I made a fan controller. Want one?

haysdb · Post by **haysdb** » Tue Jan 06, 2004 6:10 pm

Fancontrol,

If anyone seriously wants to review one of these, and YOU have any interest in that happening, I will make one of my boards available so long as the reviewer agrees to either sent it back or reimburse me for it.

I'm still trying to decide how best to use mine. I currently have all of my machines running Folding@Home 24/7, so a smart fan controller doesn't offer any advantage over set-and-forget fan settings.

David

fancontrol · Post by **fancontrol** » Tue Jan 06, 2004 7:32 pm

Thanks for the offer, I'd like to keep it in the back of my pocket for the time being.

haysdb · Post by **haysdb** » Wed Jan 07, 2004 12:21 am

Consider it a standing offer.

silvervarg · Post by **silvervarg** » Wed Jan 07, 2004 3:08 am

In time I hope to do a full review of the controller, but it might take lots of time before I am done. The pros are that I got 2 boards on my own, and I got a good camera and good place to put up images.
The images I posted so far looks ko in quality. I will do more proper cutting and shrinking of images for a final review.
At the moment I am struggeling hard with the software and reading the ADM1027 data sheet.

Somehow it seems that the read file might not get all data loaded into the controller all the time. Not really sure about this. The simple read and write buttons seems to work all the time.
Perhaps it has something to do with data values with leading zero? E.g. 02 or 2 might not be treated correctly by the read file?
So far I have not loaded anything into EEPROM, just to ADM1027.

I spent some time trying to get an old and rattling NMB 92mm fan to work without extensive clicking. At 100% PWM duty cycle it works okey. At anything considerably lower is click a lot regardless of PWM frequency.
I will try a lot with some other fans. Problem at the moment is that I really need to get a good 92mm fan and a few 80mm L1A, but both of these seems hard to get by.
I might even give up on the L1A and get a few ADDA 14dB fans instead. Does anyone have a clue on how they work with PWM.

fancontrol · Post by **fancontrol** » Wed Jan 07, 2004 10:40 am

silvervarg wrote:In time I hope to do a full review of the controller

cool.

Somehow it seems that the read file might not get all data loaded into the controller all the time. Not really sure about this. The simple read and write buttons seems to work all the time.

that may be due to timeout errors on your PCs serial port. It would be ok with small commands becaust there's no chance for the timout that kills (there are three different ones) to act up.

I'll post a new version of the code that has longer timeouts on the web site soon. The down side is it takes a little longer to load settings, which isn't that big a deal.

It's definetely not a 'leading zero' problem. I did test all the boards here, loading multiple configurations and temperature compensations. That leads me to think it's something btw your PC and the board. Please email me any specifics that lead you to think otherwise.

I spent some time trying to get an old and rattling NMB 92mm fan to work without extensive clicking.

clicking/ticking/noise in general is bad. I told semm that maybe I never had the problem because I use smaller fans. Putting a cap across the fan might help, and it's easy to add in-line with the fan. I'd like to try it here before I send something out, I just haven't had time.

//edit
Can someone with ticking try something for me? By turning off tach monitoring you may get less noise, and a more varied response when changing PWM frequency. Turning of monitoring is, for obvious reasons, a bad idea, but the results may point us to a long term solution.

If you're up for it, here's what you do:
Assuming your ADM is at 0x5C (you can find out by clicking status)
Enter 5C78 and click "read" - this gets the value of config register 3, I'll bet it returns 80.

Change the value from 80 to F0. see what happens and let me know. You can set it back by either resetting the micro or just writing 80 back to the same register. Setting it to 00, btw, will probably cause fan 3 to freak out.

fancontrol · Post by **fancontrol** » Wed Jan 07, 2004 8:56 pm

I brought home a big fan to do some testing. The results were remarkable... or should I say, remarkably disappointing.

First, in my defense, the biggest one I have at home is some kinda 60x10, which isn't much.

The fan I brought home is a 120mm 24V panaflow (FBL12G24U, for those into that kind of thing). I hooked it up and it rattled the windows!

At the risk of taking up space, here's what I found:
There are two sources of annoying noise. I'll talk about them seperately.
1) tach measurement
About once a second, the ADM applies full power to the fan until it gets a tach reading. If you're running your fans at low power, and thus very slow, it takes a while for the good reading to show up. As a result, you will hear a 'rush', but no ticks, about once a second. Annoying! Let's call this devil "throb."

You can disable this, using the stuff under 'edit' in the last post. But loosing tach measurement is a bad deal, and you still have to deal with PWM noise, which brings me to...

2) pwm itself
There is no doubt that PWM, particularly at low duty cycle, makes fans shake like mad. We'll call this menace "tick."

You can vary the frequency to try and minimize it. Higher frequency turns the rattle into a hum, but seems to increase the minimum duty cycle (i.e. you can't run the fan as slow). Reducing the frequency was most appealing to me. It's still annoying, however.

So, what to do? Here are my first thoughts:

Putting a capacitor across the fan:
I tried it with a 1uF; tick gets a little better. You will still have the issue with throb, however, and a cap could make it worse for 2 wire fans. Moreover, for large, high current fans you will need, as one of my college professors once put it, a BMFC!

Change from PWM to voltage control:
Off hand, I can't come up with a good way to do this with the boards I've designed. There's probably a way if I started from scratch, but I don't think there are enough places to tap in to make a band-aid (but I'll think about it more).

There is one thing, and it makes me cringe a little--something like a zalman fanmate. Hear me out: +12V is used only to drive the fans. by reducing the +12V feed to something lower (anything helps) you get the benefit of increased PWM duty cycle (less tick) and a smaller difference between the pwm voltage and full-on (less throb).

The down side to that, of course, is that fans can not run at full speed when things get hot. While there is no immediately appealing solution to that one, you (or I) could rig something up to the uC alarm outputs that disables the fanmate-esque thing when a thermal limit is exceeded. [for those who don't know the ADM data sheet better than anyone should, you set temperatures for each sensor above which all fans run at 100%] So what you get is a system that nominally runs slow and quiet with itty bitty ticks and throbs, hopefully less than you can hear, that kicks into blow-dryer mode if anything gets scary hot. It's another thing for me to think about more.

Off hand, I think I'm going to try getting some sort of fanmate-esque thing running on my old proto (it won't be easy; the proto is junk compared to what I've shipped) and see what happens.

discuss.
-fc

PS 120mm panaflow in a shuttle V25? hmmmm...

PPS The other fan I brought home, that I didn't try, is a Comair Rotron MC12T3H. It has power wires and 2 26AWG wires tied together. Anyone know what this is about? Is it thermistor controlled?

haysdb · Post by **haysdb** » Wed Jan 07, 2004 9:15 pm

I appreciate you being up front about all of this. People with less integrity would say things like "it works for me", "I don't experience any ticking" and "it must be something wrong with your configuration".

David

BaconTastesGood · Post by **BaconTastesGood** » Thu Jan 08, 2004 12:15 am

Fancontrol,

I can't find the schematic and specs for your board, are they in this thread and/or can you summarize?

A linear voltage controller has a lot of advantages. The main problem is that it generates a lot of waste heat potentially, and it's also more difficult to control from a uc. If there were a PWM -> vdc type chip, that would be ideal.

silvervarg · Post by **silvervarg** » Thu Jan 08, 2004 5:30 am

Thanks Fancontrol, now the fan behaviours make a lot more sense to me.
I also realised that I should have checked my 2/3 pin setting in the 0x73 register. I must check that when I get home.
I will test with turning tach measurment off to see what effect this has on PWM noise. I have done some tests with an 80mm Zalman fan and a 80mm Panaflo 24M1A.
In short the Zalman is not doing very well. But it is not very good when running through a fanmate either.
The 24M1A works ok with TACH on and at least 19% (0x30) PWM duty cycle. Below this value the vibrations gets a little bit bigger, but still acceptable.

At 0x01 PWM setting I still get 346RPM on the 24M1A. I guess this is the long tach reading pulse that brings the fan up to speed. Since the 1% PWM itself provides almost no force this is almost the same as going with one long pulse every second. So the result is close to 1 Hz PWM frequency.
Does not look to promising for the TACH signal.

The TACH reading on the datasheet (0x78) looked like 1=TACH reading on. From your post you indicate that 0=TACH reading on.
If you are correct I would assume that bit 3 (FAST) should be interpreted the same way. That is 0=FAST on. That would suggest a setting of F4 instead of F0.
I will do some testing to find out for sure.

The real say thing is that I believe the ADM1027 measures TACH in the wrong way. The correct way of measuring TACH would be something that didn't affect the PWM. The only way that I could come up with is to measure TACH pulses during each PWM on pulse no matter how short it is and then divide the measured RPM with %fan speed.
Knowing that %fan speed is sqrt(%PWM duty) this can be calculated.
In fact I might even consider implmenting this calculated TACH as speed presented in software to user from current PWM setting. This way I could disable TACH on the ADM1027 and skip the TACH cable to the motherboard.
This all depends on how my further findings about how some fans react to PWM driving and what the effects of turing of TACH in the ADM1027 turns out to be in my tests.

fancontrol · Post by **fancontrol** » Thu Jan 08, 2004 6:45 am

setting a 1 in bits 3-7 of 0x78 tells the controller to quit taking good measurements of fan speed (i.e. stop throbbing). Instead it does the best it can at the current pwm settings. Speeds go all over the place.

I'm not sure I get what you're saying about tach measurements.

the down side to calculating speed instead of measuring it, of course, is it does not indicate if the fan is dying, stalled, or even present.

fancontrol · Post by **fancontrol** » Thu Jan 08, 2004 6:47 am

silvervarg wrote:I also realised that I should have checked my 2/3 pin setting in the 0x73 register. I must check that when I get home.

actually, I use the adm to look at the jumpers to know whether its a 2 or 3 wire fan. If the jumpers are installed correctly, the setting is taken care of for you.

fancontrol · Post by **fancontrol** » Thu Jan 08, 2004 6:48 am

BaconTastesGood wrote:I can't find the schematic and specs for your board, are they in this thread and/or can you summarize?

click www below, go to downloads, and download. it's all there.

silvervarg · Post by **silvervarg** » Thu Jan 08, 2004 7:18 am

Fancontrol:
setting a 1 in bits 3-7 of 0x78 tells the controller to quit taking good measurements of fan speed (i.e. stop throbbing). Instead it does the best it can at the current pwm settings. Speeds go all over the place.

I'm not sure I get what you're saying about tach measurements.

the down side to calculating speed instead of measuring it, of course, is it does not indicate if the fan is dying, stalled, or even present.

Some of this does make sense, and some doesn't.
What doesn't really add up is:
Bit 3 (FAST) (4x long pulses per second) looks like 0/1 would have same meaning that it has for bits 4-7. According to your examples values 0x80 vs 0xF0 it doesn't.

I agree on the downside with calculated tach. But I could combine this with the erratic TACH values that you said we still get to notice if a fan stalls or is not present. That would remove most of the problems.
Not sure how a fan dying would be detected. Would the fan stop or does it just slow down and you could possibly detect this days before it actually does die?

I got yet another problem that I just remembered. I get Interups from ADM1027 "Int1: 03". Reported from fwf1.exe.
I assume this is Interrupt mask 1 register 0x74, so this would indicate error on Vccp and 2.5V.
Any clue on what effect this would have?
I will check what limit values is currently setup for these, but it would be the defaults set when the board was shiped.

fancontrol · Post by **fancontrol** » Thu Jan 08, 2004 8:50 am

silvervarg wrote: Bit 3 (FAST) (4x long pulses per second) looks like 0/1 would have same meaning that it has for bits 4-7. According to your examples values 0x80 vs 0xF0 it doesn't.

bit 7 (0x80) must be set, otherwise the ADM thinks fan 4 is stalled all the time. Since that's tied to PWM3, the third fan goes nuts.

bit 3 (0x08) is, apparently, the same as setting bits 4-7 (0xF0). If I get around to it, I will ask the developer of the chip; we've had a few chats in the past.

I agree on the downside with calculated tach. But I could combine this with the erratic TACH values that you said we still get to notice if a fan stalls or is not present.

Not true. Even with no fans present you can still get 'phantom' tach readings, particularly for 2 wire fans, due to PWM pulses.

does it just slow down and you could possibly detect this days before it actually does die?

yep.

this would indicate error on Vccp and 2.5V.

Since the ADM is not connected to 2.5 or processor voltage, those interrupts should be ignored.

Jan Kivar · Post by **Jan Kivar** » Fri Jan 09, 2004 2:33 am

fancontrol wrote:
BaconTastesGood wrote:I can't find the schematic and specs for your board, are they in this thread and/or can you summarize?
click www below, go to downloads, and download. it's all there.

The datasheets aren't in the .ZIP. Nor the board layout/schematic.

Jan

silvervarg · Post by **silvervarg** » Fri Jan 09, 2004 2:55 am

I did lots of testing with the long TACH pulse turned off. It worked exactly as Fancontrol described, and the results are very good.
With the long TACH turned off the fans rotate at a lot more evenly, removing most of the vibrations. I also thing that the fans become more silent, but it is easy to get fooled here since at low duty cycle the fan speed is heavilly affected by the long TACH pulse.
Without the long TACH pulse the slowest speeds that fans really run at becomes reasonable.
So far I have tested with the following new fans:
40mm ADDA (AD0412MS-G70 0.08A)
50mm Cooler Master
80mm Zalman PS80252H 0.25A)
80mm Panaflo 24V M1A

The 40mm ADDA works great on all speeds down to 19% PWM. No noticable clicking.
The 50mm Cooler Master clicks a bit from 13%-50% PWM. At higher PWM's the clicking reduces and eventually get drowned by other sounds from the fan.
The 80mm Zalman Clicks all the way (it clicks even when voltage regultaed and even at steady 12V). The clicking reduces to an acceptable level at ~69% PWM and above. I would say this is a low quality fan that should not be used anywhere.
The 80mm 24V M1A works great all over the register. PWM does not cause much clicking or vibrations on any setting from 30% PWM and above.

As expected the TACH readings become slightly strange when running without long TACH measurment pulses, but I still get fairly OK TACH readings on all the acceptable settings mentioned above.
The readings are not good enough to use as a presentable RPM number right away, but there is enough information to use for a program. The program will be able to notice: if a fan is present, if it has stalled and even detect prematurely if a fan is dying. It will also be enough to calculate the real speed that the fan is spinning.

I don't have to many fans at hand at the moment to test with, but so far it seems that high quality fans that don't click much works fine with PWM. Low quality fans that seems to work poorly no matter what.
Small fans work slightly better than large fans, but at least 80mm fans work fine with PWM.
The long TACH pulses are horrible to use compared to results without them, so turning them off is a must IMHO.

Jan Kivar · Post by **Jan Kivar** » Fri Jan 09, 2004 3:10 am

silvervarg wrote:Fancontrol went with Cygnal chips, and I think the choise was mostly based on that he already had programmer etc for those chips.
I did look a bit into similar things and think the PIC chips seems very nice, but so far I am just on the planning stage.
I guess there is about 5 more interesting series.

After reading few posts at our national HW forum, I've started thinking of Atmel's AVR ATmega family for the MCU. Has pretty much same features as PIC16F87x series. And from the bashing in the forums, AVR chips seem to be better than PIC... Don't know about it, don't shoot the messenger.

There is an evaluation board; AVR STK500. Pricing is around 150€, as You said. (the bare board, it comes with a basic MCU, but seems that I'd still need to buy a MCU from the ATmega line. The basic MCU don't have any ADCs. If I read correctly...)

On to the next "problem"; I don't know C, nor Assembly. I did a basic course in Java few years back, but it didn't have any serial port etc. stuff. Oh, and I did one course in Turbo Pascal, like six-seven yaers ago.

There seems to be "free" C compiler for the AVR family: AVR GCC. Or, it's a library (?) to the GCC to compile C to... something that the MCU understands.

Could I pull this off (the entire design, that is)? How "easy" it is to program the code for those MCUs? Or does this post sound like I'm going to end up with something expensive fried?

Cheers,

Jan

silvervarg · Post by **silvervarg** » Fri Jan 09, 2004 3:39 am

And from the bashing in the forums, AVR chips seem to be better than PIC

I believe that there is a small difference. To be honest I have no clue on what is better. The AVR would have been my second choise. Perhaps more invenstigations will make me change my mind to go with AVR instead. I think it will take another year before I start doing anything, so there is plenty of time to change my mind.

Could I pull this off (the entire design, that is)? How "easy" it is to program the code for those MCUs? Or does this post sound like I'm going to end up with something expensive fried?

It is very hard to manage to fry something because of program bugs since you will probably test everything carefully before actually useing it in a computer as a fan controller.
If you manage to fry anything druing testing it will probably be due to hardware errors, and if you go with a low cost microchip it will be less than 10 euros that you fried.

On to the next "problem"; I don't know C, nor Assembly.

If you can download someone elses progam for this processor and understand how it works I am sure you will be able to use parts of it as a skeleton for building your own program. You will probably need to lear a bit about binary/hexadecimal and bit handling things about the C language, but that shouldn't be big a problem.

Jan Kivar · Post by **Jan Kivar** » Fri Jan 09, 2004 4:22 am

silvervarg wrote:
Could I pull this off (the entire design, that is)? How "easy" it is to program the code for those MCUs? Or does this post sound like I'm going to end up with something expensive fried?
It is very hard to manage to fry something because of program bugs since you will probably test everything carefully before actually useing it in a computer as a fan controller.
If you manage to fry anything druing testing it will probably be due to hardware errors, and if you go with a low cost microchip it will be less than 10 euros that you fried.

There's one of the problems: money. Or, more precisely, wasting it.

I think I'd need to put at least 250€ (BTW, can You all see the Euro sign?) in the components (the board and various other stuff, transistors, resistors, connectors etc.). Then, after I can get the design working (not necessarily 100% at the time), I'd need to design a PCB. I don't have a clue how; seems that I'd need a program to do this. Which are quite expensive. Any tips? Could I use the ExpressPCB's tools, and then make the board somewhere else?

I can't manufacture the PCB by myself. I think that there are few small companies (national, of course since I don't possess a credit card ATM) who could process it for me, but I think that it's not cheap. (say, 100€ for a board should be more or less there)

And then there is the building part. Again, chip in another 100€ (at least) for a decent soldering unit and all other stuff (solder, wick). Oh, and did I mentíon yet that I'm not a seasoned solder?

So, 500€, for something that could go up in smoke once You plug it in is terrifying me. Plus, it needs lots of effort and time. As I'm still studying, time is a factor.

silvervarg wrote:
On to the next "problem"; I don't know C, nor Assembly.
If you can download someone elses progam for this processor and understand how it works I am sure you will be able to use parts of it as a skeleton for building your own program. You will probably need to lear a bit about binary/hexadecimal and bit handling things about the C language, but that shouldn't be big a problem.

I've seen few other source codes, but, as I said they seem bit gibberish to me as I don't know C/Basic/Assembly. And assembly is device-dependent, or so I've understood?.

I think I can handle them though, during the Java course I was totally into it (coding, that is). I just had to finish all the assignments, even if doing all didn't improve my grade anymore.

Silvervarg, fancontrol (also the rest of the SPCR community!), would You be interested if I'd make a separate thread about my "project"? I think that I could list some goals for my design, and get some feedback/ideas/info...

Cheers,

Jan

silvervarg · Post by **silvervarg** » Fri Jan 09, 2004 5:20 am

Sure, make a new thread. It might be time to let this huge thread go to sleep.
I will comment in the new thread instead.

fancontrol · Post by **fancontrol** » Fri Jan 09, 2004 7:13 am

Jan Kivar wrote:The datasheets aren't in the .ZIP.

True, but if you can get to the zip then you can get the mfg web sites, which have all the info you need. Mfg and Mfg p/n for anything that matters is in the BOM and in the documentation, such that it is.

Jan Kivar wrote:Nor the board layout/schematic.

Not true. The layout (.pcb) and schematic (.sch) are all there, in their native format so you can screw with it all you want. The documentation, again, points out that www.expresspcb.com has the development tools you need to edit these files.

Jan Kivar · Post by **Jan Kivar** » Fri Jan 09, 2004 8:10 am

Sorry, my bad on the schematic. I thought that it would be in expresspcb -folder. Actually, it's good (for You) not to include those datasheets, as they are probably quite large. The .zip is currently ~400k, with the datasheets it would be nearly ~5 MB.

Cheers,

Jan

fancontrol · Post by **fancontrol** » Wed Jan 21, 2004 8:32 am

if you're in the mood to tweak on the uC firmware yourself, you can buy my debugger here.

I'm off to other things. Namely, a tiny, cheap, two speed, thermistor based thermostat (hardly a controller). In hindsight I should have started (and stayed) with something that simple.

fancontrol · Post by **fancontrol** » Mon Mar 22, 2004 8:40 am

one final note, particularly to anyone thinking of building one of these boards (or something similar) themselves:

The SMSC EMC6D103 is almost exactly like the ADM (to conform with the Intel spec.) with two big differences:

1. It doesn't appear to support 2-wire fans.

That, as we know, could be overcome by additional external circuitry.

2. It offers a high speed (25kHz) PWM output. That could easily be filtered into a variable voltage output -- NOT PWM.

have fun...

silvervarg · Post by **silvervarg** » Sun Apr 11, 2004 7:57 am

I think I am set for the moment with my fan control setup.
Since noone else has been able to figure out a smart way to get the setup I really wanted I changed the setup in my computer to have 4 separate compartments. The seems to be nessecary to get real quiet cooling without an even more advanced controller.
Since I could not find a more advanced controller this seems like the next best thing. Unfortunately I am down to 1-sensor-1-fan setup, and this is a bit of a waste for an advanced fan controller.

The setup is like this:
Compartment 1: Onboard sensor only. Most critical components:
Fancontroller board, MOSFETs. 1 fan set to turn on at 39*C. Has never turned on so far.

Compartment 2: Temp 2 sensor only. Most critical components:
CPU, northbridge, southbridge. 1 fan set to turn on at 62*C. If pushed for a long time with prime95 on worst settings it will turn on at minimum speed ~700RPM. In all normal operations the fan stands still.
If I clock CPU harder for more advanced gaming this will turn on and keep me on safe levels (still needs more testing).

Compartment 3: No temp sensor. Most critical components: RAM?
No fans connected, and no place for decent sized fans.

Compartment 4: Temp sensor 1. Most critical components: HDD, PSU.
So far I have a single fan in the PSU that is constantly on at ~5.5V. It will have to go soon, but not really sure how to revbuild yet.

In case any of you out there is interested here is basicly my setup file for the controller.

Code: Select all

Comments: In section of 3 rows (disregaring first row and last 2 rows) this is:
1. Configure fans to sensors. 1-to-1 mapping.
2. Set temperature range for each sensor to 13.33*C
3. Set mimimum fan speeds. Note that PWM3 is either stopped or at full. It is attached via a fanmate set to lowest setting.
4. Set start-fan temperature for each sensor.
5. (2 rows only) Set hysteresis to 2*C of all sensors.

Now I am curious how others have setup there controllers and how they designed thir airflow and how the fans interact with each other.

Is seems to be very complex to make a perfect fan setup, so it might be easier to just stay with simple 1-sensor-controls-1-fan-channel. Then you can ofcourse connect more than one fan on the channel.

Rory B. · Post by **Rory B.** » Sat Apr 17, 2004 6:27 am

How does the fan's RPM circuit work? Does it absolutely need 12v? Or can it work at 5v if the controller knows to look for a 5-volt signal? Or whatever the operating voltage happens to be at the time?

fancontrol · Post by **fancontrol** » Sat Apr 24, 2004 7:47 am

if you're talking about fans with a tach wire, it will be fine. if you're talking about fans w/o a tach wire it *should* be fine, but you may need to change the value of the shunt resistor.

this thread just won't die.

I made a fan controller. Want one?

holy earthquake, batman!