Agilent (Varian) Cary UV-Vis

The Cary UV-Vis applications Scan, batch Scan and Scanning Kinetics save their data as .dsw, .bsw and .bsk files (.csw uses the same container). OpenReadout returns every spectrum and baseline as stored, absorbance in AU against wavelength in nm, with the scan parameters and run metadata listed below.

Derived from public files of two depositors (a Cary 4000 with Scan 3.00 and a Cary 60 with Scan 5.0) and the software’s own CSV exports. Provenance: docs/provenance/agilent-cary.md. Crate: openreadout-spectro.

format id	files	reads	confidence
agilent-cary	.dsw (Scan), .bsw (batch Scan), .bsk (Scanning Kinetics), .csw (same container)	every spectrum and baseline as stored (float32 x, y), absorbance in AU and wavelength in nm, spectrum names, collection times, operator, instrument model, application and version, scan parameters (rate, interval, averaging time, bandwidth, start/stop, source changeover, beam mode, baseline settings), kinetics schedule	medium (evidence rubric, docs/assurance.md)

Not read: the graph, report, database and baseline-information stores (listed by info --view structure with their offsets and sizes); .csw is accepted by extension on the evidence of the shared magic only (no corpus file).

Container

offset	content
0x00	0x11 then Varian UV-VIS-NIR (a length-prefixed text): detection
0x3E	the first store

A store is a u32-length-prefixed class name (TContinuumStore, TBaselineStore, TGraphStore, TReportStore, TDatabaseStore, TBaselineInfoStore), then a u32 store size counted from the store’s first byte; the next store starts at start + size. In every corpus file the last store ends 4 bytes before the end of the file; any other end is a finding (store_chain_end). A store that runs past the end of the file is corrupt (exit 4). Files over 256 MiB are refused (exit 6; the corpus files are under 200 kB and are read whole).

Spectrum store (TContinuumStore) and baseline store (TBaselineStore)

After the size field, a 1,028-byte header:

offset in header	type	content
0	u32	(not decoded; word_0 in the vendor tree)
4	u32	store version: 149 in every file; any other is refused (exit 6)
8, 12	f32	x minimum, x maximum
16, 20	f32	y minimum, y maximum
24	u32	point count
28	u32	(not decoded; word_7)

Then point count × (f32 x, f32 y) pairs; points that do not fit in the store are corrupt. The stored extremes are compared with the values’ (a finding, stored_extremes, when they differ).

Then u32-length-prefixed Latin-1 texts, up to the first bytes that are not one:

text	our reading
first text	the spectrum’s name (the CSV export’s title row)
Collection Time: 12/18/2024 10:12:52 PM	collection time, local clock, month first (every file with a day above 12 puts the month first, the Polish depositor’s too)
Operator Name : …	operator
Scan Software Version: 3.00(182), Scan Version 5.0.0.999, Scanning Kinetics Version 5.0.0.999	application and version
Parameter List : then one text per parameter	parameters: split on runs of two or more spaces into a name and values (Scan 3.00 writes fixed 302-character fields, 5.0 ␣␣name␣␣value[␣␣value])
Method Log : … End Method Modifications	method log; Method Name :, Date/Time stamp:
<SBW (nm)> , 2.000, <Current Wavelength> , 300.00, [Time] , 2.000	status texts; [Time] is the kinetics method’s scheduled time of the spectrum (minutes)

The bytes after the last text are counted (trailing_bytes), not decoded.

Axes

X Mode Nanometers → wavelength (nm). A Scanning Kinetics file names no X mode: nm is assumed (its method log’s X Start nm/X Stop nm bound the values; a note says so and the assurance profile records it). Any other X mode returns the values without a unit, with a finding (x_mode). Y Mode (Ordinate mode in kinetics files) Abs → absorbance (AU), validated against the CSV exports; %T/%R → transmittance/reflectance (%) with a finding (y_mode: no export checks them); anything else under its own lower-cased name without a unit.

What the reader returns

• Traces: one per set of spectra that share their x values (bit for bit), y mode and kind (sample spectra first, then baselines): name absorbance (baseline absorbance for baselines), one sweep per spectrum, values as stored (float32). extra.axis is the regular wavelength axis when the x values are evenly spaced, else the listed x values; extra.spectrum_names, extra.y_mode, extra.x_mode, extra.data_type UV/VIS SPECTRUM.
• Tables: for a set of several spectra (batch and kinetics files), one row per spectrum: collected_s (seconds after the set’s first Collection Time) and, in kinetics files, scheduled_min (the [Time] text). The two differ: the Scanning Kinetics example’s spectra are scheduled at 0, 2, 4, 6, 36, … min and collected 0, 0.9, 2.9, 4.9, 6.9, 36.9, … min after the first.
• Experiment (from the first sample spectrum): vendor Agilent (Varian), model (parameter Instrument: Cary 4000, Cary 60), software and version, operator, sample name (the spectrum’s name), started_at (its Collection Time, local clock, no zone), method name, and the method parameters scan_rate (nm/min), data_interval (nm), averaging_time (s), spectral_bandwidth (nm), scan_start, scan_stop, source_changeover (nm), beam_mode, baseline_correction, baseline_type, slit_height, signal_to_noise_mode, cycle_mode — all inferred.
• Vendor tree (info --view full): cary.stores (class, offset, size), chain_end, file_tail, and per spectrum its header words, texts, parameters, method log and status texts.

Validation

corpus/oracle/series/cary-*.json (oracle/series_oracle.py on each file’s CSV export): every point’s x (to 5·10⁻⁴ nm) and y (the CSV writes the float32 values to ten significant digits). The Cary 60 files have no export: checked for self-consistency (stored extremes; the parameter Start/Stop against the x range; the store chain ending 4 bytes before the end of the file).

Vocabulary

name	meaning
CaryReader	the FormatReader for agilent-cary
CARY_FORMAT_ID	"agilent-cary"
AGILENT_CARY	the assurance profile (crates/openreadout-spectro/src/assurance.rs)
spectrum_names, y_mode, x_mode, collected_s, scheduled_min	trace and table fields above
scan_rate, data_interval, averaging_time, spectral_bandwidth, scan_start, scan_stop, source_changeover, beam_mode, baseline_correction, baseline_type, slit_height, signal_to_noise_mode, cycle_mode	method parameters above
store_chain_end, stored_extremes, x_mode, y_mode	finding codes
stores, chain_end, file_tail, word_0, word_7, trailing_bytes, method_saved, method_log, status	vendor-tree keys

---

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