A Bit of A7R Sanity

Posted by

A while back I wrote a post I humbly called Roger's Law of New Product Introduction, complete with the graph shown below. The release of the Sony A7R has demonstrated the accuracy of that post as few other releases have.

A few weeks after the A7R release we seem to be following the path quite nicely.

There were a lot of unreasonable expectations prior to the camera's release, and, as usually happens, those unreasonable expectations aren't being met. Of course that has resulted in the usual strident Fanboy smack talk on forums everywhere. It's also resulted in some desperate magical thinking among the group of people that wanted the A7r to fulfill all of their dreams. So let's get some of the things the A7R is not out of the way.

  • The A7R is not a 36-megapixel Canikoleica MD900 Mk IV that will shoot all lenses from every brand amazingly well.
  • Little batteries do not carry as many watts as big batteries, and when they power big sensors and electronic viewfinders they don't last very long.
  • A good electronic viewfinder is still not an optical viewfinder.
  • While improved, Sony's menus do still seem to be created by someone leaving a 'Menu Item Suggestion Box' on the wall and letting the intern add them all in two days before production.
  • Yes, they apparently tweak the raw files when they compress them.
  • The shutter is loud and causes some vibration. The 1/focal length rule isn't going to work for most people. (Truth is, it already doesn't work for most people.)
  • There are only a few native-mount lenses, and adapters are still adapters.
  • It's slow. It wakes up slow, focuses slow, and it takes 1.5 images per second.

I'm sure I left out some more reasons it's being trashed in various forums as the worst camera ever, mostly by people who haven't even touched it and probably never will. The last time I remember a camera being trashed this hard was the Canon 5D Mk II. You remember, that overpriced camera that did nothing well - except become the best selling full-frame DSLR that revolutionized what people do with DSLRs.

So despite everything this camera isn't, I think it's still a game changer. And trust me, I'm no Sony fan. I don't like the menus. Except for the glorious 135mm f/1.8 there are few lenses in their lineup I'm even impressed with (until now). With recent improvements by Nikon and Olympus, Sony clearly has the worst factory repair service of any manufacturer (at least in the U. S., not that they actually repair anything in the U. S.). So I don't like Sony. Despite that, I think this is a fascinating camera that is going to change the industry a bit.

I'll carry the 5D Mk II analogy a bit further. The 5D Mk II disappointed a lot of people when it was released because of what it didn't do. What most of those people missed was it did something really well and really inexpensively. It produced movie-quality HDMI video through excellent lenses for a small fraction of the price of the other methods available for producing movie-quality HDMI video. It didn't do it gracefully. There were all kinds of problems and constraints, but they could largely be worked around. A whole industry sprang up to help people work around those constraints and problems.

The A7R gives us something we don't have, too. It's arguably the highest resolving camera (it may be slightly higher resolving than the D800e since it really has no AA filter, but maybe it's a tie). It's small and mirrorless, which some people really want. (Notice I don't say YOU really wanted it, but there are definitely some people who do.) And it costs $2,300. A Nikon D800e is $600 more even on special. A Canon 5D III is $1,000 more. A Leica M is $4,600 more. Let me put the good points in a list like the one above:

  • Small.
  • Relatively inexpensive.
  • Incredible resolution.
  • Easily adaptable to other lens mounts.

It's not going to be the camera for everybody. But it is going to be a camera for a lot of people. I may even (gasp) buy one for myself. Despite all of the negatives, there are some game-changing positives. The fact that the two superb prime lenses were released with the camera makes it even more interesting. The last time Sony released awesome lenses with a camera was right after they rebadged the stuff they bought from Minolta.

Since I'm a lens guy, I think it's worth discussing two things. First, is how good are those Sony lenses, really? They're great on this camera but is that because the lenses are so good, because the camera is so good, or because they're manipulating the hell out of the raw data in the camera? The second thing is what kind of performance can you reasonably expect out of an adapter, and how do you go about getting the best possible performance with adapted lenses.

How Good Are the Native Lenses

The native prime lenses have been assessed on-camera and resolution is superb. But I wanted to know why they were superb compared to adapted lenses. Are they really that good? Are they tuned for the Sony camera in a way that adapted lenses can't be? Are the Sony lenses just OK, but the adapters make the other lenses look worse than they are? Does Sony manipulate the raw data to make them appear better than they actually are?

Does any of that make a difference? Actually it does. If all adapted lenses can't perform well, then I have to look at this system based on the native lenses only, which makes it less interesting. If the raw data is being manipulated, then future raw upgrades or some third-party hack may help adapted lenses perform better in the near future. If third-party lenses are soft in the corners because they aren't tuned to the camera's sensor cover, then an optical adapter might improve adapted lenses in the future. If the adapters themselves are the problem, then people modifying adapters with optical absorbents and shims may be on the right track.

In an earlier article I compared A7R Imatest results of the Sony 35mm f/2.8 with the Zeiss 35mm f/2 and Canon 35mm f/2.8 IS lenses mounted on adapters. Since we also have an optical bench I can compare the lenses without any camera involved.

The bench we're currently using has some serious constraints when testing E-mount lenses because of the narrow backfocus distance. We can only test them about 10 degrees off-axis, which is most unfortunate since the biggest difference we saw on the A7R was in the corners. Still, it's worth a look.

Here are on-axis (center) optical bench and Imatest results for all 3 lenses at f/2.8. (Of course several samples of each lens were tested.)

Bench MTF Frequency Sony-Zeiss 35mm f/2.8 FE Zeiss 35mm f/2 ZE@f/2.8 Canon 35mm f/2 IS @ f/2.8
Imatest MTF50 on A7R132013101295

Here are the results 10 degrees off-axis (about 1/3 of the way to the edge).

Bench MTF Frequency Sony-Zeiss 35mm f/2.8 FE Zeiss 35mm f/2 ZE@f/2.8 Canon 35mm f/2 IS @ f/2.8
Imatest MTF50 on A7R118010951190

Don't worry too much about hair-splitting the numbers. Simply remember the optical bench tests are fairly pure tests of just the lens with no camera attached, while Imatest is a test of the lens-camera combination.

Imatest results and optical bench results in the center for all three lenses are about the same. which isn't surprising. Off-axis, though, the Sony on the optical bench is, if anything, a bit weaker than the other two lenses. On the camera as shown by Imatest, though, it's at least as good as the Canon lens and a bit better than the Zeiss.

So why would the Zeiss 35mm f/2 fare worse on the camera, while it is better than the Sony lens when tested on the optical bench? Well, it could be the adapter, but remember I matched best adapters for the camera and lens out of a large box full of name brand adapters. Plus the Zeiss is affected a lot more than the Canon 35mm f/2 IS.

The logical answer is that the location of the exit pupil is fairly far back in the Zeiss lens. People who talk in mathematics I can't really follow tell me that the further back the exit pupil, the more off-axis resolution (and color shift) will be affected by a thicker cover glass on the sensor. It's the same reason so many wide-angle M-mount lenses have problems when adapted to NEX cameras.

Adapter Variation and Sanity

People are going to shoot lenses with adapters on the A7R. I'm going to shoot lenses with adapters on the A7R. But some people are driving themselves insane trying to make things perfect, and that's not going to happen. So I thought I might save a few people a lot of hours by summarizing what we know about using adapters in general and on the A7R in specific.

Adapters aren't perfect, but most are just fine

We all know there is variation among adapters. Even the best and most expensive adapters. You can't add large pieces of metal between the lens and the camera without adding a little variation to tilt, centering, or backfocus distance. Here's some things I know. You don't have to accept them, but I've used hundreds of adapters of almost every brand and gotten more than a few peeks behind the curtain into places where adapters are made.

  • Even big adapter manufacturers buy their parts from some factory somewhere. Most change suppliers for components pretty regularly. Which means their adapters may change pretty regularly. There is no 'best brand of adapters'.
  • There is enough variation that an adapter that is great on this camera with that lens may not be so great on this other camera or with this other lens.
  • An adapter that really messes up laboratory testing results generally has very little to no effect on actual pictures.
  • People often try to measure the thickness of the flat part of the adapter. That doesn't matter as much as the thickness and alignment of the internal mounts (the part that locks into your camera or lens) for things like tilt.
  • The best way I know to check if an adapter is good is take some careful pictures with it, using the lens and camera you plan to use. If the four corners look the same, it's a good adapter.

So, taken to an extreme you might try a few copies of an adapter and find the best one. If you're really, really OCD, you might even match the best adapter for each lens. What doesn't work, though, is trying 16 copies of an adapter hoping the lens just gets sharper, or the corners all get better.

Blacking Out Adapters

There have been some well thought out posts and blogs from Marc Aurel and others who are concerned that light glare within an adapter might be causing increased softness in the outer areas of the image. Marc has even posted templates for cutting out black felt to line the Metabones III adapters that most people are using.

A number of reports have indicated this definitely is helpful for tilt-shift lenses, but it's unclear if it's of benefit for non-tilting lenses. I tested lenses on a Metabones III adapter, then blacked the adapter with some optically black ink and retested, then lined it with black gaffer tape (in case the roughened surface might be of further help).

Normal and taped Metabones III adapters. Not nearly as neat as Marc's but it covers the reflective surfaces. 

I won't bore you with tables of numbers, but none of the 35mm or 50mm lenses we tested showed the slightest improvement in the edges and corners. So this is probably worthwhile for tilt-shifts. It might (although I doubt it) be worthwhile for very wide-angle lenses. It's certainly not effective for standard range lenses.

So What's Left?

Whenever I don't know for certain what an issue is, I go talk to lots of people I know who are smarter than me and know optics better than I do. They were about 100% in agreement that the off-axis softening seen with the A7R and third-party adapted lenses (or, if you'd rather, the lack of softening with the native mount lenses) has to do with the thickness and composition of the sensor's cover glass.

Different manufacturers use different glass and the refraction of this material off-center causes smearing and color shifts to some degree. The effect is worsened with lenses that have an exit pupil near the back of the lens, it's less severe with exit pupils further forward.

Unfortunately, I don't have a list of exit pupil positions in all the lenses, and don't know if anyone does. In general, M mount lenses and wide-angle lenses will have a rear exit pupil, telephoto and tilt-shift lenses tend to be more forward. Hopefully as more people  report their experimentation with alternative lenses someone will start a database for those who are interested.

I'm also told it's not too difficult to make an adapter with an optical element to correct for sensor glass differences. Assuming there's enough demand I expect someone will be releasing something like that in a few months. Sony lenses designed for FE mount certainly already take this into effect, which is why they seem better in the corners and edges. It probably also explains why some NEX lenses, despite vignetting, also do well. Whether Alpha cameras had a similar sensor cover I do not know, but maybe someone out there can tell us.

Conclusion (for now)

There's no question the A7R has some issues. Some of them we can expect to be improved with a firmware update or two. Some are the nature of the camera and won't clear up until the A7X or whatever comes next. Given the limitations the options are to either trash it on forums or to learn how to work around those limitations.

The camera does some things very well at an excellent price. That will be enough to assure some people will learn how to work around those limitations. In another month or two the screaming will die down and some people will be using the camera regularly and making superb images with it. Because it's fully capable of making superb images.

It's never going to work for action photography. It may (or may not) be a great walk around camera. But it will do certain things better than any camera out there at a price that's going to attract a lot of attention.

And lets not forget what may be the biggest change Sony seems to have made: it already has at least two excellent lenses, maybe more. And more are coming. I expect when Sigma and the other third party manufacturers release FE mount full-frame lenses they will have corrected for sensor glass effects and we'll see good performance from those, too.

That doesn't help those who want to shoot native mount lenses on the A7R right now, though. Right now it seems that most people buying the camera are planning to use lenses on adapters. Despite spending a lot of time discussing the problems off-axis, lets remember that most lenses perform every bit as well adapted to the A7R in the corners as they do on their native mount cameras. They perform much better in the center on the A7R, which makes the edges seem problematic. For people shooting landscapes it's something of an issue, but for many other types of photography it's insignificant.

Sony did some things very wrong with this camera. They did some things very right. Mostly they did things very different. I applaud the differences. They may not all work, but at least they're shaking things up a bit. And I do predict when all the dust settles in a couple of years, we'll look back on this camera as one that created some changes.


Roger Cicala


December, 2013


47 Responses to “A Bit of A7R Sanity”

Thomas said:

I have the A7. I resisted it- even hated it at first pick up. That is- until I bought a Minolta MD adapter and slapped some old school lenses on it. Not getting paid? This is the way to go. These older lenses have a ton of quality if your willing to work around their flaws. That means no shooting direct pictures of the sun- sorry.

I have an EP5 as well. Another awesome camera and with a good prime is hard to distinguish from my Zeiss full frame glass. I sold my Sony A900 because of just how impressed I was with the EP5- and I do print from time to time. There still is room in my bag for an A99II but at this point- it's the small guns getting the workout.

AmateurS said:

Thank You for great reviews! I really liked the idea of a sensor attached to the lens without any unnecessary obstacles between. It works! Good bye to backfocus, frontfocus! And you can focus manually with absolute precission - if necessary. No problems with 20 mm SLR lens on quality adapter. Sony A7r is superb in many ways. However - RAW compression artifacts do exist - unfortunately. It looks like they have developed compression algorithm few sensors ago and it doesn't work as expected with this spectacular sensor (to much DR?). Landscape photography using high contrast wide angle lens (with some vigneting) or sky with tonality gradients reveal the problem easily.

Martin said:

Interesting reading. I have a Canon 5D Mark III, an A99v and I bought the A7r for doing my very own find the right thing comparison. 5D3 is bulky. I travel 250k Miles on Airplanes every year to about 10 countries. Overtime to carry the 5d3, 16-35, 24-70 and 70-200L II begins to suck. So my preference, small, lightweight and the optical performance of a full frame sensor. Here we go. A7r sounds like a great thing. FE 55 with a much better performance than the EF 50L, a 24-70, sealed, which promises to be a excellent replacement for my Zeiss 24-70ZA and a 70-200G which finally is sealed, with the new coatings and less than 2k bucks which makes it less exaggerated than the new 70-200G2 which goes over 3k. A99 was way too much. 5D3 is good for my weddings and for my Airshow work. But I need neither for my children nor for my travels a 5D3. I sold the A99 because it was not as fast as 5D3 at sports nor it was much better than A7r in direct comparison. If Sony delivery a few FE lenses 16-35/4, 24-70/4 and 70-200/4 and a 100/2.8 macro, I am am well served for my Studio work, travels and my kids. 5D3 will remain for sports. The A7r is a very good walk around cam. The 55Z is an outstanding AL performer! and FE135/1.8 would be a gem. All looks very promising to me to complement my heavyweight Canon bag and replace it for all day use.

LensRentals Employee

harold said:


Submitted on 2014/01/08 at 5:07 PM
I have the A7. Its awesome. I dumped my 6D and E-M1 for it.
The LCD does not scratch more then anything, and anyway why not use a £3 LCD cover ?
The menus are logical and reasonably easy to use. More then Olympus no ?
The build is terrific. Why do so few manufacturers use a metal top plate ?
My sensor never gets any dust.
Sony’s problem is that the two primes are so good that most people may never need more then the 35mm and the 55mm.
The EVF is excellent.
AF is fast in good light but can hunt in bad light if you can’t find a contrasty enough bit to focus on. What’s new with CSCs ? Even my E-M1 hunts quite a bit in low light.
Roger, comparing it with the best points of the D4/1Dx will bring out loads of negatives, but they seem pretty unreal to me.
Sorry to the Canikon folk lugging around their ancient mirror boxes and massive lenses and justifying it with “the mechanical AF is faster!”. Yes for some folk it is critical, but not the guy obscuring my view in front of the Greenwich Naval College with his massive rucksack, massive tripod and 30 mins setup of a camera to take a picture I reckon was no better then mine, heck no better then the girl next to me with a Canon 550D and kit lens. He certainly didn’t need fast AF LOL!
I remember the 5D as a game changing classic. But it was attacked for years about not having good AF, only because this was the only point where it was quite average actually (centre point only) and opposers felt it was a point of weakness, open to attack.
I don’t criticise anyone’s choice of camera. People love their GM1s, E-M5s, 6Ds, D3, M240s etc. for a whole host of reasons, some real, some imagined. For those not making their living from sports or wildlife and can live with only a few lenses for the next few months the A7 and A7R are pretty much unbeatable if you value its mix of IQ, handling, quality, functionality and portability and don’t have a big investment in another brand’s lenses.

Oskar Ojala said:

There are two very big issues with lenses and adaptability. The biggest one is that there are no wides natively and DSLR wides are large, making the size advantage much smaller, especially since pro DSLR are easier to hold. There is the kit zoom, but it's big, slow and based on samples I've seen, the wide end is too bad to justify going full frame anyway. All this leads to is that smaller (=M mount) wides need to work decently.

The second biggest issue with lenses is that the high quality native primes (35/2.8 and 55/1.8) are quite expensive. Combine that with the fact that this is a new system and initial investment and risk goes up a lot compared to buying e.g. Nikon or Canon DSLRs or even a micro 4/3 camera. I'm hesitant to put north of three grand into something that I may have to sell later with a considerable loss. Buying into the two major brands carries less risk.

Art said:

I'm dubious that a 4k screen is going to show me something my 30-inch IPS panel isn't showing me right now. And I don't need a 4k screen to tell me what a print looks like. If it makes a good print right now, then I don't need to wait for a 4k or 8k or whatever screen to see if my images will 'stand the test of time.' Yes, I would like Sony to release a firmware update to give us a true lossless option (if for no other reason than to let us prove to ourselves if there are any issues with the compression). But in the meanwhile I'm not going to let that stop me from taking advantage of an interesting photographic tool :)

Samuel H said:

Ok, let's look at the numbers you posted...

The brightest stop of light is encoded with 2048 different values in 14-bit linear, Sony is using 256 values. For a single stop of light, that's way more than I need.

The second-brightest stop of light is encoded with 1024 different values in 14-bit linear, Sony is using 356 values. For a single stop of light, that's way more than I need.

The third-brightest stop of light is encoded with 512 different values in 14-bit linear, Sony is using 400 values. For a single stop of light, that's way more than I need.

From the fourth-brightest stop of light, Sony is using just as many values as you get with pure 14-bit linear.

I really fail to see how this could be a problem for anybody. Also, the table you posted is clear indication that any possible issues related to this 10-bit-log conversion could only be found in the three brightest stops of light captured by the camera. Most of that range is only visible when you push real hard on the highlight recovery slider of your RAW image processor.

Max said:

The problem may not always be apparent in the first iteration, eg compression out of the camera. But when you go to push or pull in post if the values are missing (cut out by Sony's algorithm) then you're set up for poor IQ because you can't manipulate the truncated values. Uncompressed truly 'lossless' RAW is something all of the 'lesser' cameras do including Nikon, Canon, Sigma, Leica, Fujifilm etc

Why doesn't Sony offer true 14 bit output as an option?

You're paying for it and memory cards and processing are cheap.

Some may not see the effects today but they will be visible in your 4k screens next year and 8k screens of the future.

It's an easy fix for Sony for this 'landscape' camera. Even if it makes the FPS < 1 it's worth it to get a camera that can be trusted to produce images that will stand the test of time.

Samuel H said:

That sounds like 10 bit log, which is absolutely perfect if you ask me. Encoding images in linear light is awfully inefficient, because all those values in the highlights are useless and because difference encoding doesn't work as well in linear as it does in log.
I seriously doubt 10 bit log is the source of any IQ issues here.

A Ronald Gallant said:

I rented the A7R with the Sony 35mm f/2.8 lens to test it for street photography. It was a bust. Auto focus was inaccurate and high ISO performance was poor. The focus problem was most pronounced when the subject was at about 20 feet or less from the camera; less of a problem for shots at 60 feet or more. The closer distance focusing errors were not minor; they were gross errors. Previously I rented the Olympus OM-D E-M1 with M.Zuiko 17mm f/1.8. In my hands, the Olympus worked far better for street photography than the Sony A7R. But it did not please me enough to buy one. Too many tiny buttons that got pressed accidentally and the limits of a M4/3 sensor were what put me off. For me, so far, what works best for street photography is a D800 with Zeiss 21mm f2.8 lens with a guess at the focus distance and cropping to about 35mm equivalent to fix composition errors.

Gary Morris said:

First, this was a well presented analysis. Thanks.

When I was younger, I used to stress over all manner of little detail. Not anymore. I'm now way too old to worry. And I don't make a living as a photographer, so a little softness in the corners isn't a big deal. In fact, I've shot maybe a few hundred images with my A7R, a Voigtlander M-E adapter and my slightly dented Noctilux .95. In a word the images are superb. Way better than anything my M9 ever captured. The detail is amazing. The focus peaking is a feature sent from heaven for old eyes. And the corners aren't much of a concern because I usually do a little cropping anyway. Add to this the fact that I almost always shoot between .95 and 2.8 which means corners are almost always soft. I can unequivocally recommend the combination of the A7R and the Noctilux (as well as Leica's little jewel lens… the 50mm Summicron). Next up is experimenting with Leica's other Summicrons. The fun has just begun!

Art said:

Max, there is a consensus that Sony's raw compression is not lossless. But as has been explained, that doesn't mean they are equivalent to 8-bit files, and it certainly doesn't mean that image quality is "a disaster." This kind of hyperbole is just what Roger was talking about. Is the camera perfect? No. Is IQ a "disaster?" Certainly not. If we are to discuss the flaws of this camera, then let's do it truthfully and with sanity :)

I have not been able to find any samples to support the notion that there are any obvious issues showing up with the raw files. The so-called "orange peel" texture is something I have seen in D800 files as well when oversharpened, and it looks exactly the same as when it happens in oversharpening A7r files. Other people have reported seeing it if they oversharpen files from other cameras. This leads me to believe it is not an issue exclusive to the A7r (though since A7r files are already sharper out of the camera, we can speculate the effect might appear sooner than with other cameras when sharpening is applied in post.)

As far as the other issues, again, I haven't seen anything yet to convince me the A7r raw files are not as good if not better than any other camera I've used. I have seen no banding or any of the other issues you purport to see. I can make the files fall apart if I push them to crazy extremes, but the same holds true for every other camera I've used. I will hold out the caveat that because the compression is lossy there theoretically could be issues I just haven't come across yet. But saying such issues are widespread in A7r files and creating disastrous image quality is wild exaggeration at best. Sure, I'd still be more comfortable with a true lossless raw file. But in the absence of evidence, I'm not going to try and convince myself into finding a problem that isn't there.

Max said:

I spent 10 days looking at several hundred shots and there were all sorts of tonal issues.

One analysis shows that Sony does the following in their compression


0-800 contains 801 unique values ​​(ie is continuous)
801-1424 contains 320 unique values ​​(skips 1 out of every 2)
1424-1427 contains one unique value (skips 2 of 3)
1428-2023 contains 149 unique values ​​(skips 3 of 4)
2024-2029 contains one unique value (skips 5 of 6)
2030-4093 contains 258 unique values ​​(skips of 8)


Claus said:

the 5DmkII uses ITU T.61 lossless compression (the standard is available in PDF on the net, but it basically takes the difference between neighbouring pixel and makes a Huffmnan-compression. Canon adds its own remapping of the sensor to this, but it is easily disentagled. I wrote my own version of an unpacker some years ago). Apparently Sony uses a logaritmic compression (similar to a or u-law in the fixed telephone voice digitizing) This is not "rounding to 8 bit" as Max implies, but it do introduce some loss of information. If this turns out to be a major problem, it can be repaired in Firmware (unless Sony has contractual reasons e.g. with Nikon not to that)

GeorgK said:

NEFs from a D7000 are 20MB average, DNGs from a K5 about 19MB. Both are 16MP cameras. There is a lot more redundancy to be expected in 36MP pictures (esp. when used with "average" lenses), so 37MB for a compressed RAW seems quite reasonable.
The "lost bits" argument is crap, sorry. There is a little bit more in compression mathematics then "/" and "*".

Samuel H said:

I've just shot some pics with a 5D2. 21mp*14/8=37mb RAW files, but these CR2s are around 24 MB each. Where did the other 13MB go?

Again, not saying those issues are not there, just that it's very unlikely that these are 8-bit files (though it would be nice if I could see some samples, I googled a7r banding and found notihing).

Leave a Reply