diff options
Diffstat (limited to 'docs/library')
| -rw-r--r-- | docs/library/machine.PWM.rst | 31 |
1 files changed, 31 insertions, 0 deletions
diff --git a/docs/library/machine.PWM.rst b/docs/library/machine.PWM.rst index f2273d8b4..4c72255d8 100644 --- a/docs/library/machine.PWM.rst +++ b/docs/library/machine.PWM.rst @@ -77,3 +77,34 @@ Methods With no arguments the pulse width in nanoseconds is returned. With a single *value* argument the pulse width is set to that value. + +Limitations of PWM +------------------ + +* Not all frequencies can be generated with absolute accuracy due to + the discrete nature of the computing hardware. Typically the PWM frequency + is obtained by dividing some integer base frequency by an integer divider. + For example, if the base frequency is 80MHz and the required PWM frequency is + 300kHz the divider must be a non-integer number 80000000 / 300000 = 266.67. + After rounding the divider is set to 267 and the PWM frequency will be + 80000000 / 267 = 299625.5 Hz, not 300kHz. If the divider is set to 266 then + the PWM frequency will be 80000000 / 266 = 300751.9 Hz, but again not 300kHz. + +* The duty cycle has the same discrete nature and its absolute accuracy is not + achievable. On most hardware platforms the duty will be applied at the next + frequency period. Therefore, you should wait more than "1/frequency" before + measuring the duty. + +* The frequency and the duty cycle resolution are usually interdependent. + The higher the PWM frequency the lower the duty resolution which is available, + and vice versa. For example, a 300kHz PWM frequency can have a duty cycle + resolution of 8 bit, not 16-bit as may be expected. In this case, the lowest + 8 bits of *duty_u16* are insignificant. So:: + + pwm=PWM(Pin(13), freq=300_000, duty_u16=2**16//2) + + and:: + + pwm=PWM(Pin(13), freq=300_000, duty_u16=2**16//2 + 255) + + will generate PWM with the same 50% duty cycle. |