DxOmark Canon 5DIII

[Slobodan Blagojevic]

Simply not true. What might be true is to say that the 5D MK III was never going to compete in DxO's game.

But the fact remains - and fact is the word - that the results people are getting from the Canon are fantastic, regardless of what that bloody website might have to say about it...

Thanks for the good chuckle this Saturday morning! The best illustration for sour grapes I've seen lately.

[MrSmith]

Compared to the D800, the 5D3, particularly in the shadows, is a noisy piece of crap.

so is a MFD back at 800asa, so is a D800 at 3200asa and especially if you open them up by 2 stops.
i guess if you like to shoot lots of shadows and lift the exposure 2 stops in processing the 5DIII isn't for you and you may need to source another manufacturers camera to suit your needs and desires.

if anyone wants to offload their obviously flawed and limited MkIII then get in touch

[Sheldon N]

Seen this? http://www.fredmiranda.com/5DIII-D800/index_controlled-tests.html

The DxOMark scores are reflected in reality. Compared to the D800, the 5D3, particularly in the shadows, is a noisy piece of crap.

Sad but true. The difference is quite stunning.

I hope the 1D X is slightly better, but somehow I doubt that it will be.

[marcmccalmont]

the reason i thought the MkIII had less noise at 160/320/640 was previous experience with the II where the files are cleaner at these iso's and i only ever move between these iso's so going from 160-400 would be worse than going 160-640.
and from this graph http://home.comcast.net/~NikonD70/Charts/RN_ADU.htm#EOS%205D%20Mark%20III_14

maybe i am interpreting the graph wrongly (i'm a humble photographer not an academic) but the results of using the 5dII daily for the last few years have shown better results at these iso's. i believe the moving image MkII users have formed similar conclusions

Your reasoning is correct if you expose for the scene but if you ETTR then you are better off with 100, 200 etc, The graphs are misleading
Marc

[bclaff]

from this graph http://home.comcast.net/~NikonD70/Charts/RN_ADU.htm#EOS%205D%20Mark%20III_14

maybe i am interpreting the graph wrongly (i'm a humble photographer not an academic) but the results of using the 5dII daily for the last few years have shown better results at these iso's. i believe the moving image MkII users have formed similar conclusions

My graph does not distinguish between gain that is entirely analog and gain that is partially or entirely digital.
As Emil points points out, certain Canon cameras do digital scaling to achieve intermediate ISO values.
This type of graph helps identify this situation.
Since the values have not been converted to electrons (something I'm working on) you cannot see in this chart that the intermediate ISO values have the same read noise in electrons as the full ISO values.
In other words, ISO 160, ISO 200, and ISO 250 are the same value in electrons they are just reported differently in ADUs.

Regards,
Bill
P.S. - I now have a preliminary release of the appropriate chart:
Read Noise in Electrons (with 5D3 preselected)

[Guillermo Luijk]

regardless of what that bloody website might have to say about it...

Why bloody? it's because you don't like what they have to say about your favourite brand?

[marcmccalmont]

i guess if you like to shoot lots of shadows and lift the exposure 2 stops in processing the 5DIII isn't for you and you may need to source another manufacturers camera to suit your needs and desires

Yes I do raise my shadows 2 stops a lot to bring out details and yes I will spend $3000 to do this instead of $3500 not to do this!
Marc

[uaiomex]

What Canon did, was to recycle the 5D2 with a pro-grade AF and charge you $1,300 usd for it.

Canon  photo equipment is not made at a factory anymore. That's the problem. It is made at a farm. 

Eduardo

Why bloody? it's because you don't like what they have to say about your favourite brand?

[BernardLanguillier]

It all comes down to the optics in the future, I feel that Nikon has a edge in the wide and Canon in the longer end. As a photographer, I use the longer optics more and it makes sense for me to stay with Canon.

I totally agree that switching brand makes little sense, but for what it is worth the current Nikon line up of long primes is really outstanding. Canon may be even better, but both are probably very close.

I have been using a one gen old 300 f2.8 VR and it may be the sharpest lens I own... and it pretty much delivers that level of performance at f2.8-4. Add to that very fast AF, excellent VR silky smooth bokeh and I have a very hard time understanding how any lens could be significantly better.

There are few factual tests available out there for those super lenses. The only one I could find is the 200f2:

http://www.photozone.de/nikon_ff/499-nikkorafs200f2vrff?start=2

It pretty much speaks for itself.

Cheers,
Bernard

[MrSmith]

Your reasoning is correct if you expose for the scene

years of shooting transparency and a modified zone system/spot meter use means i'm used to 'exposing for the scene' 

[Ray]

As a photographer who has used only Canon equipment in the digital age until recently, when I started buying Nikon equipment, I can sympathise with those who might have the dilemma of whether or not to switch brands.

Lenses are just as important as camera bodies. You need both to take a photo. If one has accummulated a few good Canon lenses over the years, it doesn't make much sense to start all over for the sake of an improved sensor in a couple of camera bodies.

However, if Nikon has a particular lens or two with more desirable characteristics than any Canon lens the photographer currently owns, and the photographer knows that such lenses are of a focal length that he will use a lot, then it might make sense to either switch brands or use both brands.

This was basically my situation. I wanted a good wide-angle zoom, and the Nikkor 14-24/2.8 seemed to be the best available. I bought my Nikon/Canon adapter first, then the lens which I used on my 5D for a while. Unfortunately, the adapter was not fully functional. There were too many annoying quirks, one of which was the refusal of the Canon 5D body to fully respond to the 'off' button when the lens was attached. Consequently, if either the lens or the battery was not removed after using the camera, the battery would go flat in a couple of days.

As a result of a certain degree of frustration with such quirks, I bought the D700 which I considered a worthwhile improvement in its own right, compared with my 5D. I had two reasons to switch to Nikon, but I continued to use my Canon 50D whenever I needed longer focal lengths than 24mm. I still use the 50D with my 100-400mm IS zoom, although I now also have a D7000 and another Nikkor lens. The extra weight of the 50D body seems fairly trivial compared with the weight of the 100-400, and the Nikon equivalent to the Canon 100-400 does not appear to be better, and perhaps even not as good.

I'm stuck with two systems, but don't think that I'm complaining. 

[EgillBjarki]

I totally agree that switching brand makes little sense, but for what it is worth the current Nikon line up of long primes is really outstanding. Canon may be even better, but both are probably very close.

I have been using a one gen old 300 f2.8 VR and it may be the sharpest lens I own... and it pretty much delivers that level of performance at f2.8-4. Add to that very fast AF, excellent VR silky smooth bokeh and I have a very hard time understanding how any lens could be significantly better.

There are few factual tests available out there for those super lenses. The only one I could find is the 200f2:

http://www.photozone.de/nikon_ff/499-nikkorafs200f2vrff?start=2

It pretty much speaks for itself.

Cheers,
Bernard

The results on the samples and MTF are very impressive indeed! I guess in some instances Nikon's longer lenses are top level. The same could be argued with Canon's recent 24mm and 17mm tilt and shifts, that they are catching up in the wide end. I guess this comes down to the workflow and different needs of each photographer.

But judging from the review from Fred Miranda, it seems clear that the D800 has a big advantage in recovering shadow details. I can see that I need to make multiple exposures with my 5DIII to compete with the DR in D800.

[hjulenissen]

Seen this? http://www.fredmiranda.com/5DIII-D800/index_controlled-tests.html

The DxOMark scores are reflected in reality. Compared to the D800, the 5D3, particularly in the shadows, is a noisy piece of crap.

I am a bit surprised by this comment in the text. I'd have thougth that the trade-offs involved in choice of exposure was well known and debated?
"I know this is disappointing for Canon shooters but on the bright side, I figured out a workaround if you shoot RAW. Start by overexposing by about 1 stop above the correct exposure before taking your shot and then underexpose by the same amount later in software, basically normalizing the image. For example, in Lightroom, you would move the exposure slider to -1. This gives you the correct exposure but the shadow detail is much cleaner just in case you need to push it a stop or two. In other words, using this technique banding and noise subsides considerably."

[BJL]

I am a bit surprised by this comment in the text. I'd have thougth that the trade-offs involved in choice of exposure was well known and debated?
"I know this is disappointing for Canon shooters but on the bright side, I figured out a workaround if you shoot RAW. Start by overexposing by about 1 stop above the correct exposure before taking your shot and then underexpose by the same amount later in software ...

Yes, it seems almost defensive to suggest this as a solution. Apart from being useful only when the subject brightness range from metered midtones to brightest highlights is low enough, and the fact that this only shifts  the usage of the available DR rather than increasing it, this is also what is done far more easily with the 5D3's extended low ISO speed setting of 50, so there is probably no need for that fiddle.

[EDIT:] I take part of that back. With Canon's noise behavior at low to moderate EI levels, this partial implementation of ETTR could be also useful at higher shutter speeds than the "ISO 50" setting allows. For example, instead of using an ISO speed setting of 400 with normal exposure settings androcessing, one could use the same shutter speed and aperture but ISO speed setting of 800, and then lowering the levels in post. This gathers the same light, but applies twice the analog gain, helping to lift the shadows further relative to the noise that enters after that gain is applied. So long as no highlights get amplified up to the point of clipping, one does not in any practical sense lose an dynamic range. What one is doing is in effect producing a "high key" raw file. [END EDIT]

[shadowblade]

I am a bit surprised by this comment in the text. I'd have thougth that the trade-offs involved in choice of exposure was well known and debated?
"I know this is disappointing for Canon shooters but on the bright side, I figured out a workaround if you shoot RAW. Start by overexposing by about 1 stop above the correct exposure before taking your shot and then underexpose by the same amount later in software, basically normalizing the image. For example, in Lightroom, you would move the exposure slider to -1. This gives you the correct exposure but the shadow detail is much cleaner just in case you need to push it a stop or two. In other words, using this technique banding and noise subsides considerably."

This is a well-known method (exposing to the right) but actually *decreases* your useable dynamic range - you can't do it when the highlights are already close to maximum, while areas of deep shadow still remain.

[ejmartin]

the reason i thought the MkIII had less noise at 160/320/640 was previous experience with the II where the files are cleaner at these iso's and i only ever move between these iso's so going from 160-400 would be worse than going 160-640.
and from this graph http://home.comcast.net/~NikonD70/Charts/RN_ADU.htm#EOS%205D%20Mark%20III_14

maybe i am interpreting the graph wrongly (i'm a humble photographer not an academic) but the results of using the 5dII daily for the last few years have shown better results at these iso's. i believe the moving image MkII users have formed similar conclusions

The proper interpretation of the graph comes with the additional knowledge that ISO 160/320/640 etc are obtained by pulling the data 1/3 stop from 200/400/800 etc, after digitization, and 125/250/500 etc by pushing it 1/3 stop from 100/200/400 etc.  Clearly you don't change the signal-to-noise ratio when you multiply both by the same factor, which is what you do when you pull/push an exposure, and so 160/320/640 etc do not have better noise characteristics (more precisely, better S/N) than 200/400/800 etc if you shoot both at a fixed exposure (shutter speed and aperture).  By looking only at read noise, one is only looking at half the equation; noise gains relevance in the context of a signal (exposure to light), and one needs to know how changing the ISO will affect the noise relative to the signal.

What is different is the metering.  160 meters for 1/3 stop more exposure than 200, so if you are exposing according to the meter's suggestion, then at 160 you are getting the exact same file as if you had shot at 200 and dialed in 1/3 stop more exposure than the meter tells you.  And so if you are using the camera's meter to set your exposure, then you are getting a better file at 160 than at 200, because the metering gives you 1/3 stop more signal (exposure) for the same noise, and therefore better S/N.  What the meter is not telling you however, is that you are losing 1/3 stop of highlight headroom in the process, as with any ETTR.  Personally, with my Canons, I set the custom function that restricts ISO's to only integer stop increments, and dial in EC when the situation calls for it, so that I am quite conscious of where my exposure is relative to clipping.

[BJL]

... ISO 160/320/640 etc are obtained by pulling the data 1/3 stop from 200/400/800 etc, after digitization ...

What is different is the metering.  160 meters for 1/3 stop more exposure than 200, so if you are exposing according to the meter's suggestion, then at 160 you are getting the exact same file as if you had shot at 200 and dialed in 1/3 stop more exposure than the meter tells you.  ... you are losing 1/3 stop of highlight headroom in the process, as with any ETTR.

So these are doing what Fred Miranda proposes in the quote above, except with just a 1/3 stop pull instead of a full stop, and with the shift back down after ADC done automatically in the camera. I appreciate your preference for total control and understanding of the process, but I can also see why some people would prefer the convenience of both this and the "ISO 50" pull processing setting over having to manually correct later.

[LKaven]

So these are doing what Fred Miranda proposes in the quote above, except with just a 1/3 stop pull instead of a full stop, and with the shift back down after ADC done automatically in the camera. I appreciate your preference for total control and understanding of the process, but I can also see why some people would prefer the convenience of both this and the "ISO 50" pull processing setting over having to manually correct later.

The entire business of digital pushmi-pulyu gain could be handled as metadata tags -- regardless of how one prefers to meter -- and should be.  There's no reason to go throwing data away before it ever gets written to the raw file.

[BernardLanguillier]

This is a well-known method (exposing to the right) but actually *decreases* your useable dynamic range - you can't do it when the highlights are already close to maximum, while areas of deep shadow still remain.

Correct ETTR should never decrease your usable DR.

Cheers,
Bernard

[madmanchan]

The entire business of digital pushmi-pulyu gain could be handled as metadata tags -- regardless of how one prefers to meter -- and should be.  There's no reason to go throwing data away before it ever gets written to the raw file.

Agreed.  This is what the DNG BaselineExposure tag is for.  A vendor can decide how "hot" or "dark" to record the raw data, and write the desired digital compensation into this tag.