You want an asymptotic approach with minimal overshoot, if possible, but you need to adjust the timings and bands and nudge sizes according to your setup. So you start the cycle at 50% wait 30 secs, if measured is 49 (11 off) so increase pwm by 25 % wait 30 secs, measure 45 (7 off) so increase by12 %, wait 30 secs, measure 43, (off by 5 so another 12 % (now at 99%} wait 30, measure 41 (3 off) so add 4% (99 +4 must top out at 100) so run at 100, wait 30secs temp is now 36 (so 2 off, reduce by 1 - > 99) wait 30, temp is still 36 reduce by another 1 etc down at less than 1 degree offset make no adjustment Having a target of 38 when if you run the fan at 100% it only gets down to 44 is pointless so you need to set realistic targets and have a fan big enough to achieve it. So let’s assume a target of 38 degrees, and bands at 10, 6, 3 and 1 degrees with each having a percentage of 25, 12, 4 and 1, Okay so your pwm is (say) 100 hz, so 100% means that the 1st pulse is on for the whole 1/100th second and is immediately followed by the next pulse. I’ve worked with even some large scale multi-variable control systems that haven’t performed ‘too well’ and later been replaced with ‘nudge and wait’.Īll these numbers are arbitrary, start with ‘some’ numbers and adjust as you go along. Not sure how you’d implement PID on the hardware you mention, probably by ignoring any Integral and Derivative components and just using the Proportional to determine ‘Nudge’ Direction (though that would work, but it’s hardly PID ? any counter to this I’d be impressed with). You are also ignoring that this is Tobi’s solution and he has obviously not had any issues with it. There would be no harm in Jan putting a blocking diode in and you putting a reverse diode in (you weren’t arguing the same point) but as an Electrical Engineer I would say that neither would add any benefit and his would just reduce the voltage (that AND PWM would increase iron losses). Then rinse and repeat with the transistor.īut then you want to exacerbate the switching situation by moving to PWM, It will work, but it does undermine your argument.

Power a 5v fan (with a volt meter across the 5v leads) then disconnect the power, see what happens. Have you worked much with decelerating DC motors, especially in a mechanical load context like this ? (fan drag) Do you know what the english translation of his phrase was ?Īnd how do you think the transistor will react in that circuit ?

Flirc raspberry pi case wireless