Waters Empower ASCII exports (.arw)

Waters Empower keeps raw data and results inside its own database, so data leaves Empower only through exports. The ASCII raw-data export, saved with the .arw extension, holds one chromatogram as text. OpenReadout returns it as one trace with the sample, method and injection fields the export includes. The AIA/ANDI netCDF export (.cdf) is read by the andi-chrom reader. Sony .ARW camera files share the extension but are TIFF and never start with a quote.

Format id empower-arw. Code: crates/openreadout-chrom/src/empower_arw.rs. How each fact was established: docs/provenance/empower-arw.md.

Layout

Text (ANSI/UTF-8), rows separated by CR, LF or CRLF, cells by tabs:

1. the names of the fields the export method includes, each in double quotes ("SampleName", "Channel", "Sample Set Name", "Instrument Method Name", …);
2. their values, in double quotes, as many as names;
3. one row per point: retention time (minutes) and the detector value, unquoted decimal numbers.

The value’s unit is not in the export (it is the detector’s: mV, AU, EU for fluorescence, …). Times are printed to 7 significant digits, so an evenly sampled run shows steps that differ in the last digit; points within 10⁻⁶ of the grid (relative to the time) are a regular trace.

Detection: a first line of tab-separated, double-quoted cells (with or without the .arw extension). Rows of more than two columns (a multi-wavelength PDA export) are refused (exit 6); a header whose value count differs from its name count, or a data row that is not two numbers, is corrupt (exit 4).

Mapping to the data model

• One trace named <SampleName> / <Channel> (or the channel alone), one channel named after Channel (else value), dtype float64, no unit. Evenly spaced times: sample_rate_hz, start_s and extra.axis (retention time, minutes); otherwise sample_rate_hz 0 and a first channel time (minutes).
• extra: fields (every exported name and value), and from them sample_name, channel, sample_set, instrument_method, processing_method, vial, injection, injection_volume, acquired_by, acquired_at when exported; line_ending, x_start_min, x_end_min, time_channel.
• check: time_not_increasing (warning), irregular_times, no_channel_field (info).

Validation

• Six exports of Appia’s test data (MIT, two fluorescence channels of three samples, 6,601 points each): every value equal to Appia’s own reading (processed-tests/…hplc-wide.csv, an independent reader) and the times within Appia’s single-precision rounding (2·10⁻⁶ min).

Known gaps

• Multi-column (3D PDA) exports: no licensed public example; refused.
• Empower’s native data (database) and its report exports are not read; AIA/netCDF exports are read by andi-chrom.

Vocabulary (every public identifier in empower_*.rs must appear here)

identifier	meaning
EmpowerArwReader, EmpowerArwDataset, EMPOWER_ARW_ID, open, export	reader, opened export, the id empower-arw, open by path, the parsed export
MAX_ARW_BYTES	largest export read (512 MiB)
ArwExport, fields, times, values, line_ending, field, regular_step	a parsed export: header names and values, points, line ending; a named field’s value; the step of an evenly spaced export
ArwError { Corrupt, Unsupported }, parse_arw, looks_like_arw	why an export was not read; the parser; detection

---

How this reader was derived, file by file: provenance log.