Fans demolished; lessons learned

Felger Carbon · Post by **Felger Carbon** » Sat Jul 21, 2007 7:11 am

In the past 2-3 months, I've torn apart some fans. By taking a careful look at what I tore apart, I've learned some things I otherwise wouldn't know. Useful things? Hmm...

I decided the MaxOrb would be a wonderful low-height low-air-flow HSF if only the lousy fan was replaced. Then I discovered the fan could be replaced with, at worst, medium difficulty. This led me to cut the outer casing off some 120mm fans. Two of the fans I cut the casing off of were the YL D12SL-12 and the GW NCB. Now, these two fans have a lot of friends here at SPCR, and I'm one of them where the NCB is concerned.

I almost wish I hadn't learned what I did about those two fans, because what I learned is politically incorrect. The affection and outright love of the YL here at SPCR, especially in its high-priced Nexus guise, is almost astonishing. At equal RPMs at or near 1000RPM, they are reputed to be the quietest thing around.

I discovered the YL has cause to be the quietest thing around, for the simple reason it pushes less air at a given RPM than anything around. The fact that accurate CFM measurements are very difficult to make; impossible for the average experimenter, has helped the YL keep this secret.

But when I cut the casing off the YL and the NCB and placed the remainder - the naked blades especially - side by side, a major difference was immediately obvious to me: the YL blades not only had a smaller area than the NCB, but the blades obviously had a much lower pitch.

This latter point - the low pitch - explains why the YL is quiet. Simple: fan noise is highly proportional to CFM, because the moving air column's energy is proportional to the cube of its velocity - and the end result of that energy is noise. Less CFM, a lot less noise. And since folks at SPCR (including me) compare fan noise at equal RPMs, the YL is always going to be the quietest because it's always going to be the one pushing the least air!

I think I mentioned this once, in passing, in the regular forums. I didn't make a big deal about it because micturating against the wind is never a good idea. SPCR folk are almost religious in their devotion to their quiet Nexus 120s. I don't want an angry mob of righteous quiet-lovers coming after me with tar, feathers, and a rail.

Sometimes ignorance is bliss.
----------------------------------

Yesterday I tore apart 4 fans. This time not just the casing; I dremeled down to the bearing and motor drive PC board. I was stealing the cables off a couple of $5 A-C 80mm PWM fans (which I bought for the cables, not the fans), so the fan itself was eligible for teardown. What I was looking for was the complexity of the motor control integrated circuit.

I discovered the A-C PWM IC had 16 pins, two (power and ground) being exceptionally wide. Dunno what the name of this package is. It's a complicated IC with a lot of smarts (for a fan) inside.

The advent of a 3-wire 120mm fan with reliable starting at low voltages (which the A-C PWM fan is when the PWM pin is ignored) made my Scythe S-Flex D obsolete. So, next I tore that down and discovered the IC had ten leads, all skinny.

I had remembered regular 3-pin fans as having motor ICs with only 4 leads, so I also tore down a GW NCB (ex-PSU; I don't use PSU fans anymore) and a 92mm Evergreen low-rpm fan (obsolete for the same reason as the S-Flex D). Both ICs had 4 leads, but the NCB also had a 3-lead transistor to drive the tach signal. The Evergreen (Evercool EGF-92) did not have a separate transistor; the tach seemed to be driven by one of the four motor IC leads... Huh?

Gee, if the IC has two leads for power and ground, and two leads to drive the 4 motor coils (two coils per lead), then how can one of the leads be used to drive the tach signal? The NCB was conventional; the tach signal is a buffered copy of the drive signal to one of the coils (in case you ever wondered why two tach pulses per RPM) with a transistor needed to provide the open-collector tach drive.

The EGF buffer transistor is in the I.C., so how is the IC powered? Well, the IC drives both sets of coils, but only one set at a time. The voltage on the other line is always high. Since the IC uses little power itself, two (internal) diodes from the two drive lines will always assure that the IC always has power.

That's a lot of work to save one transistor, but they make a lot of fans and if you can save a couple pennies a fan, you're golden. Transistors cost nothing these days; packaging, testing, placement and soldering are all expensive. Putting that transistor in the IC at the "cost" of two free diodes is a big win.

I discovered a couple other things. For instance, all four fans have a big diode in series with the +12V lead. That way, if any klutz accidentally connects the wrong polarity voltage to the fan (perhaps by reversing ground and +12), nothing happens except the fan won't turn. That means all the fans run off +11.3V nominally because of the ~.7V diode drop. In three of the fans, the diode is a surface-mount device but an actual 1N400X is used in the EGF fan - that's an ancient-history thru-hole component.

There were a few other mundane but mildly interesting (to a retired design engineer) aspects of the motor drive circuit board assemblies. My biggest surprise had nothing to do with the board. Hello, bearings!

You see, by breaking things down to the motor drive circuit board and the central driveshaft, all I had to do was pop the seal off the bearing shaft and see all the lubricating oil. Ho ho ho!

The A-C and Sflex had a double-thick seal (as I recall, certainly the Sflex) to hold "all that reserve oil" in the shaft. Fact: The A-C 80mm PWM shaft was dry. Only the rotating parts had any oil at all on them. There was absolutely no excess. The Sflex was almost as dry; there was very little - very little - excess oil.

Only the EGF fan seemed to have anything like reserve oil, and at that not enough so that when you tipped the open shaft upside down, a drop would fall. I tried it; no oil fell. But the EGF fan had lots more lubricant than the A-C PWM or the expensive Sflex fan. The NCB was, in this group of 4, middling in its lubrication.

Hey, I know Saudi crude is headed for $80 a barrel but...

sanse · Post by **sanse** » Sat Jul 21, 2007 8:28 am

thanks for these realistic down-to-earth remarks.

Bluefront · Post by **Bluefront** » Sat Jul 21, 2007 10:27 am

Yeah...interesting stuff. It's too bad most computer fans cannot be taken apart without destroying the thing. If there was an easy way, we could re-pack the bearings, maybe quiet down bearing noise. I've tried to remove the shafts on several fans....no luck. I've tried to drill a small hole in the clip that secures the hub. Couldn't do it.....thought a hole would relieve tension on the clip, allowing the shaft/blades to come out.

I did cut apart a few fans....I was interested in how the blades were balanced....I found some substance that looked like a crude balancing attempt, located in one spot near the base of a blade. I cannot conceive of a method that could be used to balance blades.....other than a crude balancing cone with a center bubble (like an old-time wheel balancer).

As far as the YL fans and airflow goes.......I don't know. If I need more airflow I try to find a medium or high speed version of the fan. Then hope the thing is quiet at a low voltage setting. Actually most low-speed/low airflow fans to me are of little use. I always try to use very few fans, but which have the ability to speed up a bunch if necessary.

I know some people use a lot of low-speed fans in their setups.....not me. The fewer fans, the better......in my book.

Felger Carbon · Post by **Felger Carbon** » Sat Jul 21, 2007 11:02 am

Bluefront wrote:I know some people use a lot of low-speed fans in their setups.....not me. The fewer fans, the better......in my book.

I you need X CFM, you can get it with one fan, or with 2 fans, each providing X/2 CFM. Method 2 is 15dBA quieter. That assumes there's room for the two fans, and that gzins and gzouts (for airflow) remain equally balanced, of course... and that's where the bun gets sticky.

Bluefront · Post by **Bluefront** » Sat Jul 21, 2007 11:55 am

There are so many factors at play when setting up a fan configuration. My cases are usually smaller than most. And I like to use interior dampening materials.....not to mention I almost always design a filter into the intake setup. So I have a limited space for intake fans. That means a fewer number of fans.......only one would be nice. But two are easier to deal with, if you like to cool your hard drives with airflow.

Besides.....at idle or low-CPU usage, one fan at low speeds is usually sufficient. It's only at a small percent of the time, that I need high airflow. For that short period, one fan that can speed up enough to provide sufficient airflow, is fine by me......even though two fans providing the same total CFM would be slightly quieter. I can easily deal with two fans.......more than that gives me a head-ache.

jaganath · Post by **jaganath** » Sat Jul 21, 2007 2:13 pm

I discovered the YL has cause to be the quietest thing around, for the simple reason it pushes less air at a given RPM than anything around.

Sorry Felger, but this is a big claim to make without bulletproof evidence; as a counter argument, witness the Madshrimps comparison of the Nexus flavour of the D12SL-12 with the Globalwin; at 7V the two fans are virtually indistinguishable, both for noise and cooling, but as the Globalwin has a higher top speed one would expect it to be rotating about 100-150rpm faster than the Nexus at 7V; yet the Nexus is able to match it. admittedly it drops off at 5V but then again the Nexus may be rotating as much as 200rpm slower than the Globalwin at that point. If you have a Nexus to hand it would be great if you could document the rpm of it and the Globalwin at 12/7/5V to enable a more transparent comparison.

http://www.madshrimps.be/?action=getart ... rticID=421

Felger Carbon · Post by **Felger Carbon** » Sat Jul 21, 2007 4:27 pm

jaganath wrote:If you have a Nexus to hand it would be great if you could document the rpm of it and the Globalwin at 12/7/5V to enable a more transparent comparison.

What I don't have at hand is my Extech $29.95 CFM meter. Just point it at a fan, press the button, and there's the CFM readout to lots of decimal places.

What I hope I have is a logical mind. I suggest you cut the casing from a YL and an NCB. Note that the NCB has greater blade area. Now gaze long and hard at the pitch of the two fans' blades, and the blade cross-sections. Explain to yourself how the NCB could fail to have a greater CFM at equal RPM. Without, mind you, invoking magic or miracles.

nick705 · Post by **nick705** » Sun Jul 22, 2007 4:53 am

Felger Carbon wrote:Now gaze long and hard at the pitch of the two fans' blades, and the blade cross-sections. Explain to yourself how the NCB could fail to have a greater CFM at equal RPM. Without, mind you, invoking magic or miracles.

Is it really always as simple as saying higher angle of attack == higher airflow at a given RPM? The finer points of fluid dynamics are beyond me, but I do know that propellors (or wings) can be very complex things...

Bluefront · Post by **Bluefront** » Sun Jul 22, 2007 5:54 am

I also agree that just looking at a blade design, cannot tell you everything. Things like submarine propeller design, are treated like a national treasure. Any yet.....the newest/greatest design, looks just like the older version.

Trying to analyze the performance/noise of computer fan designs, is difficult. You can only go so far with our meager testing equipment.

jaganath · Post by **jaganath** » Sun Jul 22, 2007 8:48 am

Is it really always as simple as saying higher angle of attack == higher airflow at a given RPM?

in a word, no. in general, steeply-pitched (high angle of attack) blades are optimised for (relatively) low RPMs, while shallow blades are optimal at high RPMs. The reason for this is that if the angle of attack is too high the blades will stall (just like a wing) and you will get little or no airflow. you would need a very good CFD program to tell which fan blade shape is best at any given rpm. (although you can tell just by looking at the Noctua, for example, that it will have poor pressure capability, because there is a large tip clearance between the blades and the housing and the fandisc is very transparent, ie if you look at it head-on there is a lot of daylight; compare with the fan on a jet engine, where there is virtually no daylight).

Felger Carbon · Post by **Felger Carbon** » Sun Jul 22, 2007 9:13 am

What I stated was: I looked at the blades, sans case, side-by-side. From this I felt I could reach a definite conclusion.

A number of you feel that you can reach a contrary conclusion without bothering to look at the evidence. So be it. You are entitled to your methods and conclusions.

nick705 · Post by **nick705** » Sun Jul 22, 2007 11:18 am

No, don't misunderstand, if the Nexus/YLs really are moving less air then fair enough (personally I couldn't give a rat's arse about hardware cults, whether it's fans, Ninjas, P180s or anything else).

You point out yourself though the difficulty of getting accurate CFM measurements, so we're left with looking at things like blade pitch and surface area, and drawing conclusions from that alone. I'm just wondering if it necessarily works out like that in practice - if for instance you increase the pitch of the blades on a ship's propellor in isolation without considering the overall design (eg the shape of the blade cross-section, where a small variation can have a big impact), you're likely to end up creating more turbulence but less thrust.

Taking the argument to an (admittedly silly) extreme, a fan with blades pitched at 89 degrees (or one degree to the axis of rotation) should shift more air at the same RPM than one pitched at 30deg, or 45deg - that clearly won't be the case, but the behaviour at lower angles of attack might not be entirely predictable either.

Maybe it's relevant on the small, slow, relatively unsophisticated scale of a PC case fan, maybe not... I'm not impugning your investigations, I'm genuinely interested.

*prepares tar and feathers for FC....

colin2 · Post by **colin2** » Sun Jul 22, 2007 11:52 am

FC you might want to look at the link Jaganath courteously provided above:

http://www.madshrimps.be/?action=getart ... rticID=421

which *measures* cooling effectiveness. That's evidence, and evidence about the results that people actually want from a fan!

The reviews on this site include cfm figures. Perhaps those figures are wrong, but I think I'd want my own cfm readings before making that charge.

All *you* have, strictly speaking, are some observations about different fan pitches, plus a hypothesis that there's a close relationship between that and airflow and hence cooling effectiveness. This is inflated with a lot of self-praise.

Tom P · Post by **Tom P** » Mon Jul 23, 2007 7:04 am

I read posts written by someone genuinely interested in the subject and who wanted so share his observations, not "a lot of self praise".

Bluefront · Post by **Bluefront** » Mon Jul 23, 2007 7:32 am

Heh......I'm afraid some of you guys are completely missing FC's brand of humor. That's just the way he talks all the time, on these forums anyway. Get used to it. The guy has a lot of energy concerning many aspects of quiet computing......and has a lot of information to share. He does it somewhat differently from the norm......

jaganath · Post by **jaganath** » Mon Jul 23, 2007 7:54 am

yeah, colin was a bit harsh; still, I think the thing that has caused the misunderstanding is that FC presented a hypothesis as an uncontestable fact; this recently caused some ruckus in Silent Storage where someone started a thread titled something like "Warning: 2.5" drives will die if used as boot drive" or similar. This led to some rather unpleasant exchanges when it turned out this was not a proven fact, but rather an extrapolation based on 2.5" HDD susceptibility to repeated head load/unload cycling. Every fan has a fan map (PQ curve) that relates airflow and static pressure, it may well be that the higher-pitched fan hits the "sweet spot" in a system with an impedance X at rpm Y; unfortunately neither Yate Loon nor Globalwin publish these for the two fans we are talking about AFAIK, so we are kind of groping in the dark. With that in mind, it seems unwise to say the Yate Loon produces the least CFM per rpm of any fan in the world; after all, there are some pretty poor performers out there!