Scan data

Each SpectrumRecord yielded by iter_spectra has its mz/intensity arrays filled from one of two on-disk payload files, chosen by the SpectrumParamsType block(s) recorded for that scan in MSScan.bin:

• MSProfile.bin (profile mode) - LZF-compressed int32 ADC/intensity counts. The reader decompresses the block and converts it to f64 m/z / f32 intensity pairs using the scan's declared mass range (MinX/MaxX) as the axis. See MSProfile.bin.
• MSPeak.bin (centroid or MRM) - already-calibrated peak lists. Q-TOF centroid data is stored as parallel arrays (m/z array, then intensity array); QQQ MRM data is one (mz, intensity) pair per scan. See MSPeak.bin.

If a scan has both a profile and a centroid block (seen on some Q-TOF acquisitions), the reader prefers profile data and falls back to centroid only if the profile decode is empty.

Precursor information

MS2+ spectra populate SpectrumRecord.precursor:

• Q-TOF Auto-MS/MS: target_mz is read directly from the scan record (MSScan.bin offset 84) - the isolation window center Agilent's acquisition firmware selected for that MS2 scan.
• QQQ MRM: there is no isolation-window target in the same sense: each MS2 scan corresponds to one MRM transition, identified by a channel ID. The reader surfaces this as precursor_native_id (mrm_channel=<id>) rather than a target_mz.

Retention time

retention_time_sec is MSScan.bin's per-record RetentionTime (stored on disk in minutes) converted to seconds. For QQQ MRM data, all channels within one acquisition cycle share the same retention time - the instrument fires through its transition list once per chromatographic point, not once per unique retention time.