esp32: Fix PWM not allowing frequencies < 611 Hz.

[robert-hh]

Tested with duty_u16(), duty() and duty_ns() for frequencies down to
1 Hz.
Fixes issue #8189

[robert-hh]

Changed, rebased & squashed. The names of #defines chosen by the initial implementation could be more distinctive, but left them as they were.

[IhorNehrutsa]

Bug is here:

micropython/ports/esp32/machine_pwm.c

Lines 87 to 92 in 30b6ce8

	// Possible highest resolution in device
	#if (LEDC_TIMER_BIT_MAX - 1) < LEDC_TIMER_16_BIT
	#define HIGHEST_PWM_RES (LEDC_TIMER_BIT_MAX - 1)
	#else
	#define HIGHEST_PWM_RES (LEDC_TIMER_16_BIT) // 20 bit for ESP32, but 16 bit is used
	#endif

Algorithm is developed in assumption that HIGHEST_PWM_RES must be 16 or 14
but LEDC_TIMER_BIT_MAX and LEDC_TIMER_16_BIT is enum (not a macro) and we got
HIGHEST_PWM_RES == 20

I tried to fix with SOC_LEDC_TIMER_BIT_WIDE_NUM, but SOC_LEDC_TIMER_BIT_WIDE_NUM also is not difined in Micropython compile time

This is right version:

// Possible highest resolution in device
#if CONFIG_IDF_TARGET_ESP32
#define HIGHEST_PWM_RES (LEDC_TIMER_16_BIT) // 20 bit for ESP32, but 16 bit is used
#else
#define HIGHEST_PWM_RES (LEDC_TIMER_BIT_MAX - 1)
#endif

and then lines

micropython/ports/esp32/machine_pwm.c

Line 299 in 30b6ce8

return ledc_get_duty(self->mode, self->channel) << (HIGHEST_PWM_RES + UI_RES_SHIFT - timers[TIMER_IDX(self->mode, self->timer)].duty_resolution);

micropython/ports/esp32/machine_pwm.c

Line 315 in 30b6ce8

int channel_duty = duty >> (HIGHEST_PWM_RES + UI_RES_SHIFT - timer.duty_resolution);

are right too.

[robert-hh]

With:
#define UI_RES_SHIFT (UI_RES_16_BIT - HIGHEST_PWM_RES)
the expression:
HIGHEST_PWM_RES + UI_RES_SHIFT - timer.duty_resolution
results in:
UI_RES_16_BIT - timer.duty_resolution
So care must be taken about duty_resolution being larger or smaller than 16.
By limiting HIGHEST_PWM_RES to 16 and consequently limiting duty_resolution to 16 that is forced with the previous code. Or the scaling is adapted like in my code.

[IhorNehrutsa]

@robert-hh
Yes, you are right, there are two ways. May try to compare both.

[robert-hh]

Tested with your version works fine as well on a generic ESP32. The Frequency deviation of 1001.6 Hz instead of 1000Hz stay. GENERIC_C3 and UM_FEATHERS2 show better frequency resolution and granularity.

The previous code worked also well on a UM_FEATHERS2 and a GENERIC_C3.
At the GENERIC_C3 down to 10Hz, at the UM_FEATHERS2 reliable down to 2 Hz, with a strange hiccup when changing from 10Hz -> 1Hz, but not when changing from 2Hz -> 1Hz.

[robert-hh]

In favor of @IhorNehrutsa's PR, being the initial contributor for this PWM module.