Convert#

The Convert screen (sidebar 🔄, Ctrl+5) batch-converts recordings (.db) and BDF files into analysis formats. Controls are on the left, the matching FILES list on the right; a status line reports “{sel} selected – {vis} visible” and conversion progress.

The ⟳ Refresh button in the header rescans the input directory so files added since you opened the screen appear in the list. The list also auto-rescans whenever you return to the Convert tab.

Converting files#

  1. Input directory — where to look for recordings. “Default” points at your recordings folder; “Open” reveals it in the file manager.

  2. Output directory — where converted files are written.

  3. Search pattern — a glob, searched recursively. * matches everything; something like sub-*/**/*.db narrows to a subject’s subfolder.

  4. SOURCE FORMATS — tick .db and/or .bdf.

  5. TARGET — pick the output format:

    Format

    Use with

    FIFF/FIF (default)

    MNE-Python pipelines.

    SET

    EEGLAB.

    BDF

    General biosignal interchange.

    CSV

    Spreadsheets, custom scripts.

    BIDS

    A shareable EEG-BIDS dataset (see below).

    The BrainAccess Sample channel (the device’s sample number) is exact in FIF, BDF and CSV. FIF stores samples in double precision, so files are twice the size of single precision; MNE splits a FIF file larger than 2 GB into parts (…-raw.fif, …-raw-1.fif) and reads them back as one recording. BDF stores 24-bit integers, so Sample is written modulo 8,388,608 (2^23): it restarts at 0 after 9.3 h at 250 Hz (1.2 h at 2 kHz) — add 8,388,608 at each restart to recover the full count. SET stores single precision, which rounds Sample once it passes 16,777,216 (9.3 h at 500 Hz).

  6. Tick the files you want in the FILES list (“Select all” / “Clear” helps for large batches).

  7. Click the green “Convert selected files”. Progress is shown per file — “Converting {n} of {total} files ({pct}%)” — and a running conversion can be stopped with “Cancel conversion”.

Options#

  • “Synchronize start time across streams” — when a recording contains several streams, crop all of them to a common start time (t = 0) in the output.

BIDS datasets#

With TARGET set to BIDS, the output directory becomes the root of an EEG-BIDS dataset. Each recording is written as sub-<subject>/ses-<session>/eeg/sub-…_ses-…_task-…_run-…_eeg.bdf with its sidecars (_eeg.json, _channels.tsv, _events.tsv, electrode positions, scans.tsv, participants.tsv, dataset_description.json). Converting into an existing dataset replaces recordings that end up with the same subject/session/task/run, so convert a dataset in one go.

  • Subject, session, task and run come from Board’s recording names (subj-1_ses-S001_task-rest_run-001_…_eeg.db → sub-001/ses-001, task-rest, run-001) or from sub-…/ses-… folders. BIDS labels keep letters and digits only, so the task eyes-closed becomes task-eyesclosed (the sidecar’s TaskName keeps eyes-closed).

  • “Task for files without one” — BIDS needs a task; it is used for recordings made without one.

  • “Power-line frequency” — 50 Hz, 60 Hz or Unknown, stored as PowerLineFrequency.

  • A recording with several devices gives one file per device, told apart by acq-<device> (e.g. acq-BAMINI001).

  • Repeated run numbers. The Record screen does not advance the run number, so several recordings often share one. When recordings of the same subject/session/task share a run, that task’s recordings are renumbered run-001, run-002, … in recording-time order.

  • Same recording twice. Files with the same name parts and date/time are one recording: a .db is converted and its BDF exports (…_eeg_<device>-raw.bdf) are skipped; device exports without their .db become one acq-<device> file each. The finished status says how many runs were renumbered and files skipped.

  • Markers become events.tsv. The recording start is not an event: it goes to scans.tsv (acq_time, from the file name’s date and time, converted to UTC with this computer’s time zone).

  • Electrode positions are the standard 10-05 positions for each label, not measured ones; channels with labels outside 10-05 keep their name and get no position. Accelerometer channels are ACCEL, contact, sample-counter and battery channels MISC.

  • Choose an output directory outside your recordings folder: other files inside the dataset root make it invalid BIDS. Fill in the dataset’s authors, licence and README yourself before sharing.

The conversion does not start if a selected file has no subject, or has no task while the task field is empty; the status line names the files. Deselect them (or enter a task) and convert again — “Dismiss” clears the message.

If something is greyed out or empty#

  • “Convert selected files” disabled: you need valid input and output directories, at least one selected file, a chosen target format, and at least one source format ticked (“Select at least one source format.”).

  • “No files match the current directory, pattern, and source format filters.” — widen the pattern to *, check the input directory, or tick the other source format.