Take a look at the PDAF stripes across the female model's face in this image (cropped from an original vertical shot). The bad news is that it looks terrible. The good news? There's a fix.

When we published our original Sony a7 III sample gallery, we and our readers had some concern over odd striping artifacts viewable in some images, such as the one above.

It's known as 'PDAF striping' and it's not limited to a single camera, or even a single brand. It's likely due to light reflections off the metal masks of on-sensor phase-detect pixels. Masked pixels, as opposed to a split dual-pixel design that obviates the need for metal masks, is to the best of our knowledge a design choice: the approach enables high performance AF even during 10 to 20 fps bursts.1 The unfortunate side effect is the potential - albeit rare - for single pixel stripes in transitions from blown areas to darker ones. Depending on your set of lenses and the kind of photography you practice, you might never see this issue, but if you do it can be unpleasant and distracting. Particularly if you print large, or pixel peep.

The striping does seem to be limited to certain lenses: the Sony FE 85mm F1.8 is particularly prone to this issue

In our shooting with the a7 III, prominent examples of bright stripes tended to show up up in images that contained a significant amount of flare or backlit subjects. It seemed to mostly be an issue with shots taken with the FE 85 F1.8, when subjects were captured with the intent of deliberately inducing flare (we saw it in a couple 100-400mm F4.5-5.6 GM shots, but it wasn't offensive).

We investigated the issue further with the help of our very own knowledgable forum members, and are pleased to announce that as of today there is at least two fixes available.

See the results for yourself:

JPEG processed from original Raw
JPEG from corrected (fixed) Raw

This particular fix is courtesy of Professor Hank Dietz, who is none other than the fellow who developed a fix for Sony compressed Raw artifacts. His tool is available here. For now, it only works on compressed Raw files, but it works with Raw files from any Sony camera exhibiting the issue. That's important, because this issue is not isolated to the a7 III by any means (though its predecessor the a7 II doesn't appear to have any issues). It occurs with the a7R II/III, a9, R100 V - i.e., most cameras with masked pixels for on-sensor phase-detect AF (and not just limited to Sony).

Higher resolution bodies are less problematic because a single pixel stripe is a smaller proportion of the image. The a7 III and a9 also have more phase-detection pixels than any other camera Sony has introduced, which makes the issue slightly more pronounced on these cameras.

Striping does appear to be isolated to certain lenses though: the Sony FE 85/1.8 is particularly prone to creating this issue, and we were easily able to induce striping even in modest backlit shots like the one above (zoom into 1:1 on the flare above the toddler's hat). However, we've had less luck recreating the striping with many other lenses like the 55/1.8, FE zooms, or the various 35mm primes we've tested. There doesn't appear to be much aperture dependence, based on Jim Kasson's analysis of our test shots.

You need not worry... just run problematic Raws through Prof. Hank Dietz's repair tool, and you'll get a Raw file largely devoid of any issues

What does all this ultimately mean? Well, if you shoot a lot of backlit subjects and love flare, don't worry: you don't have to steer clear of the a7 III or Sony mirrorless cameras. Instead, you may wish to steer clear of certain lenses like the 85mm F1.8 or 50mm F1.8. But even then, you need not worry too much: just run your problematic (compressed) Raws through Prof. Hank Dietz's repair tool, and you'll get a clean Raw file in return.

To take advantage of Professor Dietz' work, you simply drag-and-drop a compressed Sony Raw file into the web application, wait a few seconds, and download the resulting file (you'll want to add '.ARW' to the end for your Raw software to read it properly). And voila - almost every trace of the striping is removed. In extreme cases, you may see some remnants of striping, but out of the files we tested, this only happened once. Oh, and if you have any feedback, please PM Hank via his DPReview profile.

There's another fix independent of Hank's work. We've been in touch with one of the contributors - forum member pippo27 - for the open-source RawTherapee Raw converter, and he's integrated a fix in the Raw processing pipeline based on an understanding of the exact locations of the masked PDAF pixels.2 It's already available in development versions you can download here, and will be included in the next major release. There's even a Photoshop fix developed by DPR forum member Magnar W, if you're past the Raw development stage or have a problematic JPEG.

While we're still trying to nail down the root cause of these issues to help us predict when, where, and with what lens this will be an issue, rest assured that you needn't worry too much: the occurrence is rare, and when it appears, you'll likely have a (quick) workaround. We do hope though that Sony itself addresses the issue in-camera, especially as Prof. Dietz suggests it shouldn't even be too computationally burdensome.


1Both Sony and Canon executives have hinted to us the higher performance of the masked approach, and the computational burden of the dual-pixel approach, respectively.

2Thanks to extensive investigations led by Jim Kasson, Bill Claff and forum member Horshack.