Skip to main content

LC-MS IT-TOF (.lcd)

Status: Partial​

Payload run-length encoding: confirmed. Index-to-m/z calibration: confirmed. Scan metadata prefix: still open.

The TTFL Raw Data storage in IT-TOF .lcd files holds the primary mass-spectrometry data. It uses an indexing scheme built around 4 interleaved acquisition channels per retention-time point (e.g. rapidly alternating ionization polarity or MS level - dataset naming like ..._pos-neg_NN.lcd implies this, though the reader does not yet decode which channel maps to which mode).

Retention Time and Data Index​

Retention Time is a plain u32[N_RT] array in milliseconds. Data Index is normally N_RT * 64 bytes - one 64-byte entry per retention-time point, split into four 16-byte subsets (one per interleaved channel). Each subset carries an absolute byte offset into MS Raw Data, the retention-time entry index it belongs to, and a global event counter.

Two corpus-derived corrections to keep in mind when reading this stream:

  • A subset's own entry_i field must be read directly, not assumed to equal its physical block position - files with only 2 real interleaved channels per retention-time point pack two RT points' subsets into one 64-byte block.
  • Data Index is not always an exact multiple of 64 bytes; a trailing partial block (any positive multiple of 16 bytes) is valid when the real-subset count isn't a multiple of 4.

MS Raw Data scan header (64 bytes)​

Every scan starts with a 64-byte header of 16 u32 fields. The confirmed fields include: a global scan sequence number, the RT entry index, retention time in milliseconds (an exact copy of the Retention Time stream value), the channel/subset index, and several file constants. A field once hypothesized to be a per-scan peak counter was disproved by direct measurement - it's better explained as a sub-cycle acquisition timestamp, though that reading is not fully confirmed either. The true per-scan peak count is only recoverable by fully decoding the payload.

Payload: run-length-encoded sparse profile (confirmed)​

Every scan payload splits into an undecoded, variable-length metadata prefix followed by a run-length-encoded tail: repeating [marker: u16 = 0x8000 | run_length] [skip: u16] [run_length raw u16 intensity values], terminated by the marker word 0x8000 alone. Reconstructing (position, intensity) pairs is a matter of walking the runs and accumulating skip into a running position.

This decoder was verified byte-exact (zero leftover bytes, no parse errors) across every scan in every locally available IT-TOF .lcd file from two accessions - 109,336 scans total, 100% clean. The reconstructed position axis can reach values in the hundreds of thousands for some scans, so the reader does not clamp or assume any upper bound on it.

Index-to-m/z calibration (confirmed)​

The RLE payload's reconstructed position is a raw digitizer/TOF time-bin index. The reader converts it to physical m/z using a per-file calibration parsed from TTFL Tuning/Tuning Result NN: this stream stores a reference calibrant mass ladder (identified as sodium formate cluster ions, [Na(HCOONa)n]+ - a standard, publicly documented ESI calibration solution, identified by its exact spacing matching integer multiples of the public monoisotopic mass of HCOONa) alongside its measured flight times. Fitting the standard TOF flight-time law time = a*sqrt(mz) + b by least squares gives a residual at the level of floating-point round-trip noise (about 1 part in a million) across every IT-TOF file in the local corpus, with only 4 distinct (a, b) pairs across 81 files - consistent with a handful of real per-instrument tuning sessions, not coincidence.

Since vendor software cannot be used to check this (per this project's clean-room policy), the evidence that this also calibrates the RLE payload's own index axis is independent and internal: applying it to real scan data recovers plausible small-molecule m/z for the bulk of real signal, and - most convincingly - the theoretical sodium formate cluster masses recur at a tightly clustered predicted index position (within a few index units, well under 0.1 Da) across dozens of independent scans spanning an entire run, concentrated in the channel pair independently expected to be positive polarity. See docs/format/03-lcd-ttfl-msdata.md section 3c for the full derivation.

Applying the calibration to the rare, very large noise-tail index values (into the hundreds of thousands) yields implausibly large m/z - this reflects those positions being electronic noise, not a flaw in the calibration; the reader does not filter or clamp them.

Not yet resolved​

  • Scan metadata prefix: the bytes before the RLE tail (~194 bytes in most scans, more when extra MS/MS precursor metadata is present) are not decoded.
  • Per-channel polarity/MS-level: not resolved - see Known limitations.

See docs/format/03-lcd-ttfl-msdata.md in the repository for the full byte-level record and verification methodology.