Flash! Fans not a primary noise source! Really!

Felger Carbon · Post by **Felger Carbon** » Sun Jun 10, 2007 10:31 pm

I just spent all afternoon writing a very careful and thorough explanation of the sources of deviation from the perfect model of fans, SLMs, and noise measurements.

Long story short: turns out fan noise isn't fan noise. I proved it with a simple experiment, one anybody with a $19.95 Radio Shack SLM can duplicate. We're not talking about subtle effects here, folks!

Because of this, my careful explanation was hogwash. When I plugged my actual measurements in, nothing fit. I decided that either I was the sloppiest person ever to take closeup fan noise measurements, or something basic was wrong.

My explanation depended on the fan being the source of the fan noise that's being measured. Let me describe the results of a simple experiment I just performed:

I have the GW NCB fan hanging on a twine "clothesline". My SLM is on a tripod with the microphone aligned on the fan hub. With the fan at 12V, I measured the "fan noise" 140mm from the fan hub, on the intake side of the fan. Got 43.3dBA.

Without touching the fan or turning it on/off, I moved the SLM/tripod around to the other side of the fan, 140mm from the rear of hub's reverse (the motor assy). The SLM measured 62dBA! WHOA!!

The noise level didn't drop back to ~43.3dBA until I moved the microphone over 300mm from the rear of the fan. Since I was using a 300mm scale, I estimate the distance at 310-315mm.
---------------------------------

Here's the explanation: The fan draws in and pushes out a column of air. The particles have kinetic energy added. The column of air is pushed out the back of the fan by the spinning fan blades. About now the column of air, at first at the outer edges only, begins to encounter the room's resident air particles. As the column advances, more and more of the resident air particles are bumped into. In the process of bumping into the resident air particles, the kinetic energy is given up.

Energy is never destroyed! It can change form, but it never simply disappears. The energy output of the fan is not noise, it's kinetic energy stored in air particles. The process of the column of air hitting the resident room air is highly turbulent, with the result that the kinetic energy turns into acoustic noise energy (whoosh). This process reaches its peak somewhere around 140mm behind the fan!

Therefore, when the SLM is 140mm in front of the 25mm-thick fan, the SLM is actually ~205mm from the noise source! And my careful writeup that I spent all afternoon on is hogwash, because I described the fan as the noise source... and did some trig calculations based on that. And my calculations didn't fit my data(sigh). ;(
---------------------------------

The collision of the column of air is highly turbulent. The SLM, stuck into this process, can only measure a part of the turbulence at a time. Despite being on "slow" response, the sound level was jumping around typically +/- 2 full dBAs! Frequently higher!

The fan does make a small amount of noise - motor noise, mostly. But the biggie is the whoosh, which is the turbulent collision occuring well behind the fan.
-----------------------------------

One reason I fought the "50LOG10" formula so hard is that Comair et al say all the kinetic energy does not get turned into noise. If it was all turned into noise, the correct formula (even for linear weighting) would be 60LOG10. I now accept that 50 is correct for linear weighting, but I still don't know what happens to that other energy, the part that doesn't turn into noise. Heat??

Das_Saunamies · Post by **Das_Saunamies** » Sun Jun 10, 2007 10:44 pm

Noise is air pressure, we know that much. Also, they put wind hats on microphones on TV because they pick up monster noise if air gets blown into them... could this have caused your readings?

I'm just an Arts student, so I have no expert scientific capability, but the above thought just popped into my mind.

Felger Carbon · Post by **Felger Carbon** » Sun Jun 10, 2007 10:50 pm

Das_Saunamies wrote:Noise is air pressure, we know that much.

Noise is randomly fluctuating air pressure. A steady pressure on a microphone diaphragm doesn't cause an output signal of any kind, noisy or not.

Das_Saunamies · Post by **Das_Saunamies** » Sun Jun 10, 2007 10:59 pm

Alrighty, got that cleared up then.

jaganath · Post by **jaganath** » Sun Jun 10, 2007 11:50 pm

when I saw that you positioned your SLM on the intake side of the fan, I though you were aware of this effect and were compensating for it.

The fan does make a small amount of noise - motor noise, mostly. But the biggie is the whoosh, which is the turbulent collision occuring well behind the fan.

this is why jet noise is so hard to reduce, a lot of the noise is produced outside of the engine nacelle as the extremely high-speed jet mixes with the (relatively) stationary freestream air.

this is also why any obstacles (such as restrictive grills) near the exit increase the noise level, they make the flow even more turbulent.

jessekopelman · Post by **jessekopelman** » Mon Jun 11, 2007 12:01 am

Felger Carbon wrote:One reason I fought the "50LOG10" formula so hard is that Comair et al say all the kinetic energy does not get turned into noise. If it was all turned into noise, the correct formula (even for linear weighting) would be 60LOG10. I now accept that 50 is correct for linear weighting, but I still don't know what happens to that other energy, the part that doesn't turn into noise. Heat??

But why does the data from SPCR fan roundups correspond more to 60logX than 50logX?

jaganath · Post by **jaganath** » Mon Jun 11, 2007 12:16 am

why does the data from SPCR fan roundups correspond more to 60logX than 50logX?

SPCR measurements are A-weighted,which follows 60log10 (according to FC, and this appears well-supported by the data).

Felger Carbon · Post by **Felger Carbon** » Mon Jun 11, 2007 1:27 am

jessekopelman wrote:But why does the data from SPCR fan roundups correspond more to 60logX than 50logX?

The actual formula is 60LOG10(rpm2/rpm1). We are talking about one fan, two RPMs. If the rpms are different, the noise spectrum produced by the fans is slightly different. The lower rpm produces somewhat lower frequencies. The A-weighting network in a SLM drastically minimizes the effect of lower freqencies. As a result of this filtering effect (and that's what it is), A weighting "penalizes" the lower-rpm noise level, so that what is 50LOG10 unweighted becomes 60LOG10 weighted. For a 2-1 difference in rpm, you get a 15dB unweighted change in noise level, or an 18dBA A-weighted change in noise level.

Bluefront · Post by **Bluefront** » Mon Jun 11, 2007 2:23 am

If I'm reading this correctly, it confirms what I've heard all along......fans blowing into a case, sound quieter than the same fans blowing outward. This is caused by the muffling effect of the case itself. If you've ever tried to record sound outside on a windy day, you'll duplicate FC's experiment. And I think this is part of the reason for the foam cover over a microphone....trying to break up/reduce the effects of a moving column of air.

How to make use of this effect you say.....turn your setups into positive pressure, blowing all the fans inward, letting the exhaust escape passively through vents. Sounds simple.....but there's more to it than that.

Felger Carbon · Post by **Felger Carbon** » Mon Jun 11, 2007 5:19 am

Something else gets explained as well. If the actual noise source is well behind the fan (and it is), then a fan mounted in a wall will prevent any sound from reaching the front from any source other than through the fan opening. In my current experiment, sound can pass around the outside of the fan case - and it does. This explains why I measure so much more noise from my GW NCB than is predicted if the fan itself were the actual source of the noise. And it explains the 19dBA - it was measured with the fan mounted in a wall, as some (many? all?) of the noise-measurement procedures require. By allowing noise from the actual noise source at the rear of the fan to pass around the fan casing, I measured 6dBA more noise than was measured with the wall.

When SPCR measures the noise of a fan at one meter that's mounted on a Ninja (or other HS), is there a wall? Nope. Does SPCR commonly measure higher SPLs than the manufacturer's specification? Yep.

Felger Carbon · Post by **Felger Carbon** » Mon Jun 11, 2007 9:55 am

I spent a few hours making some very interesting measurements. First, I switched back to the Yate Loon D12SH-12 fan to stay comfortably above the ambient noise level. Second, I rotated the mounting of the fan 90 degrees so I could access 1 side of the fan, or either the front or back, using the mike/tripod.

I discovered the extreme instability in SPL I had observed was a function of using the wrong (lowest) measurement range. The internal amplifier was clipping on the higher noise peaks. This went away when I selected the correct range. So whenever the reading started to look unstable, I'd try moving up or down a range.

I went looking for the highest reading I could find. This turned out to be 97.1dBA, and it was hard to find that level - it was very position-sensitive. But I was able to locate that reading on either the front or rear of the fan, at a zero distance from the fan case and just behind (in front of) the outer edges of the actual fan blades.

This is where the column of air was fastest, and where the very first encounters of the fast column with the ambient air could occur. 20mm behind the fan, I saw 92.0dBA, about 5-7mm in from the outer edge of the fan blades. 45mm behind I got 89.3dBA, almost halfway in to the hub. 70mm behind it was 88dBA, with the max moving back out (!) to 1/3 of the way in to the hub. 100mm behind, 84dBA again 1/3 of the way in. 200mm behind, 72dBA and suddenly the horizontal position resumed being very sensitive again - the maximum was 1/3 of the way in to the hub. Huh??

300mm back I got 60dBA and it wasn't horizontally position sensitive at all. Here's my interpretation of this mess:

The locus where the kinetic energy changes to noise energy isn't a point, or even a "ball". Instead, the locus is the skin of a plastic drinking cup, getting slightly smaller on the way down, and then there's a bottom. 200mm away from the fan, the bottom of this "cup" hasn't been reached. 300mm down, the bottom has been passed.

I made no effort to locate the "bottom of the cup" more accurately, since what I wanted was to establish general principles. I think I succeeded.

The location of the highest noise at the exact edge of the column of air and at the edges of the fan makes sense. The "cup" tapering makes sense, as the outer part of the column gives up its kinetic energy, the next collisions occur progressively inward. The narrowest part of the cup seems to be ~45mm back, so the cup is coke-bottle shaped. I don't understand the extreme lateral position sensitivity 200mm back. Something strange and (to me) unexpected there.

This means the "fan noise source" is a thin cyclinder whose maximum width is the blade width, ~110mm, and whose on-edge "height" appears to be over 200mm but not 300mm.

This makes me believe the "fan noise source" will not act anything like a point source, and it will be hard to make a math model of the dBA level vs distance from the front of the fan that makes any sense when close to the fan. That confirms my initial measurements, when I was blaming the odd results on mismeasurement (by me) of the close distances from the fan.

The locus of highest noise levels I found confirms that the noisemaking process is the exchange of kinetic energy for acoustic energy. None of this happens in the fan because there's no lazy ambient air particles in the fan. It does start at the very edges of the fan case tunnel, at the outer edge of the fan blades. It starts there because that's the earliest opportunity for high-energy particles to hit low-energy particles.

The noise source is entirely outside the fan, mostly in the back of the fan.

IsaacKuo · Post by **IsaacKuo** » Mon Jun 11, 2007 11:06 am

I'm surprised you needed to make fancy measurements to come to this conclusion. The fact that fan noise is highly directional should be rather obvious just from listening to one while moving your head or rotating a fan in your hand. It's NOT a subtle effect!

The extreme directionality of fan noise is something I've long considered the elephant in the room when it comes to fan noise measurements.

I'm skeptical about your ultimate conclusion, though. In my experience, airflow "whoosh" noise can occur ANYWHERE, depending on the exact geometry of the case and airflow. In particular, when a case has restrictive intakes, there's often a whoosh noise generated at the intake grills/filters, even though the "cause" of the noise is in the opposite corner of the case in the form of one or two exhaust fans.

Your experiments trying to determine the exhaust geometry of the whoosh noise profile in the case of an open air fan are interesting, but ultimately of limited practical use because airflow geometry is so radically different in any practical situation.

I definitely think that it's impossible to draw a conclusion that generally fans blowing into a case create less noise than airs blowing out of a case. I've seen it go either way. Whether blowing into or out of a case, I find that most of the airflow noise comes from turbulent airflow within the case. There are just so many corners and obstacles and walls in a case which result in turbulent airflow patterns. It seems like blowing air into a case results in even more turbulent internal airflow, while pulling air out of a case results in somewhat smoother internal airflow.

There's only one reliable rule of thumb I've come up with--to get rid of airflow noise, get rid of airflow. That's why I use a single undervolted 80mm fan for my silent builds. Even at minimum voltage, a 120mm fan generates too much airflow to effectively silence (given my low noise floor).

Post by **Rusty075** » Mon Jun 11, 2007 1:14 pm

A couple of points:

-saying "fans don't make noise, its the air turbulence that does" is a bit like saying "speakers don't make noise, its the air movement that does". (or "guns don't kill people, people kill people"). They're true statements, but not really surprising. All noise is from air movement, that's the only thing we can hear after all. The idea that the noise from fans does not all originate from the fan itself is probably something that most people haven't thought about though.

-the measurements for where the turbulence noise originates are valid for only that particular fan model, and at a specific rpm. Change the fan diameter, depth, blade geometry, and even (to a lesser extent) its speed, and all of those data points become meaningless.

-you've just discovered why all fan testing specifications call for measurements to be made from 1m. At the close ranges you are trying to measure at you are going to get all sorts of nearfield interactions that will make your readings wonky. Some of them are caused by the fan not being a point-source of noise, and some are the result of interferences within the SLM mic, which is not a point-recording source either.

-I may have missed this in an earlier post, but what SLM are you using? If you are getting down close to its absolute minimums you are going to lose resolution in the results. There's a danger in putting too much faith in measurements that are probably really only +/-10-20%.

Felger Carbon · Post by **Felger Carbon** » Mon Jun 11, 2007 1:31 pm

Rusty075 wrote:-you've just discovered why all fan testing specifications call for measurements to be made from 1m.

An outstanding idea in an anechoic chamber. A lousy idea in a live room with undefined reflections IMHO. Including in my room, which is why I prefer close measurements to get the intrinsic noise caused by the fan.

I'm using an Extech 407738, rated noise floor 26dBA with a 1/4" microphone. My room ambient measures 26.9 when things are quiet and in the 27s when traffic picks up. It's quietest in midwinter at 3AM. There's nothing moving or doing then!

edit: typos, 4x.x instead of intended 2x.x

Post by **Rusty075** » Mon Jun 11, 2007 5:29 pm

Felger Carbon wrote:An outstanding idea in an anechoic chamber. A lousy idea in a live room with undefined reflections IMHO. Including in my room, which is why I prefer close measurements to get the intrinsic noise caused by the fan.

But moving the mic closer to compensate for a poor testing environment does not unnecessarily give you better results. It may give you more consistent results, but they may be consistently wrong. An equivalent statement would be, "I don't have a microscope, so I just leaned in real close and squinted. Here are all protozoans I saw..."

I'm not trying to be negative about your testing, please don't take it that way. I think its a great thing, and very useful, but just that you need to be realistic about the results that can be achieved under those conditions.

Just look at SPCR...after 5 years of pretty much full-time work, and with access to expert opinions and financial resources that (I'm guessing) probably outweigh yours, Mike is still trying to find a way to measure fan noise accurately. It's a very complicated subject, particularly when trying to do it on a shoestring.

Howard · Post by **Howard** » Mon Jun 11, 2007 7:04 pm

Did you come to a conclusion about how to minimize the noise?

Felger Carbon · Post by **Felger Carbon** » Mon Jun 11, 2007 9:39 pm

Rusty075 wrote:But moving the mic closer to compensate for a poor testing environment does not unnecessarily give you better results. It may give you more consistent results, but they may be consistently wrong.

You are 100% correct; I may indeed be consistenly wrong. But two days ago, I thought that by measuring very close to the fan hub I was measuring very close to the noise source. Now I know that's totally wrong. An important part of getting better results is to know what it is you're measuring. I think I'm making good progress, and thought it would be amusing (to most) and thought-provoking (to some) to point out that the fan isn't the actual source of the noise, nor is it the primary locus of the noise.

jaganath · Post by **jaganath** » Tue Jun 12, 2007 1:48 am

I thought that by measuring very close to the fan hub I was measuring very close to the noise source. Now I know that's totally wrong.

I wouldn't say it was totally wrong. certainly for all the fans I have access to the noise is loudest when my ear is right up against the fan hub (not so close as to endanger my earlobes however!

) yes, turbulent interactions with the air are what cause the noise, but these will be strongest at the blade tips where air velocity and shear stress is highest and the eddies and turbulent vortices would be most energetic.

Bluefront · Post by **Bluefront** » Tue Jun 12, 2007 2:50 am

By measuring close to the hub/motor, you are separating the motor bearing noise from the air movement noise. That's why when you put your ear very close to the fan motor, the sound changes somewhat......because the air movement noise is reduced at that close distance.

There are various methods to reduce air movement noises. I'm convinced that sucking air through a low restriction filter, reduces intake air noise. This is like a wind-screen over a microphone.......testing needs to be done in this area. (Hint)

cmthomson · Post by **cmthomson** » Tue Jun 12, 2007 2:54 pm

At the risk of offending you, I need to point out that when you put the microphone that close to a fan, you are primarily measuring the buffeting of the meter's microphone. Ever tried to record outdoors in a wind?

Not mentioned in the previous posts is that the most common measurement point for fans is 1m away, 45 degrees from the fan plane. The reason for the large angle is of course to reduce the buffeting that would occur if the microphone were placed directly into the air stream...

BTW, 47 dBA is an extremely high number for a very quiet environment. It should be in the mid to high 20's for an indoor suburban room (with of course no AC, furnace, refrigerator, TV, traffic, weather, spouse, dogs/cats, etc, etc making noise; ie, Sunday 6am on a calm day)

Devonavar · Post by **Devonavar** » Tue Jun 12, 2007 3:21 pm

I'm not sure that the wind screen idea is quite what we're after. Wind screens for microphones are intended to block airflow while remaining acoustically transparent. Some attenuation is inevitable I suppose, but the design goal is the opposite of what you're looking for: Blocking noise while keeping airflow unrestricted.

However, bearing in mind the point that airflow (flucuating air pressure) and noise are two sides of the same coin, this really comes down to reducing air turbulence in the immediate vicinity of the microphone.

I think the appropriate analogy for a wind screen would be earmuffs, not a filter at the noise source.

Felger Carbon · Post by **Felger Carbon** » Tue Jun 12, 2007 4:18 pm

cmthomson wrote:BTW, 47 dBA is an extremely high number for a very quiet environment.

I apologize. After stating that the usual quiet ambient I measure with a SLM having a 26dBA electrical base noise is 26.9dBA, I then wrote "47" instead of 27. I'll correct that by editing.

Alas, there is apparently no 1" microphone option for my Extech, which would provide a far lower electrical or SLM noise floor. Under the circumstances, I think 26.9dBA is reasonable.

Felger Carbon · Post by **Felger Carbon** » Tue Jun 12, 2007 4:28 pm

Devonavar wrote:However, bearing in mind the point that airflow (flucuating air pressure) and noise are two sides of the same coin, this really comes down to reducing air turbulence in the immediate vicinity of the microphone.

I have previously (in this thread) pointed out that the large fluctuations of SPL I had encountered were in fact my simply not selecting the correct measurement range on the SLM, with the result that the microphone amplifier circuit was overloaded. That problem went away by simply selecting the correct range. This was the first occasion I'd encountered with this SLM where the lowest range was not the correct one! I'm still a newby on this specific piece of gear...

disphenoidal · Post by **disphenoidal** » Tue Jun 12, 2007 7:40 pm

Devonavar wrote:I'm not sure that the wind screen idea is quite what we're after. Wind screens for microphones are intended to block airflow while remaining acoustically transparent. Some attenuation is inevitable I suppose, but the design goal is the opposite of what you're looking for: Blocking noise while keeping airflow unrestricted.

However, bearing in mind the point that airflow (flucuating air pressure) and noise are two sides of the same coin, this really comes down to reducing air turbulence in the immediate vicinity of the microphone.

I think the appropriate analogy for a wind screen would be earmuffs, not a filter at the noise source.

I think a windscreen would be a solution, or part of a solution. The microphone and SLM don't respond to the buffeting in a predictable way--e.g., perhaps the buffeting causes the case of the meter to vibrate and amplify the sound? The only reasonable way to deal with this, I think, is to move the meter out of the fan's air stream. Also, considering no silent-PC enthusiast would sit in the stream of a fan, I don't quite see why an off-axis measurement wouldn't be preferable.

jaganath · Post by **jaganath** » Tue Jun 12, 2007 10:35 pm

the usual quiet ambient I measure with a SLM having a 26dBA electrical base noise is 26.9dBA

would it be safe to assume that we can infer from this that your best ambient is below 26dBA? If your ambient was 26 or above, one would automatically assume 26+3=>29dBA.

Felger Carbon · Post by **Felger Carbon** » Tue Jun 12, 2007 11:21 pm

jaganath wrote:would it be safe to assume that we can infer from this that your best ambient is below 26dBA? If your ambient was 26 or above, one would automatically assume 26+3=>29dBA.

Best hell, my average ambient, in daytime, at this time of year, is ~27.2dBA when there's no traffic within 2 blocks, which is usually (that is, when I cannot hear a specific vehicle with my own biological SLM). And refrigerator off, of course.

In Dec-Jan, at 3AM, my place is dead quiet.

Flash! Fans not a primary noise source! Really!

Flash! Fans not a primary noise source! Really!

Re: Flash! Fans not a primary noise source! Really!

Re: Flash! Fans not a primary noise source! Really!