[Feature request] NiceRF LoRa2021 (LR2021 Gen 4) variant support

[c03rad0r]

Use Case

I have working MeshCore firmware running on the NiceRF LoRa2021 module (Semtech LR2021 Gen 4 chip) paired with an ESP32-C3_Mini_V1 development board. I would like to contribute this as a new board variant.

Hardware

• NiceRF LoRa2021: 19.72x15mm module, Semtech LR2021 Gen 4 chip
• Dual-band: Sub-GHz (150-960 MHz) + 2.4 GHz LoRa/FLRC/GFSK
• ESP32-C3_Mini_V1: Maker go, 22.52x18mm, USB-C, 4MB flash, ~$4
• Total cost: ~$8 (MCU + radio module)

RadioLib Support

The LR2021 uses RadioLib's LR2021 class (extends LRxxxx → PhysicalLayer). This is already in RadioLib v7.6.0 which MeshCore depends on.

Important: This is a different RadioLib class from LR1121 (extends LR11x0). The LR2021 and LR1121 are separate chips. PRs #2696 and #2198 add LR1121 support — this request is for LR2021, which is additive.

Test Results

I have this running on real hardware:

• ✅ Radio init succeeds
• ✅ Noise floor: -102 dBm stable
• ✅ Companion connects via meshcore-cli v1.5.7
• ✅ Ed25519 identity generated
• ✅ Advert TX verified
• ✅ All build targets compile (companion, repeater, chat, KISS, room server)
• ✅ No regressions on reference builds

Implementation

Following the established pattern (CustomLR1110.h + CustomLR1110Wrapper.h + variants/):

class CustomLR2021 : public LR2021 {
    // irqDioNum = 9 (LR2021 uses DIO9 for IRQ, not DIO1)
    // getPacketLength() override (header error handling, like LR1110)
    // std_init() with tcxoVoltage=0 (XTAL, not TCXO)
};

Key differences from SX1262:

• IRQ on DIO9 → irqDioNum = 9 + setDioFunction(9)
• setRxBoostedGainMode(uint8_t 0-7) vs bool
• No DIO2-as-RF-switch (NiceRF module handles RF switching internally)
• Crystal oscillator, not TCXO

One Notable Deviation

Arduino's SPIClass returns all-zero responses on the ESP32-C3_Mini_V1 with non-default SPI pins. Direct ESP-IDF spi_bus_initialize() works fine. My variant includes a custom EspIdfHal.h that uses ESP-IDF SPI directly. I am happy to investigate this further or remove it if there is a better approach.

Ask

Would you accept a PR adding this variant? I have clean code ready targeting the dev branch, with all builds passing and hardware testing done.

Related: #2678 (2.4 GHz — LR2021 also supports this), #2441 (multi-radio — LR2021 is dual-band)

[c03rad0r]

Update: I have successfully tested the LR2021 variant with the Berlin MeshCore community mesh. I discovered community relays and contacts, and interacted with users multiple hops away via the public group channel. This confirms multi-hop flood routing and interoperability with the existing MeshCore infrastructure on standard EU PHY settings (869.618 MHz, SF8, BW62.5, CR4/5).