# MAG160C C SDK (csdk/)

Pure C reimplementation of the MAG160C thermal camera SDK, written from the
recovered protocol (see `analysis/protocol_spec.md` for the full reverse
engineering report).  The previous C++ implementation was archived to
`analysis/legacy-cpp/`; this tree replaces it.

## Layout

```
csdk/
  include/mag160c/mag160c.h   public C ABI (error domain, magic codes, frame
                              format, temperature API, TCM framing)
  src/
    mag160c_internal.h        shared helpers (last-error)
    mag160c_error.c           TLS last-error + error names
    mag160c_frame.c           0x1bb1b11b frame parse + streaming assembler
    mag160c_temp.c            Calibration / ConvertResponse2Temperature / T2E
    mag160c_tcm.c             FTDICommand 0x7e framing + rotate/light builders
    mag160c_ir.c              libusb session (link/start/stop/ffc/info)
    mag160c_tables.h          T2E + E2TAccQ10 tables (generated from binary)
  tools/mag160c_cli.c         CLI: ir-info, ffc, start, stop, tcm-rotate,
                              tcm-light, frame-test, temp-test
  tests/                      test_frame, test_temp, test_tcm, test_api
```

## Build

```
cmake -B build -S csdk          # auto-detects libusb-1.0 via pkg-config
cmake --build build
ctest --test-dir build
```

Without libusb-1.0 the IR session APIs return MAG160C_ERR_NOT_SUPPORTED;
the pure helpers (frame parse, temperature math, TCM framing) and their tests
work with no dependencies.  Threads (pthread) are used when available.

## Recovered protocol summary

- VID 0x833C, config value 2, interface 0.
- Commands: EP OUT 0x03, 8-byte packets `{u32 magic, u32 param}` (0x3c-byte
  with 0x38-byte payload for DDT).  Response: EP IN 0x82 (0x1000 max, 2000 ms).
- Magic commands: 0x6bb6b66b/66c prepare, 66d/66e DDT, 66f info, 670 version,
  672 FFC, 673 start, 674 stop, 676/677 set params.
- Magic responses: 0x5bb5b55b..55f (0x38-byte info blocks / 0x10 pair),
  0x5bb5b57b file header.
- Frame stream: EP IN 0x81, marker 0x1bb1b11b @0, counter @4, data length @8,
  type @0xc (0 response / 1 raw), shutter @0x10, pixels @0x1c (uint16 LE),
  trailing 0x1bb1b11c @0x1c+len, frame size 0x38+len.
- Temperature: per-pixel piecewise-linear Calibration with baseline,
  ConvertResponse2Temperature auto-gain, T2E curve (0x112 entries) with Q13
  band interpolation.
- TCM: FTDICommand 0x7e framing with additive (mod-256) checksums; command
  table main 0x02 (rotate 0x77, light 0x31/32/33, ...).