Skip to main content

Calibration

Feature Request

I have used T20 and T74 both of which are colour cameras.

With T74 the calibration files are 3-channel which I expected, but with T20 they are single channel. I have been able to achieve reasonable results with T20 using Cosmetic Correction to get rid of hot pixels, but why not provide 3-channel Calibration files like you have with T74. The sing channel bias master for T20 is seriously over-exposed which is really strange as it should be readout noise.

Posted by Martin Tymms ·
2 comments

Log in to comment and vote

Comments2

  • Mladen Dugec

    •

    Sep 4

    Thanks Martin for pointing this out.

    I’ll need to check the new calibration Master workflow to see where I went wrong.

    My apologies.

  • Martin Tymms

    •

    Sep 3

    The Optolong L‑Ultimate is not a broadband filter with a uniform response across the visible spectrum. It has two separate 3 nm transmission bands: O III at approximately 500.7 nm, recorded mainly by the green and blue pixels of a colour camera, and Hα at 656.3 nm, recorded by the red pixels. Optolong’s specifications confirm this strongly wavelength-dependent transmission.

    Because it is an interference filter, its transmission also varies with the angle at which light passes through it. Across the field of view, changes in incidence angle can shift the narrow passbands slightly. Vignetting, reflections, sensor microlenses and the camera’s Bayer filters may therefore produce different illumination profiles in the red, green and blue data. The red-channel Hα flat can consequently have a different gradient and degree of corner fall-off from the green and blue O III flats.

    A single monochrome flat formed by averaging all three colour channels assumes that the spatial illumination pattern is identical at Hα and O III. With an L‑Ultimate filter, that assumption is not necessarily valid. Such a flat may correct dust shadows and general vignetting reasonably well while leaving colour-dependent gradients, red or cyan corners, or unequal over- and under-correction across the image.

    For accurate calibration, each colour component must retain its own flat-field information:

    Rcorrected=RlightRflat,Gcorrected=GlightGflat,Bcorrected=BlightBflat.

    There is, however, an important technical distinction: for a one-shot-colour camera, this does not normally mean debayering the flats and constructing a conventional RGB image. The preferred master flat remains an undebayered CFA/Bayer master. It contains separate measurements from the red, green and blue photosites and is applied to the raw CFA light frames before debayering. Calibration software must preserve or appropriately normalise these CFA components rather than collapse them into a single averaged monochrome flat. Siril, for example, explicitly provides CFA-flat equalisation to handle the different mean intensities of the colour components. Siril calibration documentation

    A strong concluding formulation would be:

    Because the L‑Ultimate transmits two widely separated and extremely narrow wavelength bands, its red Hα and green/blue O III illumination patterns cannot automatically be assumed to be identical. The master flat must therefore preserve independent red, green and blue—or, more precisely, CFA—flat-field information. Collapsing the flat into a single averaged channel can discard real wavelength-dependent variations and leave residual colour gradients after calibration.