I’ve been wondering perhaps a bit too much about PWM motor speed control of brushed DC motors. Does it really work? Why should it work? Won’t it not work as well as motor voltage control? After lots of reading and thinking through this carefully, I think I can explain this in a way that is sure to sway skeptics like myself.
Earlier, I’ve already read that the Raspberry Pi uses some sort of PWM control to generate its built-in analog audio output. Yeah, that makes sense, as the original Macintosh did that too. Just earlier today, I’ve read in detail exactly how it works, and I’ve explained why it is better than the original Macintosh’s way of doing so. But, here is the most interesting specific part pertinent to our discussion on motors. A PDM audio signal can be trivially converted to an analog signal via an RC (resistor-capacitor) low-pass filter circuit. Indeed, simply by storing up the pulse charges in a capacitor, you can blur out the waveform by averaging its values over a larger area to get a nice analog signal.
Okay, so that’s the key that you need to know related to motors too. Yes, we do know that motors, being moving mechanical devices, have inertia and momentum such that once a motor is set in motion, it will keep moving. So, if you have some jittery pulsed signal to slowly get a motor started moving, it will keep moving at a certain pace that is slower than maximum speed when you keep feeding it the same jittery signal. But why should we think that this jittery signal will spurious vibrations in the motor that you otherwise wouldn’t experience were you to feed it a continuous, lower voltage signal? Well, as we’ve said before, a low-pass filter circuit can be constructed using resistors and capacitors, the capacitor’s function being to store up charge so as to smooth it out. A brushed DC motor is also an inductor, and as it turns out, that means that it also has the capability to store up energy on its own, albeit in a different form. Namely, it forms a residual electromagnetic field that causes current to keep flowing even when you switch off power to the motor. Likewise, a motor is also a form of resistor due to its high current draw and energy dissipation. There you go, you’ve got the basic components of an RC low-pass filter right inside your motor.
With that in place, your only other concern for constructing an RC low-pass filter circuit to convert a PWM signal to an analog signal is that the PWM frequency is well matched with the roll-off frequency of your low-pass filter. Suffice it to say, a motor has a relatively low roll-off frequency, so if you have a PWM that is good enough to generate an analog audio signal, it is more than good enough to get smooth motor motion.
Now, we know that motors must be connected to microcontrollers via H-bridge motor controller integrated circuits, and this is where things get interesting. An L293D H-bridge motor controller has built-in diode protection to allow the back-emf, the current that wants to keep traveling that is caused due to the stored electromagnetic field, to travel through an alternate safe path rather than blowing out the microcontroller that tried to switch off the motor. When you simply disconnect the power and allow the back-emf to flow freely, this is called coasting since the motor is allowed to continue rotating relatively freely on its own. But, an H-bridge motor controller circuit can also be configured to provide a sense of drain or resistance to the back-emf, wnich is called braking as it causes the motor to slow down faster. So, the point here is that if you use PWM to switch back and forth between coasting and powering relatively quickly, you should get some pretty nice smooth motor motions without spurious vibrations introduced. But, if you switch back and forth between powering and braking relatively quickly, that is going to cause much more of a vibration problem. Additionally, another useful control scheme is to use PWM to switch back and forth between coasting and braking, in effect giving you variable control over the braking power. Of course, with electric motors, you can also power in the opposite direction as an even stronger form of braking.
Notably, the choice of diode also affects the degree of coasting that the motor experiences. A diode that allows a more generous flow of back-emf will allow for the greatest coasting with least resistance, but you can also select a diode that does not allow as much back-emf through. As stated in the Wikipedia article on the Flyback diode, this results in a speedier slowdown, but it also causes more stress on your microcontroller’s switch.