p.7 #1 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
For some cameras we can see in the DxO graphs that they are marked as "smoothed", like the Sony RX 100 II and III. It is obvious by just looking at the jump in the graphs. For the A7s, as snapsy says, we can see the same kind of jump, but they did not mark it as smoothed. There is little transparency regarding any correction they do in the data before presentation.
There are other types of in camera "Noise reduction" before writing to the raw file that are not harmful to the validity of the comparisons. Like long exposure noise reduction by frame subtraction. That kind of noise reduction is performed on a single pixel basis and does not affect detail reproduction.
p.7 #2 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
snapsy and Anders, thank you both for your patience, insight and explanation.
The information you have provided has enabled me to get an insight into sensor testing and how some measurements are made, answered some Q's and most importantly highlighted some of the caveats.
p.7 #3 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
+1 @ caveats ... the (often) untold story ... and rarely by pundits, marketing or opposing fan-boys.
I suspect the old "80/20" rule is in play (or more like 90/10 or 95/5) where the propagation is targeted toward the larger group or more common application / utility / interest. Then when someone wants to delve into the 20/10/5/1 percentile issues, it becomes evermore challenging to secure accurately pertinent information ... often inundated by reiteration of the already known 80.
Those two images, beach houses and that roundy thing with the deep blue sky and the little guy outside ....especially the beach houses. That looks like a beautifully warm rendering with natural pleasing colors. Almost film-like. I'm beginning to believe that I will LOVE the output from this camera. Now, beach-shots of colorful little cottages in sunset are quite forgiving of course, very hard to mess up colorwise. But there's certainly something about the rendering and overall color balance...
p.7 #5 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
Another two panorama samples that's combined from Canikon 5Ds and D800 official JPG samples at ISO100. Note that they are not the same subject so our eyes can only judge on the smoothness transition between highlights to shadows. Which one, do you think that has more digital look (less DR) compared to other ?
p.7 #6 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
mttran wrote:
Note that they are not the same subject so our eyes can only judge on the smoothness transition between highlights to shadows.
Unfortunately we cannot judge anything about that with these images either. I see only a huge difference in the subject and the light, in addition to the jpg processing, which alone makes it impossible.
p.7 #7 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
Regarding the diff @ jpg vs. RAW ... is there a similar (i.e. an internal processing diff exists) diff between RAW vs. TIFF @ pre-cooked RAW files from the mfr's?
I guess the question would really be is TIFF vs. TIFF an even more equitable (reduced variables) comparison than RAW vs. RAW ... in the same spirit that RAW vs. RAW is more equitable (reduced variables) than JPG vs. JPG?
Having not used TIFF yet, I'm not versed on the diff @ TIFF vs. RAW (which iiuc, is not quite the same as raw). So, it would seem that if mfr's are "pre-cooking" the files differently, is it possible to get each @ "uncooked" files. I'm just wondering if part of the diff we see in DR is because we are comparing different levels of "pre-cooking" that generates more DR via pre-cooking algorithms, among mfrs, i.e. suggesting that the starting points @ "uncooked" might be much more similar than when we see them @ "pre-cooked".
p.7 #8 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
RustyBug wrote:
Regarding the diff @ jpg vs. RAW ... is there a similar (i.e. an internal processing diff exists) diff between RAW vs. TIFF @ pre-cooked RAW files from the mfr's?
I guess the question would really be is TIFF vs. TIFF an even more equitable (reduced variables) comparison than RAW vs. RAW ... in the same spirit that RAW vs. RAW is more equitable (reduced variables) than JPG vs. JPG?
Having not used TIFF yet, I'm not versed on the diff @ TIFF vs. RAW (which iiuc, is not quite the same as raw). So, it would seem that if mfr's are "pre-cooking" the files differently, is it possible to get each @ "uncooked" files. I'm just wondering if part of the diff we see in DR is because we are comparing different levels of "pre-cooking" that generates more DR via pre-cooking algorithms, among mfrs, i.e. suggesting that the starting points @ "uncooked" might be much more similar than when we see them @ "pre-cooked"....Show more →
TIFF is a file format that can hold both raw data and demosaiced images. Don't confuse the file format with what data it holds. TIFF files as we know them contaion demosaiced images, like jpg imaegs minus the lossy compression. It does not make sense to think of TIFF files as more generic than raw files for noise measurement purposes.
p.7 #12 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
gdanmitchell wrote:
That was my take on the Chuck Westfall interview, where he speaks about the DR (and noise) issue in some very circumspect ways. Some wanted to read that as an "admission of guilt" and that the new camera wasn't all that great in this regard.
I read it as setting the expectations to a low level until production camera specs are finalized.
Accurately speaking, there is no such thing as "no noise" in a digital camera. There is such a thing as very low noise, and there is such a thing as even lower noise.
"Practically speaking," both cameras might well produce noise so low that it is not an issue in many cases at all. ("Practically speaking"... "no noise" from both.) I think the point of the post here is that it is a report suggesting that the 5DS might combine very low noise and larger dynamic range than expected for photographers whose desire is to achieve this a relatively low ISO levels.
p.7 #13 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
alundeb wrote:
From what I've heard, the read noise character of the 7DII is much less objectionable than from the 5DIII.
According to CPN europe, the 5Ds uses the same 16 channel readout architecture as the 1DX.
p.7 #15 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
Just thinking if using magic lantern dual ISO feature then the 50MP become 25MP, still more than enough. Correct me if I'm wrong. If so high resolution is not only benefit on resolution.
p.7 #16 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
RAW files come from the camera itself, are (or )used to be the unprocessed photo site data, today usually incorporate some camera data processing, can only be displayed by running them through a software converter, and are not intended to look best — for example, they _must_ be sharpened, contrast is low, the colors are flat, etc.
A TIFF file usually is created even later in the process — at a minimum it is the result of a conversion from a raw file, and more often it is the end result of a post-processing workflow, so it can include all of the usual post-processing variables. It would be extremely unusual (and very inefficient) to display tif files on the web, as they are unnecessarily huge.
To some extent, I contend that the "cooked" (to use your term) files is provide the most interesting and useful results. I really don't care if the raw file from camera X is "better" or "worse" by some standard than that of camera Y. I care about what I can do with the file and the eventual final photographic output that I'm striving for. While the desire to level the playing field and apply the exact same standard to each type of file is supposed to create a more objective comparison, it also leads to comparing things that have less real-world applicability, or from which that real-world meaning must be guessed.
I believe that the most interesting and useful comparison is between prints produced after the files had been taken through idealized post-processing workflow by expert photographers, in each case optimized to produce the best possible output from the files to be compared...
More than once I have learned extremely important things about what difference matter, what differences don't matter, and what can be done with files by such a test.
(Added: I've described elsewhere the surprise among of group of very good photographers who were presented with prints by one among them (a former LF film photographer and current 80 MP Phase One shooter) that came from a 36MP FF camera and the 80MP Phase One back on a Mamiya 645 system and were scaled for 30" x 40" printing... and were unable to consistently pick which came from which camera. No matter what logic tells us about the truly excellent IQ of the Phase One files, our eyes told us a lot about the meaning (or not) of these differences in real world photography... and forced us to deal with the distance between that and what confirmation bias would have led us to expect.)
Dan
RustyBug wrote:
Regarding the diff @ jpg vs. RAW ... is there a similar (i.e. an internal processing diff exists) diff between RAW vs. TIFF @ pre-cooked RAW files from the mfr's?
I guess the question would really be is TIFF vs. TIFF an even more equitable (reduced variables) comparison than RAW vs. RAW ... in the same spirit that RAW vs. RAW is more equitable (reduced variables) than JPG vs. JPG?
Having not used TIFF yet, I'm not versed on the diff @ TIFF vs. RAW (which iiuc, is not quite the same as raw). So, it would seem that if mfr's are "pre-cooking" the files differently, is it possible to get each @ "uncooked" files. I'm just wondering if part of the diff we see in DR is because we are comparing different levels of "pre-cooking" that generates more DR via pre-cooking algorithms, among mfrs, i.e. suggesting that the starting points @ "uncooked" might be much more similar than when we see them @ "pre-cooked"....Show more →
p.7 #17 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
gdanmitchell wrote:
I believe that the most interesting and useful comparison would be between prints produced after the files had been taken through idealized post-processing workflow by expert photographers, in each case optimized to produce the best possible output from the files to be compared...
More than once I have learned extremely important things about what difference matter, what differences don't matter, and what can be done with files by such a test.
Dan
Comparisons of prints tend to lead to conclusions like "I can't see any difference between a Hasselblad Medium Format camera and a Canon PowerShot with a 8x6 mm sensor and 14 MP."
Prints are no good indication of relative performance between cameras, because they are very forgiving in terms of noise. Some small noise may in fact help the prints look sharp and detailed.
We can learn that excellent prints can be made with cameras that technically perform different, wich is important to know though.
p.7 #18 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
gdanmitchell wrote:
I care about what I can do with the file
+1
This was a noticeable difference I saw when I switched from Nikon to Canon several years back ... that the Canon files were a bit more "malleable" than the Nikon's had been despite (or likely as a result of) coming out of the oven a little less baked. Initially, I was disheartened by my switch to Canon since they didn't look like the Nikon's did ... until Chuck Westfall explained to me their intent for professional files was different from that of consumer files, by design. Once I got my head around that a bit more (and learned more about processing), I came to appreciate it differently.
Accordingly, making "out of the oven" comparisons can be a bit deceptive, even @ RAW, I suppose ... since neither is truly raw. Canon just seems to be a bit "more raw" than others ... and I think that's the way they WANT it to be.
p.7 #19 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
RustyBug wrote:
+1
This was a noticeable difference I saw when I switched from Nikon to Canon several years back ... that the Canon files were a bit more "malleable" than the Nikon's had been despite (or likely as a result of) coming out of the oven a little less baked. Initially, I was disheartened by my switch to Canon since they didn't look like the Nikon's did ... until Chuck Westfall explained to me their intent for professional files was different from that of consumer files, by design. Once I got my head around that a bit more (and learned more about processing), I came to appreciate it differently.
Accordingly, making "out of the oven" comparisons can be a bit deceptive, even @ RAW, I suppose ... since neither is truly raw. Canon just seems to be a bit "more raw" than others ... and I think that's the way they WANT it to be.
At CES I learned that vinyl records are making a huge comeback - one company said that production is booked solid for the entire year. Another had a product that streams a record player over bluetooth. I asked one rep what was behind the resurgence and he said people like to hear the pop and crackles of the player, that it makes the music more "real" to them. After hearing that nothing will surprise me with respect to individual preference for raw file malleability.
p.7 #20 · 5DSR Low ISO DR 1.5-2 stops better than 1DX ...
RustyBug wrote:
+1
This was a noticeable difference I saw when I switched from Nikon to Canon several years back ... that the Canon files were a bit more "malleable" than the Nikon's had been despite (or likely as a result of) coming out of the oven a little less baked. Initially, I was disheartened by my switch to Canon since they didn't look like the Nikon's did ... until Chuck Westfall explained to me their intent for professional files was different from that of consumer files, by design. Once I got my head around that a bit more (and learned more about processing), I came to appreciate it differently.
Accordingly, making "out of the oven" comparisons can be a bit deceptive, even @ RAW, I suppose ... since neither is truly raw. Canon just seems to be a bit "more raw" than others ... and I think that's the way they WANT it to be.
...Show more →
I'm not sure about Canon files being "more raw" than other files. But even if they were, that doesn't mean necessarily mean they are more malleable. Case in point would be what happens when you push the shadows on Canon files. That is certainly a circumstance where you can do a lot less with the Canon file.
That aside, I don't believe that the way raw files are "cooked" generally makes one better or more malleable than the other...it just means that you can't apply the same processing parameters to each to get the same looks. But you can apply different processing parameters in your raw conversions and likely get a similar look. The real difference in my view between cameras comes when you can't apply any kind of processing to get a similar result because of sensor limitations.
