DR and RAW bit depth on Canon's forthcoming...

[Flake]

The human eye is a precision instrument, it will detect meaningful differences if they exist. If it can't detect them, then they are irrelevant, no matter what tests tell you. It's not much use having an extra stop of DR, if your meduim of choice can't show it. I suspect that you'd need an expensive Spectraview type monitor to see any differences.

See the above video and you'll understand why you're thinking too quickly. The human eye has MASSIVE dynamic range. We can see out of a window and still see every detail in the dark room we're in. When taking a photograph out of a window, the result is that the room will be severely underexposed. We need to push up its exposure in post-processing to recover any detail in the underexposed area. Just simply overexposing shadows will show you a difference in dynamic range, which by the way is a real-life-situation thing, as that's extremely common in postprocessing images. The fact that cameras lack this dynamic range is the whole reason why HDR photography exists.
Cameras are still very far from having the dynamic range or sensitivity of the human eye.

Except that the human eye does not work in the same way as a camera! Persistence of vision means that we are able to scan a scene and the iris is able to adjust its aperture to allow less light to reach the retina, giving the impression that we have a huge dynamic range when we do not. In addition rods and cones in the retina allow difference sensitivities to light, and we can to some extent see in very low light, albeit only in black & white.

The human eye has an approximate focal length of 50mm and around 50MP equivalent, having two of them on the front of the head means we have stereoscopic vision and can perceive depth and 3D, again without two images a camera cannot replicate this.

 

[Tijn]

Except that the human eye does not work in the same way as a camera! Persistence of vision means that we are able to scan a scene and the iris is able to adjust its aperture to allow less light to reach the retina, giving the impression that we have a huge dynamic range when we do not.

I'm aware of that. And yet, we do have a very large "dynamic range" even when counting in that difference. We can see stars and we can see outside on sunny days, but not at the same time. Still, we can see outside a window and at the same time keep seeing detail of the room inside. Without pupil adjustments. My camera won't do that as well as I "do", at least not without post-processing.

 

[tt]

Except that the human eye does not work in the same way as a camera! Persistence of vision means that we are able to scan a scene and the iris is able to adjust its aperture to allow less light to reach the retina, giving the impression that we have a huge dynamic range when we do not.

I'm aware of that. And yet, we do have a very large "dynamic range" even when counting in that difference. We can see stars and we can see outside on sunny days, but not at the same time. Still, we can see outside a window and at the same time keep seeing detail of the room inside. Without pupil adjustments. My camera won't do that as well as I "do", at least not without post-processing.

Out of curiosity - what would be the resolution of the foveal vision?

 

[KeithR]

As long as I can get from this:

to this:

And from this:

to this:

with my 7D, I don't see what the problem is - there's pretty wide DR in those.

Personally though, DR at the top of the histogram is what matters to me, and that seems to be Canon's priority too, I'm pleased to say:

(No idea why the "recovered" bike image is presenting small here, it's the same size as the underexposed one...)

And I suppose this post will get me some more negative karma from the children here...

 

[Marsu42]

with my 7D, I don't see what the problem is - there's pretty wide DR in those. Personally though, DR at the top of the histogram is what matters to me, and that seems to be Canon's priority too, I'm pleased to say:

... but there's a catch: I used to use the htp mode on my 60D on every shot until I realized that even the expanded iso range compresses the shadows so much that recovering them brings in much noise. Nowadays, I simply underexpose 1/3ev and shoot raw - the only drawback that highlight recovery in lightroom torpedoes clarity when used at the same time.

However, I agree with you that the current 18MP sensor has enough headroom for most shots, and if not it's in typical hdr situations when a bit higher dr wouldn't be of any use anyway.

 

[Mt Spokane Photography]

As long as I can get from this:

The human eye can see much of the fine detail missing in the images, even after they are adjusted. Things can always be better.

 

[thatcherk1]

I think the main point of this thread isn't that we want to shoot the sunset without filters and expect perfect exposure from shadows to the sun itself. I think what people are wanting is Canon to produce a sensor with DR that other manufacturers have already achieved. It's more of a comparison to other brands and a worry that Canon will continue to fall behind with DR. In a day and age of "raw" workflows, people like the idea that they can shoot in the field, and spend hours and hours perfecting a digital grad filter rather than be stuck with a poor decision made when taking the photograph. Sometimes I'd prefer to just get it in the ballpark in the field, and not have to touch it much in post. But I certainly can't fault anyone who wants to perfect it later. With digital raw workflow we can achieve some beautiful (and realistic looking) images that would have been difficult or impossible in the days of film with only "in the field" and limited darkroom tools.

Also a note about monitors/photo paper, etc. In practical terms, people want a couple more stops of DR not so they can get some muddy looking image that shows 14 stops of latitude. The purpose is to smartly make adjustments in LR and PS to display those 14 stops in a beautiful way with things like the curve tool, or the grad/brush tools in LR. Or for others such as photojournalists, occasional under/overexposure happens, and more photos will be recoverable if there is more DR to work with especially with LR4(beta)'s exposure adjustment which is highly accurate to actual in-camera exposure adjustment.

Adobe's latest raw processing is exceptionally good at rolling off highlights similar to the way film does. It makes clipping sections of the image look much more pleasant. As soon as I start pulling exposure in LR, that nice roll-off begins to lessen until the clipping looks like crappy old digital. So any number of stops that are added to a camera's capability will only improve the beautiful film-like rolloff.

And as to a comment about how some images you want washed out, it's insanely easy to take even a 30 stop HDR and crush it down to 1 stop of range. It's never a problem to limit DR in post.

 

[LetTheRightLensIn]

Personally though, DR at the top of the histogram is what matters to me, and that seems to be Canon's priority too, I'm pleased to say:

Most DSLR sensors are linear capture so there is no such thing as "at the top of the histogram" and the fact is you can't get out of a 7D image (and yes I have one) wht you can from a D7000 at ISO100.

 

[LetTheRightLensIn]

I've said before, that lab tests are pretty meaningless, unless they reflect the real world, after all, as photographers, we are judged by what we achieve, not by what could be theoretically achieved.

But they do! Have you seen what a heavily pulled D7000/D3x ISO 100 shots looks like compared to one from 50D or 5D2?

Flake mentioned the use of HDR and grad filters. As a landscape photographer, I know that even with an extra stop of DR, there is no way that I could capture most scenes without using grad filters, unless I make some compromises. In fact, photography is all about making compromises, that is what sets the greats apart from the merely good.

But once you are into nearly 3 stops of usably better DR then it does start to make much more a difference in what things you can capture or not.

It's not much use having an extra stop of DR, if your meduim of choice can't show it. I suspect that you'd need an expensive Spectraview type monitor to see any differences.

Not at all. In fact most of them are IPS and don't even have very huge contrast ratio compared to some cheap SPVA screens. But it makes a big difference either way.

And seriously if Canon had 3 usable stops better DR like it did back when, compared to Nikon, would all of you still be finding every excuse in the world why we don't need it and who cares??

Why would anyone, most of all a Canon user, NOT want Canon to fix this up???
I care because I have Canon equipment.

 

[LetTheRightLensIn]

You don't need to do the Stoffer Wedge Test, its been done many times. You can see the usable DR range of competiting camera bodies here:

5D MK II 8.4 EV
Nikon D700 7.8 EV

http://www.dpreview.com/reviews/CanonEOS5DMarkII/page25.asp

You can get more by manipulating RAW files, but the images tend to look flat and ugly.

I'd also like more DR, but, so far, no one seems to match film.

DPR test is ridiculous, it depends on nothing other than what tone curve is in the jpg setting.

You still get a tone curve processing in RAW. you have to manulipate it to get a little more DR. Then images tend to look flat. DPR tells you the real world info. Its like a statement of what you get off the showroom floor whan you buy a automobile, its what many, if not most get.

No you don't have to make it flat, you just apply special curves, maybe even different ones to different parts of the image and you can end up with rich contrast and color and nice detail without too much mush/noise/poor color and without weird over the top HDR look if you go about it right.

DPR doesn't come close to telling you the real world info. Even without using just a simple curve the in cam jpgs don't come close to the real story from RAW. You could easily chose a jpg style that would make any DSLR look better than any other from DR and then you coudl pick another one and reverse the results.

 

[LetTheRightLensIn]

Reading some of these posts makes me think that there are some photographers who think it should be realistic to take photos of the setting sun and still keep details in the shadow areas! Hardly any mention of HDR, and no mention at all of specialised filters.

Back in the days of film - and DR comparison of film and digital is incredibly difficult, companies like Lee & Cokin were producing the same filters as they are today. That's photographic technique to improve the dynamic range, and even with the dynamic range improvements being mentioned on the thread, it's still not enough to negate the use of ND grads .

It also makes me wonder if this is a fault of the camera or the monitor? Often I have images which appear to have a blown sky, and increasing exposure in PP of course makes it worse, but using the ND grad tool reveals that the camera has indeed captured all the detail in the sky, it's just that the exposure level needs to be reduced so it can be viewed. Here is a piece from Wikipedia on the subject:

The most severe dynamic-range limitation in photography may not involve encoding, but rather reproduction to, say, a paper print or computer screen. In that case, not only local tone mapping, but also dynamic range adjustment can be effective in revealing detail throughout light and dark areas: The principle is the same as that of dodging and burning (using different lengths of exposures in different areas when making a photographic print) in the chemical darkroom. The principle is also similar to gain riding or automatic level control in audio work, which serves to keep a signal audible in a noisy listening environment and to avoid peak levels which overload the reproducing equipment, or which are unnaturally or uncomfortably loud.

You are not reading the posts correctly then.

 

[jrista]

I think everyone is largely missing the point of my original post. I want improved headroom in my photographs. I'm not looking to produce photos that are dull and flat and expose every last scrap of dynamic range a sensor is capable of. I am looking to have some buffering and room within which to push my exposures around. Its a dynamic range...not a constant range. The whole point is to attenuate the raw image to look the way you want it to look within that dynamic range (which means the final result will exhibit LESS tonal range than possible when utilizing the full, unattenuated DR the sensor is capable of), and have the freedom to shoot in scenes with broad contrast without too much fear of encountering the worst problem with digital photography: the hard highlight cutoff.

Film is an analog device, and as such, when it comes to highlights, you have a lot of room to "push" exposure without fear that your going to hit a brick wall after which all additional exposure information is worthless and causes unrecoverable blown highlights no matter your skill in post. Digital sensors, however, DO have a brick wall, and its all too easy to blow out highlights with Canon cameras. Two additional stops of DR, or four times as much sensitivity to light, would greatly increase the amount of tones that could be allocated to highlights, allow smoother and broader highlight "shoulders" when it comes to default tone curves applied by cameras or RAW processors (better simulating film), etc. The benefit is not in what you get strait out of the camera...the benefit is in greater freedom to correct in post, and less need to correct with filtration in-camera because you have more highlight headroom.

Canon seems to have stopped pushing the envelope some 3-4 years ago when they reached about 11.7 stops of maximum DR (thats according to the DXO tests, which measure every level possible from only a few photons to the point where the sensor just begins to blow out highlights.) During the same time period, Nikon and Sony have continued to push the envelope, such that they are now able to utilize almost all of the theoretical headroom possible with 14-bit RAW images. As someone who is pretty heavily invested in Canon equipment, as someone who likes Canon gear and gear options, as someone who is thoroughly invested in my future as a photographer through Canon...I want them to compete and innovate on all fronts. Not just ISO. Not just sensor resolution. Not just lens quality and variety. Not just camera features, AF, fps, etc. Dynamic range matters too, particularly to nature photographers who shoot out in bright daylight frequently, where high contrast scenes and blown highlights (or the potential for them) are common. Its been years since we've seen any real improvement on that front from Canon.

I'm not some wishy-washy whiner who complains about jumping ship to Nikon (or vice versa when the tables are reversed) any time the other brand announces something new and better. I LOVE Canon glass. I LOVE how Canon still innovates in the lens space, with concepts like diffractive optics, built-in teleconverters, specialty lenses like the MP-E 65mm that no other brand has any counterpart for at all. I don't think there is anything better or more versatile than Canon glass in a general sense (although there are the occasional specialized lenses from the likes of Zeiss that certainly do incredible things, such as a lens that can resolve nearly 400lp/mm wide open). I don't want to have to go looking to another brand in a few years because they have pushed the envelope to 15- or 16-bit RAW, and are achieving nearly another two full stops of dynamic range...while Canon simply claims 16-bit RAW, but is still only achieving 12 stops of DR. That would plain and simply suck, and Canon would have less excuse for that than they do now when they claim 14-bit RAW and still only get 11.7 stops of DR.

I own the 7D, and Canon has thoroughly touted their 14-bit image processing pipeline and 14-bit RAW images it can produce...its rather miffing not to be able to fully utilize all 14 bits on every level...when its clear that its possible because other manufacturers have done it.

 

[dtaylor]

Have you ever manipulated a low ISO D3x or D700 file and then gone an manipulated one from Canon? The difference is every bit as real as DxO says it is.

Yes I have. No it is not. Shoot a transmission step wedge with just about any body and compare your results with DxO. Then you'll understand why I laugh at their DR claims.

 

[wockawocka]

I'm very surprised that Canon hasn't gone to 16 bits yet. I've been testing out an Aptus 22 back out and it blows away my 1Ds3 for IQ, but that's all it does.

 

[dtaylor]

You still get a tone curve processing in RAW. you have to manulipate it to get a little more DR. Then images tend to look flat. DPR tells you the real world info. Its like a statement of what you get off the showroom floor whan you buy a automobile, its what many, if not most get.

When you process a RAW for maximum DR you don't leave the image flat and walk away. And you don't lose detail if your output device can't handle 10 or 11 stops. You're basically getting all the detail out of the RAW that you can, compressing it down for monitor or print, and restoring the missing contrast and color without blowing back out the shadows and highlights (i.e. LCE, vibrance, saturation...sometimes with masks to protect specific areas).

 

[dtaylor]

I think the main point of this thread isn't that we want to shoot the sunset without filters and expect perfect exposure from shadows to the sun itself. I think what people are wanting is Canon to produce a sensor with DR that other manufacturers have already achieved.

LOL! No one else has achieved this.

See KeithR's posts for dramatic examples of recovery on Canon's 18 MP sensor. Nikon's 16 MP sensor is better than this, but not by much, certainly not by the amount people make it out to be.

Go back to the introduction date of the 7D, the first body with Canon's new 18 MP sensor. I'm pretty sure at that moment Canon had the widest DR crop sensor. Whoever has the newest sensor is likely to have the lowest noise / widest DR until a newer one comes out.

 

[Kernuak]

The human eye is a precision instrument, it will detect meaningful differences if they exist. If it can't detect them, then they are irrelevant, no matter what tests tell you. It's not much use having an extra stop of DR, if your meduim of choice can't show it. I suspect that you'd need an expensive Spectraview type monitor to see any differences.

See the above video and you'll understand why you're thinking too quickly. The human eye has MASSIVE dynamic range. We can see out of a window and still see every detail in the dark room we're in. When taking a photograph out of a window, the result is that the room will be severely underexposed. We need to push up its exposure in post-processing to recover any detail in the underexposed area. Just simply overexposing shadows will show you a difference in dynamic range, which by the way is a real-life-situation thing, as that's extremely common in postprocessing images. The fact that cameras lack this dynamic range is the whole reason why HDR photography exists.
Cameras are still very far from having the dynamic range or sensitivity of the human eye.

I'm not disagreeing that the human eye has much higher DR than any camera, simply that current displays and paper aren't able to match it. If the increases in a sensor's DR isn't replicated in the medium we choose to see the image, then the advantages are limited. HDR compresses the dynamic range, but consumer displays are unable to see the range, until it is tonemapped, which is one of the reasons it looks so unnatural in many cases. Even the use of grad filters can make a scene look unnatural, simply because of the smaller difference between the highlights and shadows (although it isn't compressed in the same way).

 

[Kernuak]

I've said before, that lab tests are pretty meaningless, unless they reflect the real world, after all, as photographers, we are judged by what we achieve, not by what could be theoretically achieved.

But they do! Have you seen what a heavily pulled D7000/D3x ISO 100 shots looks like compared to one from 50D or 5D2?

While I've seen the D3, I haven't seen results from those cameras, but have any real comparisons been made? I'm talking about side by side comparisons in the real world, so that as many variables as possible can be ruled out. We can't judge comparisons, if we don't know the conditions at the time of taking. Simple comparisons of unrelated images is simply unscientific and less meaningful than lab tests under artificial conditions (where you can of course control things at least). I'm not saying that there aren't differences, but it's very easy to jump to conclusions. It's very rare, that reviews make direct side by side comparisons, taking near identical shots. Even when you do make comparisons though, Nikon and Canon cameras seem to have different compromises, aiming for more latitude at opposite ends of the spectrum. I've seen too many cases when I was doing research, of papers that put a spin on something that wasn't actually true, when looked at in detail or repeated.

Flake mentioned the use of HDR and grad filters. As a landscape photographer, I know that even with an extra stop of DR, there is no way that I could capture most scenes without using grad filters, unless I make some compromises. In fact, photography is all about making compromises, that is what sets the greats apart from the merely good.

But once you are into nearly 3 stops of usably better DR then it does start to make much more a difference in what things you can capture or not.

Three stops is obviously a different matter, but that is a huge jump, in excess of 30%, so is it realistic?

It's not much use having an extra stop of DR, if your meduim of choice can't show it. I suspect that you'd need an expensive Spectraview type monitor to see any differences.

Not at all. In fact most of them are IPS and don't even have very huge contrast ratio compared to some cheap SPVA screens. But it makes a big difference either way.

And seriously if Canon had 3 usable stops better DR like it did back when, compared to Nikon, would all of you still be finding every excuse in the world why we don't need it and who cares??

Why would anyone, most of all a Canon user, NOT want Canon to fix this up???
I care because I have Canon equipment.
[/quote]
I didn't say that I wouldn't want more DR, quite frankly, any photographer would always be looking for improvements, whatever they are. However, the type of photography I do would probably benefit more from other improvements, such as improved ISO sensitivity (i.e. cleaner). I can already compensate mostly for shortcomings in DR, but noise at higher ISO isn't as easy and if I want to shoot wildlife around dusk, that is a priority for me. Also, some were offering film as having more DR than Canon's sensors, which certainly isn't the case for slide film and is debatable for negative film. If we want improvements, then we have to be realistic in what we compare it too.

 

[jrista]

Have you ever manipulated a low ISO D3x or D700 file and then gone an manipulated one from Canon? The difference is every bit as real as DxO says it is.

Yes I have. No it is not. Shoot a transmission step wedge with just about any body and compare your results with DxO. Then you'll understand why I laugh at their DR claims.

You should take a look at the videos on this page:

http://testcams.com/blog/2011/05/03/nikon-dx-vs-canon-aps-c-dynamic-range/

DXO's results are absolutely real. It boils down to noise floor, and Sony sensors have the lowest floor of any on the market today...by more than two full stops above and beyond what Canon has to offer. The scenario in the link above is not as rare as it may seem. I shoot wildlife and birds a lot, and when it comes to most of the wildlife I shoot...the deer family and canines, they are usually out during golden hour. There isn't much light during that time of day, and the necessity of pushing exposure in post is frequent. I've encountered the read-noise banding issue demonstrated in the videos from the link above frequently...its a detail destroyer like you wouldn't believe, and its impossible to eliminate without significant downscaling. Increasing ISO helps reduce the read noise (for some reason read noise drops as ISO increases with Canon designs), however every stop of ISO increase is also a stop of DR loss, so its a trade-off.

You can see the difference in read noise levels at sensorgen.info:

Nikon D7000:
http://sensorgen.info/NikonD7000.html
Consistent read noise at 3 e-
Mostly linear drop in DR as ISO increases
Max DR at 14 stops (13.9 really), dropping to 12.6, 11.7, 10.7, etc. as ISO increases


Canon 7D
http://www.sensorgen.info/CanonEOS_7D.html
Max read noise of 8.6 e- (almost three times that of Nikon)
Average read noise of 3 e- after increasing ISO by three stops
Loss of DR for ISO 100 and ISO 200, ISO 400 is right on the money, ISO 800 loses 1 stop as expected
Max DR at 11.2 stops at ISO 100, 200, and 400

Canon 5D II
http://www.sensorgen.info/CanonEOS_5D_MkII.html
Max read noise of 27.8 e- (almost 10 times that of Nikon!!)
Loss of DR through the first four ISO levels, 100-800
Max DR of 11.2 stops at ISO 100, less than expected DR at ISO 200, 400

(NOTE: The 5D II can somewhat handle the higher read noise at low ISO because its pixel capacity is higher than either the D7000 or the 7D because its Full Frame. There are still a lot of complaints about the 5D II's banded read noise though, so its still a visible problem.)

I think the main point of this thread isn't that we want to shoot the sunset without filters and expect perfect exposure from shadows to the sun itself. I think what people are wanting is Canon to produce a sensor with DR that other manufacturers have already achieved.

LOL! No one else has achieved this.

See KeithR's posts for dramatic examples of recovery on Canon's 18 MP sensor. Nikon's 16 MP sensor is better than this, but not by much, certainly not by the amount people make it out to be.

Why don't you take a look at the videos in the link above. Sony (and Nikon via Sony sensors) HAVE INDEED achieved better DR and FAR, FAR lower read noise than Canon. The difference is unbelievable. For many photographers, it may not matter, but for those who shoot in lower light, and bleed from their eyes every time they work a photo in post and have to apply an ungodly amount of noise reduction, watching their precious detail disappear as they do so...it matters a LOT.

Go back to the introduction date of the 7D, the first body with Canon's new 18 MP sensor. I'm pretty sure at that moment Canon had the widest DR crop sensor. Whoever has the newest sensor is likely to have the lowest noise / widest DR until a newer one comes out.

Yes, but that was three years ago! Canon hasn't innovated a sensor since (except the 1D X sensor, which we still have yet to see if it improved DR and read noise or not. It could be exactly the same as its always been, without any improvement at all!) It's been three years since Canon put something other than the same old 18mp sensor in any camera other than their 1D line (and the 1D IV still has the same read-noise banding that all of Canon's other sensors have...search for one of the many 1D IV noise videos on Vimeo for real-world examples.) Post-process shadow recovery is has become an important part of photography for many photographers...not just from the standpoint of fixing a botched exposure, but for those who simply don't have available light, can't or choose not to use flash (for obvious reasons), and need shadow recovery in post to fully realize the shot they took.

It is indeed possible to build a sensor that is capable of using over 99% of the theoretical dynamic range possible with 14-bit data. Sony's done it, they've been doing it for a couple years, and there are many visual examples that prove the point on the net if you care to actually see actual evidence of the difference. For those who know of the issue, and for those to whom it matters, Canon will really start losing business over the next couple of years if they continue to demonstrate an inability or lack of desire to fix the problem in the face of the innovation their competition is showing.

 

[LetTheRightLensIn]

Have you ever manipulated a low ISO D3x or D700 file and then gone an manipulated one from Canon? The difference is every bit as real as DxO says it is.

Yes I have. No it is not. Shoot a transmission step wedge with just about any body and compare your results with DxO. Then you'll understand why I laugh at their DR claims.

You just don't get it.