Change Timer1 to Timer3 on ATmega32u4-based dev boards

[CoolBots]

As of version 1.2.1 of the library, ATmega32u4-based boards are defined to use Timer1 raher than Timer3. This configuration conflicts with millis() and micros() functions. While not too difficult to change, it does make the library more difficult to use in educational settings (we're a robotics/programming school for kids). If possible, could Timer1 be changed to Timer3 for these devices? The specific device we are using is an Adafruit ItsyBitsy 5V.

It would be no issue to submit a PR, of course, provided there isn't a strong reason to keep it as Timer1, as technically this would be a breaking change (although it is likely 32u4 users of this library make this modification anyhow, as millis() and micros() are useful functions...). Likewise, if there is a workaround that is fairly straightforward and does not require modification of the library that we are not aware of, please let us know.

It is worthwhile mentioning that we have seen other examples of this choice of Timer1 instead of Timer3 being an issue on 32u4 boards, most notably from Pololu Robotics and Electronics: https://www.pololu.com/docs/0J57/8.b

CodeRef:

Servo/src/avr/ServoTimers.h

Lines 41 to 43 in 4def8a2

	#elif defined(__AVR_ATmega32U4__)
	#define _useTimer1
	typedef enum { _timer1, _Nbr_16timers } timer16_Sequence_t;

[per1234]

Hi @CoolBots. Thanks for your report.

As of version 1.2.1 of the library

This statement seems misleading to me. No change to the timer used by ATmega32U4-based boards was introduced in the 1.2.1 release. In fact, the last relevant change was made back in 2012:
arduino/Arduino#102

and released in Arduino IDE 1.0.2 (at that time the Servo library was distributed as a part of the IDE installation), and in the 1.0.0 release of the standalone Servo library after the library was split out from the IDE project.

[CoolBots]

My apologies, I merely meant to tie the issue to the version of the library I discovered it in. You're right though, it's misleading, as it sounds like a new issue in this version, which is not the case. I'll be more careful with my wording.

After reviewing the change you referenced in 2012, I realize that Timer3 was removed to avoid conflict with tone() on ATmega32u4. Simply changing to that timer will fix one issue at expense of another. I'm open to suggestions; if no practical solution exists due to lack of timers, this issue should be closed as unresolvable.