[NT7S]

Version: 2.4.0 (head 8472d61). library.properties, library.json and the README changelog are set to 2.4.0.

Summary

Adds I2C bus selection so the Si5351 can sit on Wire1 or any other TwoWire bus (ESP32, RP2040, SAMD, STM32, Teensy and similar boards).

• New constructor Si5351(TwoWire *wire = &Wire, uint8_t i2c_addr = SI5351_BUS_BASE_ADDR).
• Integer address overloads (uint8_t, int, unsigned, long, unsigned long, long long, unsigned long long) keep Si5351(), Si5351(0x61), Si5351(0) and Si5351(NULL) compiling on the address path, as before.
• A null TwoWire * (nullptr or (TwoWire *)0) falls back to Wire.
• init() (begin() plus the device probe) and every register read and write go through _wire. Both constructors share one setup_obj() helper, so there is no delegating constructor.
• README: new "Selecting an I2C Bus" section and a v2.4.0 changelog entry.
• Separate fix: examples/si5351_vcxo now calls get_correction(SI5351_PLL_INPUT_XO). The old call failed to compile on every core.

Usage

#include "Wire.h"
#include "si5351.h"

Si5351 si5351(&Wire1);           // or Si5351 si5351(&Wire1, 0x61);

void setup()
{
#if defined(ARDUINO_ARCH_ESP32)
  Wire1.setPins(13, 15);         // ESP32: pick SDA/SCL before init()
#endif
  si5351.init(SI5351_CRYSTAL_LOAD_8PF, 0, 0);
  // Wire1.setClock(400000);     // set the bus speed after init()
}

Compatibility

• Default bus unchanged: the I2C transcript on the default bus is byte-identical to master (1ddd43e) across 693 scenarios x 2 register-file patterns.
• Si5351(97.0) (a floating-point address) is now a compile error because the overloads make it ambiguous. It used to convert silently.
• The constructors are not explicit, the same as before.
• begin(): init() still calls begin() with no arguments.
  • On ESP32 this keeps pins set with setPins(). If the bus is already running, it is left alone; the core logs "Bus already started in Master Mode", which is harmless and also happens on master.
  • On other cores begin() runs again and may reset an earlier setClock(), so call setClock() after init().

Cost

Uno, basic si5351_example: +138 B flash, +2 B RAM.

Verification

• Compile matrix: all examples plus a constructor-matrix sketch compile on the 9 boards Test Lab compiled for. This includes uno, esp32, mkrzero, rpipico, STM32 F411, nona4809, nano33ble and esp8266, plus ESP32/RP2040 sketches that use the second bus.
• Tests: issue-106 tests (t106) 23/23 pass, and the must-not-break suite 776/776.
• Hardware: verification by the maintainer is still pending.

Independent review

• Review 1 (fresh reviewer, 69ead04): CHANGES REQUIRED. There was 1 blocking finding (Si5351(0)/Si5351(NULL) were ambiguous) and 6 non-blocking ones (null guard, delegating constructor, README notes on begin()/setClock() and ESP32-only pins, changelog wording, extra tests). All were fixed in ee18a6b.
• Test Lab re-check of ee18a6b: PASS.
• Test Lab re-check of a4cfc9a (README wording and the vcxo example fix): PASS.
• Re-check of abf9b2a (version bump to 2.4.0 and changelog only, no code changes): PASS (Test Lab; only README, library.json, library.properties changed; arduino-lint 0 errors; builds clean).
• 8472d61 (changelog wording only): docs-only change, no re-review needed.

Related

PR #76 (I2C interface abstraction for STM32) overlaps with this one, but this PR does not replace it: #76 abstracts away the Wire library itself.

Closes #103
Closes #106