Fans: comparisons vs absolute measurements

Felger Carbon · Post by **Felger Carbon** » Wed Oct 31, 2007 10:36 am

A good figure of merit for fans is noise vs airflow. Sure, price, MTBF, quality control etc are all important, but noise vs airflow seems the bottom line. Noise meters - SLMs - are readily available for purchase (if not cheap). CFM meters are not available at any price.

Those of us who have been around the block a couple of times have favorite fans. When a new fan appears on the market with interesting features, it would be good if a fan-to-fan comparison could be made of noise vs airflow.

Consider a child's toy: a pinwheel. Blow air softly, it spins slowly. More air movement, it spins faster. With a stroboscope, its RPM can be measured precisely. A "jig" can be made to mount various 120mm fans in precise locations WRT a pinwheel.

Mount a new fan - a Scythe 800RPM SlipStream 9-blade, for instance. Run it at 12V. Measure the noise at 3 inches from the hub. Measure the pinwheel's RPM.

Now, mount a GW NCB fan. Adjust the voltage until the pinwheel is turning at the same RPM as before. Now you have equal CFMs. Measure the noise at 3 inches. Voila: you know which fan wins the noise/airflow contest, and by how much.

A more difficult task is devising a similar jig to compare fans when blowing air thru a heatsink's cooling grid. The jig would have to include the heatsink in question (but not hooked up to a computer). Still, a comparison can be made.

I happen to have an SLM and a stroboscope. What should serve as the "pinwheel"? A model-airplane propellor? Any suggestions?

Felger Carbon · Post by **Felger Carbon** » Wed Oct 31, 2007 1:39 pm

Gee, stroboscopes are expensive, aren't they? This one costs $17 , and that doesn't include the pen flashlight you'd need for a lightsource.

Pinwheels are hard to find, too. Here's one for $4. You want a wider selection? Here are 4, 5, 6, and 7 inch plastic propellors and mounting wires and propellor bearings. Tower carries similar stuff from other companies, but I didn't want to confuse you with too many choices.

The $17 tachometer is designed for 2 or 3-blade propellors. Leave it set to 2, and you can use it to measure fan RPM with the penlight (other room lights off). For 7-blade fans, just divide by 3.5 to get the true RPM. So you get double service: as a CFM comparator, and as a fan RPM monitor. Works best with the fan outside a computer case since the penlight has to be on one side of the fan and the tachometer on the other side.

You'll need a Rheostat to provide voltages for the fans, a 12-13.8V source for the Rheostat, a digital multimeter (Xoxide, casemod carry them for ~$8) to record the voltages associated with the "CFMs" or fan RPMs. You can blow over $50 for all this stuff easily. Making actual fan measurements ain't cheap.

The SLM? Use two channels on the Rheostat to run the two fans simultaneously. Once you've determined the fan voltages that result in the same "CFM" (propellor RPM) hold both fans up to your head and listen. It will be obvious if one fan is louder or more unpleasant-sounding than the other.

mcoleg · Post by **mcoleg** » Wed Oct 31, 2007 1:39 pm

good thinking.

i am thinking, you could even use another fan instead of a pinwheel. as long as you use the same part for each measurement, it should be fine.

Felger Carbon · Post by **Felger Carbon** » Wed Oct 31, 2007 2:18 pm

mcoleg wrote:i am thinking, you could even use another fan instead of a pinwheel. as long as you use the same part for each measurement, it should be fine.

Nope. The fan's magnets don't allow the fan, when not powered up, to spin freely. Try it with any fan.

Bluefront · Post by **Bluefront** » Wed Oct 31, 2007 3:59 pm

FC.....you could get one of those fans with a removable blade/shaft assembly. Remove the blades, and rip the windings/guts out of the motor. You'd have a fairly precision propeller.

Felger Carbon · Post by **Felger Carbon** » Thu Nov 01, 2007 1:27 pm

I have no faith in the ability of my face to provide matching airflows. Us Carbons just don't come with 5-decimal-digit CFM meters built in. But I just did the best I could:

The GW NCB I had been using for test was previously well-used in PSU cooling duty. So I substituted a little-used NCB. I connected an 800RPM Slipstream to one channel of a Rheobus and set it to full voltage. Then I connected the new(er) NCB to another channel and did my best to match airflows. As before, the NCB sounded louder... a lot.

This time I went one step further: I measured the noise level on each fan at a distance of 3.5 inches. I got 45dBA even on the NCB, and 36.7dBA on the SlipStream. Since the Slipstream is closer to the ambient than the NCB, the difference is even greater than the mathematical 8.3dBA.

This experiment is really about my face: when my Tower shipment comes in a week from now, I'll repeat the experiment using a model airplane propellor's RPM as the device to equalize airflow between the two fans, undoubtedly far more accurately than with my face.

I also ordered the $17 tachomenter, so I'll be able to compare results with it (and a penlight) vs my Extech stroboscope.

Felger Carbon · Post by **Felger Carbon** » Thu Nov 01, 2007 1:42 pm

More measurements: Same conditions as above.

NCB RPM 1243, Slipstream 864. (That explains the large noise difference!)

I decided to measure the noise 1/4" from the fans' hub center:

NCB 54.2dBA, SlipStream 45dBA even - a 9.8dBA difference. This is about 15dBA above the ambient, so those are good numbers. The question is, were these numbers taken at the same airflow?

Trekmeister · Post by **Trekmeister** » Thu Nov 01, 2007 4:09 pm

Interesting method for measuring airflow. The rpm of the pinwheel might not be a linear match to the airflow, but since you are looking for an equal airflow that won't matter.

The only possible problem I can think of at the moment is that the type of flow might affect the rpm of the pinwheel. I'm thinking that maybe the airflow from a fan with steep angled blades such as the noctura could create more of a vortex in the chamber than a fan with more common blade geometry would. (I assume that this whole thing would be placed in some kind of tube) This vortex could then affect the pinwheel to show a higher airflow than it really is if the pinwheel spins the same direction as the fan, opposite otherwise...

Now that is only a thought and a whole lot of maybe:s, no idea if this would be something happening in real life or not. Just trying to help you find possible problems!

lm · Post by lm » Thu Nov 01, 2007 5:46 pm

Otherwise good, but equal rpm different fan = different airflow, because the fan blade shape differs.

Felger Carbon · Post by **Felger Carbon** » Fri Nov 02, 2007 2:48 am

lm wrote:...equal rpm different fan = different airflow, because the fan blade shape differs.

Yes yes yes! The different airflow is exactly what I propose to measure. Because of the differences in blade shapes between the NCB and SlipStream fans, noise should be measured at equal airflow and not at equal RPM. In fact, I attempted to do that above by using my facial CFM meter to "measure" equal airflows at 864RPM for the SlipStream and 1243RPM for the NCB.

But something more accurate than my face is needed to measure airflow. I propose using a model-airplane propellor.

CyberDog · Post by **CyberDog** » Sat Nov 03, 2007 9:00 am

I was thinking that too that fans put more or less turbulence. So your propeller might work differently on different fans. So I suggest that you either build duct with air flow straightener or a box where a fan blows or sucks air and a propeller on other opening.

N7SC · Post by **N7SC** » Sat Nov 03, 2007 12:58 pm

FWIW, don't big wind tunnels use all manner of vanes and airflow management devices to straighten out and remove the turbulence from the air flow their turbines/fans produce?

If there is concern that the differing fans being tested are producing different types of swirl and turbulence that could be effecting the measurement of CFM via FelgerCarbon's rather interesting proposal, then the airflow can be straightened out and "de-turbulenced" by a similar methos. Cheaply and easily too.

jaganath · Post by **jaganath** » Sat Nov 03, 2007 1:55 pm

http://sln.fi.edu/flights/first/makesimple/index.html

Felger Carbon · Post by **Felger Carbon** » Sat Nov 03, 2007 2:45 pm

A wind tunnel directs air past an airfoil (etc). With a wing's airfoil, the direction of the airflow is critical.

What I need to do is get a propellor RPM that represents a particular airflow. The direction of the airflow is not critical, since if off on half the propellor, it will go "the other way" on the other half. Further, it is not at all clear to me that a change in the direction of airflow can be made without reducing the airflow, since the change in direction necessary consumes power. Subtracting power from an airflow changes its velocity.

I need a propellor RPM that's a particular value for a particular magnitude of airflow, not direction.

A wind tunnel doesn't care about the power drop since they just crank the power higher to get the test velocity they want. Apples and oranges!

jaganath · Post by **jaganath** » Sat Nov 03, 2007 4:56 pm

it is not at all clear to me that a change in the direction of airflow can be made without reducing the airflow, since the change in direction necessary consumes power. Subtracting power from an airflow changes its velocity.

yes, but this effect should be the same for all fans. thus you get a relative measurement including a little airflow resistance; this is a more realistic measurement, as a case is not a "in free air" environment anyway.

Felger Carbon · Post by **Felger Carbon** » Sat Nov 03, 2007 5:43 pm

jaganath wrote:yes, but this effect should be the same for all fans.

If ther effect is the same for all fans, then the deviation from perfect direction is the same for all fans, and will result in the same offset to the propellor RPM for both fans... and hence changing the airflow is pointless.

What case? I'm not including a case in this setup, since a valid criticism would be that I'm testing in Case X, model 5b, and so the result would be invalid for Case Y, model 37q.

I am absolutely opposed to the thought that to measure airflow you must change the airflow before measuring it. Therefore, I will test the unchanged airflow. Anyone who disagrees is welcome to change their airflow any way they like when they measure their airflow. I am not trying to impose my opinions on anyone. Honest!

Felger Carbon · Post by **Felger Carbon** » Mon Nov 05, 2007 7:17 pm

The initial fixture is completed and pics plus test results of the new Scythe SlipStreams and three other fans are posted in the User Reviews forum. Not a bad 3 hrs work! Pant pant pant...

Felger Carbon · Post by **Felger Carbon** » Thu Nov 08, 2007 8:09 pm

Question: if the CFM doubles/halves, does the sensor propellor respond accordingly?

Answer: I dunno. The question is irrelevant if we're only comparing fans at identical CFMs. I suspect that the sensor, being lightweight and pretty much friction-free, is pretty linear with CFM.

It's well known that although fan CFM varies with RPM, it does not do so linearly. Cutting RPMs in half results in cutting CFM more than that.

Since I had a perfectly good CFM sensor sitting there, I decided to take some numbers. This is the raw data:

GW NCB fan:
RPM 1450, sensor 858, ratio 1.690
RPM 1001, sensor 577, ratio 1.735
RPM 740, sensor 400(honest), ratio 1.850

SFlexE fan:
1268RPM, sensor 860, ratio 1.474
925RPM, sensor 603, ratio 1.534
603RPM, sensor 367, ratio 1.643

As I said, this is the raw data. The different ratios of the two fans are the natural result of differently pitched blades. Is there some information buried in the change in ratio as the fan RPM decreased? I'm just letting you share info that I've gathered.

jaganath · Post by **jaganath** » Fri Nov 09, 2007 12:46 am

It's well known that although fan CFM varies with RPM, it does not do so linearly. Cutting RPMs in half results in cutting CFM more than that.

are you sure? the fan law regarding CFM seems pretty linear:

http://www.gorhamschaffler.com/fan_laws.htm

Felger Carbon · Post by **Felger Carbon** » Fri Nov 09, 2007 12:58 am

Jaganath, I'm testing fans that aren't mounted in infinite walls, where air can't get from the front to the back of the fan. In my test setup, that's an easily available path that is taken!

In the "infinite wall" setup, the leakage path is between the end of the fan blades and the beginning of the fan case. This distance is different for different fans. Some time back in the User Review forum I noted that 4 fans had distances in the ratio of 1, 2, 3, 4 in that blade-case distance. The S-Flex is a "tight" fan, the GW NCB a "loose" fan as noted in my long-ago review. Look for something like "A physical examination of 4 fans". Something like that.

With no "infinite wall" the leakage is pretty obvious. Isn't it?

Bikeman · Post by **Bikeman** » Sun Nov 11, 2007 1:36 am

Mount your fan on a 30x30cmÂ² plate and you have a very decent approximation of an "infinite plate". Or add a 12cm diameter sewer duct to the front or the back of the fan.

I think the "infinite plate" is needed to eliminate direct airflow shortcircuiting around the edges of the fan. Any of the above proposed solutions eliminates this.

Nice work you're doing here, though.

Felger Carbon · Post by **Felger Carbon** » Tue Nov 27, 2007 6:39 pm

Here's a sampling of plastic model airplane propellers:

Left to right: 6" dia, 5", 4", 3.25", 2.9", and 2.5". The 5" prop is the one I use in my 120mm-fan test fixture. The 2.5" prop is the right size for a 60mm-fan fixture, but the center hole is smaller than the 1mm of the other 5 fans, so I'd need a different support wire than what I have.

I performed a quick-and-dirty test on the black 2.9" fan, because it's a pretty low-pitch fan. The Nexus Frizzbee, 60mm and 1600RPM, would not turn it. But the Logisys HC-102 (1600RPM at ~5V) does turn it at 6.5V adequately, and vigorously at 8V. If that's a problem, I can try a cut-down 4" prop since is has a much higher pitch (and so would be easier to turn in a breeze).

Some combination of these propellers will allow me to replicate the extremely simple 120mm-fan test fixture for 140mm, 93mm, 80mm, and 60mm fans. I do like testing fan noise under equal-airflow (CFM) free-air conditions.

Felger Carbon · Post by **Felger Carbon** » Thu Nov 29, 2007 3:31 am

Here's the 80mm version of the equal-airflow test fixture. You have to look carefully to see the white styrene base-plate against the larger block of white rubber being used for vibration isolation:

This is much smaller than the 120mm fixture - obviously - but the principle is the same, and the proportions are about the same. The little dot on the sensor fan blade is a tiny piece of modelling clay I applied to balance the sensor prop. It does double duty to help using the Extech strobe correctly to read RPM.

I tested 3 80mm fans: a 2+ year-old SilenX, a brand-new Sunbeam Silent Core, and an 18-month-old GW NCB. Here's the result (fan, fan RPM, sensor RPM, fan voltage, lowest steady dBA reading, mike distance = 1/4", room ambient 27.9dBA when no traffic):

Code: Select all

SilenX_ 1566RPM 1140RPM 12.0V 47.3dBA (1/4")
GW NCB_ 1545RPM 1144RPM 9.04V 50.4dBA
Sunbeam 1249RPM 1142RPM 9.55V 47.3dBA

The SilenX has the lowest pitch by a small margin, the Sunbeam the highest. The NCB 80mm is not known for being an exceptionally quiet fan, and these results confirm that. Though 2+ years old, the SilenX has never been used - I bought it so I'd have a quiet 80mm fan if I ever needed one.

I "designed" (laid out) the fixture using my Vector Engineering $20 CAD package. It went together easily with the toughest task being waiting for superglue to dry at each step before proceeding. The sensor prop is an unmodified 3.25" unit.

The reason the sensor prop RPM varies is that I was also using the strobe (with its main control and vernier) to measure the fan RPM, and it's not possible to exactly restore the sensor RPM to the original value. The fan RPM is only useful to rank the fans according to blade pitch; the point is, what's the noise level at equal airflow?

andyb · Post by **andyb** » Thu Nov 29, 2007 6:34 pm

I performed a quick-and-dirty test on the black 2.9" fan, because it's a pretty low-pitch fan. The Nexus Frizzbee, 60mm and 1600RPM, would not turn it. But the Logisys HC-102 (1600RPM at ~5V) does turn it at 6.5V adequately, and vigorously at 8V. If that's a problem, I can try a cut-down 4" prop since is has a much higher pitch (and so would be easier to turn in a breeze).

As a side topic, have you considered a bit of reverse thought. Using a fan @ X CFM to find out how easy it is to force a "standing" fan to start to spin. As you have noted some fans have problems making the propeller spin due to the blades pitch, likewise this could be tested on the fans themselves. This is a largely pointless exercise as the outcome has no worth, only general interest, or at least my interest as to how much air would be needed to force-start a fan spinning, the only real worth would be to test the proppelers themselves and with my next idea.

Another idea could be to try more than one propeller design per fan test (not at the same time of course), you could identify aproximate fan pressure by doing this as of course by using one fan that is difficult to turn, and one that is easy to turn. This knowlege would be as useful to SPCRers as the CFM itself.

Andy

Felger Carbon · Post by **Felger Carbon** » Thu Nov 29, 2007 6:58 pm

andyb wrote:This is a largely pointless exercise as the outcome has no worth, only general interest, or at least my interest as to how much air would be needed to force-start a fan spinning, the only real worth would be to test the proppelers themselves and with my next idea.

Andy, the propellors and styrene are dirt cheap at any major on-line hobby store (such as Towerhobbies). It takes a half-day to make a fixture, most of which is spent waiting for superglue to dry (i.e. you can catch up on yer browsing during that half-day). And then you can perform any test whatever that you want!

I'm glad the work I'm doing on these things is of interest to somebody! I thought testing for noise under equal-airflow conditions was a really good way to compare fans, honest.

But I'm not willing to spend 5 minutes of my time performing somebody else's experiments. If the data interests you, go get it! I will read your resulting data with interest.

andyb · Post by **andyb** » Fri Nov 30, 2007 1:01 pm

But I'm not willing to spend 5 minutes of my time performing somebody else's experiments. If the data interests you, go get it! I will read your resulting data with interest.

Thats fair enough, but dont get your hopes up.

Andy

PS: I do find your results interesting.

Felger Carbon · Post by **Felger Carbon** » Sat Dec 08, 2007 9:25 am

I have the 92mm test fixture up and running. I now have 120, 92, 80, and 60mm fixtures. I could easily make a 140mm fixture, but testing Yate Loons against Yate Loons doesn't seem useful.

The first two fans on this list are 2.5 years old, and got a lot of use the 12-15 months. The GW is not only new but a newly released product. The Arctic Cooling 92x25mm fan is 2000RPM nominal; I just bought it. Data is fan ID, fan voltage, fan noise level at 1/4" from center of intake hub, the sensor RPM, fan size, and the range used for measurement on the manually ranged SLM. Ambient 29.8dBA.

Code: Select all

SilenX"11dBA"  12.0V 48.2dBA 601RPM 92mm 30dBA range
SilenX"14dBA"  9.69V 49.4dBA 601RPM 92mm 30dBA range
GW NCB         8.63V 48.4dBA 601RPM 92mm 30dBA range
A-C AF9225     6.60V 47.6dBA 601RPM 92mm 30dBA range

What surprises me is the very tight grouping. I haven't added a subjective listening test of what little noise these fans have; I wanted this to be a completly objective test of fan noise under equal-airflow conditions.

Fans: comparisons vs absolute measurements

Fans: comparisons vs absolute measurements

Sensor propellor linearity with varying CFM