pcb heatsinking areas.

[Simon]

I'm looking to use an LTC3637 SMPS chip. It has a pad underneath it to put heat into the PCB ground trace. But how big does the copper area need to be ? How do I calculate the area of copper needed to dissipate a certain amount of heat ? I would expect copper weight to make quite a difference as it means more cross section area to conduct heat.

[Precipice]

It's a heat spreader, not a heat sink that magically makes heat go away...
Or: What's getting the heat off your board, once the copper's spread it around a bit?
Datasheets concentrate on the copper area, and assume that the board is cooled, because they can't do any better. You have to!

[Simon]

Well heat spreader then, it's a surface of copper, if it is at a higher temperature than the ambient around it will dissipate heat into the ambient. It won't be a great heatsink because of the small cross sectional area.

[Precipice]

Yeah, and the 'surrounding area' is your PCB. How are you getting the heat out of that?

Or - what 'ambient temperature' are you going to invent, to tell the copper area calculator?
The real answer is - this is fantastically complicated, and borderline impossible to model. Overengineer it to hell and gone from the start, or make many tedious revisions to your design...

Cutting it close, on thermal design, never works out well.
Is there other hot stuff on the board?
Is there thermally sensitive stuff on the board?
Is there airflow?
Is there a conduction path out of your board, to the box?
Is there a conduction path from your box, to the outside world? Convection? Radiation? How hot can ambient be?
Will anybody promise you anything on environmental, and keep those promises in the field?

These all matter a huge amount more than finessing the copper area under a chip.

And, when it comes down to it - why would you do anything other than make it as big as possible, and cover the board with as much copper as possible, as a ground fill? Ground is (generally) good, copper is generally good, and vias for stitching are cheap.

[Simon]

It's not hugely complicated I would hope for my collegues who work on heat transfer can help but wondered if there was already some good rules of thumb. Yes there are a lot variables, since when does engineering not have lots of those.

Yeah, and the 'surrounding area' is your PCB. How are you getting the heat out of that?

It may surprise you to know that any two differences in energy try to balance themselves, things roll down hills, things cool down and well just to make our nearest and dearest field of engineering work and be more exiting than a piece of ice at -10C electrical charges will balance whenever they get a good shot at it and we spend our time trying to control that and make them do something useful for us while they are at it. If something is hotter than something else it has more energy and will try to balance it.

So you nee to look at how quickly the cross section area can conduct heat and how quickly the surface area of the copper will dissipate into the cooler air.

Yes i have an ambient figure: 86C that's why over engineering it might not cut it. Chances are the chip won't make hardly any heat as I'm only drawing 20-50mA tops and it's designed for 2A and 160C

[Jeroen3]

Well I guess there must be some documents around.
http://www.ti.com/lit/an/snva419c/snva419c.pdf
http://www.ats.net/wp-content/uploads/2014/04/Advanced-Thermal-Management-Solutions-on-PCBs-for-High-Power-Applications_APEX2014.pdf

Someone must also know a rule of thumb in C/mW per square cm?

[wraper]

It actually quite depends a bit on circumstances. If you are using multilayer board, that won't matter so much as the board will conduct heat quiet well. If only 1-2 layers, that pad will need to dissipate most of the heat by itself. If heat is significant, there are usually a lot of vias (often filled with solder) in that pad to conduct heat to other layers.

[Precipice]

OK, rule of thumb - that chip, dissipating the power it will to generate the power you're asking, will not need any heat spreading at all. It'll be within a couple of degrees of the PCB temperature.  Work out the power it's dissipating (you've got the efficiency curves), and the Tja numbers. Remember the diode will be generating heat too. That'll give you the degrees above ambient. 

So, how are you going to control the PCB temperature?

^^^ is the real question. And it's not answered by the 'how much copper area do I need under the chip' question.

(Last I heard, you were doing automotive things. In some automotive situations, you're hard mounted to a lovely great heatsink... In others, you end up sandwiched between a hot plastic body and a carpet, with dreadful heat paths and stagnant air insulating you. Whether your board is hard-mounted to a case or shock-mounted, will make an order of magnitude of difference to your heat path. Whether there's airflow or not, likewise. Since we can't guess, and you won't say, then you're not going to get firm answers, or even firm rules of thumb.
It's entirely possible your design is easy, and thermals can be trivially managed - but, without information, it's hard to guess.)

[Simon]

So, how are you going to control the PCB temperature?

^^^ is the real question. And it's not answered by the 'how much copper area do I need under the chip' question.

I am asked to design for a top ambient of 86C so providing it can live in 86C it will live at anything lower than that. How or why my ambient could see 86C is not my problem. The chip will supply up to 1A and I'm asking for not more than 20mA on average which is 2% of capacity so likely i will be ok.

OK, rule of thumb - that chip, dissipating the power it will to generate the power you're asking, will not need any heat spreading at all. It'll be within a couple of degrees of the PCB temperature.  Work out the power it's dissipating (you've got the efficiency curves), and the Tja numbers. Remember the diode will be generating heat too. That'll give you the degrees above ambient. 

Are you saying that the FR4 is acting as a heat sink/conductor ? I have no good how well FR4 conducts heat. It's not a hot board so influence from other components is limited.

[Precipice]

Yes, FR4 conducts heat pretty well. Copper is better, of course, but if you're just trying to spread half a Watt over a PCB, FR-4 won't struggle. The copper's a bonus.

Hmm, see if you can borrow a thermal camera. Go play, look at similar boards. It's fascinating.

(Is that 86oC your PCB temperature, or the ambient temperature outside your PCB & box?)

[Simon]

It's 86C ambient outside of the box which is aluminum so should move heat pretty well. A thermal imaging camera ? what are you thinking ? I only said heat transfer company, not "good heat transfer company"   

[Precipice]

Dumping your tiny amount of generated heat into an aluminium box (whose temperature is someone else's problem) - yeah, you'll be fine. Although don't touch it!

(dumping a couple of watts into a tiny plastic box tucked into the car's head-lining, with direct sunlight streaming through the windscreen, has maybe made me a bit paranoid about this stuff...)

[Simon]

Well the space the control box is in is relatively enclosed and i have a 1KW motor with it's controller cooking away but I was told 86C so that is what I'm sticking to. If they want to go higher then they need to look at their side of the design not expect the impossible. As it is I've gone from linear regs to SMPS to try and keep up with ambient temperature requirements.

[T3sl4co1l]

The thermal conductivity of FR-4 is pretty poor, and through-plane (it's not isotropic!) is just abysmal.  However, if you have multiple layers of copper (even thin foil stuff) spanning across the board, you can improve the lateral conductivity by several times.  And with a forest of vias through the board, you can improve the through-plane conductivity by orders of magnitude!

A device with, say, 1 cm^2 pad, can quite reasonably be heatsinked to the tune of 10W or so, using a combination of thermal vias (filled with solder, if possible) and using grease or thermal pads to attach a heatsink to the PCB.

Tim

[Simon]

I have seen a fan speed controller where the switch mosfet and back emf diode had vais under them and a small heatsink soldered to the other side.

[Precipice]

These things are also quite handy -
http://www.aavid.com/products/standard/573300d00000g
 shortest possible thermal path from the pad to the heatsink, easy to assemble and cheap. If you just need a bit more dissipation than the PCB can offer and have some airflow, they're great. I often lay them out the mounting pads next to DDPAKs and stepper drivers, just in case. The come in reels for surface mounting, as just another component.

However, for Simon's tiny heat load and probable lack of airflow, I think there's no need...
(Making it possible to fit a thermal pad between the PCB and the box might be reassuring, though. Mounting screws need to be positioned carefully to avoid flex, though, and I still think it'll be unnecessary. Measuring temperature rise with a couple of taped-on thermocouples, or even a temperature sticky or two, should be convincing enough)

[Simon]

I'm using a TSOP (LTC3637) or will be as it seems to be the only option at such a temperature range I should be ok as it's 20mA sporadic out of a chip that is designed to supply at least 50-100mA with a top limjit of 1A I can't sink to case as I can't use it for electrical connections and it would be very impractical with this package.

[Precipice]

squishy gap pad, like
http://www.bergquistcompany.com/thermal_materials/gap_pad/gap-pad-5000S35.htm
lets you dump heat without an electrical connection, and without super-careful machining and clearances.
Still not saying you need it for this - but it's a useful thing in your design toolbox.

[KerryW]

Rule of thumb: For single or double sided boards, 1000/cm^2 degrees C per watt.