SpenceKonde
commented
3 years ago Second trace is correct behavior, when reducing the duty cycle, changes will never take effect until the start of the next pwm cycle.
First trace is the sort of glitch I thought might occur in one of the other threads, but he specifically never went all the way to 0, so that particular situation (non-pwm to pwm) was excluded. While there may be some simple half-measure mitigations to make the 0-to-non-0 case non-perverse. possible in time for next release, any sort of serious mitigation will need to wait until the one after that - lot of demand for fixes in current release to be rolled out to board manager, and cannot justify such a delay.
ldalep
This is a follow on to issue #261 I set up the following sketch using PIN_PA3 as a PWM pin with a trigger/marker digital pin PIN_PA1. I ran the sketch twice changing the the PWM pin sequence from 0,1,200 to 1,2,200 with delays of 25ms, 5msec, 5msec. This was done at 20Mhz and Default timer, Code 2.1.5. In both cases I ran a trace (Sampled at 12Mhz.... kinda slow) which are shown below
0,1,200 - Note that extra Pulse width on the transition from 0-1. Also there is an immediate transition to 0 level when going from 200 to 0.
1,2,200 - Note the difference from 200 -1
unsigned long t = 0; byte output_pwm_pin = PIN_PA3;
void setup() { pinMode(output_pwm_pin, OUTPUT); pinMode(PIN_PA1,OUTPUT); }
void loop() { // put your main code here, to run repeatedly:
PORTA.OUT |= PIN1_bm; // High Marker and Trigger analogWrite(output_pwm_pin,0); // Change here for "0 to 1" t=micros(); while ((micros() - t) < 25000) ;
PORTA.OUT &=~PIN1_bm; // Low analogWrite(output_pwm_pin,1); // Change here for "1 -2" t=micros(); while ((micros() - t) < 5000) ;
PORTA.OUT |= PIN1_bm; // High Marker PORTA.OUT &=~PIN1_bm; // Low
analogWrite(output_pwm_pin,200); ///blink at about 200 pwm t=micros(); while ((micros() - t) < 5000) ;
}