental issue is to always keep the
insulators very thin, mica does well at this but is fragile. Same with
light anodize, heavy anodize can be a good compromise if sufficient care
is taken not to break the anodize (it is still very thin). Grease helps
but has a bad tendency to leak / vaporize away. Epoxy and silicone
usually hurts thermal transmission. It is quite simple to look up the
thermal properties of the various materials and calculate the effects.

?-)