Dynamic Range - Try it for yourself, conclude for yourself: 5D III vs. A7r

[Zv]

It surprises me that no-one has tried to put a Sony sensor into a Canon 5d3 and make the perfect camera.
Swapping engines and transmissions in cars is not uncommon, so why not swap sensors in cameras?

I don't think it's a simple as changing a lens or a focusing screen. The sensor is the guts of the camera and it's hard wired in. I'd imagine Sony uses completely different components than Canon so you can't just stick it on and expect it to work. Then you need to programme the digic processor to work with the new sensor. So you'd need the electrical engineering skills from both Sony and Canon plus the coding skills of the ML team to stand a chance. How many attempts would it even take? Do you wanna buy a bunch of $3000 cameras just so you can play Dr Frankenstein with them?

I bet there is some lunatic out there that is probably doing this very thing as I type this!

 

[Rofflesaurrr]

@Khalai: No, there are no screens that can natively display 14 or more stops at once. However, it is for that very reason that we push shadows. A RAW is rendered to the screen with a tone curve. That tone curve compresses the information in a RAW, crushing the blacks and bleeding out the highlights. The middle part of the image fits on the screen...the middle 8 stops.

The purpose of shadow recovery is to bring back the shadows that were crushed and the highlights that were bleed out. The highlights aren't clipped, they are just rendered such that they appear clipped. The shadows aren't pure black...they are just rendered such that they appear pure black or nearly so. In the actual RAW file, in linear space, all that information is decidedly NOT black or white. It's all non-zero/non-clipped information that can be utilized.

Having more dynamic range, such as you get with an Exmor, means that more of that information that falls outside the display range of your computer screen is recoverable. Canon sensors are eating away at a large chunk of that information, then scattering banding (a semi-random or non-random factor) into the signal that reaches right up into the lower midtones, which are already displayed on the screen.

So, no, there are no screens that can natively display 14 stops of DR, which is the very reason we push shadows. I took some photographs of what appears to be a dark room. In reality, the room was not dark at all, it was much more like the second set of photos I shared. The information all those "blacks" as rendered in the first set of photos was all there, in the raw. My 8-bit screen, your 8-bit screen, most people's screens, cannot display the entire range of information found within the RAW, so I had to push the shadows up to make it fit. I compressed a wider dynamic range into a narrower dynamic range. In the third set of photos, I compressed the data even more, bringing in another stop of recoverable information in the A7r file that simply wasn't there in the 5D III file. (Hell, the 5D III file doesn't even have three stops of recoverable data, let alone four or five.)

It's actually because of the limitations of display devices and print media that we push and pull the digital signals of our RAW images around. Even when 10-bit computer screens become common, and 12-bit screens start hitting the marketplace, we will still be pushing shadows for print.

Pardon me if I'm incorrect, but although the bit-depth of a display can limit it's dynamic range, it's mainly about the levels of color gradation that are possible. Contrast ratio is the equivalent of dynamic range and that's mostly limited by the display technology itself. My main monitor (HP DreamColor) is an IPS display with a 10-bit panel, and can display 1024 levels of gradation per sub pixel. It's measured contrast ratio is slightly over 1000:1, and that's nearly 10 stops of dynamic range which is decent. Lower end IPS displays are around 9 stops . My previous monitors were VA panels with 3000:1 contrast ratios (~11 stops of DR). Now my reference monitor, which is a professional plasma display, is 12-bit over dual link HD-SDI with 16 bit A/D converter. It has a 30,000:1 static contrast ratio enabling it to display approximately 14.3 stops of dynamic range. OLED displays, which are not widely available yet in anything other than phones are capable of much, much more.

ANYWAY....

I'm a big fan of natural lighting whenever possible. The ceiling bounce flash in Lee Jay's photo worked extremely well and turned out a great photo. However, if that same flash were used in your living room scene, you'd lose all the soft shadows in the room from the window light and it would look rather... odd. Same thing with Keith's cockpit photo. The lighting is clearly artificial and it makes it seem as if there is bright sunlight pouring in from behind, which is not the case. All the shadows are also at the wrong angle for the light to be coming in the front glass. Sorry Keith, I'm not trying to bash your work, just critiquing.

I think if I had to shoot your living room scene without a flash on a Canon camera, I would blow the highlights a bit and then pull them in post. I have better luck recovering highlights with Canon RAWs than I do shadows.

Thanks for the RAW photo samples. I do wish Canon sensors had more DR, but like others have said. There's more to a camera system than the sensors. IMO Nikon ergonomics are crap, and Sony is full of bleeding edge technology with botched usability. While the A7R and A7S do deliver some outstanding results, I couldn't have one or the other as my day to day camera.

 

[privatebydesign]

@Khalai: No, there are no screens that can natively display 14 or more stops at once. However, it is for that very reason that we push shadows. A RAW is rendered to the screen with a tone curve. That tone curve compresses the information in a RAW, crushing the blacks and bleeding out the highlights. The middle part of the image fits on the screen...the middle 8 stops.

The purpose of shadow recovery is to bring back the shadows that were crushed and the highlights that were bleed out. The highlights aren't clipped, they are just rendered such that they appear clipped. The shadows aren't pure black...they are just rendered such that they appear pure black or nearly so. In the actual RAW file, in linear space, all that information is decidedly NOT black or white. It's all non-zero/non-clipped information that can be utilized.

Having more dynamic range, such as you get with an Exmor, means that more of that information that falls outside the display range of your computer screen is recoverable. Canon sensors are eating away at a large chunk of that information, then scattering banding (a semi-random or non-random factor) into the signal that reaches right up into the lower midtones, which are already displayed on the screen.

So, no, there are no screens that can natively display 14 stops of DR, which is the very reason we push shadows. I took some photographs of what appears to be a dark room. In reality, the room was not dark at all, it was much more like the second set of photos I shared. The information all those "blacks" as rendered in the first set of photos was all there, in the raw. My 8-bit screen, your 8-bit screen, most people's screens, cannot display the entire range of information found within the RAW, so I had to push the shadows up to make it fit. I compressed a wider dynamic range into a narrower dynamic range. In the third set of photos, I compressed the data even more, bringing in another stop of recoverable information in the A7r file that simply wasn't there in the 5D III file. (Hell, the 5D III file doesn't even have three stops of recoverable data, let alone four or five.)

It's actually because of the limitations of display devices and print media that we push and pull the digital signals of our RAW images around. Even when 10-bit computer screens become common, and 12-bit screens start hitting the marketplace, we will still be pushing shadows for print.

Pardon me if I'm incorrect, but although the bit-depth of a display can limit it's dynamic range, it's mainly about the levels of color gradation that are possible. Contrast ratio is the equivalent of dynamic range and that's mostly limited by the display technology itself. My main monitor (HP DreamColor) is an IPS display with a 10-bit panel, and can display 1024 levels of gradation per sub pixel. It's measured contrast ratio is slightly over 1000:1, and that's nearly 10 stops of dynamic range which is decent. Lower end IPS displays are around 9 stops . My previous monitors were VA panels with 3000:1 contrast ratios (~11 stops of DR). Now my reference monitor, which is a professional plasma display, is 12-bit over dual link HD-SDI with 16 bit A/D converter. It has a 30,000:1 static contrast ratio enabling it to display approximately 14.3 stops of dynamic range. OLED displays, which are not widely available yet in anything other than phones are capable of much, much more.

Exactly!

There is so much bad information in this thread it is depressing.

 

[privatebydesign]

To the OP: let's see you capture a scene that includes the actual sun disc recorded in the EV range and lift shadows from a shaded area with the A7r. This is the only situation in which my Canon gear cannot cope.

Of course you will fail. I know because I have already tried with a D800. That's why myself and hundreds, nay thousands of 'landscape' orientated photographers are still using Canon.

you mean like this one?..
www.canonrumors.com/forum/index.php?topic=8105.msg161888#msg161888
good luck using a Digic 4 or 5 Canon for a shot like the first one, you would not be able to re-tone the image to look like it does to the eye.
How does a Canon user cope with such a shot? Cry and go home? Spend many hours in post, trying to tame the stripes and fix the muddy dark tones to have some sort of semi-realistic texture? HAHAHA! Bracket?!? ;D

the last 2 in that post could even be done with a 7D, not that I'd chose one if I had an ABC alternative available.

Yeh, you can't do that with a Canon..........

 

[jrista]

Well, I can see the insults have started to fly. Sad. This community is really crummy at times. Some truly nasty people here.

On my front, today was a bit of a dud, multiple times. The A7r has a continuous bracketing mode, up to 5 frames. I tried it a bunch, and could never get it to work...it just reproduced the same exposure for all the frames. I ended up trying the manual bracketing, but I had to touch the camera, and there was no delay option. I know for a fact that the camera moved between frames, maybe even during the exposures, so who knows if they will turn out. I really love the IQ from this camera...but yeah, it has some major flaws (at least with Canon lenses...I cannot comment how good it is with native E-mount lenses, I suspect better.)

Hauled back to a site along the indian peaks range where I was hoping to get some nice mountainous landscape shots, but it turned out someone built a ranch on the land, so I could no longer get up there. Bust for the real landscapes...go ZIP today.

My parents live up here, and I set up in their yard to do some astrophotography. For some reason, I simply couldn't get the mount polar aligned earlier in the night. Took me a couple of hours to zero in on a good alignment, maybe another 5 minutes left of tweaking to fully dial it in, and clouds rolled in. ALMOST had it, and the night was over.

The worst was yet to come, though. While packing everything up, I caught my foot on something, knee went down hard on solid rock. Gouged the crap out of it, it's bleeding rather profusely right now...and it's swelling up like a balloon.

I can barely walk...not sure if I really damaged something or just banged it up good. I don't know what will happen tomorrow. I am not sure if I'll be able to walk, certainly in a lot of pain. I have a couple more sites that I want to visit tomorrow. Lighting was TERRIBLE today...really ugly, flat, direct, BORING. I managed some aspen shots, none of which I really like. Absolutely no color in the sky at sunset...blue faded to faint yellow-orange then gray right at the horizon, and the light just sucked, even during golden hour. I don't know that tomorrow will be any different. If I can walk, I'll try a couple more spots, see if I can get something useful. Otherwise, the aspen shots will have to do for comparison purposes. I'm not sure how well they will demonstrate things...

:\ Some days just suck.

 

[Aglet]

Yeh, you can't do that with a Canon..........

HAHA!

Do you expect me to take your example seriously?!?

C'mon, in my example the "pushed" shadow areas are still darker than the shadow areas in your "before" shot! ;D
You're not in a DR-limited scene. Other than the sun being in frame, that's some might flat dull scenery.

A large part of the lower right corner of my before shot is rendered RGB (0-255) 0,0,0 by the camera's internal jpg engine. In the after shot it's brought up the the mid-20s on average.
Your before shot dark area is about 11,17,20 brought up to about 22,35,42 in the after shot.

My Sol glint is not even clipped in the before shot; 248,247,218 is about the brightest pixel in the small image. In the after shot, it's still 249,246,212 so I'm not pulling my highlights down like in your example either.
My example is a proper ETTR shot to take advantage of the Exmor capabilities; maxing out the exposure at the brightest point in the image and re-curving the entire thing to bring back the tonality that is close to how it was visually perceived.

my after

vs your before

If you want to come up with a good comparison, leave the 1DS3 at home, grab a current Digic 4 or 5 body and see if you can replicate this much DR and recovery in one shot.

Thanks for playing tho.

EDIT: bonus point for picking a similar composition

 

[Aglet]

Of course there will be those that say my Aldeburgh beach shot didn't hit the noise floor ? Is that the term that has been bandied about ?

Here is a frame that is from the faster exposure, this one not actually part of a pano sequence. The first is the original, converted from the raw, the second is the push from the same frame, the third is a 100% crop of the push. There has been no noise reduction applied, this is a straight conversion from DPP. It is a blue boat which doesn't make the job easier. The fourth is a very quick brush together of the two conversions to show that the 5DII has enough dynamic range to produce a stupid picture where the shaded foreground is lighter than the sky.

I do agree that the 5DII is not as good in this respect as cameras such as the 1DsIII, and certainly nothing like as good as the 6D, but it is more than adequate for what I wish to do.

You silly insult from prior post is mere water on my Exmor duck. Don't waste your energy, save it for the photoshop work you need to do on your images.

Your dark areas, as presented, in this "dark" shot are still varying from a low of 1 or 2 to 5 or 6. So you may not have quite hit the noise floor.
AND you might have a good copy of a 5D2. Jeez, my early model had MIDtone banding when viewed 1:1!

 

[DarioVE]

@ Sporgon

it was already mentioned that DPP is better at develop cr2 raw expecially in the shadows, from this point your examples of the blue boat is very interesting even more with no noise reduction ...

it would be very useful if you would post your version of the 5D3 jrista file developed with your workflow so we have a direct comparision (i think jrista used lightroom since is similar to my results). In some way the windows area could be the clouds and the furniture the dark beach ... so even it is not intended to be an artistic shot i think it is informative.

I think it is probable that your DPP conversion would be better and i'm curios to see how much ...

In the end if not better sensor Canon could realise a better/powerful software or why not help Adobe improve the CR2 raw converter ??? Their DPP is free, they do not need to protect sales in this regard, maybe it is the opposite, there are many adobe customers that could get more out of canon so no need to search elsewhere.

 

[tomscott]

A lot of these examples are worst case scenario pushing the tech as far as it can go.

 

[Sporgon]

A lot of these examples are worst case scenario pushing the tech as far as it can go.

That's exactly right. The only reason I defend the Canon sensor is to refut the outrageouse comments that have been made against it by a few people. There seem to be thousands of people read CR but aren't members, and I guess many are inexperienced in photography, and it annoys the hell out of me thinking of those people reading some of the asinine comments made here, by people who are more versed in sounding like they know what they are talking about than actually producing pictures.

 

[Sporgon]

@ Sporgon

it was already mentioned that DPP is better at develop cr2 raw expecially in the shadows, from this point your examples of the blue boat is very interesting even more with no noise reduction ...

it would be very useful if you would post your version of the 5D3 jrista file developed with your workflow so we have a direct comparision (i think jrista used lightroom since is similar to my results). In some way the windows area could be the clouds and the furniture the dark beach ... so even it is not intended to be an artistic shot i think it is informative.

I think it is probable that your DPP conversion would be better and i'm curios to see how much ...

In the end if not better sensor Canon could realise a better/powerful software or why not help Adobe improve the CR2 raw converter ??? Their DPP is free, they do not need to protect sales in this regard, maybe it is the opposite, there are many adobe customers that could get more out of canon so no need to search elsewhere.

In this case this morning it was actually converted in an old version of DPP. No noise reduction, shot at 100 ISO ( which is actually 73 on the 5DII). I'll have a look at the mkIII file but I'm guess it is chronically under exposed. You can't push zero data and get an acceptable result, whether it is riddled with read noise or not.

 

[ChristopherMarkPerez]

Exactly.

Push the tech far enough and you'll find something you don't like.

I think what the original poster shows me is that the things Sony has over Canon is, literally, lost in the noise.

For every technical limit, I'll bet we can find an equally technical way of getting beyond that limit.

DR range not sufficient? Shoot three different exposures of the same scene and use luminosity masks to blend the exposures. The results will be much better in extreme dynamic ranges than anything you could take in a single shot and you'll have a lot more control over the end results.

Shadow noise got you concerned? They take multiple exposures of the same scene and then average. Noise reduction never had it so good.

Not enough mpixels? Shot multiple sections and stitch them together later. Ever interested in what a 25,000x10,000 pixel Canon 7D image might look like? I can tell you it's GLORIOUS!!! I have a series on iron and rivets that will soon be released and the details in la tour Eiffel is more amazing that you might realize.

I completely understand this is a "gear" forum and that I many times I ask why people aren't out taking photos. It can become monotonous at times. I get it.

What I'd ask is why people demand the technology at the edges deliver on their Pet Peeve when what they have is more than sufficient to just about any task you throw at it?

No one has yet been able to show me how in the Real World these little technical edge differences will make my life as an image maker any better than it already is. That's now small the differences currently are in what is being discussed in these kinds of threads. Big differences that I can see? That's a different matter. Then we'd have something to discuss.

My offer still stands: I bring a stack of large prints made using cameras and lenses of different makes. I'll throw in Sony A7r/Zeiss/Leica/Canon/Nikon/Fuji/AF/manual focus. You get time to consider DR, resolution, MTF, or anything else you feel you know something about and know (in your heart of hears) really matters. You look at the prints and each time you can tell me what camera/lens made which image, I'll buy you a beer. Each time you miss, you buy me a beer.

This offer is open to one and all. Pro. non-Pro. Know-it-All. Newbie. Doesn't matter. Deal?

A lot of these examples are worst case scenario pushing the tech as far as it can go.

 

[dtaylor]

* This is a far better test scene then a Coke box or anything else produced here for this topic. Kudos to jrista for that.

* Once again we see a pair of images where the total dynamic range is nearly identical between Canon and Exmor (sorry DxO), but the noise in the Canon shot means the Exmor sensor has more shadow latitude. (Yes, I'm going to be a pain about using correct terminology.)

* Canon noise/banding overwhelms RAW NR with this combination of underexposure and a +4 or +5 push. No surprise there. I've had no trouble at +2.5, and can usually manage +3, but if highlight retention is driving your shadows deeper you're going to have noise, banding, and detail loss with Canon.

* That said, this test scene demonstrates something I've said all along: the range between both sensors being OK and both sensors needing HDR is rather narrow. Despite the severe underexposure the highlights were not preserved and are not recoverable. If they don't matter then you don't need to underexpose this scene to this degree and the 5D3 will work fine. If they do matter then neither sensor can handle this scene without lighting or HDR...and the 5D3 will work fine.

Sometimes Exmor is easier to work with, or produces the better shadows in print, or lets you work with one frame where you would want two on Canon. Just not as often as some imagine.

* I don't think jrista deserves snark for this thread. He started a DR specific thread with fair samples.

* At the same time, I don't think anyone deserves to be called a "Canon fanboy" for discussing the practicality of this shadow latitude difference, especially in light of some of the hyperbole seen on this forum.

 

[Don Haines]

At the same time, I don't think anyone deserves to be called a "Canon fanboy" for discussing the practicality of this shadow latitude difference, especially in light of some of the hyperbole seen on this forum.

+1

Personally, I don't understand how someone can be accused of being a "Canon Fanboy" for pointing out that another company does something better. Could someone please explain the logic to me?

 

[privatebydesign]

At the same time, I don't think anyone deserves to be called a "Canon fanboy" for discussing the practicality of this shadow latitude difference, especially in light of some of the hyperbole seen on this forum.

+1

Personally, I don't understand how someone can be accused of being a "Canon Fanboy" for pointing out that another company does something better. Could someone please explain the logic to me?

Don,

You are falling into that fault of yours, hoping that any more than 1-3% of the replies here are based on logic.

I try to have "the discussion" but any attempts are derailed, disallowed, disputed, or the best one, taken in a completely different direction, to such a degree that it just isn't worth the effort.

There are some really smart people here with some very bad ways of putting things. I appreciate the effort jrista tried to make here, but the subject is so corrosive whatever approach is taken it seems doomed to failure.

From my own point of view I was genuinely interested in the topic, not least because I shoot a lot of images like the first post, interiors where I need something outside. What the files illustrate to me is the Exmor cannot do what I, personally, would need any better than my Canon, the shadow "detail", whilst being noise free, still has no tonality.

And that is just a simple issue of maths and what happens when the output from a linear capture "curve" is changed to a dramatic gamma style tone curve after a dramatic gamma tone curve has already been applied. If you take the bottom 5% of the tones and stretch it to fill the bottom 35% of the tonality then you have less than 1/7th the tonality, and that doesn't cut it for what I shoot in DR limited scenes. Further, the blooming is as bad in both images, and that doesn't work either.

So, whilst I appreciate for others the Exmor might be the answer, for me, who does regularly shoot DR limited scenes, the Exmor is not the answer and Jrista's images have proved that beyond any shadow of a doubt.

 

[LetTheRightLensIn]

Don't you think the irony of this is that many of the images shown on CR match or even surpass those on other sites ?

I try not to hang out on sites with mediocre photography I like perusing the sites and portfolios of others that make me strive to become a better photographer

Surely, there are many talented photographers on this forum. However, there are many people that run their mouths yet never post sample images. Gotta wonder how much their gear is really holding them back.

A lot of people like to keep their photography separate from all the tech/Canon marketing mess pushing controversy and feel freer to just call it all as they see it and so on.

And for the record, when you do come across the portfolios of those who ask for more DR, I have to say they generally are among the more impressive and comprehensive ones you'll find while many of those attacking them and talking about going out and shooting end up having a portfolio consisting of three, non-artistically, blurred photos of a cat in the back yard (not in your case though, just to make that clear though and for all I disagree with what and how he posts here, Sporgon does have a good portfolio too).

I mean heck did you miss the flickr account linked to earlier in this thread?

 

[LetTheRightLensIn]

It surprises me that no-one has tried to put a Sony sensor into a Canon 5d3 and make the perfect camera.
Swapping engines and transmissions in cars is not uncommon, so why not swap sensors in cameras?

I don't think it's a simple as changing a lens or a focusing screen. The sensor is the guts of the camera and it's hard wired in. I'd imagine Sony uses completely different components than Canon so you can't just stick it on and expect it to work. Then you need to programme the digic processor to work with the new sensor. So you'd need the electrical engineering skills from both Sony and Canon plus the coding skills of the ML team to stand a chance. How many attempts would it even take? Do you wanna buy a bunch of $3000 cameras just so you can play Dr Frankenstein with them?

I bet there is some lunatic out there that is probably doing this very thing as I type this!

You need to replace almost all of the electronics, design new custom chips, write a new firmware from scratch.

 

[LetTheRightLensIn]

To the OP: let's see you capture a scene that includes the actual sun disc recorded in the EV range and lift shadows from a shaded area with the A7r. This is the only situation in which my Canon gear cannot cope.

Of course you will fail. I know because I have already tried with a D800. That's why myself and hundreds, nay thousands of 'landscape' orientated photographers are still using Canon.

you mean like this one?..
www.canonrumors.com/forum/index.php?topic=8105.msg161888#msg161888
good luck using a Digic 4 or 5 Canon for a shot like the first one, you would not be able to re-tone the image to look like it does to the eye.
How does a Canon user cope with such a shot? Cry and go home? Spend many hours in post, trying to tame the stripes and fix the muddy dark tones to have some sort of semi-realistic texture? HAHAHA! Bracket?!? ;D

the last 2 in that post could even be done with a 7D, not that I'd chose one if I had an ABC alternative available.

Yeh, you can't do that with a Canon..........

That example was only like a half stop or stop lift of lower mid-tones, not near black areas.

 

[LetTheRightLensIn]

I never get noise or banding in images taken with the 5DII, not in blue skies, dark shadow - nowhere. Either you had a real dud of a camera or a real dud of a brain. Maybe both, I'm being generous.

If you've never once seen noise or banding in 5D2 shadows then you never push it or use DPP which now apparently mushes shadows to nothing, maybe not much banding or noise, but raw mush, no details at all either.

I mean come on, you can say it doesn't matter for what and how you shoot, but to try to imply that nobody ever sees banding or noise in 5D2 low ISO shadows unless they have a defective copy, come on man.

 

[LetTheRightLensIn]

@ Sporgon

it was already mentioned that DPP is better at develop cr2 raw expecially in the shadows, from this point your examples of the blue boat is very interesting even more with no noise reduction ... In the end if not better sensor Canon could realise a better/powerful software or why not help Adobe improve the CR2 raw converter ??? Their DPP is free, they do not need to protect sales in this regard, maybe it is the opposite, there are many adobe customers that could get more out of canon so no need to search elsewhere.

That is debatable. DPP shows less banding and noise in shadows now, but it also gets a very mushy, sometimes almost posterized, very fake, digital, waxy, no detail look IMO. Personally I'd call that putting lipstick on a pig.