PWM to Analog converter circuit

[Tealc]

OK had another thought. What about this sort of concept.

Combined PCB with a line of PWM straight and a line of converted to 3 pin. Best of both worlds? A much neater solution.

PCB would be a little wider than my normal size but I think it won't be too much bigger, maybe 25mm wide or so.

And of course tach feedback off the 3 pinners which I didn't draw in.

 

[Tealc]

That's the idea. Gives you a bit of scope towards adding extra fans.

I've never made a version like this before. It shouldn't present any issues provided you don't use fans that load down the PWM signal like certain Corsair fans have been known to do.

 

[Tealc]

Quick on the knee board mock up.

Saves a few mm if the DC section is at the top and PWM at the bottom.

 

[Tealc]

Let me know if you'd like me to build this variant for you, or maybe if someone else might find it useful.

 

[Tealc]

Decided to go ahead and build the thing....

Quite a useful board this one. Negates the need for PWM splitters and fan controllers. The fan response curve for the 4x DC fans can be tailored with the separate RPM feedback.

Board dimensions 75 x 22 x 14mm.

 

[Tealc]

Almost.

I chose to put the actual PWM fan array's tach output on the 4 pin PWM connector and the DC section on a separate 3 pin connector.

 

[Tealc]

Those proper 4 pin PWM fan headers are quite a bit more expensive than the ones I've been using. I did buy a few of them early on but they were probably the most expensive part in the build. Most of the cost of the boards I sell, and it's only 1 every other month or so these days, is for the time it takes me to build.

Im sure they could be built in China for next to nothing but I'd need to stump up cash for thousands of units. I just don't have the spare money or belief that they'd sell all that well.

 

[Tealc]

Tealc would you be able to make one of these to run 12 fans? Just wondering as an option.

Hi Adam. The circuit as it stands isn't ideal for quite so many fans. I did do some ground work a while back on boosting the power capabilty and had some encouraging results but seeing as there was no outlet for these boards I didn't develop further.

My normal method of working around this is to make multiple circuits to spread the load of the fans. In this case 3 x 4 fans. The easiest way to do this would be to make a single PCB with 3 lines of circuits, similar to that imaged a few posts ago but with 3 lines of analog rather than a mixed digital and analog solution. A 12 fan solution would obviously involve a larger board, maybe something like 75mm x 40mm x 14mm (LWH). Load on the PWM channel would be minimal and should not pose a problem. Fan tach output could easily be made on all 3 channels to achieve balance. Power from Molex, obviously.

Was it all 3 pin fans you were interested in using or a mixture of PWM/3 pin?

 

[Tealc]

Good question. It's all about the tach feedback.

If the 12v is pulsed, the whole electronics of the fan are switched off between the pulses. What this means to the hall sensor that sends a pulse back to the motherboard twice per revolution is that it'll work and then not work several thousand times per second and this will cause absolute havoc to speed monitoring. I've seen crazy 125,000 rpm readings using 12v PWM on PC fans and the fan was only spinning at 600rpm or so. When a 2 pin fan is employed in say a PSU it doesn't matter about fan monitoring as they'll design the fan to fit the function.

What a 4 pin fan does is disengage the drive only and keeps the remainder of the electronics active, so the tach will still work.

With the converter circuit I have used there is always some residual voltage in the circuit to keep the electronics active, no matter the current draw, so the hall sensor provides a reliable result, down to about 3v or 4v.

 

[Tealc]

So in theory then you could pulse between 12v and 5v, instead of 12v and off?

It's just that I'm reading elsewhere that some mobos, Asus and Gigabyte etc, do use PWM on pin 2 for 3-pin fans...? And others are saying they vary the voltage...

Pulsing between 12v and 5v would provide only a 7v potential difference. This is how the 7v mod trick works. It would work but the fan would be limited to 60% speed.

Motherboards will likely use PWM to drive a DC-DC buck converter. I've not ever looked into it though but expect this to be the case as it's more efficient that way.

 

[Tealc]

For the tach to work it'll have to be a smoothed voltage, at least to some extent. To be honest I've never bothered to view the fan voltages or PWM signal on my computer to actually see what they look like on my oscilloscope. I may do one day.

 

[Tealc]

Carlos, who I sent a build off to a while back, decided to make a few PCBs and sent me a few to try out...

So I loaded up the components on one of them to see how it looked and how it changed the build time and quality.

Overall it makes it much easier to build. It's just a matter of shoving a few leads in holes, following the polarity in a few cases and soldering the leads from around back.

I bent the 4 pin PWM connector off to the side thinking it might be nice to have it come out the side, but decided to go back to straight up for the remainder. It is intended for a panel mount Molex but I couldn't find any. I also didn't bother to fit a heatsink to the PNP.

The holes in the PCB were quite tight for the resistors and the larger 330 Ohm resistor barely fit and had to be pulled through with pliers and due to its size had to be vertically mounted.

It certainly sped up the build time, at the expense of a approx £1 per PCB vs a 10s of pence for a length of strip board which requires cutting with hacksaw and drilling of tracks to lay out the circuit. It was much easier to solder and also limits risk of solder bridges and has far less stray capacitance.

If I were to sell these I'd probably have to make a 4 pin extension cable to go with each, otherwise an extra lead would be required.

So many thanks to Carlos for sending me these all the way from sunny Spain.

 

[Tealc]

FQP27.... something. I can't remember but it's written on the bag in the video where I control 10 fans. I took a screenshot but I can't read my own writing fully... As for the rest of the components it was largely the same, although I do seem to remember swapping in resistors somewhere.

It was selected because of low rds on and logic level operation. Goodness knows where the bag is now though to verify.

I'm currently looking at the Phanteks hub and seeing how that operates on heavy loads. Well, I will be when I get a chance.

 

[Tealc]

Good thinking....checks email.

FQP27P06.

P-Channel QFET® MOSFET -60V, -27A, 70mΩ

This P-Channel enhancement mode power MOSFET is produced using Fairchild Semiconductor’s proprietary planar stripe and DMOS technology. This advanced MOSFET technology has been especially tailored to reduce on-state resistance, and to provide superior switching performance and high avalanche energy strength. These devices are suitable for switched mode power supplies, audio amplifier, DC motor control, and variable switching power applications.

They are about 90p each and so are much more expensive than the B772.

I would like to point out that I picked this one only because of low RDS(on) and logic level and was just an experiment. If I'd spent longer researching I may have found a better component for the job.

As an aside the later versions of the Phanteks PWM hub use a D1M1425 P-ch Mosfet which is also logic level and has an even lower RDS(on) of 14mΩ.

I was trying to minimise voltage drop across the transistor and have less heat. A lot of the specifications of components are beyond my comprehension having never really studied electronics design properly.

 

[Tealc]

Sorry it is actually DTM4425. My eyesight isn't what it used to be.

Ooh that's a nice idea for an Arduino project. Might make one for myself. I have a great display for rpm and duty cycle.

There are rheostats out there and will cover several watts but they are a lot more expensive. It's generally better to use low power potentiometers and amplify using transistors.

 

[Tealc]

Hi. From time to time, yes.