TR Forums

I have one of those motherboard with "fake" PWM system fan outputs, ie. they are regular voltage controlled outputs with "PWM fan support".

I thought I'd remedy this in a relatively clutterfree manner, so I bought a couple of these:
0-10V -> PWM converter

I guess I have to divide the 0-12V down to 0-10V with a couple of resistors. The rest, like circuit power, fan power, etc, I should be able to use the 12V straight from the PSU, right?

Is there anything I need to consider?

Assuming my interpretation of the Chingrish specs is correct, yeah I think that would work. No guarantees.

I do have to ask though: Why bother? Yeah, maybe the fan speed curve will be a little more accurate, but then again maybe not.

Another idea is to smooth out the PWM signal itself using capacitors instead of completely avoiding the PWM.

Here's an actual implementation from somebody with more analog knowhow than me: http://hackaday.com/2011/06/16/beginner ... wm-output/

chuckula wrote:

Another idea is to smooth out the PWM signal itself using capacitors instead of completely avoiding the PWM.

Here's an actual implementation from somebody with more analog knowhow than me: http://hackaday.com/2011/06/16/beginner ... wm-output/

Sounds to me like he's trying to do the reverse -- i.e., he has a motherboard fan output with voltage control, and wants to convert it to a PWM output.

(My apologies to the OP if I've got it backwards.)

just brew it! wrote:

chuckula wrote:

Another idea is to smooth out the PWM signal itself using capacitors instead of completely avoiding the PWM.

Here's an actual implementation from somebody with more analog knowhow than me: http://hackaday.com/2011/06/16/beginner ... wm-output/

Sounds to me like he's trying to do the reverse -- i.e., he has a motherboard fan output with voltage control, and wants to convert it to a PWM output.

(My apologies to the OP if I've got it backwards.)

Oh, yeah, nevermind. Need moar covfefe coffee.

just brew it! wrote:

Sounds to me like he's trying to do the reverse -- i.e., he has a motherboard fan output with voltage control, and wants to convert it to a PWM output.

(My apologies to the OP if I've got it backwards.)

You are correct .

The reasons being: a) PWM-fans are not designed for voltage control, and may even take damage from it, and b) you don't risk the fans stopping when the voltage dips too low.

Oh, and c) the fun of it.

FubbHead wrote:

PWM-fans are not designed for voltage control, and may even take damage from it

PWM fans are also designed to take a PWM signal at 25kHz. If the frequency of that converter's PWM is significantly different, I could see that potentially causing problems as well. I doubt physical damage would result, but it may not work as intended, or could cause audible whining if the PWM frequency is in the audio range. So there are potential pitfalls either way.

If you end up using resistors for a voltage divider, check that you're not exceeding their wattage rating, commonly 1/8 or 1/4W.

meerkt wrote:

If you end up using resistors for a voltage divider, check that you're not exceeding their wattage rating, commonly 1/8 or 1/4W.

Since the divider is just to provide the control voltage (not the supply voltage), it should be possible to use relatively high-value resistors. With a 2K ohm and 10K ohm (to provide the necessary divider to scale 0-12V to 0-10V), the combination of the two resistors together would need to dissipate only 12mW.

It would be wise to check the voltage swing of the fan header with the chosen resistors though; it may not behave as expected if it doesn't see a load similar to that of a typical fan. If the output needs to be loaded, then lower resistance (and higher wattage) resistors may be needed to simulate the loading of a fan motor.

You might find this doc useful. The other issue here is that PWM fans expect a 3.3/5V PWM signal, not 12V. There's a pull-up to probably 3.3 in the fan, so this way the mobo can PWM between ground and open with a basic 3.3V I/O and everything Just Works.

This should be mostly solvable by putting another 4:1 voltage divider on the adapter -> fan PWM line, but I'm not entirely comfortable with that solution, because the spec isn't clear enough on whether this stuff is 3.3 or 5V. It looks like it really is designed to have an open high where the fan's pull-up does the work rather than being driven high. To emulate that, maybe just put a diode in the line so that the adapter can pull the fan's PWM pin to ground but not go the other direction. I'd still be more comfortable putting a 12 -> 5V divider ahead of that, not knowing what weirdness may be lurking in cheap fans.

Yeah, if it doesn't draw current there it doesn't matter.

just brew it! wrote:

It would be wise to check the voltage swing of the fan header with the chosen resistors though; it may not behave as expected if it doesn't see a load similar to that of a typical fan. If the output needs to be loaded, then lower resistance (and higher wattage) resistors may be needed to simulate the loading of a fan motor.

With only a cheap multimeter to work with, how would I go about to test this? Measure the output over a small resistor?

synthtel2 wrote:

You might find this doc useful. The other issue here is that PWM fans expect a 3.3/5V PWM signal, not 12V. There's a pull-up to probably 3.3 in the fan, so this way the mobo can PWM between ground and open with a basic 3.3V I/O and everything Just Works.

This should be mostly solvable by putting another 4:1 voltage divider on the adapter -> fan PWM line, but I'm not entirely comfortable with that solution, because the spec isn't clear enough on whether this stuff is 3.3 or 5V. It looks like it really is designed to have an open high where the fan's pull-up does the work rather than being driven high. To emulate that, maybe just put a diode in the line so that the adapter can pull the fan's PWM pin to ground but not go the other direction. I'd still be more comfortable putting a 12 -> 5V divider ahead of that, not knowing what weirdness may be lurking in cheap fans.

Yeah, it doesn't say. As above, how do I measure this (the output voltage) from the converter?

Seeing how it also can be set to 0-5V conversion, maybe that is the PWM voltage it puts out? Then I guess it would be better to divide the input to 5V?

FubbHead wrote:

just brew it! wrote:

It would be wise to check the voltage swing of the fan header with the chosen resistors though; it may not behave as expected if it doesn't see a load similar to that of a typical fan. If the output needs to be loaded, then lower resistance (and higher wattage) resistors may be needed to simulate the loading of a fan motor.

With only a cheap multimeter to work with, how would I go about to test this? Measure the output over a small resistor?

I'd just hook a ~10K ohm resistor between the output and ground, and fiddle with the BIOS settings to change the speed at which it is trying to drive the "fan". If the voltage across the resistor seems to vary in a "reasonable" manner based on the settings and CPU/system temperature, it should be OK.

FubbHead wrote:

synthtel2 wrote:

You might find this doc useful. The other issue here is that PWM fans expect a 3.3/5V PWM signal, not 12V. There's a pull-up to probably 3.3 in the fan, so this way the mobo can PWM between ground and open with a basic 3.3V I/O and everything Just Works.

This should be mostly solvable by putting another 4:1 voltage divider on the adapter -> fan PWM line, but I'm not entirely comfortable with that solution, because the spec isn't clear enough on whether this stuff is 3.3 or 5V. It looks like it really is designed to have an open high where the fan's pull-up does the work rather than being driven high. To emulate that, maybe just put a diode in the line so that the adapter can pull the fan's PWM pin to ground but not go the other direction. I'd still be more comfortable putting a 12 -> 5V divider ahead of that, not knowing what weirdness may be lurking in cheap fans.

Yeah, it doesn't say. As above, how do I measure this (the output voltage) from the converter?

Seeing how it also can be set to 0-5V conversion, maybe that is the PWM voltage it puts out? Then I guess it would be better to divide the input to 5V?

I would assume (just a WAG) that the PWM output pulses will be driven to the positive supply rail, or close to it. Even though you only have a cheap multimeter, you should still be able to measure this. Set the adapter for 100% duty cycle (control voltage at max), and see what the output voltage is. If it's an old-school analog multimeter you can even (sort of) measure the pulse duty cycle once you know what the output is at 100% duty cycle; the inertia of the needle in the meter will average things out for you. Digital units probably have some sort of filtering as well to smooth things out and give an average reading, but that'll depend on the meter design.

I like synthtel2's diode idea; it eliminates the need to worry about the voltage the adapter is putting out, since it'll only pull the PWM line down, allowing the fan's own internal pull-up to handle the positive side of the cycle. Just hook a diode between the PWM output of the adapter and the PWM input of the fan, '-' (cathode) end of the diode towards the adapter, '+' (anode) end going to the fan. Best bet is to use a diode with low forward voltage (e.g. Schottky diode), as this will allow the PWM input of the fan to be driven closer to ground on the negative part of the PWM cycle (ordinary silicon diode could conceivably result in an ambiguous logic level to the fan depending on what the threshold is).

Another thing occurs to me... the adapter may be designed to drive a powered load, not a control input. You may need a pull-down resistor on the PWM output to ensure that the negative part of the cycle goes to ground.

TBH this is really starting to sound like more trouble than it is worth. I do think it is workable, but it may require a fair bit of tinkering to get it right depending on the exact characteristics of the adapter and motherboard fan output.

AFAIK the boards with the "fake" PWM fan headers do have a fully functional 4-pin header for the CPU (just like previous boards, though then only the CPU had a 4-pin and all others were proper 3-pin headers). The problem is none of the other 4-pin headers seem to control speeds properly with either 4 or 3-pin fans. If your CPU header works, you can get the PWM signal from there!

The most expedient but messy looking solution though would be to use a y-adapter to connect all your PWM fans to the CPU header for speed control, and reserve the other headers only for where you can use fixed-speed 3-pin fans, with an external speed control like a fan-mate or even soldering a 20 to 82ohm resistor inline (I use a fan mate to find the value of the resistor). After all not every fan needs speed control.

A crude but effective voltage control would be to simply solder a thermistor inline instead of the resistor, but you'd have to attach it somewhere that gets hot under load because the circuit itself doesn't draw enough power to help get the thermistor hot. You could get all complicated with the thermistor controlling a FET but why bother? High wattage thermistors are cheap.

There are probably enough problems with noisy factory PWM circuits on motherboards already without trying to reinvent the wheel. In general it is far more difficult to find a PWM fan that doesn't tick or click at low speeds than a 3-pin one, so for external fan controller I'd prefer a voltage controlled one with all 3-pin fans.

bfg-9000 wrote:

AFAIK the boards with the "fake" PWM fan headers do have a fully functional 4-pin header for the CPU (just like previous boards, though then only the CPU had a 4-pin and all others were proper 3-pin headers). The problem is none of the other 4-pin headers seem to control speeds properly with either 4 or 3-pin fans. If your CPU header works, you can get the PWM signal from there!

I've done that before, sharing the CPU PWM signal, but I'd rather not have the whole case taking off because the CPU is working.
And I have had the opposite experience; I haven't had one noisy PWM fan yet when using a proper PWM signal. When having them voltage controlled on the other hand...

But sure, most motherboards 1-2 years back seem to use all proper PWM outputs, and hopefully it won't be long before my next upgrade, so the necessity is questionable at best. But at least I gotta try!

It's the midsummer holidays now, so all the measuring and testing will have to wait. But I'll pop in here afterwards, with a success story or (most probable) more questions.

FubbHead wrote:

I've done that before, sharing the CPU PWM signal, but I'd rather not have the whole case taking off because the CPU is working.

Most motherboards I've seen seem to use the CPU temperature to control all of the fan headers, so I'm not sure what your modification accomplishes.

A surprising number of people here have the Abit IP35 Pro, which has about the most customizable fan headers there are. You can pair any of six headers to any of three thermal sensors, and set minimum + maximum speeds (voltages or PWM duty cycle) and temperatures individually for each header. Essentially you can create a custom fan curve, just like with a video BIOS editor, for each header and the only thing lacking is a hysteresis setting.

So if they could do that 10 years ago, it's sad to see newer boards that can't even properly control voltage, only PWM on a single header.

bfg-9000 wrote:

A surprising number of people here have the Abit IP35 Pro, which has about the most customizable fan headers there are.

It was that exact feature, profiled in the June 2007 roundup review here, that led me to purchase an IP35 Pro.

Most of the boards I've seen can cue case fan control off of either CPU temps or chipset temps, but I'm not sure that chipset temps are that much more useful. It's been a while since I've seen one that doesn't allow you to spec out custom fan curves, though some are much easier to use than others. To JBI's point, what you can do is spec different curves for different headers, so the case fans may technically be cued off CPU temp, but their curves can be very flat in the relevant range.

I definitely prefer PWM control when everything is working right, if nothing else for its ability to control fans to lower speeds without stalling, but some cheap PWM fans do seem a bit janky.