I’ve said that I wasn’t going to test the Sony a7III. The testing I do takes a while, do I need to purchase cameras I test, and I don’t see much of a place for the a7III in my own photography since I already have an a9 and an a7RIII. But working with files that others have sent me made me curious, and I’ve about decided to get rid of my a7RIIs (nothing wrong with them, but I want to deal with only one kind of battery, and the user interface differences between the second and third generation cameras are driving me nuts), so, when Adorama got the a7III in stock, I ordered.
The camera is here now, and I’ve started testing two odd things about the camera:
- The read noise frequency response at ISOs of 100 and just above, and 640 and just above
- The PDAF striping
I’ll be reporting soon, but the PDAF striping artifacts bring up a larger issue: Internet tribal behavior in the face of camera idiosyncrasies. So far, most of the digital ink about PDAF striping has been passionate advocation of one of the following positions.
- 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.
I’ve seen this movie before. Two cases come to mind, both involving Sony cameras.
When the Sony a7R came out, there was a lot of noise about “shutter shock” (which was kind of a misnomer, since the root cause of most of the vibration was from the motor that wound the shutter). Mike Collette, Joe Holmes, Huntington Witherill, and I, along with others, spent a lot of time and effort analyzing the cause of the problem, quantifying its effects, and seeking mediation strategies. For our troubles, we were attacked by two groups of people who said:
- The a7R shutter shock 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 shutter shock on the a7R occurs only under an extremely rare set of clearly-identifiable circumstances. It’s a nearly-perfect camera, and everybody who says something different has a hidden agenda.
Eventually, we came up with a set of circumstances where shutter shock was significant, another where it was inconsequential, and – unfortunately – a set where the behavior was not well defined. A few people seemed to care about that. Finally, Sony released a replacement for the a7R which didn’t have the problem, and the hoopla died down.
When the NEX line of cameras came out, and especially when the first alpha 7 cameras were released, there was a big fuss about Sony’s lossy raw compression algorithm, which was always operating in those cameras. There were two schools of thought:
- The a7x lossy compression is a fatal flaw. You can never know when and where it will affect the final image, 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 lossy compression on the a7x affects images only under an extremely rare set of clearly-identifiable circumstances. They are nearly-perfect cameras, and everybody who says something different has a hidden agenda.
A few people reverse-engineered the Sony compression algorithm. I wrote software to simulate its effect on images. Eventually, we came up with a set of circumstances where raw compression artifacts were significant, another where it was inconsequential, and – fortunately – a very small set where the behavior was not well defined. A few people seemed to care about that. Finally, Sony released a firmware update and new cameras which – with what I’ve called malicious obedience – allowed uncompressed raw but not losslessly compressed raw, and the hoopla died down.
I’ve emphasized the similarities with some cutting and pasting above; I apologize if that made it a boring read.
I find this tribalism distinctly unhelpful. The arguments for the extreme positions obscure complexities that are important to photographers, and the filling of fora with extreme comments makes it hard on people who just want to get information about what camera to buy and how to use it to best effect. Not only is it difficult to separate the wheat from the chaff, it is unpleasant to wade through all the name-calling, ad-hominin arguments, and outrageously false statements.
Having said all that, I expect the cycle to repeat with the a7III PDAF striping. Undaunted, I will continue to work on it.
Here’s what we know:
- It occurs with the a7III with some lenses in backlit situations with lens flare present.
- The a9 had similar issues, but people don’t seem unhappy with the a9.
- It is related to off-axis light impinging on the sensor.
- It is related to the sensor rows with PDAF pixels.
- It is not strongly dependent on f-stop.
Here’s what we don’t know:
- If there are any lenses that can’t generate these artifacts on the a7III.
- If you can get objectionable artifacts without visible flare (we do know you can get measurable and visible artifacts in parts of the picture with no visible flare, and measurable artifacts with hardly any visible flare present at all).
- The exact mechanism by which the artifacts are generated.
- If we can make a short list of situations to avoid if you are worried about PDAF striping.
I’ve already done some preliminary tests, and have proven to myself that, with some lenses at least, I can generate the effect at will. Some of the images that started the furor over these artifacts were made by DPR’s Rishi Sanyal using the Sony 85 mm f/1.8 lens on the a7III. I don’t have that lens, but Horshack is sending me his to test; I should have it shortly. I would like to start by doing a series of tests of 85 mm lenses, including the Sony 85/1.8, the Zeiss Batis 85/1.8, the Otus 85/1.4, and maybe a few others. I would like to see if I can pin this lens-dependence characteristic down.
So stay tuned.
Jim:
Since lens flare seems to be part of the equation for generating the banding problem, and since the Sony is mirrorless, well suited to adapted lenses, it’d be interesting to also test out, if you have any handy, some old single-coated and multicoated 85mm optics. I’m thinking of the tail end of single-coated optics and the beginning of multi-coated ones from the major manufacturers of the time, late 60s – early 70s.
In some cases, such as the 85mm f1.9 Super-Takumar versus the 85mm f1.9 Super-Multi-Coated-Takumar, the optical formula is supposedly the same, only the coating differs, according to https://www.pentaxforums.com/lensreviews/Super-Multi-Coated-TAKUMAR-Super-Takumar-85mm-F1.9.html
That’s probably also the case for the pre-ai Nikkor 85mm f1.8, HC versus H versions. According to http://www.photosynthesis.co.nz/nikon/serialno.html, http://www.photosynthesis.co.nz/nikon/lenstype.html,
http://www.destoutz.ch/lens_85mm_f1.8_265911.html
http://www.destoutz.ch/lens_85mm_f1.8_390085.html,
the Nikkor HC is multicoated, but otherwise the specs and appearance are virtually identical, so one might infer that the optical formula is probably the same.
There may also be analogous pairs in Canon, Minolta, or Leica, I don’t know.
A thought;
The microlens array on the sensor is usually a die-stamped thing with spherical or aspherical lenses cut into a square aperture for each pixel. They tend to be in the vicinity of f/1.8 to f/2 for current FF cameras. It is unclear to me if the microlenses are tilted towards the center of the sensor for the corner of the picture. That IS the case for Leica, but I do not know about other manufactures. Larger apertures, or flare which has similarly large angles, exceeds this and has to go somewhere else. Only about 4% is reflected away towards the lens. Since the CFA is underneath the microlenses, and under that is some shiny silicon and wiring/bonds, my guess is that there is total internal reflection that can carry light quite far from something like a flare.
Because the DPAF columns disrupt the normal microlens array, the would remove one of the surfaces facilitating the carrying of the light a la a fiber or light pipe. In effect, this would cause an immediate end of any “bloom” effects from flare.
I believe a reasonable way to test this theory is to get a small diameter collimated beam of light. Putting something like a lightbulb at the focus of a camera lens is a pretty good way to do that if you don’t have an alternative.
Shine it on the sensor with no lens on the camera head on and see what you see. Increase the angle of obliquity more and more and more and see if eventually you can see a reasonable amount of light not inside the beam, and then a vertical line where it was disrupted.
The flange of the camera may inhibit you doing this at very large angles.
But in the sample shot Jim posted, the PDAF columns are brighter, rather than darker, than the surrounding ones.
That’s a good point. Maybe whatever they do for the DPAF pixels (I do not know) manages to collect the light instead of sending it e.g. back towards the lens.
My comment was more an idea than a definitive answer 🙂
The collimated light is a good idea although the only way to get that would be with a laser and seeing as lasers can damage sensors you’re going to have to be brave to test that out (perhaps Jim has access to some low powered ones). I suspect that the DP pixel is masked to light entering at one set of angles and thus is more sensitive to off axis light from flare resulting in a brighter than expected result.
All good but when the affected 11+7 bit photo is perhaps the best photo you’ve ever taken of your wife – one tends not to forgive nor forget.
I’ll admit the ‘one handed’ click the shutter style of photography I was using that day wouldn’t have been possible with the Nikon gear at that time – too heavy, no tilting screen but still Sony lied (they advertised 14- bit) and didn’t do anything to straighten it out other than – here’s a new camera spend another $3000 and trust us.
Well as you know the A7rii dabbled in non 14 bit also.
“There’s an old saying in Tennessee — I know it’s in Texas, probably in Tennessee — that says, fool me once, shame on — shame on you. Fool me — you can’t get fooled again.”
Arthur I had the same experience. Was well aware of the issue but had taken thousands of shots and never noticed anything so I was quite dismissive of it. Then one evening I set up a meticulously planned cityscape where I had to get special permission and was devastated when I reviewed the photos the next day. Thankfully I’d done some shots with my Nikon 1 when setting up which turned out much better. The Sony is still a great camera and almost perfect for most of my needs but since then I don’t have the same trust in it I used to have especially as Sony has only fixed it in newer models.
I’ve shot more than 20000 (haven’t checked lately, could be a much larger number) pictures with my A7RII and not once had an image ruined by this. I’ve looked on the Web for examples of the phenomenon and all examples seemed to be backlit with a clear shadow area where the effect manifested itself. From my own archive I only managed to find one photo that with extreme adjustments showed the effect and this had no practical implications since the effect was too weak.
Thus, I don’t personally feel that striping is in any way a significant issue for me, at least in the R-series cameras. However, it’s interesting to follow the topic, DPReview did a service when finding out that the Sony FE 85 is prone to striping; I don’t have that lens and the equivalent Batis has been completely problem-free so far.
Thank you and well said.
These tribal battles take areal-world toll on those of us who just want to enjoy photography and get he most out of our gear. In my case the lack of clarity on compression lead me to keep using compressed until very recently. My hobby is bird photography and feather detail is very important. My first two outing shooting uncompressed raw (A7rII) gave me what on first impression are the best feather detail I’ve ever captured. Not sure if that result will hold up when I spend more time comparing to past work and as I shoot more – but I will be very disappointed to have waited so long to test if the results really are better.
On that front I applaud the “malicious obedience” description. I held off because Sony gives only a “false choice” between obscenely large files and lossy compression. I find it unlikely that here is no lossless algorithm that can be implemented on available hardware with little or no processing lag time difference betweeen uncompressed and lossless compressed (trading off more compute to compress vs r less write-time from smaller files). If there is an observable quality difference in real-world shooting (like feather detail on birds) Sony should be ashamed they haven’t measured that difference and addressed it through lossless compression. If I’m hallucinating and there is no difference I sure wish the “tribes” hadn’t fogged up the issue so much that I am left to doubt and have to invest in drawing my own uncertain conclusions.
This site is one of very few I look to for sound information and I treasure the resource – even when I can’t quite understand how the results (of in this case compression testing) would apply to my use-case. Still better than the noisy speculating masses – if it weren’t for this site concluding lossy compression could be significant in some circumstances I’d have never even tried uncompressed. Thank you (again), know that your efforts are deeply appreciated.
John (above)
Not trying to challenge you but what is your reference point for “best feather detail” ?
Almost all of the bird and nature pros I know use the 400mm to 600mm Nikon lenses and one of the high end Nikons (or lately the very affordable D500).
When I’ve compared these to Sony’s best the Sony falls far short.
Arthur
Hi Arthur,
I compare my results to the best I can find on the Web. No one in particular as so far it is not hard to find images better than my best – but virtually all are Canon or Nikon. In my experience w/ compression I just don’t get the fine detail throughout the in-focus portion of an image , some feathers always blur into each other and look ‘muddy’ – but w/o compression my results so far look closer to the best of what I see on various online sites w/ bird photos, feathers stay distinct w/o the ‘muddy’ blurring. Nothing I can prove yet, I’ve not spent enough time comparing closely to be certain I’m not just seeing what I hope to see. Perhaps I’ll post again in a few months when I have more confidence one way or the other.
BTW – I shoot w/ the (manual focus) Nikon 500 f4p via adapter. I only get birds patient enough for me to focus, but the f4p is light enough to hand-hold when shooting 1/000th territory. I sometimes even get blur-free shots at 1/250th w/ the in-body stabilization. Manufacture date of my lens is around 2000 so optics shouldn’t be too far from state of the art – so I kind of expect I should be able to at least occasionally come very close to the best images I can find online. In a way I’m glad to hear you say the Sony falls far short – I’d never blame my gear but that won’t stop me from taking some pleasure in the thought that perhaps I am at a disadvantage. I’m in it for the time outdoors and the challenge not for fame or money and certainly not for brand loyalty. so I’ll keep plugging away even if my gear isn’t the best there is.
Cheers,
John
John –
Fair enough and I understand. We all have to draw the line somewhere and make compromises in life. It’s the only key to happiness. I am sure my best Nikon setup (D850, 300 2.8 and a 1.4TC) are a compromise to my friends 600mm lens on his D500.
I’ve been looking at Porsches lately and the 911 I want is a compromise to all of the GT3 and Turbo drivers out there – so I get it.
But even this AM, I see on Lloyd’s site the most disastrous bird photo I’ve ever seen ‘proudly published’. (Even alarms me to some degree that this passed his normally critical eye as he suffered a TBI about a month ago).
This is not a good photo and I don’t need to qualify that with ‘my opinion’ or any other gating terminology.
A good bird photo clearly shows not only the barbs but the barbules of the feather, at least for some area of the photo based on DOF.
Birds are tough, they breathe and vibrate at faster rates than 1/250th and in my estimation 1/400th + is needed to eliminate motion blur of the (resting) bird subject itself. Of course, no form of stabilization helps with this matter.
I don’t care for birds one way or another but they are fun to photograph and they make excellent subjects for honing your skills and understanding of how to photograph.
My insistence of almost always sticking with ISO 64, the high shutter speed and (I’ve been there and ultimately gave in to AF and it’s cost) AF lenses allow me to get defined barbules when taking bird photos.
Last but not least, all of my bird photos were when I could afford the time to set up a blind, bait with seeds and nuts and wait for the birds to come. It does take a lot of time and patience to draw them in close.
Lloyd’s photo –
https://diglloyd.com/blog/2018/images/2018-0419-_DGL0269-YellowHeadedSomething-f2_8.jpg@auto.html?dglySz=2304×1728&dglyD=-C_ogRl4-XZrCDjJg6LRj-Teo3y0aMvqIqGdn8MvrLfEoYDHu3AptEdbRwmSjGKubqTaGWam8elxTCS2Qzg4q0ADvlj2KNZK1ct8CsBeG1iRIC6XJuCwVLd–zYhIw_dmCt6krJdlXsVTPe_FQgHDAuWAxa78PZHOkxt0WoHJN7cplA5iLc82FF-uLH7Cur2zlV8dZtYoiZJCDwtKgJRCLIs9V9ylvQUMxpcBFgcichylDqlt5tDT3RpyOXfFZWPplS6AhXMBGyx
Hi Arthur,
Since you have been so kind in comments I will share some 8MP images I sent to a birder who saw me shooting and asked. Would love to have you take a look and give your honest opinion of technical quality. If you want me to post raws I am willing, but not tonight. These are from uncompressed raw.
I am shy about my work so never post, but I have to agree lloyd photo leaves a lot to be desired. I don’t think I would have considered that a keeper when I started w/ a Nex 7 and Leica 560 6.8 (back when I thought I could be really cheap about this hobby). Definitely not a keeper by my current standards.
Below is link to images that are at high end of my current standard for keepers:
https://drive.google.com/open?id=14YEb8maEMg91opL7YUpEE4T9sNyCj295
Cheers,
John
Excessively sharpened in post is my first observation.
Your ‘technique’ is good as far as getting close enough to the bird to get some detail.
I’ll see if I can dig out some bird photos from my stored. Give me a few days.
It turns out I deleted all of my Imatest and bird photos about 2 years ago but I did find this one JPEG. Relatively sure I did not apply any sharpening in post.
The DOF is shallow due to this being a 300m @ 2.8 – I want to try using the Otus 28 and 55 with a remote in the future but even then, I think at 5.6 I am still having DOF issues at ultra close distances. 2-4ft.
https://drive.google.com/file/d/1vykJJlrdYsrvC3WvbGyxyCU1dyVsPHGE/view
Just noticed that you will need to download and open with paint or other program to see at 200% + . Google’s online viewer is crap. File should be 7.06mb
Apologies for dropping off for a bit – life intrudes sometimes.
First big clue from the very nice jpg you sent is that if I want great detail I need to get closer. Although I appreciate your compliment on my birding technique, the reality is that I shot the images I shared at Great Meadow in Concord, MA. The birds there are so co-operative that I believe the forest service must be giving them a cut of the gate in modeling fees. I don’t do as well with less accommodating subjects, definitely still learning how to get close.
Even with co-operative subjects, I believe all the images I posted are crops, not sure the distance but not right up against he close focus limit of the 500 f4p which I believe is ~16′ – birds in images I shared are probably over 20′ away. I’m also partial to f5.6 for high shutter speed so DoF can be less than ideal – if I front-focus even a tiny bit it really shows in loss of detail.
On sharpening, the heavy hand is mostly out of frustration – I so want detail that I’m not getting that I over do the sharpening trying to pull out detail that isn’t there. My aging eyes don’t help either. I’ll try to rein myself back in. Will have to spend some time looking very closely and figuring out best sharpening.
Really appreciate your feedback and sample image. Gives me a benchmark to strive for.
Thank you.
Hi Arthur,
Thank you for the great blog I am currently using the A7Rii and was looking at replacing it with the Aiii
Do you think the low light ISO would stand up against the Sony A7sii?
Thanks
Kevin
Part of the problem is the mods in DPR, I havent noticed the same toxicity in other photography forums. In the suggestions forum of DPR, I posted a request for a new forum particularly for people who want to discuss both Canon and Sony. As justification, I stated that anytime I try to compare anything from one brand to another in either of these forums, I get jumped by “fanboys”. A DPR mod deleted my suggestion post, saying I could repost it without using the term “fanboy”. I didnt repost it as I am now of the opinion that DPR is not a forum I wish to contribute to and try to improve, at least not with mods like that. It wasnt like I was directly calling any particular person a “fanboy”. I guess I was pointing out a problem with the forums using terminology which the mods basically didnt like, so they deleted my post. Well, fine, I am not going to waste my time trying to improve their forum if they get all bent out of shape and have to use the mod hammer to protect their fragile little egos.