CA / purple fringing and causes?

[Lars Hansen]

I'm curious to know if the phenomenon of cromatic abberation/purple fringing is always related to poor/cheap lens design or a faulty lens - or if this phenomenon can also appear even when using well designed lenses.

I've read that digital sensors (as opposed to film) are much more prone to exhibit this phenomenon especially in light situations where the wavelengths of the light contains more violet/UV which provokes it - edges between e.g. very dark and bright white or edges where hightlights are blown. Are certain sensor designs even more prone.



 

[Bjørn Rørslett]

The situation is complex and the lens only plays a part in what is going on.

First of all, with a light source lacking a continuous output spectrum (think fluorescent or their derivatives), the usual assumption behind the colour spatial autocorrelation in the scene breaks down. The image will typically show purple fringing with nearly all lenses. As a collateral effect focusing can also be tricky. Do note that purple fringing also can result from an overloaded sensor.

Secondly, the correction of chromatic aberrations in a lens, in particular wide angle and zoom optics, might be less efficient at certain focal settings (zooms) or towards the corners of the image (wide angles in particular, but all lenses can have this problem). For lenses with a very long focusing range, such as Micro-Nikkors or equivalents, chromatic aberrations can be more visible in the near range. Even the legendary Voigtländer 125 mm f/2.5 APO-Lanthar suffers from this issue near 1:1. Lenses with internal focusing (IF, RF designs) very often are troubled with longitudinal chromatic aberration (axial colour) making the image exhibit differently coloured fringes in front of and just behind the plane of focus.

Thirdly, as you already noted, harsh transients in contrast can trigger colour fringing of which kind the red/cyan is the most common. The cause is again to be split between subject  (excessive transients), lens (residual chromatic aberration in particular the lateral kind), and camera (inadequate handling and modelling of colour spatial autocorrelation).

[Lars Hansen]

Thanks Bjørn - I'm still digesting your reply but I realise there no simple answer - as with so many things.. 

It puts things into perspective when one reads a lens review saying that CA is minor or absent - such a statement probably applies to the most commonly tested situations but does not rule out that it can sneak up on you in other situations.

Using filters like UV, Haze and skylight etc. was very common back in the film days (probably still are..) - would such filters potentially mitigate color fringing? E.g in light with extra UV/violet.

The sensor in my Fuji X-E1 has no AA filter (which I assume is irrelevant wrt. color fringing) but it also has a different color array design and your last line made me wonder if that could play a part:

..., and camera (inadequate handling and modelling of colour spatial autocorrelation).

       


             

[Bjørn Rørslett]

You cannot in any meaningful way alleviate the fringing issues with filters - unless the filter is narrowband and/or the camera is modified to have its internal filter pack removed.

To put the latter into perspective: a lens that does not perform very well used for ordinary visible light photography, be this conducted with or without filtration, can suddenly become a terrific performer for a UV or IR camera. A good example is the humble and often shunned 43-86 mm f/3.5 Zoom-Nikkor, which is a really good lens for IR. The equally humble but much less common 35-135 mm f/3.5-4.5 Zoom-Nikkor is my first choice for false-colour IR work, where it shines, yet it is a bland and non-distinguished lens otherwise. And the list goes on.

As to the Fuji "X" sensors, they still use in principle Bayer colour filtration to build the final reconstructed RGB file, so the considerations are equal to the traditional Bayer-matrix sensors.

[Bjørn Rørslett]

To elaborate an aspect less familiar to many photographer: The apparently oddly named concept 'Colour spatial autocorrelation' boils down to the fact that in a scene, a point (pixel in the captured file) with a certain colour is likely to have neighbours with the same or similar colours to them. This is the foundation of the Bayer matrix approach for reconstruction colour information in a digitally captured image.

[David H. Hartman]

Is purple fringing more of a problem with the D800 than a lower megapixel camera?

Dave