Possibly some active noise control technique could be considered.

Hi John,

I suggest you first to reduce the vibrations of the blades. Change its material, add some viscoelastic materials to damp it.

Why are you trying to reduce noise inside this machine ? I have always seen serers in dedicated room, so it will be better and cheaper to reduce the noise in the rrom than in the computer.

1. Active noise reduction is an option but an expensive one - it means more electronics. Still to be investigated.

2. This is a specification from a customer. We can make all the recommendations, however, if you don't meet the noise spec, you may not get the job to start with.

Hi John,

You could find very nice noise reduction algorithms that have been uploaded on the File Exchange - Matlab Central at the following link :

1- the open-loop Wiener filter with decision-directed (from one of my paper published in 1993 at ICASSP)

2 - the Two step-noise reduction TSNR) technique (also one of my paper)

3- the harmonic regeneration noise reduction (HRNR) method (also)

The first algorithm can be found at http://www.mathworks.com/matlabcentral/fileexchange/7673-wiener-filter

The second and third ones at http://www.mathworks.com/matlabcentral/fileexchange/24462-wiener-filter-for-noise-reduction-and-speech-enhancement

I hope that it will help you. Tell me if this is the case.

Sincerely,

Pascal SCALART

Dear Mr. Tremblay,

I have the experience to significantly reduce noise of cooling fans by different passive low cost measures.

Please contact me to learn more.

Best regards,

Lothar Gaul

I would go for a thin layer of micro-perforated elements (metal sheets with submillimeter holes), placed at a certain distance from the wall (maybe the case containing the fans), depending on the frequency range where you want to be more effective. Those panels might be costy, but there are cheaper versions like Acoustimet micro-slits. Soon a new kind of such panels will be presented by my team, with a number of advantages compared to the previous ones. Just contact me if you want to know more.

"Enjoy the silence"! (D.M.)

Fabio

Is your fan noise problem tonal or broadband ?

If it is tonal, I have patented a technology based on the interaction mechanisms between a flow control obstruction and the fan propeller. Optimal axial and circumferential positioning of the flow control obstruction generates a secondary noise of equal magnitude but opposite in phase with respect to the primary tonal noise. Important reductions occur without affecting the fan efficiency.

If it is broadband, no need of this technology !

You can contact me if you want more information.

Cordialement,

Anthony

thanks for all the feedback. I will check out the MATLAB algorithms. The fans used are variable speed counter-rotating units, which produce significant tonal noise as well as the typical broadband "wind" noise. The tonal noise, I believe, is likely the culprit in terms of an annoyance factor, but obviously, in order to meet a noise specification, the entire noise emission must be at or below a certain level. We are still doing tests here and I will come back with more info in a few days.

Hi John,

What is the spectral characteristics of your tonal noise (I mean, the frequency range where most of the energy is concentrated)? In the case the main tonal components lie in the low frequency range, I fear you cannot prevent from employing active noise control solutions.

However, there might be cheaper alternatives to "standard" active noise cancellation, employing active sound absorbers (the same as passive materials, but far more absorbent in the low frequency range) rather than using transducers to produce anti-noise (which is, I agree, a very demanding technology in terms of hardware/signal processing).

May I suggest you have a look at the following article:

http://asadl.org/jasa/resource/1/jasman/v129/i5/p2968_s1

to make up your mind?

Kind regards,

Hervé

the majority is in the range 500Hz to approximately 3kHz.

You mean, the global noise (tonal + broadband)? Usually, it is assumed that the annoyance is mainly due to the tonal part of noise (assuming the levels are significant compared to the broadband noise). So my first reflex would be to accurately identify the main annoying tonal frequencies (let's say the ones in the sound power spectrum that present sound power levels above a certain threshold) to decide whether it is worth or not to employ ANC (roughly speaking, if the energy is mainly concentrated below eg. 500 Hz).

if those tonal components do not go too much below 500 Hz, I would rather recommend design duct silencers, lined with passive absorbers such as resonators or Micor-perforated panels combined with air plenum.

Another technique is to design side-banches along the duct silencer to realize quarter-wave resonators, that act as a notch filters (more tricky to design than Helmholtz resonators, and only functional if the tonal frequency matches the dimensions of the derivation).

But I also suspect that you can actually find off-the-shelf silencers for this kind of noisy devices (I remember my system manager showing me such silencers for computer fans a few years ago...).

regards,

Hervé

Hello John!

I have read your question and the following discussion with a great interest! Considering the application, i.e. cooling of an electronic apparatus (e.g. a computer processor), where the low-cost of the complete solution (including research efforts in design, manufacturing, etc.) has the highest importance, I would definitely recommend a passive noise control approach. Hervé has recommended several great options.

Since you have not described the problem in detail, it is difficult to offer any definite or optimal answer. On the other hand, I would look at the problem from a more practical perspective. If the noise generated by fans is really an issue for your customers, what about to eliminate fans completely. ;) I would suggest to use a water colling system with the placement of heat exchangers (with fans) at places, where the generated noise might not be an issue for your customers. Another possibility is the use (or development) of some novel and efficient heat exchanger without fans at all. ;)

I know it is a cheating from the point of view of the acoustic research, but it may help you to solve your problem in an elegant and efficient way. ;)

With my best regards,

Pavel

Actually - I did not specifically mention the equipment involved but only that it is similar in size to that of a server unit, and utilizes the same kinds of cooling fans (variable speed, counter-rotating). The units are telecommunications power electronics. We do indeed make liquid-cooled versions, however, there is a power level below which it is desireable to have only forced-air cooling. I have indeed considered liquid-cooling, but space constraints and ease of installation make this impractical (at this time). I continue to enjoy the feedback from everyone. Being prompted to think outside the box is refreshing.

jt

To find out a solution to the presented problem, firstly one should know the geometry of the flow channel. Secondly, if there are two counter-rotating fans, the numbers of their blades are needed (are they the same type of fans with the same geometries ?). Thirdly, the fans rotating speeds are to be presented. Fourthly, there should be more info about the dimensions of the blades. Next, the data in frequency domain from noise measurements is to be analysed.

Or... specific acoustic adaptation on the casing walls is to used.

Besides the suggested bigger fans with lower rotation speed I would also suggest to have a look at the installation of the fans. Whenever possible try to decouple the fans from other installed components to prevent transmission of the vibration to the housing, where it is transmitted as audible sound. In some cases the direct sound of the fan is not the real problem but the not known tranfer paths of sound energy. So, using rubber bearings and some absorbing material at the case could minimize the problem

Choosing a quiet, remote fan, connected by means of a hose can be an important improvement already. A silencer hose would increase the effect. If there are several fans in the original apparatus, you can replace them all by hoses connected to the one quiet (centrifugal) fan.

Acoustic Panels the distances, and the dimensions should be taken into consideration.

Do you mean the techniques used in noise cancelling headphones? 

http://www.wearableinear.com/best-noise-cancelling-in-ear-headphones/

Hi John,

You could use Lattice Boltzmann based Methods to simulate the fan alone or in its complex environment (including chassis, etc). Then you can analyze the CFD data and identify the noise-generating mechanisms to reduce both tonal and broadband noise. You can also reduce experimental testing by numerically designing the fan and numerically certify it. 

Regarding our solution PowerFLOW (LBM based), we have published many papers in association with our customers/academic partners on the topic (low/high subsonic axial/centrifugal fan noise assessment and minimization). Let me know if you want more information.

https://www.youtube.com/watch?v=rHSYjsROcpU

https://www.youtube.com/watch?v=83pVUmWVJcw

Conference Paper HVAC Blower Aeroacoustics Predictions Based on the Lattice B...

Conference Paper Direct aeroacoustics simulation of automotive engine cooling...

Conference Paper Axial fan noise aeroacoustics predictions and inflow effect ...