JPEGs on the A7, from a JPEG shooter's perspective

Started 6 months ago | User reviews thread
ForumParentFirstPreviousNextNext unread
Flat view
quezra
Senior MemberPosts: 2,062Gear list
Like?
JPEGs on the A7, from a JPEG shooter's perspective
6 months ago

If you've come across DPReview's A7 review, you'll probably have been told that:

The areas that need the most improvement are related to JPEG image quality.

They 'hope' that a 'firmware upgrade' will address some of these issues, which we will come back to later in the review. So this review is a response to their assessment of JPEGs on the A7. It should be pointed out just how much they're implicitly complimenting this camera, if the area they think is most in need of improvement is the JPEGs (which I will show you are marginal issues even by DPR's own admission). It means the rest of the camera is pretty darn good if these minor issues are the biggest fault they can find in the camera!

How can you tell they're minor? Well let's look in detail at the faults they find with the JPEG engine:

  • JPEG quality disappointing compared to peers – crude sharpening, over-aggressive processing and occasional posterization

Ok, these sound like very serious problems! And for those of us who've been using JPEGs primarily on our A7, shocking indeed, given the extraordinary claims Sony had made about their new JPEG engine. How did we all miss such terrible issues? Well let's look at each of these areas in turn.

Crude Sharpening

A perennial complaint of JPEGs of yesteryear, the return of sharpening artifacts should indeed be a serious issue for the JPEG engine. Looking at the review, they show an example at ISO 16,000(!) and go on to say:

On the plus side, the JPEG has done a good job of preserving detail in the face, while smoothing the beige wall to the left. Our processed Raw file uses the same noise reduction over the entire image, meaning a balance has to be struck between smoothing the background and retaining detail. As such, our chosen settings, while appearing to retain more detail, have neither the well-defined edges or smooth background of the camera's JPEG.

Amazingly, it seems to be they're acknowledging it did a better job than them. And don't forget, we're talking about an ISO16,000 picture. If the JPEG engine is doing a better job in defining edges on an ISO 16,000 image than you, most people would put that down in the "Wow, amazing!" category. But were they merely using a high ISO to demonstrate the issue reaches all the way down?

Where the context-sensitive approach falls down is at transitions between areas it concludes to be smooth and those it considers detailed - leaving a pronounced 'halo' of noise at the edge of smooth regions. This, combined with the camera's tendency to sharpen edges, can leave a rather unpleasant result if you zoom in. Our test scene suggests the problem becomes visible around ISO 12,800.

Ok, so... it starts at ISO 12,800. And we're helpfully informed it only occurs in very specific areas. So how would it look in print results?

We printed the central 19 x 13" of a similar image to the one shown above (such that the whole image would have been 30 x 20") and, although the effect was clearly visible, and the radical difference in noise levels across the image is slightly unusual, the results were pretty usable.

Verdict:

DPReview have clearly tried to make a mountain out of a molehill. When we think it's barely been a couple of years since we were impressed with usable ISO 1,600 images, the fact that they can nitpick with a straight face about edge detail at ISO 12,800 viewed at 100% size shows how far JPEG engines have come. It can also now be exclusively revealed that Sony's JPEGs are actually very usable up to ISO 6,400 (one stop less than the limit where they identified sharpening artifacts appearing, unsurprisingly).

If you are coming from a smaller sensor camera and this is your first full frame, the 7 full stops of usable ISO range from JPEGs no less will feel positively luxurious. If you've been shooting FF cameras for a while, you probably aren't using JPEGs anyway, but in case you are, bear in mind you "only" avoid sharpening artifacts below ISO 12,800.

But wait, there's one last thing. You can actually edit your JPEG settings very simply in-camera. From Fn > Creative Style (or Menu > Camera > 4 > Creative Style if you have customized your Fn button), you can adjust sharpness by +/-3 (that's 7 different levels) to get it 'just right'. Sony have imported the full plethora of JPEG customization options from their NEXes, and you can store two different settings of each option (yes, you can have two different "Standard" settings!).

Sony: 1 - DPR: 0

Over-Aggressive Processing

Another serious problem that has afflicted JPEGs in the past. DPR asserts:

Overall, the noise suppression is too aggressive in the JPEGs, making particularly the a7 images seem artificial at times. Naturally this can be turned down, but the default settings look way too much like a brush stroke filter has been applied. This gets worse as ISO rises.

Well naturally this can indeed be turned not just down but also off! Amazing! You'll find three different levels (Normal, Low, Off) under Menu > Camera > 5 > High ISO NR. I recommend Low (it is still very good, while giving a much closer rendition to 'film-like' grain at higher ISOs than previous NEX camera JPEGs), or if you're planning to PP a lot, Off. Ok, so problem solved.

Verdict:

Yes, the default is quite strong, but you can also think of it as the strongest setting (it is). Some people really like all their noise eliminated and have no plans to pixel peep. Personally, I found Sony's old JPEG 'default' to be intolerable and their 'low' to still be too strong (there was no option to turn it off on my NEX-5N). But I also think back to the days when I'd first bought the NEX-5N and was not in any way a pixel peeper - it was in fact the default settings that blew me away and made me fall in love with the camera in an instant. It was only years later when I started to pixel peep all the time that I started to become dissatisfied with default NR, and then later even their low NR in the old engine.

You see, Sony assume their JPEG users are most likely that sort of shooter - not particularly a pixel peeper and expecting the camera to get the image 'right' for common 4x6 prints, and while that assumption might not fit into DPR demographics, it is probably correct for the real world. The mistake DPR made was in assuming the 'default setting' user is also as much of a 100% view pixel peeper as they are. Today I am indeed a (slightly paradoxical) pixel-peeping JPEG user, yet I am fully satisfied with the results I get from low NR, even though it can be a very subjective thing.

We'll be nice to DPR and call this one a draw.

Sony: 2 - DPR: 1

Occasional Posterization

Now this one was bizarrely the one they spent the most time illustrating, given that it's also the least likely you'll be able to notice. Let's see what they say:

An odd artifact that we noticed several times in the JPEGs is the tendency toward posterization in subtle gradations, again, often in out-of-focus areas. It's certainly not a problem that appears in every image but it's something we noticed often enough to be concerned about it.

Ok, fair enough and well-done for spotting something everyone else missed! Ok let's see the evidence:

As you can see, the problem isn't image-destroyingly bad - indeed, whether you can even see what we're talking about will depend to a large degree on the device you're viewing the site with. But, once you've looked at the image's red channel, it should be much more obvious what we're talking about. Look closely again at what should be a smooth background of the portrait on the Experience page, and you'll see it in a real-world setting.

Wait what? We can barely see it and probably only if you look in the red channel? Do you really think a standard default-settings-only JPEG user has software that can isolate his red channel for him?

Any users hoping to post-process their JPEGs will be particularly disappointed, since these hard steps between tones become exaggerated as soon as you start adjusting the image brightness. Obviously we'd recommend shooting Raw if you plan to do any significant post-processing (with any camera), but potential a7 owners need to be aware that its JPEGs offer even less processing latitude than usual.

See here's the thing DPR. "Significant post processing" and "JPEG" and "default settings only" could certainly be realistic choices of different types of users, but put them all together for a moment and think. Does there really exist a JPEG user who also is heavily into post-processing (a very technical art indeed) yet unaware of his in-camera settings? I can imagine a RAW user heavily into PP not being aware of JPEG settings since he never uses them. I can imagine a default-settings JPEG user whose extent of PP work is "Auto Correct" in Picasa. I can even imagine a JPEG-only pixel peeper who does moderate post-processing but how could he have got to 2014 without knowing there are in-camera settings outside of "default" or the standard limitations of JPEG? In your efforts not to make your 'findings' a total irrelevance, you had to concoct a chimera of a photographer that picks all the suboptimal choices in every aspect of digital manipulation. Such a mythical person needs to read up on post-processing, not blame the camera.

So what's a simple workaround if you do spot these things cropping up in your camera?

Turning the cameras' 'High ISO NR' to 'Off' improves matters (at all ISOs), because the standard processing appears to be supressing noise to such a degree that you lose the 'dithering' effect that noise usually brings to tonal transitions. The downside is that you have to put up pretty noisy high ISO images (noise reduction, in both JPEG and Raw is still being applied in the 'Off' setting).

Oh jeez, the same thing that reduces heavy post-processing also removes posterization. Well then.

Verdict:

Posterization, if you can ever see it, usually requires carefully set up conditions to see, usually with the help of software tools. If you are deluding yourself that you can do heavy post-processing from JPEGs, you will be disappointed, just like you'd be disappointed trying to do heavy post-processing of JPEGs from any other digital camera ever made. This part of the review looks to me like DPR spent a lot of time (=money) investigating the issue, actually discovered it was practically irrelevant, but rather than report to their editors that they'd wasted their time, concocted a fantasy photographer for whom this might be an issue.

Sony: 3 - DPR: 1

Conclusion:

After I read DPR's review, I went through all my 1,000+ photos taken with my A7 to look for hints of these issues. I only could find some posterization issues in images I had post-processed (duh!). Not one from a straight-out-of-camera image. Not one instance of heavy noise reduction (I had set mine to low from day 1). Not one instance of sharpening artifacts (possibly coinciding with the fact that I barely had any ISO 6,400 images and no ISO 12,800 images!).

Now remember how DPR ended their review calling for a firmware update to 'fix' these issues? This is where it gets confounding. DPR writers got slammed in the official review for failing to notice that there were simple switches to toggle if you really didn't want to risk these issues. They defended their decision, insisting that these as 'default' settings should be judged because that's what people would pick the camera up using. But if DPR were actually aware of options available, why would they be calling for a firmware update to fix the issues when the solutions already existed in-camera? This is the baffling part that suggests they really didn't explore the options, made more puzzling by the length they went to illustrate these "issues".

Usually DPR reviews are seen as valuable and help to illustrate a lot of the different dimensions of a new camera. But they have failed here because of their over-emphasis on non-issues, and give a very skewed picture (did they post-process?!) of the A7, and missed an excellent chance to tell you other stories about the image quality. If you are a JPEG user interested in the capabilities of a FF sensor, you can safely discount everything DPR said about JPEGs in their review. Before this review, I'd been telling friends this was the best JPEG engine I've ever used, and despite what DPR 'experts' have said, those of us who've spent time with the JPEGs have stood by what we said (see some expert views here and here). Just take the 10 minutes or so to tweak your settings at the start, and I hope you will enjoy many years of beautiful images from a splendid camera.

 quezra's gear list:quezra's gear list
Sony Alpha 7 Sony FE 55mm F1.8 Sony FE 28-70mm F3.5-5.6 OSS Sony Alpha NEX-5N Voigtlander 35mm F1.4 Nokton +7 more
quezra's score
5.0
Average community score
4.8
bad for good for
Kids / pets
good
Action / sports
acceptable
Landscapes / scenery
excellent
Portraits
excellent
Low light (without flash)
great
Flash photography (social)
great
Studio / still life
excellent
= community average
Sony Alpha 7 Sony Alpha NEX-5N
If you believe there are incorrect tags, please send us this post using our feedback form.
ForumParentFirstPreviousNextNext unread
Flat view
Post (hide subjects)Posted by
ForumParentFirstPreviousNextNext unread
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow