The bench
Each lesson opens one board with a lamp on its PWM pin and a frequency counter (0.1 s gate) clipped on the same pin, plus a second counter where a second output matters. Every frequency a step checks is read by a counter, so a step passes on what the pin really does. The project opens all four boards at once, each with a lamp and a counter on its tone pin.
Take the course at Timers, PWM and capture on four chips on /learn.
What each sketch does
tmrbasics(Uno): Timer1 in CTC mode toggles pin 9 once a second and interrupts on every match.tmravr(Uno): Timer1 fast PWM mode 14 at 1000 Hz and 25 percent on pin 9, beside analogWrite()'s 490 Hz on pin 3.tmrledc(ESP32-C3): LEDC at 5 kHz and 10 bits on GPIO 4, a 440 Hz tone on GPIO 5, and a frequency and resolution the hardware refuses.tmrpico(Pico): PWM slice 1 by register, GP2 and GP3 sharing a frequency, and the 8.4 fixed-point divider.tmrstm(Black Pill): TIM3 channels 1 and 2 on PA6 and PA7, and center-aligned counting.tmrcapture(Uno): Timer1 input capture on pin 8 with an overflow count, as a reciprocal counter, measuring a tone from pin 3.tmrdrvuno,tmrdrvc3,tmrdrvpico,tmrdrvstm: the project, the same pwmBegin, pwmDuty and pwmTone calls on each chip's own timers.
Honest limits of the simulator
- A function generator's square wave reaches an MCU pin here on a grid of about a millisecond, so input capture of a funcgen signal reads wrong in the simulator. The capture lesson measures a digital signal from the Uno's own pin 3 instead, which is exact, and keeps the frequency counter as the reference.
- The Uno's Timer1 external clock input (T1, pin 5) is not modeled, so no lesson counts edges with it.
- The RP2040's phase-correct mode and edge-counting mode are not modeled; the Pico lesson uses the fractional divider instead.
- The frequency counter reads edges exactly, with no noise and no time base error.
Related
Bare-metal STM32 builds TIM3 one register at a time, and Scheduling and RTOS uses timers to schedule tasks.