PC motherboard: PWM 4 pins to 3 Pins digital converter.

[Kiriakos-GR]

It must be on my luck ... the latest motherboard that I got looks unable to control the speed of any Non PWM  fan .

And so I must ask your help .

From my current searches on the net , found this interesting link  
an prologue at the 4 pin fans technology.
http://www.pavouk.org/hw/fan/en_fan4wire.html

And found also someone who made an circuitry , that he said that it works.
Another user looks to had confirm that, and another few was debating about it.
Because it uses the negative line to control the output voltage .
( picture at the bottom )




Two companies they had build this in to  (PCB) solution, and its marketed but for a price,
I like to make my own..  with your help ..

First (not wide)  marketed solution.
http://www.paqt.co.uk/docs/PaQ_PWM.pdf

Second .. wide marketed solution .. ( Nanoxia PWMX 3-Pin to 4-Pin PWM Converter )
http://www.svc.com/pwmx.html


Schematic found ...

  

[Mechatrommer]

just buy the fan Kiri. its not that expensive compared to if you blown out your MB. even cheaper, just buy the non PWM version, just connect to 12V-gnd, no hassle, just $3 or so. i dont understand your circuit, sorry.

[Kiriakos-GR]

i dont understand your circuit, sorry.

This circuit gets PWM pulses , and converts them to DC , and so the output drives the common fan,
and the fan speed  changes.

About the fan, I got the best ever 140mm for my cooler , but the damn motherboard, does not look to have speed control for both,  4 pin fans & 3 pins fans .
 
Here is the white paper , about PWM
http://www.formfactors.org/developer/specs/4_Wire_PWM_Spec.pdf
 

[fsleeman]

You might want to check out this thread: https://www.eevblog.com/forum/index.php?topic=1366.0

I was trying to use a 555 timer to generate a PWM signal for a 4-wire CPU fan. I was able to generate the required signal but then realized I had a 3-wire fan instead! Although nobody proposed a solution for converting the 555's PWM signal to true DC voltage, the fan works even if accepting a PWM signal. The PWM frequency might need to be a lot lower than the 25-30k those fans usually uses depending on the physical properties of the fan and its electronics. Let me know if you find a simple PWM to DC solution, 555 or not.

[Time]

This seems like a strange way to do it but I don't see why it wouldnt work.  Changing the duty cycle would change the speed.

I might get corrected when I say this but I think this is a type of rudimentary chopper circuit.

Why not just breadboard it and find out if it works?

[scrat]

I don't know if I'm right at understanding what you need.
So please correct me if I'm wrong.
For a 4-wire fan, you have a 12V source which gets modulated (inside the fan) by the PWM signal (TTL).
Now, since old 3-wire fans were linearly regulated, you just have to convert your PWM signal into a proportional 0-12V source.

If until here all I said is right, the problem has at least two simple solutions, but since they seem to trivial to me, and you are all but stupid, I assume I have to be wrong somewhere.
However,
1) You can filter the PWM signal and go to the base of an npn BJT (emitter to ground), and linearly regulate the 12V supply (fan between +12V and collector). Speed sense feedback will provide control by the motherboard, even if transistor characteristic isn't linear. I'd put a
2) You can either put PWM signal directly to the gate of an NMOS (source to ground) and drive the fan in PWM, which shouldn't be a problem.

In both cases (but especially for the second), a back connected diode (cathode towards +12V supply) in parallel to the fan would be needed.
Open collector PWM output from the motherboard means you have to provide a pull-up from the 12V line (a resistor + 3.3V zener?).

I'd prefer the second solution, I think that motor phases inductance will do necessary filtering, although some LC could be needed.

[scrat]

I was making the same "mistake" as the circuit it was posted above (image): the sense signal is not referred to ground, which will cause wrong speed detection.

With two transistors (a small one and another a bit larger), it should work better.

[Kiriakos-GR]

Dear scrat  .. I am not an electronics engineer , so to use your generic info about making something.

I have post enough links at the first post, so to even see what I am looking for .
And not just that , the working circuitry of it, its pictured on this PDF ( Link also at my first message )
http://www.paqt.co.uk/docs/PaQ_PWM.pdf

All that I am asking are an clear simple schematic , with common to find parts , that will work.

In about 50 threads worldwide , all those people they stack in theory, and no one produces an simple working idea , like the schematic  posted above , or like the one pictured on the PDF .

That is four transistors , and few resistors.

If some one has the knowledge and the free time , to make something usable , he will become an public hero in the EEV and world wide .

If there is no new ideas on the table , I will had to try out the first schematic , and see if it works.

 

[Time]

in other words do it for him and don't worry about explaining it

[cyberfish]

First of all aren't your electrolytic capacitors backwards?

And you have a short from PWM input to ground through the base and emitter of the transistor.

[scrat]

Yesterday it was a long day, woke up at 5 and posted after midnight... and sorry for being long-winded as usual.

My idea was quite simple, I doubted it was too simple, indeed, and so I thought I was missing something.
With a few adjustments the posted schematic should work, if you have the patience to wait for me to look at it later. Unfortunately I don't have a fan by hand and the time to give it a try, but if "theory" is worth, it has to work. It could be even simpler, as I'm reading now about the tacho signal on the white paper, which I missed last night.

[Kiriakos-GR]

in other words do it for him and don't worry about explaining it

The correct description are ... do it for him , before Dave does it , and blog about it..  
Millions of young ones, that are exploring the computers hardware , have an major need for an practical solution, about this matter .

And I am included ..  as young one ...      

And, at the end , I will have just an spinning fan , and you .. the fame .  

[oPossum]

L1 and C1 may not be necessary for some fans.

[Kiriakos-GR]

Is that tested , or just an fresh idea  ? 

[oPossum]

Not tested. Just an idea. May require some tweaking.

[PeterG]

I have not bothered with the WPM controller my MB has. I simply use the fan from the heatsink connected to the 12v power with a diode in series. The diode drops the voltage just enough to silence the fan without loosing too much performance.

Its quick, easy, reliable and not over complicated.

Just my 2 cents.

Regards

[Zero999]

Not tested. Just an idea. May require some tweaking.

It's a buck converter, nothing new. The frequency, duty and inductance (not optional) need to be optimised for the load. There's no feedback so the output voltage will change as the load current changes.

[Mechatrommer]

Not tested. Just an idea. May require some tweaking.

this is the problem i think. even the 1st person has tested it and working, the second guy might not makes it work and start to blame and complaint, who knows, he may used different type part and fan? as opossum stated, tweaking is required, and to increase the success rate of the tweak.... experience is required

[Kiriakos-GR]

experience is required

Yes I agree, but from the part of the designer ...
He must be the experienced .

So to design an PWM demodulator ( 25KHz ) to DC  5 - 12 V .

And the funny part are , that this it has be made all ready , from some people in UK , and in China too.
 
And So, they are the experienced and capable , and all the rest on this planet are good,
only to just criticize or offer cheap rands.

And so ,  Shafri, save all those comments , or get to work with it ..

Every modern motherboard , has an PWM controller build in .
Every INTEL  fan  with four cables ,  has build in an demodulator .

I could probably  destroy the housing of an INTEL fan, so to get the PWM PCB , and connect just the wires that drives the motor it self , with to the cables of the New 3 Pin fan ...   ( but I hate the idea of destroying an good fan , just for that. )
      
And about the single - simple demodulator s ,  there is an portion of risk , as their output becomes like an switching power supply ,  that if it works out of control , it can possibly  fry one expensive  25$  140mm fan .

  

[jahonen]

If that buck converter design shorts out the switch transistor Q2, then voltage at the output is same than the supply voltage. Or if it becomes open, then output voltage seen by the fan is 0 volts. Either failure mode should be safe from fan perspective (of course assuming that fan can handle supply voltage).

Regards,
Janne

[oPossum]

All that I am asking are an clear simple schematic , with common to find parts , that will work.

That is what I posted. It will work. The uncertainty is how well it will work.

If there is no new ideas on the table , I will had to try out the first schematic , and see if it works.

That circuit is very poor design. Switching the ground lead of the fan will prevent the tach from working properly. There is no smoothing, just chopping. Using a 78L05 to bias the transistor is quite absurd - a resistor divider from the 12 v supply would be adequate.

It is unclear if the second design is supposed to be linear or chopper. There are no values on the cap and resistors. The high gain of the NPN prevents it from working well as linear, and the cap on the base prevents it from working well as a chopper. Will the fan tach work with a chopped supply?

If a circuit works for someone, that does not mean it will work for you or that it is a good design. Asking for something that has been built and tested is rather pointless. Reasonable design is far more important than an ugly hack that works for someone.

[Mechatrommer]

I'm looking forward to see an electrician with "thick hard" skin to get mess around with delicate electronics

[Kiriakos-GR]

If that buck converter design shorts out the switch transistor Q2, then voltage at the output is same than the supply voltage. Or if it becomes open, then output voltage seen by the fan is 0 volts. Either failure mode should be safe from fan perspective (of course assuming that fan can handle supply voltage).

Regards,
Janne

Thanks for the info ..   I feel a bit better by knowing this.

[Kiriakos-GR]

Is there any prediction , about the consumed energy of this buck converter ?

The on-board  fan header , looks in specs with:
12V 0.20A  ( 1700rpm) 90mm INTEL fan.
Or 12V 0.20A (1000rpm)  140mm.
 

 

[scrat]

It is unclear if the second design is supposed to be linear or chopper. There are no values on the cap and resistors. The high gain of the NPN prevents it from working well as linear, and the cap on the base prevents it from working well as a chopper. Will the fan tach work with a chopped supply?

I must admit my fast "design" has those big issues. And I was not considering the need for a tacho and phase-switching logic power supply, which can't be chopped. I was thinking at the motor windings only, which inductance is usually considered to be sufficiently high to filter out switching. The too high gain of NPN BJT can be avoided by putting a proper emitter resistance.

Your buck circuit is too much, in my opinion, I'd rather make it linear, avoiding the use of inductances. Let me make some calculation and a simulation to validate my circuit with the necessary modifications. Could this be accepted as a proof?

[scrat]

Kiriakos defied me (and the matter was very simple), so I couldn't remain indifferent

Yesterday I had some spare time to try a very simple circuit made with what I had by hand.
Since I was trying to make something that didn't require tuning, instead of struggling with transistors I used an opamp (one of the cheapest and widely spread, LM358).
The circuit is a very simple power amplifier, the opamp drives the npn BJT to translate the 0-3.3V reference into a 0-12 V output with a max 300mA capability.
I simulated the PWM input with a potentiometer, while the output was given to a real fan (12V, 0.11 A). It works (of course!), but the output can't swing up to 12V, as expected, it only goes little above 10V. A diode should solve the swing limitation.

I'll make some other tests these days.

[oPossum]

The circuit is a very simple power amplifier, the opamp drives the npn BJT to translate the 0-3.3V reference into a 0-12 V output with a max 300mA capability.

With a 12 volt supply and and 6 volt output with 300 mA draw, the transistor will have to dissipate almost 2 watts - more than 3 times it's absolute max rating.

That is an obvious problem with your "working" circuit. There are more problems.

A diode should solve the swing limitation.

How??? Why not PNP transistor like TIP42?

[oPossum]

Is there any prediction , about the consumed energy of this buck converter ?

The on-board  fan header , looks in specs with:
12V 0.20A  ( 1700rpm) 90mm INTEL fan.
Or 12V 0.20A (1000rpm)  140mm.
 

The circuit itself uses very little current - less thn 10 mA.

[Kiriakos-GR]

Well , just from fear , to not damage the expensive motherboard and fan, I did the move and payed the price,  for this  Nanoxia " PWMX"  PWM controller.

It will arrive in a week or so.
This controller  and the company it self , called as German technology ...  what ever it called it did cost to me allot , because the shipping cost was a bit high, and I had to buy and some other little things , ( fan grills and dust filters ) so the shipping cost to become worthwhile.

Still , I care about the up come of this thread , and I will test the " publicly approved " circuitry  ,
my board has two PWM headers .

There is no need to say , that I am dying to see the inner world of this Nanoxia " PWMX"  PWM controller.
  

[scrat]

With a 12 volt supply and and 6 volt output with 300 mA draw, the transistor will have to dissipate almost 2 watts - more than 3 times it's absolute max rating.

You're right in the fact it can't regulate at 300mA on all the range with that particular BJT.
The fan doesn't draw its rated current at lower speeds (and so lower voltages) because its torque increases with speed, but it could happen that it gets braked (even stuck) and then the BJT can burn.
I was using the components I had by hand, and my fan (as I said above) is rated at 110mA. The transistor wasn't even warm.

An PNP will work better on one hand, but I hadn't one to test. It could have the same (but "reversed") problem, since it could be unable to switch off. One (or two) diode(s) should solve (partially, at least) the problem by adding a voltage threshold to opamp's output (see attached). A rail-to-rail opamp would definitely solve it.

I know it's a really stupid circuit, and I also know the buck converter will be much better (best efficiency, smaller components). Maybe I'm going to try that too, but I was looking for a simple solution that anyone could make without any tweaking.

@Kiri: this last move lets me down... Don't you rely on us?
Joking aside: can you tell us how much is this German robbery?

EDIT: the resistance above the diode has a "random" value, it must be reasoned a little.

[Kiriakos-GR]

This German robbery  cost about 6 - 7.50 EUR In Germany ..

Funny enough , I found the same part in Cyprus retailed at Just 4 EUR !!  
http://www.eshopcy.com.cy/show.phtml?id=PER.807855

But there was no easy shipping to Greece as available option. ( I had contact them )
The same shop has base also in Greece, but they do not cooperate.
I had to use only an Courier service , that would cost an crazy amount of money (Cyprus to Greece) .  

And so I got it for 6.88 EUR   (plus some other stuff) = 7 EUR shipping  ( 30 Total )  From the mother Germany.   
 

[scrat]

@Kiri: Funny... And then we complain about pollution, costs, Far East competition... If transportation is such a crazy thing...

@oPossum: I made a bit of reasoning, and discovered that the diode there is not feasible So PNP is needed. I think the diode should work for PNP, without the resistor, to avoid rail-to-rail opamp.

[oPossum]

A simple linear circuit.  Not built or tested.

[oPossum]

@oPossum: I made a bit of reasoning, and discovered that the diode there is not feasible So PNP is needed. I think the diode should work for PNP, without the resistor, to avoid rail-to-rail opamp.

A simple voltage divider will work well. For example 1k from base to emitter, and 3.3k from base to op-amp.

[scrat]

@oPossum: I made a bit of reasoning, and discovered that the diode there is not feasible So PNP is needed. I think the diode should work for PNP, without the resistor, to avoid rail-to-rail opamp.

A simple voltage divider will work well. For example 1k from base to emitter, and 3.3k from base to op-amp.

Should work, too.
Can you please explain which is your "thought process" to design transistor circuits with feedbacks, like the one you posted above? I know it's simple, and I can analyse it, but I won't be that good in producing it myself, so I ask.

[Kiriakos-GR]

Just find an nice document explaining about  PWM signals and fans ...

http://www.addausa.com/ppt/PWM.ppt

And this its an ready solution.
http://www.diodes.com/zetex/_pdfs/literature/pdf/SCMCBR1.pdf



I have to admit , that the PWM control system its a bit complex.
No one can see how complex it is , if he does not study just a bit , the principals about it.  



.

[TheDirty]

I used one of those chips a few years ago, well not the same chip, but an Analog dbCool chip that does the same thing.  It had configuration and reporting over USB to control three fans using PWM with two temp probes and a stall warning buzzer.  It controls standard 3 pin fans over PWM and has a method to check fan speed pulses only during the PWM 'on' pulse.

[sonicj]

Kiriakos,
im still not grasping the original intent of this circuit....

is it:
- the motherboard won't run without a pwm fan attached?
- do you want the motherboard to control fan speed or is a fixed speed adequate?
- do you want to be able to manually control fan speed?

btw, what fan did you purchase? is this to be a case fan or cpu?
cheers!
-sj

[scrat]

Just find an nice document explaining about  PWM signals and fans ...

http://www.addausa.com/ppt/PWM.ppt

And this its an ready solution.
http://www.diodes.com/zetex/_pdfs/literature/pdf/SCMCBR1.pdf

I have to admit , that the PWM control system its a bit complex.
No one can see how complex it is , if he does not study just a bit , the principals about it.

As sonicj says, this last link is somewhat confusing: this chip is designed to make a standalone controller for a fan, not an interface for 4-wire to 3-wire standard motherboard fan.

PWM principal (especially in this application) is simple, you're just partializing voltage to obtain, as an average, a fraction of it. If you see the fan as a DC motor, to regulate its speed you just increment or decrement the width of the slice of 12V you give it. The average voltage will increment or decrement in the same manner. The speed sense is taken from the tacho signal's frequency.

[Kiriakos-GR]

Kiriakos,
im still not grasping the original intent of this circuit....

is it:
- the motherboard won't run without a pwm fan attached?
- do you want the motherboard to control fan speed or is a fixed speed adequate?
- do you want to be able to manually control fan speed?

btw, what fan did you purchase? is this to be a case fan or cpu?
cheers!
-sj

1) The motherboard are capable to run  with  4 or 3 pins fan.
2) With three pins fans , the fan will work always at full speed . ( 12V input no speed variation )
3) The 4 pin fan , will use the PMW  signal ( pin No4 ) , so to adjust the fan speed as needed .

The PWM to DC converter , it will effect the voltage input of the three pin fan , and so it will start to have variable speed .

All motherboard headers have an 200mA DC output , enough current for specific fans , that does not need more than that.
And above I had posted an guide about  fan size VS rpm = 200mA

The top wanted in this wish list ,  its an easy to build PWM to DC converter.
This one will be able to handle ( or just about)  the 200mA  load.

The best ever solution it would be the one, that can supply more current up to 600 - 700mA ,
enough current to run  even the faster 120mm fans  ( 2400 - 2800 ) rpm  .

At the most PWM to DC converters , it looks that the addition of an potentiometer that limits the voltage output to an specific range , its an easy to do .

Well this looks useful , because I can use an monster size Panaflo FBA12G12H  12V 0.6A ,
to operate at lower rpm speed as max , and still to through large amount of silent CMF .  

The design posted by @TheDirty, looks great about the usability factor at list.
Three outputs, plus current amplifiers ( drivers ) that could handle anything, and possibly fans above the 200mA .    
    

Basically , anything that it can handle 800mA to 1A as current output , it would be called ,
the ultimate in specs  PWM to DC converter.
In this scenario , the output transistors  ( alike the 2N3055 ) they could draw amps by be powered,
by molex connectors.
With this design you can run any monster 120mm fan , at full speed with no fear.

My next move : To email Nanoxia about the max mA  output, that their controller are rated for.
 
  
I am aware of , that some of  my requirements, are possible to be made, and some not . 
 

 



    

[Mechatrommer]

2) With three pins fans , the fan will work always at full speed . ( 12V input no speed variation )

makes me confuse... if its constant 12V, why dont just use 2 pins (+ve -ve)? whats the 3rd (sense) pin is good for?
looking back at your OP, you said you want to control non PWM fan? and i ask... a non PWM fan got a sense output? if yes, then will it be compatible with 4pin sense input?

All motherboard headers have an 200mA DC output , enough current for specific fans , that does not need more than that.
The best ever solution it would be the one, that can supply more current up to 600 - 700mA ,
enough current to run  even the faster 120mm fans  ( 2400 - 2800 ) rpm  .

i suggest you stick with the standard, if the MB provide 4pins only and capable of 200mA only, then better stick to it, its not designed that makes the MB/CPU crash i think it means, the CPU will only need such amperage/wattage to do the cooling. earlier you made a thread on saving electricity, now you want to waste it by pushing to 600mA/higher limit. i admit that i've added some more fans inside my CPU (12V constant), but not to cool the CPU further, but to control dust entry and departure point so it will not enter randomly and stuck on top of the MB or other electronic stuffs. i trust the fan that comes with my CPU will do the job better enough.

now you want to make a tweak to the stuff you are not familiar with and expect somebody comes out with bulletproof and proven solution. i think you fight is not worth enough. thats why i still reluctant to struggle to google on better solution for you and keep rambling nonsense'ly, even if i have the capacity and knowledge. not only the electricians have blown things up you know! electronicians too, but more dangerously, in a silent manner, sometime even more difficult to trace the blast point. you adviced me/us something... that i will give it back to you... you said "dont give the advice to kids, they can do some dangerous stunt" sounds something like that. and i know you are not kid, so i'm advicing... with respect, its for your own good. its your MB you are playing with! unless you are willing/ready to scrap that? Cheers

ps: any info here are good though, given the right purpose for it and you really know what you are doing, especially increasing (overclocking) the wattage of the circuitry involved.

[Kiriakos-GR]

2) With three pins fans , the fan will work always at full speed . ( 12V input no speed variation )

makes me confuse... if its constant 12V, why dont just use 2 pins (+ve -ve)? whats the 3rd (sense) pin is good for?
looking back at your OP, you said you want to control non PWM fan? and i ask... a non PWM fan got a sense output? if yes, then will it be compatible with 4pin sense input?

shafri dot not kill your self ... with generic thinking ..
The motherboards PWM  circuitry its an design, designed by INTEL , the  " how-to "  its described on the first post links.
http://www.pavouk.org/hw/fan/en_fan4wire.html

The three pin one , same story ....  + /-/ Sense/  
At the old three pin header (CPU header) ,  this was able to control the DC ..
Today the default  (CPU header) its the PWM , that partially supports the  three pin fans. 

  

All motherboard headers have an 200mA DC output , enough current for specific fans , that does not need more than that.
The best ever solution it would be the one, that can supply more current up to 600 - 700mA ,
enough current to run  even the faster 120mm fans  ( 2400 - 2800 ) rpm  .

i suggest you stick with the standard, if the MB provide 4pins only and capable of 200mA only, then better stick to it, its not designed that makes the MB/CPU crash i think it means, the CPU will only need such amperage/wattage to do the cooling. earlier you made a thread on saving electricity, now you want to waste it by pushing to 600mA/higher limit.

All that I suggested - wished , was the PWM to DC converter to was able to amplify "some how" the output from 200mA to 700mA ...  is this a Sin ?    

Plus , one such capable circuitry, has lesser chances to burned out.

[Kiriakos-GR]

What I plan to try , its to drive with the Nanoxia PWMX an BD135 , if this is able to cooperate, I just got up to my target.



 

[Mechatrommer]

What I plan to try , its to drive with the Nanoxia PWMX an BD135 , if this is able to cooperate, I just got up to my target.

if you'd be more specific, then it will be easier for the rest of the "geeks" i think

[scrat]

What I plan to try , its to drive with the Nanoxia PWMX an BD135 , if this is able to cooperate, I just got up to my target.

if you'd be more specific, then it will be easier for the rest of the "geeks" i think

Maybe it's not so easy to the greeks to be easy for the geeks.
Sorry for the nonsense, perhaps I'm suffering for low glucose level in my blood...

[Kiriakos-GR]

No problem mate ,
I have too much sugar in my blood , and so I drink coffee with out sugar,
and this causes to me , an over excitement    

[scrat]

Ok, after a good dinner one recognizes the madness of previously done things... I was just wandering your reaction (my sentence involved the Greek, but just because of a rhyme reason). I enjoy seeing you're a tongue-in-cheeck (a composite word I never have read, but Wordreference suggests it) one too. And you must be sweet (not that this moves my interest, let's state it!), if you have so much sugar in your blood.
Enjoy a good coffee, but beware of over excitement!

[sonicj]

1) The motherboard are capable to run  with  4 or 3 pins fan.
2) With three pins fans , the fan will work always at full speed . ( 12V input no speed variation )
3) The 4 pin fan , will use the PMW  signal ( pin No4 ) , so to adjust the fan speed as needed .

The PWM to DC converter , it will effect the voltage input of the three pin fan , and so it will start to have variable speed .

...............

ok, i understand now!

that panaflo looks like it has quite the following! i have older media center box with nothing but FBA08A12L's. oldskool 

i recently built a box with a Noctua 120mm case fan. silent computing has come a long way since i last messed with those panaflos! the noctua is truly amasing! the little bit of noise it does make is like nothing you've ever heard before.

i don't have the time to try this, but what about a lm317 with appropriate RC to convert the pwm to dc reference?
-sj

[Mechatrommer]

i don't have the time to try this, but what about a lm317 with appropriate RC to convert the pwm to dc reference?

There is no try, either do or dont to be able to enter "Kiriakos Design Contest", you must provide a sound and "proven to be working" schematics. he dont want anything that involve any explosion.

[PeterG]

The speed of a 3 pin fan header can be adjusted. The mobo bios must be set to manual fan speed and an app like Speedfan used to control the speed of the fans. The third pin is used for tach only. The speed on the 3 pin fans is controlled with voltage rather than pwm.

Regards
 

[Mechatrommer]

The speed of a 3 pin fan header can be adjusted. The mobo bios must be set to manual fan speed and an app like Speedfan used to control the speed of the fans. The third pin is used for tach only. The speed on the 3 pin fans is controlled with voltage rather than pwm.

simple but nice explanation! i was just confused when Kiri said, its 12V constant only and cannot be controlled. now its become clearer that a "voltage averager" can do the job. and oppps! i dont have any schematics... sorry

[scrat]

simple but nice explanation! i was just confused when Kiri said, its 12V constant only and cannot be controlled. now its become clearer that a "voltage averager" can do the job. and oppps! i dont have any schematics... sorry

He was only saying an obvious thing: if you directly connect a 3-wire fan to a 4-wire output the fan will be at 12V supply all the time, since the PWM control isn't connected. They've certainly made the 4-wire connection like that for compatibility reasons.

[PeterG]

CPU fans generally use a 4 pin header, case fans use a 3 pin header. The believe the 4 pin header is used for the cpu because it is easier for the bios to control the fan speed accurately and also detect failure.  Case fans do not require such control. Most 3rd party cpu cooling fans come with adapters when required to connect to either the 3 or 4 pin header. Better quality cpu fans come with a 4pin connector, some with there own fan speed controller.

Regards

[Mechatrommer]

@scrat: maybe its just my ignorance

[scrat]

@scrat: maybe its just my ignorance

Or you're just too angry with Kiriakos
"     "   "    "     "   inattentive, like me.

[Mechatrommer]

Or you're just too angry with Kiriakos
"     "   "    "     "   inattentive, like me.

no dont get me wrong, i like him! he is one of the cool guy with alot to talk ... or complaint
ps: and he can transfer a thread from low energy level, to "high" energy one, didnt you noticed?
ps: if i angry at him, i will simply ignore him and let him blow his MB the sh*t out!

[Kiriakos-GR]

CPU fans generally use a 4 pin header, case fans use a 3 pin header. The believe the 4 pin header is used for the cpu because it is easier for the bios to control the fan speed accurately and also detect failure.

That's True , and I had emailed Gigabyte , asking about the actual RPM scaling against temperature raise.
As it looks that the PWM controller , it will offer an base voltage (min rpm to 700-800 INTEL fan) ,
and the controller it will increase the speed of the fan if the temperature raise above an specific limit.

The reply that I got from Gigabyte was , that if the CPU temp gets more than 40C , the fan will work at full rpm ( 1700) , and if it temp  drops down, the fan rpm will fall back to the base minimum (700-800 ) .

I bet that this was an quick answer ...  If I believe this .. it translates to, that the PWM its an stupid non scalable way about handling speed of motors .  

I know that shafri hates me , because I am an Olympus digital fun boy, all the Canon fun boys are the same..    

 
  

        

[Mechatrommer]

I bet that this was an quick answer ...  If I believe this .. it translates to, that the PWM its an stupid non scalable way about handling speed of motors .  
I know that shafri hates me , because I am an Olympus digital fun boy, all the Canon fun boys are the same..    

no you are utterly wrong about the PWM. makes me thinking, the actual circuitry that you are looking for might resides in the 4pin fan, and most likely the fan is the DC, its just the PWM->DC converter already embedded in side it.

and why you still understand me wrongfully?! i'm a practical man, i dont care what brand it is! i only care what comes out of it, for the sake of humanity (technologically and psychologically)! i'm not a "fun boy"! and i dont care if you are!

ps: I choose Canon because it serves my purpose! I wish i could convert to Nikon, but there is no turning back anymore at least not... for now.
fyi: I'm building TTL remote flash trigger compatible to most popular camera brand (TTL capabled), and i would loved to include the Olympus, its just i dont have the neccesary gear to R&D.
here! if you dont believe me! need to register i think to see the pictures.

[sonicj]

canon is cool for chdk.

[Mechatrommer]

ooo! i didnt notice there is "Canon Hacker's Development Kit"! o well, i will not bother! since my machine is more than twice the price of rigol, if it broke, i'm not sure someone will be able to fix that other than Canon Service Centre. as i said... it serves my purpose already

[scrat]

I'm pretty sure things on that famous fan controller work like this:

- The mother-board decides (in its maternal wisdom) at what speed her son /the fan /has to run (a little bit of verse).
- The speed controller on the motherboard measures fan's speed from the tacho signal: frequency of tacho signal = rotations/sec.
Worst issue here is the motherboard speaks English, while tachos are Mexican, and so they only speak Spanish.
- As in all feedback regulators, if the frequency of tacho signal is lower than the desired spinning rate, PWM duty-cycle will be increased, while it will be decreased in the opposite case.
- The 4-wire fan controller takes the PWM signal and 12V supply. With the 12V supply it feeds the control electronics (minimal, but there is one, with 1 to 3 Hall sensors to detect rotor position). A PWM switcher just executes chopping (slicing) of the 12V supply, just applying the PWM to it, ending to have a 12V square wave with the same shape (duty-cycle at least) as the input PWM coming from the motherboard fan controller. The control electronics on the fan then gives this PWM signal (power) to the motor, connecting in the proper direction 2 of the 3 phases at a time, depending on the rotor position (sensed by Hall sensors, since the rotor has a magnet). The fan is in fact a BLDC synchronous motor.
Hall sensors could be avoided (sensorless BLDC), but I doubt they do it into a computer fan, since it will require using a more complex controller (this is usually done with an MCU).

As I said I'm not sure at 100% of this, as I never opened one of those, but this is the explanation I give myself, and it seems quite straightforward.

[Kiriakos-GR]

Interesting explanation that targets about wide understanding for the masses .

But the true hierarchical pyramid goes differently.

Intel builds the CPU  and the primary motherboard design,
with all the extras that would help the specific CPU design, to be functional and safe .
    
Intel selects the CPU fan specs : rpm / duty cycle / electrical requirements.

Intel sends white papers of the original design to all the motherboard manufacturers,
so to follow the original design unchanged !!  ( For compatibility reasons mostly about the supplied Intel stock cooler and motor )

On this document http://www.formfactors.org/developer/specs/4_Wire_PWM_Spec.pdf
page 14 , its all clear about the base theory.


Result : the PWM circuitry of any motherboard that supports an specific INTEL CPU family are all alike.
a) Motherboards that supports both CPU generations 65nm & 45nm  , are capable probably from Bios,
to supply an different duty cycle info, so the motherboard to correspond correctly , with the two versions  of the INTEL stock coolers. ( different motor specs )

So far I was aware, the fan speed detection wire (3 pin), was there to just trigger an Bios alarm ,
about damaged CPU fan , and nothing more.

( And so dear shafri , its not the motherboards brand that makes any difference about PWM )

About the 3 pin  CPU headers and motors (INTEL CPU coolers) :  
a) Its another INTEL design , that was followed some years back by the motherboard manufacturers,
and had as main target to support again the INTEL CPU coolers.
( this design it was had his own logic (automated control circuitry)  able to be bypassed and controlled also by software , the famous speedfan software reviled that it was an scalability  " min to max rpm " of five steps.

The marriage of  PWM 4 Pin  with the automated control circuit of  3 pin fans :    
There is no such thing , the motherboard manufacturers in order to do that, they need to build both speed control circuits , plus an setting about manually activation from Bios.
By having the PWM one, active by default .
This dual solution gets the cost up , and it starts to not be widely in use any more.

Today : Every computer owner , who cares to add an third party cooler  ( heat-sink fan compo)
needs this  PWM--> DC board , so to be able to use and choose an fan , with out to be limited,
from the current few PWM, aftermarket choices.

There is few high quality 3 pin fans , that they do worth having them.
And there is no equivalent in the PWM marketed ones.  


Additional info about the PWM signal.
It can used to trigger more than one fan , but its one fan, it must be connected on different header.
My motherboard has two PWM headers , and another two 3 Pin type ( case fans)    

( bad news from the German that I ordered my controller, he will had to delay shipping,
because he had no stock of something in the list ... but promised that I will receive an extra gift)
Estimated arrival about 12 of November.    
          

    
  


  

 

[sonicj]

ooo! i didnt notice there is "Canon Hacker's Development Kit"! o well, i will not bother! since my machine is more than twice the price of rigol, if it broke, i'm not sure someone will be able to fix that other than Canon Service Centre. as i said... it serves my purpose already

its only for their P&S cameras afaik. the "firmware" is non-invasive. it runs on top of the stock firmware so chances of killing your camera are nil. if you are into creative photography, i highly suggest giving it try. you can pick up a compatible model for less than $100 on ebay. the motion, intervalometer, and RAW scripts allow for some pretty amazing shots with a consumer P&S.



sorry for the threadjack! at least its about cameras!  carry on!
-sj

[Mechatrommer]

nice workout Kiri! kudos!
@sonicj: thanx 4 d info.
ps: looking at the vidz reminds me of Karl von Moller's, when will his SoE be out? hmmm.

[PeterG]

I am happy with my EOS 500D for now.

Yeah ok , i am a canon fan.

Regards

[Kiriakos-GR]

Yeah ok , i am a canon fan.

How many rpm ?      

[TheDirty]

But the true hierarchical pyramid goes differently.

Intel builds the CPU  and the primary motherboard design,
with all the extras that would help the specific CPU design, to be functional and safe .
    
Intel selects the CPU fan specs : rpm / duty cycle / electrical requirements.

Motherboard manufacturers only uses the basics of the Intel original reference design now, it's not a straight copy.  They definitely do not use the reference design for fan control.  Speedfan constantly has to update itself to handle all the different fan control methods used by different manufacturers.

Many newer motherboards allow both PWM control for 4 pin fans or VDC control for 3 pin fans using the same 4 pin connector.  The Gigabyte can select Auto Mode, PWM mode (4pin), or VDC mode (3pin) from within the BIOS.

[Mechatrommer]

The Gigabyte can select Auto Mode, PWM mode (4pin), or VDC mode (3pin) from within the BIOS.

ahhh! software mod!

[Kiriakos-GR]

The VDC control for 3 pin fans , by just software control, from windows, its not something that great.

Yes I agree that some motherboard manufacturers, does not use the reference design for fan control,
but they do that,  so to use alternative chips = cutting corners & cost.  


Mark , how much it cost ?  just the part list of your above design ?

[TheDirty]

Using alternate fan control methods isn't cutting costs or methods at all.  Some of the fan control on non-Intel motherboards far surpasses that of Intel reference designs.

The design I used uses an Analog ADT7460, looks like they sold this line to ON Semi.  In theory you can connect these directly to the SMBus on your motherboard and talk to it with a driver and you wouldn't need any other interface hardware.
http://www.onsemi.com/PowerSolutions/parametrics.do?id=2126

In my design I have the ADT7460, with a cheap microcontroller talking to it over I2C and I have a FT232RL to allow the PC to talk to it to reprogram the temp/fan speed maps and to get the current fan speed data.  Looks like the ADT7462 is $8 and the FT232RL is $4.50...  with the uC and MOSFETs, I would say it's about $20.  If I was doing it now, I'd use a uC with integrated USB, maybe I'd get rid of the ADT chip all together. 

[Kiriakos-GR]

By just looking the PCB that you had made ,
I had the feeling that this was what I am looking for.
But, by looking in use SMT parts , I lost my interest about replicating this design.

Even so, I feel the need to congrats you, because you got that far.  

[Mechatrommer]

But, by looking in use SMT parts , I lost my interest about replicating this design.

there is always DIP replacement?

[alm]

there is always DIP replacement?

As long as they're old parts, few new parts are brought out in DIP anymore. If you want to use modern parts, it's hard to ignore SMT.

[Mechatrommer]

As long as they're old parts, few new parts are brought out in DIP anymore. If you want to use modern parts, it's hard to ignore SMT.

then you should award Reply #1 as a winner Kiri!

[Kiriakos-GR]

My brain had stuck to this NANOXIA PWMX ..

How its possible, no one to have even open one, and take some pictures !!

 
( I just finished the build of my new quad core computer,
I had to format and all , but now its just perfect " super fast and silent " ,
but I like the idea of having an ace under my sleeve for the Summer time.    )

[TheDirty]

There's a thread on SPCR about this, with a link to this Maxim app note:
http://www.maxim-ic.com/app-notes/index.mvp/id/3149

Plus pictures of the Manoxia opened up:
http://www.silentpcreview.com/forums/viewtopic.php?t=52555


Looks like it's really cheaply made.  Soldering is terrible.

[oPossum]

I think it is rather poor design. The microcontroller controls a crude 4 bit DAC, so no more than 16 speeds - maybe less. The transistors may not share current equally (no emitter resistors). Linear operation is inefficient and will drop about 1 volt at max speed.

[Kiriakos-GR]

Alleluia  ...     speechless ..  Thanks .  

As soon it arrives , I will verify the schematic.  


About the pictures I had some difficulty to find them on the thread..
It is external links, and so here they are...

http://www.xs4all.nl/~joeribl/images/pwmx-front.jpg
http://www.xs4all.nl/~joeribl/images/pwmx-back1.jpg
http://www.xs4all.nl/~joeribl/images/pwmx-back2.jpg

[Kiriakos-GR]

Plus pictures of the Manoxia opened up:
http://www.silentpcreview.com/forums/viewtopic.php?t=52555


Looks like it's really cheaply made.  Soldering is terrible.

Who cares .. 

[sonicj]

wow! thats pretty bad!

[Mechatrommer]

wow! thats pretty bad!

maybe its that hard for him to find a good flux or reflow machine supply (like me ). so need to DIY from what he already has (solder only). the PCB is nice though.

[oPossum]

Build quality is typical for a low cost product.

PCB is thin and slightly warped. It does not appear to be FR4.



Wave soldered (not reflow), flux residue on board.



Some flux on top side too. Pot is smooth and tight.



Connectors are keyed. Heat shrink, but no strain relief.

[Mechatrommer]

Wave soldered (not reflow), flux residue on board.

not hand soldered huh! well ok... thats UGLY! my hands are better than that.

[Kiriakos-GR]

I do not believe this ... even more pictures .  

Can any one predict the output in Amperes  ?  If I use as 12V input from molex  ( so to not stress the fan header )  and the rest cables  normally connected !!  

It looks that there is four output transistors  named as J3.


The PWM reference circuitry ,  have as info , that the output circuitry it must be capable to offer even 2A ,
for 1-2 ms ( starting current ) .

edit:  Those must be called as JFET , found one description about output as 35mA/V .
At 12V X  0.35 = 420mA  its one   By four times = 1680mA   ( yep that's sounds close enough ) 

[oPossum]

Q1 to Q4 are NPN as shown in the schematic. They may be S9013 or MMS9013. There is no standard for marking SOT-23 package components.

Q5 and Q6 carry current to the fan. I estimate they will be warm at 100 mA, hot at 200 mA and dead at 400 mA (continuous @ 6V to fan).

[Kiriakos-GR]

What about replacing those with better ones ...  is there any suggestion.  

I need  600mA  continuous @12v , and 2A peak for 2 milliseconds.
(the size does no matter )

[oPossum]

The fan will draw less current at lower voltage, so it may be OK as-is. Try it and see what happens. If it blows up then mod it.

A CFP (Sziklai) output using a PNP transistor (Q6) in a TO-220 package should be good for 2A or more with a small heat sink.

Modified schematic...

[Kiriakos-GR]

Nice     

In the primary design Q5+Q6 works in parallel .
In the last  the Q5 drivers the larger Q6. 

I will have a look about the BD part specs on my vrt books. 

[oPossum]

I built a circuit to characterize PWM to voltage converters.

Schematic


Source code, schematic, and captured data

The first test showed that the Nanoxia PWMX was not working correctly. The problem was bad solder on the emitter of Q3.
Plot showing Voltage (left axis) vs. PWM % (bottom axis).



After fix. Pot set to max fan speed (full CW).


Pot set to middle of rotation.


Pot set to min fan speed (full CCW).


Pot set to middle voltage.


Comparison. Max (violet), middle rotation (red), middle voltage (green), min (blue).


Comparison of output with no fan (blue) and fan that draws 220 mA @ 12 volts (red).
Noise in the plots is due to fan commutation. There is no noise with no fan connected.

[oPossum]

Tests at 20, 25 and 30 kHz. It measures both the on and off duration of the pulse. Very good.

[Kiriakos-GR]

oPossum,
if you had get in to the "Win a UEi DM391 Multimeter" thread,
I wish with all my hart, to get it  

( its an long shot, but this is my wish )




 

[oPossum]

I built this on a breadboard...



Tested with 220 mA fan. Results...





Note: Max current with Q2 2N4403 or similar is ~300 mA. Use larger Darlington PNP like TIP125 for more current.
Use R5 if fan must run below 20% PWM.

[scrat]

Designed, built, tested, indication of possible modifications. Great job, oPossum!

[Kiriakos-GR]

I can see only one part missing from the last design,  the pot that I will control the final DC output even at 100% PWM .

(Like the PWMX does.)  

( About the 2N4401  i have the equivalent BC337 next to me X2 )   
 

[oPossum]

I built this on a breadboard...



Tested with 220 mA fan. Results...



Fan runs at 25%, but tach does not work until 32%


Increase R1 to move start of curve up (so fan will run at lower PWM %). Decrease R6 to move entire curve down (so max speed is reduced).

[oPossum]

PWM % vs RPM for genuine Intel 4 wire fan supplied with boxed Pentium 4 LGA775

[Kiriakos-GR]

The actual dream or wish , of any one who needs this design.
Are to use an 120mm fan with this circuitry .
The only problem are , that even the best ever 120mm fan it will start to become audible above the 1000-1200rpm .

And and so the  20%PWM  ( start point it should represent the 600rpm ) and the 100% PWM ,
the 1400rpm mark.

I bet that with this design , and the adjustment points R1 & R6 , everything looks possible.

I think that I am going to build it.  

The German that I ordered the PWMX , he did not shipped anything after all this days, with the excuse that he run out of stock about the PWMX , and he waits to restock.
I had kick his **** today ( after of two weeks of patience ) , and put the Paypal on his tail, so to get my payment back.  

[oPossum]

R1 = 820  R6 = 1k






R1 = 820  R6 = 820






R1 = 820  R6 = 680

[oPossum]

Violet:  R1 = 470  R6 = 1k
Red: R1 = 820  R6 = 1k
Green: R1 = 820  R6 = 820
Blue: R1 = 820  R6 = 680

[oPossum]

Adjustable version.

[Kiriakos-GR]

ok,  just did the parts list ...  For the Adjustable last version.

Resistors
R1= 390 ohm
R2 / R4 =  2.7k
R3 = 1K
R5 = 220 ohm
R6 = 560 ohm

Timers pots
R6A / R1A = 500 ohm

Capacitors
C1 = 10 uf (mF)  ( Capacitor electrolytic 16v )  
C2 = 1n  ( ceramic )

TR  
Q1 = 2N4401  
Q2 = TIP42  



I am getting out to get them.. ( home town)
Its best that way , so the original design to be replicated,
with out the use of equivalent parts.  

[TheDirty]

Great work oPossum.

Sadly I doubt you need to buy parts.  Now that the schematic is on the internet you will be able to buy it all as a product in another couple days for $3.50 at dealextreme.

[Kiriakos-GR]

I will handle the marketing in Europe ...            

Many poor souls will find some joy ... but they must get the basics...

I had some parts in my boxes, but spent 2EUR to get the trimmers + transistors + some 1/2Watt resistors.
I did not find the 2N4401 , I will had to use as alternative one, the BC637.


.

[scrat]

Great work oPossum.

Sadly I doubt you need to buy parts.  Now that the schematic is on the internet you will be able to buy it all as a product in another couple days for $3.50 at dealextreme.

The strange thing is why the Nanoxia board, besides the poor production quality, was quite complicated for the task (it used a digital IC and 4 transistors), while not woking too good. Nonetheless, they have their market...

[oPossum]

The strange thing is why the Nanoxia board, besides the poor production quality, was quite complicated for the task (it used a digital IC and 4 transistors), while not woking too good. Nonetheless, they have their market...

German Design 

[Kiriakos-GR]

The strange thing is why the Nanoxia board, besides the poor production quality, was quite complicated for the task (it used a digital IC and 4 transistors), while not woking too good. Nonetheless, they have their market...

German Design  

You deserve to get " your blood back "     ( About Nanoxia )


But, the troublesome solder point on your PWMX , it can not be called as plague !!  
But the fact remains ... those Germans they cooperate with "build for low budget" partners ..
That's a poor corporate decision .    

In few hours, I will assemble the American design ..   

[scrat]

BTW, I'm curious about which IC they used. Was it an MCU?
It seems a case of "making difficult the easy by means of the unuseful" (as my father used to say).

[Kiriakos-GR]

Are you talking about the Nanoxia ?

If so the SOIC 14  it is the name ...  but it looks that there is about five IC's from other makers.
Just the same part ... I do not remember the code , it must be like  CD74XXXXX.   

[oPossum]

I assume it is a microcontroller. I looked at it with a 10x lighted magnifier and didn't see any markings.

SOIC-14 == Small Outline Integrated Circuit with 14 pins (package type, not part #)

[Kiriakos-GR]

Yes , even the pictures does not show ID markings ... I bet that they had soldered it, upside down.


As about me , I have do the half circuitry , soon I will do and the rest.
Placing the parts on PCB , the parts orientation its a small puzzle.
I have  3-4 small PCB's, all at the same dimension , but the length falls sort, so the design to have the best look.
I am still cooking it ..    

[oPossum]

High current mod for Nanoxia PWMX



The higher gain of the CFP provides higher voltage and the ability to user higher current fans (with a heatsink on the PNP transistor).

Before (red) vs. after (blue)

[oPossum]

DIY clone of Nanoxia PWMX




Source code for PIC

[oPossum]

Watch the 12HV609 decode PWM.

[Kiriakos-GR]

Great  work  oPossum ..    ( your suggestions will help many )


I managed to clear the mess with the German seller , and my Nanoxia is on the way.

I had some urgent tasks this week, and I have left my small project unfinished. 

But now with your latest addition  " The moded Nanoxia " , I think that is best to wait and see what will be my final needs.

By the way , why you had use on the moded Nanoxia  an TIP32C  and not TIP42 ? 

[Kiriakos-GR]

The Nanoxia arrived ,  and I start testing it ...

I must admit that it works !!     ( no i did not open it up to check the PCB yet )

I must also to admit , that after truly testing it , I am a bit disappointed .
I was count that this adjusting pot , it will  make the deference that I was looking for ,
so to adjust with 100% PWM ,  the final DC output under my own likes .
( Or better said ,  to be able to adjust the fan speed at an specific low RPM value)

I found one software made by Gigabyte  ( officially not compatible with my motherboard )
called as Easy Tune 6 ...   ( It supposed that my motherboard are compatible with the older version Easy Tune5) ...
Any way  the new works fine for me and has an amazing tool .... I can control the PWM signal of my CPU Fan header  from  10%   to 100 %   ..

And so I got lots of playing around with Nanoxia + The Big fan +  PWM signals .

I am not that high-tech , graphs and Oscilloscope , but I managed to make some pictures that shows how this thing responds ...


My main complain about it , are the scalability of the Pot ( potentiometer ) it self  !!
The adjustment  like  Low - High  are very poor for my taste ...
According my tests   it represents  just  200 rpm change from low to High ..
This poor behavior ( un-moded )  makes it, look like a poor choice !!  
Who cares for just 200 RPM change  ?    

I am happy and sad together ...    


Pictures of my tests  ,  and pictures of the box ( for the collectors out there)  , that actually its not carton, but its made by strong nylon. http://en.wikipedia.org/wiki/Nylon  

  
  

[Kiriakos-GR]

And more info about the scalability issue ....

I call it as issue , because I dislike the results ..

Its probably the motherboard that its not made  " smart " enough ...  

Any way I continued playing with the Gigabyte software and the motherboard PWM controller,
And I did the amazing  , I got step by step from  10 - 100%  so to find  exactly the actual scalability points.

And I did ... ( picture below )

Well the point is that no matter how smart or damn made , the Nanoxia is made,
the scalability of the motherboard it self , are set by this specific way .
With just five fan speeds scaling  ( or five PWM levels of pulses as output).

So far so good .

Well the PWM-X starts to feel hot at the max Fan speed ( Fan specs 12V 600mA ) , I think that now it gets less voltage from the PWMX , and it must use about 500mA .
I did not test it at full speed for more than 40 seconds , and so far I did not damage anything.

I will take a big breath , and think of what I will do tomorrow, with clear mind .



    

 

[fitnessfan]

Adjustable version.

Sorry, but I know that the thread is 3 years old. %-B I need a 4-pin pwm to 3-pin voltage converter. I am new to this.

Could someone help me to make a circuit on curcuitlab? I don't know the parameters for TIP42C. (Can we use TIP42C here)? Please simulate it with DC Sweep.

Click on the image!

[TonyPh]

Here's my effort.



It's basically copied from a Phanteks converter with a few modifications to make it adjustable. This one was put together for a GPU fan header to convert it's PWM to 2x 3 pin PC fans

It basically, I think (I'm new to electronics), just turns the 3.3v or 5v PWM into 12v and uses the capacitor to smooth out the 12v at the fan end, working a bit like a filter.  It's good for around 800mA of connected fans right down to 60mA or less and keeps voltage near 12v quite well, at least until you get to the high current draw anyway.  It uses a BS170 Mosfet and a B772 PNP medium power transistor, along with a few passives. It's the simplest implementation I've seen and costs me less than £1 in parts.

I don't have a schematic at the moment.

[eikei]

fitnessfan, if you still need it, here is the layout of my device that uses oPossum's scheme (see attachment). It was necessary to make it as small as possible and so the PCB dimensions are only 29x13mm. Here's what it looks like (made two units for CPU and case intake fans):



Capacitors and resistors are smd 1206 type. The Ground line should be made on the opposite side of the pcb, or Ground points can be connected with a wire.

The device works fine - tested with some 120mm and 80mm fans. oPossum, thanks for a simple and great scheme!

[hur]

Adjustable version.

I just built this. Works great I am running Two 350mA 120mm 2600rpm fans with it.

I am using TIP42G and 1/4 watt resisters. TIP42 is heating up.

Am I using too many Fans?