temperature sensing/fan control over smbus

quix · Post by **quix** » Sun Mar 02, 2003 4:06 pm

Hey!
i'm planning to build a simple cicuit around a LM85 to monitor temperatures and control fans and hook that up to the smbus on my asus p2b (would probably also work on most other mainboards...). i would also code the software for that.

my question, has anyone of you done something like that already, or could give me links to similar projects?

jpsa · Post by **jpsa** » Sun Mar 02, 2003 7:39 pm

yup! done already, go to the motherboard monitor site mbm.livewiredev.com and click on "extensions" PWM controlable by software for a lot of fans

fancontrol · Post by **fancontrol** » Mon Mar 03, 2003 8:18 am

I'm working on something very similar using an Analog ADM1027. The advantage over the National parts is that the ADM does closed loop fan control by itself; it doesn't require constant monitoring.

I use a micro-controller to create a seperate SMBus on the board, but included a header to connect that bus to anything else. The uC could be left off (or disabled or used in slave mode) for connections directly to the motherboard.

Read more here, if you like.

Riffer · Post by **Riffer** » Mon Mar 03, 2003 4:00 pm

I want something even simpler.

Just a way that I can plug one or two more of those common thermistors into the SMBus header on my Abit BD7 so that Motherboard Monitor can read them.

I have seen a few articles, but not simple enough for a newbie in electronics to follow.

fancontrol · Post by **fancontrol** » Mon Mar 03, 2003 8:33 pm

Riffer wrote:I have seen a few articles, but not simple enough for a newbie in electronics to follow.

Can you point me to them? I'm curious, mostly, but maybe I can help?

Post by **MikeC** » Tue Mar 04, 2003 1:51 am

fancontrol --

The AMD1027 is the chip used in the Intel thermal fan speed controller embbeded in some of their newer mobos. My experience with their implementation: TERRIBLE! The PWM causes the fans to jerk/pulse constantly until almost full speed is reached. It is worse than no fan speed reduction at all. I don't know enough about the circuitry or about PWM to tell whether its an intrinsic quality of the 1027 or their muffed implementation, but I'd expore this aspect of performance before proceeding too far, if I were you.

quix · Post by **quix** » Tue Mar 04, 2003 2:17 am

Heya!
here an update on my project, well, it's basically stopped. since it's hard to get these lm85 or something similar here in switzerland, i dont even have a credit card (thats not so usual here in swiss).

to mike's post, i dont think this would be a problem, the pulse-frequency can be set (on lm85 at least), and i think adding a capacitor should help/fix that.

i was looking throguh the internet for some similar projects and here's what i found, i'm sure some of you will find this interesting:

http://home3.inet.tele.dk/ersholt/Fanbu ... us_CK.html
this is something i could build, as it uses very standard components, but it's not very useful. well, i could use it to switch some fans from 5 to 7v when under load, but, it's too much work, i'm too lazy :-)

http://www.cpemma.co.uk/index.html
this page also has some simple and interesting circuits

http://www.overclockers.com/tips1059/
http://www.voidyourwarranty.net/review/ ... index.php3
two similar projects using a max6657

http://www.madhacker.org/mbmfansensors.htm
http://www.madhacker.org/mbmsensors.htm
two circuits using maxim chips for monitoring fan speed and temperatures. these chips are very easy to use, really, soldering them might be a problem, but otherwise, very simple.

and then, there are the datasheets to the lm85 (btw, supported by speedfan 4.07):
http://www.national.com/ds/LM/LM85.pdf datasheet of lm85
http://www.national.com/an/AN/AN-1260.pdf application note for the lm85, there's a very cool cicruit inside, basicly showing how to wire everything this chips can do, check it out.

and for those that only need temperature monitoring, without fan control, check the lm83:
http://www.national.com/ds/LM/LM83.pdf there's also a circuit inside, uses all it's functions, very simple!

hmm, it's a pity i cant get them... would be soo cool (hehe), you do your stuff, some heatsinks here, some there, put some temperature diodes, and let it do the rest. you tell it to turn the fans on to 5v at 50 degrees, and rise to 12v at 65 degrees... hmm, kinda high in this example, but convection would work better at those temperatures...

btw, while posting all those links, check my page at http://n.ethz.ch/student/mdjokic/pcmod3.html

fancontrol · Post by **fancontrol** » Tue Mar 04, 2003 11:20 am

MikeC wrote:My experience with their implementation: TERRIBLE! The PWM causes the fans to jerk/pulse constantly until almost full speed is reached.

Well, I've built the boards and installed them in PCs and have had none of the trouble you mention. The part will 'jerk' fans if the tach inputs are not configured correctly. It's trickier than I first thought; you have to pay close attention to the requirements laid out in the data sheet (I spent a weekend reading/prototyping/testing to get it right).

That could be the source of your problem. A quick check to see if that's it is to set the DCX bit for the corresponding tach in Configuration Register 3. That will eliminate the startup, or "pulse stretching", problem you may be having. Note that if you are not using the 4th tach input you must set the DC4 bit or PWM3 will flip out all the time.

I hope that helps. Thanks for the warning. Trust me, I wouldn't have gone this far if I had the problems you describe; I would have chucked the boards and found another way to waste my time