SX1262 sub-GHz radio support for HMS/HMT inverters (LilyGO T-Connect-Pro)

[chakphanu]

SX1262 as CMT2300A alternative for HMS/HMT inverters

I've implemented SX1262 sub-GHz radio support for OpenDTU, running on the LilyGO T-Connect-Pro board
(ESP32-S3 + SX1262 + W5500 Ethernet + ST7796 display).

This enables HMS/HMT communication without a CMT2300A module — the SX1262 handles the same GFSK protocol
directly.

What works

• HMS-1600-4T (serial prefix 0x1421) tested, HMT should also work
• Full DC/AC/temperature data via RealTime, Alarm, and DevInfo sub-commands
• Frequency hopping with phase prediction — 100% RX success rate (41/41 bursts)
• MODE_TH (Thailand 920–925 MHz), EU/US/BR modes defined
• Web UI: SX1262 config card, radio stats, D6 counters
• ST7796 480×222 split-panel dashboard for T-Connect-Pro

Key technical details

• RX bandwidth 156.2 kHz (0x1A) eliminates FEC decode failures (was 30% failure at 117.3 kHz)
• TCXO warmup reduced from 2000 µs to 300 µs — root cause of fragment loss during hopping
• DIO1-based fast channel cycling: 4 ms timeout per channel, 13.5 ms full 3-channel cycle
• Phase prediction: next = (current + fragCount) % 3 — proven via SDR ground truth
• Extracted HoymilesRadio_SubGhz base class so CMT and SX1262 share the same interface

Architecture

HoymilesRadio (NRF24)
HoymilesRadio_SubGhz (abstract base)
├── HoymilesRadio_CMT (CMT2300A — existing)
└── HoymilesRadio_SX1262 (SX1262 — new)

HMS/HMT inverter classes use SubGhz* pointer, so they work with either radio backend without modification.

Fork & release

• Branch: https://github.com/chakphanu/OpenDTU/tree/feature/sx1262-subghz-radio
• Release (firmware binaries): https://github.com/chakphanu/OpenDTU/releases/tag/v25.9.11-sx1262
• Based on upstream OpenDTU v25.9.11

Hardware

Component	Detail
Board	LilyGO T-Connect-Pro (ESP32-S3)
Radio	SX1262 sub-GHz
Ethernet	W5500
Display	ST7796 480×222
Inverter tested	HMS-1600-4T (Thailand 920 MHz band)

Not submitting as PR since I don't have CMT2300A hardware to verify the refactoring doesn't break existing
setups. If anyone with CMT hardware wants to test the SubGhz base class extraction, the branch is there.

Related: #1546 (asked about alternative RF modules for HMT)