FEATURED Building PWM Controller for 4 wires PWM fan

jalyst said:

You might be speaking too soon...
xXSebaSXx hasn't shown us what his Nidec fan/s can do yet!

Click to expand...

Give me a week or two... The components for the PWM controller are on the way, but now I have to figure out the PWM frequency on that Nidec before I start building the controller so I can generate the right signal to it... Once I have a definite answer on the frequency I will order one for testing; can't order 4 like LennyRhys did with the San Aces ~~~the Nidecs are $100 a pop and I'm not flush with cash by any measure~~~ So one will have to do for now, but if it turns out to be all that it is advertised to be; I will for sure get at least one more in the future.

xXSebaSXx said:

@ bing...
And since we're getting technical... I'm reading up on the Nidec Servo fan that I'm planning on buying and can't find the PWM frequency that the thing is rated for anywhere on the datasheet.
This is the only mention of the PWM speed control on the entire thing...
Can you make out the PWM frequency from that diagram? I can't.

Click to expand...

About the freq, Bobnova is right and I missed that part, I was assuming this monster is working on Kilo hertz range, my mistake and really sorry since you've made the purchase, I should have spot it earlier.

I put few markings at the pic here from it's datasheet, that big red arrow is pointing the "t" is the time length for single period, and from the standard formula -> Freq = 1 / PeriodLength , we got frequency range from 500Hz up/to 5 Khz. This is quite low frequency and really curious on how is the noise profile at slow speed when using low freq pwm.

From G1238B12BBZP-00 datasheet
View attachment 88969

With this much lower frequency from 25KHz which I set at the original circuit, yes, it needs adjustment at the capacitor value. Is there RadShack nearby ? I think they have these kind of cheap discrete components right ?

For the 2-98% version , the C1 needs to be replace with :

• For 500 Hz -> 22 nF
• For 5 KHz -> 2.2 nF or 2200 pF

While for the 0-100% circuit, the R1 & R2 also the C1 need to be replace with these following values :

• For 500 Hz -> R1 = 20K , R2 = 33K and C1 = 33 nF
• For 5 KHz -> R1 = 20K , R2 = 33K and C1 = 3.3 nF

Just change the C1 value between those two values to get the freq between 500 to 5KHz if you want to experiment at other freq.

Actually for this kind of circuit driving the pwm fan, the frequency accuracy is not that important, so imo for the 0-100% version, if you want to keep using the old value for R1 and R2, just in case radshack doesn't carry them, while changing only the C1 part is fine too, the frequency range will be about 460 up to 4.6 KHz which is quite there.

If you see the chart for the PWM vs rpm, you don't need the 0-100% version, at lowest speed it only needs approx 10% duty cycle, while the top speed at 90%, so the 2-98% version is more than enough and much simpler to build too.

LennyRhys said:

Thanks BF. It's really intriguing that despite out circuits being virtually identical your setup behaves in an entirely different manner from mine. My controller won't stop the fans, and the whining noise is especially bad at the lowest duty cycle.

All the fans are the same - after trying the first fan (at this point I was using the 680pF cap) I thought that maybe there was something wrong with the motor, so I tried the others. Same noise.

I tried your suggestion of 820pF in series, and 390 is still a lot better. I've tested all my caps from 330pF right up to 1K pF, and 390 is by far the best, however as you say it is not the most efficient. One compromise would be using the 470pF cap but running the fan at around 10-15% duty cycle; there will be more air/motor noise, but no whining, and better efficiency.

Re. RPM sensor, it's totally borked - up to 3K rpm it's fine, but above 50% duty cycle my software gives me a reading with seven digits LOL

I will hopefully get started on my article next week, and I'll try and get another video together in the meantime.

Click to expand...

Lenny, this is really strange, again really sorry, I just realized at that video, that fan didn't stop at lowest duty cycle, and usually 2% is pretty low enough to make it stop like BrutalForce has.

I suspect the circuit is not working as it should be, especially you can't read the rpm accurately and those high pitch whine, assuming the fan's internal is working, I suggest you re-verify all component's value and recheck the wiring/routes at the breadboard.

Also you can try manually force the duty cyle to perfect 0% by connecting the PWM wire to Ground wire to see if the fan works as it should, this effectively will stop the fan as it's specification. Becareful though, do not accidentally connect the pwm wire to the 12 Volt power line, worry this might toast the fan's internal circuit.

While leaving the pwm wire left unconnected will be the same as at full 100% duty cycle, and it should rotate at it's max speed.

Thanks bing. Here's a thought - remember I fried the last circuit? Well I took the 0.01 uF cap from that circuit because I had none left. What I'll do right now is replace that cap with a new one. All the other components are brand new.

I'm going to get another 100k pot at some point and build a circuit with the 556.

LennyRhys said:

Thanks bing. Here's a thought - remember I fried the last circuit? Well I took the 0.01 uF cap from that circuit because I had none left. What I'll do right now is replace that cap with a new one. All the other components are brand new.

I'm going to get another 100k pot at some point and build a circuit with the 556.

Click to expand...

Yeah, that sounds like a really good plan.

Keep us informed.

Just tried it with a new 0.01 uF cap, no difference - still really bad whining noise at low rpm, and I can't stop the fan with the pot (curretnly using a 470 pF cap).

I also forgot to mention that I have only one 100k pot, the one that was connected to the PWM circuit when I fried it - might it be the pot that is causing the problem? I checked all the solder joints and components etc and everything looks absolutely fine and works well.

I already checked the pot with a DMM and seemed to be working fine with a range of about 0 - 92.

This stage of circuit building is very enjoyable, but at times extremely frustrating!

bing said:

Also you can try manually force the duty cyle to perfect 0% by connecting the PWM wire to Ground wire to see if the fan works as it should, this effectively will stop the fan as it's specification.

Click to expand...

Just tried this, and the fan stopped.

I will try to put another circuit together with the 556 on Monday!

Did you include the inverted schmitt buffer?

Before I installed that I ended up with a weak PWM signal. Bing had me build in a Schmitt buffer, hence the TLC 556 instead of a 555.

I only see one IC 555 in yours. Just thought I would throw this out there.

Hey thanks for the explanation on the PWM frequency of that Nidec bing... I have a Ratshack just a few blocks away from my house... Still waiting on parts from Newark. They should have arrived today, but no such luck. I'll pass by the ratshack tomorrow and see if they have those parts in stock so I'm ready for when the rest of the stuff comes in... I needed to go there to get a breadboard and some jumper wire to set a test run of the controller before making it into a permanent circuit.

Brutal-Force said:

Did you include the inverted schmitt buffer?

Before I installed that I ended up with a weak PWM signal. Bing had me build in a Schmitt buffer, hence the TLC 556 instead of a 555.

I only see one IC 555 in yours. Just thought I would throw this out there.

Click to expand...

Yeah dude - the circuit has two 555s. At breadboard level I used only one 555 because at that point I didn't have the San Ace fans

Lenny, I've been thinking bout troubleshooting your problem, assuming the circuit is wired correctly and all components are in good shape, here few things :

Try change the R2 10K resistor to 1 K and if you still having the annoying whine and borked rpm reading, pull the R2, yes, no R2 at all and see the result. I watched your video repeatedly and realized there is no way sanyo made this quite a high end fan whines with that annoying high freq like that, it could be the pwm signal's fault.

Remember, one of the main purpose of using 4 wires pwm implementation at fan is to reduce noise significantly at lower rpm while maintain quite good efficiency compared to 3 wires fan

Other thing to do is, disconnect the pwm wire and measure the fan's current draw and the rpm, and it should be around 6000 rpm (+/- 10%), cause the current draw at your last measurement even using the right cap value (16Khz) is at quite low, it should be near 4 Amp at max speed and usually it will be around 3.9 - 4.1 A, that 3.8 Amp looks suspicious.

Btw, how did you measure the current draw ? Was that DMM wired in serial with the positive power line ? Cause this method sometimes at cheap DMM will not report an accurate value when measuring at quite high current, and also the fan might not received a full 12 volt, yep, the DMM was stealing (dropout) some voltage at it self in order to measure the current passing thru it.

Bing - thanks for the suggestions; I'll work on them today. I bought two more 100k potentiometers this morning so will try them too.

I did indeed wire the DMM in series with the +12v wire. Will also have a look at the current draw with PWM disconnected

Re. R2 - I don't have an R2 in the circuit because I'm using two 555s, not a 556. I'm breadboarding the 556 circuit now though and will let you know how I get on

OK I have built the 556 circuit at breadboard level. It works fine, but it is exactly the same as the dual 555 circuit in that it produces the same high pitched whine with various capacitor values.

I did as you suggested with R2:

• replacing R2 with 1K = no difference
• taking out R2 = PWM bypassed, fan at full speed

I also tested the power of the fan outside of the PWM circuit: it draws 3.48A max, measured again by DMM in series. In this scenario, rpm is reported at approx. 5300.

Something tells me the fans are to blame, not the circuit.

Tested all four fans just to be sure, and one of the fans is faulty but I will try to fix it later. It turns ok and works absolutely fine, but it makes a really bad clicking noise like something is loose inside.

Anyway, changing to a polypropylene or polypropene 680pF cap helps a little bit; it's definitely better than the ceramic disc caps, less whine. Do you think it could be the quality of the caps to blame? Unfortunately I have only 680pF of the better kind of cap; all my other values are ceramic disc.

I did use silver mica 470pF caps with mine, and the other one was a electrolytic.

Sad to hear that one of your fan is faulty, is that noise sort of crackling or grinding noise ? If it is, that is mostly the sign of damaged bearing , probably it was dropped hard or bumped at hard surface cause those precision ball bearing is really sensitive to high g shock.

I'm starting to suspect and worry that probably those fans were severely abused by previous owner since you bought them from e-bay, hoping this is not the case, while BrutalForce's fan was bought fresh & new.

About the cap, that ceramic type is fine for this particular application, as long they're within the correct value and not broken of course. Usually ceramic is not suitable for oscillator circuit, since they're not accurate (within 5 to 20% range) and also their value will drift as temperature change, but this PWM circuit doesn't need that kind of accuracy and stability. That polypropylene cap is definitely better, of couse the best is still silver mica type for it's stability against temperature, but again, this it is not utterly important.

Since you have EE in house , if I were you, I'd persuade him to bring this circuit to an oscilloscope which I guess he has it, or at least has access to it , cause its the ultimate tool to troubleshoot this circuit to see if the pwm signal is ok, yeah, abit overkill, but once its confirmed that the pwm controller is working, than it is easier to proceed to the next step.

Three of the fans are definitely brand new bing; the "faulty" one I am suspicious about, because it definitely looked used when I took it apart.

I didn't know exactly what I was doing so I took apart one of the good fans to see how it was different from the faulty one. The problem turned out to be relatively simple to fix: the magnet had come loose inside the rotor hub, and it was rubbing against the inside of the hub making a nasty scraping noise. I applied some strong craft glue and left it overnight and now the fan works like new

NAKED San Ace monster fan





Here is the magnet inside the hub after repair. You can see that the inside of the magnet (a tough rubber ring) has been slightly chewed from coming into contact with the electromagnets on the stator.



In the new year I will pick up some silver mica caps to try in the 556 circuit - I think that if they are better than the polypropylene caps, the whining noise might disappear



And here's the 556 on the breadboard

nice shots

Nice work Lenny, interesting fix. Sorry to hear one of the fans was a dud.

Thanks, really nice pics of that beautiful inner side of that beast, love the look of that tidy wounded 12 poles stator.
You might blow that dusty bearing with your camera's lens air blower.

See that components at the circuit board, compared to classic 3 wires fan, this is good example on the reason why pwm fan version is more expensive than their 3 wires brethren. Also for high speed with 12 poles stator here, the driver circuit is even more complex. Majority normal DC fan out there only have 4 poles, this design itself makes this fan a different league than those sissy fans out there.

Glad to hear you've fixed it.

Been a while since this thread was updated

I found time to make the circuit again, and I shrunk the 556 circuit to a more manageable size. I also got silver mica caps to see if they would help with the whining noise, but unfortunately it's still there.

For the time being I'm going to use a 330pF cap, which is fine because I hardly ever have the controller powered up (using a different CPU fan at the moment because I'm too busy with work and have no time to play )



And a vid of two monster fans controlled in tandem:

LOL that's beastly, I wonder what the 5400RPM Nidec's I'm considering will sound like at 75%?!
I'm thinking of getting the non-PWM version (Scythe GT) now, & maybe even slightly lower RPM.