thermal coupling of SMD

[Malcolm Reeves]

Hi,

I want to thermal couple 3 SOT223 parts. At first I thought of using
a DIL style heatsink over the top of all 3 and bonding it with thermal
epoxy. However, the thermal coupling in the SOT223 is to the tab
rather than the top of the case so I would have pretty poor thermal
coupling this way. Tab is 15C/W, case top is 80C/W, perhaps less,
certainly more than 40C/W so much worse than the tab. Of course the
epoxy will also bond to the tab but the thickness is so much the
thermal impedance is very high.

My next though was use the PCB copper, so tab solders to copper, then
screw metal bar with silpad to couple to the copper. But in 2oz
copper a track 3mm wide (the tab width) going 3mm is 36C/W thermal
resistance (by my calculations). I could solder a piece brass or
copper right at the tab so as to beef it up but this is getting messy.

I also looked at using high value resistor just for the alumina
substrate for thermal coupling but best I could get was about 40C/W
with a 1206 4 pack (i.e. connections on the long sides).

Does anyone have any suggestions?

TIA

Malcolm

--

....malcolm

Malcolm Reeves BSc CEng MIEE MIRSE, Full Circuit Ltd, Chippenham, UK
([email protected], [email protected] or [email protected]).
Design Service for Analogue/Digital H/W & S/W Railway Signalling and Power
electronics. More details plus freeware, Win95/98 DUN and Pspice tips, see:

http://www.fullcircuit.com or http://www.fullcircuit.co.uk

NEW - Desktop ToDo/Reminder program (free)

 

[John Larkin]

Hi,

I want to thermal couple 3 SOT223 parts. At first I thought of using
a DIL style heatsink over the top of all 3 and bonding it with thermal
epoxy. However, the thermal coupling in the SOT223 is to the tab
rather than the top of the case so I would have pretty poor thermal
coupling this way. Tab is 15C/W, case top is 80C/W, perhaps less,
certainly more than 40C/W so much worse than the tab. Of course the
epoxy will also bond to the tab but the thickness is so much the
thermal impedance is very high.

My next though was use the PCB copper, so tab solders to copper, then
screw metal bar with silpad to couple to the copper. But in 2oz
copper a track 3mm wide (the tab width) going 3mm is 36C/W thermal
resistance (by my calculations).

That's about right.

I could solder a piece brass or
copper right at the tab so as to beef it up but this is getting messy.

I also looked at using high value resistor just for the alumina
substrate for thermal coupling but best I could get was about 40C/W
with a 1206 4 pack (i.e. connections on the long sides).

Does anyone have any suggestions?

TIA

Malcolm

Will the parts be dissipating much power themselves? Must the tabs be
insulated from each other? *Why* do you want them to be isothermal?
Can you give more details on the application?

John

 

[red rover]

What is your objective of adding the thermocouple?
If you are trying to see how hot the device will run
in actual use then some of the techniques you are
talking about could alter the results from reality.
Or are you trying to make this part of a control system
or overload protection that remain in the final product?

If the heatsink, metal bar or brass tube were only added during
measurement then the measured results would differ
from reality. A thermal imaging camera could help here.

Keep in mind also that the thermocouple can act like a
heatsink so you may want to go with small wires if you
are trying to not affect the measurement.

Although it is a good idea to keep the thermal resistance
to the probe low, keep in mind that there shouldn't be
much temperature drop along this resistance as long is
there is no significant heatflow along this path.

Steve

 

[James Meyer]

Hi,

I want to thermal couple 3 SOT223 parts.

Does anyone have any suggestions?

TIA

Malcolm

If you put the devices close together and then isolate them from the
environment, you will get good thermal tracking between the devices. You won't
track the environment, but then you didn't say that was a requirement.

For example: a small PC board to hold the devices with very thin wires
connected to the board and an overall covering of something like styrofoam.

All you have to do to get good thermal tracking is to make the coupling
between the devices (no matter how poor) much better than the coupling between
the devices and the environment.

Think of the mechanism that produces temperature differences, heat flow,
as a current. If I have two 1k resistors that are connected in series with a
10k resistor between them and I want to have the same voltage on all their
terminals, then I connect them to the outside world with 100 Meg resistors so
that there is only a very small current flow.

Picture a 10 volt source connected to a 100 Meg resistor connected to
our 1k and 10k resistors connected finally to ground with another 100 Meg
resistor. The voltage at any of the middle resistor's terminals will be very
close to 5 volts. Even if the voltage supply changes and the inner resistor's
terminal voltage changes, those inner voltages will always track each other very
closely.

Jim

 

[Mac]

Hi,

I want to thermal couple 3 SOT223 parts. At first I thought of using
a DIL style heatsink over the top of all 3 and bonding it with thermal
epoxy. However, the thermal coupling in the SOT223 is to the tab
rather than the top of the case so I would have pretty poor thermal
coupling this way. Tab is 15C/W, case top is 80C/W, perhaps less,
certainly more than 40C/W so much worse than the tab. Of course the
epoxy will also bond to the tab but the thickness is so much the
thermal impedance is very high.

My next though was use the PCB copper, so tab solders to copper, then

This can be very effective for cooling purposes, so why not for coupling?

screw metal bar with silpad to couple to the copper. But in 2oz
copper a track 3mm wide (the tab width) going 3mm is 36C/W thermal
resistance (by my calculations).

Why not make the connection much wider? I would think you could at least
go up to twice the width of the tab. Also, maybe you can have the
copper solder-coated, in which case it might conduct a little better than
the copper alone. Or you can put copper on several layers, if it is a
multi-layer board, and tie the layers together thermally vias.

I could solder a piece brass or
copper right at the tab so as to beef it up but this is getting messy.

I also looked at using high value resistor just for the alumina
substrate for thermal coupling but best I could get was about 40C/W
with a 1206 4 pack (i.e. connections on the long sides).

Does anyone have any suggestions?

Finally, I have to ask, is the part available in any other packages? For
example, if it is available in a TO-220, you could mount the three pieces
together on a common aluminum block. In this case, you probably don't want
a heat-sink, because you only want the heat to flow from chip to chip.

TIA

Malcolm

Mac
--

 

[Malcolm Reeves]

Hi,

Let me answer all the posts in one.

The application is Vbe compensation in a power amplifier, not thermo
couple measurement as someone thought or anything temporary. This is
for final production (volumes in the 100s).

The arrangement is a complementary follower so only the drivers need
to be compensated. The drivers dissipate 0.3W max so thermal error is
4.5C just for the 15C/W junction to tab. I don't want to add much to
this so I guess I want better than 15C/W for the coupling.

I did think about using, say T0126 parts but I already use in the
design the SOT223 parts and their performance seems superior to any
T0126 parts I can find. Probable a case of new silicon designs using
new packaging and older designs (i.e. poorer performance) using older
packaging.

As far as I can see, since the tab IS the only good thermal path, then
I have to use that. So it looks like I HAVE to use copper shim,
unless anyone knows a manufactured part.

Malcolm

Hi,

I want to thermal couple 3 SOT223 parts. At first I thought of using
a DIL style heatsink over the top of all 3 and bonding it with thermal
epoxy. However, the thermal coupling in the SOT223 is to the tab
rather than the top of the case so I would have pretty poor thermal
coupling this way. Tab is 15C/W, case top is 80C/W, perhaps less,
certainly more than 40C/W so much worse than the tab. Of course the
epoxy will also bond to the tab but the thickness is so much the
thermal impedance is very high.

My next though was use the PCB copper, so tab solders to copper, then
screw metal bar with silpad to couple to the copper. But in 2oz
copper a track 3mm wide (the tab width) going 3mm is 36C/W thermal
resistance (by my calculations). I could solder a piece brass or
copper right at the tab so as to beef it up but this is getting messy.

I also looked at using high value resistor just for the alumina
substrate for thermal coupling but best I could get was about 40C/W
with a 1206 4 pack (i.e. connections on the long sides).

Does anyone have any suggestions?

TIA

Malcolm

--

....malcolm

Malcolm Reeves BSc CEng MIEE MIRSE, Full Circuit Ltd, Chippenham, UK
([email protected], [email protected] or [email protected]).
Design Service for Analogue/Digital H/W & S/W Railway Signalling and Power
electronics. More details plus freeware, Win95/98 DUN and Pspice tips, see:

http://www.fullcircuit.com or http://www.fullcircuit.co.uk

NEW - Desktop ToDo/Reminder program (free)

 

[James Meyer]

Hi,

Let me answer all the posts in one.

The application is Vbe compensation in a power amplifier, not thermo
couple measurement as someone thought or anything temporary. This is
for final production (volumes in the 100s).

Let me see if I've got this right. You have two devices that dissipate
heat and you want to thermally connect a third device so that its junction's
temperature track the junction temperatures of the first two?

If that's what you want to do, then you can't do it with any sort of
simple thermal link.

On the other hand, you *can* make the third device's junctions very
accurately track the temperature of the heatsink that the first two devices are
connected to. All you have to do is to mount the third device to the common
heatsink and then insulate the third device from the ambient environment.

The whole point about "thermal resistance" is that if there is no heat
flow, there will be no temperature drop. That's what I tried to point out in my
first post.

Jim

 

[Malcolm Reeves]

Let me see if I've got this right. You have two devices that dissipate
heat and you want to thermally connect a third device so that its junction's
temperature track the junction temperatures of the first two?

If that's what you want to do, then you can't do it with any sort of
simple thermal link.

On the other hand, you *can* make the third device's junctions very
accurately track the temperature of the heatsink that the first two devices are
connected to. All you have to do is to mount the third device to the common
heatsink and then insulate the third device from the ambient environment.

A good point but I can see practical difficulties. Tj-tab is 15C/W so
at 0.3W I have a 4.5C error to the heatsink. My sensor dissipation is
peanuts so the error is just 4.5C. The tab is the best coupling to
the junction. Suppose I bond an Al plate to the top of the cases. It
looks like Tj-top is about 80C/W. The heatsink is say 22.5C/W so
45C/W per power device. Th-a is 13.5C.

So I have two paths, power junction to tab to heatsink to air, and
power junction to case top to Al plate to case sensor to sensor
junction to sensor tab to air (no heatsink). In order for this to be
better than the 4.5C error of just sensing the heatsink, the thermal
resistance from Al plate to air must be >> 3 x 80C/W and the sensor
tab to air must be >> 3 x 160C. The combination must be greater than
3 so roughly 6 each. That's quite high and I think would be very
difficult to achieve. The area is much more and 1mm of case epoxy was
on 80C/W. Ok there is the air but pot this lot and you have a block
3x3x21 plus the pot depth on each side. And of course the tracks will
take heat as well as current. I think this would give a larger error
than measuring the heatsink.

The best that you could do would be to make all the connections to the
tab and use the c leg as the thermal coupling to the sensor tab. But
you still need to get this to low thermal impedance to overcome the
problems highlighted above, loss to the air giving an error.

Malcolm


--

....malcolm

Malcolm Reeves BSc CEng MIEE MIRSE, Full Circuit Ltd, Chippenham, UK
([email protected], [email protected] or [email protected]).
Design Service for Analogue/Digital H/W & S/W Railway Signalling and Power
electronics. More details plus freeware, Win95/98 DUN and Pspice tips, see:

http://www.fullcircuit.com or http://www.fullcircuit.co.uk

NEW - Desktop ToDo/Reminder program (free)

 

[Jeff]

After reading several of the posts, how about just potting the parts
together? Just design the PCB so the parts are really close together, and
keep lots of copper around them. Next, after all the parts are mounted,
create some type of box/barcade to fill with highly thermally conductive
epoxy.

www.mgchemicals.com offers thermally conductive epoxies ands silicones that
are in the order of 0.65 W/m*k. Due to your small space requirements and
ablity to entirely coat the parts, this may be acceptable.

 

[Jim Meyer]

A good point but I can see practical difficulties. Tj-tab is 15C/W so
at 0.3W I have a 4.5C error to the heatsink. My sensor dissipation is
peanuts so the error is just 4.5C.

What I am trying to say is that the 4.5C error is the best you
can do and that it is easy to achieve. All the other mucking about
with alternate thermal paths won't buy you anything. Simply mount the
sensor device to the heatsink and don't expose the sensor to any air
flow or connect excessively large leads to it.

Jim