Marlin 2.0 Fan PWM frequency too low
Posted by fdumi
|
Re: Marlin 2.0 Fan PWM frequency too low October 25, 2020 08:21PM |
Offtopic:
A RC low pass filter would presume the existance of a considerable amount of series resistance. The ESR of the capacitor and the wire resistance is not that much; demonstration below.
So let's say one implements a capacitor in parallel with a motor, which is driven by a low-side MOSFET, as it is in the case of the MELZI board. The diagram qould look something like this:
Here, L_wires and R_wires would represent the resistance of the wires from the fan to the power supply (not measured but the values should be in this range). V1 is the power supply and V2 is the controller output pin. I don't know what MOSFET is exactly on the PCB as it has a heatsink glued on top of it; I used a generic IRFZ44N for this simulation. Before I show the simulation results, let's do some calculations. Let's say the nominal supply voltage is 24V and the minimum voltage for the fan to operate is 8V (as per the datasheet). The current consumption is around 40mA measured and the frequency is around 976.6Hz measured. Therefore the required capacitor would be (calculations done in Mathcad, so the units are correct):
2.5uF is not a standard value and the next higher standard value is 3.3uF, but let's go with 2.5uF. Let's consider now that the 2.5uF capacitor is an electrolytic cap and has an ESR of 2 ohms (usual for a "normal" quality electrolytic capacitor).
Disregarding the ESR and considering a slow MOSFET switch-on time of 100ns, we get the following theoretical maximum current spike that the capacitor would draw for a charge from 0V to 24V, and from 8V to 24V:
Indeed these current values are theoretical because they assume ideal comoponents. With this calculation I just wanted to demonstrate that good quality capacitors can gulp a lot of current.
Now, for the simulation which is a bit closer to real life:
As you can see, even with "real" components we still get current spikes in excess of 10A. Yes, there are capacitors on the PCB that can supply that short current pulses, but at the cost of faster component degradation due to heating.
Does a capacitor in parallel with the fan solve the issue? Maybe, I have never tried something like this and I don't consider it a solution, as it is a bad idea.
Ontopic:
I still believe the SW implementation is not good. I actually searched on google for papers on PWM frequency for fan control and found that in PCs they use either really low frequencies (sub 100Hz) of higher than 20kHz so the audible noise is eliminated.
Quote
Ohmarinus
Won't it be more like a low-pass filter? Or like averaging out the voltage?
A RC low pass filter would presume the existance of a considerable amount of series resistance. The ESR of the capacitor and the wire resistance is not that much; demonstration below.
So let's say one implements a capacitor in parallel with a motor, which is driven by a low-side MOSFET, as it is in the case of the MELZI board. The diagram qould look something like this:
Here, L_wires and R_wires would represent the resistance of the wires from the fan to the power supply (not measured but the values should be in this range). V1 is the power supply and V2 is the controller output pin. I don't know what MOSFET is exactly on the PCB as it has a heatsink glued on top of it; I used a generic IRFZ44N for this simulation. Before I show the simulation results, let's do some calculations. Let's say the nominal supply voltage is 24V and the minimum voltage for the fan to operate is 8V (as per the datasheet). The current consumption is around 40mA measured and the frequency is around 976.6Hz measured. Therefore the required capacitor would be (calculations done in Mathcad, so the units are correct):
2.5uF is not a standard value and the next higher standard value is 3.3uF, but let's go with 2.5uF. Let's consider now that the 2.5uF capacitor is an electrolytic cap and has an ESR of 2 ohms (usual for a "normal" quality electrolytic capacitor).
Disregarding the ESR and considering a slow MOSFET switch-on time of 100ns, we get the following theoretical maximum current spike that the capacitor would draw for a charge from 0V to 24V, and from 8V to 24V:
Indeed these current values are theoretical because they assume ideal comoponents. With this calculation I just wanted to demonstrate that good quality capacitors can gulp a lot of current.
Now, for the simulation which is a bit closer to real life:
As you can see, even with "real" components we still get current spikes in excess of 10A. Yes, there are capacitors on the PCB that can supply that short current pulses, but at the cost of faster component degradation due to heating.
Does a capacitor in parallel with the fan solve the issue? Maybe, I have never tried something like this and I don't consider it a solution, as it is a bad idea.
Ontopic:
I still believe the SW implementation is not good. I actually searched on google for papers on PWM frequency for fan control and found that in PCs they use either really low frequencies (sub 100Hz) of higher than 20kHz so the audible noise is eliminated.
Quote
fdumi
Has someone had this issue, or is there someone with more experience with the Marlin source code who can guide me to where/what should I change in the source code?
Sorry, only registered users may post in this forum.