Since this post about the Sony a7III PDAF striping, I haven’t written anything on this blog about the issue. That’s not because I haven’t been on the case. I’ve been working understanding what’s going on all this time. But, because the effect is so variable, hard to quantify, and varies with what seems like everything but the phase of the moon, I’ve been progressing in fits and starts. I’ve been posting my progress on DPR, and getting help in avenues for further exploration, as well as a generous helping of abuse for what some folks perceive as bias.
Although you haven’t seen it on this blog, I’ve done a lot of work on the PDAF artifacts, and I have reached some conclusions as to how serious a problem this is. I’ve also learned what the variables are, and they are daunting in their breadth; so much so that given what I consider the severity of the issue – more on that down the page – I won’t be exploring all of them in detail.
Those of you who are vaguely uncomfortable with numbers, get a cup of coffee and rejoice: this is the kind of post you’ll like.
First off, I’m going to compare PDAF striping to several other Sony a7x problems:
- Lossy raw compression
- Shutter shock in the a7R
- The ever-morphing “star-eater” digital filtering
All of these consumed a lot of Internet bandwidth in their day. None of those materially affected most people’s photographs most of the time. All of them affected some people’s images some of the time. All of them have partial or complete workarounds or mitigation protocols. Two of them have been fixed by Sony in subsequent cameras and/or firmware releases.
In my analysis, I‘m going to consider three things:
- How likely is it that image damage is noticeable
- How easy is it to know in advance whether the damage will likely occur
- How objectionable is that damage
- How onerous and effective are mitigation strategies.
Lossy raw compression is highly unlikely to materially damage images, the situations where such damage may occur are describable to and recognizable by a moderately-skilled photographer, the artifacts can be fairly ugly, and, unless you have a camera that lets you turn off the compression, mitigation can be moderately difficult.
A7R shutter shock is moderately likely to damage image, the situations leading to damage are moderately difficult to predict, the damage is usually not particularly objectionable (to the point where many think there is none), and highly effective mitigation is a real pain.
Star-eater filtering is no problem at all except for those photographers who engage in esoteric pursuits like making tracked night-sky images and maybe photographing fabrics with long exposures. It is easy to know if you’re likely to have a problem, the damage is only objectionable to a few, and one kind of mitigation is hard (the other, avoiding the problem shutter speeds, is an inconvenience to some, and too much trouble to consider to others).
PDAF striping is not likely to materially damage images. Predicting exactly when it will occur and how badly is difficult, but it’s pretty easy to describe a broad range on circumstances where it will never occur. The damage is at least as ugly as the raw compression artifacts. We already have some mitigation techniques, and I expect we’ll have more in the near future.
If I had to rank-order the above Sony vicissitudes in terms of their effect on most people most of the time, from worst to best, my list would be:
- A7R shutter shock
- Sony raw compression, PDAF striping (tie)
- Star-eater filtering
So, without further ado, as Jack Webb used to say, “Just the FAQs.” (I may have misremembered that line.)
Do other cameras have PDAF striping? They do indeed. In full-frame Sonyland, PDAF striping first appeared on the a7RIII. It also occurs on the a9, and now the a7III. Some other cameras show these kinds of artifacts.
Is it worse on the a7III? It’s about the same as on the a9, and both those cameras are worse than the a7RIII.
Those other two cameras have been out awhile. I never heard about this before. What’s the big deal with the a7III? Hard to say. I suspect that the a9, being optimized for sports and wildlife photography, is less likely to be used in the circumstances that provoke the striping. It may be that the a9, expensive specialist camera that it is, just shipped in far lower quantities than the a7III. Maybe the circumstances at Sony’s a7III launch event were especially conducive to triggering the striping.
Is it worse with some lenses? Indeed it is. The Sony 85 mm f/1.8 is especially problematic.
Are some lenses immune? If you really push things, probably not. I’ve not found a lens in which the right amount of flare in the right place pushed hard in postproduction won’t cause striping.
Are some lenses so good with respect to striping that I can not worry about it? Yes. I put the Sony 12-24/4 in that category, but it is possible to make it stripe on the a7III if you want it to, and I have seen an image in which it looks like the striping was caused accidentally.
Can I get striping without lens flare? I don’t think so.
If I see lens flare, does that mean that there’s striping? No.
Is there any place in the image where it’s especially problematical to put a bright light source? Technically, yes, but it varies with lens, and is probably not worth trying to track down.
Does increasing the ISO setting reduce the striping? Yes, it does. By ISO 3200 or so, it’s hardly ever noticeable, being masked by noise.
I just post images to the web. Do I have anything to worry about? No.
Is this something people will see looking at full frame images on a 4K screen? Unlikely.
How big do prints have to be before you see the artifacts? It depends on how bad it is, how good your vision is, and how close you get, but by and large you’re safe at 8×10 inches and smaller. At 17×22, you might take some care. Bigger than that, you should have a good look at images taken in problem situations before you spend a lot of money on a print.
How likely is an amateur who’s not paying attention to wander into a situation that causes striping? Not very likely. There are a few exceptions: sunsets, wide-angle shots with the sun in the frame, strong flare from stage lighting, and the like.
How likely is an amateur who’s not paying attention to see the striping if it’s there? Not very likely. See the discussion about prints and screen viewing above.
How likely is a striping-clueless pro advertising photographer to provoke the striping? Since PDAF striping came up, I’ve been looking at advertising images with an eye to answering that question. I see a lot of images where striping could be a problem. It seems that strong backlighting with the light sources visible in the frame, and pronounced lens flare is quite fashionable these days, and pros with deep pockets have no difficulty affording powerful strobes.
How likely is a pro advertising photographer to be using an a7III or an a9? Not very likely, especially now, unless they never employ the kind of lighting that causes the problem. I think that, if they’ve recently upgraded their gear, they’re more likely to be using a D850, an a7RIII, a GFX, or a high-end Canon. If they haven’t, they’re probably happy with their D3x.
How easy is it to fix the problem after the fact? If there’s not a lot of detail in the area with the striping, it’s simple to fix it in Photoshop. If there is, pretty good fixes are not hard. It is best fixed before or during raw development. Raw Therapee already has a fix in the current beta version. There’s another browser-based fix available on the web. I think more raw developers with be taking their cue from Raw Therapee.
If it’s so easy to fix, what’s the issue for the pros? They may not find the striping before the client sees it. It’s not all that easy to see sometimes, but once you’ve seen it, it’s distracting, artificial-looking, and frankly, ugly. Not something you want to have get in the client’s hands.
This is making my head hurt. Net it out for me. Did Sony make a good tradeoff when the increased the PDAF areas? I think they did. In the case of the a9, that is super important to their target customers, and the likelihood of striping for those folks is very low. In the case of the a7III, with an almost-entirely-amateur customer base (my guess), it makes really desirable features like eye-AF work much better, and the chances of that group of people simultaneously triggering the striping, noticing it, and finding it objectionable are low.
Maybe I should just buy a DSLR and forget it. If you are otherwise a candidate for the a7III, I think that would be throwing out the baby with the bathwater. There are a host of downsides to a DSLR (like lens AF tuning), as well as some upsides.
I’ve had some more questions, so I’m adding them to the list.
Is PDAF striping as much of a concern as aliasing in all it forms? Not even close.
Is PDAF striping specific to Sony cameras? No.
Is PDAF striping endemic to Sony cameras? No. The 33×44 sensor in the GFX and X2D doesn’t even have PDAF, much less the problem. The sensors in the Nikon DSLRs don’t have on-sensor PDAF, so they can’t have the issue.
Is PDAF striping endemic to on-sensor PDAF? No.
Does turning off EFCS fix the problem? No.
Does turning on the electronic shutter fix the problem? No.
Erik Kaffehr says
If you compare PDAF striping with aliasing in all it forms, which issue would be more significant?
With regard to the raw compression used by Sony, I think that reviewers handled that poorly. It was blamed for a lot of artefacts it could not be the cause of. On the other hand, I think that Sony should not have employed the delta compression but the tone curve compression is basically sound.
Sony has solved the problem using a 16-bit file format with two unused bits. I ask myself why they didn’t do a <<2 operation like a famous company in Copenhagen?
Constructive discussions may have led to some remedy dropping the delta coding, but keeping tone curve and perhaps adding some Huffman kind compression.
The shutter shock issue was a serious one. As far as I could see, it was a real effect, affecting every picture shot with moderate shutter speeds. My impression is that careful shooters, like tripod users, were most affected.
Another question may be if the PDAF striping is specific to:
– Sony cameras?
– Sony sensors? Just as an example, would also Nikon D850 be affected?
– Generic use of on sensor PDAF?
I would hope that the root cause of the issue will be found and a workaround in hardware can be developed.
Thanks, Erik. I’ve added some more questions and answers based on your comment.
After testing the striping issue, will you make a general review of A7III like the articles about A7RIII?
I care about the SOOC JPG color science and the performance of the ACR/LR start point profile. It seems to me that the SOOC JPG of A9 and A7RIII looks more comfortable than that of A7III (more greenish skin tone in shadow area under artificial light). I’ve seen tons of SOOC JPG samples of those three cameras but they were not shot under same conditions.
I see purple flare in many of pictures when people get PDAF striping. Is that some kind of sensor reflections?
I wonder, since it seems so lens-specific, if the root cause of that specificity is the distance of the rearmost lens element surface to the sensor surface and the angle of incidence of the light coming from that rearmost element. If so, could you predict other suspect lenses based on that information?
I don’t think it’s anywhere near that simple, because it is associated with lens flare. In fact, as far as I know, it can’t occur unless there is lens flare. You can’t predict that from the position of the rear element. And, by the way, you can’t predict the ray angles of the non-flare bundles from the position of that element, either. For that you need to know the location, shape, and diameter of the exit pupil.
Oskar Ojala says
I did see striping in one of my photos taken with the A7RII, but I have to push contrast far beyond reasonable to make it visible. Suffice to say that I saw no real-world impact of striping, but was once technically able to reproduce it. I think it can happend with any PDAF sensor, the likelihood and impact just varies. And the more cameras out there, the more likely someone is going to dig up some weird edge case behavior.
I was not able to get the a7RII to stripe.
“At 17×22, you might take some care. Bigger than that, you should have a good look at images taken in problem situations before you spend a lot of money on a print.”
It’s always a very good idea to inspect a photo thoroughly before sending it of to be printed in a large format.
Griffin Moores says
I noticed the A7III striping pretty heavily with the Sony 28 f2 shooting into the sun. In looking to mitigate flare, are there any lens filters beyond UV that could help with this problem, as I haven’t found quick solutions that adequately address this without degrading other elements of the image quality. I’ll try to keep in mind using hoods where possible, but not much you can do when shooting directly into the sun. Here is a link to an (edited) image where the problem was fairly dramatic with the position of the sun and striping marked: https://drive.google.com/file/d/1g4PxUGZiGyPNB5caIE8Bhrkl4yNBIHkG/view?usp=sharing