Debunking false claims about Sony a7III PDAF striping

A week or so ago, I published a FAQ about the Sony PDAF striping phenomenon that people noticed with some Sony a7III images. I had previously posted a little essay on tribalism in which I talked about two tribal reactions to the a7III striping.

• The striping on the a7III is a fatal flaw. You can never know when and where it will occur, and all the fixes are a PITA and impair quality. Don’t buy this camera, and everybody who says something different has a hidden agenda.
• The striping on the a7III occurs only on a specific set of lenses under an extremely rare set of clearly-identifiable circumstances. In the incredibly-unlikely circumstance that it should occur, there are fixes that will remove it perfectly.  It’s a nearly-perfect camera, and everybody who says something different has a hidden agenda.

Since the publication of my FAQ, events have unfolded as predicted, and several false claims have emerged. From the fatal flaw camp, the ones predominate are misidentification of real or perceived anomalies in images as PDAF striping. This is a familiar pattern; it occurred with depressing regularity with respect to bogus compression artifacts in the era before Sony offered uncompressed raw files on a7x cameras. In the case of compression artifacts, some of the confusion appeared to be genuine, and many did not take the time and effort to understand what the “bar-code” artifacts from the delta-modulation portion of the craw compression algorithm looked like, or the circumstances under which they were likely to occur. Another false claim is the exaggeration along the lines of “all sunset pictures will be striped.”

From the “move along, nothing to see here” group, the falsehoods have been more varied.

Only one or two lenses can cause striping.   Not remotely true, although Some lenses are more prone to this issue than others. In fact, I haven’t yet found a lens that can never cause the artifacts. I thought that I had one in the Sony 12-24/4 zoom, but right after I posted that information, someone showed me a striped image made with that lens.

To get them, you must intentionally light the scene to provoke the artifacts. The backlit lens-flare look that allows the PDAF striping to occur is not commonly employed by amateurs, but now that the PDAF striping issue has raised my consciousness, I see it again and again in advertising and professional portrait and wedding work. Either it is trendy now, or it’s always been there and I haven’t noticed.

It’s easy to fix the artifacts, so you shouldn’t worry about them. There are two levels on which this is deceptive. First, although it’s easy to fix the stripes even after demosaicing if there is little detail in the striped areas, a robust solution needs to take place before demosaicing. There are at least two beta-level pre-demosaicing fixes, but one is browser based and the other is part of an excellent, but little-used raw developer, Raw Therapee.  But the important fact left out of the above claim is this: you must find the artifacts before you can fix them. Often, the stripes are not readily apparent at the magnifications normally employed in image editing. (I have written a program to find striping in images, but it’s not available to the public and is prone to false positives in real-world images.) Absent unusual vigilance on the part of the photographer (eg, inspecting each candidate image at 1:1 or greater), it would be easy to send flawed files or prints to a customer. If you’re an amateur you can reprint or resend if you or whoever sees your work discovers a problem. If your reputation and livelihood depend on perfect files every time, you can’t take that chance. Once an art director finds a flaw in a file (or worse, doesn’t find the flaw before it gets to the press) she’s not going to trust you. In a competitive environment, she’ll have many other photographers to pick from next time. Bad news. She’ll probably network with other ADs. Worse news.

You need to print big to see the striping. If the striping is bad and you look carefully, you can see it in 8×10 inch prints. In 13x20s, it’s not hard to see at all, although it may not jump out at you.

This whole PDAF striping thing is fake news, and not a big deal. There’s truth there, but the statement is overgeneralized. For the preponderance of photographers, almost all the time, this is nothing to worry about. You’ll probably never provoke the striping, you probably won’t see it if it’s in your files, even if it’s there it won’t show up at web resolution, and if you find it in a print, you can always fix the file and reprint. But there are people for whom the striping is not only a big deal, but it’s a big enough deal to make them choose another camera for some of their work.

My recommendation is that photographers who meet all the following criteria not use the a7III (or the a9, or, to a lesser extent, the a7RIII):

• Your livelihood and your reputation rests on nearly-perfect results
• You regularly employ lens flare in your images
• Your output medium has greater than web resolution.

If you don’t meet all the criteria above, then you are likely to enjoy the image quality

As I’ve said before, in terms of impact on most photographers in comparison to some of Sony’s other artifact-inducing quirks, I rank PDAF striping as follows:

1. A7R shutter shock. The a7R shutter shock earns the dubious distinction of first place here because, although the effects are blurring, as opposed to out-and-out ugliness, they occur in many situations. Both craw compression and PDAF striping produce ugly artifacts, with PDAF striping the more objectionable.
2. Sony raw compression, PDAF striping (tie). Both craw compression and PDAF striping produce ugly artifacts, with PDAF striping the more objectionable. The tie results from the fact that it’s easier to inadvertently generate compression artifacts than PDAF striping. Neither one is particularly easy to trigger.
3. Star-eater filtering. The star-eater filter only eats stars if there are single-pixel (or nearly so) stars to be eaten, and a missing or wrongly-colored stars is not something most folks care about. That means that the filtering only adversely affects serious astrophotographers, who should probably choose another camera.