Ideal PWM frequency?
- Thread starter boriz
- Start date
BeanieBots
Moderator
Not really, because it's never as simple as that. Depends what you want to do. For example, you might want to either be spot on or completely avoid resonance.
With motors, the inertia of the motor/load, backlash and brush peak current also come into it.
As a general rule, avoid audio frequencies (when possible) with mechanical devices such as motors because the noise is simply annoying.
With lighting, avoid low frequencies which produce flicker.
With a peltier pump, avoid low frequencies because it reduces thermal shock.
And the list goes on.....
What's the application?
With motors, the inertia of the motor/load, backlash and brush peak current also come into it.
As a general rule, avoid audio frequencies (when possible) with mechanical devices such as motors because the noise is simply annoying.
With lighting, avoid low frequencies which produce flicker.
With a peltier pump, avoid low frequencies because it reduces thermal shock.
And the list goes on.....
What's the application?
And to add to that - if you go too high with the frequency then you may end up with switching losses due to things like input capacitance on mosfets. 10Khz is a place to start and maybe go to 20Khz if the whine is annoying.
For low pass filters, ballpark figures for a 1Khz lowpass filter are 10k and around 0.02uF. Smoothing depends on a compromise between filtering out the PWM vs the response time. You could use 10k and 22uF and it would filter just fine but would take a second or more to adjust values.
Look up the formulas for R C L and filters.
It depends on the application - driving a motor, creating a voltage for a sample/hold, driving a lightbulb etc. As BeanieBots says - what is the application.
For low pass filters, ballpark figures for a 1Khz lowpass filter are 10k and around 0.02uF. Smoothing depends on a compromise between filtering out the PWM vs the response time. You could use 10k and 22uF and it would filter just fine but would take a second or more to adjust values.
Look up the formulas for R C L and filters.
It depends on the application - driving a motor, creating a voltage for a sample/hold, driving a lightbulb etc. As BeanieBots says - what is the application.
No particular app.
I have used PWM several times for different things, with success. I tend to stay in the audio range because my MOSFETs are slow. But every time, I wonder how to establish the ideal frequency. It’s probably no big deal, just interested is all.
EG:
A motor coil is a bit like an inductor. Too low a frequency and the coil gets saturated during the first part of the duty, then the rest of the duty is just wasted power heating up the coil through it’s resistance. On the other hand, too high a frequency and the coil may never get fully charged up. The fact that it’s a motor with three coils + iron formers + brushes + rotary inertia etc. makes it more complicated. But there must be a way.
A similar situation arises with capacitative loads. Even a simple filament torch bulb must have some ideal frequency. Even if it’s ok to be an order of magnitude out.
I would be interested in finding a general method for establishing the best frequency for any app.
I have used PWM several times for different things, with success. I tend to stay in the audio range because my MOSFETs are slow. But every time, I wonder how to establish the ideal frequency. It’s probably no big deal, just interested is all.
EG:
A motor coil is a bit like an inductor. Too low a frequency and the coil gets saturated during the first part of the duty, then the rest of the duty is just wasted power heating up the coil through it’s resistance. On the other hand, too high a frequency and the coil may never get fully charged up. The fact that it’s a motor with three coils + iron formers + brushes + rotary inertia etc. makes it more complicated. But there must be a way.
A similar situation arises with capacitative loads. Even a simple filament torch bulb must have some ideal frequency. Even if it’s ok to be an order of magnitude out.
I would be interested in finding a general method for establishing the best frequency for any app.
You are right - it probably doesn't matter so much. The audible whine would be one of the bigger factors.
Re motors though, all the power ends up doing work, even if the frequency were 1Hz. What happens is that when the motor first starts the coil is a resistor (and an inductor), and yes, there are losses from resistive heating. But as the motor spins up to speed, there is a voltage that appears called back emf which counters the volts going in. You can even measure this in between power pulses. Take a 3V motor unloaded, spin it up to full speed and it will have almost 3V of back emf. This means the motor hardly draws any power, which is what you expect for an unloaded motor. So with pwm driving a motor you can't really have a frequency that is too low. (well maybe 0.1Hz with the motor stopping and starting but even 10Hz will work fine).
A motor is happiest at a certain load and speed and efficiency graphs are available for most motors. But if you are running a motor at its most efficient, it won't really matter if you are driving it with a 100Hz or 1Khz or 10Khz pwm.
Re motors though, all the power ends up doing work, even if the frequency were 1Hz. What happens is that when the motor first starts the coil is a resistor (and an inductor), and yes, there are losses from resistive heating. But as the motor spins up to speed, there is a voltage that appears called back emf which counters the volts going in. You can even measure this in between power pulses. Take a 3V motor unloaded, spin it up to full speed and it will have almost 3V of back emf. This means the motor hardly draws any power, which is what you expect for an unloaded motor. So with pwm driving a motor you can't really have a frequency that is too low. (well maybe 0.1Hz with the motor stopping and starting but even 10Hz will work fine).
A motor is happiest at a certain load and speed and efficiency graphs are available for most motors. But if you are running a motor at its most efficient, it won't really matter if you are driving it with a 100Hz or 1Khz or 10Khz pwm.
BeanieBots
Moderator
As with almost anything when you are not sure, consider the extreme cases.
You are correct in thinking that there must be an "ideal" in so far as infinitely low and infinitely high won't work. So there is an ideal somewhere in between.
Unfortunately, that "ideal" relates to many things other than just the load you are driving. Hence, the "ideal" is only ideal for the EXACT circuit you have.
For motors, it is as the Dr. describes. Don't get motor coil saturation and inductor saturation confused. They are quite different things. You shouldn't be able to saturate a motor if you keep within its current specs.
For an incandeswcent lamp, you need to consider the thermal properties of the filament. How long does it take to get to temperture and then cool down again? Quite slow for a 50W 12v halogen but quite fast for a 50W 240v mains bulb. After a few hundred Hz, there will be no perceived difference but as you get faster gate capacitance will play a larger part. So, probably anywhere between several 10's of Hz and several 10's of kHz depending on lamp and FET.
For LEDs it a different game all together.
You are correct in thinking that there must be an "ideal" in so far as infinitely low and infinitely high won't work. So there is an ideal somewhere in between.
Unfortunately, that "ideal" relates to many things other than just the load you are driving. Hence, the "ideal" is only ideal for the EXACT circuit you have.
For motors, it is as the Dr. describes. Don't get motor coil saturation and inductor saturation confused. They are quite different things. You shouldn't be able to saturate a motor if you keep within its current specs.
For an incandeswcent lamp, you need to consider the thermal properties of the filament. How long does it take to get to temperture and then cool down again? Quite slow for a 50W 12v halogen but quite fast for a 50W 240v mains bulb. After a few hundred Hz, there will be no perceived difference but as you get faster gate capacitance will play a larger part. So, probably anywhere between several 10's of Hz and several 10's of kHz depending on lamp and FET.
For LEDs it a different game all together.
BeanieBots
Moderator
Or maybe the highest frequency that does not cause other complications