A distilled summary of fundamental design elements and key factors that affect the acoustic performance of DC axial fans in the role of PC cooling. This information is presented apart from our fan test methodology or actual fan review articles to keep information overload at bay. NOTE: A Postscript Q&A features Dorothy Bradbury, UK fan maven extraordinaire.
Nov 12, 2006 by Mike Chin
The electric, multi-bladed, axial fan has been around since the late 1800s, starting originally as a ceiling fan for well-to-do households in the U.S. Today, they are visibly ubiquitous as air-blowing devices for human comfort all over the world, especially in warmer climates and seasons. Electric axial fans are also employed in almost every sector of industry, mostly for gaseous transportation and cooling ventilation of machinery. Their first application in the personal computer was to provide cooling airflow in the power supply. Now, axial fans can be found on every type of computer component.
Not a reproduction of an early ceiling fan. This one is inspired by Wright… Wilbur and Orville. (Photo courtesy of www.ceilingfan.com)
The body of information about electric fans is voluminous. With a history that extends back to the first practical use of electricity, the scientific and engineering knowledge about electric fans seems endless. There is no desire or advantage in trying to reproduce or reference all of this extensive information. The focus in this article is on those aspects of DC axial fans that are particularly relevant for quiet operation in cooling computer components.
A simple way to think about an axial fan is that it is the marriage of a impeller (or propeller) and an electric motor. The impeller has blades extending from a hub that is spun by the motor. The electric motor has a shaft with a wire armature electromagnet that connects to the impeller’s hub. A rotary switch or commutator reverses the direction of the electric current twice every cycle, to flow through the armature. The electromagnets push and pull against permanent magnets on the motor housing. As the poles of the armature electromagnet pass the poles of the permanent magnets, the commutator reverses the polarity of the armature electromagnet. During that instant of switching polarity, inertia keeps the motor going in the proper direction.
Two key mechanical parts in a DC axial fan are critical to its acoustics: The bearings that hold the shaft in place, and the design of the blades that interact with the air. Other factors include structural aspects (such as rigidity, dynamic balance, and the internal self-damping characteristics of the materials the fan is made of) and the commutator frequency. All of these factors will be touched upon in this article.
I. Ball vs. Sleeve
The main types of bearings used in axial fans are ball bearings and sleeve bearings. When choosing between them, the engineer usually considers the following factors shown in the table below (from Ball vs. Sleeve: A Comparison In Bearing Performance (pdf) by Melody Williams of NMB Technologies):
|Criteria||Ball Bearing||Sleeve Bearing|
|Fan Longevity||Longer life||Shorter life|
|Fan Mounting Options||Vertical, Shaft Center Line Parallel, Perpendicular||Vertical|
|Noise Emission||Quieter at High Speeds||Quieter at Low Speeds in Early Life|
|Lubricant||Less Evaporation||More Evaporation|
|Cost||More Expensive Per Unit||Less Expensive Per Unit|
Ms. Williams favors the ball bearing for many reasons, the primary being long term reliability. For silent computing enthusiasts, the issue is not so cut and dry. Key highlights are cited from the comparison article.
“Temperature: The chart below uses the L10 method to illustrate fan performance at various temperatures. L10 refers to the point of time at which 90 percent of a large population of these fan types will continue to work. On average, a fan that operates for 50,000 hours will run continuously for over five years.
|Temperature||Ball-Bearing Fan||Sleeve-Bearing Fan||% Difference|
|95K Hours||80K Hours||18%|
|75K Hours||52K Hours||44%|
|63K Hours||40K Hours||58%|
|54K Hours||30K Hours||80%|
“Fan Mounting: Fan orientation does not affect the longevity of ball-bearing fans because of preloading, the procedure by which manufacturers build an initial side load, such as a spring or a wave washer against the balls of the bearing. When mounted in vertical positions, sleeve-bearing fans can maintain life spans comparable to their ball-bearing counterparts. However, when sleeve-bearing fans are mounted in any position other than vertical, the fan’s life span decreases.
“Noise: Sleeve-bearing fans run more quietly than ball-bearing fans. This is true for applications that have very low fan speeds. The noise advantage of sleeve bearings is much reduced in applications that require faster fan velocities.“Parts: Typically, sleeve bearings deteriorate under high temperatures because they are made from porous, powered metals from a sintering process. Also, sleeve-bearing fans can develop a high micro-hardness that makes secondary machining difficult. Steel parts make ball-bearing fans more exact. They are precision-ground and super-finished.
“Lubrication: Sealed-for-life ball-bearing systems use thicker lubricants that have more additives, and are less subject to evaporation. While the lubricants within sleeve-bearing fans have a greater concentration of oil, the sleeve-bearings’ bushings can only hold a fixed amount of lubricant. Since there is no periodic recharging of the oil, the lubrication within a sleeve-bearing system is more likely to evaporate.
“Point of Contact: Sleeve bearings are line-contact bearings; there is broad contact between the shaft and bearing that generates a good deal of friction. Ball bearings are point-contact bearings, which generate minimal friction. Previously, opponents of ball-bearing fans argued that the ball-bearing system could lead to brinelling or indentations in the raceway. Yet, defenders of the ball-bearing system believe that if a ball-bearing system is assembled correctly, each component will fit perfectly, eliminating the potential for parts damage.”
Comair-Rotron, another fan manufacturer, has a somewhat different perspective on the issue of ball versus sleeve:
“Most ball bearing fans are noisier by 1 to 3 dBA over their counterpart sleeve-bearing fan. Also the additional noise is somewhat pure tone in nature. Therefore, the annoyance level is considerably higher than with the sleeve-bearing fan. This higher noise level is also in the higher frequency ranges, which makes it even more annoying.
“Sleeve bearing fans, generally speaking, can easily sustain multiple shocks of 80 g’s with duration of 11 msec without impacting noise at all. This is not true for ball bearing fans. Figure 3 shows what can happen to ball bearing fan noise if the fan is subjected to 40 g’s (11 msec duration). This is a very important factor since the equipment manufacturer has no control over how this equipment is treated after the fan is installed, particularly in shipment. It is quite common for a ball bearing fan to be noisy before it is even used just from the handling of the equipment it is installed in.
“Typically, sleeve bearing fan noise does not increase due to life. This remains true up until the system begins to fail due to loss of oil. However, ball bearing fans can begin to get noisy in a very short time. This increase in noise is due to many facts, such as grease channeling, loss of grease, damaged bearing camouflaged by the grease, etc. Also, as time goes on, the grease may begin to dry out which allows for a very noisy fan, but it will continue to run for a long time. This brings up an interesting point: the reason for the use of ball bearing fans is to extend the fan life past sleeve bearings. However, if usable life were defined to end when the fan became noisy, it is quite possible the sleeve bearing fan would out live the ball bearing one.”
SPCR’s Take on Sleeve vs. Ball
The advantages of ball bearings for general application in
computer equipment are clearly spelled out in the NMB comparison. For silent computing, the advantages of ball bearing fans are only relevant if the acoustics at low and very low speed are very good. The advantages of sleeve bearings are certainly relevant for silent computing. The Comair-Rotron article shows why PC silencers might prefer sleeve bearing fans. They are…
The caveats against sleeve bearing fans are…
II. Modified Sleeve Bearings
Many manufacturers have developed variants of sleeve bearings that try to combine the positive aspects of sleeve bearing fans along with merits of ball bearing fans. The better known of these different bearing types include Hydro Wave by Panaflo (now part of NMB), Hypro by Adda, Sintec by EBM/Papst, and most recently, Fluid Dynamic by Sony. The Hydro Wave and Sintec bearing types go back a number of years; the FD bearing is relatively new for fans, although it has been used in hard drives since around 2001.
While precise details differ, and engineers from different camps would probably debate relative merits with some conviction, the broad goals of these modified sleeve bearing technologies is to…
Details of how these bearing types work are contained in company-sourced documents, which, as you would expect, contain strong partisan opinions. Nevertheless, they are well worth studying if you want to learn more about fan bearing design and the impact they have on fan acoustics.
“HWB addresses the traditional weakness of sleeve bearings… axial friction. By utilizing a unique system, the thrust plate “floats” on a circulating film of oil, which greatly reduces the axial friction and the bearing’s deterioration. The HWB pattern on the bearing virtually eliminates thrust plate contact which is depicted above. As the shaft rotates it acts as a pump constantly circulating the oil in the radial, as well as the axial direction. The sealed system and reservoirs of oil maintain the proper lubrication to minimize mechanical contact and prolong the fan’s life expectancy. The radial direction analysis shown above is essentially the same as a traditional sleeve bearing. Panasonic’s unique HWB reduces the axial contact.”
Fig. 1: Lubricant circuit in Sintec bearing
“In fact, the single most important factor affecting the reliability of a fan is the composition and reliability of the lubrication system and not the fatigue life of the bearing itself… Papst has conducted comparative life tests which show that fans with particular sleeve bearing systems can now achieve similar reliability to those with ball bearings. i.e. an L10 lifetime of 80,000 hours of operation at an ambient temperature of 40°C.
“Behind this improvement in sleeve bearings lies significant developments in sintering technology, in particular the unique Sintec process used by Papst. This process uses a special metal powder that is pressed into the required shape at very high pressure and then sintered at high temperature. The material created in this way is porous, with the pore volume taking up about 15 to 30% of the bearing. The pores are then filled with lubricant using a vacuum soaking process and a re-circulating reservoir is created to ensure continuing operation over long periods.
Fig 2: The compact Sintec bearing guarantees excellent alignment and a large oil reservoir.
“This unique design of Sintec bearing ensures that lubrication is maintained even when the fan is at rest, so that the bearing is never dry, even during the crucial startup phase. At rest the capillary effect forms a film of lubricant between the shaft and bearing so that lubricant never drains away from the bearing. Then when the fan starts to rotate, a slight pressure difference causes a hydrodynamic pressure wave to be set up in the bearing gap as a result of the rotary movement of the shaft causing the lubricant to circulate around the bearing. This hydrodynamic state produces a lubricating bulge at the narrowest part of the bearing gap, exactly where friction should be greatest, ensuring that the shaft and bearing no longer touch and lubricant is always supplied to exactly where it is needed.
“To ensure re-circulation, the oil pressure that is greatest at the narrowest place of the bearing gap forces the lubricant into the pores of the sintered bearing. To compensate for this, oil flows out from the sintered metal in areas of lower pressure and circulates to the areas of high pressure where it is needed. This re-circulating circuit creates a stable state as in Fig 1.
“To allow the Sintec bearing to take up the radial load of the fans, two bearing positions with sufficient axial spacing are required. The ideal solution is to use double sintered bearings with two separate bearing positions in a single sintered part. The main advantages are improved alignment and a considerably larger oil reservoir due to the larger bearing volume.
“The design also features an additional axial bearing disc which ensures that a stable position is reached in both axial directions. By taking up the axial lateral forces in either direction, fans can be mounted with the shaft in a non-horizontal plane. Under this condition a film of lubricant is formed between the rounded end of the shaft and the disc and a state is achieved similar to that normally seen in a radial bearing.”
Fluid Dynamic Bearing (PDF document, linked directly from Jaro Thermal) There appears to be a handful of brands offering fans that employ the FDB bearing, and they generally identify Sony as the maker or inventor of the bearing itself. No information about FDB could be found on any of Sony’s English-language sites. The fan brands we found that use FDB bearings are Jaro Thermal, Arctic Cooling, and Scythe. It’s clear that Scythe does not manufacture its own fans. Whether Arctic Cooling also employs OEMs is not clear. However, it is clear that Jaro does manufacture its own fans; it’s possible that Jaro may be the actual manufacturer of all the FDB fans mentioned above.
The text below is quoted from the web site Green Supply Line:
“The FDB fan’s unique design eliminates surface contact between the shaft and bearing, which translates into lower noise, higher reliability, and a longer operating life. The fan also uses a patented “vacuum impregnating oil” that eliminates air bubbles and oil leakage, which means no oil leakage due to changes in air pressure.
“The FDB bearings are manufactured with an extremely thin, herringbone-shaped groove on the interior wall to ensure a constant oil layer. This provides an exceptionally low noise level as low as 19.7 dB for a 92 x 92 x 25 mm fan, said the company. Life expectancy is guaranteed for 80,000 hours at 25 degrees C. Click the link for a look at the FDB bearing’s structure.
“The life of this model is achieved by a V-shaped groove that generates the dynamic fluid pressure on the inner surface of a sintered metal sleeve allowing dynamic pressure to build up in the center of the shaft, said Dennis Eisen, vice president of sales for Jaro Thermal. In addition, the less vibration caused by the bearing results in the lower noise, he said.
“The FDB fans may replace conventional dual ball-bearing systems, offering a higher maximum operating temperature up to 90°C compared with 70°C for fans with dual-ball bearing systems. The operating temperature range is -10°C to 90°C. The FDB fans are RoHS compliant.”
The following table was extracted from Jaro’s brochure, which suggests that the FDB is the most advanced and best performing of all fan bearings.
Jaro’s FDB-promotional comparison.
Note: 1B1S = “1 ball, 1 sleeve” bearing; 2BB = two ball bearing
This is the end of our look at fan bearings for now, but the list of fan bearing types does not stop here. There are different types of ball bearing fans, such as dual and single ball bearing fans, mixed sleeve / ball bearing fans, ceramic ball bearing fans, magnetic levitation bearing fans, and so on. Our experience has taught us to be wary of so-called revolutionary designs, and most ball bearing fan variants don’t sound as good as the best sleeve and modified sleeve bearing fans. When we run across quiet fans using other types of bearings, we’ll tell you about those bearings and update this article.
FAN BLADE DESIGN
Looking into propeller, impeller, or fan blade design can sink you quickly into aerodynamics, physics and other disciplines that are arcane and mind-boggling for the average person, certainly for this writer. The topic of spinning axial blades has received intense, prolonged study for over a century, in aviation (propellers for airplanes and helicopters, turbines in jet engines, etc.) and in marine-related applications — just think of the money and intellectual resources lavished on the development of silent propellers for nuclear submarines alone. Not having as much military value, it’s not likely that blade design for axial fans has received quite as much attention, but the technology does go back over 120 years.
Elsewhere, we’ve mentioned the difficulty of trying to correlate aspects of blade design to practical results. That comment bears repetition. The best we can really do is to try and identify the various elements of blade design. Changes in any one of these elements can have an impact on airflow / noise performance, and there are endless variations.
Much thanks go to long time SPCR supporter Neil Blanchard, for creating the following illustrations. Here, Neil compares the variations in the design of two relatively quiet fans.
On the intake side, the blades on both fans are swept forward, but to different degrees; the NMB’s sweep further forward. The blade area on the NMB is greater, and the pitch (angle of attack) of the leading edge is lower. The Nexus has a larger radius on the leading tip, and the edges of the blades are more blunt. The edge of the hub on the Nexus has a larger radius.
On the exhaust side, the frames are quite different; the struts on the NMB are rounder, but they are nearly parallel to the trailing edges. The struts on the Nexus, on the other hand, intersect with the tailing edges only at one point, which would seem to reduce air turbulence. Also, the strut with the wires is quite wide on the NMB, which probably creates a larger “shadow” in the air flow.
An article by Comair-Rotron on Fan Acoustic Noise identifies several primary causes of noise in fans, including some related directly to the blades:
“Vortex Shedding Broad band noise generated by air separation from the blade surface and trailing edge. It can be controlled somewhat by good blade profile design, proper pitch angle and notched or serrated trailing blade edges.
“Turbulence Turbulence is created in the airflow stream itself. It contributes to broad band noise. Inlet and Outlet disturbances, sharp edges and bends will cause increased turbulence and noise.
“Speed The effect of rotational speed on noise can best be seen through one of the fan laws:
- dB1 = dB2 + 50 log10 (RPM1 / RPM2)
Speed is an obvious major contributor to fan noise. For instance, if the speed of a fan is reduced by 20%, the dB level will be reduced by 5 dB.”
“The stator motion is a square wave that is switched on and off before and after the peak torque position. This motion causes a small amount of undulation in motor torque, producing an audible noise caused by the lower frequency commutation operation. Each small torque causes a minute contracting of the entire fan structure and results in an audible clicking noise while the fan is operating.”
This noise is most often heard as a rapid clicking or buzzing. JMC’s solution to commutator switching noise is to put the frequency up to 25KHz, where it is inaudible to human beings. This is the main feature of their PWM fans.
“Stator excitation is a square wave that is switched ON 45° before peak torque position, and switched OFF 45° after peak torque position. This excitation causes a small amount of ripple in motor torque at the frequency of commutation. Each small torque “burst” causes a minute flexing of the entire fan structure, and results in a faint (but audible) “ticking” noise while the BDC fan is operating (Figure 2A). Acoustic PWM noise is generated in exactly the same way. When the PWM pulse turns on, a step change in torque occurs within the fan, the profile of which matches the rise time of the PWM pulse (Figure 2B). This impulse torque is articulated by the fan structure as audible noise. This is true mostly in larger fans (i.e., fans with operating currents in excess of 300 mA), since they generate a greater amount of torque and have larger size and mass. This effect is more pronounced at low operating speeds (i.e., low PWM duty cycle): the lower the PWM duty cycle, the greater the percentage of time the fan is OFF (quiet), and the more noticeable the acoustic noise caused by the PWM becomes.”
TelCom Semiconductor’s solution to PWM induced fan noise is to “slow the slew rate of the PWM drive signal to the fan”. In essence, the edges of the square wave pulses are rounded, and the voltage spike reduced. As a result, the acoustic click or spike is dramatically reduced.
The broad transition in CPU heatsinks from fans under 80mm size around the year 2000 to fans as large as 120mm by 2005 was driven mostly by the need for greater cooling capability without further escalating noise, which had reached absurdly high levels in “performance” PCs. 120mm (and now larger) fans also appeared as PC case ventilation fans during that period. These developments have helped to lay the foundations for a much broader realization of quiet computers.
All the factors discussed thus far have been internal to the fan itself. There are external factors that also impinge on fan acoustics. The major ones are noted here:
Fan Load (Electrical) Noise varies as the system load varies. This variation is unpredictable and fan dependent.
Impedance refers to vent openings that may not be large enough to allow 100% of the fan’s airflow, restrictive dust or protection grills, or very densely packed heatsink fins that represent high resistance to airflow. Fans make more noise with greater impedance.
Altitude could also be said to be a part of the load. When the air is thinner, airflow is reduced for a given RPM.
Vibration-induced noise may originate from the fan, but can be exacerbated and amplified by resonant panels to which the fan is mounted. Mechanical isolation or decoupled mounting techniques are often required for the quietest fan operation.
It is difficult to sum up all of the various complexities around DC axial fan design and implementation into an intelligent paragraph or two; nonetheless, it’s a task expected of the author of an article such as this. Keeping it simple to avoid revealing too many of the holes in my grasp of the topic, I would suggest that the Ideal Silent Fan has the following characteristics:
Any fan makers who have products that come even close, please contact me asap. 😉
* * *
* * *
* * *
POSTSCRIPT, Nov. 13, 2006
** Dorothy Bradbury ** on “What determines the rated speed of a fan?”
During the usual round of proof-reading and editing that takes place for each article at SPCR, Russ Kinder, a senior member of the editorial team, asked…
I was only sure that it’s not a resistor in series or parallel at the input voltage; if it was, the power ratings of all the models in a fan series would be the same, and they are not. Beyond that, I could only think it would have to be the commutation frequency because making and stocking different types of armature windings for all the fans that manufacturers make would be a complete headache. But that was only a conjecture.
I had the opportunity in the last couple of hours (ah, the wonders of the web!) to exchange a couple of emails with Dorothy Bradbury, the British fan expert mentioned in our first 80x25mm fan roundup. She runs the web site http://www.dorothybradbury.co.uk/, subtitled “SILENT & INDUSTRIAL FANS”. I decided to forward the comments from Russ to her. Her thorough response came back promptly, in somewhat shorthand form.
I’ve posted Ms. Bradbury’s response in the original form except for minor [identified] clarifications to ensure I don’t introduce any misinterpretations. Thank you, Ms. Bradbury! And, Russ, you will be happy to finally have your answer.
Dorothy Bradbury’s answers to Russ’s questions:
What makes one fan spin at 2000 rpm’s and another at 4000 when the same voltage is applied?
I’ve seen more than one person in the SPCR forums suggest erroneously that the difference between say a [Panaflo] H1A and an L1A is only a resistor hidden somewhere inside the hub.
* * *
Dorothy went a couple steps further, and posed and answered two of her own questions to further clarify the above.
So how does an AC fan work?
What do you mean by DC fans with motor-ICs use phasing?
* * *