manual PWM fan control server HP-ProLiant

[adelage]

Hi all, I hope to be in the right section
I've recently got a beastly ProLiant DL580 g5, quick specs:
4 x E7330 quad 2,4ghz
16GB 667 ECC
2 X 73gb SAS 10000rpm
6 X DELTA PFC1212DE -6J32 12v 3.24A reverse PWM

now, the problem is that the fans are quite loud, at the startup looks like a jet engine is inside my room (kind of 5000rpm or more), but even after they slow down, they keep on being noisy and annoying (I think they are around 2000/2500 rpm). I've asked at the HP forum if there's a way to control them via software, but no answer yet, and even after installing the HP Management Suite (huge bundle of software) I haven't been able to find a solution. At this point I was wondering if I can manually control them through a PWM controller like this (edit:as BruceUSA says below, this solution doesn work) , considering that I have to keep some kind of input (I think the speed wire) to the server otherwise it thinks there are no fans and shuts down "by design"...unless there's a way to disable this feature...
Any Suggestion?

EDIT:UPDATE, SUMMARY OF PROPOSED SOLUTIONS "Work in Progress"

Okay, as the thread is starting to become full of posts (I really appreciate all of the ideas that you guys have proposed, and I'm sorry if I've gone a little bit OT), I list a quick summary of all the informations I've collected until now:

Things to know before approaching to the potential solutions:
- The fans are 5 pin reversed pwm, the 5th pin is a bridge on the ground wire that act as a locked rotor check up
- The system starts at full speed (around 5000rpm) then slows down keeping around 40% of the speed (HWinfo reading), as soon as one of the six fans is removed/faulty the speed rise up to max, with 2 or more fan removed/faulty the system automatically shuts down
- The speed of the fan doesn't go under 1500rpm via manual pwm control (thnks LennyRhys)
- Using a standard voltage control the fans require a kick start of 8v and barely keep the voltage of 5v, acting in a weird way due to a sophisticated DSP speed control embedded inside the fan (thnks Lenny and NiHaoMike)
- Temps under occt are at around 40/50C for the cores (note that for xeons the tCase is 68C, so max temps allowed are a bit lower than for desktop cpu) and 70/80C for the RAMs with the default cooling system, note that this test is very rough as I stopped it after 3min (read max temps in hwinfo).

Potential solutions:

1) Control the speed via software: This would be the ideal solution (if possible), as it doesn't have any side effect, but I still haven't found out which software of the huge HP suite (if there is one) can control the speed of the fans. I've tryed with third party softwares as speedfan but it doesn't work, simple the software doesn't recognize the fans. According to the HP support there's no way to control the speed of the fans via software

2) Use (build) a pwm fan controller like this one: This is not a bad option, but requires some diy (and tools that at the moment I don't have) and a mod of the fan cabling, that would be cutting the pwm cables from the system board and leaving the rest of the circuitry intact, to avoid problems (hopefully) with system check up.

3) Use a voltage controller (a pot or possibly something more efficient): Same as for the pwm controller, but with the need of a kick start and a bit more important mod of the fans cables (depending on the controller used.

4) Use both the solutions 2 and 3: In case the speed obtained by the two methods would be still loud, it may be worth a try to use both pwm and voltage control.

5) Use Arduino:well, here I still can't say much (see the posts below and at page 2 by NihaoMike and Bobnova for details), but apparently there would be a way to control (and maybe program) the sophisticated DSP circuitry embedded inside the fan using Arduino.

6) Substitute the fans: This is a solution that would be optimal from the point of view of silence, especially having a lot of spare fans getting rusty in the drawer, but would require an adapter (diy) 5pin->3/4pin classic, the installation of fans direcly on the cpus heatsink and over the ram modules to compensate the reduced airflow, and a severe check of the temps.

To end here there are some pictures (taken with my phone, sorry for the quality) of the system to better understand the thermal design:
1 - Whole build
2 - Chassis + CPU/RAM board (sorry, I had to censor my pants!)
3 - Detail CPU/RAM board
4 - Detail chassis fans 1
5 - Detail chassis fans 2
6 - Detail chassis fans 3 (fans' headers)
7 - Fan connector
8 - ICS chip (maybe pll....but I couldn't see the crystal)

 

[BruceUSA]

The controller you link here that won't work. You can't buy pwm controller to control high speed pwm fan because they don't exist. I don't know anything about your server mbo but on my experience with these fans, I was able to controlled via software, Asus Fan Xpert but can handle only 3 of it. When I add more fan, the pwm signal go crazy and won't work anymore. I am running 10 of these bad boys on my watercooled system and I am using a custom built pwm booster to drive all of my 14 delta fans. You should contact Bobnova member of this forum. He can build you a manual pwm controller for pretty cheap. I got a pwm controller and pwm booster from him.

This fan are 4.8a each, at least this is on mine. I don't think they are 3.24a that you listed. PS. correction. I guess there is a version of that 3.24a.

Check out the above link. Watch the video See the fan ramping up, is this noise level is acceptable to you. I was using 3 of Delta fans 5500rpm @252cfm pwm fan.

 

[adelage]

I am pretty sure there have to be a way to control them via software, but there are to many "system management" softwares in the HP website, and until now I've been just able to get info like "yes, your six screaming fan are working and the system is cool", but I can't even find the speed readings..I'm sure the fans are 3.24A as it's printed on the fan label, this one, I think the one at 4.8A is a similar model sold with the classic 4pin pwm header, mine instead has got a proprietary 5pin header, and also looks like the pwm signal input is reversed. About Bobnova, the problem is that I live in the UK, both of you are from the States, so I don't think is worthing the shipping/customs cost...What a shame for that motor controller, I was so happy and I didn't understand why so many people were wasting time and money building diy pwm controller when with less than 20$ you can get one ready online....I'm starting to consider to buy a welder...
BTW if it's you in the video, cool rig and nice kid, looks like he can't wait to play!

 

[BruceUSA]

than, try Lenny Rythn, he can also help, I think that is his name. He is from the UK. Yup that was me and my kid.

 

[Bobnova]

Shipping to the UK is fairly expensive. If you can get in touch with Lenny I'm sure he can figure something out though.

 

[adelage]

cheers guys, I'm going to send a PM to Lenny, even if I wouldn't dislike to build it on my own, especially considering that I have to mod the headers to keep speed,12v and ground wire on the server board...and in case I would use one of this beasts on my main rig with 4770K and H80(sorry for the blasfemy but a 360cm rad is hard to fit inside the SG-05), can i still keep on using the pwm signal from the mobo, right?

 

[adelage]

About controlling PWM fan, without any circuit, when you connect the PWM pin to the ground (negative), that is equal to 0% PWM duty cycle, the fan will slow down to it's rated minimum speed, and if you leave that PWM pin disconnected, it will automatically equal to 100% aka max speed, easy & nice isn't it ?

guys, perhaps someone can confirm this? if it's true I could give it a try,while I'm waiting to get/build a pwm controller...

 

[Bobnova]

Yup, it's true.
Bing is one of the main forces behind the PWM fan control movement around here

Do note that the minimum speed for your fans may be zero RPM.

 

[adelage]

well, considering that the pwm is declared reversed, I wonder what's gonna happen, maybe if I link it to ground the fan runs at max speed, and if I disconnect it the fan runs at minimum(or 0rpm)...and what if I try to control the pwm signal given by the mobo with a potenziometer? Would it work?

 

[Bobnova]

The PWM signal needs to be a square wave, so just a pot won't work.

If it's an inverted PWM device, 5v (100%) will be zero fan speed, and 0v (ground, 0%) will be full fan speed.

 

[adelage]

ah OK, sorry but my electronic competence is limited to the use of some resistors and diodes
so, as far as i understand, the pwm signal is an impulse at 5v and 25hz,isn't it? and what regulates the speed? the variation of frequency or of the variation of voltage? from what you said I guess the frequency is constant and the 0/5v range determines the speed...

 

[LennyRhys]

Hey adelage, thanks for the PM.

I have six of the exact same fan (-6J32) and they don't go much below 1500rpm, so controlling them by PWM will not make them any quieter, sadly. The 5v signal is a pulse signal which controls the speed of the fan via some fancy internal circuitry. The frequency of, say, 25khz doesn't change but the length of the pulses changes and causes the fan to speed up or slow down. At about 20% duty cycle, this fan starts spinning at about 1500rpm and is pretty loud. Anything below 20% and it grinds to a dead stop.

To get them quiet you would have to use voltage control, and building something like that is quite a task. Despite the rating of 3.24A, I measured the fans at about 2A at full throttle. You could control them with a Lamptron or something similar, but you'd have to keep an eye on how much power was being drawn in total.

 

[adelage]

Hey adelage, thanks for the PM.


To get them quiet you would have to use voltage control, and building something like that is quite a task. Despite the rating of 3.24A, I measured the fans at about 2A at full throttle. You could control them with a Lamptron or something similar, but you'd have to keep an eye on how much power was being drawn in total.

Thanks to you for joining us!
Anyway, the problem is getting tricky, but I have to find a solution, as soon I'll need to keep the server on almost 24/7, and I sleep/live in the same room.
About voltage control, I thought it was less efficient (I'm also concerned about power consumption, and it looks like this fans are quite power hungry) and that the rpm range was shorter than pwm control..once I come back home I'll give it a try with 5v and 7v

 

[wagex]

Thanks to you for joining us!
Anyway, the problem is getting tricky, but I have to find a solution, as soon I'll need to keep the server on almost 24/7, and I sleep/live in the same room.
About voltage control, I thought it was less efficient (I'm also concerned about power consumption, and it looks like this fans are quite power hungry) and that the rpm range was shorter than pwm control..once I come back home I'll give it a try with 5v and 7v

be careful cross loading the psu with that much amperage, ive done it on much lesser fans with fine results but depending on how many fans that seems like it could be unhealthy.

 

[adelage]

be careful cross loading the psu with that much amperage, ive done it on much lesser fans with fine results but depending on how many fans that seems like it could be unhealthy.

mmm, actually I don't even need to try 7v as, even if it works, it's gonna be still loud..I'll try 5v and 3.3v to see what happens using an external psu, even because I don't know where to get this voltage from the 2 beefy server psu..

 

[LennyRhys]

The only thing you might find with such low voltage is that the fans don't have enough power to start spinning. I can check it out for you.

Edit: the fans seem to have a speed delay built in to the supply voltage. At approx 5v the fan doesn't spin up (as expected); if I start the fan at 8v and turn it down to 5v, it reaches a certain speed, slows down a bit, and repeats the cycle of speeding up, slowing down, etc. VERY strange fan behaviour. Sadly these 6J32 models of the PFC seem to be designed specifically not to spin much slower than 1500rpm or so.

In the long haul it may be better to just replace the fans with something like Yate Loons, but you'd want to be sure that the decrease in airflow wouldn't compromise the cooling.

 

[adelage]

The only thing you might find with such low voltage is that the fans don't have enough power to start spinning. I can check it out for you.

wow, I have just come back from work, and honestly I feel a bit tired to deal with noisy fans without even having the proper instruments, so thank you very much!
I think it's exactly like you said, they are not designed to work at low speed...at this point I'm a bit confused, as buying other fans it's gonna be problematic: 1 because this was a budget solution to have a powerful rendernode (less than 300£ including 4 e7450 hexacore that I'm still waiting to be delivered), and I didn't want to spend money anymore 2 because the fans have a proprietary 5pin connector, with a kind of locked rotor check (that is nothing other than a bridge pin on the ground wire) 3 if I want to use other fans I can't use pwm to control (or make the system control) the speed. Further, if there are only 5 fans plugged, the system make them run at max speed, if there are less than five fans it shuts down.
the only thing I can do is try to understand if I can bypass the 5 fans limit (by design 3 fans are more than enough, the other ones are only for redundancy) and see if I can recycle some of fans I got in the drawer (at least as temporary solution), testing properly the thermal response...I was also thinking to put some fans on the cpu heatsinks (they are strangely passive) to compensate the reduced airflow in the system...anyway, tomorrow I'm off, so I'll spend some time to understand what can be done

Edit: the fans seem to have a speed delay built in to the supply voltage. At approx 5v the fan doesn't spin up (as expected); if I start the fan at 8v and turn it down to 5v, it reaches a certain speed, slows down a bit, and repeats the cycle of speeding up, slowing down, etc. VERY strange fan behaviour. Sadly these 6J32 models of the PFC seem to be designed specifically not to spin much slower than 1500rpm or so.

BTW I've seen a similar behaviour in a psu fan, a tiny (80x15mm) protechnic electric from my almost deceased sfx(aka itx) be quiet 350w, it used to react i very weird ways2v/3v med freq buzz @3v/4v high freq buzz @4v/5v starts "hiccough" spinning (similar to the one you described, but with some stop-starts @>5v normalize speed

 

[NiHaoMike]

Edit: the fans seem to have a speed delay built in to the supply voltage. At approx 5v the fan doesn't spin up (as expected); if I start the fan at 8v and turn it down to 5v, it reaches a certain speed, slows down a bit, and repeats the cycle of speeding up, slowing down, etc. VERY strange fan behaviour. Sadly these 6J32 models of the PFC seem to be designed specifically not to spin much slower than 1500rpm or so.

The variable speed Delta fans use digital logic (DSP) to do the motor control and like all digital logic, behave erratically with too little supply voltage. The PFC series are designed for high static pressure so I'm not surprised they won't go really low the way the regular ones do. The minimum speed is set in firmware but unfortunately, there's no way (I'm aware of) to reflash a fan. (The DSP will be running from an internal oscillator so no way to underclock it, either.)

That said, a common desk fan runs at about 3600 RPM so 1500 RPM shouldn't be particularly noisy. What might be an issue is that the director vanes might act like a whistle. Swapping them for fans without director vanes can help but I suspect the heatsinks are probably very restrictive.

 

[xsuperbgx]

can you lower the voltage and lower pwm frequency at the same time? Would it go below the 1500 rpm?

 

[adelage]

The minimum speed is set in firmware but unfortunately, there's no way (I'm aware of) to reflash a fan.

Hell, I didn't know than even a fan can have a firmware, I guess is due to this particolare kind of speed control, but

That said, a common desk fan runs at about 3600 RPM so 1500 RPM shouldn't be particularly noisy.

Well, if it was possible to keep them at around 1500rpm would be acceptable, but how?
I'd prefer not to alterate too much the cabling, but how can i reduce the impulse of the pwm signal sent by the system board, if a pot doesn't work?

or, 2nd alternative, replacing the fans (and adding some on the cpus' heatsink, as, how you said, their are a bit poor), how can I trick the circuitry of the system to make it think that all the fans are fine?

Tomorrow I'll post some pictures to better show the thermal design of the beast...