PWM-->Analog Fan controller for conventional 3-pin fans

dinofx · Post by **dinofx** » Sun Jun 04, 2006 5:12 pm

I have created a custom circuit similar in spirit to the one created by DirtyLude/Mark Higgins. But, I think we are each targeting different use cases. In my case, I'm depending on modern motherboards which already have PWM drivers built-in. My circuit takes a PWM input signal, and creates a steady variable output voltage. It was originally designed for a water pump, but it can power any two devices with normal 3-pin fan connectors on them. Mark's circuit does just the opposite. PWM is not present in the system, and pulsed power is generated by the circuit.

The other difference is that my circuit makes no attempts to interpret temperatures. CPU, GPUs, and motherboards already contain temperature monitoring embedded in the chips themselves. So, you can use SpeedFan or your motherboard's BIOS to interpret temperature levels, and adjust the PWM signal to your fan headers. The circuit simply turns PWM into a steady variable voltage.

The total size of the circuit is 1"x1.5". I have ordered the PCBs already and created a few of the circuits:

Kizz · Post by **Kizz** » Mon Jun 05, 2006 12:04 pm

Looks pretty good - am I right in thinking this does a similar job to the attenuator available with the T-balancer to smooth out PWM into a more analogue voltage signal? If so that's exactly the kind of thing we need more of (IMO) to fix PWM controllers. Do you have any plans to market these, and if so at what cost? Regardless, it's more than I could make, so good job.

frankgehry · Post by **frankgehry** » Mon Jun 05, 2006 7:12 pm

What size are the surface mount resistors in your project - 1206? TIA

dinofx · Post by **dinofx** » Mon Jun 05, 2006 7:41 pm

Kizz wrote:Looks pretty good - am I right in thinking this does a similar job to the attenuator available with the T-balancer to smooth out PWM into a more analogue voltage signal? If so that's exactly the kind of thing we need more of (IMO) to fix PWM controllers. Do you have any plans to market these, and if so at what cost? Regardless, it's more than I could make, so good job.

This is not like the attenuator. The MOSFET is precisely controlled to provide the exact amount of resistance needed for the desired voltage level, regardless of load. Also, it can handle up to 20 watts (as with water pumps). I have many of these available, $18 as a kit, or $35 soldered, assembled, and tested.

If the attenuator is just a big capacitor, it might help, but it would mess up the linearity of the response to input.

dinofx · Post by **dinofx** » Mon Jun 05, 2006 7:42 pm

frankgehry wrote:What size are the surface mount resistors in your project - 1206? TIA

Yes, they are 1206. It was not as hard to solder as I expected.

Bluefront · Post by **Bluefront** » Tue Jun 06, 2006 1:51 am

I admire your ability to solder those surface mount resistors. However I doubt many modders attempting to assemble a kit, would find it so easy. Ruin one and finding another would be difficult.

Which brings up the cost of an assembled unit...not bad. But you are competing with some devices with similiar abilities(changing PWM to analog). M-Cubed has the Fan-Amp and the miniNG. Both in the price range, both come assembled with cables and temp sensors. Tough competition.

I'm past the point in my life where I would attempt such a soldering project....eyes not so good, hands not so steady. Maybe some people on SPCR might be braver......

dinofx · Post by **dinofx** » Tue Jun 06, 2006 7:51 am

Bluefront wrote:I admire your ability to solder those surface mount resistors. However I doubt many modders attempting to assemble a kit, would find it so easy. Ruin one and finding another would be difficult.

Which brings up the cost of an assembled unit...not bad. But you are competing with some devices with similiar abilities(changing PWM to analog). M-Cubed has the Fan-Amp and the miniNG. Both in the price range, both come assembled with cables and temp sensors. Tough competition.

I'm past the point in my life where I would attempt such a soldering project....eyes not so good, hands not so steady. Maybe some people on SPCR might be braver......

It is easier than you might imagine. You simply put down a drop of solder on one of the pads, then line up the resistor and re-heat the pad. But, there is a 3rd option. SM components pre-soldered, you do the through-hole stuff. $25 +S&H.

Quite frankly, there is nothing else like this on the market, or I would have bought one. M-Cubed is huge, occupies a PCI slot, requires a USB port, and it has no way of monitoring your CPU's embedded temperature sensors. On a modern system, who gives a crap about adding analog temperature monitors? Your Motherboard, CPU, GPU, and even Hard Drives already report their temperatures to SpeedFan. M-cubed has no access to these temperatures, unless perhaps with the use of their software and a USB cable. Also, they use PWM for controlling fan speed, which means you get tachometer problems, and some people say more noise.

On the other hand, this circuit is small, elegant, and can function properly without ANY software, external wires, or USB ports. For example, in my Shuttle XPC, the BIOS can be set to "Smart" fan speed, in which case it increased my water pump and radiator fan speed with no software. I have a PWM fan for the radiator, and a second non-PWM fan also powered by this circuit.

Bluefront · Post by **Bluefront** » Tue Jun 06, 2006 12:00 pm

Well I think you are talking about an M-Cubed T-Balancer......like what you describe. The M-Cubed FanAmp and miniNG are not USB connected....they are stand-alone fan controllers that can both run analog voltage.

I have a few of each. The Fan-Amp costs less than your unit....the miniNG costs slightly more. Neither require any software to run, in fact don't come with any. They use their temp sensors to control fan speed.....you cannot read the temps of these sensors.

The surface mount resistors would be the hard part of the soldering. The other things look easy....

disphenoidal · Post by **disphenoidal** » Tue Jun 06, 2006 3:53 pm

Neat idea, I like it. Is the amplifier (LM741) set up to integrate the pulses and provide a proportional output, with a MOSFET to provide extra current?

Also, soldering 1206s isn't all that difficult, if you have a reasonably good iron, tweezers, and solder-wick to clean up any mistakes. Of course, I have younger eyes.

I am kind of confused, however, by why you couldn't just use the PWM output straight from the motherboard?

dinofx · Post by **dinofx** » Tue Jun 06, 2006 4:50 pm

Bluefront wrote:I have a few of each. The Fan-Amp costs less than your unit....the miniNG costs slightly more. Neither require any software to run, in fact don't come with any. They use their temp sensors to control fan speed.....you cannot read the temps of these sensors.

Those sound similar. The only difference being that you are limited to input from two sensors that have to be installed in a less-than-ideal place. (I'm defining ideal as on my CPU

Also, tuning the reponse I'm sure is more difficult. You have to start ramping up the speed at some lower temperature to ensure that it runs at 100% when your "max" temp is reached.

dinofx · Post by **dinofx** » Tue Jun 06, 2006 5:11 pm

disphenoidal wrote:Neat idea, I like it. Is the amplifier (LM741) set up to integrate the pulses and provide a proportional output, with a MOSFET to provide extra current?

Also, soldering 1206s isn't all that difficult, if you have a reasonably good iron, tweezers, and solder-wick to clean up any mistakes. Of course, I have younger eyes.

I am kind of confused, however, by why you couldn't just use the PWM output straight from the motherboard?

It's an RC filter, which converts the PWM into a low-side analog signal between 0 and 5 volts. The signal is then inverted, multiplied, shifted, etc. using the op-amp as an adder/subtractor circuit. The output of the MOSFET (rather than the opamps own output) is used for feedback.

I just used my index finger to hold down the resistors. That way I knew they weren't getting too hot to be damaged

What do you mean by using the PWM directly? A PWM fan can do this because it is always receiving power for the Tach. Even if I didn't care about tach, my water pump has an internal capacitor, which means you can't get a linear response from the duty cycle. At 50%, the pump would basically draw twice as much current half the time, and run off the capacitor when the current is cut.

dinofx · Post by **dinofx** » Wed Jun 07, 2006 11:14 am

Here is a review of the MiniNG. It has some drawbacks, including difficult configuration, slow response time, and tricky temp. probe placement.

I know that my Pentium 4 with EIST and SpeedStep can change thermal output considerably over a very short period of time. I didn't get a good feeling for how quickly MiniNG adjusts its fan speeds, but I hope its not on the order of minutes.

http://forums.silentpcreview.com/viewtopic.php?t=31588

Bluefront · Post by **Bluefront** » Wed Jun 07, 2006 12:09 pm

Actually the FanAmp is more similar to your device than the miniNG. I have been using two in different computers for a while now. I never did a real review, because of the boards I was using.

The FanAmp has a connector to a MB fan header, a header that has MB fan control. If your board can automatically adjust fan voltage from 7-12V, the fan amp can use that signal to automatically send an analog voltage to your fan. It gets power from a molex connector, and has a temp sensor if your board doesn't support fan control.

It is a simpler device than a miniNG, but with some similarities. Your device is also similar in it's operation. Unfortunately, I never tried the FanAmp with a MB that supported fan control, so I don't know how well that function works....

dinofx · Post by **dinofx** » Thu Jun 08, 2006 6:05 am

Bluefront wrote:The FanAmp has a connector to a MB fan header, a header that has MB fan control. If your board can automatically adjust fan voltage from 7-12V, the fan amp can use that signal to automatically send an analog voltage to your fan. It gets power from a molex connector, and has a temp sensor if your board doesn't support fan control.

The more recent motherboards are switching to PWM to control fan speed using 4-pin headers. My circuit is only suited for these newer motherboards. Almost all socket 775 MBs have at least on such 4-pin PWM fan header, since Intel ships such a fan with their CPUs.

So, fan amp is designed for motherboards which actually varry the output voltage at a 3-pin fan header. If a motherboard has this capability, it essentially has a FanAmp built-in and I don't see the purpose of buying a fan amp for a fan amp. I guess the only purpose is to boost the power handling and possibly lower the bottom end range using manual tuning of the pots.

By changing the resistors on my circuit, you can get linear control from 6-12V, instead of 7.4-12V. By coincidence, replacing every resistor with 100K does the trick. I could probably detemine the resistors required for any desired control range, but I only have 47, 68, 82, and 100 in my possession to work with.

Bluefront · Post by **Bluefront** » Thu Jun 08, 2006 2:54 pm

You're right about the FanAmp....it can handle a lot more current than the MB header. And you can manually adjust the idle speed. And using the second pot and the supplied sensor, you can manually set a temperature that will turn the fan to full speed. And it does work well not connected to the MB....that's how I use mine.

dinofx · Post by **dinofx** » Fri Jun 09, 2006 9:01 am

I remember reading specs stating 1 amp, or 12W, for each fan header, but I don't plan on testing that limit any time soon.

Bluefront · Post by **Bluefront** » Fri Jun 09, 2006 3:26 pm

Here's the specs from the M-Cubed site...

"FanAmp is designed specifically for 1-channel fan systems, for semi-passive cooled systems or for safety fans in passive cooled systems. The FanAmp is easily adjustable with 2 poties. FanAmp makes it also possible to convert the mainboard signals (7-12V) into a wider range of 0-12V. This features make zero fan speed controllable with any mainboard tool. Performance: 25W permanent
Package contents:
1 analog temperature sensor, control unit, motherboard cable"

dinofx · Post by **dinofx** » Fri Jun 09, 2006 6:40 pm

To clarify, I was referring to motherboard specs, not the fanamp. According to Intel, a motherboard fan header is good for 12W. The older version of the Laing water pump was 11W, but it would be a mistake to try to plug it into a fan header.

dinofx · Post by **dinofx** » Mon Jun 12, 2006 11:00 am

Special offer: $28 assembled, tested, and shipped, to first person who PMs me. The only limitation is that you must be ready to use the circuit in your existing system with PWM fan headers.

Bluefront · Post by **Bluefront** » Mon Jun 12, 2006 2:41 pm

Well I'd be glad to try/buy one, but I don't own a MB with PWM. Too bad.

leem · Post by **leem** » Sun Jul 02, 2006 11:52 am

Dinofx, sorry for newbie question, but will your circuit work on 3-wires input from the motherboard or is the PWM the fourth wire?

I'm trying to find a mini-ITX motherboard which will enable me to control case and CPU fan speeds using Speedfan and the on-board sensors, but am having no luck. From what I read I assume that I need a motherboard with 4-wire fan outputs, i.e. it cannot control fan speeds just using 3 wires?

I'd love to be able to use your circuit to quieten the fans in my Hoojum case...

Thanks

cmthomson · Post by **cmthomson** » Tue Jul 04, 2006 4:57 pm

Just curious, why two-pin headers instead of three-pin?

One nice feature of the NMT-2 and -3 is pass-through of the RPM sense signal on the third pin. That way you can use standard fan connectors.

Your device with 3-pin headers would be ideal for PWM motherboards and standard fans (no need to replace connectors).

I was all set to order one until I noticed this...

sjschwinn · Post by **sjschwinn** » Wed Jul 05, 2006 1:55 pm

Bluefront wrote:Here's the specs from the M-Cubed site...

"FanAmp is designed specifically for 1-channel fan systems, for semi-passive cooled systems or for safety fans in passive cooled systems. The FanAmp is easily adjustable with 2 poties. FanAmp makes it also possible to convert the mainboard signals (7-12V) into a wider range of 0-12V. This features make zero fan speed controllable with any mainboard tool. Performance: 25W permanent
Package contents:
1 analog temperature sensor, control unit, motherboard cable"

I'm new here, but I can comment a bit more on the FanAmp since I've been struggling with it for several weeks, and working (at a snails pace) with m-Cubed tech support (and a language barrier) to figure it out....

The FanAmp DOES NOT CONVERT PWM TO DC ANALOG! I thought it did, but after several e-mails I discovered it does not. It is designed to do DC analog to DC analog range conversion.

I was looking around the 'net to find some circuits to do PWM to analog conversion, and there are several. The problem was building it. I think Dinofx's widget fills a great need that nothing yet solves.

dinofx · Post by **dinofx** » Thu Jul 06, 2006 7:47 am

cmthomson wrote:Just curious, why two-pin headers instead of three-pin?

One nice feature of the NMT-2 and -3 is pass-through of the RPM sense signal on the third pin. That way you can use standard fan connectors.

Your device with 3-pin headers would be ideal for PWM motherboards and standard fans (no need to replace connectors).

I was all set to order one until I noticed this...

So, I take it since your assumption is flawed, you're ready to place your order

. There are two 3-pin headers for two normal fans. The 2-pin headers are for plugging in the power and MB. It only takes a single pin to send the RPM signal to the motherboard. So, the tach singal from one of the fans *is* passed-through.

The two 2-pin headers are basically the typical 4-pin connection to the motherboard. They are Tach, PWM, Ground, and +12V. Because they are side-by-side, you can use a single 4-pin terminal housing, or two 2-pin housings, to allow the wires to be separately (dis)connected. Because water pumps would require a lot of juice, it is possible to use your PSU for power source, and the motherboard only for the in/out signals.

Myth! · Post by **Myth!** » Thu Jul 06, 2006 1:53 pm

you also have a market in those who have the m-Cubed T-ban, and want proper DC outputs to some of their devices. The T-ban digital doesn't do this as stock, and the attenuators only help a little. The T-ban kicks out PWM but it isnt a square-wave, they says its like a sine wave from 0v-12v-0v etc. I think the attenuator changes this to say 5v-10v-5v or something similar. Do you know if your circuit will handle this unique (i think) type of PWM and give a solid DC output?
The T-ban can quite simply be hooked into Motherboard Monitor to use the on-board sensors or both the onboard and the externals (highest counts) to give protection during a crash. The other advantage i found is putting a sensor in the PSU and hooking up that fan to the T-ban giving me a much quieter zalman 300w at idle/normal than using the thermistor.

dinofx · Post by **dinofx** » Thu Jul 06, 2006 5:55 pm

The circuit expects an input with an average voltage of between 0.0 and 4.3 volts, since that is what Intel's PWM spec states. The resistors could be swapped out to support some other range.

cmthomson · Post by **cmthomson** » Fri Jul 07, 2006 3:14 pm

dinofx wrote:So, I take it since your assumption is flawed, you're ready to place your order . There are two 3-pin headers for two normal fans. The 2-pin headers are for plugging in the power and MB. It only takes a single pin to send the RPM signal to the motherboard. So, the tach singal from one of the fans *is* passed-through.

Doh!

I definitely should have looked closer... BTW, you might want to edit that (much better) picture into the initial post.

Are the two tach pins of the two 3-pin connectors wired together, or is the upper one left NC? I.e., do you need to cut/pull out one of the tach wires in a two-fan system (and be careful which one)?

dinofx · Post by **dinofx** » Sat Jul 08, 2006 3:14 pm

You do not need to modify your fan's wires. In my case, the top fan is a Panalfo 80x15mm that did not come with tachometer. Only the bottom position fan/device can report tach to the motherboard. The other is NC.

dinofx · Post by **dinofx** » Tue Aug 22, 2006 7:04 am

I have been using this circuit without problems for many months. Another modder recently used one in his AMD-based motherboard. It worked perfectly, although that particular motherboard required a fan with tachometer signal or else you could not reduce the speed. I have 3 assembled boards ready for use.

Post by **MikeC** » Tue Aug 22, 2006 7:40 am

dinofx -- This is a fantastic device. I'd like to get a hold of a couple to review! Pls email me about this.

Also, you should know that I've seen in a recent Gigabyte(?) AM2 board fan controls in the BIOS that includes a switch between PWM and voltage control. This is on a board with a 4-pin CPU fan header. I think the board makers are taking the fan control chip developed and specified by Intel for their boards and putting them in AM2 boards -- probably to save on $ & time.

sjschwinn -- You said The FanAmp DOES NOT CONVERT PWM TO DC ANALOG! I would like to hear more about this. We have a reviewer who is about to start work on mCubed fan controllers...

Ditto, Myth! -- I'd like to hear more about your knowledge of the T-ban.