In my March post, I introduced the concept of
“thermal resistance” (Rthj-a) as a measure of how easily a system is able to lose its heat
energy to its ambient environment. We saw that, for a MOSFET device mounted on a single-sided PCB, Rthj-a
depends heavily on the amount of PCB copper to which the device is soldered. See Figure 1.
Figure 1 Rthj-a v PCB copper area for an LFPAK56
device
I then went on to pose the question: “Is there another way in which we could assist the device thermal
energy in its passage toward ambient? Would it help, for instance, if we put a heat sink on top of our
device?”, as illustrated in Figure 2.
Figure 2 Adding a heat sink to the top of the LFPAK56
device, and its associated thermal network
So can we predict the thermal performance of the new system (i.e. with heat sink attached) simply by calculating the
new equivalent Rthj-a – which we’ll call Rthj-a(new) – from the
series/parallel combination of Rthj-a, Rthj-c and RHS? If we carry out these
calculations, we end up with the second curve (“Predicted performance”) shown in Figure 3, gray line.
Note that Rthj-a(new) is found using the equation:
where
Rthj-c = 34K/W
RHS = 25K/W
Rthj-a is read from the graph of Figure 1
Figure 3 Rthj-a and calculated
Rthj-a(new)
Wow! It looks like adding that heat sink is really going to improve the thermal performance of our system,
doesn’t it? Let’s run the simulations now, just to confirm this impressive improvement in performance
(Figure 4, orange line)…
Figure 4 Rthj-a and simulated and calculated
Rthj-a(new)
Well that was unexpected. The results from simulation do not agree with those from the simple calculations.
Although the simulations do show some benefit from adding the heat sink, the benefits are not as great as the simple
calculations would lead us to believe[1].
So why the big difference between what we expected and what we achieved? There are several reasons why, but probably
the most important is that the ambient temperature “experienced” by the heat sink is not the same as
the ambient temperature for the system as a whole. If you cast your mind back to Figure 2 of my January 2016 post, you will
see an illustration of the blanket of hot air which appears above a PCB-mounted device when it is dissipating heat
energy. When we place a heat sink on top of the device, the blanket of hot air also envelopes the heat sink and
degrades its ability to further remove heat from the device. To put it another way: the heat sink is itself being
heated up by hot air rising from the PCB surface. I’ve illustrated this phenomenon in Figure 5.
Figure 5 Thermal simulation plot of an LFPAK56 device with heat sink on a
PCB.
In effect, the heat sink now has its own local ambient temperature (Tamb(HS) in Figure 5) which is higher
than the system ambient temperature Tamb. So going back to the thermal network of Figure 2 (top), we can
see that the network is not quite as simple as we first thought, and needs to be redrawn as shown in Figure 6
(bottom).
Figure 6 The original thermal network (top) and modified version
(bottom).
Sadly, predicting Tamb(HS) really is beyond the scope of simple calculation as it depends on the transfer
of heat energy to the surrounding air in three dimensions. So, whilst attaching a heat sink to the top of a
dissipating device will usually yield some benefit, the degree of benefit may be difficult to predict and will be
less than simple theory suggests. If we apply some forced-air cooling to the scenario then manual predictions will
become even more difficult – this is a topic I may return to in future blog posts.
I’m out of space now, but thanks for reading and I hope to see you again next time for more thermal ramblings!
[1] I
should also add that we have observed this phenomenon in real-life testing, so the results shown in Figure 4
cannot simply be explained away as some quirk of the simulation software.
