Is Canon now two generations behind Nikon?

[jrista]

To be fair, I think jrista posted two reasonable examples: one was the room interior with bright window, and the other was a stream with bright sky.

No one has posted a reasonable example. That would require both sensors shot so that all other factors are equal, and RAW files provided for everyone to evaluate.

<snip>

jrista doesn't actually know this. Neither do you or I. You have to actually test both at the same scene.

I'll grant that it hasn't met scientific standards, and I'd like to see the side-by-side you describe. My opinion that jrista's examples were reasonable was based on two criteria: my personal experience with my 60D (I'm aware it's not Canon's best) and jrista's history of being careful about his assertions. I.e., he's earned the benefit of my initial trust (as if he cares), though I would be pleased to see scientifically valid tests to support the assertion.

Thanks for the vote of confidence.

I would honestly post direct comparisons if I had the ability to. I think I'm going to rent a D800 next chance I get (I'm just coming off a week vacation, and I hadn't planned on renting a D800 or D810 before hand...next vacation I will.) There ARE D800 raw files available on the net that demonstrate the differences fairly well. I don't think I've really encountered all that many that truly show of the true shadow lifting power of having two additional stops of dynamic range well enough...and even if I did manage to find some NEFs that did, they would of course be suspect.

The only way the issue can reasonably be settled is if I do as you say...take some comparison images myself in identical scenarios, be open about how I did my testing, and freely provide my RAW files (which I have done on several occasions, including recently with the sunflower RAW.) I have never expected miracles, but in my experience with D800 NEFs, you don't "just" have more editing latitude...the data is very clean, the tonality and color fidelity, is better than in any Canon CR2 file that deep into the shadows, and even if you don't like the lower contrast that heavy shadow lifting often results in, you still get better, cleaner, softer, smoother falloff into the shadows than you do with a Canon file.

I'd really love to have that kind of deep shadow image quality in my Canon RAWs...

 

[V8Beast]

Yes, it's not the be all and end all, but I can guarantee when Canon match the Exmors, the tone in here will change as people start to see the benefits across the ISO range, not just at low ISO.

That always happens, but that's the case in both camps. Canon fanboys subscribed to the resolution/IQ school of thought during the 5D2/1Ds3 era, then converted to ISO/FPS school of thought during the 5D3/1DX era. Nikon fanboys subscribed to the ISO/FPS school of thought during the D700/D3 era, then converted to the resolution/IQ school of thought during the D800/D4 era. Both camps are equally guilty of fanboyism

 

[jakeymate]

I'd really love to have that kind of deep shadow image quality in my Canon RAWs...

Jon, for what it's worth, here's the first quick and dirty test I did when my D800 arrived, alongside my Mk3.

27th June 2012. I'd read about the D800, but was a Canon fan, but thought it worth buying one to test out after all
I'd read, even back then when it was new.

I got the only D800 available in Aus at the time as they were in very short supply, so I figured I'd just Ebay it and the 24-70 F2.8 it came with after I'd tested it, and go back to my Mk3.

In all honesty, I never ever expected to leave Canon, but in around 20 mins of taking shots on both, my mind was made up and I decided to become one those people that switched and forever got crap for it

I had one FB photographer friend (a pro from WA) tear me a new arse on Facebook and de friend me when I posted these images one night!

Over two years later, I still am getting torn a new arse.

This wasn't my intended test anyway. I just grabbed a Diet Coke carton and put it on a smooth grey background and took two test snaps to match exposure between the two.

That was all this was for, to make sure they both were exposed the same before I started doing tests, so I came upstairs loaded the RAWs and noticed the difference straight away.

Mk3 pushed 3 stops in ACR

D800 pushed 3 stops in ACR

Mk3 pushed 3 stops in ACR 100% crop

D800 pushed 3 stops in ACR 100 crop

My thoughts then, as now, were the smoothness of the bg gradation on the D800, the lack of read noise, the rendition of colour pulled from deep shadows, and the detail retained in the rosette print pattern on the box.

The raws are here if you want them.

http://www.deanagar.com.au/raws.zip

Please don't link these anywhere but here, I don't want my website bandwidth using up in 24 hours

 

[jakeymate]

To offset the 'why do you have to push exposure 3 stops, only bad photographers get exposure wrong' posts, here is the red channel of both, NOT pushed, just at 0 adjustment.

That noise is in your channels, even you don't push them at all.

Do I even need to label these?

Mk3 red channel

D800 red channel

 

[jrista]

To offset the 'why do you have to push exposure 3 stops, only bad photographers get exposure wrong' posts, here is the red channel of both, NOT pushed, just at 0 adjustment.

That noise is in your channels, even you don't push them at all.

Do I even need to label these?

Mk3 red channel

D800 red channel

Yeah, that's pretty much my experience with D800 NEFs. That shadow tonality is just phenomenal. THAT is what I want. It doesn't matter if you lift by stops...it's just...beautiful. NO BANDING! I really, really, cannot wait until Canon figures out how to get rid of their banding noise...because it's just the most hideous thing I've ever seen.

 

[Aglet]

Maybe someone with more technical knowledge can chime in, but my understanding is that the effects of read noise diminish as ISOs go up. I've read (can't recall where) that lifting shadows becomes roughly equal between D8x0 and 5D3 somewhere between ISO800 and ISO1600.

Again, this is not my personal experience, and I can't cite you a source. If you have access to both cameras maybe you could post side-by-side samples at ISO1600.

you can see how the d800's shadows are much cleaner than the 5d3's up to iso 800. The d800's read noise takes a sudden jump at 1600 so it's similar to the 5d3. All are pushed 4 stops so you can easily visualize the noise patterns.

http://a2bart.com/tech/allcamdknz.htm

Neuro, if you think you have a better, faster way to demonstrate read noise, spit it out.
Constantly trolling with snark about lens cap on shots for this purpose is just dumb and frankly, makes you look bad if you can't top the technique.

 

[jakeymate]

Yeah, that's pretty much my experience with D800 NEFs. That shadow tonality is just phenomenal. THAT is what I want. It doesn't matter if you lift by stops...it's just...beautiful. NO BANDING! I really, really, cannot wait until Canon figures out how to get rid of their banding noise...because it's just the most hideous thing I've ever seen.

When you actually look at it technically Jon, the Canon's give up (if you look at the RGB channels) nowhere near the shadow areas.

When you put 80% of RGB on top go each other, you hide a lot of sins in what appear to be shadows.

In my experience, the Canon sensors starting giving up detail to read noise at around 65%.

If you load up the Mk3 channel in PS and look at the point where read noise appears and becomes problematic in that BG gradation, it's starts at 64%.

That's another 36% of tonality that's essentially useless.

I had a big talk today with another photographer who came in for a session, and I was explaining to them how my style changed after going to Nikon.

I used to crush blacks in my work and expose accordingly when shooting, automatically knowing where my read noise cut off point would be.

I used to do very high contrast images, which certainly looked cool, but I did that to crush out poor shadows.

My style was defined by my kit in truth.

My work has become less and less contrasty in the past 2 years and now I favour a more matt look, because I can do that, and not worry that the lower 36% of my RGB channels are polluted and need to be hidden somehow.

The issue is, so many Canon photographers think that 64% RGB is almost black, and don't worry about it.

It's not. There's 36% more usable info in those channels on Exmor. That's not insignificant.

When the 2-3 stop advantage is touted and then dismissed by Canon users, it makes me laugh.

I've displayed an image where the Canon gives up resolving smooth detail at 64% RGB.

I'm no mathematician, but isn't that 36% more usable information?

 

[jakeymate]

Maybe someone with more technical knowledge can chime in, but my understanding is that the effects of read noise diminish as ISOs go up. I've read (can't recall where) that lifting shadows becomes roughly equal between D8x0 and 5D3 somewhere between ISO800 and ISO1600.

Again, this is not my personal experience, and I can't cite you a source. If you have access to both cameras maybe you could post side-by-side samples at ISO1600.

That is a great chart, thank you for posting.

As a D800/D810 user, I would have happily said the Mk3 was cleaner at 3200 (really just to stop the crap I get from Canon guys and throw them a bone) but that chart is very interesting.

The D800 is right there with the Mk3 up to 6400, and I'll admit, I never shoot over 1600, so I'd accepted the CR forum wisdom that the Mk3 surpassed the D800 over 3200.

Apparently not.

 

[mb66energy]

@jakeymate:
Thanx for the shoots which make both cameras IQ a little bet more comparable. I think a better comparability which excludes the different MP counts of the sensors without resizing the final image files might produce the following procedure: Two shots where the object has the same size in the FINAL NON-SCALED image. This might exclude some aliasing artifacts between object and sensor structure.

For me it shows that Sony sensors are better in smoothness (thanks to the absence of disturbing read noise) compared to Canon sensors. Every bit of (1) coming closer to the real scene and (2) enlarged editing latitude is welcome.

But your images prove to ME that both cameras are tools to get absolutely stunning photos ... if, yes if the subject matters, the photographer has a vision of light and good technical abilities to put that vision into a jpg or a print, finally.

So I will stay with my so inferior 40D, 600D and EOS M and some great lenses for a while, taking photos and enjoying the race between camera brands ... and the audiences reactions!

 

[jakeymate]

@jakeymate:
Thanx for the shoots which make both cameras IQ a little bet more comparable. I think a better comparability which excludes the different MP counts of the sensors without resizing the final image files might produce the following procedure: Two shots where the object has the same size in the FINAL NON-SCALED image. This might exclude some aliasing artifacts between object and sensor structure.

For me it shows that Sony sensors are better in smoothness (thanks to the absence of disturbing read noise) compared to Canon sensors. Every bit of (1) coming closer to the real scene and (2) enlarged editing latitude is welcome.

But your images prove to ME that both cameras are tools to get absolutely stunning photos ... if, yes if the subject matters, the photographer has a vision of light and good technical abilities to put that vision into a jpg or a print, finally.

So I will stay with my so inferior 40D, 600D and EOS M and some great lenses for a while, taking photos and enjoying the race between camera brands ... and the audiences reactions!

Why exclude the megapixel count?

The D800 manages to trounce the the Mk3 sensor in that lower 36% of tone, DESPITE it having 14 more megapixels.

The raws are there if you want to look at NON SCALED images, although the 1:1 crops are non scaled.

The fact that the Mk3 still looks so poor even taken down to 1920 wide is even more telling. If you can see it at 1920, a third of it's actual resolution, then all you get on the full res is more detailed read noise.

The lack of detail in the print pattern is nothing to do with resolution and everything to do with the Canon sensor simply not seeing the detail at all.

If it did, the lower resolution would simply mean the detail was smaller, not not there at all.

To compare res for res, take the 100% Nikon crop and take it down by 77% and technically, it should look like the Mk3.

It doesn't.

 

[mb66energy]

@jakeymate:
Thanx for the shoots which make both cameras IQ a little bet more comparable. I think a better comparability which excludes the different MP counts of the sensors without resizing the final image files might produce the following procedure: Two shots where the object has the same size in the FINAL NON-SCALED image. This might exclude some aliasing artifacts between object and sensor structure.

For me it shows that Sony sensors are better in smoothness (thanks to the absence of disturbing read noise) compared to Canon sensors. Every bit of (1) coming closer to the real scene and (2) enlarged editing latitude is welcome.

But your images prove to ME that both cameras are tools to get absolutely stunning photos ... if, yes if the subject matters, the photographer has a vision of light and good technical abilities to put that vision into a jpg or a print, finally.

So I will stay with my so inferior 40D, 600D and EOS M and some great lenses for a while, taking photos and enjoying the race between camera brands ... and the audiences reactions!

Why exclude the megapixel count?

The D800 manages to trounce the the Mk3 sensor in that lower 36% of tone, DESPITE it having 14 more megapixels.

The raws are there if you want to look at NON SCALED images, although the 1:1 crops are non scaled.

The fact that the Mk3 still looks so poor even taken down to 1920 wide is even more telling. If you can see it at 1920, a third of it's actual resolution, then all you get on the full res is more detailed read noise.

The lack of detail in the print pattern is nothing to do with resolution and everything to do with the Canon sensor simply not seeing the detail at all.

If it did, the lower resolution would simply mean the detail was smaller, not not there at all.

To compare res for res, take the 100 Nikon crop and take it down by 77% and technically, it should look like the Mk3.

It doesn't.

You argued with resolution of fine structures so I presented my idea to get that comparable.

I assumed 100% crops because that is what i usually expect for comparisons like that - my fault.

 

[jakeymate]

You argued with resolution of fine structures so I presented my idea to get that comparable.

I assumed 100% crops because that is what i usually expect for comparisons like that - my fault.

Sorry, if that's how I came across, and I can see the issue there.

The fine structure present is not due the resolution, it's the sensor resolving the shadow detail.

The Canon sensor doesn't even begin to resolve that detail as it's in shadow and it actually can't even see it.

Sure the D800 would see a bit more of it due to it's higher resolution, but it's not even visible at all in the read noise on the Canon.

 

[jakeymate]

Neuro, if you think you have a better, faster way to demonstrate read noise, spit it out.
Constantly trolling with snark about lens cap on shots for this purpose is just dumb and frankly, makes you look bad if you can't top the technique.

Is there a better way to isolate the read noise and see just the read noise?

I'm not too techy but shooting zero light would do that perfectly wouldn't it?

 

[mmenno]

The reasoning of downplaying the importance of DR to the point of 'We don't need a better DR sensor because canon doesn't have one' eludes me.
Back when canon's sensors were better at higher iso's it was all you ever heard, and now when nikon's are better at low iso's it's suddenly of no importance whatsoever..

 

[Lawliet]

Is there a better way to isolate the read noise and see just the read noise?

If you want clean data & isolated variables - no, nothing that could be done at home.
If you want to stick the fingers into your ears and hum really loud...well, you want a test that does just the opposite.

 

[jakeymate]

The reasoning of downplaying the importance of DR to the point of 'We don't need a better DR sensor because canon doesn't have one' eludes me.
Back when canon's sensors were better at higher iso's it was all you ever heard, and now when nikon's are better at low iso's it's suddenly of no importance whatsoever..

The thing that puzzles me, is that Canon's aren't the best at hi ISO either.

Right now, the A7s, the D4s, and even the 2009 D3s are pretty much the best hi ISO cameras out there.

I know quoting DXO is a crucifiable offence round here, but their first Canon in the hi ISO chart is the 1DX at number 10.

The top 9 are all Sony's Exmor sensors, except the D3s which has some mystery around who made it apparently.

So all the talk about Exmors being the best low ISO sensors as if somehow they suck at high ISO's does not seem to be borne out in the real world.

 

[Loswr]

5DIII can do 132MP/sec
D810 can do 180MP/sec

D4s can do 176 MP/sec
1D X can do 252 MP/sec

Did you have a point?

Yes, neither of the cameras that you mentioned are the 5D3 or D8x0 (which are the topic of this thread.)

Try and stay on topic next time.

So, what is the relevance of MP/sec? Besides demonstrating your ability to use a calculator, that is...

5DIII can take 6 full frame pictures per second
D810 can take 5 full frame pictures per second

Try and stay with specifications relevant to photography next time.

 

[Keith_Reeder]

To be fair, I think jrista posted two reasonable examples: one was the room interior with bright window, and the other was a stream with bright sky.

Yep - but with all due respect to him, he's not the final arbiter of what can and cannot be achieved with a Canon file - I've been a number of examples from Jon which immediately made me think "Hmmm... That's not a very good job", because, like a lot on here, he seems wedded to a single converter. Some of the stuff on his own website has made me scratch my head about the image quality he's achieving with his bird photography, too.

The simple fact (and I use the word advisedly) is this: DPP (for example) can pull clean detail out of the shadows of Canon low ISO files in a way which - blind tested (been there, done that) - isn't that far from similarly "cranked" Sonikon sensor files.

There's more than one way to skin the low ISO DR cat...

But even the examples we're discussing are - in the great scheme of things - unlikely to represent the kind of photograph which most of us are taking on a regular basis, and it's actually good design and good business for a manufacturer not to expend time, money and effort building in capability which - realistically, whether the DR whiners like it or not - is only going to be of "niche" value.

 

[Keith_Reeder]

Your words.....just about describe your rather aggressive over the top rant apparently in "defence" of Canon!

Good God - basic reading comprehension really does seem to be at a premium in this thread. Nothing I wrote "defends" Canon! It criticises the bashers, and explains why some people (to use your word, not mine), "defend" Canon, but not a single word of that post can be characterised as me "defending" Canon.

I bet Canon have never heard of you.

I can guarantee they haven't - what's your point?

Just try and be a bit more open minded and less fanatical in your next post.

You still don't get it, do you? It''s other people's lack of open mindedness, and their fanatical obsession with DR, that's the problem here.

Do try to keep up...

 

[Keith_Reeder]

The reasoning of downplaying the importance of DR to the point of 'We don't need a better DR sensor because canon doesn't have one' eludes me.

The reasoning doubtless eludes you because nobody's actually doing that.

More disingenuous "spinning" to push an agenda and score cheap points...