Tracking down sound sources in fans is not an easy task. Whereas rela­tively advanced aero­dy­namic simu­la­tion programs are now avail­able and have become estab­lished on the market, the simu­la­tion of aeroa­coustics is still very much at the research stage. The number of cells needed for the required spatial reso­lu­tion of small turbu­lent struc­tures is far higher than in the case of aero­dy­namic simu­la­tion. For a fan in partic­ular flow situ­a­tions the figure may even be as high as tens or hundreds of millions.

Another neces­sity is high time reso­lu­tion – typi­cally with inter­vals of around 10 microsec­onds. This demands great compu­ta­tional resources and the asso­ci­ated time and finan­cial outlay. That explains why, in the case of fans for example, only the larger turbu­lent struc­tures (of rele­vance to acoustics) are resolved. Even with these limi­ta­tions the outlay is consid­er­able, and work is currently in progress to find ways of reducing the amount of compu­ta­tional work involved. Exper­i­mental processes are helpful in this regard.

Beam­forming with micro­phone array

One example of an exper­i­mental method of local­izing sound sources on a rotating fan used by the ebm-papst motor and fan special­ists to supple­ment complex aeroa­coustic simu­la­tion is the so-called beam­forming process. At the heart of this is a circular micro­phone array (Fig. 1) with 80 micro­phones arranged on two levels.

The micro­phone array is used on the intake side of the fan test stand to measure the differ­ences in prop­a­ga­tion time of the sound waves to each of the micro­phones. Sophis­ti­cated algo­rithms then eval­uate the data obtained over a period of 30 seconds at a known fan speed. The results show that the beam­forming method detects the same trends as aeroa­coustic simu­la­tion (Fig. 2). The exper­i­mental find­ings thus also permit checking and opti­miza­tion of the numer­ical simu­la­tion. For local­izing sound sources ebm-papst uses the so-called beam­forming process as a supple­ment complex aeroa­custic simu­la­tion.

The eval­u­a­tions reveal two domi­nant sources of noise for a typical axial fan: the tip gap flow between blade and fan housing and the so-called inflow turbu­lence (Fig. 3). At the tip gap, the differ­ence in pres­sure between the outlet and intake sides causes air to flow over the fan blades at the blade tip. The flow inter­acts with the edges present there, in other words the blade surface and the surrounding housing wall. Vortices form which can increase the sound level by up to 10 dB on sepa­ra­tion.

Inflow turbu­lence is partic­u­larly an issue when the fan is enclosed. A box such as those used for heat exchangers, for example, was chosen for testing with the micro­phone array. Back­flow areas with corre­sponding circu­la­tion, in other words air turbu­lence, occur at the housing walls. This is then drawn towards the points at which the gap between fan and housing wall is at its narrowest. The turbu­lence from both sides merges here.

These “vortex strings” then produce great turbu­lence (Fig. 4). As a result, consid­er­able fluc­tu­a­tions in pres­sure and velocity occur at the front edge of the blade, resulting in some­times dramatic addi­tional noise in the low-frequency range in partic­ular. This takes the form of both broad­band noise and narrow band tonal sound compo­nents, also known as blade passing noise. Everyone has at some point come across the unpleasant “humming” noise typical of this.

From finding the cause to tack­ling the noise

Once the sources of noise have been local­ized, action can be taken to improve the aeroa­coustics of the fans: The size of the gap between the blade tip and fan housing was found to have a consid­er­able influ­ence on noise behavior. The noise level decreases with a smaller gap, but produc­tion-related neces­si­ties mean that the gap dimen­sion cannot be reduced beyond a certain point because of the risk of the blade tip catching on the fan housing.

This is where winglets can help. These curved end caps added to the blade tips influ­ence the tip gap flow and the vortices that form and so signif­i­cantly reduce the noise level (Fig. 5). This has a posi­tive effect on the tip gap flow, thus reducing the inter­ac­tion of the flow with the edges. The sound power drops by up to 10 dB as a result.

Geometric modi­fi­ca­tions to the fan alone are however not suffi­cient to reduce the inflow turbu­lence, as this results from the instal­la­tion situ­a­tion. Addi­tion­ally insu­lating the housing tends not to be partic­u­larly successful either, as the insu­la­tion panels used normally only take effect as of higher frequen­cies. A different approach is more promising: Improving the inflow of air to the fan lessens turbu­lence and thus also the annoying low-frequency noise it gener­ates. For this purpose ebm-papst has devel­oped a special air inlet grill (Flow­Grid) that acts like a flow straight­ener on the intake air. It thus dramat­i­cally reduces noise-gener­ating distur­bance in the inflow and is equally effec­tive with both axial and centrifugal fans (Fig. 6).

Regard­less of the struc­tural condi­tions and the instal­la­tion situ­a­tion in the housing, the fans then attain noise values compa­rable with those for oper­a­tion under labo­ra­tory test condi­tions. Aeroa­coustic testing has thus proven its worth as a means of opti­mizing fans. It will be inter­esting to see what the future holds in store. One thing is certain, energy-effi­cient fans from ebm-papst will become even quieter still.

---

“Numbers alone say nothing”

The level of noise emis­sions is a deci­sive quality factor for fans. Dr Marc Schneider, ebm-papst’s Group Leader for Acoustics, explains how the company keeps its prod­ucts quiet

At what point is fan noise perceived as disturbing?

That is not an easy ques­tion to answer. There are of course phys­ical char­ac­ter­is­tics like the noise level that you can measure on a test bench. But numbers alone often say nothing about how the human ear perceives this kind of noise. For a subjec­tive assess­ment, what is impor­tant is how “raw” the noise is perceived. This percep­tion can happen when the signal is given a temporal struc­ture through a change in the frequency or ampli­tude.

A lot of noises also have tonal compo­nents which can be extremely disturbing. This percep­tion differs from person to person, which makes assess­ment even more compli­cated. One person reacts nega­tively to low-frequency noises, another winces at higher-frequency noises.

How do you measure this personal percep­tion?

At ebm-papst, we have the “AudiMax”, a so-called psychoa­coustic labo­ra­tory. In this noise-insu­lated facility, we have space for up to eight test listeners to whom we can play the noises of our prod­ucts in different config­u­ra­tions.

How does this method help achieve usable results?

After hearing the noises, our staff ask the test subjects about how they perceived them and thus create a data­base with scien­tific consid­er­a­tions. Using this data­base, we can eval­uate together with our colleagues from produc­tion which measures need to be taken and which not. Our overall aim is to create a fan that is perceived as comfort­able by as wide a spec­trum of test subjects as possible.

---

Sound effects

Analyzing noise to provide figures relating to physics and people.

A product’s quality in terms of its sound emis­sion is usually deter­mined by its dB(A) value, which refers to the sound pres­sure level gener­ated by an acoustic source at a certain distance. Depending on the level of the alter­nating pres­sures from which the sound pres­sure level is deter­mined, expo­sure to noise can cause effects ranging from slight reduc­tions in mental perfor­mance to pain and even uncon­scious­ness.

The concept of dB(A) also includes a means of assessing acoustic measure­ments to help answer the ques­tion, “How does noise affect people?” This ques­tion has kept scien­tists busy for decades and is now of increasing interest to engi­neers. For example, the following find­ings have been compiled in psychoa­coustic studies: For phys­i­o­log­ical reasons, humans do not perceive every frequency equally. Figure 1 shows the results of hearing exper­i­ments with tones of different frequen­cies. The curves show, depending on frequency, what sound pres­sure level is needed in order to be perceived uniformly by humans. This is called loud­ness.

This diagram provides a number of insights. For one thing, it can be seen that the human ear is most sensi­tive in the range between two and four kilo­hertz. Much higher or lower frequen­cies are perceived as quieter, even at the same sound pres­sure level. It can also be seen that this char­ac­ter­istic of hearing is not only depen­dent on loud­ness. The isophone at 40 phon was used for the dB(A) weighting (Figure 2). In addi­tion to dB(A), there are other frequency weighting methods. For example, dB(C) has been proposed as a better alter­na­tive for high sound pres­sure levels. The dB(D) weighting has become estab­lished for aircraft noise.

The dB(A) value and its reduc­tion from one product gener­a­tion to the next are sales argu­ments for indus­trial prod­ucts. This is insuf­fi­cient when the effect of noise reduc­tion on people is deter­mined by a signif­i­cantly different percep­tion than that of pure loud­ness. For example, other research has shown that a noise reduc­tion of approx­i­mately ten dB(A) is perceived as a halving of the loud­ness.