The great Freezer 64 Pro 2-stage fan duct experiment

Felger Carbon · Post by **Felger Carbon** » Sat Mar 03, 2007 4:25 pm

The idea is, you can double the pressure (but not CFM) by using two fans in series, with only a slight increase in noise, which can then be compensated for by dropping the RPM 11%. The theory is, the added pressure will help force air through the tightly spaced cooling fins. I decided to give it a try, and do it right!

This is the completely assembled duct, ready to be mounted and run. Two matched 100mm Scythe "L" fans, .062" acrylic housing, put together with double-sided tape and sticky foam. You can see the 6 holes for the heat pipes:

The bottom isn't sealed, but that's a necessity. This part of the duct sits really close to the motherboard:

This is how it all fits & sits on a Freezer 64 Pro/mobo:

Ready to run in my #1 computer. The weight of the two fans and the acrylic rests on the three gray foam strips, and friction holds the duct in place:

A 12" x 12" sheet of .060" extruded acrylic (McMasters.com 8589K11) is $2.31 for enough to build 2 of these with lots left over. The double-sided tape (76405A11) is $4.03 for 36yds of 1/2", and holds the assembly together firmly. No super glue anywhere!

The experimental results:

Compared to one identical fan, there was no change with the fan controller at 11V (max) and only a .3C improvement at 6.6V (min). Another wonderful theory that didn't pan out!

I conclude that the 2mm fin spacing of the Freezer doesn't impede the cooling air flow even at ~560RPM! Hmm...

bonestonne · Post by **bonestonne** » Sat Mar 03, 2007 5:27 pm

being so close and also with the fans being set identically, i wouldn't expect much change. if they were high performance fans rather than low powered i might expect a difference, but even with that, two aren't better than one in this case, as its just moving the same amount of air with twice the power. not really necessary really.if there were vents for air to escape the sides of the heatsink instead of being forced to move in a straight line, you might see a better change, but right now, its just that you aren't moving any more air, and now you have two motors heating up.

jaganath · Post by **jaganath** » Sun Mar 04, 2007 2:21 am

maybe the fans need to be in series on one side of the heatsink? or maybe the chosen fan doesn't provide much static pressure?

Felger Carbon · Post by **Felger Carbon** » Sun Mar 04, 2007 3:20 am

jaganath wrote:maybe the chosen fan doesn't provide much static pressure?

The point is, whatever pressure a fan provides, placing two fans in series doubles the pressure (if there are no pressure leaks). I compared two fans in series with just one fan of the same fan type, and saw absolutely no cooling improvement. Not even the pressure leak at the base of the duct accounts for no improvement!

Either the 2mm fin spacing is not impeding the airflow, or one fan wasn't turning for some reason. I'll look at that later today, when I'm fully awake in broad daylight, but I don't think that was the problem.

J. Sparrow · Post by **J. Sparrow** » Sun Mar 04, 2007 5:17 am

I tried a similar setup some time ago, putting an exhaust duct between the Freezer Pro and my 120 mm exhaust fan. I saw no difference, too, but I was quick to blame the duct for not being tight enough and letting air flow outside the cooler.

However, temperature dropped with an intake duct + an intake fan.

One possible explanation is that the Freezer is designed to quickly cool the heatpipes down by letting air out on the sides, which is impossible when you duct it.

Felger Carbon · Post by **Felger Carbon** » Sun Mar 04, 2007 12:44 pm

Both fans were turning fine. The new duct, and the old fan, are both held in place by gravity and friction, so they can be replaced without moving the computer except for removing its left side, so I'm using my #1 computer.

I'm going to repeat the experiment, with one and two fans, this time at 12V, 9V, 7V, and perhaps 5V. Perhaps because those fans won't start at 5V, although they should run, and the data (which I'll post) might be useful.

Either way, the computer I regularly use is being cooled just fine, and very quietly.

Kremmit · Post by **Kremmit** » Mon Mar 05, 2007 12:38 pm

J. Sparrow wrote: One possible explanation is that the Freezer is designed to quickly cool the heatpipes down by letting air out on the sides, which is impossible when you duct it.

That duct does look like it would stop any airflow in or out of the sides. A single fan, no-duct configuration would almost certainly have some side-flow, either in or out based on whether the fan was a in a push or pull orientation.

When you ran the single-fan test, was the same duct installed?

Felger Carbon · Post by **Felger Carbon** » Mon Mar 05, 2007 3:54 pm

Kremmit wrote:When you ran the single-fan test, was the same duct installed?

No. Today I unplugged one of the two fans, but left the "off" fan in place. The die temp rise went from 17.0C to 18.4C, at 6.6V (controller minimum) on the fan or fans.

I most likely won't pursue this further, because I discovered today how to run the Aerocool Dominator passively at 3.6V, with an 18.4C die rise above room ambient. That meets my requirements, with a 30C max summer temperature. So, I don't need multistage cooling on my CPU at all.

Kremmit · Post by **Kremmit** » Tue Mar 06, 2007 9:55 pm

Felger Carbon wrote:
Kremmit wrote:When you ran the single-fan test, was the same duct installed?
No. Today I unplugged one of the two fans, but left the "off" fan in place. The die temp rise went from 17.0C to 18.4C, at 6.6V (controller minimum) on the fan or fans.

If the duct wasn't present with the single-fan test, but was present with the 2-fan test, then the sideflow means you're not doing apples to apples. Additionally, the disconnected-but-still-attached 2nd fan is gonna cause some restriction that the single fan has to fight, and will also likely cause more air to flow through the sides, which, again, it couldn't do in the 2-fans-with-duct test. Those differences might account for your unexpected results.

Felger Carbon · Post by **Felger Carbon** » Tue Mar 06, 2007 10:38 pm

Kremmit wrote:If the duct wasn't present with the single-fan test, but was present with the 2-fan test, then the sideflow means you're not doing apples to apples.

You're 100% correct. I figgered out that what I have to do is build a duplicate duct with just the one fan. Not modify the original; if I did that I couldn't perform further tests if they seemed indicated. Problem is, I'm lazy and I have a thing called a Dominator to look at too. I'll get there, because the underlying issue - fin spacing - is common to both.

Bluefront · Post by **Bluefront** » Wed Mar 07, 2007 3:38 am

Most of my duct experiments prove to me that a tight duct that routes all airflow completely through a fin pack, provided superior cooling.......but also requires a faster fan speed and more noise. But there are so many variables, that YMMV with any particular setup.

Felger Carbon · Post by **Felger Carbon** » Wed Mar 07, 2007 12:05 pm

I think I've figured out what's going on.

My 2-fan duct has the same cross-sectional area from start to finish. This means the velocity of the air is uniform from the entrance to the exit. It seems to me that the first fan has to do the heavy lifting, and the second fan gets pretty much of a free ride.

A better approach would be where fan #1 accellerates the case air to V1 and the second fan to V2 - perhaps double V1. This requires fan #1 to have twice the area of fan #2. If fan #2 is 100mm, then fan #1 should be 140mm! It also means that no one fan has to increase the speed of the air from zero to V2. This seems to me to have a better chance of working more quietly (which is the objective) than one fan, or two fans of equal diameter.

I still need to test a one-fan duct against the 2-fan on the Freezer, and I will. That data should be added to the data base.

It also suggests I take a careful second look at the 220mm-120mm 2-stage fan data that I've already posted on the Andy Samurai, in addition to further tests of the Andy in passive mode with a proper duct. For higher-dissipation CPUs, a second quiet fan may be needed on the Andy.

On the other hand, while the Dominator seems to be working well passively with my Sempron64 2800+, there is no slow speed, quiet 140mm fan available at this time. So I don't see how a quiet 2-stage fan system can be based on the 220mm fan and the Dominator. This could change if a low-speed 140mm fan appears.

Post by **NeilBlanchard** » Wed Mar 07, 2007 1:00 pm

Hello,

I have a question: does the duct with one fan work better or worse (or the same as) the same fan mounted conventionally? Does it work better pushing air, or pulling air through the HS?

Oh, and are the two fans louder than one, or the same?

Felger Carbon · Post by **Felger Carbon** » Wed Mar 07, 2007 1:39 pm

NeilBlanchard wrote:I have a question: does the duct with one fan work better or worse (or the same as) the same fan mounted conventionally? Does it work better pushing air, or pulling air through the HS? Oh, and are the two fans louder than one, or the same?

Neil, I'll answer the first question after I build the second duct with one fan and perform some measurements.

At the same RPM(s), 2 fans are 3dBA louder than one fan. However, if the two fans in the 2-fan duct are both run at 89% of the one-fan RPM, then the noise levels will be equal. The question is, is the 2-fan approach (in the search for higher pressure) efficient enough to more than compensate for the reduction in RPM? [This leaves aside dollar and space considerations entirely.]

andyb · Post by **andyb** » Wed Mar 07, 2007 3:57 pm

Its not just the fin spacing that a low airflow problem but also the fin depth, I dont suppose you would be interested in hacking off most of the back of the heatsink in the name of silence

(the back as in the curved end).

My theory is that if the distance that the heat has to travel through the fins is reduced, you reduce its overall cooling potential, however the cooling potential is not present a low rpm's because of the added resistance due to the fin length.

The answer would be to reduce the fin length to relieve the pressure, and once the pressure is reduced more air passes the fins aiding cooling. Overall I would guess a slight improvement with 1 fan and a bigger improvement with 2 fans.

Andy

Felger Carbon · Post by **Felger Carbon** » Wed Mar 07, 2007 4:30 pm

andyb wrote:I dont suppose you would be interested in hacking off most of the back of the heatsink in the name of silence

No, but if you do it I'll be pleased to read the measured results, whatever they may be.

Kremmit · Post by **Kremmit** » Thu Mar 08, 2007 12:19 am

Felger Carbon wrote:I think I've figured out what's going on.

My 2-fan duct has the same cross-sectional area from start to finish. This means the velocity of the air is uniform from the entrance to the exit. It seems to me that the first fan has to do the heavy lifting, and the second fan gets pretty much of a free ride.

My understanding of the original experiment was:
Let V1 equal the air velocity Fan1 creates when it's spinning in free air. When ducted into the heatsink, the impedance would lower it to (V1 x 0.5) (or whatever speed the heatsink slows it down to). Fan2 is then added to pull against that impedance and bring airflow back towards or up to V1. Unless you were hoping that Fan2 could accelerate airspeed beyond V1, then your test was valid. Fan2 wasn't getting a free ride; it's job was to create the pressure needed to allow Fan1 to reach it's full potential in spite of that impedance. And, of course, that translates into the ability to run a lower fan voltage and speed, which is where the quiet comes in.

--------------

Felger Carbon wrote: On the other hand, while the Dominator seems to be working well passively with my Sempron64 2800+, there is no slow speed, quiet 140mm fan available at this time. So I don't see how a quiet 2-stage fan system can be based on the 220mm fan and the Dominator. This could change if a low-speed 140mm fan appears.

But you're only insisting on a 140mm fan because of an arbitrary spec:

Felger Carbon wrote:A better approach would be where fan #1 accellerates the case air to V1 and the second fan to V2 - perhaps double V1. This requires fan #1 to have twice the area of fan #2. If fan #2 is 100mm, then fan #1 should be 140mm!

Is there a reason V2 can't equal (V1 x 1.5) or (V1 x 1.7) or whatever? Perhaps a 120mm fan isn't out of the question.

But I'm not sure the premise is sound; if I'm right above regarding Fan2's job, then making Fan2 a fraction of the size of Fan1 just means that Fan2 creates less pressure for Fan1 to utilize.

On the other hand, if you're right about Fan1 doing all the heavy lifting, and Fan2 getting a free ride, then how does ducting from a large fan down to a small fan give Fan2 less of a free ride? Seems to me you're increasing the velocity being fed into Fan2, thereby giving it even less work to do. At a high enough velocity from Fan1, Fan2 would stop doing any useful work at all, and instead start creating drag!

Either way, I can't see mismatched fan sizes working out. It's the same scenario as mismatched fan voltages on a pair of same sized fans in series- the slower fan just ain't pulling it's weight.

Felger Carbon · Post by **Felger Carbon** » Thu Mar 08, 2007 1:46 am

Kremmit wrote:
Felger Carbon wrote: On the other hand, while the Dominator seems to be working well passively with my Sempron64 2800+, there is no slow speed, quiet 140mm fan available at this time. So I don't see how a quiet 2-stage fan system can be based on the 220mm fan and the Dominator. This could change if a low-speed 140mm fan appears.
But you're only insisting on a 140mm fan because of an arbitrary spec:

There is a danger we're talking past each other here. The 140mm fan on the Dominator is because the cooling grid on the Dominator is 140mm square.

Now, in the second paragraph above, I had moved on and was talking about fan1 of a 2-stage fan duct that had nothing to do with the Dominator. If that was what you meant as well, then we're OK.

I agree that something is not right WRT the Freezer duct. In fact, I brought the subject up in the first place when actual measurements did not meet my expectations. Actual measurements have a way of doing that!

Some more actual measurements may clarify things, or they may provide further mystification. We'll see. Let me politely point out that I'm the only one performing actual measurements in addition to doing a lot of handwaving!

jaganath · Post by **jaganath** » Thu Mar 08, 2007 3:11 am

there is no slow speed, quiet 140mm fan available at this time.

silencio sells the D14SL-12 (link) cost about $26 inc shipping.

Felger Carbon · Post by **Felger Carbon** » Thu Mar 08, 2007 4:00 am

I think I've more better figured out what's going on.

Consider a room of still air. Introduce a fan in the middle of the room, and turn the fan on. It sucks air from one direction of the room, where the air is still. This requires a pressure drop at the intake of the fan. It also pushes a column of air into the still air in the other direction, requiring a pressure rise at the output of the fan. The pressure provided by the fan is the sum of the input pressure drop and the output pressure increase. A pressure difference is what moves air.

If a cooling fin stack has resistance, then there is a pressure drop across it due to the moving air (no moving air, no pressure drop needed).

What I can do now is prove that a fin stack has resistance, rather than wave my hands and simply assert that it does.
-------------------

My 2-fan duct has the same cross-sectional area from start to finish, right? Wrong! Right smack in the middle is this stack of cooling fins. With 2mm fin-to-fin spacing and .4mm fin thickness, the cross-sectional air space is 80% of the rest of the duct. And that means, since the number of air molecules per second passing any point in the duct is constant, the air through the cooling fins has to move 1.25 times faster than the air outside the cooling fins! To accelerate the air at the input to the fin stack requires a pressure differential, and there is yet another pressure differential at the back of the stack. The "resistance" of the cooling fins is proportional to the sum of these two pressures divided by the air flow.

If the fin stack is spaced 1.2mm fin-to-fin, with .4mm fins, then there's only 67% the air space in the fins, and the air has to move 1.5 times faster while passing through the fin stack. This proves that closer spacing of the fins produces more resistance to air movement in the fin stack.

This accounts for one of the reasons for fin stack resistance. I sincerely hope (and you should too) that the air flow through the fin stack is turbulent instead of laminar. That produces better cooling at the cost of higher air flow resistance. How much? I don't know, I don't have the math for turbulent air flow.

But I have proof that the 2mm spacing of the cooling fins on the Freezer must provide resistance to the air flow in the duct. And I've shown a mechanism whereby tighter fin spacing has more resistance and wider fin spacing (Ninja, anyone?) has less. A big reason is that the fins all seem to be about .4mm thick, regardless of the spacing.

jaganath · Post by **jaganath** » Thu Mar 08, 2007 4:50 am

some useful links:

http://www.sunon.com.tw/english/wealth/tech/tech-04.htm
http://powerelectronics.com/mag/Ufnal%2 ... %20PET.pdf

I sincerely hope (and you should too) that the air flow through the fin stack is turbulent instead of laminar.

http://www.rotronmilaero.com/systemImpedance.cfm

n will assume a value from 1 to 2 depending upon whether the flow in the system is completely laminar (n=1) or completely turbulent (n=2). For most electronic equipment, n will be found to be nearly equal to 2, and this value may be assumed for calculation purposes in absence of other data.

I don't think anyone can doubt that the heatsink causes resistance to flow, I'm just puzzled why the increased static pressure of the fans in series didn't provide cooling benefits; an initial conclusion might be that the impedance isn't enough to extract any significant increase in flow from a series arrangement.

kaange · Post by **kaange** » Thu Mar 08, 2007 5:37 am

Felger Carbon wrote: My 2-fan duct has the same cross-sectional area from start to finish, right? Wrong! Right smack in the middle is this stack of cooling fins. With 2mm fin-to-fin spacing and .4mm fin thickness, the cross-sectional air space is 80% of the rest of the duct. And that means, since the number of air molecules per second passing any point in the duct is constant, the air through the cooling fins has to move 1.25 times faster than the air outside the cooling fins! To accelerate the air at the input to the fin stack requires a pressure differential, and there is yet another pressure differential at the back of the stack. The "resistance" of the cooling fins is proportional to the sum of these two pressures divided by the air flow.

If the fin stack is spaced 1.2mm fin-to-fin, with .4mm fins, then there's only 67% the air space in the fins, and the air has to move 1.5 times faster while passing through the fin stack. This proves that closer spacing of the fins produces more resistance to air movement in the fin stack.

Actually those are only the average increases in airspeed as they only allow for the reduction in cross-sectional area. The amount of surface area presented by the fins also contributes to the air resistance in the form of skin drag, which reduces the effective cross-sectional area further through the boundary layer. The more fins there are, the more boundary layers are created (one on each side) so the more constricted the air path is. Also, the longer the fins are, the thicker the boundary layers become, again increasing constriction.

These boundary layers will begin as laminar which mean that less of the fin is able to pass heat to the airflow but at some point, the boundary layers will interfere with the airflow enough for separation to occur so that turbulent flow begins and the fin area behind this point will again be able to pass heat to the airflow.

This is why deep finned heatsinks with tight fin spacings work well with faster airflow. Separation begins earlier along the depth of the fins so they become more efficient. With low airspeed, the laminar boundary layers can basically meet between the fins, constricting the airflow enough to prevent separation from occuring so most of the fin depth is wasted.

A more open fin structure allows for the laminar boundary layers to become thick enough for separation at low airspeed so more fin depth is effective but with higher airspeeds, there is less surface presented to the airflow so efficiency does not ramp up as much as for a more tightly packed heatsink.

Finding the most efficient ratio of fin spacing and depth would be pretty much impossible to model and a large amount of experimentation would be used in developing CPU heat sinks.

jaganath wrote:
n will assume a value from 1 to 2 depending upon whether the flow in the system is completely laminar (n=1) or completely turbulent (n=2). For most electronic equipment, n will be found to be nearly equal to 2, and this value may be assumed for calculation purposes in absence of other data.

I think that statement refers to forced airflow in 'normal' electronic equipment where noise is only an afterthought and the airpath is restricted and tortuous (rackmount servers anyone?). For a quiet system, I would think most of the airflow would be laminar as turbulence also generates noise. This is why so many people try to duct air to the PSU to reduce its fan speed as the airpath though a PSU is one of the most (if not the most) constricted in a PC so the same flow rate through the PSU would generate far more noise than through a case exhaust fan with a similar outlet size.

Felger Carbon · Post by **Felger Carbon** » Thu Mar 08, 2007 7:14 am

kaange wrote:These boundary layers will begin as laminar...

I built model airplanes (my own designs, of course) in my misspent youth. If I had built their wings out of rectangular blocks, rather than with rounded and tapered leading edges, I would have had spectacularly non-laminar airflow over the wings!

The reason the cooling fins are not tapered at the entrance and exit to the stack is that there would be blood all over the insides of PC cases from people like me cutting themselves on the sharp cooling fins. So they're left with rectangular edges. And the airflow is non-laminar immediately upon hitting that square edge. IMHO, YMMV.

jaganath · Post by **jaganath** » Thu Mar 08, 2007 11:09 am

Thank you for that very informative post kaange, especially the discussion of how the boundary layer evolves. I take it your dayjob involves aerodynamics in some capacity?

It seems we at SPCR are in somewhat of a quandary; we want laminar flow for least noise, but you need turbulent flow for best heat transport!

Anyway, this discussion has made me appreciate the job that heatsink designers do a lot more; the optimal size and spacing of the heatsink fins varies wildly according to the available flow and static pressure, in this domain one size really doesn't fit all.

We're still left with the puzzle of why the series fans didn't seem to improve cooling?

Felger Carbon · Post by **Felger Carbon** » Thu Mar 08, 2007 12:52 pm

jaganath wrote:We're still left with the puzzle of why the series fans didn't seem to improve cooling?

Now that I've taken care of a few housekeeping details, I can concentrate again on the Freezer ducting. I'll test a one-fan duplicate duct, and a one-fan simply SM'd to the fin stack. And I'll compensate for RPM of the two fans vs the one fans for equal noise. In other words, with the appropriate variables controlled as closely as I can. Perhaps some decent data will help us understand what's happening.

J. Sparrow · Post by **J. Sparrow** » Thu Mar 08, 2007 1:27 pm

jaganath wrote:We're still left with the puzzle of why the series fans didn't seem to improve cooling?

It look very much like so; in my highly unscientific tests, I also tried to duct a bigger fan (120 mm) to the Freezer (92 mm - almost the 2:1 area ratio which Felger was suggesting) and it didn't produce any spectacular results.

kaange · Post by **kaange** » Thu Mar 08, 2007 6:52 pm

Felger Carbon wrote:And the airflow is non-laminar immediately upon hitting that square edge. IMHO

With slow enough airspeed, the edge vortex will be so small that the turbulent eddy energy would dissipate almost immediately so that laminar flow reassumes and the boundary layer reattaches to the parallel surface.

At the airspeed required for flight, this reattachment wouldn't happen. No one here would want airspeed that high through their heatsinks.

Plus the thickness of the section contributes to the edge vortex energy as the amount of air that is redirected to tangential flow (relative to the main airstream) at the aerodynamic mid point increases with section width so the tangential airflow energy must also increase. I doubt your model airplanes had 0.4mm thick wings (mine certainly did not).

kaange · Post by **kaange** » Thu Mar 08, 2007 6:54 pm

jaganath wrote: I take it your dayjob involves aerodynamics in some capacity?

No, it has to do with involvement in motor and water sports.

Kremmit · Post by **Kremmit** » Thu Mar 08, 2007 8:50 pm

Felger Carbon wrote: a danger we're talking past each other here. The 140mm fan on the Dominator is because the cooling grid on the Dominator is 140mm square.

Aha, gotcha. That's the think about online discussion; sometimes something gets lost in translation between writer and reader.

Now, in the second paragraph above, I had moved on and was talking about fan1 of a 2-stage fan duct that had nothing to do with the Dominator. If that was what you meant as well, then we're OK.

Yup, that's wat I was on about.

Actual measurements have a way of doing that!

Ain't that the truth.

Let me politely point out that I'm the only one performing actual measurements in addition to doing a lot of handwaving!

Understood, and if I came off as a jackass, my apologies. I'm just trying to help get to the bottom of this mystery. Sometimes an outside point of view can help, but sometimes it's just annoying.

At any rate, I anxiously await the results of the next round of testing.

Felger Carbon · Post by **Felger Carbon** » Fri Mar 09, 2007 9:21 am

Here's the Freezer being cooled by one 100mm Scythe fan, 1000RPM nominal. I'm at 4100' elevation here. At 12V, the fan runs 1095RPM, and 665RPM at 7V. Temp rise over room ambient, with 3.6V on the 220mm fan and 12V is 15.4C; at 7V it's 19.2C. The temp rose 24.7% from 12V to 7V with this very low air flow.

SPCR tests with somewhat higher airflow. From 12V to 7V, here's how some larger HSFs fared:

Ninja 14C to 17C, a 21.4% increase
Ultra-120 15C to 21C, a 40% increase
Andy Sam 16C to 24C, a 50% increase
- only the Ninja beat the Freezer's 24.7%!

By using % temperature increase instead of the temperature delta, differences in power levels etc cancel out, and we can see the effect of airflow and fin spacing. Here's the setup:

This is stage one of three. Stage 2 is a ducted single fan, and stage 3 a retest of the 2-fan duct.

Note there's no effort (yet) to improve cooling by some larger ducting to direct the flow from the 220mm fan, not letting it freely bypass the Freezer entirely.

Yes, I know SPCR's reported numbers don't have much precision, but they're all I have to work with. You can interpret these results any way you wish. The Freezer temps above were measured for 3+ hrs at each voltage, and several readings were taken and averaged.