70D and Dxomark....

[LetTheRightLensIn]

I shoot regularly with two photographers using D800's, and I've yet to see either of them produce an image I like as well as what I get from my 5dIII. ... and they can't touch me on video.

Bear in mind that Canon's FF sensors are a different universe than its crop sensors DR-wise. Before I moved to a FF camera, DR was a sore spot for me on an ongoing basis. Now, it is much less so.

I don't know that I'd really say that. All of the Canon cameras have generally been stuck at a pretty similar DR, yeah some difference for FF but not a lot, for low ISO although a few models do better for usable DR to a somewhat larger degree (1Ds3,6D,1DX but not 5D2 or 5D3 or 5D, etc.) since they have a touch more engineering DR and are less banding prone. It's the higher tone SNR where the more universal differences of a larger degree have been or DR at very high ISO between aps-c and FF.

My old XTi wasn't even close to that ideal, and from what I've read, the 60D and 70D are only slightly better. My 6D gets me most of the way there. YMMV.

The 6D is the first model they have made that almost got back to old 1Ds3 performance levels for low ISO DR. It does do better than all the rebel, xxD,7D,5D3,5D2,etc. stuff. That said the difference, and note that you already easily noticed it, between the 6D low ISO DR and the XTi DR is a fair amount less than the difference between the 6D and a D600. It is a shame that they didn't manage to give the 5D2/5D3 the 1Ds3/6D DR usability at least. For some reason after the 1Ds3 and 40D they went backwards in terms of pattern banding for a bunch of years and only got back to where they had been again with the 6D.

 

[LetTheRightLensIn]

So perhaps Nikon gives some extra room for lifting the shadows. Good to have when you need it. Perhaps Canon gives some extra room for pulling the highlights. Also good to have. Not a big deal either way, and neither prevents a photographer from making great photos.

These are linear capture sensors so there is no such thing as giving more room for pulling highlights for the most part (CFA differences might slightly change the degree at which a channel blows out relative to another) so you are mostly just getting confused by metering differences and taking them as sensor differences when it comes to highlights. If you properly expose Nikon shots to save the exact amount of highlights as a Canon shot then you will have the same highlights but the lowest tones will be a lot cleaner on the Nikon.

 

[Sporgon]

I expect to be as good (as far as DR is concerned) as Nikon, Sony, Pentax, Fuji, Oly, Panasonic.... Did I miss something? Ideal - no. I do not want anything better than any other brand can do, in terms of DR.

Just out of curiosity; why did you choose Canon ?

 

[LetTheRightLensIn]

If the next round of cameras don't deliver then it's not like you'll just have been stuck with old tech for DR for a few months or a few years but for more than a decade. Just think of how much you have shot how many places you have been to over a decade. Who wouldn't want the option to have had expanded photographic possibilities for a whole decade?

Yes, I would want the perfect camera that gives me the option to do anything and everything with any image in any situation. So I'm 100% in favor of technical advancements in sensors and everything else about cameras. But I'm not going to be perpetually disappointed when each camera is introduced and somehow falls short of that ideal goal. Cameras today are capable of doing a lot.

Sometimes it almost feels like that so long some rabidly defensive users shout down anyone everywhere who asks for more DR and manages to portrays those who want more as just a few random nutters or incompetent photographers going on about nothing that matters enough to bother caring a whit about, Canon will never bother to deliver more DR (unless maybe DxO alone starts blabbing and raising such an embarrassing ruckus that they feel like they are loosing to much face).

People raised a ruckus for years (and D300 existed) and then eventually the 7D. People raised a ruckus about AF (and D700 existed) and then finally the 5D3. People raised a ruckus about manual video control and then the new firmware (and the fact that it took such an insane level of ruckus from even the biggest names and players in that case was kind of scary considering it seemed mind boggling for them to not have thought to have had that in their to begin with and how resistant they were to admit they should have). We had the D800/D7100/D7000/D600/etc.etc.etc. now we just need the DR ruckus to be let to live and maybe before another decade is out we get more DR (also for something like DR you need to start the ruckus early since stuff like that takes more time to deliver, which is why some were harping on it already half a decade ago).

 

[Aglet]

I've finally posted a screenshot of the pushed ISO levels from 100 to 12,800.

www.canonrumors.com/forum/index.php?topic=16713.0

70D's looking better than previous generations.

Will add full-size crops later.

 

[Pi]

So perhaps Nikon gives some extra room for lifting the shadows. Good to have when you need it. Perhaps Canon gives some extra room for pulling the highlights. Also good to have. Not a big deal either way, and neither prevents a photographer from making great photos.

These are linear capture sensors so there is no such thing as giving more room for pulling highlights for the most part (CFA differences might slightly change the degree at which a channel blows out relative to another) so you are mostly just getting confused by metering differences and taking them as sensor differences when it comes to highlights. If you properly expose Nikon shots to save the exact amount of highlights as a Canon shot then you will have the same highlights but the lowest tones will be a lot cleaner on the Nikon.

Also, the D600 can capture about 15% or so more photons than the 6D (sensorgen), which makes it less likely to get overexposed with the same amount of light.

 

[zlatko]

canon has no more room for pulling in high lights than nikon or other sensors
if you think so that is a myth, you count the stops from middle grey and up and the same exposure, time / f-stop , which also reveal the QE in the sensor= sensivity and gain

I found Nikon to more prone to overexposure, for whatever reason. Could be just the metering.

 

[zlatko]

It is a shame that they didn't manage to give the 5D2/5D3 the 1Ds3/6D DR usability at least. For some reason after the 1Ds3 and 40D they went backwards in terms of pattern banding for a bunch of years and only got back to where they had been again with the 6D.

For me, the 5D3 has better usability than the 6D, as it should for the price. It sounds like your definition of usability is very narrow if pattern banding when pushing Lightroom sliders trumps everything else.

 

[Aglet]

I've finally posted a screenshot of the pushed ISO levels from 100 to 12,800.

www.canonrumors.com/forum/index.php?topic=16713.0

70D's looking better than previous generations.

Will add full-size crops later.

what is looking better? visual results? noise reduction, less banding etc etc

less banding. Overall average noise levels are not much improved but there's now more randomness and less structure to the noise so it should be less objectionable than previous recent generations of Canon's crop sensors. This kind of noise is more readily handled by NR software and anyone having to lift dark areas should not be bothered by banding as badly as before.

100% crops are going up in a minute.

 

[dtaylor]

There is a kernel of truth in this DR "debate" but most of the "examples" posted just don't stand up and do not actually illustrate the "problem" at all, that is why there is such a staunch defense of Canon gear from some people here. You think your sensor failed you at that concert, it didn't, you didn't educate yourself on a competent way to use your equipment.

Couldn't have said it better myself.

 

[dtaylor]

Second, you didn't ETTR, you did use +0.33 EC but that wasn't enough for an optimal RAW exposure.

Are you kidding me? This is a live subject. What ETTR?

What is: exposure compensation?

Third, your post processing exposes the fact that you just don't know what you are doing. Try this, I just did to your jpeg file. Exposure up 1.35 (that is where you should have been exposing by the way) Highlights down -37. That is how to post process your image, you have detail where you wanted it, but you have zero noise and banding even at 300% (with zero noise reduction).

The guy would not stay still, you know.

What does that have to do with the instructions on how to best post process this image? Did you even read what he said? Do you think he's telling you to shoot multiple exposures? (Hint: ETTR means Expose To The Right.) I just tried those two steps on your unprocessed JPEG and there's zero noise or banding on a properly calibrated monitor. (Key point I'm going to address in a moment.)

Fourth, even at 100% and badly processed the banding is not serious enough to destroy the image, a little noise reduction and it is gone anyway, try +35 NR in Lightroom and it disappears.

100% chroma NR does not put a dent on it. 80% lumina hides most of it. 35% lumina hides the random noise but leaves the pattern noise in place.

I can confirm that in ACR8.1 +35 NR eliminates the noise and banding, again on a properly calibrated monitor.

Why do I keep throwing in that phrase? Because if you shove the brightness on a monitor up (or stand up and stare down at an extreme angle) you can see some noise again. That's not how a print will look. And that's not how it will look on any normal monitor. But I've seen monitors with brightness / contrast / saturation shoved up so high that it's like adding 3EV to a photo while pushing the colors into neon. Since we're actually picking apart an example with instructions it's critical to view the results properly. If your monitor is not calibrated...and that includes more steps then just using the color pucks...fix it.

Nonsense. But even if I was guilty of not exposing correctly ("Hey, you, would you stay still for a moment? Thank you!") so what?

You're not exposing or processing correctly and you're not listening. The instructions had nothing to do with multiple exposures. Please re-read the post.

I went one step further in that I used the instructions with the unprocessed file and added 35NR and 50% sharpening to clear any noise but retain detail in the face, hands, etc. Guess what? There's not really any noise to see even if I shove the brightness up on my monitor. Even your poorly processed example would not show any noise in print. The properly processed one doesn't show any noise at all.

Now do you understand why we're sick of hearing about Sony sensors? Yes, they have cleaner shadows. Yes, that means they can take more abuse. But it's difficult to even imagine an edge case where it matters. To come up with such edge cases you have to imagine the absurd, like shooting with a broken flash. (You don't realize the flash isn't firing? You don't have a spare? You can't crank up the ISO when you realize...on the first shot...hey the flash is broke?)

Yet the people who push this in canonrumors (and I don't see this come up nearly as often any where else) act like the difference in shadow noise and pushed DR is literally the end of the world. Throw your Canon cameras away because it's just impossible to work without a "modern sensor."

That phrase also ticks me off. We all know that the difference in shadow noise comes down to a Sony patent. That doesn't mean Canon sensors aren't "modern." The sensor in my M is cleaner then the sensor in my 7D, and the 70D sensor is cleaner still. And I would say the 6D has a 1 stop high ISO noise advantage over the D600 with a normal exposure (as opposed to a +5EV no NR torture test zoomed 200% into a shadow). But somehow these sensors and improvements are not "modern"?

Whatever. Learn how to process your files, then maybe complain about a sensor.

 

[dtaylor]

This is completely irrelevant. Did you even pay attention what post I replied to? Somebody posted a high ISO image, processed who knows how, and asked "where is the banding"? I posted a real world image, not pushed by 4 stops which shows banding. I could have posted a shot of an empty stage to make that point.

The banding is only evident after sub-optimal processing. And even then only barely so at 100% on a monitor. It would never show in print or at normal viewing sizes.

BTW, are you sure you understand what ETTR means? In light limited situations like this, the only benefit of ETTR may have is ... well, to compensate for the deficiency of the sensor (with minor, probably invisible tonality gain which diminishes greatly for low light)

I will concede that in this situation ETTR is not nearly as important as properly processing the file. And that it would be difficult to do consistently with rapidly changing lights.

But it's trivial to produce a version of your file with no apparent noise or banding.

Why would you "process with competence" shots from a sensor which has no problems with the shadows? Shouldn't t the banding be not there in the first place? You are missing the point again. This is not a thread about pp techniques.

Why would you "process with competence" shots from a sensor which has no problems with color? Shouldn't the color be pleasing in the first place? You are missing the point again. This is not a thread about pp techniques. When will Sony/Nikon have "modern sensors"? 8)

 

[dtaylor]

You do not "improve" on the RAW capture, you convert it.

Then improve your conversions.

BTW, this image in my computer is with black background. I am not a fan of pattern noise, and yes, I can see it even without 100% zoom. It was an example meant to counter the funny talk about 4 stop push, and "what, 4 stops only?" remarks, etc.

If you can see it on your monitor at <100% then you are pushing 4 stops, you're just doing most of the push with the brightness control on your monitor. Calibrate your monitor. I'm picky about this because I actually make prints, so my monitor is as close as it can possibly be to Epson 3880 prints under good viewing lights. And I have to close the blinds in my office to clearly see the banding you're complaining about even at 100%.

 

[dtaylor]

The thing is Canon very much seems to have become a company that simply won't give about anything unless forced to,

Stop with the imaginary drama. You know...or should know for all you talk about DR...that the shadow noise issue traces back to a Sony patent. Canon can't do the same thing Sony can in designing their sensors. Do you expect Canon to close their fabs and become dependent on Sony over this issue?

so until everyone makes a big stink they will just toss us old, old tech sensors when it comes to low ISO performance for who knows how many more years or half decades.

Patents last 20 years. I don't know if Canon will be able to work around the patent, or if some other technology will make it moot, in less time. Probably the latter, but who knows when.

If the next round of cameras don't deliver then it's not like you'll just have been stuck with old tech for DR for a few months or a few years but for more than a decade.

Nice hyperbole, but in 2003-2005 Canon sensors were measurably better in all respects. (And yet I still told my Nikon lens owning friend to buy a Nikon DSLR. Imagine that!)

Just think of how much you have shot how many places you have been to over a decade. Who wouldn't want the option to have had expanded photographic possibilities for a whole decade?

Poorly processed concert photos and brick walls =/= "expanded photographic possibilities."

 

[dgatwood]

The thing is Canon very much seems to have become a company that simply won't give about anything unless forced to,

Stop with the imaginary drama. You know...or should know for all you talk about DR...that the shadow noise issue traces back to a Sony patent. Canon can't do the same thing Sony can in designing their sensors. Do you expect Canon to close their fabs and become dependent on Sony over this issue?

Any idea what patent number this is? I can't imagine Sony having a patent on something as obvious as the mere concept of consolidating the preamps and ADCs onto the sensor silicon (or at least not a valid patent that would hold up in court), so there must be a lot more detail to Sony's patent that makes it patentable. As such, there must be many, many other ways to perform such consolidation without violating Sony's patent.

If there's no way to work around the patent, that almost always means that the patent is too broad, and thus invalid. In that case, the first thing Canon's legal department should do is threaten to sue to invalidate the patent on obviousness grounds, and use that threat as a bargaining chip to get a dirt-cheap patent license.

 

[Pi]

I just tried those two steps on your unprocessed JPEG and there's zero noise or banding on a properly calibrated monitor. (Key point I'm going to address in a moment.)

That is brilliant. I cut the shadow detail/color that I wanted to keep and converted it to JPEG. Then you tell me - see - no problem, I got rid of it. How come I never thought of that? Wait, I did. What you posted is not my processing - it was done to show what I am missing and why I did not process it that way.

I can confirm that in ACR8.1 +35 NR eliminates the noise and banding, again on a properly calibrated monitor.

You cannot confirm anything because you do not have the RAW file. The values I mentioned are in the RAW converter.

Why do I keep throwing in that phrase? Because if you shove the brightness on a monitor up (or stand up and stare down at an extreme angle) you can see some noise again.

That was a wild but wrong guess

That's not how a print will look. And that's not how it will look on any normal monitor. But I've seen monitors with brightness / contrast / saturation shoved up so high that it's like adding 3EV to a photo while pushing the colors into neon. Since we're actually picking apart an example with instructions it's critical to view the results properly. If your monitor is not calibrated...and that includes more steps then just using the color pucks...fix it.

It is. A decent NEC, well calibrated, the whole flow color managed.

I went one step further in that I used the instructions with the unprocessed file and added 35NR and 50% sharpening to clear any noise but retain detail in the face, hands, etc. Guess what? There's not really any noise to see even if I shove the brightness up on my monitor.

Why would you add 35% NR? This is an image with some bright spots almost blown. It is nothing like 4 stops underexposed image. Why would you have to apply 35%? Are you saying that noise can show up even if you do not underexpose by 4 stops? That would be a big news.

Even your poorly processed example would not show any noise in print. The properly processed one doesn't show any noise at all.

What print? Do you really believe that I would print that?

Now do you understand why we're sick of hearing about Sony sensors? Yes, they have cleaner shadows. Yes, that means they can take more abuse. But it's difficult to even imagine an edge case where it matters. To come up with such edge cases you have to imagine the absurd, like shooting with a broken flash. (You don't realize the flash isn't firing? You don't have a spare? You can't crank up the ISO when you realize...on the first shot...hey the flash is broke?)

I see, my flash did not fire. I get it now.

Whatever. Learn how to process your files, then maybe complain about a sensor.

Learn to read. This image was processed for the purpose of posting it here to bust the myth that higher DR is useful only when you push by 4 stops. Why don't you just state the obvious without teaching me how to process my images. You have no idea of how I actually did process this image.

 

[LetTheRightLensIn]

The thing is Canon very much seems to have become a company that simply won't give about anything unless forced to,

Stop with the imaginary drama. You know...or should know for all you talk about DR...that the shadow noise issue traces back to a Sony patent. Canon can't do the same thing Sony can in designing their sensors. Do you expect Canon to close their fabs and become dependent on Sony over this issue?

so until everyone makes a big stink they will just toss us old, old tech sensors when it comes to low ISO performance for who knows how many more years or half decades.

Patents last 20 years. I don't know if Canon will be able to work around the patent, or if some other technology will make it moot, in less time. Probably the latter, but who knows when.

Maybe if CR had posted the DR patent part of Ankorwatt's find and had put it on the front page along with the security sensor whatever that they did put on the front page (as they do with basically every other patent under the sun that someone comes across) you'd know that Canon has high DR patents, including ones using column ADC, a that a new one was just filed relatively recently, Ankorwatt brought it to light, and then others brought to light that they had actually had others for column ADC varients in 2007 and 2006 too and it has even been said that they didn't care enough to file one for a completely different potential method to increase DR back earlier on too (although maybe something was lost in translation in the telling of that story).

The problem is that with most of those patents, and certainly not with the latest one, they can't make use of any of it on their ancient 500nm process (where they make all their DSLR sensors). Other makers invested in new advanced fabs or moved DSLR production to advanced fabs that can handle more advanced sensor designs (not the that the sensor sensors themselves are bad from Canon, they are among or even the best, but the associated electronics parts of their sensors are very out of date so they are tossing out many stops of performance at low ISO that the sensor sensor actually has). Canon has invested as much in it and what they do have that is more advanced they keep reserved for P&S sensors.

Also even with old style tech, Nikon managed to do half way to Exmor in improvements, nothing patented there at all, and Canon stayed the same.

It was also said that they DSLR division in Japan was sent something to patent for improving DR that was supposed to be able to work on their current fabs but marketers/MBAs at DSLR division were said to have just tossed it off and were supposedly like who cares, shocking the other Canon division. It never got filed much less used, who knows what ever happened to it. Although maybe something got lost in translation between the two Canon divisions in the story that was told was not as absurd sounding for real.

If the next round of cameras don't deliver then it's not like you'll just have been stuck with old tech for DR for a few months or a few years but for more than a decade.

Nice hyperbole, but in 2003-2005 Canon sensors were measurably better in all respects. (And yet I still told my Nikon lens owning friend to buy a Nikon DSLR. Imagine that!)

[/quote]

What hyperbole? If the next round in 2014 doesn't deliver will will have had old tech for DR for over a decade, remember that the round after the next round doesn't happen the day after the next round is released but another good three years later on top.

Poorly processed concert photos and brick walls =/= "expanded photographic possibilities."

RIiiight sure because all we want is expanded possibilities to shoot brick walls. Stop making nonsense up.

 

[LetTheRightLensIn]

It is a shame that they didn't manage to give the 5D2/5D3 the 1Ds3/6D DR usability at least. For some reason after the 1Ds3 and 40D they went backwards in terms of pattern banding for a bunch of years and only got back to where they had been again with the 6D.

For me, the 5D3 has better usability than the 6D, as it should for the price. It sounds like your definition of usability is very narrow if pattern banding when pushing Lightroom sliders trumps everything else.

Wow where the did heck did I ever say that the 6D has better usability than the 5D3 or that the 6D trumps the 5D3?!!!? I said I sure wish they had given the 5D3 the better resistance to pattern banding that they gave the 6D and the slightly lower read noise as well. 5D3 has vastly better video, better AF, better controls, better fps, better reaction time, etc.

It's pretty easy to make us all seem silly when make stuff up and keep going around saying that all we care about it shooting brick walls, or underexposing every shot by 5 stops, or claiming that 6D is easily the most usable and best Canon body. I'm sure I could make you sound silly if I asked you to stop saying that the 75-300 IS has far superior AF and image quality than the 300 2.8 IS II couldn't I? So stop saying that. Don't be silly and say such things. You keep saying the 300 2.8 IS II is the worst lens every made. Why do you keep saying that?

If I was a 100% pure landscape shooter I probably would sell my 5D3 and get some cash back and use the 6D for Canon shooting. But I shoot all sorts of things so no I did not trade in my 5D3 for 6D.

 

[Pi]

Another one of those images "underexposed by 4 stops and pushed back". The highlights are already blown and the shadows are noisy as hell. Click on the image below.

I know, I know, I can hide the evidence. 75% luminance NR, and the noise is gone, well, most of it.

 

[Aglet]

Another one of those images "underexposed by 4 stops and pushed back". The highlights are already blown and the shadows are noisy as hell. Click on the image below.

I know, I know, I can hide the evidence. 75% luminance NR, and the noise is gone, well, most of it.

That IS a bit noisy.
I haven't paid attention to this place for a while, what are you shooting with?
What ISO were these outdoor crowd shots taken at?

And, if you think THIS is bad..
I loathed my early model 5d2 at times for how noisy it was - so I sold it.
The day I sold it, I quickly shot some 1 EV steps of smooth gray background at 100 ISO in studio.
Pushing up shades lower than metered 0 EV is an easy way to show noise.
BUT - I took a +3 EV shot, lowered it 2 EV and, if you enhance the contrast, can STILL see FPN!
I can see FPN at 0 EV, I can see FPN at EVERY EV if I move the region back to near 0 EV (even using DPP)

If there were smooth tones anywhere near midtone in an image it was possible to find V & H patterns unless using a lot of NR which then smeared luminance details and made it pointless. At least the heaps of chroma noise could be dealt with.

FWIW, IMO, my 5d2 was a PoS! most disappointing camera I've ever owned, still managed some great shots with it, nonetheless. Can mask some of the noise when printing large by using rough or textured papers or canvas.
I could likely tolerate a 6D or 70D, they've improved enough.