Circuit to simulate a computer cooling fan?

[tongro]

I have a server computer which contains three cooling fans. The fans' speed is controlled by firmware, but even the lowest speed is very loud, even though all the internal components are always very cool to the touch.

So, I want to find some way to reduce the noise level. When I removed one of the fans, the machine's BIOS objected loudly, refusing to boot. So then I took a spare fan and hacked off the rotor housing and connected it, but this seemed to screw up the feedback signal, and the fan's "failed" light came on and the BIOS objected again. Another option would be to simply cut off the rotor blades, but they are very difficult to get at with ordinary tools.

The next thing I would like to try is to construct a simple circuit which would simulate the presence of a fan. There are three connections; 0V, supply (presumably variable voltage), and feedback. I'm not an electronics engineer, but I do have some experience of amateur electronics and can build a simple circuit from a schematic.

So, can anybody point me in the direction of a suitable circuit or discussion of the technology involved, or provide helpful hints of any description?

Thanks very much.

Tony.

 

[melone]

Are you sure that eliminating one of the fans will allow for adequate cooling of your system? Seems like you are tempting fate on this endeavor.....

 

[VEBill]

Perhaps you can set your BIOS to not check the fan. It makes sense that the BIOS would have to allow for those fans that do not provide the feedback signal.

 

[itsmoked]

For starters the fan's speeds are fed back to the MB for monitoring.

If you remove them IT KNOWS.

If you cut blades off it will probably speed up dramatically causing an Absurd speed error.

Furthermore there is probably no way to cut the blades off in a manner that will not cause severe vibration at the speeds involved.

Most bios's I've seen allow you to control which fans are monitored so you may be able to turn off monitoring for a certain fan and THEN remove it.

Alternatively you may be able to purchase a much lower CFM fan for the fan size you have (40mm 80mm etc etc). Lower CFM equals lower noise. You could swap out all the fans for lower CFM units. That would allow redundancy for when the fan does fail.(which absolutely will occur) Then you would be down to say two fans instead of removing one then being down to one fan.

Do please pay attention to the cooling air flow patterns!!

Removing one fan or stopping it is foolish as then air will come in right there and short circuit right out the neighboring fan causing a very large drop in air flowing through the enclosure. So if you remove a fan BLOCK the opening.

BTW: the bios should be adjustable because silent operation often means removing all fans for water cooling. This scenario is quite common and so the mutherboard's bios MUST usually be adjustable.

 

[IRstuff]

Sure, it's cool NOW...

just wait until summer and there's more dust and less A/C...

TTFN

 

[felixc]

Perhaps you can find more silent fans in the aftermarket?

 

[Warpspeed]

I was once tasked with the problem of designing in some quiet as possible cooling fans. And the best solution I could come up with at the time was to fit temperature controlled variable speed fans.

The idea is that the fans must be placed on the hot exhaust side of the equipment to extract the hot air, and the fan speed will self adjust to maintain some reasonably constant internal temperature.

In winter the fans are barely turning over, on a really hot day they made a fair bit of noise, but that is just too bad... At least the temperature inside the equipment was kept fairly constant, and the airflow adjusts to the minimum required to do the job.

Multiple extra fans will run slower, and fan redundancy takes care of itself too, so it is not a bad system.

Some of those fan monitoring chips are quite smart, and measure fan current, and ripple frequency from the commutator. They are designed to detect almost any imaginable fan problem, including air blockage and are not that easy to fool. Best bet is to get into the software as mentioned previously.

 

[MikeHalloran]

Put the server in a room by itself, where it should be anyway to assure physical security.



Mike Halloran
Pembroke Pines, FL, USA

 

[tongro]

OK, thanks for all those contributions.

I suppose the only solution is to look for quieter fans, but at 90x38 mm, they might not be easy to find.

 

[itsmoked]

90x38 ??!?!?

Indeed.. Check out

http://www.coolerguys.com/

for other quieting possibilities.

I have a colleague who has a server next to his head in his office. He wanted absolute silence so he changed all his rack mount PC supplies to DC input (no fans) and put water cooling on all processors and bridge chips. It is quiet.

 

[tongro]

That's 90x90x38 of course. I wish I could use other options, but there's the pernickety BIOS to overcome first.

 

[felixc]

If you're ready to go to such an extreme, there is an article in Tomshardware where a computer has been stripped off of all of its fans, floppy and CD drive, and has been plunged into a bath of cooking oil, as the cooling means.


The BIOS should allow for a certain range of speed for the fans before crying out. A way to turn the speed down a bit and reduce noise, hopefully, is to insert a resistor, or a couple of power diodes in series with its power input. These fans normally have 3 pins, power, ground, and sense out.

 

[itsmoked]

Is this a custom proprietary server or does it have an off the shelf motherboard in it?

 

[tongro]

felixc:

A resistor - great idea! I'll try various values and see what happens.

itsmoked:

It's a standard HP Proliant; there are two sockets for each fan position, to allow for redundancy and hot-plugging.

 

[lcsjk]

Reducing the voltage with a resistor and or diodes will help reduce the noise.
Search the internet for cooling fans. Look at Nidec. Search the major distributers such as digikey, allied, newark and mouser. There's lots of others.
If the fan blades are too close to the grill or slots, they will make more noise. Sometimes a small separation will help immensly.
All fans are not alike. Some are three blade, some 5 blade and some have different airfoil shapes to the blades. All have different noise levels.
I don't know if your system was designed or just made by "add more fans to keep it cool" However, do not remove the fans, and don't reduce the airflow very much or you will probably be replacing more than just the fans. THe CPU will probably not be your problem.

For simulation: The fan's speed sensor is normally an open collector (or drain) logic signal. You can simulate it with a pulse train at the fan's RPM, one pulse per revolution.

 

[tongro]

I think I'll have to drop the resistor/diode idea; it will reduce speed but also affect feedback which will make the BIOS complain.

I'll get hold of an oscilloscope to find out exactly what the feedback signal looks like and see if I can make a little circuit to produce the same output.

Thanks for all the contributions.

 

[ko99]

With a rheostat (like

http://www.frostytech.com/articleview.cfm?articleID=883)

you should be able to find a setting where the fan is sufficiently quiet before the BIOS considers it "failed".




ko (

http://www.ecooling.biz)

 

[itsmoked]

I can't believe there is nothing commercial for this...

 

[ko99]

The Zalman rheostat I mentioned is made precisely to reduce fan noise

ko (

http://www.ecooling.biz)

 

[itsmoked]

Yes it slows the fan but it doesn't lie to the MB.