Thanks for putting this all together with the detailed crop of each of the same photos as I recommended. I agree that there is a subtle difference even I see also some of the artefacts in the D810 file if I look closely. Not sure how to interpret this - post processing difference by each or coming from the compression itself? Very hard to say IMO.
retrofocus wrote:
Thanks for putting this all together with the detailed crop of each of the same photos as I recommended. I agree that there is a subtle difference even I see also some of the artefacts in the D810 file if I look closely. Not sure how to interpret this - post processing difference by each or coming from the compression itself? Very hard to say IMO.
You're welcome. Here is another closeup.
The artifact you see on the D810 is an actual scuff on the bottle. You see it on all the shots. Reason I know this is because I've seen these artifacts enough where I can recognize it.
I actually don't mind this. It pushes the development cost of peripheral features to the people who actually use the features, rather than every single person who buys the camera.
The problem is its not a very functional intervalometer. When shooting longer exposures at night, your exposure times jump in 10 second intervals once you get over 20 seconds, so if your ideal exposure time is 25 seconds you have to pick a shorter time and bump ISO which introduces noise. That's a pain when you're dealing with thousands of images. The built in intervalometer on my NX1 and D750 are far less restrictive.
So what you're paying for is a half-baked consumer oriented app that lacks the functionality of a real intervalometer. The better option is just to pony up the extra cash for something like the Pixel TW282.
GMPhotography wrote:
I live in one of the sunniest parts and hottest parts of the country maybe the world for that matter. So high contrast scenes are the normal mode of operation. I'm in Arizona the Sonora Desert and it's very bright almost always but still I seem to get around without to much recovery modes in C1.
Shadows and highlights are all relative. Lets say I'm shooting a band playing in a dimly lit bar during the day. There's a small window which I don't want to be completely blown out white. So I expose for that window, and the rest of the image looks black. In my favourite RAW editor I bring out the image of the band using the shadow recovery tool. In this case the entire image is the "shadow".
So when people talk about pushing shadows, they're not literally talking about areas where the sun is blocked by something, but merely dark parts of the image, that you would otherwise simply exposure for, except you wanted to protect some highlight somewhere. Sony's amazing dynamic range makes this sort of thing possible, only it's limited, to an extent by the compression scheme.
Sony's lossy compression scheme is designed so that if you don't push the image around significantly in post you just wont see any artifacts, at all. They will be hidden in the shadows. That is why I believe they are slow to do anything about it.
If they had designed it slightly better it wouldn't be a problem and you could still have almost the same compression ratio. Even a better converter is possible - I'm working on a way to automatically fix these artifacts (or at least massively reduce their impact) directly in the RAW file. Early results are promising.
johntruong wrote:
You're welcome. Here is another closeup.
The artifact you see on the D810 is an actual scuff on the bottle. You see it on all the shots. Reason I know this is because I've seen these artifacts enough where I can recognize it.
Thanks again - here now I clearly see the difference. Assuming that the post processing was identical (which you confirmed above), yes, this could be an issue with the compression. First time I ever seen it, thanks for sharing!
johntruong wrote:
Like I said earlier, the issue may be moot for a lot of users (which is great), but unfortunately I am not one of them.
I see this more often than I care for. Most of the time I am clone stamping it out in Photoshop.
Would you be willing to send me your Sony RAW files? As I mentioned earlier I'm working on some software (in my very limited spare time) to automatically remove these artifacts. Here are some early results, shadows have been boosted massively to make the artifacts easier to see, I'm not claiming this is a realistic example.
retrofocus wrote:
Thanks again - here now I clearly see the difference. Assuming that the post processing was identical (which you confirmed above), yes, this could be an issue with the compression. First time I ever seen it, thanks for sharing!
It's definitely the compression. What you get is furry or splodgy looking areas up to 32 pixels horizontally away from any strong highlight. The example above was shot portrait so the artifacts run vertical instead of horizontal.
johntruong wrote:
You're welcome. Here is another closeup.
The artifact you see on the D810 is an actual scuff on the bottle. You see it on all the shots. Reason I know this is because I've seen these artifacts enough where I can recognize it.
GMPhotography wrote:
We should start a thread about color management and monitors and such so folks understand that whole process. When we talk about banding look for what color space these images are in and work with real photographic wide gamut monitors. This is a big topic here that would help a lot of folks understand color management
I would enjoy reading more about color management, here or at another forum. <cough> That said, my challenges have mostly been in making colors pleasing from the camera, not in getting accurate prints from my processed files, although it helps that I use a local printer with whom I can chat about what I want from my prints.
---------------------------------------------
retrofocus wrote:
I looked at your link above and I have a hard time seeing any difference other than the 5D MkIII which appears a lot more noisy. Could you maybe make a red circle around the areas which you refer to? I might miss something.
I see compression artefacts in the Jackson-Triggs bottle, on the left. Look between the two vertical highlight lines. I see both added color and a reduced contrast gradient in the Sony files. It helps to magnify the images.
Now, that it requires magnification beyond 1:1 to see it clearly means that it isn't a big deal as it. But, if those particular values fall into the high contrast ranges when I'm adjusting an image's curves, it can be visible in what would be an arm's length print. Just like optical aberrations, once you see them it is hard to not see them. Since high contrast edges are particularly important to my style of photography, and indeed are one of the key things that the eye/mind sees when first looking at an image, faults here are even more critical than losses in resolution or rendering of bokeh. I consider it a show-stopper not just because of its rendering issues but because it contributes to my overall sense that Sony is still a consumer electronics company rather than a photographic company.
Note that my primary camera is a Sony, so I don't think it makes the cameras unusable, just puts them beyond consideration for a next purchase unless at a substantial discount on the used market.
Over at Naturescapes, a fellow just got his 5ds and states that his raw file is a whooping 52 mpix (don't know the complexity of the photos). If true Canon seems to be compressing the raw just like Sony 11/7; my A900 which is a 24 mpix camera produces 35 to 40 mpix uncompressed raw.
mogul wrote:
Over at Naturescapes, a fellow just got his 5ds and states that his raw file is a whooping 52 mpix (don't know the complexity of the photos). If true Canon seems to be compressing the raw just like Sony 11/7; my A900 which is a 24 mpix camera produces 35 to 40 mpix uncompressed raw.
I assume you mean megabytes, mpix is not a measure of file size. Canon compress all RAWs using the lossless JPEG algorithm.
It's not like Sony's 11/7 compression at all because it is lossless.
mogul wrote:
Thanks for the explanation Matt...if the Canon way of making much smaller files is better, why doesn't Sony use it?
Why indeed
mogul wrote:
There must be some trade offs.
Lossless compression results in variable filesizes and in the worst case can make files larger (though I don't think this will ever happen in practice with RAW files). Basically the amount of compression achieved depends on the content of the image.
Sony's lossy scheme always gives the same size files (actually they vary very slightly, due to the lossy compressed embedded thumbnail!) which is advantageous when designing buffers etc, you can more consistent performance.
Canon's lossless compression is more computationally expensive than Sony's lossy scheme which only requires a few instructions and a very small buffer, however with Canon most of the heavy lifting can be done in hardware.
Matt Grum wrote:
I assume you mean megabytes, mpix is not a measure of file size. Canon compress all RAWs using the lossless JPEG algorithm (encode as JPEG, take the difference between the compressed and original and store that using entropy encoding).
It's not like Sony's 11/7 compression at all because it is lossless.
I have decoded Canon CR2 file at one time, the raw-part of the file was encoded according to ITU T.81 Loss Less Compression and does not involve JPEG. It compresses the raw-image a factor of 2 (approx).
Tariq Gibran wrote:
It's actually a bit disappointing in one major respect. The EVF resolution is the same as the previous A7r at 2,359,296 pixels.. The optics are improved to give the higher magnification. It's a bit unfortunate that Sony did not also improve the resolution to something like what Leica offers in the new Q, 3,680,000. When zooming in for magnified view, a higher resolution EVF would make a noticeable difference. I guess that will be next years upgrade for Sony as they have been using this resolution EVF for many years now and it's starting to show it's age (the higher finder magnification might even make the resolution of the EVF appear even worse/ more noticeable since you still have the same number of pixels as before)
Ammm 2,359,296 pixels is enough... Id rather have a bigger viewfinder with less lag than more pixels, which is exactly what they did. The Fuji x-t1 has a 2.36million dot OLED display and its the best i've ever used up to this point. Seems the Sony one will be even better due to slightly higher magnification.
juvx wrote:
Ammm 2,359,296 pixels is enough... Id rather have a bigger viewfinder with less lag than more pixels, which is exactly what they did. The Fuji x-t1 has a 2.36million dot OLED display and its the best i've ever used up to this point. Seems the Sony one will be even better due to slightly higher magnification.
The EVF on the new Leica Q is the cat's meow...wonder who makes it since Sony has pretty much cornered the market (after buying Epson's plant).
