:: Calculate SMD thermal sinking / layout requirements (using VIAS)
This write up is to simplify the entire process of calculating what kind of vias should be used to maximise tiny heatsink foot prints of PCB pads of tiny packages. the example used here is a 12pin DFN with a heat sink pad.
The important reference we need to refer to are from widely available PDFs searchable online which describes the common thermal conductivity factor (K) characteristics of materials used in circuits (described in W/mK)
copper = 355
FR4 = 0.25
Solder 63/67 = 39
first we divide factor characteristic K by the thickness of the PCBthe heat needs to pass thru. In my case is 0.6mm. This gives W/m-2.
FR4 = 0.0004
Solder 63/67 =0.065
Then we need to do the math involved with heat going thru just 1 via. Assuming the via plating is 0.5oz (0.01778mm thick), the exposed surface area of the various parts of a 0.6mm via are :
via (big circle minus small circle area) = 0.03252mm2
solder in the inner via cavity circle = 0.2502mm2
the surrounding FR4 material = 0.676mm x 0.676mm minus the via = 0.1742mm2
multiply all the above with the corresponding thermal characters of W/m-2
copper = 0.01924 W/C-1
FR4 = 0.00007 W/C-1
solder = 0.01626 W/C-1
combined total = 0.0358 W/C-1.
so for 1 watt of power going thru this 0.6mm via, the temp rise is = 1/0.0258 = 28.1C
We repeat the same procedure and find that a 0.3mm via does 77.2C for 1 watt.
And we try different holes and layouts
0.45mm surrounding 0.5mm
best estimate ~ 2.59C/W
full of 0.3mm
best estimate ~ 2.57C/W
the key problem is the limitations of not hitting the pins, and not going out of range of the PAD.
But that is not all, you still need an actual sink stuck on the rear to get it right. If the sink adds on another 1.2C/W, you are really adding alot of impedance to the pile of steaming sh*t.
If we take a chip like LTC3122 with a JC of 5C/W and add to it this 2.57C/W and the sink, per watt will see the junction go up by 8.8C. A 10watt dissipation gives ~ a rise of 88C + ambient of say 30C, and we have a total of easy 118C. This is not a happy way to make the LTC3122 work. If then this requires specialized PCB and specialized sinks, this will increase the production costs significantly and it makes the entire design process a waste of time, which I seem to be discovering now.