Filter for PWM controllers

kaange · Post by **kaange** » Thu Dec 21, 2006 7:28 pm

Given the comments about PWM fan controllers making many fans noisy, has anyone tried filtering the PWM signal to reduce the stepping to a DC ripple?

In theory, it should just be a matter of placing a small resistor (say 5 ohm) between the controller and the fan and then placing a capacitor across the positive and ground on the fan side - obviously the value of the capacitor would have to be determined by experiment to find how much ripple needs to be removed for quiet operation and it could well be that different fans will need different capacitance to stabilise them enough for quiet operation.

I'm asking because their seem to be many nice fan controllers that operate by PWM.

Greg

kater · Post by **kater** » Fri Dec 22, 2006 1:54 am

ppl have used attenuators and claim them to be ve helpful - you might want to search the forums around here as this was discussed

Bluefront · Post by **Bluefront** » Fri Dec 22, 2006 1:55 am

M-Cubed makes a device called an Attenuator......for this purpose. Has solved every PWM noise problem I ever tried it on.

kaange · Post by **kaange** » Fri Dec 22, 2006 5:53 am

I did search before posting

and saw discussion about the m-cubed attenuators but I was wondering if anyone had tried a simple low pass filter like I described. I'm interested in simplicity and minimal cost - buying a bunch of the m-cubed attenuators may well work but it adds a considerable amount to the cost of a PWM controller.

Greg

kater · Post by **kater** » Fri Dec 22, 2006 5:58 am

do some reverse engineering if you feel up to it

i mean, go see an attenuator, find some technical specs in the web, see what it is actually and it will probably turn out to be some stuff you can put together at home with parts you bought for the cost 2 beers

EDIT - oh, my new sig works, great

kaange · Post by **kaange** » Fri Dec 22, 2006 6:17 am

I'd have to private import one to see it in the plastic and the web shots don't show any details.

I guess I'll just buy one of the PWM controllers and do some experiments of my own.

Thanks for you er... help

kater · Post by **kater** » Fri Dec 22, 2006 6:33 am

a few links

http://www.williamson-labs.com/attenuator.htm

http://www.fmsystems-inc.com/eng_tee.htm

http://en.wikipedia.org/wiki/Attenuator

http://www.performance-pcs.com/catalog/ ... 80814f3485

i'd also ax bluefront to make some pics of his attenuators. personally, i don't think i'd manage to produce one, but you sound like you know your way around in electroncis so it probably won't be a huge problem - hopefully

Felger Carbon · Post by **Felger Carbon** » Fri Dec 22, 2006 8:44 am

kaange wrote:Given the comments about PWM fan controllers making many fans noisy, has anyone tried filtering the PWM signal to reduce the stepping to a DC ripple? In theory, it should just be a matter of placing a small resistor (say 5 ohm) between the controller and the fan and then placing a capacitor across the positive and ground on the fan side.

Greg, if the PWM signal were coming from a voltage source, your scheme would work with the appropriate value of R and C. However, the PWM source is an open-collector PNP transistor or its FET equivalent, meaning it pulls electrons up but does not pull electrons down (as a voltage source would). This means that the PWM signal would pull the R-C to nearly full voltage and keep it there.

kaange · Post by **kaange** » Fri Dec 22, 2006 3:32 pm

But wouldn't the effect be similar to a voltage source since the PWM current output is stopped when the transistor switches off? (I know that transistors act as current sources and the output is placed across a resistor to provide an output voltage) The resistor upstream of the capacitor would allow for a voltage difference between the transistor and the capacitor so when there is current flow, the capacitor charges up, reducing the voltage seen by the fan while when there is no current flow, the capacitor would discharge while maintaining voltage to keep current flowing to the fan, smoothing the operation (hopefully).

Otherwise, energy would be created somehow, wouldn't it?

Greg

Bluefront · Post by **Bluefront** » Fri Dec 22, 2006 3:32 pm

You won't see anything much by looking at them. It's sort-of a plastic piece with one fan connector going in, and two coming out. You can hook up two fans to each attenuator. They're not that expensive, considering how well they work.

Ah the price we pay for quiet stuff...

kaange · Post by **kaange** » Fri Dec 22, 2006 3:38 pm

kater wrote:a few links

http://www.williamson-labs.com/attenuator.htm

http://www.fmsystems-inc.com/eng_tee.htm

http://en.wikipedia.org/wiki/Attenuator

What do voltage dividers have to do with my query?

kater wrote:http://www.performance-pcs.com/catalog/ ... 80814f3485

Isn't this the same thing as the m-cubed one?

Greg

kater · Post by **kater** » Sat Dec 23, 2006 6:19 am

greg, the few links i pasted in my reply were to help you - either by providing you with at least some info on attenuators or by giving you ready access to other solutions or just by showing that it's pretty easy to find all kinds of info just by googling for it - especially that you said you don't feel like buying a couple of them attenuators as this will be pretty costly
you sound like you know about a zillion times more about electronics than me and you could take advantage of the links - but i think i was mistaken, oh well, happens to me more often than i'd like that
the last link i provided just shows that ther's not only mcubed on the market and that's more than probable that you can find a thingy called attenuators pretty much everywhere, be it us, europe or the beatiful land of australia
all in all i tried to share what little i could and give any kind of help - that's what forums mean to me - lotsa of folks trying to help each other
ok, that's all, i'm done, over & out & merry christmas

Qwertyiopisme · Post by **Qwertyiopisme** » Sat Dec 23, 2006 10:22 am

If the transistor in the PWM controller has an issue with back-voltage applied to it, how about having something like this? (you will need to increase the delivered voltage as the diode has a voltage drop and the resistor will take a bit of power).

http://img381.imageshack.us/my.php?image=temp1ic0.png

where:
the leftmost resistor gets rid of any residual voltage when the controller goes into its LOW state

the diode stops voltage from being applied over the controller while in LOW state

the rightmost resistor limits the current draw from the controller

the cap stores the energy

jaganath · Post by **jaganath** » Sat Dec 23, 2006 12:47 pm

ok, that's all, i'm done, over & out & merry christmas

Merry Christmas kater!

kaange · Post by **kaange** » Sat Dec 23, 2006 2:28 pm

Qwertyiopisme wrote:If the transistor in the PWM controller has an issue with back-voltage applied to it, how about having something like this? (you will need to increase the delivered voltage as the diode has a voltage drop and the resistor will take a bit of power).

http://img381.imageshack.us/my.php?image=temp1ic0.png

where:
the leftmost resistor gets rid of any residual voltage when the controller goes into its LOW state

the diode stops voltage from being applied over the controller while in LOW state

the rightmost resistor limits the current draw from the controller

the cap stores the energy

Yep, that was a circuit that I had in mind as a further option (I'd prefer to keep it as simple as possible but this is close enough). In fact, I was thinking that having a diode in series would be a good thing just to ensure that no damage could occur to the PWM controller.

Greg

cpemma · Post by **cpemma** » Tue Dec 26, 2006 2:57 pm

Another way proposed by Micrel (who make the MIC502 fan-control IC) is to soften the switching action making the PWM signal less of a square wave. It does work but means modifying the controller rather than the wiring to the fan.

]Micrel paper.

kaange · Post by **kaange** » Tue Dec 26, 2006 3:13 pm

The only problem is that the output transistor is then disapating the voltage across it while there is current flow so it will heat up far more than with on/off switching. This is why rheostat controllers have heatsinks across the output transistors while the PWM ones don't. In fact the effect of the Micrel mod is to convert the controller from PWM to a rheostat-like controller.

Plus the surface mounted components make it hard to mod most of the controllers directly.

Greg

cpemma · Post by **cpemma** » Tue Dec 26, 2006 5:06 pm

kaange wrote:In fact the effect of the Micrel mod is to convert the controller from PWM to a rheostat-like controller.

It's far too mild an effect for that, see the Fig 4c traces. I've tried it, the transistor doesn't get noticeably warmer.

Maybe not easy to do on a commercial controller, but worth considering if you build your own.

kaange · Post by **kaange** » Wed Dec 27, 2006 5:59 am

Ah yes, I see - it just smooths the falloff. I wonder how effective this is for PC fans? If such a small amount of smoothing is enough to prevent clicking noise, then in the circuit I propose, only a small amount of capacitance would be required.

Experimentation beckons! I'd better choose a PWM controller

Greg

MikeHunt79 · Post by **MikeHunt79** » Tue Apr 03, 2007 2:01 pm

Would something like a 1uF cap be ok to smooth things out, or would more components be needed?

disco stu · Post by **disco stu** » Wed Apr 11, 2007 5:52 am

I have had the same problem as the topicstarter and I did some research on it. If you use a capacitor, the problem is that it will make average voltage higher then intended. The bigger the capacitor, the smoother the output line will become, but also the higher the average voltage will become.
So if you adjust your PWM controller to make the fan work at e.g. 50% (6V) the actual average voltage can be e.g. 9V (75%). If the capacitor is bigger, it's possible that the average becomes 12V/100% regardless in which position you turn the adjust button of the PWM controller.
This is because the capacitor smoothens voltage. What you do need is an inductor (a coil) which smoothens current instead of voltage. This way you will be able to smoothen the square outputline and thus eliminate noise of the PWM signal in fan, and also maintain full rpm adjustment control of the fan.

cmthomson · Post by **cmthomson** » Wed Apr 11, 2007 5:53 pm

This whole discussion misses a vital point.

It is not the short-duration square waves of PWM that causes the fans to click. The frequency of these main pulses is very high, typically 10-20KHz.

Early PWM controllers put out only this high-frequency variable-width square wave. This worked fine to control the fan speed with high efficiency, but it screwed up the tachometer (speed) readings. This is because the tach needs an "on" voltage for at least a quarter turn (longer for some fans).

To fix this problem, modern PWM controllers put out occasional long full-voltage pulses (several per second) each about one rotation time in duration. This is what causes the clicking: these full-voltage pulses cause the fan motor to accelerate abruptly, which distorts the frame and makes it click.

A capacitor won't have any effect on this periodic acceleration unless it is very large. But then the voltage would be way off.

disco stu · Post by **disco stu** » Thu Apr 12, 2007 10:52 am

Frequency tested on the Noise Isolator fancontroller: 100Hz.
Can you specify "several per seconds"? I assume it's not going to be very much, because you can here every separate click of the clikcing sound. With this information it seams that also a coil (inductor) will not help to solve the problem (or it has to be huge)...
Shortly PWM-controllers are just nog good.

Or you have to make them yourself and live with the fact that you don't have tachosingal.

cpemma · Post by **cpemma** » Thu Apr 12, 2007 12:58 pm

I've just assembled a Velleman kit PWM controller which has given some interesting results. The circuit allows PWM frequency to be adjusted from under 50Hz to around 6.5kHz. A 120mm Akasa Amber produced typical PWM growl on low speed at low frequency; increasing the frequency resulted in an increasing pitch whistle, but above a critical frequency silence reigned (apart from wind noise).

This was not a frequency above my ears' audio range; I think the motor inductance value is the significant factor, producing a filter effect.

The SG3525 looks a handy (and cheap) IC and the Velleman circuit can be greatly simplified for PC fan control. Experiments continue.

disco stu · Post by **disco stu** » Thu Apr 12, 2007 2:12 pm

Does the Velleman IC / SG3525 IC use the long full voltage pulses as cmthomson described?
If so: what is the frequency of it comparing to retail PWM fancontrollers?
If not so: then there's just filtering a normal high frequency current by the fans coil, but this does not give more information about how to filter the low frequency long full voltages pulses.

cpemma · Post by **cpemma** » Thu Apr 12, 2007 3:21 pm

cmthomson wrote:To fix this problem, modern PWM controllers put out occasional long full-voltage pulses (several per second) each about one rotation time in duration. This is what causes the clicking: these full-voltage pulses cause the fan motor to accelerate abruptly, which distorts the frame and makes it click.

Maxim wrote:PWM modulation is simple, inexpensive, and efficient, but it applies a series of pulses to the fan's power supply. Accompanying each PWM pulse is a slight physical perturbation of the fan assembly, which can cause an increase in fan noise, usually in the form of a "clicking" sound at the PWM frequency. Depending on the fan's design, this noise may be nearly inaudible, or may be moderately loud and distracting to users.

I've found type and amount of noise to be fan-dependent rather than controller-dependent. And given how I think the speed sensing works, I can't see how occasional 'long pulses' will allow the firmware to give an accurate speed reading. Do you have a link with info?

cmthomson · Post by **cmthomson** » Fri Apr 13, 2007 5:52 pm

Going by memory, sorry no link.

My current system is typical of modern PWM motherboards. My CPU fan is a Nexus, controlled by the motherboard via SpeedFan. It puts out a faint clicking sound of roughly 3 per second.

The "growling" referred to by other posts would be related to the PWM primarly frequency being quite low. With more recent PWM controllers, the main frequency is very high, and does not produce any audible noise in the fan.

Filter for PWM controllers

Filter for PWM controllers

Re: Filter for PWM controllers