DR from 5Ds will be 2 stop better then 7D mk II

[jrista]

The noise floor is generally the computed RMS of noise from all pixels from dark frames (usually, you would want to compute it from many dark frames, and generate an average RMS). That would include banding, but bands tend to be outliers, so they don't impact the RMS all that much. The difference between the 5D III DR (10.97 stops) and the possible 5Ds DR (maybe around 11.23 stops) could very well be the difference in banding. That isn't much of a difference, though, assuming the 5Ds does arrive with read noise and FWC similar to the 7D II.

Thanks for explaining! So this would at least account for some of the confusion over "more" or "less" dr, "usable" or "banding".

The key thing here is that noise adds to the signal. So, if you have a read noise floor of, say, 20e-, but your bands are 80e-. The bands are infrequent, so they don't actually change the RMS much...but they add 80e- to the signal. That band is going to shift the ENTIRE signal in that row of pixels up by 80e-. Such a band would be visible with very light shadow pushing, well before you actually got to the read noise floor.


Once that band is gone, you could push the shadows a fair bit more before the actual 20e- noise floor became a problem.


There is also the personal tolerance thing. Canon read noise is ugly, plain and simple. It's got bright red blotches, splotches, short little horizontal juts, green speckle. It's the nastiest stuff. I think most people stop pushing before they get very far, because once they see that, they don't want to go any farther.


That is in contrast to even A7r read noise, which is stored in a lossy-compressed RAW file. A7r read noise is primarily luminance, and primarily gaussian. There is hardly any color noise to start with, and what color noise there is still follows a mostly gaussian distribution. There are very few hot pixels. I have no problem lifting an A7r four stops. The deeper shadows are noisy, but they clean up really nicely, and with just a bit of luminance NR slider in LR.


I think personal tolerance and noise characteristic plays a really big role. There is no question Canon read noise has improved. Compare the 5D II to the 5D III to the 6D, and you will be amazed at the 6D, for sure. But compare the 6D to the A7r or D810...and you'll understand the true value of ultra low read noise. Theoretically Canon could gain DR by increasing FWC as well...but their read noise is just ugly. Even with more FWC, their poor read noise characteristic is still going to give people pause as they start to dig deeper.


Thus, I think it's going to be really difficult to get any kind of objective commentary about the 5Ds data quality until someone does some objective, mathematical testing on the raw files. It's really difficult for people to decouple their perceptions when visually evaluating, and past tolerance limits could be playing a big role in what people are saying about how much more dynamic range the 5Ds has. It may simply be that lower banding and maybe more gaussian read noise is making people feel they can lift more...but in an objective comparison, things may not be much different.

 

[Troja]

The DR of the 5DR/S IS THE SAME as the 5D MkIII, there is no discussion on that, the only point of interest is the Canon comment that there is improved shadow and highlight latitude. Some say those two comments contradict each other, I think the first is worded ambiguously enough that both might well be true.

I'm still puzzled about the definition of dynamic range (I'm sure we can agree it's about what we find in the raw image data, and not somewhere up the pipeline or at the sensor before it's usable).

Is dr the noise floor (however that is defined?) *excluding* banding, or must it include banding because it's very hard if impossible to remove in post and thus cannot be counted into the "dynamic range"?

Depending on this definition, the 7d2-based 5ds probably will have more dr than 5d3 (as in "less banding") or it won't (unlike the ff 6d/1dx which really improve upon the 5d3). Or am I mistaken? What do the dr experts say?

The noise floor is generally the computed RMS of noise from all pixels from dark frames (usually, you would want to compute it from many dark frames, and generate an average RMS). That would include banding, but bands tend to be outliers, so they don't impact the RMS all that much. The difference between the 5D III DR (10.97 stops) and the possible 5Ds DR (maybe around 11.23 stops) could very well be the difference in banding. That isn't much of a difference, though, assuming the 5Ds does arrive with read noise and FWC similar to the 7D II.


My problem with these statements (paraphrased) is they conflict:


A. The 5Ds has the same dynamic range as the 5D III
B. The 5Ds has more editing latitude

hello every one

if you are counting on the file size and downscaling it to DXO 8Mp you have 0.66 stop of improvements to 7dmk2
it means about 12 stops of dynamic range in 5ds
Not bad


You can't have both at the same time. Maximum editing latitude is ultimately determined by dynamic range. To speculate a bit. One thing that could be going on is that testers who are saying "more dynamic range" are feeling more freedom from lower banding. They may FEEL they have the ability to push shadows more, because the noise they see has a better characteristic. You often didn't need to push shadows at all in the past to see banding...it was often visible right there in the midtones. That would be zero stops shadow pushing, if that was the case. Without banding, the difference between not pushing at all (rendering the image to the 8 stops of an 8-bit screen), or pushing a small amount (maybe a stop before banding stopped you)...and pushing up to the three stops allowed by ~11 stops of dynamic range, would seem quite huge.


Perceptually, that would seem like a huge difference. Mathematically, things wouldn't have really changed. If, mathematically, the 5Ds really does have "the same" dynamic range as the 5D III, then logically the claims of more shadow pushing ability are more perceptual than mathematical (and, it would be unlikely that the 5Ds has over 13 stops of engineering DR.)


I don't know what the case is, but trying to reconcile the two statements A and B above, this is one of the conclusions I can draw. You either have the same dynamic range as the 5D III (or slightly better, more like the 7D II)...or you have more dynamic range. To have more dynamic range, you have to have either lower read noise or a larger FWC, or both. You can't have the same dynamic range and lower read noise at the same time...because lower read noise with the same FWC would mean you have more dynamic range.


At this point, to know any more, we need to see literal tests.

if you are counting there are 0,66 stop of more dynamic range in 5ds compared to 7dmk2, which means around 12 stop of dynamic range
better but not good as others

 

[jrista]

if you are counting there are 0,66 stop of more dynamic range in 5ds compared to 7dmk2, which means around 12 stop of dynamic range
better but not good as others

Where is your the 0.66 stops (2/3rd stops) more DR for the 5Ds coming from? Are you just guessing, or do you actually have concrete data?

 

[PureClassA]

....His first post on CR.. hmmm... Maybe one of Keith Cooper's Canon pre-tester contacts : ;D

If it has 12 stops. Great. I'm buying the 5DS mostly for controlled lighting situations anyway. Two Paul Buff Einstiens and two Alien Bees. Outdoor work too but still portraiture. 6D has been great. Just bought a 5D3 on Ebay for $2200 as a back up full frame because I have a ton of dancers to shoot in few weeks.

The DR debate is fascinating and I hope Canon steps it up, but I'm real happy with what they deliver now although I understand for some work it isn't the best. C'est la vie folks.

 

[K-amps]

Dynamic range won't help you become a better photographer!

Usually shoot with a pinhole camera?