Latest / Mechanical Engineering Made Simple / Forced Convection Physics For Better Cooling

Transcript

Forced Convection Physics For Better Cooling · Mechanical Engineering Made Simple

0:00
Welcome back to Mechanical Engineering Made Simple.
0:02
You know you can spend weeks designing the perfect mechanism.
0:06
Oh. Absolutely you can.
0:07
Right. You were sitting there at your
0:08
workbench, staring at a brilliant circuit or, you know,
0:12
beautifully complex mechanical system that you were building.
0:15
The math checks out, the logic is flawless, and you power it
0:19
up. And then, well, you are suddenly
0:21
facing the ultimate enemy of all engineers, catastrophic
0:25
overheating. It really is the great humbling
0:27
force of engineering. Yeah, it is.
0:29
Because you can get everything right dynamically, but if you do
0:32
not account for the thermal load, your beautiful machine
0:35
turns into a very expensive, very short lived space heater.
0:38
Yeah, you see so many designs that just look incredible on
0:41
paper, but the moment they encounter real world resistance,
0:44
they quite literally meltdown. Exactly.
0:47
So today we are decoding the physics of force convection to
0:51
fix that. Because heat transfers not just
0:54
abstract math, it's physical. Very physical.
0:57
And it is entirely manageable once you actually understand the
1:00
leverage you can pull. So that is the mission for this
1:03
deep dive. It's a great topic.
1:05
We are taking a master class of porch to this, pulling from some
1:09
incredibly comprehensive engineering texts on 4th
1:13
convection and external flows. And we are translating dense
1:16
thermodynamic mathematics into a practical real world lesson.
1:21
Basically, we're taking a shortcut to being well informed.
1:24
We are going to answer the absolute most popular questions
1:27
you, the learner, are asking online about how to actually
1:32
pull heat off your objects using fluid flow.
1:35
Right. And to do that effectively we
1:37
have to start at the foundation. We have to look at the
1:39
morphology of flow, specifically, you know, forced
1:42
convection. Let.
1:42
Me make sure I have the ground rules straight for this class.
1:44
Forced convection means we are looking at fluid movement past a
1:48
heated object, but that movement is explicitly induced by an
1:52
external agent. Exactly we.
1:54
Were talking about slapping a fan on a chassis or using
1:56
mechanical pump or a blower. We are not just sitting around
1:59
letting hot air rise naturally through buoyancy, we are forcing
2:03
the issue. Right, that is the defining
2:04
characteristic. You are paying an energy
2:07
penalty, like running a fan or a pump to force the fluid to move.
2:11
And the key take away from the source material is that the
2:14
physical geometry of your heated object completely dictates how
2:18
that fluid behaves. The fluid does not just wash
2:22
over an object uniformly like some magic cooling blanket.
2:25
Right, it interacts with the surface.
2:27
Yeah, it creates incredibly complex behavioral zones.
2:31
OK, let's unpack this because trying to visualize invisible
2:35
air doing complex things is, well, it's tough for me.
2:40
I need a real world analogy here.
2:42
Sure hope for. Let's think of the fluid flowing
2:44
over your hot component like traffic on a massive multi lane
2:47
highway. OK, I can work with that.
2:49
Now the surface of your component, like the metal casing
2:52
of the chip, is the solid concrete barrier on the side of
2:55
the road. So what happens in the lane
2:57
right next to that solid barrier?
2:58
Well, right at the surface that shoulder lane, the velocity of
3:02
the fluid is 0. The fluid molecules are
3:04
literally sticking to the surface due to friction.
3:07
It is a complete traffic jam. Nobody is moving, but as you
3:11
move outward away from the surface and into the middle
3:14
lanes, the fluid speeds U layer by layer until you hit the fast
3:18
lane. That fast lane is what we call
3:20
the free stream speed, where the air is moving as fast as the fan
3:24
is pushing it. And that entire zone of changing
3:26
speeds from the dead stop at the surface up to the fast lane is
3:31
called the hydrodynamic boundary layer.
3:33
OK, so now how does the heat play into this?
3:35
Because we are trying to cool this object down, right?
3:38
Let's say the heat is like a rumor trying to spread through
3:42
the traffic in the bumper to bumper traffic.
3:44
Right at the surface, the cars or, you know, fluid molecules
3:48
are packed tightly together. Oh, I see where you're going.
3:52
So the rumor The thermal energy passes very easily from car to
3:55
car, but as you get out to the fast lane, the cars are zipping
3:59
by and they're too far apart to effectively pass that rumor
4:02
along. That is a highly effective way
4:04
to visualize it, actually. Because the molecules are stuck
4:07
in that traffic jam at the surface, they are absorbing the
4:09
maximum amount of heat from the object.
4:11
Then they slowly transfer that heat to the slightly faster lane
4:15
next to them, and so on. So just like you have a gradient
4:19
of speed, you also get a gradient of temperature.
4:22
That is your thermal boundary layer.
4:24
Got it. Here's where I always get
4:26
tripped up though. Do the temperature boundary
4:28
layer and the velocity boundary layer grow at the exact same
4:32
rate? Good question.
4:33
Because intuitively my brain wants them to just perfectly
4:37
mirror each other. Like if the traffic jam is an
4:39
inch thick, the heat rumor should also extend exactly an
4:43
inch out. Well, if we connect this to the
4:45
bigger picture, they absolutely do not always grow at the same
4:48
rate. Assuming they do.
4:49
As a fundamental trap for a lot of system designers, the
4:52
relationship between how those two layers grow is defined by a
4:55
dimensionless parameter called the prant Lello number.
4:58
The Prantle number. Yeah, it is essentially
5:01
comparing how easily physical momentum spreads through the
5:04
fluid due to friction versus how easily thermal energy spreads
5:08
through it. OK, give me a concrete example.
5:10
What if the fluid we are pumping is highly viscous?
5:13
Like, I don't know, a heavy engine oil.
5:15
So engine oil has a massive plant fold number.
5:18
The momentum diffusivity is huge because the oil is thick and
5:22
sticky. It wants to grab onto itself,
5:23
right? So your velocity boundary layer,
5:26
that physical traffic jam we talked about, it grows huge very
5:30
quickly. The drag extends far out into
5:33
the flow OK. That makes sense.
5:34
But oil is a relatively poor conductor of heat compared to
5:38
its stickiness. So the thermal boundary layer,
5:41
the zone where the heat is actually penetrating the fluid,
5:45
it stays incredibly thin and tightly hugged against the hot
5:48
surface. Wait, I need to push back on
5:49
that. Sure.
5:50
If oil is so thick and sticky and it creates a massive zone of
5:54
friction and drag, wouldn't all that physical friction actually
5:57
generate more heat? How does pumping oil help us
6:00
cool anything down? You are confusing the viscous
6:04
heat generation with the convective heat removal.
6:06
Oh am I? Yeah, yes, extreme friction can
6:09
generate heat, but in typical engineering cooling flows, the
6:12
heat coming off the electronic component completely dwarfs the
6:15
frictional heating. I see.
6:17
The real problem with oil is that you could have a massive
6:20
expensive pump pushing gallons of this stuff over a component
6:24
and you might look at that flow and think you are getting
6:26
massive cooling. Because there's a lot of fluid
6:29
moving. Exactly.
6:30
But because that thermal boundary layer is so so paper
6:33
thin, the actual transfer of heat is heavily constrained
6:36
right at the boundary. The fast moving oil in the upper
6:39
lanes isn't carrying any heat away at all, it is just wasting
6:43
your pump power. You have to know your fluids.
6:45
Print tilde number. Which brings us perfectly to the
6:48
most searched question on engineering forums when it comes
6:51
to thermal management. Let's hear it.
6:53
Let's say you are an engineer. You've designed a custom shaped
6:57
component. It is not a textbook flat plate.
7:00
It has weird curves. How can you quickly estimate the
7:03
heat transfer for this custom shape without running a massive
7:07
expensive computational fluid dynamic simulation ACFD every
7:11
single time you tweak a dimension?
7:14
The dreaded CFD bottleneck. Those simulations are incredibly
7:18
powerful, but they take hours or sometimes days to set up mesh
7:22
properly and compute. In early stage design, when you
7:26
are changing dimensions by a millimeter every 10 minutes, you
7:30
cannot wait three days to see if your chip is going to melt.
7:33
Which is why our source material spends a significant amount of
7:36
time detailing a stroke of mathematical genius from the
7:39
early 20th century, the similarity transformation
7:42
technique. I won't understand this because
7:44
it sounds like a cheat code for physics.
7:46
It essentially is a scaling cheat code.
7:49
Imagine mathematicians like Blasius and Faulkner scan
7:53
looking at boundary layers over 100 years ago, long before
7:56
computers existed, they realized they had a huge problem.
8:00
To calculate heat transfer, they needed to know the fluid
8:03
velocity at every point downstream along the plate.
8:05
Let's call that distance X. But they also needed to know the
8:08
velocity at every point moving vertically away from the
8:11
surface. Let's call that distance Y.
8:13
So they were trying to map a massive 2 dimensional grid of
8:17
changing temperatures and speeds completely by hand.
8:19
Exactly, but they noticed a pattern.
8:22
While the boundary layer gets thicker the further downstream
8:25
you move, the actual shape of the velocity profile stays
8:29
proportionally the exact the same.
8:30
So the curve of how the speed changes from the surface to the
8:34
free stream, it just stretches out.
8:36
You've got it. So instead of having to solve
8:38
for both how far down the plate you are and how far away from
8:41
the surface you are, they combine them.
8:42
Oh. That's brilliant.
8:43
They squished X&Y into a single mathematical variable, which is
8:48
denoted by the Greek letter ETA. By doing that, they created
8:52
standard mathematical building blocks.
8:54
So it means that whether you are looking at a tiny one inch metal
8:59
wedge or a massive airplane wing, the proportional math
9:03
behaves the exact same way. And the reason this matters to
9:06
you, the listener sitting at your workbench, is that it gives
9:08
us the standard models. We can use the equations for a
9:11
flat plate or a cylinder to quickly estimate heat transfer.
9:15
For complex shapes, we just need to calculate a few key numbers.
9:19
Right, the non dimensional parameters.
9:21
This is where the real practical engineering happens.
9:23
Let's walk through them starting with the Reynolds number.
9:26
The equation in the text is Reynolds number equals rho times
9:30
U * L divided by MU. What does that actually tell me
9:34
about my system? Well, let's look at the physical
9:37
meaning behind those variables. Rho is your fluid density, U is
9:41
the free stream velocity, how fast your fan is blowing.
9:45
L is your characteristic length, like the length of your circuit.
9:48
Board, OK, so all those on the top.
9:50
Right. All those on the top represent
9:51
your inertia forces. That is how much the fluid wants
9:55
to just keep barreling forward like a freight train.
9:58
And on the bottom. On the bottom divided by MU.
10:01
That is your dynamic viscosity. That represents the sticky
10:04
dragging forces holding the fluid back.
10:06
So the Reynolds number is just a battle between inertia and
10:09
stickiness. That is a great way to view it.
10:12
A low Reynolds number means the sticky, viscous forces are
10:15
winning. The flow is smooth and layered,
10:18
what we call laminar. Got it.
10:20
A high Reynolds number means the four inertia is so overwhelming
10:23
that the fluid trips over itself, starts turning, and
10:26
becomes turbulent. And then we get the payoff, the
10:29
Nusselt number. The text provides this equation
10:31
for a flat plate in laminar flow.
10:34
Nusselt sub X = 0.332 times Reynolds sub X to the one half
10:38
power times Prantle to the one third power.
10:41
That's the one. Why is that equation the Holy
10:44
Grail for thermal engineers? Because the Nusselt number tells
10:47
you if you are fluid flow is actually doing the work you are
10:51
paying it to do. It is the ratio of convective
10:54
heat transfer to conductive heat transfer.
10:57
OK, break that down for me. Let me put it plainly.
11:00
If your Nusselt number is 1, your fluid is acting like a
11:03
solid brick. It is just sitting there
11:05
conducting heat. But if your Nusselt number is
11:07
100, the movement of the fluid is pulling away 100 times more
11:11
heat than if it were perfectly still.
11:14
Oh wow, so that formula tells you exactly how much extra
11:16
cooling your fan is buying you. You just plug in your fan speed
11:19
and fluid properties, find the Nusselt number and boom.
11:22
Know if your component survives no days long.
11:25
See if D simulation required. Exactly.
11:27
It gives you the immediate physical heat transfer
11:29
coefficient. It empowers rapid prototyping.
11:32
Let's take that empowerment and do a physical walkthrough.
11:35
Let's step out of the math and onto the workbench.
11:38
Sounds good. Picture this.
11:40
We have a printed wiring board, a standard green PWB.
11:44
Right in the middle, sitting flush against the board, is a
11:47
high power integrated circuit package in ICA.
11:51
Classic scenario. Maybe it is a high end processor
11:54
or a power amplifier. It generates intense localized
11:57
heat, and it needs a heat sink, right?
12:00
We consult the text and it compares 3 different methods of
12:02
cooling using heat sink arrays. The first method is plate
12:06
stacks. Just a series of tall, parallel
12:08
flat metal plates sticking up off the chip base.
12:12
Now my instinct tells me that more plates equal more surface
12:15
area, and more surface area equals better cooling.
12:18
I want to pack as many plates onto that chip as humanly
12:20
possible. And your instinct would cause
12:22
your chip to catch fire. Wait, really?
12:24
Yeah, the text explicitly warns about this trap.
12:27
There is a strict mathematically defined optimum number of
12:30
plates. If you have too few, you are
12:33
right, you lack of the surface area to shed the heat, but if
12:36
you pack them in too tightly. The air gets choked out.
12:38
Yes, the flow resistance skyrockets.
12:41
The air coming from your fan realizes it takes way too much
12:44
energy to push through those microscopic gaps between your
12:47
tightly packed fins. Fluids are lazy.
12:50
They take the path of least resistance.
12:52
Exactly. So the air just bypasses the
12:54
heat sink entirely. It flows around the outside and
12:57
the center of your plate stack basically bakes in its own
13:02
trapped heat. OK, so a basic plate stack
13:04
demands a perfect balance between surface area and floor
13:07
resistance. That brings us to method 2,
13:10
which I look at and think who came up with this?
13:13
It is called offset strips. Offset strips.
13:16
Instead of 1 long continuous plate running the length of the
13:19
chip, you chop the plate up into short strips and stagger them.
13:22
It looks like a metallic checkerboard.
13:24
Why intentionally chop up a perfectly good heat sink?
13:27
Doesn't that ruin the smooth flow of air?
13:29
That is precisely the objective you want to ruin the smooth
13:32
flow. Remember our traffic jam
13:33
analogy? As the air hits the front edge
13:36
of a solid fin, that boundary layer, the velocity traffic jam
13:40
starts to build up and get thicker the further down the
13:43
plate it travels. A thick boundary layer acts like
13:47
an insulating blanket. It traps the heat.
13:49
Oh I see. So by chopping the fin into a
13:51
short strip, the boundary layer doesn't have enough physical
13:53
distance to get thick. Exactly.
13:55
The strip suddenly ends, the fluid is dumped back into the
13:58
open channel, it mixes, and then it slams into the leading edge
14:02
of the next staggered strip, right.
14:04
The boundary layer is forced to constantly restart at 0
14:07
thickness at every single leading edge.
14:10
And a thin boundary layer means massive heat transfer.
14:13
We are weaponizing the physics against itself by constantly
14:17
slicing the boundary layer down to 0.
14:18
It is brilliant engineering, but there is a catch.
14:22
There's always a catch. Offset strips heavily rely on a
14:25
known consistent direction of airflow.
14:28
You have to point your fan exactly down the channels of
14:30
those staggered strips. If the air hits them diagonally,
14:34
the whole system chokes. Which naturally brings us to
14:36
method 3, because what if you don't know where the air is
14:39
coming from? Let's say your circuit board is
14:42
shoved inside a weirdly shaped chassis.
14:44
There are ribbon cables blocking the path and the air is just
14:47
bouncing around randomly before it hits your chip.
14:50
When the environment is constrained and the flow
14:52
direction is unpredictable, you abandoned flat plates entirely
14:56
and you look at pinfin heat sinks.
14:58
Pinfins, yes. Instead of plates you have a
15:01
dense forest of little metallic cylindrical pillar sticking up
15:05
from the base. So what does this all mean for
15:07
the designer? It means that intelligent
15:09
geometry can actually beat raw fan power.
15:13
You do not always need to upgrade to a louder, faster
15:16
power hungry fan. You just need a smarter physical
15:20
shape to intercept the air. Exactly.
15:22
A cylinder is a cylinder from every single angle.
15:25
Pinfin heat sinks are incredibly insensitive to the approach
15:28
angle of the air. It doesn't matter if the air
15:30
flow comes from the north-south, east or West, the flow dynamics
15:33
around a cylinder remain consistent.
15:35
Now the. Text mentions they're especially
15:37
effective with top down axial fan.
15:39
Why is that? Well, think about a fan blowing
15:41
directly down onto the top of this pin forest.
15:44
The pins in the dead center of the sink experience longitudinal
15:48
flow. The air washes straight down the
15:51
vertical length of the pillars, but that air has to escape, so
15:54
it turns 90° and moves outward. Therefore the pins on the outer
15:58
perimeter experience cross flow. The air hits them horizontally.
16:02
I see the pin fin geometry is the only design that excels in
16:06
both those completely different flow regimes simultaneously.
16:10
So we have optimized our pin fins.
16:12
We have carefully routed our airflow.
16:15
But what happens when the chip is so ridiculously powerful,
16:18
like an extreme laser diode, that even a perfectly designed
16:21
heat sink can't strip the heat away fast enough?
16:23
We literally run out of thermal capacity in the ambient air.
16:27
We need to completely change the medium.
16:28
Exactly when standard force convection reaches its physical
16:31
limits, you bring in the heavy artillery turbulent jet
16:34
impingement. This sounds aggressive.
16:36
We're talking about blasting the component with a high velocity
16:39
jet of fluid. Let's walk through the anatomy
16:42
of a jet, because, as the text shows, it is vastly more complex
16:45
than just pointing a garden hose at a hot rock.
16:47
It really is. The physics of a jet happened in
16:50
three highly distinct spatial zones.
16:52
First, you have the nozzle exit. When the fluid first leaves the
16:56
nozzle, it has a perfectly uniform velocity.
16:59
That pristine center area of untouched speed is called the
17:03
potential core. But it doesn't stay pristine for
17:05
long. Right as it shoots out through
17:08
the ambient air, the outer edges of the jet start scraping
17:11
against the still air surrounding it.
17:13
Right, It creates drag. It starts tearing and mixing
17:16
with the engineers call entrainment.
17:18
And because of that tearing, that perfect potential core
17:21
start shrinking down into a cone the further away it travels from
17:25
the nozzle. Which is why a nozzle distance
17:27
is critical. You want your target to sit
17:29
inside that potential core before it fully erodes.
17:32
Then the jet finally strikes the target surface.
17:34
This is the third zone, the stagnation zone.
17:38
This raises an important question.
17:40
What exactly is happening physically at that stagnation
17:43
point? Because, well, stagnation sounds
17:46
calm, but blasting a high pressure jet into a wall sounds
17:51
incredibly violent. Oh, it is thermodynamically
17:53
violent. Imagine the mass of that fluid
17:56
hitting the solid wall. It undergoes an extreme rapid
17:59
deceleration in the vertical direction.
18:02
It hits the wall and completely stops.
18:04
Right? But mass cannot just disappear.
18:07
All of that high pressure fluid has to go somewhere, so it
18:09
undergoes massive acceleration transversely.
18:12
It shoots horizontally outward along the surface at incredible
18:15
speeds. And what is that violent
18:17
horizontal squish due to our boundary layers?
18:20
It absolutely crushes them. Because of that brutal change in
18:23
direction and severe horizontal acceleration.
18:25
The hydrodynamic and thermal boundary layers are squashed
18:29
their absolute minimum thickness right at that stagnation point.
18:32
And tying it all back to our earlier lesson, a near 0
18:35
thickness boundary layer means an astronomically high rate of
18:38
heat transfer. It provides surgical targeted
18:41
cooling right at the dead center of your hottest hotspot.
18:44
But the text draws a major distinction between two types of
18:47
jets, submerged jets and free surface jets.
18:51
Help me visualize the difference.
18:53
OK, so a submerged jet is discharging a fluid into an
18:56
environment made of the exact same fluid.
18:59
Think about taking a water hose and turning it on while you were
19:02
completely underwater at the bottom of a pool.
19:05
Oh gotcha. Or blowing a jet of air into a
19:07
room full of air. The edges of the jet interact
19:10
very heavily with the surrounding fluid of the same
19:12
density, which dilutes and tears apart the potential core very
19:16
quickly. By contrast, a free surface jet
19:18
is something like a liquid jet shooting through a gas.
19:21
Imagine standing on the edge of the pool and spraying a fire
19:24
hose through the open air to hit a target.
19:26
Exactly. There is very little friction
19:28
between the dense water and the lighter.
19:30
The liquid jet maintains its velocity, its shape and its
19:34
potential core. Much further, it delivers a much
19:36
more pristine, forceful impact at the stagnation zone.
19:40
That makes sense. If you have an electronic
19:41
component that is pushing the absolute boundaries of heat
19:44
flux, liquid free surface jet impingement is at the very top
19:47
of the engineering food chain. It is fascinating how the exact
19:50
same underlying rules, boundary layers, momentum, thermal
19:55
diffusivity govern everything from a gentle computer fan
19:59
washing over offset strips all the way up to literal liquid
20:03
artillery. The mechanisms scale perfectly.
20:05
The math holds up. It does.
20:07
But before we wrap up today's deep dive, I want to leave you,
20:10
the listener, with a final kind of provocative thought,
20:14
something from the text that might make you rethink your
20:17
entire approach to thermal design.
20:19
I know where this is going. We have spent this entire
20:21
session obsessing over the air. We talked about boundary layers,
20:25
heat sink fins, pin fins and turbulent jets.
20:28
Naturally, convection is where the active energy exchange
20:30
happens. But in obsessing over the air
20:33
above the chip, we often completely ignore the board
20:35
sitting directly underneath it. The text highlights a phenomenon
20:38
called conjugate heat transfer. The hidden thermal path.
20:42
Most engineers just treat the circuit board as an insulator.
20:45
They do, but researchers Nakayama and Park conducted some
20:48
fascinating studies highlighted in the sources.
20:51
They proved that if you establish a truly excellent
20:54
thermal bond between your hot component and a highly
20:57
conductive substrate within the board, for example, mounting
21:01
your chip directly to a 1mm thick solid copper plate
21:06
laminated inside the PWB right, that solid substrate itself can
21:10
carry more than 50% of the total heat dissipation.
21:14
Over 50%. Think about the implications of
21:17
that for your design. You might be agonizing over CFD
21:20
simulations, spending thousands of dollars optimizing complex
21:23
offset strip geometries and sourcing the loudest, fastest
21:27
axial fans on the market. Yeah, meanwhile, half of your
21:30
thermal battle was it's actually being fought and won by a
21:33
silent, perfectly still piece of copper sitting underneath the
21:37
silicon. It aggressively pulls the heat
21:38
away through solid conduction, spreads it out over a massive,
21:42
wide area of the board, and then allows the ambient air to gently
21:45
carry it away far from the crowded critical hotspot.
21:48
It's incredible. So the provocative question I
21:49
want to leave you with today is this.
21:52
Are you so intensely focused on moving the air above your
21:55
components that you are entirely ignoring half of your total
21:59
cooling potential lying completely dormant right beneath
22:03
them? It is a critical blind spot.
22:04
Do not neglect the solid mechanics of your board while
22:07
you are busy chasing fluid dynamics.
22:09
A perfect lesson to closeout the class.
22:12
Remember, design junkies master the critical few or the trivial
22:15
many will master you. If you're leaving this episode
22:18
with something you didn't arrive with, drop a comment.
22:21
Takes you a few seconds, but it tells me this workbench is worth
22:23
keeping open.