SPCR's Fan Testing Methodology

Want to talk about one of the articles in SPCR? Here's the forum for you.
Devonavar
SPCR Reviewer
Posts: 1850
Joined: Sun Sep 21, 2003 11:23 am
Location: Vancouver, BC, Canada

SPCR's Fan Testing Methodology

Post by Devonavar » Thu Nov 02, 2006 10:34 pm


rpsgc
Friend of SPCR
Posts: 1630
Joined: Tue Oct 05, 2004 1:59 am
Location: Portugal

Post by rpsgc » Sat Nov 04, 2006 12:40 pm

That's a lot of fans! :shock:
Good luck.

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Sat Nov 04, 2006 1:47 pm

From the CFM footnote:
All the other methods gave us considerably lower CFM numbers
What other methods were tried, out of interest?

I was a little disappointed that the fan laws didn't get a mention, after all they are very relevant for this article.

McBanjo
Posts: 671
Joined: Thu Dec 08, 2005 1:40 pm
Location: Sweden
Contact:

Post by McBanjo » Sat Nov 04, 2006 3:03 pm

I'm not sure this would be involved in this project directly but more indirectly, kind of a side-project :-)
Could you measure different fan-guards/grill to see the effect in SPL and CFM reduction?
Not needed with every fan ofcourse but 3-4 different fans would be good. Might show something about how much power the engine has

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Sat Nov 04, 2006 3:23 pm

jaganath wrote:I was a little disappointed that the fan laws didn't get a mention, after all they are very relevant for this article.
Are they? I'm not positive about that.

From your first link....
Fans of the same basic design and proportions operate theoretically in accordance with certain fan laws. In practise, these laws do not apply exactly because of design considerations and manufacturing tolerances, but they are useful in estimating approximate outputs of similar fans of different diameters and speeds as applied to normal ventilation work, and can be summarised as follows:

a Volume of air flow varies as (fan diameter)³ and as rpm
b Pressure developed varies as (fan diameter)² and as (rpm)²
c Power absorbed by the fan varies as (fan diameter)5 and as (rpm)³

It is important to note, however, that these Laws apply to the same point of operation on the fan characteristic. They cannot be used to predict other points on the fan’s curve. These laws are most often used to calculate change in flow rate, pressure, and power of a fan when the size, rotational speed or gas density is changed.
We're not predicting anything; we're measuring and listening. I guess you could say we're ignoring the "laws" for now so as not to be encumbered by them.
Could you measure different fan-guards/grill to see the effect in SPL and CFM reduction?
Topics such as this and fan laws do deserve discussion, probably in another article that might be entitled, "Making the best use of quiet fans" or "How to use fans effectively".

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Sat Nov 04, 2006 4:21 pm

Are they? I'm not positive about that.
Well, the key lesson that we can highlight from the fan laws is that CFM for any given fan is directly proportional to RPM, and varies with the cube of the diameter, whereas the relationship between noise and RPM is more like a semi-exponential function; this explains why SPCR promotes big, slow fans over small fast fans. This could be incorporated into subsequent fan-centric articles.

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Post by NeilBlanchard » Sat Nov 04, 2006 4:59 pm

Greetings,

I'm looking forward to the results from first group of fans!

Mike, are there any plans to cull out the less quiet fans? Maybe, a quick measurement at 12v and 5v, and if there is no chance that it will make the top tier, then just note it as such, and put it on a list of fans to avoid? If you have two or more samples, and one is good and the other(s) not-so-good, then these should get put through the full battery of tests, just for the data points...

ultraboy
*Lifetime Patron*
Posts: 547
Joined: Thu Aug 26, 2004 5:02 pm
Location: Bangkok Thailand

Post by ultraboy » Sat Nov 04, 2006 6:52 pm

Wow. This excellent article reflects on how much time, effort and careful thought SPCR editorial team have invested in gathering relevant data so that we, the readers, can make our informed choice. :D :D

Congratulation!!!, and Thank You SPCR.

paapaa
Posts: 198
Joined: Wed May 04, 2005 1:24 am
Location: Finland

Post by paapaa » Sun Nov 05, 2006 12:53 am

If I understood correctly you are only measuring the performance on open air with no back pressure. AFAIK the reality is different when the fan has to fight against the lower case pressure etc.

Do you have any plans to test the fans with a case or with a heatsink to reflect real world performance?

PS. Thanks for actually doing this! We finally get som insight of whether there are any significant differences between fans of similar design and RPM.

niels007
Posts: 451
Joined: Mon Nov 22, 2004 2:18 am

Post by niels007 » Sun Nov 05, 2006 5:02 am

Thats an Herculian (sp) effort!

The only thing missing and often talked about in the forums is that in actual applications, especially on heatsinks, the heatsink 'impendance' is important. Alledgedly fans with more blades provide more pressure making them work better.

Although probably too much to ask for, some way of measuring the pressure, especially at low speeds, might be interesting.

Good luck with the testing though :) (you might get a well known coffee brand to sponsor? :))

Edit: Paapaa beat me to it! argg slow sundays.. I need some coffee!

ATWindsor
Posts: 285
Joined: Mon Apr 18, 2005 11:53 pm

Post by ATWindsor » Sun Nov 05, 2006 7:46 am

It will be interesting to see your results, I once tried to measure about 8 fans in an anechoic chamber with a SPL-meter. I was unable to get a reliable reading of wind speed, however all the fans I tested had the same amount of noise at the same fan speed (within measuring errors), and followed at 18 db/per octave line almost exactly if i remeber correctly. (18 db higher sound per doulbling of fan speed).

Its also woth mention that the sound from turbulence isn't "real" sound, you are measuring the sound of turbulence "hitting" the microfone, which may or may not translate into the sound made of noise hitting some part of the computer, and/or ear.

It will be especially exiting to se if you can measure any major differences in flow, most fans are extremly similar in desing, same number of fins who look the same, and so on.

http://www.hardware.no/tester/kjoling/a ... er/22935/6

Here is the little test i did, its unfortantly in norwegian :( But the black line is the line for 18 db/per octave (with correction for background noise which was abit over 14 db at the time). One of the fans had a noisy motor, but the rest is more or less equal, except for operating range. (It must be noted that nexus was the only ones beeing anyway close to the truth when it comes to the sound-level they rate their products with).

I wish you good look, and I'm looking forward to the test :)

AtW

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Sun Nov 05, 2006 7:57 am

To address concerns about pressure...

Given fans of 25mm depth spinning at typically <1500rpm, the pressure is always very low, and differences between models is trivial. To wit -- we have interchanged Nexus, Scythe, and Globe 120x25mm fans running inaudibly on Ninja HS with no significant differences in cooling. I've done the same with 80x25mm Panaflo, Nexus and other fans on earlier HS -- again without any notable change in temperature.

No, we have not confirmed it with an empirical experiment, but we will.

People should remember why HS like the Ninja are favored here: Low impedance. It means a low airflow, low pressure fan can work very well.

The reason for the absence of pressure differences has to do with the thin depth -- 25mm. The blades cannot have enough depth for any to exhibit real differences.

I quote from http://www.acal-radiatron.com/fanselection.php:
Static pressure increases exponentially with airflow. So the actual cooling system operating point is where the plot of a prototype system measurements intersects the fan performance curve, and designers should aim to select a fan where this falls within the ‘sweet spot’ area.

This is if you want the best cooling/noise ratio. It's my contention that given a specific size 25mm depth fan, you hardly need to consider cooling -- they will almost all be the same at the same rpm. It's the noise that really differs.
Why are there so many different fans?
Fans can be provided for a wide range of voltages, including DC & AC, from 25mm through to 254mm square frames although AC types are usually limited to fans 80mm square, and larger. Clearly a wide range of operating voltages simplifies matching the fan to existing power supplies, for example Micronel offer a 80mm fan with a unique 5V option, while Airflow Tech offer both 48 and 53 V DC models. Apart from considerations of supply voltage, the three key physical parameters are:

1. Frame size
2. Speed
3. Thickness

Larger frame size and higher fan speed generally increase airflow, but since higher speed also creates more pressure and noise, the optimum selection will depend on a balance between space available and acceptable noise levels. It is worth noting, however, that it can be difficult to achieve high reliability within the smallest of frame sizes not least because of the lack of space for adequately sized ball bearings (ie less than 10mm thick usually results in very high loading).

The thickness of the fan determines its pressure capability, but note that the datasheet pressure rating is for stalled air (i.e. no airflow), so it is important to check the performance curve to find the airflow delivered under your expected system static pressure conditions. All other things being equal selecting a thicker, and so higher pressure rated fan will normally deliver more airflow against a given system pressure and this may be a more practical alternative for increasing the effective cooling than fitting a higher speed or a larger framed fan.

rwerp
Posts: 12
Joined: Wed Mar 15, 2006 1:56 pm
Location: London, UK

Post by rwerp » Sun Nov 05, 2006 9:26 am

Have you considered using the Bernoulli's effect to measure the air speed? Just make the fan blow into a long tube with a pressure meater built in in the side.

mgh
Posts: 2
Joined: Sun Nov 05, 2006 10:53 am
Location: Johannesburg, South Africa

Possible improvement for CFM measurement

Post by mgh » Sun Nov 05, 2006 12:07 pm

I've been thinking about the problem of CFM measurement and have come up with the following. It solves three problems with the current solution: positioning of anemometer, effects of turbulence and measurement of the open area of the fan.
My proposed solution is a (cube shaped) box (see image) with a pvc pipe about the length of the box fitted through a hole in the side of the box and a hole in the top of the box to position the fan on.

Image

The pipe is about the length of the box and inserted two thirds into the box. The diameter of the pipe is about 1.5 times the diameter of the biggest fan to be tested. The pipe creates a uniform airflow, with a diameter big enough that the fan doesn't feel much resistance. The anemometer is positioned in this pipe at about two thirds from the left. The area is fixed now and equals the circle area of the inner diameter of the pipe.
The fan is positioned in a hole in the top of the box, about two thirds to the right. To be able to measure different sizes of fans, make the hole in the box big enough for the biggest fans to be measured. Then use pieces of board with a hole suitable for each of the smaller fans and put this between the fan and the box to measure the smaller fans.

I hope my idea may help you.

Best wishes,

Marcel

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Sun Nov 05, 2006 2:02 pm

MikeC wrote:Given fans of 25mm depth spinning at typically <1500rpm, the pressure is always very low, and differences between models is trivial. To wit -- we have interchanged Nexus, Scythe, and Globe 120x25mm fans running inaudibly on Ninja HS with no significant differences in cooling. I've done the same with 80x25mm Panaflo, Nexus and other fans on earlier HS -- again without any notable change in temperature.
There is another interaction between the fan and a heat sink that I think is probably more interesting: the change in sound when a fan is placed against a heat sink vs its sound in free air.

When I was experimenting with AcoustiFan DustProof fans, I found that they made a distinct thrumming sound mounted on a Ninja, which was not present in free air. When I tried a Nexus, this thrumming was completely absent.

Close examination of the fans revealed that the Nexus has a larger gap between the downwind edges of the blades and the plane of the housing than the AcoustiFans do.

So it might be a helpful step in a review to listen to the fan up against a heat sink...

Rusty075
SPCR Reviewer
Posts: 4000
Joined: Sun Aug 11, 2002 3:26 pm
Location: Phoenix, AZ
Contact:

Post by Rusty075 » Sun Nov 05, 2006 2:43 pm

cmthomson wrote:There is another interaction between the fan and a heat sink that I think is probably more interesting: the change in sound when a fan is placed against a heat sink vs its sound in free air....
So it might be a helpful step in a review to listen to the fan up against a heat sink...
Too many variables to test, in my opinion. It's similar to the idea of testing the fan with heatsinks attached to measure for backpressure specs.

You'd have to test each fan, at multiple speeds, with multiple heatsinks, since the noise probably changes with the fin configuration of the heatsink.

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Sun Nov 05, 2006 6:12 pm

Rusty075 wrote:
cmthomson wrote:There is another interaction between the fan and a heat sink that I think is probably more interesting: the change in sound when a fan is placed against a heat sink vs its sound in free air....
So it might be a helpful step in a review to listen to the fan up against a heat sink...
Too many variables to test, in my opinion. It's similar to the idea of testing the fan with heatsinks attached to measure for backpressure specs.

You'd have to test each fan, at multiple speeds, with multiple heatsinks, since the noise probably changes with the fin configuration of the heatsink.
Well, I suppose if you meant to do a really thorough job if it. I was only suggesting a single test at some reasonable speed, such as 7V. This would be adequate to reveal the difference between the Nexus and the AcoustiFan.

Rusty075
SPCR Reviewer
Posts: 4000
Joined: Sun Aug 11, 2002 3:26 pm
Location: Phoenix, AZ
Contact:

Post by Rusty075 » Sun Nov 05, 2006 7:46 pm

Yeah, but what is it really revealing? If it's only revealing that Brand Q fans may have an interaction with tower heatsinks that have fins X mm thick spaced Y mm's apart when mounted Z mm's from the face of the fan...that info might not be that useful.

Don't get me wrong, I think it's an interesting topic, but I would be more interested in learning why that particular fan has a reaction with that particular heatsink, instead of testing every fan just in case it might have a reaction.

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Mon Nov 06, 2006 1:50 am

cmthomson@Well, I suppose if you meant to do a really thorough job of it.
Part of the reason this fan roundup has taken so long to get to this stage is that the editorial team have tried to be as thorough as possible; the fan market is constantly changing, with new fans being introduced almost every week, so there is a trade-off between thoroughness and timeliness.

Besides, there is an easy solution to fan-induced resonance or a change in fan sound quality in proximity to a heatsink: move the fan further away! :wink: This may require some ducting and jerry-rigged fan supports, but then that's nothing new to most SPCRers.

JayJay
Posts: 14
Joined: Thu Oct 26, 2006 6:02 am

Post by JayJay » Mon Nov 06, 2006 1:51 am

Could you also make a list, lets say for your system you would need 14cfm to keep the cpu under "ok" condition.

And then list the fans at what dBA +rpm they will have while blowing that specified amount of air volume?

Or have 2 different columns, one for 14cfm and one for 25cfm for eg. So us readers easier can see the difference.

looking forward for the reviews.

nomoon
Posts: 210
Joined: Thu Apr 14, 2005 8:35 pm
Location: Allen, TX US
Contact:

Post by nomoon » Mon Nov 06, 2006 9:19 am

I like the article very much, though I'd like so offer the following idea regarding the measurement of airflow (it looks like "mgh" has already made a similar suggestion).
Image
  • This would eliminate turbulent proximity effects of having the anemometer placed next to the fan.
    The resistance would also be identical for all fans.
    There would be no need to extract area of the fan hub, thus reducing a source of error.
The absolute CFM numbers still may not be exact, but the comparisons between dissimilar fans would be accurate and repeatable.


Jason

mgh
Posts: 2
Joined: Sun Nov 05, 2006 10:53 am
Location: Johannesburg, South Africa

Re: Fan airflow measurement apparatus

Post by mgh » Mon Nov 06, 2006 11:29 am

Thanks Jason,

I see you're thinking along the same lines as me (and your drawing is much better).

However, I see three drawbacks of your design.

1. The anemometer will restrict the airflow more.
2. The anemometer will be designed expecting axial airflow, while in your design the air exiting the box will mostly come from the sides of the anemometer instead of flowing straight at it (much like when you pull the plug in a bathtub), reducing the readings. My pipe design creates a linear air flow.
3. The anemometer will be designed to operate in open air, that is, the design takes air flowing around the meter (because of the obstruction the meter forms) into account. Readings might be too high if the air is forced through the meter. Some space around the meter corrects this.

And of course a practical one, it's easier to lay the fan over a hole on top of the box then to mount it on the side.

Best wishes,

Marcel

Sooty
Posts: 625
Joined: Mon Sep 22, 2003 5:15 am
Location: UK

NoiseBlocker BlackSilent

Post by Sooty » Mon Nov 06, 2006 11:30 am

Can I ask whether the NoiseBlocker BlackSilent series is lined up for testing? If not, I’ll mail them to get them to send samples. They’ve also got a cracking CPU cooler, see if they can’t send that as well.

paapaa
Posts: 198
Joined: Wed May 04, 2005 1:24 am
Location: Finland

Post by paapaa » Mon Nov 06, 2006 11:49 am

At least the over-hyped Noctua fans should be tested :)

Ackelind
Posts: 467
Joined: Sun Mar 20, 2005 1:18 pm
Location: Umea, Sweden.

Post by Ackelind » Mon Nov 06, 2006 1:02 pm

I'm looking forward to see a comparision between the four main fans available here in sweden (120mm ones), namely Adda (hypo bearing? ones), Nexus, Papst and Scythe. There is always endless discussions about which one is the best.

nomoon
Posts: 210
Joined: Thu Apr 14, 2005 8:35 pm
Location: Allen, TX US
Contact:

Post by nomoon » Mon Nov 06, 2006 7:44 pm

Marcel (mgh),

Thank you for the kind words regarding my drawing. :) This is a good discussion. I think that there are some trade offs with each design. Let me comment on the following:

Absolute CFM Measurement: The “pipeâ€

Devonavar
SPCR Reviewer
Posts: 1850
Joined: Sun Sep 21, 2003 11:23 am
Location: Vancouver, BC, Canada

Post by Devonavar » Mon Nov 06, 2006 8:29 pm

No, the anemometer can get quite loud. But, there's nothing that says we have to measure noise and airflow simultaneously...

MikeC
Site Admin
Posts: 12285
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Post by MikeC » Wed Nov 08, 2006 9:03 am

Regarding CFM measurements -- at the fan blades vs through a test plenum. The latter concept is documented in detail in Standard ISO 10302 – Method for the measurement of airborne noise emitted by small air-moving devices. I have a copy of this document; here's the drawing of the reference plenum:

Image

One of these plenums was built to 50% scale by a mech eng. UBC student in a project instigated by me a couple of years ago to test fans. (The 50% scale is considered a viable option, according to the standard; it was necessary because the full sized one would not fit through the door of the anechoic chamber.) Due to time limitations, only a handful of Panaflo 80L fans were actually tested, for sound power, airflow and SPL at different voltages. This test plenum is actually still in the UBC sound lab, and I am trying to obtain permission to use it, or to purchase it outright.

Meanwhile, the question remains why we did not build one.

The main reason, as mentioned earlier, is that the results of all other test methods gave us CFM numbers considerably lower than ANY fan maker's specs. This includes and/all schemes where the airflow is measured some distance away from the fan (through a flow guide or tunnel). It does not appear to matter whether that distance is 12" or 3'; it always gave us much lower numbers. I was interested in a measurement scheme that would make it possible to make some kind of comparison against published mfg's figures. The at-the-fan-blades airspeed + cubic flow calculation method gave us the closest matches.

Someone mentioned turbulence effects -- yes they are there, but they are there in the vast majority of heatsink applications as well. Fans on HS are usually used right atop the blades. But not in a case cooling application, you say. Right. We can't always have everything... and my hunch is that the small measured CFM difference in a case cooling app is very trivial. It's true that small differences in airflow make little difference in cooling on a low impedance HS; it makes even less difference for case cooling.

I want to repeat again that although we're obviously checking for CFM in free-air, zero impedance conditions, every fan of the same size/depth has virtually identical pressure at the same airflow. It's only when the depth is greater than 25mm that there's even a possibility of greater pressure.

So if a Scythe 120mm fan on a specific HS/CPU @ 7V gives us 40C, then work backwards and find that we measured this fan at 20cfm, then you can be assured that any other 120mm fan we tested set for 20cfm will give exactly the same cooling result. What I am saying is that you can trust our results apples-to-apples as long as the fans are the same size.

sun.moon
Posts: 174
Joined: Wed Sep 21, 2005 5:57 am
Location: Germany

Post by sun.moon » Fri Nov 24, 2006 2:22 pm

Hello MikeC,

A very interesting article. I also spent some time a year ago doing some figuring about fans. Looking forward to your results and appreciate your efforts.

Just a thought from my side concerning CFM. Have you considered approaching a couple of the most common fan manufactures used in the silent community to see how they test CFM and maybe come to an agreement to employ their (or a similar) testing method, if feasable? You say that all the other methods you tried resulted in regular lower CFM results than the manufacturers specs.

Kind regards,
sun.moon

tg401
Posts: 8
Joined: Sat Nov 04, 2006 4:57 pm

Post by tg401 » Sat Dec 02, 2006 1:33 am

MikeC wrote:To address concerns about pressure...

Given fans of 25mm depth spinning at typically <1500rpm, the pressure is always very low, and differences between models is trivial. To wit -- we have interchanged Nexus, Scythe, and Globe 120x25mm fans running inaudibly on Ninja HS with no significant differences in cooling. I've done the same with 80x25mm Panaflo, Nexus and other fans on earlier HS -- again without any notable change in temperature.

The reason for the absence of pressure differences has to do with the thin depth -- 25mm. The blades cannot have enough depth for any to exhibit real differences.
Would this reasoning still apply to the Noctua fan that was reviewed? Although the depth is the same, the geometry and size of the blades is quite unique.

I am trying to make a final decision between the Noctua and Nexus and I am curious if the this is an issue worth considering.

Post Reply