My first home PC, an 8086 clone, only had one fan (in the power supply) but it was NOISY. I used a very quiet 286 at work, so hoped things could be improved. After under-volting the fan with a couple of diodes and opening up the psu inlet grille, things were very much improved.
These pages are intended as guidance for anyone else who wants to calm down their PC fans. I'm aiming at users who have a smattering of electrical/electronic knowledge, who can use a soldering iron, and who'd prefer to build something themselves rather than buy a ready-made solution.
I also live in the UK, so suppliers and costs are UK-based where applicable.
The chart shows how much (or little) noise various fans make. All the data has been culled from the manufacturers' specs. and is for the common 80mm square x 25mm deep 12-volt units.
One obvious feature is that noise goes up with fan flow rating. If you want a gale, it will howl!
Secondly, comparing like flow rates, there's not all that much difference between different brands, though Papst do come out as consistently quieter than the rest across the full range. (As they also cost up to 4 times as much as the others it's good to see some justification.) And YS-Tech's reputation as noisy looks well-deserved.
Fans with sleeve bearings are lower-noise than a comparable ball-bearing fan but one common penalty is a shorter service life, though the better-quality sintered metal sleeves, like Papst's Sintec® bearings, don't suffer this drawback.
Another important point is that the noise figure you see quoted is measured with the fan suspended in the middle of an empty chamber; the noise you actually hear is with the fan fastened to a metal box, a grille of sorts on one end, and various cables and components in the flow path.
This always increases the fan's noise level, sometimes more than doubling the perceived effect. Try removing a noisy fan from your case and running it hand-held – the difference may surprise you! But you can take steps to reduce this added noise back towards the "ideal" level, using the methods to be described.
Finally, I've not found any genuinely "low-noise" fan designs available, only low-flow fans, quiet by virtue of the low output alone. By that I mean no maker produces a "low-noise" model with the same flow as their "standard" model. State of the art is such that the main difference between a cheap generic fan and a more expensive one is quality – better bearings, better finished mouldings, which have a minor impact on as-new noise levels but may well help keep the level consistent for longer.
That said, some fan models, like the Papst range in the above chart, do have an edge – for the latest favourites see the recommended list at SPCR.
So, what does the seeker after a quiet system do?
For example, a fan outputs 37cfm with 12 volts. Noise is quoted at 30dBA, and you want 20dBA. Following the chart, flow at that noise level will be about 24 cfm.
So 37:12 = 24:V, where V is the new voltage. V = 12 x 24 / 37 = 7.8 volts.
Most 12v fans will start every time with 7 or more volts supply, some are even OK down at 5v, but check first. If your calculation shows you need below 7v a better solution would be to start with a less powerful fan.
A few fans will not tolerate low supply voltage – fans with built-in electronics for speed control (by temperature sensor or other means) need the full voltage. Fans that send a speed signal to monitoring hardware should always be connected such that the fan ground wire goes directly to 0v, and with some BIOS there may be problems at low RPM..
The graph shows the air flow v pressure for a Papst Variofan®, which supplies more power to the motor as the temperature rises. So the set of curves from 30ºC to 50ºC can represent voltage curves from about 6V to 12V.
In free air at full power the fan airflow is about 45m3/h. However, the action of blowing air through a case produces a back-pressure represented by a nest of curves in the blue area, the path of each determined by the case restriction. With a 'low' resistance case the flow drops to about 38m3/h but a 'high' resistance case drops it to nearer 26m3/h.
Reducing the fan's voltage (and speed) will reduce the pressure created, but the similar performance hit is dropping the flow from 23m3/h to 15m3/h on the 30ºC line
So how big is a Pascal? If you tried to use a "U" tube water manometer to measure a pressure of 1Pa over atmospheric pressure, the water level difference would be only 0.1mm. Blowing into a sealed case, at 12V the fan can create a pressure of about 17Pa, but only around 7Pa at the low setting.
So it's very important to ensure free air flow through a case when low-voltage fans are used, or their efficiency drops drastically and you're producing noise but little cooling.
With multiple fans the situation gets more complicated as they are each contributing to the case pressure and affecting air flow. When under-volting, I think it helps to have a fan on the case front sucking air in for every fan on the back blowing air out – the front fan creates a small pressure in the case, the back fan reduces it. The end flow will only be about the average of the two fans (not the sum) but each fan will be nearer to its free-air rating than if they were both on the same side, when the actual flow from each fan could easily be less than half the spec. value. And the noise apparent from two or more low-revving fans is usually less than from a single similar fan on full-power.
Adding up the noise from several sources is complicated, as it's not a linear relationship. The formula's given here but for an example, look at this chart.
(There's also a little program written by Khaled Gaben, Korndog from the Overclockers.com forums, does the maths for you, download here.)
Using two identical fans increases the fan noise by 3dBA over the single fan, but adding a third increases it by a bit less than another 3dBA.
So taking a fan with a flow of 45cfm and a noise spec of 37dBA, halving the flow should bring the noise down to about 20dBA. Adding another similar fan restores the original flow (in a free-flow case) but the noise only rises to 23dBA.
(You're not getting something for nothing - halving the voltage of a fan also roughly halves the current, so gives only a quarter the power - two fans at half-speed are, together, half the power of a single fan at full speed, so we can expect the pair to be quieter.)
But you'll also see from this table that if the overall noise from your system is already say 38dBA, changing the odd item to a low-noise version may make only a minor difference. You have to find the worst offender and tackle that first.
The next sections cover Ways & Means.