Skip to content

Processing, analysis, and render profiles

The processing pipeline that runs on observation data is configurable through three kinds of profile attached to your observation:

  • ObservationAssetProcessingConfigProfile — how raw data is reduced and calibrated.
  • ObservationAssetAnalysisConfigProfile — how the reduced data is analyzed (photometry, WCS solving, image-quality metrics).
  • ObservationAssetRenderConfigProfile — how the data is rendered for display (stretch, midtone, saturation handling).

You can leave these on the platform defaults — which is what most observations do — or override them per observation when the science needs custom processing.

Processing profile

Controls the reduction pipeline that turns raw frames into science-ready products.

For optical imaging, this is the bias / dark / flat workflow:

  • Which calibration set to use (the telescope's daily set, your own if you requested calibrations, or a specific named set).
  • Whether to apply pixel rejection (sigma clipping, cosmic-ray rejection).
  • Whether to coadd, and if so, with what alignment / weighting.

For radio observations, this is the on/off subtraction, baseline removal, and gain calibration step.

Each run produces an ObservationAssetProcessingRun row with its inputs, outputs, parameters, and timing — so the pipeline is reproducible and auditable.

Analysis profile

Controls what gets measured on the reduced data.

For optical imaging, this includes:

  • WCS solution — plate-solve the image against a reference catalog. Output is captured in ObservationTaskAssetProcessingRunWcsSolution.
  • Image-quality metrics — FWHM, sky background, source counts, saturation flags. Stored in ObservationAssetAnalysisRunImageQualityMetrics.
  • Photometry — aperture photometry on a target source list. Stored in ObservationTaskAssetProcessingRunPhotometry.

Each run produces an ObservationAssetAnalysisRun with the inputs, results, and any header metadata (ObservationAssetAnalysisRunHeaderCore).

Render profile

Controls how the data is converted into display imagery — the thumbnails on the results / files panels, the inline previews on public-facing galleries.

For optical imaging:

  • Stretch function — logarithmic, midtone-transfer, asinh, linear.
  • Black point / white point — the brightness range to map.
  • Saturation handling — clip vs. preserve color.
  • Background handling — model-subtract or pass through.

Each render produces an ObservationAssetRenderRun row with the chosen parameters and the rendered output asset.

When to override

Stick with the defaults unless:

  • You need specific calibration handling. Custom flat sets, dark scaling rules, or interpolation strategies the pipeline default doesn't cover.
  • You're doing high-precision photometry. Override the analysis profile to lock specific aperture sizes, background annuli, or source detection thresholds.
  • You're building publication-ready imagery. Override the render profile to control stretch and color handling for visual presentation.

Profile overrides attach to the observation; they don't propagate to other observations on the project. To share a profile across many observations, configure it at the project level (forthcoming) or duplicate it per observation in the API.

Reference