DPAK and D2PAK power dissipation

Dec 12, 2006 22 Replies

Hi everyone, I am trying to figure out which is the best way to make a GND plane act as power dissipation area for a DPAK/D2PAK device (the L78M10ABDT from ST for example). Apart calculation of the area used, which I found in their AN1703, I am asking myself how to design this area: I have components only on the top side, and tracks for 90% on the top side too of a 2 layer PCB. So I have part of the top layer and almost all the bottom for GND plane. My specific questions are:



1) how must I act with the solder stop layer? is it better to leave this "heatsink" area uncovered? both on top and bottom? or not?
2) vias: is it better to connect top and bottom with multiple vias or with few (I don't have limits from my pcb manufacturer)? big or little? (this providing a gnd plane also on top layer)
3) I have also, of course, a gnd plane on the rest of my pcb for the other devices (1 PIC , 4 relays and 4 other ICs, total current max
300mA@3.3V). should I make all a big plane, or make 2 polygons connected only in 1 point?

THese are the questions in my mind, of course all other suggestions are more than welcome.



thanks since now,



Liuc



I assume you mean solder mask. Thermally, it doesn't matter.

Lots of vias will conduct heat from the topside pour (around the dpak) to the bottom ground plane. Six or eight 30+ mil vias, close around the dpak, should work... something like that.

One big plane!

John

But watch out for relay coil noise on the digital ground!

Cheers! Rich

Digital ground? Ground is ground. And there can't be noise on ground; you just define it to be zero volts!

John

I predict that you are going to find that life is full of surprises. ;-)

Yes. in theory you can define a 'ground' point as being at zero.

However any real real circuit that has currents flowing through wires will have non zero voltages due to the resistance of the wires, inductive effects when the currents change (i.e. relay noise or other switching noise effects) and cross talk due to magnetic field coupling between signals, etc.

At high frequencies there are other problems that have to be considered related to signals bouncing off impedance mismatches, etc. Ask anyone that works with RF, microwave, or modern high speed computer circuits. (FM radio is at 100 Mhz, this is a low frequency by modern computer standards.)

Yes. in theory you can define a 'ground' point as being at zero however if you look at any other point in your circuit, the voltages will not be zero even on wires connected to your 'ground' point.

["Followup-To:" header set to sci.electronics.design.] On Tue, 12 Dec 2006 08:05:01 -0800, John Larkin wrote in Msg.

close around -- or right beneath it. Anything wrong with that?

robert

Does anybody sell that ground? I've been looking for it on some of my recent projects.

Oh, life certainly is. And splitting ground planes multiplies the surprises nicely.

Exactly!

Sure. Just deal with it.

John

That's kinda what I was getting at - watch that the supply return to the sensitive digital stuff doesn't cross/share paths with the 24V relay supply return. and that neither crosses/shares with the sensitive analog stuff. I guess this can be done by visualizing current paths - it seems a split plane is kinda the lazy person's way out. ;-)

Chers! Rich

Go to an aviation boneyard, and pick up one of those boxes of dirt that they put in the tails of airplanes to ground their avionics to. ;-)

Cheers! Rich

Send me a purchase order, and I'll make you a deal on 1000.

John

I read in an app note somewhere that suggested instead of using large holes, use the smallest hole size that would at least 85% fill with solder to make a better thermal connection. You should be able to ask your board house what this size is.

Marc

Solder filling isn't particularly reliable esp with lead-free.

A larger hole introduces more copper which is a good thing (tm). I use about 2mm holes for this.

Graham

A 20-mil drill (1 mm to you furriners), plated to 1 oz, through a

0.062 thick pcb, has a thermal resistance of about 70 k/w. Theta falls as diameter increases, and drops hugely if you can fill the bore with solder, which would usually have to be done by hand.

So maybe half a dozen 30 mil holes would get down below 10 k/w net, pretty good considering the thermal resistance of the surface copper itself. Using internal planes helps a lot, as the theta to them is less and they can help spread the heat laterally, especially if you can include a second ring of holes further out.

I haven't figured out if it's better to use a few big vias or a lot of smaller ones, or what the optimum pattern might be.

Nowadays, of course, the board houses seldom actually give you the copper thickness you spec.

John

2mm
0.8 mm in fact.

Graham

2mm
0.508 mm in fact.

John

about 2mm

No.

0.787mm if you want me to be more accurate but 0.8mm is a standard drill size over here. I've never seen anyone claim to drill smaller than 0.6mm without complications btw.

Graham

about 2mm

over here.

btw.

25.4 mm to the inch, right? So 0.02 inches must be 0.508 mm. Google agrees.

The smallest vias we regularly use are 10 mil drills, 0.254 mm. That's good enough for a 456 pin FBGA package, or a very dense board in general. Heck, some bga balls are on 0.8 mm centers. I've heard of people using 4 mil vias.

John

about 2mm

over here.

btw.

?? drilled via sizes of 15-20 thou are pretty common with. 4 thou or less is possible these days (0.1mm) but I don't think they are conventionally drilled (laser).

Consider these drill sizes:

#77 = 0.018" (0.45mm) #78 = 0.016" #79 = 0.0145" #80 = 0.0135" (0.34mm)

Which make up roughly 50% of the sizes in an assortment of carbide bits which I picked up, presumably representing how frequently they are used.

I'd hate to think of 2.5, 3 or 4 mil lines and spaces with 31 mil vias.

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

many small holes have more surface surface area, and so without solder filling will have more conductivity,

with it's probably faily close.

Ofcourse the best is one huge hole and solder a copper heatsink into it. :)

Bye. Jasen

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required