CAN frame export and resampling

outputs.can_frames exports original frames from one CAN source, or reconstructs frames at requested times using explicitly selected continuous fields. See Outputs for common scopes, publication, and optional device labels.

Complete minimal and grid examples

export-can-minimal.yaml defines the joint.asc/joint.dbc source, a voltage binding, and frame export. It exports the three original frames at 0, 1, and 2 seconds within [0,3). The voltage integral is 13 V·s.

export-can-grid.yaml adds a 500 ms frame grid. It reconstructs DBC field Voltage while retaining other bits from a base frame. Of six target positions, 0, 0.5, 1, 1.5, and 2 seconds produce five frames. The 2.5-second point has no right-hand frame inside the range and is skipped. The integral remains 13 V·s.

export-can-all-windows.yaml defines voltage above 5 V as an activity over [1,2), then creates an activity window and another extending 1 second earlier. can_frames uses scope: {windows: all} to export the two windows separately. Run chronaxis run FILE.yaml --output OUTPUT_DIRECTORY. Follow exports.index_file in this run’s manifest to its timestamped export index, then inspect window/segment identities.

Export original CAN frames

With one DBC-configured CAN source, source can be omitted. Message IDs and fields must match the actual DBC, and the export label requirement applies.

Start from dual-grid.yaml, find root outputs, and add or replace only its can_frames child. Keep other output settings and root fields; do not paste a second outputs key.

outputs:         # Additional result files to publish
  can_frames:    # Export the selected original CAN frames
    format: asc  # Text CAN log; use blf for binary BLF output.

Add a reconstructed frame grid

Replace the preceding can_frames child with this version. The two fragments are alternatives, not blocks to append together.

outputs:                         # Additional result files to publish
  can_frames:                    # Export the selected original CAN frames
    format: asc                  # Output message file format
    resampling:                  # Optional output grid and value selection
      period: 500ms              # Independent of the signal CSV grid.
      max_gap: 2s                # Maximum actual gap used for interpolation.
      messages:                  # CAN message definitions selected for output
        - frame_id: 0x100        # Message ID from the DBC.
          continuous: [Voltage]  # DBC signal names, not analysis aliases.

These fields are relative to outputs.can_frames:

Without resampling, export retains original frames in range, including IDs absent from the DBC and distinct frames at the same time. The public execution path writes frames.asc or frames.blf and an export manifest.json; it does not add a frames.jsonl original-frame copy.

With resampling, only messages listed in messages and fields listed in their continuous arrays are recomputed. Here Voltage is a DBC field name, not the analysis alias voltage.

These fields are relative to outputs.can_frames.resampling:

A mapped period accepts only period: {mode: dbc}. It reads each listed message's own period and does not accept the signal CSV driver option. A missing or invalid DBC period fails initialization; it does not silently fall back to a fixed value.

The seven fields in each messages item

These fields belong in outputs.can_frames.resampling.messages[]. Message identity follows CAN signal binding, but there is no signal field here; put DBC fields to reconstruct in continuous.

protected is an explicit guard against selecting a counter or checksum as continuous. Even when it is empty, all bits outside continuous retain their base-frame values; an empty list does not make every field interpolatable. The example reconstructs only Voltage. Counter and Checksum are illustrative names and must be replaced with actual DBC fields if used.

For each reconstructed frame, the program:

1. Chooses the original frame nearest the target time as the base, choosing the earlier one on a tie.
2. Estimates the selected continuous fields from observations on both sides and replaces only their bits.
3. Preserves every other bit from that same base frame, including states, counters, checksums, and bits not defined by the DBC.

For a target at 100 ms with original frames at 80 and 120 ms, the 80 ms frame is the base. Voltage can be estimated, while its counter and state remain those from 80 ms.

Boolean, enum, and multiplexor-selector fields cannot be continuous. Interpolation is also unavailable when endpoints use different multiplexing branches, reliable records are missing, or the gap exceeds max_gap.

Integer DBC fields use nearest_even-v1: an exact half-step rounds to the even integer, so 2.5 becomes 2 and 3.5 becomes 4. Out-of-range raw codes are rejected, not clipped. Checksums are retained, not recalculated, so reconstructed frames are not guaranteed safe to replay to a physical device. The manifest records observed/estimated counts and reasons for skipped frames.

If one automatic rule matches two actual frame types, it asks you to specify the relevant type option. It does not turn a standard frame into an extended frame, or classic CAN into FD because the DBC says FD. Message names come from the loaded DBC; do not supply message.

Export relative-time data as BLF

BLF requires calendar time. If the experiment is known to have started at 08:00 Beijing time on 2026-10-23 and original timestamps start at zero, use the following fragment. In dual-grid.yaml, replace only the existing outputs.can_frames child and retain other settings.

outputs:                        # Additional result files to publish
  can_frames:                   # Export the selected original CAN frames
    format: blf                 # Output message file format
    capture_date: '2026-10-23'  # Actual capture date from experiment records.
    output_offset: 28800s       # Eight hours after that date's midnight.
    timezone: +08:00            # Capture timezone.

The date sets midnight; adding 28800 seconds gives 08:00. Use your actual capture date. If original times already count from 1970-01-01 UTC, choose epoch and omit capture_date and timezone.

If original time-of-day is already 14:05:40, do not add another 14 hours. Add the known start-time offset only when timestamps start at zero at experiment start. Analysis time.offset aligns sources and does not automatically shift exported original frames. Type and combination checks are documented in FrameExportSpec (Chinese implementation reference).

Complete example using DBC periods

Download analysis.yaml, grid.asc, and grid.dbc into one directory and run chronaxis run analysis.yaml --output grid-result.

The DBC gives Power a 1000 ms period and Voltage 500 ms. Inputs update at 0, 1, and 2 seconds. The signal CSV has one grid: period.mode=dbc with driver=power selects 1000 ms. Removing the driver is an error because two physical messages are candidates.

CAN export uses period: {mode: dbc} without a driver, giving Power a 1-second grid and Voltage a 0.5-second grid. These choices are independent. The observation range is [0,2); inspect data and manifest generation/skip counts together.

The signal CSV has four rows: power=10/20 W and voltage=20/30 V at 0 and 1 seconds. The reconstructed ASC has five frames: Power at 0 and 1 seconds; Voltage at 0, 0.5, and 1 seconds, with 20, 25, and 30 V. Voltage at 1.5 seconds needs the excluded 2-second endpoint, so is skipped instead of borrowing an out-of-range sample. The two row counts reflect their independent grids.