Add LP5562EVM USB-I2C control tool, document reverse-engineered protocol

Reverse-engineered the MSP430 serial protocol on the LP5562EVM:
- I<addr><reg> reads, O<addr><reg><data> writes, status 00=OK / 02=NAK
- EN pin must be jumpered to VDD (MSP430 doesn't drive it after reset)
- CONFIG (0x0B) is write-only, W_CURRENT (0x09) not writable via bridge
- LED cycling verified: all 4 channels respond to direct PWM control

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

docs/plans/2026-03-05-lp5562evm-protocol.md

@@ -0,0 +1,90 @@
+# LP5562EVM MSP430 Serial Protocol (Reverse-Engineered)
+
+**Date**: 2026-03-05
+**Status**: Working — I2C read/write verified, LEDs cycling
+
+## Hardware
+
+- **EVM**: LP5562EVM (SNVU203A) with MSP430F5510 USB-to-I2C bridge
+- **USB**: CDC device, VID:PID `2047:0300`, appears as `/dev/ttyACM0`
+- **Baud**: 115200, 8N1
+- **EN pin**: Must be jumpered to VDD externally — MSP430 does NOT drive EN after reset
+
+## Protocol
+
+Commands are ASCII text, no line terminator needed. The MSP430 parses character-by-character and responds after receiving the expected number of hex digits.
+
+### Commands
+
+| Command | Format | Response | Description |
+|---------|--------|----------|-------------|
+| `?` | `?` | `TI Sep 12 2012 08:19:25\n` | Firmware version |
+| `I` (read) | `I<addr2><reg2>` | `SS DD OK\n` | I2C register read |
+| `O` (write) | `O<addr2><reg2><data2>` | `SS OK\n` | I2C register write |
+| `R` (read) | `R<addr2><reg2>` | `OK\n` | I2C read (no data returned) |
+| `A` | `A` | `OK\n` | Unknown (possibly internal) |
+
+- `addr2`: 2-char hex I2C 7-bit address (e.g., `30` for LP5562)
+- `reg2`: 2-char hex register address
+- `data2`: 2-char hex data byte
+- `SS`: 2-char hex status (`00` = success, `02` = I2C NAK / device not responding)
+- `DD`: 2-char hex data byte read from device
+
+### Other Command Prefixes
+
+These accept input but are not fully understood:
+
+| Prefix | Chars consumed | Response | Guess |
+|--------|---------------|----------|-------|
+| `S` | 2 hex chars | (silent) | GPIO/setting control? |
+| `V` | 2 hex chars | (silent) | Voltage/value setting? |
+| `C` | 6 hex chars | `OK\n` | Config/control? |
+| `P` | 8 hex chars | (silent or `OK`) | Program memory? |
+
+### Invalid Commands
+
+Single characters not recognized as command prefixes return `?\n`. Extra characters after a complete command also return `?\n` each.
+
+## LP5562 Register Quirks via EVM Bridge
+
+| Register | Address | Writable | Read-back | Notes |
+|----------|---------|----------|-----------|-------|
+| ENABLE | 0x00 | Yes | Yes | |
+| OP_MODE | 0x01 | Yes | Yes | |
+| B/G/R/W_PWM | 0x02-0x05 | Yes | Yes | |
+| B/G/R_CURRENT | 0x06-0x08 | Yes | Yes | |
+| W_CURRENT | 0x09 | No | Always 0x00 | Write ACKs but no effect |
+| CONFIG | 0x0B | Functional | Always 0x00 | Write-only: takes effect but reads 0 |
+| STATUS | 0x0C | N/A | Yes | Read-only |
+| LED_MAP | 0x70 | Yes | Yes | Default 0x39 (engines mapped) — must set to 0x00 for direct PWM |
+
+## Working Init Sequence
+
+```
+O300DFF     # Reset (write 0xFF to RESET register)
+            # Wait 500ms
+O300040     # ENABLE = 0x40 (chip_en)
+            # Wait 1ms (>500us startup)
+O300B01     # CONFIG = 0x01 (internal oscillator) — write-only
+O307000     # LED_MAP = 0x00 (all channels = I2C direct)
+O300632     # B_CURRENT = 0x32 (5mA)
+O300732     # G_CURRENT = 0x32
+O300832     # R_CURRENT = 0x32
+            # W_CURRENT left at default 0xAF (0x09 not writable)
+O3002FF     # B_PWM = 0xFF (full brightness)
+```
+
+## LED Hardware on EVM
+
+- **D1**: LRTB_G6SF (Osram RGB LED) — driven by B/G/R channels
+- **D2-D5**: LW Q38E (Osram white LEDs) — driven by W channel
+- **D7**: LSL296 (red status LED) — driven by MSP430, not LP5562
+- All LP5562 LED outputs pass through **P5** header (4x2 dual-row with jumpers)
+
+## Key Discovery: EN Pin
+
+The LP5562 EN pin is controlled by the MSP430's GPIO (EN-UC), connected through the P6 header. After USB reset, the MSP430 holds EN low. The TI Windows GUI software presumably sends a command to assert EN, but we were unable to identify which serial command does this.
+
+**Workaround**: Jumper EN directly to VDD (3.3V) on P1 or P6 header. This is acceptable because in the final xblink design, EN is controlled by a wired-AND circuit (MCU GPIO + hall sensor), not the EVM's MSP430.
+
+Without EN high, all I2C reads return status `02` (NAK) with data `00`.