FAN FAQ Sticky!

[V12|V12]

Okay folks, as someone (Tempus) previously requested, figured I'd be the starter of a much NEEDED Faq about FANS, one of the MAIN forces in water cooling (unless you've got some exotic underground setup)

Here's my contribution of some VERY nice industrial fans

http://www.comairrotron.com/datasheets.html I have a Whisper XL, MAJOR DC 172mm and a Feather 115Volt monster! They are SERIOUS fans... pretty cheap at depots!

------------------------------------------------------------------------------------
jafb2000
-----------
PWM noise is a function of 3 components:
o Bearing thrust & slippage noise from PWM on/off pulses
o Motor IC noise from PWM "average" voltage re rise/fall
o Harmonics as a result of the PWM frequency

1) Bearing thrust & slippage noise from PWM on/off pulses:
o PWM frequency applies voltage as "1 0 1 0 1 0" re on/off
o So there is a resulting mechanical kick to the bearings
o This will result in some noise from acceleration
o Similarly, some noise from deceleration
o The net result is a mechanical addition to rotation noise

PWM relies on mechanical slippage to work - that the rotor-fan
& magnets keep on turning during the "off" period. Thus the
fan rather than using a continuous analog signal is using a
digital signal whose *average* voltage (like r.m.s to a.c.) is
less that the peak voltage of "1" (12V, 24V, 36V, 48V etc).

2) Motor IC noise from PWM "average" voltage rise/fall
o Motor ICs for all DC fans are self-starting
o Many self-starting circuits trigger at a certain voltage
---- and involve a cut-out + delay + re-start process
o This inevitably results in noise varying by fan & PWM specs
---- varying by fan re form of self-start circuit
---- varying by PWM frequency
---- varying by PWM voltage (not all are Vs re fan rating)
---- varying by PWM duty-cycle re 50-70%
o Motor ICs are also designed for V-continuous not chopped
---- the chopper frequency can affect motor-ic quiet-operation
---- quiet hard drives phase head movement into start/move/stop
---- a similar technique is applied to "new low noise motor ic's"

This is usually one of the loudest noise inducers:
o Because self-start circuit involvement results in mechanical noise
o Which has it's own amplitude, frequency & natural frequency

Vis., some fans perform terribly under PWM.

Taking Papst System 3000 PWM 12V controller & Papst 6212NM:
o This sucks re noise quality at lower PWM settings
o The fan will *growl* at a volume *louder* than it's full-Vs use
o The "PWM growl" negates the benefit of the PWM at <80% duty

Most systems aim to run fans at 70% duty cycle, since that is a
good realisation of cfm & static pressure, whilst at the same time
having a disproportionate benefit in terms of bearing longevity.

Taking Papst System 3000 PWM 24V controller & Papst 6224NM:
o Blissful smooth operation from 50-100% duty cycle
o No PWM noise, but no "halving of dB(A)" at 50% duty cycle
o Reason being that the fans have very large ball-bearings

This feature of very large fans isn't unusual:
o Noise is a component of rpm - and also bearings
o Big fans like Comair Rotron Caravel don't half dB(A) at 50%
---- taking a 115/230vac model 450cfm & 45dB(A)
---- at 50% duty cycle you will not see 22.5dB(A) but 30-35dB(A)

Can't recall the figure off hand as I don't use it, since PWM gets
a different dB(A) for reasons 1-2-3 on that fan. However, on the
very big fans there is a base-level noise due to bearing rumble.

It's small - and not unpleasant - just something that shows up in
the dB(A) figure in absolute levels, and subject to preference.

3) Harmonics as a result of the PWM frequency:
o This can be a biggee
o PWM frequency of 1-2kHz on most fans means run
---- they motor and *amplification thro fan mechanics* will growl
---- specifically they will emit a 1-2kHz noise during PWM
o PWM frequency of 10-20kHz is more preferable
---- and tends to increase the number of fans PWM "works" with
o Some work is being done into ultrasound
---- however noise damage in watts in ultrasound is *tiny*
---- so whilst we can't hear it, it can damage us - so vital research

As a function of this PWM frequency, you have bearing interaction.

Overall, there is a critical path of PWM frequency, motor ic type
& self-start circuitry, bearing type & size, bearing/magnet
mechanical characteristics re slippage, and mechanical slippage
of the overall system. Each varies according to each fan. Some
fans seemingly identical bar voltage *sound* very differently.

This is why PWM implementation in industry ends up specifying
a fan which has been proven - another can be unpredictable,
and why PWM is disliked in that production lacks substitution.

Personally I prefer:
o Proven fan/controller combination = PWM if cost permits
o Variable fan in use/replacement = use 2-voltage operation
---- this requires a consideration of hysterisis re control

Hence for the latter I use 24M1A at 12/24V operation, since for
a 128 fan application the PWM array to control it is >400$US. In
comparison a 12/24V controller using industrial DIN 12V PSU in
either series of parallel (designed to permit it) is cheap, 6$US.

My comments were & are re PWM.
However, the bigger fans do have bigger bearings - whilst the
rpm may be far lower (just 1350rpm or less for 254mm fans), so
more of the base-level-noise can be from the bearing themselves.

This becomes especially true into the bigger (300mm+) EBM fans.

The noise, however, may not be intrusive or annoying.
Some people even like the sound of very big industrial A/C, it is
quite loud - but sufficiently white-noise to be easily masked. A
lot of smaller-scale fan "white-noise" has low-period harmonics
in it and these are more easily selected by human perception.

Within about 7-10yrs fans will integrate PWM as standard, not
as present forms, but completely serial-port programmable; may
only be possible for distributors, but is an industry end-goal. In
terms of temperature profiles, low-temp operation & so on.

Biggest gripe over PWM is quiet enough at low temp operation,
yet enough airflow at high temp operation without jumping there
too fast. This is obviously load dependent - and why a quiet PSU
with PWM control may be quiet on one PC yet noisy on another.
In such cases, the noisy PC often has higher thermal output and
often is using the PSU fan as per the ATX spec & profit-remit to
cool not just the PSU but the entire PC - hence high cfm & noise.
If 38cfm is needed at 35dB(A) for a PSU-PC-cooled solution then
offloading it to 2 fans allows 36cfm at a combined silent 17dB(A).


Nothing wrong with larger fans, but ideally one should redesign
a PC case to house them (or multiple PCs in a box similarly). Very
easy to pick up flight cases on castors, portable, rugged, dense.
Then make the PC structure out of nylon & ABS, or aluminium. A
perforated T3R5 alloy (3mm hole, 5mm pitch) is pegboard PC.
Such solutions avoid obsolescence (AT ATX BTX etc), and can be
very economic if used to house multiple PCs in self-made racks.
Vertical racks allow narrow form-factor, passive 1U heatsinks &
use high airflow from larger fans running as required. Keep the
air resistance low (blow bottom to top, blow-thro-cases) and it
can be very quiet whilst very easily and cheaply made.

BTX cases are a very short-term solution, Intel trying to play
catch-up to the Apple Dual G5 - and quite a clumsy nasty design.
That is case manufacturer/integrator/OEM profits speaking tho.

Yes sound pressure level is logarithmic, not linear.
Fan rpm is linear with voltage, the main determinant in fan noise.

For small fans a small drop in voltage gives a large drop in dB(A).
o Ex 3450rpm-2150rpm, gives 69-43cfm and 43-27dB(A)

For large fans, the bearing system is a major noise component
o Ex 1650rpm-975rpm, gives 450-228cfm and 50-38dB(A)
o For another makers same fan the 50% duty is given as 42dB(A)
---- they sound like crap too, just a technical data point :-)

Bearing noise is not necessarily linear with fan rpm/voltage,
the large fan I cite has sealed IP54 bearings & high-temp ok.

Bearing noise need not be objectionable.
Walk into a big supermarket, or even big old mainframe room
and you'll hear a lot of bearing noise that isn't too obnoxious.
At least in contrast to a PC 60mm Delta fan doing >6800rpm.

For PWM control of fans, the dB(A) figures can go out the window.
The quieter 38dB(A) large fan I cite at 50% voltage, is a dog
with PWM and growls louder than at 100% voltage operation.
The noisier 42dB(A) large fan works fine with PWM - reason is
the kinetic energy of the bearing/rotor/magnet-structure has
far more slippage and negates the PWM mechanical interaction.

Sometime someone should print a PWM-ideal-per-exact-fan test.
Instead many bi-voltage fans, eg, 24/12V switched for hi/lo as
Compaq & many more do (IBM) - far cheaper than PWM solutions.


------------------------------------------------------------------------------------
Okay lets keep the ball rolling on this one folks, it's sort of a very important part of cooling

[pHaestus]

http://www.procooling.com/Links2/Air_Cooling/Fans/

Might be some useful tidbits there

[fhorst]

Fantastic info!
Noise levels are a big issue for me (I guess for all of us)

Some more info on how much sound dbA is:

According to several sheets,
00 - 10 Threshold of hearing
10 - 20 Broadcast & recording studios
20 - 30 Bedroom at night
30 - 40 Library
40 - 45 Small theatre
45 - 50 Dishwasher next room
55 - 60 Large business office
60 - 65 Normal speech at 3 feet
70 - 75 Vacuum cleaner at 10 feet

For cooling we need loads of CFM ;-)
One extra fan (with same level of dbA) will add about 10% to the total dbA

While I was adding more fans to boost the cfm I noticed that the fans start to suck in the center (as they should do), but blow on the sides. (It’s like having the fan lying on a table)

Backpressure from the radiators seems to be quickly too much to handle.

What can we do to overcome this?

[myv65]

Sound and decibels are commonly misunderstood. Doubling sound power (by doubling the quantity of fans) increases sound by 3 dBA. This rule of thumb holds true provided the fans are the same model. It may not seem intuitive, but it is indeed 3 dBA whether going from one fan to two or going from two fans to four. Increasing dBA by ten doubles the perceived noise level at a given frequency. So going from one fan to eight roughly doubles the perceived noise if all are mounted similarly.

[fhorst]

Originally posted by myv65
Doubling sound power (by doubling the quantity of fans) increases sound by 3 dBA. This rule of thumb

You are right about the doubling of the fan's.
I should have written "doubling" instead of adding"

When dealing with low noise fan's (below 20 dbA) the rule does not apply. also with loud fans (above 50dbA) the 3 dbA won't work.

but, what are we talking about going with 20dbA fans from 1 to 8 will give 26.6 dbA, or for the thumb rule 29 dbA.....

[pHaestus]

There was an excel spreadsheet at the overclockers.com forums that would calculate dBA for your system given the fans you input. I couldn't find it with a quick search over there (no surprise). I don't know how accurate it would be anyway as hdd becomes the main noise source when fans dip below a certain point.

[myv65]

The rule always applies to sound power level as this is simple mathematics. It does not always apply to dBA, however, as this is a scale weighted on the human ear and acoustics often change perceived sounds for a given sound power level.

It's no news that the same fan will yield a different perceived sound depending on how and where you place the fan. Same goes for the room containing the computer. Hard floors and no wall coverings make for "noisier" systems.

There are all kinds in this world and there will always be people willing to put up with noise for some other "benefit". Personally I've got my main rig in the living room next to an aquarium. It is difficult to hear anything from the computer other than the optical drives spinning up over the "noise" of the aquarium's filter, which isn't very loud on its own.

It is easy today to create an air-cooled system where the noise of "quiet" hard drives may be heard over the background hum of the cooling equipment. It's only when people think they require temperatures far below the stability point of the processor or when they want to overclock the processor that noisy solutions are common.

[fhorst]

When you are right, you're right!

One quick tip that will help reducing fan noise: use 1200 or higher sandpaper to make the blades smoother. This will help (a bit)

Any idea's on how to overcome the backpressure?

Some I know of: (in order of efficiency, top being most efficient)
(all to be used with a radiator shroud)
- Use a blower
- Sandwich the rad with 2 fan's (push/pull) to stabilize the restriction
- Stack 2 fan's with a shroud (5cm)
- Use your case as a shroud, making sure that you push/pull roughly equally
- Stack 2 fan's
- use one fan (having the rad outside of the case)

[myv65]

The "easy" answer is to use a radiator so big that you do not need much pressure. Obviously this is not always practical, but then again watercooling isn't exactly practical, either. I would opt for isolation as much as possible. ie, a separate case with the radiator and fans that could have copious sound insulation. This also allows the use of big radiators, eliminating the need for much pressure.

The other consideration is a little testing at varying fan voltage. Until you check, you never know how much impact a reduced fan speed will have on your temperatures. Again, unless you're going for that killer overclock you can probably get by with a lot less cooling than most people realize.

[V12|V12]

Fhorst:
----------

I've got a simple solution, if you can do the mod: Simply SWAP the fan and the Rad's position... SO that the fan is Drawing IN air and not trying to force it through... Most fans work best with pulling air, unless you've got got an industrial fan which are most often designed to generate constant CFM under high loads of back pressure.

[WAJ_UK]

here's a link fo db calculation
http://hyperphysics.phy-astr.gsu.edu.../sound/db.html

[fhorst]

V12|V12: Thanks for the tip. I will try that one also.

(got a bunch of fan's laying arround, for some testing)