Using ESP8266_new_pwm to improve performance

[diffstorm]

Hi,

Can you add the following library to improve pwm performance?

https://github.com/StefanBruens/ESP8266_new_pwm

Problem:
I'm using 3 pwm outputs to drive an rgb led and sometimes I see kind of flickering (leds goes off/black then back to its color momentarily).

Init code:

  pinMode(PINR, OUTPUT);
  pinMode(PING, OUTPUT);
  pinMode(PINB, OUTPUT);
  digitalWrite(PINR, 0);
  digitalWrite(PING, 0);
  digitalWrite(PINB, 0);
  analogWriteRange(255);
  analogWriteFreq(100);
  analogWrite(PINR, 0);
  analogWrite(PING, 0);
  analogWrite(PINB, 0);

Drive code:

void ICACHE_RAM_ATTR SetRGB(int rr, int rg, int rb)
{
  os_intr_lock();
  analogWrite(PINR, rr);
  analogWrite(PING, rg);
  analogWrite(PINB, rb);
  os_intr_unlock();
  yield();
}

Flash: QIO - 40MHz - 1M (256K spiffs)
CPU: 160MHz

[WereCatf]

Have you actually checked if it is any faster or not? That code is a replacement for the SDK PWM-code, but the ESP8266 Arduino-core doesn't use the SDK PWM-code.

[diffstorm]

I didn't know that, I didn't test it. It claims that is better than SDK PWM.
So what can I do for the problem, please write if you have suggestions. I use only 3 pins for PWM.
Thanks

[diffstorm]

Hi, I have tried the ESP8266_new_pwm on my system and indeed it works better. I don't see any flickering on LEDs.

[WereCatf]

That's good to know. I'll give it a try myself, too, as soon as I have the time.

[diffstorm]

You can get help by reviewing the link below. I have already asked about implementation and other useful questions;

StefanBruens/ESP8266_new_pwm#5 (comment)

[ghost]

wouldn't it be better to get it included in the SDK if it is better? or do they not want it?

[WereCatf]

@jrobeson The ESP8266 SDK isn't open-source, unlike the ESP32 SDK. You can, of course, ask them if they would like to improve on their PWM-functions, but it doesn't seem likely. Also, like I said, this Arduino-core for the ESP8266 doesn't use SDK-functions for PWM, so it'd have to be modified anyways.

[ghost]

why doesn't it seem likely?

[WereCatf]

@jrobeson Because they've been told about it several times. And their SDK has been sitting idle since March.

[devyte]

@WereCatf did you ever get around to testing that new pwm?

[devyte]

@diffstorm does the new pwm still have merit vs. the current pwm implementation?

[WereCatf]

@devyte No, I completely forgot about this whole discussion. For my own purposes, the PWM-mechanism that's already included has worked fine, but then again, I haven't used it for anything demanding.

[diffstorm]

@devyte I think current PWM implementation works better than the one in the link.
BTW, I still see flickering when brightness is lower than 100% in rainbow mode. I couldn't fix that issue, that leads me to think its a HW or SDK problem.

[WereCatf]

To be honest, to properly gauge if it is better, worse or equal in performance, someone would have to capture multiple proper oscilloscope-output at different frequencies and duty-cycles on both implementations, and no, a single capture at a single frequency ain't enough. Until someone provides such proof one way or another there's no point in debating this topic.