PWM in water cooling

[King107s]

I'm in the process of upgrading everything, including my water cooling components.

Has anyone tried using PWM fans on their radiators in water cooled systems? This is of course assuming the only heat source in the loop is your cpu... What I hope is that the high flow fans are only at top speeds when they need to be to cut down on noise and if the PMW fan is being controlled by the cpu load that should work nicely... Only problem I can foresee is not having enough air pressure against the radiator for it to function properly (at lower speeds).

Any thoughts or experience with this?

[Bobnova]

It should work just fine, really.
At low speeds the fans won't put out much pressure, but being closed loop PWM it won't matter. At low temps they'll do low pressure low speed, as the temp goes up so does speed and pressure.

I don't see it being an issue, personally.

[axis01]

The issue you might run into is the amount of amperage that the MB outputs to the PWM fan socket. You could get away with one, maybe two but three is going to push it. I'm thinking of a triple rad for the CPU here.

Axis

[Conumdrum]

Some controllers have PWM ability. So sure, why not? Don't expect the Mobo to like having 3 fans in parellel on the CPU fan connector, I smell smoke. PWM fans can cost more too, so thats another consideration.

[ascl]

If my understanding of the PWM specification is correct, I think you could take the 12v and GND lines directly from the PSU, and attach the sense and control lines to the motherboard.

This would allow you to use the motherboards (bios) fan control without hitting limitations of the power supplied, and then should allow chaining of multiple fans.

If you chained multiple fans you would need to only connect one sense line.

Would love to know if someone has tried this... I think PWM or, at least, software controlled (and hence 'smart') fans are a great idea.

EDIT: PWM spec here (http://www.formfactors.org/developer%5Cspecs%5CREV1_2_Public.pdf)

[Bobnova]

You could always go hardcore and make a PWM controller out of a basic stamp or a parallel port card or somesuch.

[Diggrr]

The PWM control on my little Intel board also controls the voltage of the only case fan connector. Not that that is important here, but other motherboards might do the same.
What I don't like is that my panaflo 92mm (non-PWM) makes strange motor noises when on this circuit.

[Conumdrum]

Pulse width modulation. Thats what a PWM fan does. Try reading up on PWM etc.
http://en.wikipedia.org/wiki/Pulse-width_modulation
So acsl, your not looking at the switching circuts at all. Thats where the PWM happens.

Monitoring it means nothing if you don't power the PWM circuit which is built into the Mobo or controller. You can't just put V to the fan................

[ascl]

My understanding (which could easily be completely incorrect!) is that you supply the fan with 12v, GND, and the control line... and that the 12v is constant but the control line supplies a frequency signal so the PWM circuit knows how to regulate the voltage.

We can ignore the sensing line, thats for monitoring and unimportant for this discussion :)

I am assuming that its the fan itself taht modulates the input voltage... if this wasn't the case, why bother with the 4th pin? the mobo could just supply the modulated voltage?

[Conumdrum]

It's a circuit(chip on the controller/mobo that does all the work. It's like a simple.........ack.

Not worth explaining what PWM is in general terms.

[ascl]

Okay, someone has done exactly what I was suggesting and it has worked:
http://www.silentpcreview.com/forums/viewtopic.php?t=53713

Artic Cooling also have fans that allowing piggy-backing off each other for up to 5 PWM fans from a single mobo connector.

EDIT: I am not sure you are correct about the PWM happening on the motherboard. The motherboard definitely generates a signal (for the control line), but the voltage is modulated on the fan itself. If this wasn't the case, you could just have 3 wires!

I am not trying to be difficult, I just want to understand... I haven't seen anything explanation for why it wont work, and from what I can see, people have done exactly this and it has worked?

[Perseus]

So, these fans would be controlled by the MB? My concern is what input is being used as a regulatory item. If it's the CPU temp the fans would be constantly spooling up and down, especially if you have an i7 with C1E turned on.

[ascl]

Depends on your BIOS. But in the most common case, only the CPU fan header supports PWM fans, and so yes, it would be based on CPU temp... and yes that could get annoying. You could also use SpeedFan to control it, and that allows slow speed changes, but certainly on the SPCR forum a lot of people do not like speeds changing as the change in noise can be more annoying than a constant noise.

[Perseus]

Depends on your BIOS. But in the most common case, only the CPU fan header supports PWM fans, and so yes, it would be based on CPU temp... and yes that could get annoying. You could also use SpeedFan to control it, and that allows slow speed changes, but certainly on the SPCR forum a lot of people do not like speeds changing as the change in noise can be more annoying than a constant noise.

Understood. My radiator fans (not PWM) ramp based on coolant temp, which works fairly well. Finding a PWN controller that will fit a fluid sensor would have the dual benifits of not taking a chance on overworking the CPU fan header and fan speed being based on something that doesn't jump around like a puppy. ;)

[ascl]

What do you use to achieve that? A controller with temp sensor? It sounds like a good option, especially using coolant temp.

[axis01]

What do you use to achieve that? A controller with temp sensor? It sounds like a good option, especially using coolant temp.

That would be his TMS-200. Software based fan and pump speed controller. Nice toy to have but a bit pricey at $100+, depending on the extras.

From looking at it, it appears as though pump regulation is for their Koolance pumps only, although there may be a workaround. Not sure how much amperage each fan header puts out either. 3x fan headers seems pretty low unless you can run multiple fans/header.

Axis

[ascl]

Ah... that looks awesome... and awesomely expensive! :)

[Perseus]

That would be his TMS-200. Software based fan and pump speed controller. Nice toy to have but a bit pricey at $100+, depending on the extras.

Be careful what you recommend, axis. The TMS is a great theory that falls down in practice. I'm one of the few who are lucky enough to have a mostly functional unit, but I wouldn't suggest others buy it, which is why I didn't bring it up directly. If I had it to do over I would have waited for version 2.

[King107s]

Pulse width modulation. Thats what a PWM fan does. Try reading up on PWM etc.
http://en.wikipedia.org/wiki/Pulse-width_modulation
So acsl, your not looking at the switching circuts at all. Thats where the PWM happens.

Monitoring it means nothing if you don't power the PWM circuit which is built into the Mobo or controller. You can't just put V to the fan................


This is my understanding as well...

I'm only using one Delta AFB1212SH-PWM (http://www.newegg.com/Product/Product.aspx?Item=N82E16835213011) which puts out 113.11 CFM @ 46.5 dBA connected to a Black Ice GTX120. I dont care too much about noise but this is about as loud as I can tolerate.

I was going to just replace the stock thermaltake fan with the above fan but I could just use it as a 'pull' fan that always runs at 100% but is not noisy (to maintain a baseline flow rate). Then when the cpu comes under load, the PWM kicks in and the high performance 'push' fan gets the flow up to where it needs to be to handle the extra heat.

Sounds good in practice, but if it doesn't work out well, i can just wire the fan up to a standard molex and run it at full speed all the time.

What do you think?

[ascl]

That delta fan draws 0.53A... which may be pushing the limit of the motherboard's ability to supply power, depending on your motherboard. Of course if you wire up the 12v + GND from a molex you wont have to worry about it.

[King107s]

I'll contact the manufacture and find out what the spec is but that’s only 6.36W... I'm sure some of the high performance air (cpu) coolers (some with two fans) pull close to the same amount and perform just fine on that PWM cpu pin.

"using Pulse Width Modulation (PWM) to automatically adjust fan speed according to the CPU load. The fan can be idled down and thus be much quieter. The PWM chip in the motor allows an exact fan speed control via BIOS."

This implies the signal comes from the main board and the fan motor handles the power. So maybe asd is right, and a 12V/grnd wire job would work the case of a power hungry fan or more than one fan.

Bottom line is, there’s only one way to find out. And that will be Thursday. If anyone comes across any more info please post.

[Bobnova]

The PWM has to be in the motor.
All the voltage regs on motherboards are PWM, hell power supplies are largely PWM now too.
If the motherboard were doing the PWMing you wouldn't need the fourth pin, you can do that with two.

[King107s]

Youre right, there would be no need for a sensing line.:beer:

[Conumdrum]

If my understanding of the PWM specification is correct, I think you could take the 12v and GND lines directly from the PSU, and attach the sense and control lines to the motherboard.

This would allow you to use the motherboards (bios) fan control without hitting limitations of the power supplied, and then should allow chaining of multiple fans.

If you chained multiple fans you would need to only connect one sense line.

Would love to know if someone has tried this... I think PWM or, at least, software controlled (and hence 'smart') fans are a great idea.

EDIT: PWM spec here (http://www.formfactors.org/developer%5Cspecs%5CREV1_2_Public.pdf)

I was wrong, I dug into it more.

This id from an obscure source, it helped me. I'm a electronic tech and taught basic eletronics and radar systems, so I have a clue, kinda.....
.................................
"You will need a function generator to apply the duty cycle and frequency PWM signal. You will also need a frequency counter to count the tachometer signal returned from the fan, or the actual revolutions of the fan. JMC uses an Agilent 34970A data acquisition unit with one 43901-switch module to measure the tachometer, voltage, current, and temperature. We then use an Agilent 33220A function generator to apply the PWM signal."
.....................

So the Mobo monotors speed of the fan via the sensing wire, and modifies the signal to apply the duty cycle and frequency PWM signal. The mobo then uses it's power wires to supply raw current and volts to the fan, the fan manipultaes the PMW switching to turn the fan. Problem is the tiny runs on the Mobo might not like the power the fan demands etc.

So YES!:) I don't see why you can't use a different power input with higher current capability, but be careful that the return line isn't needed for the mobo control circuits.

4 wires.
Tach (speed sensing)
PWM signal
Hot 12VDC
Common (The tach and PWN still need a ground)

Might work!

I jumped the gun and needed to learn more.

[ascl]

Please let us know how it goes! It all seems logical... but nothing like a practical experiment! :)

EDIT: Conumdrum don't sweat it! These forums are for discussion, and you questioning me just made me research it more :) but I am glad we agree now...

[hitokiri_808]

I meant to post in here earlier today, but I forgot to hit submit on my laptop. :bang head

Yes the speed controller is built into PWM fans, and the mobo uses the 4th wire to control the speed. That's probably why PWM fans cost a lot more.

[Bobnova]

It is a pretty big step up in brainpower for the fan controller, makes sense for it to cost a good bit more.

[ascl]

This is whats kind of silly. The fan controller shouldn't *need* brain power! I mean we have an uber powerful CPU just sitting there, and software like speedfan there to control it. All I want is a simple voltage (or PWM) regulator that can be controlled via software (USB or serial or something).

If I had a clue with electronics I'd try and make one... should be very simple! I'm a software guy tho, not a hardware guy.

[Perseus]

This is whats kind of silly. The fan controller shouldn't *need* brain power! I mean we have an uber powerful CPU just sitting there, and software like speedfan there to control it.

You're right. The idea of CPU off-loading is sort of retro considering the sheer power of contemporary CPU's

[ascl]

So we need Conumdrum to hook up one of these "Agilent 33220A" PWM generators to USB (serial is much much simpler, but its kind of dead), and I'll write some software for it...

:)


EDIT:

Hmm, far from being totally insane, this actually doesn't look too difficult. These guys sell exactly the right chip:
http://www.awce.com/pak5pc.htm

It supports up to 8 PWM signals, and you can communicate with it via rs232 (which is serial, but usb -> rs232 exist, as do add-in rs232 cards)... and its $25 USD.

This guy:
http://www.semifluid.com/?p=147

Has gone a step further and programmed the micro controller himself, which reduces the price to $5 USD, again, controlled via rs232.

EDIT2: http://www.awce.com/gp6.htm for $33 its premounted on a circuit board with rs232 connector!

[King107s]

Heh, Agilent makes all of the analytical instrumentation I use at work (I do analytical chemistry for a research and development lab in a flavor company)! They are a Hewlett packard company.

Nice finds asd! Very cool :) I never thought my little idea (not a new one) would spark so much interest.

[Bobnova]

Remember the days of adjusting the fan speeds via speedfan?
That was motherboard based PWM with brainless fans :D

[King107s]

That delta fan draws 0.53A... which may be pushing the limit of the motherboard's ability to supply power, depending on your motherboard. Of course if you wire up the 12v + GND from a molex you wont have to worry about it.

Just got off the phone with Gigabyte and the mainboard is rated to 1amp @ 12V, not bad :) So power should not be an issue for one fan.

[axis01]

Be careful what you recommend, axis. The TMS is a great theory that falls down in practice. I'm one of the few who are lucky enough to have a mostly functional unit, but I wouldn't suggest others buy it, which is why I didn't bring it up directly. If I had it to do over I would have waited for version 2.

Never recommended it, just answering his question as to how you did it. Figured it would be similar to the other software based systems but maybe not. Thanks for your recommendation NOT to use though. :beer:

Axis

[axis01]

Just got off the phone with Gigabyte and the mainboard is rated to 1amp @ 12V, not bad :) So power should not be an issue for one fan.

Make sure you don't have your rear fan or a NB/SB fan connected to the MB too though.

Axis

[billb]

I never thought my little idea (not a new one) would spark so much interest.

Your Idea, basically, controlls fan speed based on CPU load. That won't work well. If you want to control fan speed it should be based on water temperature.

Here's a scenario:
* CPU under load, fans on high, CPU heats up until water reaches an equilibriun temperature.
* CPU idles, fans slow down, water stays warm (or cools very slowly) due to low air flow through the radiator.
# CPU goes back to loaded, fans on high. water is already at, or close to equilibrium, CPU starts out hot.

What is happening is the water is always at, or near, it's max temperature. It never gets to cool down, because every time the CPU idles, the fans slow down.

And thinking a little further, automatic fan speed control in rather unnecessary. A simple water temp monitor (a CompuNurse with it's sensor epoxied into a tee) will show you water temp. A rheostat (or if you want to be fancy, a fan controller) can be used to control fan speed. You'll find that the max water temp really doesn't change all that much (unless the ambient temp changes), and manually controlling is fans is more than adequate to compensate for the changes in ambient temps.

[Conumdrum]

It's just as easy to turn up the fans on a manual $30 or less controller and turn them down when done. Much less compicated. Or just overrad like me, set them on quiet all the time and no worries. And my room temp gets above 80F easy all the time.

[King107s]

Your Idea, basically, controlls fan speed based on CPU load. That won't work well. If you want to control fan speed it should be based on water temperature.

Here's a scenario:
* CPU under load, fans on high, CPU heats up until water reaches an equilibriun temperature.
* CPU idles, fans slow down, water stays warm (or cools very slowly) due to low air flow through the radiator.
# CPU goes back to loaded, fans on high. water is already at, or close to equilibrium, CPU starts out hot.

What is happening is the water is always at, or near, it's max temperature. It never gets to cool down, because every time the CPU idles, the fans slow down.

And thinking a little further, automatic fan speed control in rather unnecessary. A simple water temp monitor (a CompuNurse with it's sensor epoxied into a tee) will show you water temp. A rheostat (or if you want to be fancy, a fan controller) can be used to control fan speed. You'll find that the max water temp really doesn't change all that much (unless the ambient temp changes), and manually controlling is fans is more than adequate to compensate for the changes in ambient temps.

Well, Bill, its a good thing I thought of that way before even purchasing the fan Gigabyte's Easy Tune mainboard software can control any of the main board connected fans logically based on temperature (not cpu load). So.... its no big deal that the cpu load comes off before the water is cooled off because I have it set to ramp based on temp.

BTW, after many annoying build issues (i.e. bolts that were too short, fan leads that were too short, pre drilled holes at the top of the case were too close together and the list goes on...) I got everything put together and the PWM is working beautifully. My system is so much more quiet now and is staying cool

Wish my 4870's weren't so damn loud!

[Perseus]

It's just as easy to turn up the fans on a manual $30 or less controller and turn them down when done.


Actually no, it isn't. If twiddling knobs puts the wind in your hair it's all good, but automatic fan control is a perfectly viable choice. Again, it's all in the eye of the beholder.

[Perseus]

Never recommended it, just answering his question as to how you did it.

Okay, but the next time someone asks you about a TRUE I'll take the question. ;)