8

I see a lot of hardware people around here so I thought I'd ask..

I have 5 fans in my PC that are all voltage controlled (3 pin connector). I bought by mistake a fan hub that only supports pwm fans (4 pin connector). They are backwards compatible but the fans will run at 100%

I tried to make a circuit I found online to convert from 4 to 3 pin but it didn't work. There's a lot of comments on those forums and I honestly don't understand anything I read there... I don't read schemas well :/

I'd be very grateful if someone here knows what I need to do. I can solder well, but I need details on what parts to get (how many ohms resistors need to be, which wire goes where, etc...)

Comments
  • 2
  • 5
    My experience with PC fans is limited, but my hunch is that the 4 pin fan hub is probably using RX and TX serial communication (called Sense and Control apparently). So I don't think that a PWM fan will be likely to be fully compatible with such bidirectional drivers. Perhaps you could make a conversion board but that'd include microcontrollers and stuff, be complicated and just purchasing a compatible fan hub would be far easier I think :/
  • 2
    You need to make the PWM output regulate an analogue voltage that in turn drives the 3 pin fans:

    https://electronics.stackexchange.com/...
  • 2
    @Condor sense is the feedback from the fan, how much RPM it actually does. Control is how much it should do. No serial communication, just rectangular signals.
  • 5
    @Fast-Nop That'd be quite inefficient though, as regulators drop several volts and dissipate it as heat. Fan motors on the other hand can draw upwards of 1A easily, causing the regulator to heat up too much. Not only that but the fan would also be unable to run at full speed, because the regulator decreases the available supply voltage.

    Edit: Just noticed that it's a switching regulator apparently, rather than a linear one. Yeah that could work I think.
  • 4
    @Condor buck converters typically have above 90% efficiency, so that's no big deal. The drop could be a problem, though.
  • 6
    @Fast-Nop Personally I'd drive the PWM pin with no more than a simple PWM drive like the the 555 timer though. I mean that signal pin probably won't be responsible for sourcing the current anyway (and if the fan draws too much, just add a resistor in between). The fan's controller can then switch the contact to the supply on and off depending on that, with perhaps even a driver written for the OS that reads the temperature sensors and communicates with the timer to change the duty cycle based on that. And if the fan doesn't switch the supply connection on and off based on the PWM signal (which would be strange), a MOSFET could be employed for that I guess. Switching regulators sure are an interesting thing though.. never heard of those. I kind of assumed that all regulators would be linear :') the more you know!
  • 3
    What power rating do you fans have? If it's low enough, I'd du it linearly, because the frquency of the PWM signal is probably way too low to be used for a switching regulator directly.
  • 2
    @Condor here's the circuit I tried. All it did was make the fan run at lower speed, without reacting at all to pwm (couldn't control it from software)
  • 1
    📌
  • 1
    📌

    I fucking suck at electronics and I fucking love reading such debates.

    Kudos and kisses to all you
  • 4
    @vlatkozelka hmm, so P1's pin1 is 12V, pin 4 is PWM (Control?).. wait why the fuck are they also using the Sense pin in the 4-pin connector? Doesn't make any sense.
    (Edit: Sense pin doesn't make any sense.. hah 😛)

    Personally I don't think that there's an easy way out that doesn't imply discarding the PWM from the 4-pin connector (which would be a waste but hey, it's only compatible to some extent), and then creating your own (with something like aforementioned 555). You could manually adjust its duty cycle with a potentiometer or let the PC do it for you based on the readings from the temperature sensors. But then, it'll probably be way easier to just purchase a compatible fan hub and get fan speed control that way 🙂
  • 1
    @Condor The thing is, the 3 pin fans only have plus, ground and RPM feedback. They can't use the PWM pin to switch their voltage, that's what a 4 pin fan would do. So a 3 pin fan alsways gets the 12 V to its plus and ignores PWM - that means 100% speed. Sure, you can use a resistor to reduce speed, but then it can't crank up when it's hot.

    I'd think about just plugging the PWM output to a MOSFET that interrupts the +12V and put a capacitor as low pass, 100µ in elko in parallel with 100n ceramics. You'd have to measure the voltage level of the PWM and get a MOSFET with suitable gate voltage level.
  • 0
    @Condor the sense output goes to the mainboard which can detect whether the fan is stuck, in which case the BIOS can alert beep.

    That can happen after a few years when the grease in the fan bearing is thickened - the fan will run, but when you switch off the server and things cool down before restart, the initial mechanical resistance is too high. Then you need to push the fan a little with a pencil to get it going (BTDT).
  • 2
    @Fast-Nop: That's a great way to build a inefficient, EMI radiating circuit. If you wanted a step down, you'd need an inductor, and because the PWM frequency is very low, it'd need to be a fairly large one. That's exactly why one person in the SE link above suggested controlling a dedicated switching regulator (operating at a much higher frequency) by the incoming PWM signal and I suggested to do it linearly if the fan power allows it.
  • 0
    The PWM is 25 kHz, and even with some harmonics, that's still nearly DC. I wouldn't worry about EMI.
  • 4
    @Fast-Nop or you know, just clean it up 🙃 no electronics can solve lack of maintenance of dirty fans.

    Regarding the 3-pin fan, so the feedback gives fan speed and then the VCC voltage is adjusted by the motherboard I guess?
  • 2
    @Fast-Nop: Good to know, I would have guessed it was lower and you'd actually hear the radiation in your speakers. But still, I wouldn't feel comfortable doing that. From a losses point of view it isn't any better than a linear regulator, anyway.
  • 1
    @7400 but the point with the ineffeciency would need research. 25 kHz interrupting 1A, that's something. Usual PWM for displays has about 1 kHz if it's a good one.

    Of course, it also depends on how steep the PWM slope is because that gives how long the MOSFET will be in the non-saturated area where it actually generates heat. I can agree that this is a quick & dirty hack, but could be OK for tinkering around a bit.
  • 0
    @Condor it was clean. The grease itself was thickened because after years, some substances evaporate and leave the thicker rest behind. Changing the fan would be best then - unless it's a big corporation where buying anything take so long that you need semi-socialist tinkering.

    Yeah, that's how 3 pin fans work. Which is also why it's difficult to get 3 pin fan hubs at all. Actually, I would spare the effort, throw out the 3 pin fans and just by 4 pin ones.
  • 0
    @Fast-Nop: If you charge a capacitor over a resistor, you're going to lose half the energy in the resistor, regardless of the resistance. This is essentially what you're doing here (and it's related to the ‘capacitor paradox’).
  • 0
    @7400 true, but the energy in the capacitor itself is neglectable with 100µ and 12V. Even if half of that energy is wasted, it's still nothing. Most of the energy will not flow into the capacitor, but into the fan, and if it doesn't because the fan doesn't draw current, then the capacitor will be charged only one time.
  • 5
    @Fast-Nop aha, please discard my other comments then.

    So a PWM pin is present in the 4 pin connector, and sense is present in both. In that case I'd connect both sense pins, VCC to the drain of a FET, and the PWM to its gate.. run the source through a capacitor and a coil to smoothen the supply voltage perhaps. A dirty hack and at high frequency the FET may heat up just as much as a linear regulator would, but meh. For fans I'd probably also go for 4-pin ones if that's what the board supports. And fan hubs.. connecting 1 fan to the motherboard's supply is one thing, but connecting 5 of them.. I wouldn't be sure if the motherboard can even handle that.
  • 1
    @Condor it takes power directly from the power supply not the motherboard. The motherboard is only used to control the fans and read their speeds.

    I guess I'll just order a compatible one, they are really cheap, it's just waiting for them for 2 weeks that sucks (Amazon doesn't deliver here, we do it via delivery companies like ShopAndShip)

    Glad you guys had some fun debating this :p
  • 1
    @Fast-Nop: Let's view it from another perspective: You included tha capacitor for filtering, so because you want to keep the voltage for the fan relatively stable. That makes sense as you wouldn't want to supply it intermittently with a 25 kHz rectangle.

    Say you want to run it at 50%, so 6 V. 0.1 mF is enough to assume a constant voltage of 6 V at tge fan for simplicity. Now you have a MOSFET on 50% of the time, with 6 V voltage drop and double the current the fan needs. That's exactly the same power dissipation as a linear regulator would have.

    Granted, if you decrease the capacitance, it is more efficient. But it was there for a purpose after all, wasn't it?
  • 0
    @7400 well in that case, switching regulators with capacitor for ripple minimisation would never exceed 50% - but they do. And that's because the energy in the capacitor is only a small part of the deal. Basically, it's only the difference between the ripple peaks (down and up) that counts for charging the capacitor. The constant minimum part of the voltage gets loaded into the cap only one time.

    However, the total fan power counts from the total voltage, not only from the ripple.
  • 1
    @Fast-Nop: Well, they wouldn't if they didn't have an inductor. But they do, and that's another energy storage device to consider. In fact that matches exactly what you see with switching regulators: Switching losses increase as the inductance decreases. With the inductanve approaching zero, you have exactly the case I described above.
  • 1
    @7400 that makes sense, yeah. I just calculated what would happen in the extreme case with always discharged cap, 100µ, 25 kHz and 12 V. Turned out we'd be speaking about 180 W - I had mis-estimated the possible order of magnitude.
Your Job Suck?
Get a Better Job
Add Comment