A complete guide to what every SyQon Studio function does, when to use it and how to approach it safely.
Canvas, scientific scopes, processing inspector, neural models and layer stack in one native workspace.
Move from frames to method and output in a focused three-step workspace, with Smart Telescope and Advanced paths, calibration groups and frame review.
Current Stacking Studio method and integration settings
Unfold a solved field into catalogue distances, explore stars and galaxies in 3D, select an object to orbit around it and compare estimated separations.
Explore catalogue names and celestial coordinates over a solved image. WCS-aware overlays follow the astronomical field rather than becoming permanent image pixels.
Inspect sources, refine alignment and preview a composition before rendering. This demonstration uses overlapping crops of the supplied M106 master, not separate observations.
Current SyQon Studio workspace with the supplied M106 XISF master
© Georg Nikolaus Nyman
This chapter explains the complete Fusion Stack workflow: how to organise acquisitions, assign calibration, judge subframes, choose integration settings and verify the files Studio produces. Start with the recommended path, then open the advanced sections only when the data gives you a reason.
Keep the original frames, inspect the manifest and diagnostics, and compare important integrations against a trusted workflow. Automatic recommendations are evidence, not a replacement for judgement.
Use RGB lights. Calibration happens in the sensor domain before demosaic. Keep one Bayer pattern, camera configuration and geometry per acquisition. Choose Edge-aware development or CFA drizzle when its requirements are truly met.
Assign L, R, G, B, Ha, OIII and SII independently. Flats and dark-flats follow the selected filter. Fusion Stack produces separate linear masters; combine them later with Channel Composition or Spectral Master Fusion.
Create one acquisition for every calibration/equipment family. Choose a common reference grid. Each group is calibrated and integrated independently, then same-filter masters are registered and fused.
Create one acquisition per panel and calibration set. Choose a reference, verify shared stars or valid WCS, then inspect union coverage, overlap geometry and seams. Filters stay separate on the same output grid.
Import dataset folder scans subfolders and available FITS frame-type headers, then proposes lights, darks, flats, bias and dark-flats. Review every assignment before stacking; folder names and incomplete metadata are not proof of compatibility.
Save session stores source paths, acquisition groups, calibration assignments, settings and frame exclusions—not the image data itself. Keep sources at stable paths, then use Load session to resume the setup.
Enable Align on comet / moving target and provide a CSV containing path, x, y for the target centre in each light. Studio converts the stellar transforms into target-locked transforms before integration.
Separate equipment, nights and calibration families.
Assign lights and matching calibration to the current filter.
Set backend, sampling, rejection, resources and output.
Calibration isolation is the most important organisational rule in Fusion Stack. Darks and bias belong only to their acquisition. Flats and dark-flats belong to both the acquisition and the selected filter.
| Frame role | Scope | How Studio treats it |
|---|---|---|
| Lights | Selected filter + acquisition | Science exposures to analyse, register and integrate. Keep dimensions, CFA layout and acquisition scaling compatible. |
| Darks | Acquisition | Match exposure, gain/offset and temperature where metadata exists. Darks retain their bias pedestal; Studio handles bias-separated exposure scaling when valid inputs allow it. |
| Flats | Selected filter + acquisition | Correct dust and illumination. Raw flats are corrected with matching dark-flats first, otherwise bias, then normalised per channel or CFA phase. |
| Bias | Acquisition | Optional zero-duration pedestal reference. It is never borrowed from another acquisition and must not be subtracted twice. |
| Dark-flats | Selected filter + acquisition | Preferred correction for matching raw flats. They do not replace light-frame darks. |
| Master calibration file | Its original role | One supplied file is accepted as a master. Enable “Flats are calibrated masters” only when the flat has already received bias/dark-flat correction. |
| Synthetic bias | Acquisition | Use only a measured constant pedestal in normalised sensor units: 16-bit FITS ADU ÷ 65535. The camera OFFSET setting is not this measurement. |
When both headers provide temperature, a dark/light difference above 2 °C is rejected. Missing metadata cannot prove compatibility.
Studio can scale bias-separated dark current only when the required exposure and bias information exists. It will not silently scale an unknown pedestal.
Advanced acquisition controls can apply a measured pedestal, overscan region, cosmetic correction and an optional defect map consistently.
Analyse subframes measures the current acquisition/filter without creating a master. Review & select frames is where you inspect distributions, preview individual files and decide which exposures are allowed into integration.
Relative stellar width in registration-proxy pixels. Lower generally indicates tighter stars, but undersampling and saturation can distort it.
Stellar elongation indicator. Compare its distribution across the session rather than treating one number as universal.
Diagonal-difference noise estimate used by quality and weighting. It is not a substitute for visual inspection.
Per-channel median sky level. Sudden changes can reveal clouds, moonlight or transparency shifts.
Combined within-session ranking. It is a selection aid, not a statement of absolute scientific quality.
Number of compact stellar detections supporting quality and registration diagnostics.
Contribution derived from the chosen weighting rule after accepted-frame analysis.
Explains recommendation, decoding problems, incompatibility or registration failure.
Large nebulosity, saturation and severe undersampling can influence the metrics. They are heuristics, not automatic cloud classification.
Scientific preprocessing defaults establish a robust baseline: calibrated masters, edge-aware demosaic, cosmetic correction, dark optimisation, local background normalisation, resumable checkpoints, point-source SNR weighting and automatic rejection.
Use the normal affine path for a consistent raw sequence. Blind projective registration is for large scale, rotation or perspective differences and should not be enabled merely because one poor frame fails. Non-converging frames are rejected instead of silently relabelled.
Local background normalisation corrects spatial background differences after registration. Within a raw acquisition the default is additive background matching, not arbitrary throughput or exposure scaling. Inspect faint extended signal carefully.
Choose inverse variance, measured stellar signal or point-source SNR. Weights change contribution, not sample brightness. Optional PSF matching can equalise Gaussian variance to the broadest accepted frame before projection.
Automatic is the right first choice. Manual methods are for datasets whose failure mode you have identified in previews, maps or statistics.
| Method | Consider it for | Behaviour |
|---|---|---|
| Automatic by frame count | Recommended default | Chooses a bounded method appropriate to the number of accepted frames. |
| Weighted mean | Very small or exceptionally clean sets | No outlier rejection; every valid sample contributes according to weight. |
| Percentile clipping | Small stacks | Removes bounded low/high ranks where robust variance estimates would be weak. |
| Studentized sigma rejection | Satellites, hot residuals, isolated transients | Tests each sample against statistics that exclude its own influence. |
| Studentized sigma clamp | Transient control without complete removal | Clamps detected extremes rather than discarding their contribution entirely. |
| Linear-fit clipping | Frames with differing linear response | Uses a fitted relationship before testing sample residuals. |
| Generalized ESD | Large stacks | Detects one or more extremes with thresholds suited to a larger temporal population. |
| Min / max | Predictable fixed-rank removal | Removes a configured number of lowest and highest samples. |
| Large-scale rejection | Broad moving contamination | Targets spatially coherent rejected regions; use only when the data justifies it. |
Drizzle reconstructs sampling from real sub-pixel offsets. It does not invent resolution, denoise the stack or compensate for repeated identical dithers. Always integrate at native resolution first.
Best first integration and default when dithering is absent or uncertain.
Useful for genuinely undersampled, dithered data with manageable output size.
Heavy storage and compute cost. Requires dense, varied sub-pixel phases.
Specialist sampling mode, not six times more real detail. Coverage inspection is essential.
A positive pixfrac spreads each square drop; smaller values reduce extra smoothing but demand denser coverage and can make noise less uniform. 0 · Point assigns each sensor sample once without spatial spreading. It requires dense dither phases and does not undo correlation already introduced by demosaic.
For Bayer lights at 2×, 4× or 6×, CFA drizzle integrates original red, green and blue sensor samples directly instead of interpolating before integration. Confirm the Bayer pattern and inspect colour-channel coverage; it is not appropriate for already demosaiced input.
The memory budget is a ceiling, not a reservation. Fusion Stack caches calibrated float32 data and processes bounded output tiles, but one decoded frame and calibration buffers must still fit. Drizzle, maps, intermediate acquisition masters and checkpoints can require far more disk than the source dataset.
Automatic prefers the supported GPU path and reports the backend actually used. Decoding, robust statistics and some mosaic geometry remain CPU work. Checkpoints can reuse verified caches and completed tiles when inputs and compatible settings remain unchanged; retain enough disk space until verification is complete.
A successful run publishes a named folder atomically. The image, maps and manifest together explain what was accepted, how it was transformed and where the output is trustworthy.
Float64 linear master for each populated filter or final fused band.
Effective support or acquisition weight. Use it to find holes, weak borders and uneven drizzle sampling.
Where temporal samples were rejected. Look for expected trails/hot pixels and unexpected target-shaped rejection.
Inputs, assignments, accepted/rejected frames, transforms, gains, residuals, settings, warnings and provenance.
Use common sampling, sufficient overlap and a defensible reference. The acquisition mosaic and post-production mosaic workflows are related, but they start from different inputs.
Do not bypass it blindly. Check camera, dimensions, binning, Bayer pattern, exposure, gain, offset and temperature. Move incompatible files into a separate acquisition.
That is expected for linear data. Enable Display Stretch for inspection; do not alter the master merely to make a viewer display it brightly.
Choose a representative reference, verify overlap and comparable scale, remove frames with trails/clouds, and use blind projective registration only for genuinely large geometric differences.
The dither phases are insufficient or pixfrac is too small. Inspect Coverage, increase positive pixfrac, reduce the drizzle scale or return to native resolution.
Verify calibration per panel, overlap support, WCS/stellar geometry, local background normalisation and feathering. Controls cannot recover missing exposure.
Lower the working-memory ceiling only if one decoded frame still fits, use a fast local SSD, disable unnecessary drizzle/maps and remove retained checkpoints only after the result is verified.
Manual Use overrides quality/FWHM/eccentricity limits, not unreadable data, incompatible geometry or failed registration.
No partial final master is published. Load the saved session and reuse compatible checkpoints when inputs and settings are unchanged.
Search by outcome—try “remove green”, “walking noise”, “mosaic”, “stars” or “export JPEG”—or filter by workflow area.
Unfold a solved field into catalogue distances, explore stars and galaxies in 3D, select an object to orbit around it and compare estimated separations.
Exploring a solved field in three dimensions.
Export the scene as PNG, PDF or MP4 with the Earth as the observation origin. Choose Landscape, Vertical or Square video framing; every export carries the Studio signature.
Sharing an isolated 3D scene.
Explore catalogue names and celestial coordinates over a solved image. WCS-aware overlays follow the astronomical field rather than becoming permanent image pixels.
Identifying objects in a solved image.
Bring overlapping panels into a dedicated workspace for alignment, control points, masks, composition preview and final rendering.
Composing overlapping, already-processed panels.
Launch the focused model editor with -gui: ROI, model intensity, switchable Autostretch and Done return an image to the calling script in its original container format.
Letting an external script pause for manual model processing.
Start a stack, open an image or project, compose channels, or launch the guided tutorial.
You are beginning a new session.
Open FITS, XISF, TIFF, PNG, JPEG, camera RAW, SYQPRJ and self-contained SYQ documents.
You already have a master, photograph or saved Studio project.
Decode DSLR and mirrorless RAW data with controlled demosaic, white balance, highlight recovery and camera colour handling.
Opening CR2, CR3, NEF, ARW, DNG or another supported camera RAW file.
Zoom, fit, pan and move through large astronomical images without changing pixel data.
Inspecting structure at different scales.
Recover processing decisions and preserve the intent of a layered workflow.
Exploring alternatives or correcting a recent edit.
Reframe, resize or rotate the complete document while keeping layers, masks and WCS relationships coherent.
Correcting framing or preparing a final composition.
Heal defects, clone signal, dodge, burn and add text on editable non-linear layers.
Removing presentation defects or applying final local corrections.
Separate stellar signal from the deep-sky background and create an independent Stars Only layer.
You want to process stars and background separately.
Locally denoise and restore astronomical structure using profile-specific neural inference.
Noise or acquisition softness obscures real signal.
Reduce stars, correct stellar shapes or deblur compact structure with local neural inference.
Stars are too dominant, optically distorted or mildly soft.
Remove complex large-scale illumination from linear images through a fixed-shape neural background model.
A linear wide field contains gradients that are difficult to sample manually.
Transfer reference chromaticity, luminance or both while preserving the geometry of the source image.
Matching the appearance of another rendering of the same field.
Reveal faint linear signal with a visual-only statistical transform, then optionally make it permanent.
Viewing linear data before committing to a non-linear image.
Shape black, midtone and white points with SyQon’s iterative tonal engine.
Moving a linear master into a controlled non-linear state.
Establish a natural tonal and colour starting point through grouped development controls.
Beginning broad non-linear image development.
Shape luminance and colour response with precise control points.
Adding local contrast or targeted channel changes.
Neutralise the image background and balance white from image statistics.
A linear RGB image has a global colour cast and plate solving is unavailable or unnecessary.
Set black, midtone and white points numerically.
You need predictable global range control.
Balance sky luminance and colour without globally reworking the target.
The background feels too bright, dark, warm or cool after stretch.
Increase midtone colour while protecting shadows and saturated extremes.
Colour needs presence without clipping stars or noisy background.
Grade six hue sectors, balance RGB by tonal range and control vibrance with luminance and gamut protection.
Performing selective creative colour finishing.
Use three-way colour wheels, image picking, selective grading and optional OpenColorIO workflow.
A colourist-style interface is preferable for final grading.
Stretch an Axiom Stars Only layer using the SETI Astro transfer with optional colour boost and SCNR.
Reintroducing linear stars into a processed background.
Calibrate, analyse, register and robustly integrate classic, multi-session or mosaic acquisitions.
Starting from raw light and calibration frames.
Measure FWHM, eccentricity, background, noise, stars and relative frame quality before integration.
Rejecting weak frames or diagnosing acquisition quality.
Register multiple panels into a common spherical canvas with overlap, coverage and seam diagnostics.
Combining adjacent fields or sessions with different geometry.
Plan a rotated, overlapping sky grid before acquisition and export tile coordinates.
Designing a future mosaic sequence.
Solve an image locally with ASTAP and write a complete celestial WCS solution.
You need sky coordinates, SPCC, overlays or WCS-aware mosaics.
Calibrate linear RGB colour against local Gaia DR3 XP spectrophotometry.
A solved linear colour image needs catalogue-based colour calibration.
Register open documents using stars or manual control points and inspect distortion residuals.
Combining masters or matching layers from different sources.
Build a non-destructive document from pixel, component, channel and mask layers.
Combining processing stages without overwriting the source.
Create tonal, stellar, filament and ultra-faint diffuse-nebula protection masks.
A process must affect only selected signal ranges or structures.
Paint, erase or place linear/radial gradients directly over the image.
You need spatial control that automatic masks cannot describe.
Combine multiple masks through Multiply, Minimum, Maximum, Add and related operations.
Protection depends on both structure and location.
Import and combine RGB, luminance or narrowband masters into a colour document.
Starting from separate mono filter masters.
Adjust Axiom stellar flux, saturation, temperature, tint, RGB balance, cores and halos.
The Stars Only layer needs independent finishing.
Create new colour relationships with an RGB 3×3 matrix and composition presets.
Remapping channels or producing controlled colour combinations.
Preview and apply SHO, HSO, HOO and alternative emission-line mappings.
Working with separate SII, Ha and OIII signal.
Evaluate custom expressions for output channels and combine the active image with another document.
A channel operation requires exact mathematical control.
Assign or convert RGB working spaces, adapt white points, inspect gamut and soft-proof output.
Preparing predictable display or print colour.
Access local contrast, MLT/MMT, convolution, FFT, LHE/CLAHE, range, morphology, denoise and PSF deconvolution.
You need classical, measurable structure control rather than a neural model.
Reveal coherent low-SNR filaments with a diffuse-rejecting multiscale ridge detector.
Extremely faint filaments are buried near the noise floor.
Rebuild compact saturated stellar cores from unsaturated wings using a flux-continuous Moffat PSF.
Bright stars contain clipped or damaged compact cores.
Model and remove uneven sky illumination with automatic or sampled background extraction.
The image contains vignetting, light pollution or broad residual gradients.
Reduce statistically excessive red, green or blue chrominance while preserving luminance.
A residual colour channel contaminates the background or target.
Enhance nebular detail with noise and star protection.
The final non-linear image needs restrained output sharpening.
Build a multiscale high-pass detail stack with gray relief view and star protection.
Enhancing coherent nebular structure across selected spatial scales.
Enhance coherent real structure and add controlled sky microtexture for final presentation.
The processed image needs crisp structure and a restrained finishing texture.
Recover highlight and bright-core detail without flattening the entire image.
Galaxy or nebula cores dominate the available range.
Compare the current result with its source and inspect model behaviour in a movable native-scale region.
Evaluating whether a process genuinely improves the image.
Monitor histograms, channel relationships, luminance, local signal and map diagnostics.
Making decisions that should be supported by measurements.
Measure pixels, local noise, profiles and stellar PSF without modifying the image.
Checking background quality, star shape or a suspicious structure.
Display sky coordinates and WCS-aware information directly over a solved image.
Identifying position or validating astrometric geometry.
Write scientific linear data or presentation-ready images with explicit format, precision and colour handling.
Publishing a result or exchanging data with another application.
Preserve editable Studio state as a self-contained scientific archive or compact linked project.
Pausing work or transferring an editable processing document.
Keep the application, neural weights and catalogues current without downloading unchanged assets.
At first launch, after a purchase or when a new release is available.
Run supported models and file workflows headlessly from scripts, PixInsight, Siril, SETI Astro or another application.
Automating batches or integrating SyQon into an external workflow.
Select the active layer, open the component, preview representative signal, decide whether to create a new layer, then Apply. Black mask pixels protect; white pixels reveal the effect.
Use Fit to understand the whole composition and 100% zoom to inspect stars, noise and reconstruction. A process should survive both views.
Prefer a new layer for meaningful processing steps, name important layers and save an editable SYQ or SYQPRJ project before final export.
Download Studio, open a master or begin a Fusion Stack session, and return to this guide whenever you need the next tool.