D PAK heat dissipation

Thread Starter

kiyoukan

Joined Nov 17, 2009
63
Hello, so a friend is building a circuit for a laser driver.
And im going to produce and assemble it.
Now i have been testing parts but have found one that is perfect in every way but its package.
http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/DATASHEET/CD00110236.pdf
for use in this circuit.
http://i48.photobucket.com/albums/f248/Cyparagon/CapstoneMultisimSchematic.png
That is not the final we have made some changes but you get the general idea.
We will be using this to drive lasers in our projectors so the driver will not just be on and off it will be sitting at in between points for different color mixing as there will be 3 drivers in each unit one red one green and one blue.
What i am worried about is the heat that will create.

Now i have a mill and thought about milling out a block to sit over it and extend the copper pad and vias out under neath the blocks feet.

but the power is hard to calculate as for some lasers you could use 5V and others will need 7.5V supply.
Some run at powers of only 300ma and other run at 1.4A.
So if you guys have any advice let me know.
 

steveb

Joined Jul 3, 2008
2,436
So if you guys have any advice let me know.
You are correct to be worried. You have to be very careful when using a DPAK and dissipate heat through the board. When you start mentioning currents over 1A and voltages over 5V, I get worried. You need to be very careful of the excess voltage dropped on the transistor and the current level.

I used to run high power lasers up to about 1 A with a 5 V supply using a MJD243, which has a DPAK package. It worked fine, but the laser was dropping a good part of that 5 V.

If you are considering doing additional heat sinking, I would recommend using a different package that is designed to mount to the heat sink rather than the board.

If your design is on the border line, you can use that special compressible heat conductive pad that lays over the part and conducts heat to the metal package that the circuit board is mounted in.
 

Thread Starter

kiyoukan

Joined Nov 17, 2009
63
the problem i have is i have no idea how other people are going to use it.
With the different kinds of lasers and different power levels.
Some only need 2.2V others need 4.6V some people may run all of them off 7.5V and let it drop the voltage.
And in case they do i want to make sure it can handle it.
I have looked for other packages that have the same high hfe gain but it seems hard.
Also i am only able to use STmicro parts.
the best i was able to find was a unit no longer in stock http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/DATASHEET/CD00172270.pdf
So i went looking for others but still cant find any in a package meant to be mounted with that hfe.
 

bountyhunter

Joined Sep 7, 2009
2,512
What i am worried about is the heat that will create.

Now i have a mill and thought about milling out a block to sit over it and extend the copper pad and vias out under neath the blocks feet.
I was the resident expert on heat transfer and power dissipation at my last job. Your block won't work at least not to any useful degree. Here are the absolute rules of heat:

1) You will get virtually no heat tranfer out the top of a plastic package. Plastic is too much insulation.

2) The traces on a PCB are shockingly linited in being able to pass heat because the copper is so thin. Power dissipation is limited to a few Watts at most.

3) Effective heat transfer requires the internal die be on a metal paddle and that metal tab must attach to the metal heatsink. For example, a TO-220 device with a metal tab screwed to the heatsing.

I wrote a good article on this but am reluctant to post it since it hasn't been published yet.
 

Thread Starter

kiyoukan

Joined Nov 17, 2009
63
Okay well thanks then.
I got the idea from seeing these.

http://www.fischerelektronik.de/uploads/pics/PM_LFPAKgr_01.jpg
And thought i could build a better version for this application.
So there is no easy way to heatsink a DPAK package short of not smt mounting it but mounting it vertical and epoxying it to a heatsink kinda like a TO-220 then solder a wire to the top exposed pad back to the board?
Man i hate this package every thing about this transistor is perfect but the package.
 

SgtWookie

Joined Jul 17, 2007
22,230
If you really want to use a linear regulation scheme like you have so far, and get rid of much of the power dissipation problem, you could use a switching buck regulator to minimize the voltage drop across the transistor - which would minimize the power dissipation within the transistor.
 

bountyhunter

Joined Sep 7, 2009
2,512
So there is no easy way to heatsink a DPAK package short of not smt mounting it but mounting it vertical and epoxying it to a heatsink kinda like a TO-220 then solder a wire to the top exposed pad back to the board?
Man i hate this package every thing about this transistor is perfect but the package.
That won't work either. Epoxy is a VERY poor heat conductor. SMT pakages simply are not designed to handle much power, like I said a few Watts. If you use a multi layer PC board with a lot of vias, you might be able to get the thermal resistance down to maybe 20C/W so you might be able to safely do 5W. problem with PCB heatsinking is there are always multiple power devices heating the board so that degrades the cooling as well. SMT is not the way to go if power is a concern, it's just a cheap solution.
 

chrisw1990

Joined Oct 22, 2011
551
FYI, That package can not be hand soldered, it has to be done using wafer solder and infra red machine.
pash, the only things that cant be hand soldered are bga and QFN packages where the pads are underneath, iv handsoldered hundreds of tight pitch SMT (QFP100 and higher pin count!) parts and that ones not even worth losing sleep over, buy some cheap solder flux off ebay (search solder paste, seems to come up with that for some reason) and makes it a cinch!.
 

bountyhunter

Joined Sep 7, 2009
2,512
pash, the only things that cant be hand soldered are bga and QFN packages where the pads are underneath, iv handsoldered hundreds of tight pitch SMT (QFP100 and higher pin count!) parts and that ones not even worth losing sleep over, buy some cheap solder flux off ebay (search solder paste, seems to come up with that for some reason) and makes it a cinch!.
The package he showed had a "heat slug" underneath which is an exposed metal surface to the dia attach paddle. To get heat out, that has to be soldered down to the PCB copper directly underneath. It's not the connection leads that can't be hand soldered, its the heat slug surface that has to sit on a solder wafer and be IR flowed. It can not be hand soldered.
 

Attachments

Last edited:

bountyhunter

Joined Sep 7, 2009
2,512
natural convection only.. no forced air.
There is no spec for still air, but looks like it would probably be about 10 - 12C/W from mounting surface to ambient (estimated from curve below). There is probably at least 2 -3C/W resistance from die to mounting surface, for maybe 12 - 15C/W total from die to ambient. 9W diss gives you about 110C rise above ambient, which is extremely hot.
 

Attachments

Last edited:

steveb

Joined Jul 3, 2008
2,436
There is no spec for still air, .
He said natural convection, not still air. There is quite a difference between the two.

It's hard to extrapolate convection cooling performance from forced air cooling plots. First of all, the orientation of the heat sink has a big effect, and the air velocity changes vertically over the heat sink.

Convection can work quite well if the system is configured correctly.

There are formulas one can look up that give some ideas, but they might not apply to that unusual heat sink design.

All in all, I'd trust measurement data. If someone does the test and measures the temperature rise, it's hard to argue with it.

I've seen mechancial engineers provide detailed finite element simulation results that said a design would overheat, and then the test engineers come over and laugh at them because they have test results showing very cool operation.
 
Last edited:

bountyhunter

Joined Sep 7, 2009
2,512
He said natural convection, not still air.
.
Same thing for heat sink specs. Either you apply airflow or you don't. Obviously air near a scorching hot component is not dead still, but it is not being pushed either and the plots for zero airflow are for natural convection, on repurtable heatsink makers who supply the data at zero airflow.

I've seen mechancial engineers provide detailed finite element simulation results that said a design would overheat, and then the test engineers comes over and laugh at them because they have test results showing very cool operation.
Funny, the 33 years I spent in power design showed the reverse about 99% of the time. People don't understand thermal design and rely on sims that tell them all is well and then call me when they don't understand why IC's dont operate at junction temperatures of 170C.


EDIT TO ADD: Here is a thermal plot from a good vendor (thermalloy). The curve approaching the left axis starts to take on a "1/R" look, rising faster than linear.
 

Attachments

Last edited:

steveb

Joined Jul 3, 2008
2,436
Same thing for heat sink specs.
So you are saying natural convection cooling and still air cooling are the same thing?

Please think carefully about this.

Look at one of the simple formulas for heat sink thermal resistance with force air flow as follows ...

\(R(v)={{0.264}\over{W\sqrt{v L}}}\)

where R is resistance, v is forced air velocity, W is heat sink width and L is the length of the heat sink in the direction of air flow.

Look at what happens as velocity goes to zero. Clearly, resistance blows up, which would not happen in the real world, but this shows that resistance can get very large when velocity is small.

When you have natural convection, velocity is replaced with an effective average velocity that depends on temperature difference between the heat sink and ambient. The velocity is no longer constant over the length of the heat sink and it is no longer constant versus temperature. A simplified formula can be provided as follows.

\(v_{eff}=0.143\sqrt{L\Delta T}\)

where \(\Delta T\) is the temp difference.

Both of the above formulas are approximations, but they display the important features.

What this means is that, with natural convection, resistance is now dependent on temperature and the hotter the heat sink gets, the faster the average flow velocity and the lower the effective heat sink resistance. In a sense this is a nonlinear feedback effect that can keep the heat sink temperature in a reasonable range.

Natural convection implies that the heat sink is oriented correctly and that the air is allowed to flow unimpeded.
 
Top