December 12, 2017, 05:08:12 PM

Poll

will be?

dead on arrival. less dynamic range, worse low light than comp. handicapped video feautures
too little too late. equal to competition but released in late 2018
blow away comp. 4k 60p(FF and crop), IBIS, 10fps, super fast buffer, 14 stops DR
canons first mirrorless camera
already preordered my a7riii or d850

Author Topic: The 5DsR mk2  (Read 15686 times)

traveller

  • EOS 5DS R
  • ******
  • Posts: 808
Re: The 5DsR mk2
« Reply #90 on: December 07, 2017, 06:35:13 AM »
The current 5Ds (sr) will take a remarkable amount of over-exposure at low ISO without loss of saturation / colour accuracy. If less noise is your goal just over-expose and reduce in post.

Doesn’t this just mean that the histogram is very inaccurate in RAW and the meter underexposes? I’m not criticising Canon alone, just about every camera (other than the Phase One XF with the latest IQ3 backs) does the same.

On my 5D3, I set the “Neutral” picture style, then customise it by lowering the contrast to minimum and bumping up the sharpness a little (4, -4, 0, 0). I find this makes the histogram closer to the RAW file, while still allowing focus evaluation on the rear screen. It has zero effect on the file imported into Lightroom/ACR (I can’t speak for other RAW converters).

Not that the meter is inaccurate. Histogram yes, and I too set “ neutral” , very flat for preview. Slight over exposure just floods the pixels a little more, and as long as you don’t overdo it then reduction in post reduces noise. So you’re not actually using the “correct” exposure. Same with negative film in the old days, in fact even now with modern emulsions like Portra it’s virtually impossible to over expose it. Transparencies over exposed where awful and thin though I’m not sure if after scanning you’d pull the colours back - all my transparencies are under exposed if anything !

If, when shooting in RAW, you call reliably pull back a full stop of highlights without losing colour information from multiple channels clipping, then the correct RAW exposure was one full stop more than the meter suggested. The camera makers have not really adjusted their meter programming from the days when they had to account for people shooting films with different exposure latitudes (actually, they probably could have fitted a mode switch back then: by the end, exposure latitude was even DX encoded on 35mm film canisters!).

How long have people been asking for an "expose to the right" metering mode, perhaps with user specifiable maximum percentage (all) channel clipping thresholds? It's my (personal) highest priority request to Canon: give users customisation controls on the camera metering, like already exist for the AF system. I would love Canon to include an ETTR meter mode that allows me to specify how much of the frame is allowed to clip and also to customise how much priority the meter gives to the active AF point.

With regard to film comparisons, digital is far more like slide film with very sharp overexposure clipping. Digital achieves more of what we now call 'dynamic range' by having even greater shadow recovery potential than slide film (even on older Canon sensors), not by having vastly more headroom in the highlights. In signal:noise ratio terms, both positive (slide) film and digital (even more so) have a better SNR overall than negative film, especially so in the shadow areas, but negative film has a more gradual drop off in the SNR in the highlights:

http://clarkvision.com/articles/exposure_latitude-1/
http://clarkvision.com/articles/digital.signal.to.noise/

Effectively, this means that the exposure strategy for both digital and positive film is the same: expose for the highlights. The difficulty is judging where the highlight clipping point is, based upon the limited information from the camera. You either use a spot meter and mathematics, underexpose "for safety" (not ideal), or with digital -use the information provided by the histogram. The issue with most current digital cameras is that the meter and histogram are both based upon the correct jpeg exposure.

RAW headroom comes as a result of the demosaicing of the bayer array data: i.e. in the same way that generating a full-res RGB image relies on the "educated guesswork" of the RAW converter's demosaicing algorithm, so the data in a fully saturated 'pixel' can be reconstructed by extrapolating its value based upon the values of the adjacent non-clipped 'pixels'. Obviously, once the adjacent 'pixels' are also clipped there is no way to accurately "guess" the colour, which gives us the upper limit of RAW headroom. The camera companies seem very reluctant to give advanced users the tools they need to extract optimum exposure from their cameras. We can use RawDigger to establish the general relationship between our camera's meter/histogram and the actual RAW values, but it would be good if we could access more of this sort of data at the time of shooting. Perhaps the camera companies feel that it is easier just to provide bracketing modes, but this doesn't help with moving subjects.

canon rumors FORUM

Re: The 5DsR mk2
« Reply #90 on: December 07, 2017, 06:35:13 AM »

Mikehit

  • Canon EF 300mm f/2.8L IS II
  • ********
  • Posts: 2185
Re: The 5DsR mk2
« Reply #91 on: December 07, 2017, 07:16:29 AM »
...
The issue with most current digital cameras is that the meter and histogram are both based upon the correct jpeg exposure.

RAW headroom comes as a result of the demosaicing of the bayer array data: i.e. in the same way that generating a full-res RGB image relies on the "educated guesswork" of the RAW converter's demosaicing algorithm, so the data in a fully saturated 'pixel' can be reconstructed by extrapolating its value based upon the values of the adjacent non-clipped 'pixels'. Obviously, once the adjacent 'pixels' are also clipped there is no way to accurately "guess" the colour, which gives us the upper limit of RAW headroom. The camera companies seem very reluctant to give advanced users the tools they need to extract optimum exposure from their cameras. We can use RawDigger to establish the general relationship between our camera's meter/histogram and the actual RAW values, but it would be good if we could access more of this sort of data at the time of shooting. Perhaps the camera companies feel that it is easier just to provide bracketing modes, but this doesn't help with moving subjects.

I sometimes wonder if we are getting to the point where the technology makes things possible and people are demanding functionality for no other reason than in theory they can do it. And comments like 'if you thought that way they would not have developed AF' type of argument.

Bracket some shots, compare the raw and the histogram and take it from there. I know on my 7D2 I can take the jpeg histogram and add a stop - does it really matter if you are 1/3 of a stop more'accurate' for ETTR? In a landscape shot the perfect light is so fleeting you will not have time to take a test under the ideal lighting, wait for the camera to crunch the raw histogram and adjust the settings.
Add to this that the histogram depends on the white balance you use so a raw histogram can actually be misleading and blow colour channels if you try too hard to get as far ETTR as possible. So you end up being conservative and no more accurate than if you take 'jpeg histogram +1 stop'.

I have read several comments by pros saying that  the dynamic range of any DSLR is now so good it makes ETTR almost redundant.

And just to be cheeky: given the profound claims about the usability of a 5-stop push to an underexposed image, and the wonderful linearity of new sensors, why on earth are people worrying about the histogram anyways?

3kramd5

  • EOS-1D X Mark II
  • *******
  • Posts: 1847
Re: The 5DsR mk2
« Reply #92 on: December 07, 2017, 10:29:47 AM »
If, when shooting in RAW, you call reliably pull back a full stop of highlights without losing colour information from multiple channels clipping, then the correct RAW exposure was one full stop more than the meter suggested.

That would only be a true statement if the metering program was ETTR, but it isn’t.

Rather than auto-ETTR, I’d just like a raw histogram in live view.

I have read several comments by pros saying that  the dynamic range of any DSLR is now so good it makes ETTR almost redundant.

And just to be cheeky: given the profound claims about the usability of a 5-stop push to an underexposed image, and the wonderful linearity of new sensors, why on earth are people worrying about the histogram anyways?

Just to be pedantic: because the ability to make a shadow brighter without introducing overpowering noise doesn’t reveal detailed in that shadow which weren’t recorded. The signal is still the signal. Exposing to the extreme end of saturation allows maximum signal and thus detail to be recorded.
« Last Edit: December 07, 2017, 10:35:27 AM by 3kramd5 »
Some Canon, some Nikon, some Sony, some Olympus

RGF

  • Canon EF 300mm f/2.8L IS II
  • ********
  • Posts: 2450
  • How you relate to the issue, is the issue.
Re: The 5DsR mk2
« Reply #93 on: December 07, 2017, 11:47:18 AM »
It seems that after some problems with the 1D Mark 2 or 3 (?) Canon has been a lot more cautious about introducing new equipment, especially bodies.  It seems to me that they want to avoid negative market reaction.  But then again, given how (pick your term, detailed oriented users, hyper-critical bunch of whiners) we are, they seems to slow with new introductions.  The 5Ds/R was the first "breakthrough" in a while.

traveller

  • EOS 5DS R
  • ******
  • Posts: 808
Re: The 5DsR mk2
« Reply #94 on: December 07, 2017, 12:19:20 PM »
...
The issue with most current digital cameras is that the meter and histogram are both based upon the correct jpeg exposure.

RAW headroom comes as a result of the demosaicing of the bayer array data: i.e. in the same way that generating a full-res RGB image relies on the "educated guesswork" of the RAW converter's demosaicing algorithm, so the data in a fully saturated 'pixel' can be reconstructed by extrapolating its value based upon the values of the adjacent non-clipped 'pixels'. Obviously, once the adjacent 'pixels' are also clipped there is no way to accurately "guess" the colour, which gives us the upper limit of RAW headroom. The camera companies seem very reluctant to give advanced users the tools they need to extract optimum exposure from their cameras. We can use RawDigger to establish the general relationship between our camera's meter/histogram and the actual RAW values, but it would be good if we could access more of this sort of data at the time of shooting. Perhaps the camera companies feel that it is easier just to provide bracketing modes, but this doesn't help with moving subjects.

I sometimes wonder if we are getting to the point where the technology makes things possible and people are demanding functionality for no other reason than in theory they can do it. And comments like 'if you thought that way they would not have developed AF' type of argument.

Bracket some shots, compare the raw and the histogram and take it from there. I know on my 7D2 I can take the jpeg histogram and add a stop - does it really matter if you are 1/3 of a stop more'accurate' for ETTR? In a landscape shot the perfect light is so fleeting you will not have time to take a test under the ideal lighting, wait for the camera to crunch the raw histogram and adjust the settings.
Add to this that the histogram depends on the white balance you use so a raw histogram can actually be misleading and blow colour channels if you try too hard to get as far ETTR as possible. So you end up being conservative and no more accurate than if you take 'jpeg histogram +1 stop'.

I have read several comments by pros saying that  the dynamic range of any DSLR is now so good it makes ETTR almost redundant.

And just to be cheeky: given the profound claims about the usability of a 5-stop push to an underexposed image, and the wonderful linearity of new sensors, why on earth are people worrying about the histogram anyways?

Exactly, why do we need autofocus, when we can just take a few shots at different distances and check the rear screen to see which is best?

I know that you were being tongue in cheek, but in all seriousness those 5 stop shadow pushes are often still not enough for contre-jour shots, especially when you haven’t properly ETTR’d the exposure, or you are much above base ISO. That being said, I find myself pushing the exposure to the right less often with my Fuji than with my Canon, at low ISO anyway. 

stevelee

  • EOS M5
  • ****
  • Posts: 224
Re: The 5DsR mk2
« Reply #95 on: December 07, 2017, 01:22:54 PM »
Several thoughts:

The histogram you see on the screen is based on the theoretical resultant JPEG rather than the linear RAW file, so it is difficult to judge how to adjust the exposure, particularly ETTR, from that readout. In a very low contrast situation, though, you can probably get away with it. I think there is a lot of value in trying things and gaining experience, the more of which you have with your particular camera, the better your judgment of what you can get away with should become.

I have read that ETTR is pointless except at base ISO. That might not be quite literally true for near-base settings, but maybe it is. Anyhow, if ETTR means you are doubling the ISO and quintupling the noise, then you've introduced a lot more problems than you've solved.

My experience has been that if there is any information at all in any channel of the brightest parts of the scene, then the Highlights slider in ACR (and presumably the same in LR) does a good job faking details in the other channels. The main need I've had for this, as I recall, is when I want some detail in clouds. Moving the slider way to the left can make the sky look downright threatening sometimes, which is vastly beyond the tweaking I need. So for my purposes, highlight recovery usually works better than boosting the shadows more than one stop. I'm usually not interested in bringing out the spider webs in the dark corners for interiors, and indeed I find too much attention to insignificant detail to be more of a distraction in the picture. I do however like to have detail in windows, particularly stained glass, while giving a good view of architectural details. The Highlights slider is usually not sufficient for that, for me anyway, so I shoot separate exposures for the windows and merge, such as in this picture of a chapel in Edinburgh, shot with my G7X II. I preferred the as-shot convergence over a corrected perspective. My goal is usually to make the picture look like what I saw when I was there. This printed up nicely on 13" x 19" paper, and I plan to frame it to hang in my hallway gallery when I get around to it.


Mikehit

  • Canon EF 300mm f/2.8L IS II
  • ********
  • Posts: 2185
Re: The 5DsR mk2
« Reply #96 on: December 07, 2017, 05:27:13 PM »

Just to be pedantic: because the ability to make a shadow brighter without introducing overpowering noise doesn’t reveal detailed in that shadow which weren’t recorded. The signal is still the signal. Exposing to the extreme end of saturation allows maximum signal and thus detail to be recorded.
My point was, the difference between jpeg histogram + 1 stop (or whatever you think is right) it surely good enough. Is the 1/3 stop (or whatever it is) by using a raw histogram really going to make or break an image?

canon rumors FORUM

Re: The 5DsR mk2
« Reply #96 on: December 07, 2017, 05:27:13 PM »

Mikehit

  • Canon EF 300mm f/2.8L IS II
  • ********
  • Posts: 2185
Re: The 5DsR mk2
« Reply #97 on: December 07, 2017, 05:29:19 PM »


I know that you were being tongue in cheek, but in all seriousness those 5 stop shadow pushes are often still not enough for contre-jour shots, especially when you haven’t properly ETTR’d the exposure, or you are much above base ISO. That being said, I find myself pushing the exposure to the right less often with my Fuji than with my Canon, at low ISO anyway.

But if its contre jour even the raw histogram becomes meaningless because the whole idea of contre jour is that the highlights blow out to the point that even a raw histogram will not tell you if the part you are interested in is within the dynamic range of the histogram.

3kramd5

  • EOS-1D X Mark II
  • *******
  • Posts: 1847
Re: The 5DsR mk2
« Reply #98 on: December 08, 2017, 08:52:01 PM »

Just to be pedantic: because the ability to make a shadow brighter without introducing overpowering noise doesn’t reveal detailed in that shadow which weren’t recorded. The signal is still the signal. Exposing to the extreme end of saturation allows maximum signal and thus detail to be recorded.
My point was, the difference between jpeg histogram + 1 stop (or whatever you think is right) it surely good enough. Is the 1/3 stop (or whatever it is) by using a raw histogram really going to make or break an image?

Maybe, maybe not, but why not make it possible to truly ETTR without trial and error?

Here’s another one: let me dial in exposure time on a touch screen rather, at let me exceed 30 seconds without an extra device.
Some Canon, some Nikon, some Sony, some Olympus

RGF

  • Canon EF 300mm f/2.8L IS II
  • ********
  • Posts: 2450
  • How you relate to the issue, is the issue.
Re: The 5DsR mk2
« Reply #99 on: December 10, 2017, 01:05:39 PM »
Wonder if the Nikon D850 will put enough pressure on Canon to up their game.

They are already in a dead heat with Nikon D5 vs 1DxM2 (some could argue one or other is winner but IMO these camera are very close in features)

The 5DM4 is a great camera, I think unchallenged by the Nikon

Nikon's D850, though lower in MP, than the 5Ds/sR offers a lot of advantages of the 5Ds.
It would be nice is Canon could match the D850 in areas other than MP which it has a clear lead.

canon rumors FORUM

Re: The 5DsR mk2
« Reply #99 on: December 10, 2017, 01:05:39 PM »