How to heat-sink properly the PWD13F60

[ali_asadzadeh]

Hi,
I have an application that I think these new devices would help a lot, but the problem is what is the best way to design it so it could achieve maximum heat dissipation, in a fan-less design! can we get 8A out of it? have you done any project with them?

[senso]

Good thermals pads and holes in the pcb to mount them against the case/pads?

And 8A is the absolute maximum current DC, I wouldn't design anything going by absolute maximum ratings..

A fly farts in another room and it all explodes.

[ali_asadzadeh]

A fly farts in another room and it all explodes.

I would agree with that, But I thought the max was 32A

[Siwastaja]

Look at the datasheet:

Absolute maximum ratings (these have little to no margin; you have to add your own):
Id, DC when "case bottom pad" is at 25 degC = 8A.

This means, if you have some imaginary perfect heatsink with some refrigeration compressor thingie running, so that you can actually force the pad to stay at 25 degC, while removing the heat, then you can run at 8A.

For a more realistic application, look at this:
Id, DC when "case bottom pad" is at 100 degC = 6.9A.

Now, 100 degC pad temperature is realistic. If your ambient (near the PCB) is at 60 degC max, you'll have 40 degC gradient to play with, when your current is at 6.9A.

You need to understand this DC number totally ignores switching losses. If your switching frequency is small compared to the edge rate, you may be able to ignore it, but OTOH, in many real-world cases, switching losses can be 50% of the total losses easily. In which case, you are down to 6.9A/sqrt(2) = 4.9A.

So, expect around 4-6A out of this puppy. Which is still around 2 kW of actual power delivered at the intended DC bus voltage, through such a small device!

Now, if you apply very thorough and well though out heatsinking techniques, and your switching losses are almost nonexisting, you might be able to push the current up to around 7A, but that has little margin left. This would be achievable with aluminium core PCB mounted to a heatsink with thermal pad material. If you can't do that, using 2oz copper and adding 0.4-0.5mm thermal vias filled with copper under the chip - then mounting the board with thermal pad to a heatsink - helps as well. If you can't have copper-filled vias, then you need to add standard vias just outside the chip area.

Some homework for you: read about
1) switching losses,
2) thermal resistance calculations so that you understand what Table 4 means and how to apply it to your design.


Finally, a sanity check:

You are expecting this part to drive motors (or function as a part of EV fast charger, etc.) at 32A * 600V = 20kW. Assuming 99% efficiency (which may be unrealistic!), you would be dissipating 200W. Is it realistic for such a small package? Maybe this 10x13 millimeter $3 SMD part is not meant to drive 20kW motors/things, which normally are driven by massive modules inside $5000 products.

[ali_asadzadeh]

Thanks Siwastaja, It helped a lot, 2Kw is certainly more than enough for my app! I wanted to know have someone attached an aluminum heat-sink to these kinds of babies!? because almost all the passives around them have a higher height!

[Siwastaja]

Attaching a heatsink on the top is little help. The plastic of the chip is too thick and conducts little heat.

You need to use thermal vias through the PCB and attach the heatsink to the PCB using thermal pad material. If you can do single-sided assembly of all components, then this is easy. Otherwise, make cutouts to the heatsink for all bottom-side components.

[ali_asadzadeh]

Thanks for the hints

[johny_oo]

Hello to everyone. I'm making a 200W frequency converter for motor driving with APFC (~85...265 AC input) at my own 2-layers PCB and have some questions about influence from one half-bridge to another half-bridge inside the PWD13F60.

Parts are driven in unipolar modulation mode: HIN1, LIN1 are driven with 23.1 kHz square-wave signals (dead-time 1 us) which represent sine wave signal as their duty cycle is proportional to sinus amplitude (lets call it HF part). HIN2, LIN2 are driven with 10...100 Hz simple square-wave (lets call it LF part), the load is connected to OUT1-OUT2 with filter (3 mH, 3 uF) and the output is sine-wave signal with frequency of LF signal.
(ST_sine_out.jpg)

VS=60V, impulse amplitude at inputs of PWD13F60 is 5V, VCC=15V, the PCB is shown below (APFC is at the right and is in OFF state, VS was taken from lab power supply).
(ST_boardview)

While measuring there is repeating interference at LF impulse with frequency of HF switching and its amplitude is dependent from VS, so at some VS value this interference become so huge that it opens LF part FET while it needs to be closed (!!!) and I have overcurrent protection enabling.
(ST_50HZ_all)

zoomed:
(ST_on-off_50Hz_peaks)

impulses:
(ST_50HZ_pulses_zoomed)

And waveform of HF signals are suitable as for "upper FET" and for "lower FET" as can be seen:
(ST_off-on_23k)

My question is:
Can it be because of inner intercoupling capacitance of IC or I need to modify my PCB and maybe add LPF to LF inputs of PWD13F60 IC? Do I need to use maybe full GND polygon under IC or there is some addition hints.

I have read about distances of 1.2mm between VS and other pads and will remake layout (was 1,05.....1,15mm) so is it need to remake something else. Or the whole PCB?

Thank you in advance!

P.S. The case temperature was about 60 C under 0.9A load and 60V VS and also dependent from VS value (VS=30V, 0.9A and only 45 deg temperature). Charge dissipation increase???

[ali_asadzadeh]

Do you have some schematics and PCB layout? it may help

[johny_oo]

Ye. Attached power part of schematic and PCB (green is bottom layer). I know that maybe my GND layout isn't correct because of very long ways from "big GND point" to gnd1 and gnd2 of pwd13f60. I think, that is not only one mistake that causes so huge interference...

P.S. On pcb layout out1 needs to be out2 and vice versa.

[cryosphere]

johny_oo,  did you succeed PWD13F60 run at what max current?

Can you help me with electronics? Am I recognize Ukraine flag? 
if so, take a look cryosphere.co

 

[johny_oo]

1) Yes! I've changed my PCB layout to divide sources into 2 branches that cause need to use 2 separate OpAmps for overcurrent detection. And now it works with about 200W motor as a load. But PWD13F60 needs active cooling with my 2-layers PCB. In this conditions we have about 45 deg C on the case with 1.1 A load current.
2) Yes, you're right with the flag. We can use some messenger or PM.

[cryosphere]

Johnny,  да, первое-второе сообщение получил.
как вам идея: две сборки PWD13F60 (использовать 1/2 каждой, они же симметричные) установить на керамическую PCB. PCB as heat sink
остальное вторым этажем
victor.zagorodnov@gmail.com