Sony Concentric Coloured Polygons

Introductory Comments

The purpose of this web page is to document the coloured concentric polygons caused by a lens correction on Sony mirrorless camera which is applied to the raw data whenever a lens is recognised by the camera firmware.

At the outset it is worth pointing out this correction is not applied when the camera is used with (non-electronic) legacy lenses or with telescopes.

Note that this is a different problem from the concentric rings and weird horizontal banding that affects the Sony A7S camera and is discussed here.

Example of Problem

A good example of the problem is this one:

The example can be found in this DPReview post: Cause of green/purple rings in fog photo?

It was shot with a Sony A7RII camera with attached Voigtlander 65mm F2 Macro APO-Lanthar lens at F/3.2

The "polygonal" rings appeared when the image was pushed in post-processing. Indeed, the rings typically appear in underexposed terrestrial images that have been "pushed" in post-processing.

Unfortunately, it is also a serious problem for deep-sky astrophotography because the processing workflow designed to reveal faint structures and nebulosity in deep-sky images will also reveal faint artifacts embedded in the raw data by crude in-camera processing such as lens corrections.

The "polygonal" shape of the rings is still a mystery but is probably due to the nature of the crude lens correction being applied

Investigation of Problem

This issue was investigated in numerous threads on the DPReview forum. Here's an example of an image of a white sheet of paper where I've opened the raw file and turned up the saturation during processing in Photoshop.

Again this was shot with the same Sony A7RII camera with a Voigtlander 65mm F2 Macro APO-Lanthar lens at F/3.2 Notice how the coloured polygons are in the same position as the earlier fog photo.

The same saturation technique was used to produce this montage of images from raw files using the same lens at different aperture settings:

The horizontal coloured bands in the left half of the images are nothing to do with the lens correction but they are triggered by the fact that the left and right halves of the sensor are independent. This is something I first discovered on the Sony A7S camera and I still don't fully understand them.

Here is a similar montage from raw files produced using a Sony FE 55mm lens on the same A7RII camera:

The next montage is instructive because it comes from a Nikkor 85mm lens that was not recognised by the Sony camera firmware:

It shows that the lens correction is not applied when the firmware is unable to recognise the lens. Other users have proved this by using the same lens with and without the electrical contacts taped over. With the contacts taped over, the correction is not applied.

Can the lens correction be switched off?

Here's a comparison of two "white sheet" images taken on a Sony A7III with a Sony FE 55mm F1.8 ZA lens at f/1.8:

It's quite clear that switching off the lens corrections via the camera menu makes no difference to the "polygonal" rings. The correction that causes the polygons is still being applied whether or not lens corrections are switched on. In both cases the raw files were saved as 14-bit with image compression switched off, which helps prove this is not a quantization or compression artifact.

Channel analysis

To get a better understanding of the correction being applied, it's best to avoid raw converters such as Adobe Camera Raw and RawTherapee because they apply a colour correction matrix (as part of the transformation to the target colour space) which mixes the colour channels.

An alternative method is to debayer the (bias-subtracted) raw data, divide one channel by another and then stretch the data to reveal the faint stuctures. Doing this on the above data gives the following result:

The left panel shows the result of dividing red by blue and the right panel shows the result of dividing green by blue. They both show a simple clear progression of "polygonal" rings which in turn indicates that those rings are embedded in the data of red and green channels but not the blue. On the other hand, dividing the red channel by the green gives the following result:

This is a mixture of the rings in red channel and the green channel, which makes it pretty difficult to understand what is driving the pattern.

Mitigation of the Problem

As already discussed, the correction causing the coloured polygon artifacts cannot be disabled when the camera firmware recognises the lens. The obvious question for astrophotographers is what can be done to reduce the impact of the coloured polygons on the final stacked and processed image.
Testing has shown that the following steps help to reduce their impact:

Switch off data compression for raw files
Use 14-bit data format and not 12-bit
Do not deliberately underexpose images. For deep-sky astrophotography this means adjusting ISO and exposure length so the back-of-camera histogram is further to the right than generally recommended e.g. shoot with the peak of the back-of-camera histogram at the half-way point.

Conclusion

For lenses recognised by the camera firmware, the Sony lens corrections are still applied to the raw image data even when lens corrections are disabled in the camera menu. This is potentially a big problem for deep-sky astrophotograpy, causing concentric coloured poygons to appear in the background. The good news is that the lens corrections are not applied for (non-electronic) legacy lenses nor for directly-attached telescopes.

I cannot recommend Sony cameras for those wishing to use lenses for deep-sky astrophotography, unless those lenses are unrecognised by the camera firmware i.e. legacy lenses or adapted lenses.