Michael, have I missed your 5D3 review?

[dreed]

The 5DIII now does what you wanted from the II and ignored the fact the World is moving forward with others like Nikon and Sony.

Yes, the 5DIII is what the 5DII should have been but is 3 years too late and thus 3 years behind already.

I have little faith in Canon even seeing the 800 as a wakeup call, they looked to be obsessed with what other Canon depts plan for their products. The result is they tiptoe around the fringes of greatness.

I don't think the wakeup call will come until early next year when they get to look at the sales of the 5DIII vs D800 from market analysts.

I'd be curious to know how Canon functions internally - just how much competition or cooperation is there between business units to deliver higher sales?

I will wait and see what the "X" delivers before buying into the Nikon system, but honestly getting a 800 and a few lenses is not much more of an investment than the "X".

I'm curious about the X for another reasons: to see if the sensor technology in it is any different/better than the 5DIII. If it is, then maybe I will hang around Canon for a while longer but if not, then no. The rationale here is that if it is better then Canon have better technology available and have simply skimped on what they've put in the 5DIII (!$@#^%)

...
but perhaps some of it is the opposite: fearful thinking due to doubts about where Canon is going with respects to the needs of "high quality, low ISO" photography. To which I say: just be a bit more patient, and wait till about Photokina 2014.

And in 2+ years, what do you expect Nikon, Sony and others will be delivering?
The same as today or something new, more, better?

Now if this were Canon, sure, I can imagine something similar to what we have today being delivered, but it is not.

There are now rumors of a $1500, 24MP FF, Nikon D600 due out later this year and if the 36MP sensor in the D800 is anything to go by, the sensor in the D600 is going to be a real kicker.

The problem that I see is that after being asleep at the wheel, Nikon seems to have woken up and have set out with a solid direction and purpose. Canon seems to be sitting back under a tree smoking a pipe...

[BJL]

And in 2+ years, what do you expect Nikon, Sony and others will be delivering?
The same as today or something new, more, better?

Snap! See the end of my post just before yours. My thought is than in two or three years, all major players are likely to have improvements to show ... but as technology diminishes imperfections towards negligible levels (like dark noise down to one or two electrons) gains in practical terms become ever smaller, and the lagard has more room to improve, meaning that the gap will tend to shrink. In particular, crossing the bridge to on-sensor high speed parallel ADC might be a bigger step in low noise and high dynamic range than anything that follows, and Canon has that obvious jump to make yet, while Sony, Nikon, Panasonic and probably Olympus and Samsung are already there.

There are now rumors of a $1500, 24MP FF, Nikon D600 due out later this year ...

I do not put much weight on rumors, and especially this one: neither the massive price drop after about six years of price stasis in 35mm format nor the idea of a new Nikon/Sony sensor of lower resolution than the one just launched makes sense, except as a repitition of speculations that keep recurring and keep going unfullfilled.

In particular, at the risk of repeating myself:
- lower resolution in the same format does not significantly reduce the cost of making a sensor
- the D800 in particular shows that lower resolution does not significantly improve noise or dynamic range when compared fairly as with the DX0 "print comparison"
- the main real reason for a new sensor of lower resolution is higher frame rates, as in the D4 and 1D X.
So a new cheaper lower level camera of the same format but with a new lower resolution sensor seems very unlikely.

[BernardLanguillier]

Another thing to consider is that the D800 was seemingly due out in June/July 2011.

From a sensor development standpoint at Sony/Nikon, this is a one year old technology and we can assume that R&D has been little impacted by the quake.

Canon may or may not be in a similar situation.

Cheers,
Bernard

[Ray]

Ray some good points there....

Well, thank you BJL  .

comparing at exactly equal focal length and then somewhat claiming a "victory" of higher pixel count for the camera with fewer, smaller, and in most respects lower quality pixels (ignoring the possible per pixel advantage of the larger photosites of the D800) seems a bit like choosing the rules to influence the outcome. With the bulk and cost of these cameras plus 400mm lenses, it would be little extra burden to use a TC to increase resolution from the D800 if needed, so exact equality of focal length seems overly rigid.

I wasn't claiming any victory. This is what I wrote: "It would be interesting to see a comparison between the 7D and the D800 in DX mode, both cameras using the same focal length of telephoto lens of equal quality."

I don't know what the outcome would be. I'm of the opinion that any increase in pixel count of less than 50%, is not worth getting too excited about. If one is comparing sensor performance, it's essential to have lenses of equal quality. If one is comparing lenses, it's essential to have sensors of equal quality and equal pixel count, at least if the methodology is similar to Photozone's.

Having determined, for example, that the D800 shot in DX mode is perhaps only very marginally less detailed than the 7D shot, using same focal length of lens of the same quality (whatever that focal length needs to be in order to achieve equal quality), one can then concentrate on the effects of differences in lens quality.

If one has a Canon 400mm that is actually sharper than the Nikon equivalent, then the matter is settled. The 7D with its effectively longer lens, actually greater pixel count, and actually better lens, should produce a noticeably better result, with or without teleconverter.

On the other hand, if one has, or is prepared to buy, a 400mm Nikkor lens which is undeniably sharper than the Canon equivalent (if there exists such a lens), then this factor alone would probably offset any minor pixel-count advantage of the 7D, so it becomes a non-issue.

For myself, I have no interest in buying a very expensive and very heavy 400mm F2.8 prime. It wouldn't suit my purposes. I limit myself to the Canon 100-400/F5.6 IS. At 400m it's marginally sharper at F8 than at F5.6. I would like an upgrade that is sharper at F5.6.

When I use this lens at 400mm and F8, I usually have to be at ISO 400 or higher to get a sufficiently fast shutter speed. At ISO 400 and above, the DR advantage of the D800 pixel is greatly reduced.

Finally, it always seems a little strange when people make hypothetical comparisons between an actual available, thoroughly reviewed product from one company against speculated possible future improvements of another company in its product offerings. At the very least, do you care to speculate on what Sony and Nikon might be working on, and have to offer at about the same time as these imagined new Canon sensors arrive? Since you now own and use a mix of Canon amd Nikon gear, surely you can speculate about what both have under development?

Canon has a tradition of a significant time-lag between matching the pixel density of its full-frame cameras with that of its earlier cropped-format cameras. For example, we had to wait for the 1Ds2 to get the pixel density of the much earlier 6mp D60 and 10D models. We also had to wait a few years to get the 1Ds3 and 5D2 with a similar pixel density to the 8mp 20D and 30D.

Nikon are obviously aware of this history of Canon development. With the D800, Nikon have simply shortened that time lag and, in a sense, stolen a march on Canon.

[BJL]

I wasn't claiming any victory.

True: you simply chose terms of comparison that favor the only advantage of the 7D (higher pixel density) and mentioned that advantage (by comparing the pixel counts of the crops) while not mentioning any of the advantages of the D800 sensor that are likely to be relevant to that comparison.

What really fascinates my though is how in all your many words about Nikon D800 vs Canon 5D3 or 7D is how little you have to say about the elephant in the room: the rather fundamental technological difference between their sensor technologies. The idea that Canon likely has other sensors under development does not address the question of why, four years after Sony and Nikon adopted the newer and overall clearly superior approach of doing ADC on the sensor in a highly parallel way, and some years after Panasonic started used that approach in the GH1 (almost certainly) and GH2 (definitely), Canon has just launched two new cameras at the top of its product line that still use an evolution of the older approach that it adopted about five years before Sony started into CMOS sensors. It is hard for me to see an explanation more likely than that Canon has a great accumulation of expertise in that "first generation" approach to CMOS sensors, and so there is more technological and financial inertia delaying its move, whereas Sony was forced to abandon its CCD technology at a time when it could "start from scratch" in its EXMOR development using a newer and better approach.

Maybe Sony was the frog dropped into boiling water that jumped out, while Canon is the frog still simmering as the water gradually gets too hot for comfort.

(As an aside, it is interesting to note that Panasonic alone has gone through all three generations of sensor technology over the last decade, from CCDs as in the original Canon 1D to "analog output active pixel CMOS" sensors as in most Micro Four Thirds sensors, to "on-chip parallel ADC CMOS" as in the GH1 and GH2. This might relate to the sheer size of the electronics operations of the companies involved: Panasonic > Sony > Canon > Nikon > Olympus.)

[dreed]

Snap! See the end of my post just before yours. My thought is than in two or three years, all major players are likely to have improvements to show ... but as technology diminishes imperfections towards negligible levels (like dark noise down to one or two electrons) gains in practical terms become ever smaller, and the lagard has more room to improve, meaning that the gap will tend to shrink. In particular, crossing the bridge to on-sensor high speed parallel ADC might be a bigger step in low noise and high dynamic range than anything that follows, and Canon has that obvious jump to make yet, while Sony, Nikon, Panasonic and probably Olympus and Samsung are already there.

I think that you're arguing that the law of diminishing returns will pop up its ugly head here somewhere, right?

And that once everyone is manufacturing sensors using a similar design, sensor output will be roughly the same, give or take a third of a stop?

I do not put much weight on rumors, and especially this one: neither the massive price drop after about six years of price stasis in 35mm format nor the idea of a new Nikon/Sony sensor of lower resolution than the one just launched makes sense, except as a repitition of speculations that keep recurring and keep going unfullfilled.

Sometimes I wonder if rumours are actually started by the companies in question as a way of manipulating consumer choice, whether they be defensive (a new camera from X will be announced in Y months so don't go away) or offensive (as the D600 rumour may be an attempt to put the purchasing of a full frame camera by others on hold until...)

In particular, at the risk of repeating myself:
- the D800 in particular shows that lower resolution does not significantly improve noise or dynamic range when compared fairly as with the DX0 "print comparison"

The problem is that sensor technology isn't static, so if they made the 12MP sensor in the D700 using the technology in the D800's sensor then maybe that would be even better again. Ultimately, we don't know and will probably never know.

I wonder if any of the websites that have done a lot of number crunching on sensor performance can deliver a set of numbers that would answer this question?

Some time ago DxO did something like this and their comment was that the sensor in the Canon S90 (or was it S95?) was better than the full-frame Nikon D3s (or was it D3x?) in terms of the performance it delivered relative to what it could collect in terms of signal. That kind of caught my eye. Then again with the G1X, Canon showed that they could deliver a really good sensor, but then you look at the 5D3 and wonder where the tech that has obviously been developed is going.

[dreed]

What really fascinates my though is how in all your many words about Nikon D800 vs Canon 5D3 or 7D is how little you have to say about the elephant in the room: the rather fundamental technological difference between their sensor technologies. The idea that Canon likely has other sensors under development does not address the question of why, four years after Sony and Nikon adopted the newer and overall clearly superior approach of doing ADC on the sensor in a highly parallel way, and some years after Panasonic started used that approach in the GH1 (almost certainly) and GH2 (definitely), Canon has just launched two new cameras at the top of its product line that still use an evolution of the older approach that it adopted about five years before Sony started into CMOS sensors. It is hard for me to see an explanation more likely than that Canon has a great accumulation of expertise in that "first generation" approach to CMOS sensors, and so there is more technological and financial inertia delaying its move, whereas Sony was forced to abandon its CCD technology at a time when it could "start from scratch" in its EXMOR development using a newer and better approach.

I would expect that Sony patented that design and implementation, so maybe the challenge for Canon is to develop and deliver something equivalent without infringing on Sony's IP. At least that's my lame attempt at an excuse for Canon.

[shadowblade]

I would expect that Sony patented that design and implementation, so maybe the challenge for Canon is to develop and deliver something equivalent without infringing on Sony's IP. At least that's my lame attempt at an excuse for Canon.

Hopefully, that means bypassing the parallel on-chip ADC stage and going straight for per-pixel ADCs. Now, that could really expand a sensor's capabilities and dynamic range, including such things as variable ISO on a pixel level.

[dreed]

Hopefully, that means bypassing the parallel on-chip ADC stage and going straight for per-pixel ADCs. Now, that could really expand a sensor's capabilities and dynamic range, including such things as variable ISO on a pixel level.

Variable ISO on a pixel level would be very interesting, however it would be a 25%(?) increase in file size to allow for RGGBI to be stored per pixel. But would that really work?

Does the noise introduced by amplifying the signal produce better quality data than not doing so?

[BJL]

@dreed,
Yes, my general point is diminshing returns, which in general means that as the gap between current technology and perfection gets smaller, the gap between competing technologies becomes ever less significant. Mike Johnston (The Online Photographer) wrote a nice piece about the ironical conflict between reducing imperfections in technology and connoisseurs putting increasing effort and emphasis onto to weighing the ever diminishing differences.

But to be far more explicit: I hope and expect that sensor technology is close to making read noise irrelevant, so that noise levels and dynamic range will be essentially all about photon shot noise, which means that they are all about how many photons are gathered per output pixel. And I say output pixel, becuase at that point, the signal and noise levels in a pixel do not depend on whether that pixel comes from one bigger photosite or by merging the signal from several smaller photosites that cover the same amount of sensor area and gathered in total the same number of photons.

My reference to the D800 is that it seems very close to that level. Moreso for the vast majority of photography in which nothing more than about ten stops below the brightest highlights is significant, because in those top ten stops or more, the dark noise of the D800 is overwhelmed by the unavoidable photon shot noise.

[BJL]

I would expect that Sony patented that design and implementation, so maybe the challenge for Canon is to develop and deliver something equivalent without infringing on Sony's IP. At least that's my lame attempt at an excuse for Canon.

I am sure that Sony has some patents and/or trade secrets on the implementation details, and for example was limited to 12-bit ADC in the A900 whereas the D800 has 14-bit so clearly work is needed beyond simply adopting the basic idea. So yes, Canon's task is not trivial. I am not sure though how "sitting by while your major competitor gets patents that box you out of using state of the art technology" is much of an excuse for Canon! I never heard this sort of excuse offered for Nikon and Sony staying with CCDs back when Canon had a clear advantage with its CMOS technology.

But as I pointed out, the core approach of on-sensor highly parallel ADC is in wide use, by almost every one of Canon's major competitors, and there is a lot of information about it in numerous published papers. In addition to Sony and Panasonic, Samsung has also been using column-parallel ADC in some sensors for even longer than Sony. In a sad irony, the oldest patent I have found on this was by Kodak!

[ndevlin]

Canon's biggest mistake is calling it a 5DMkIII.  The 5DmkII was a POS camera, with a good sensor.  This new machine is a really, really nice camera (better qua camera than the D800 in a bunch of ways) with a very, very good chip in it.

Why the hell saddle it with a name that admits only incremental improvement?? In fact, I think many Canon shooters will think the 5D3 is the cat's meow, and if you don't need monster prints, it's all the camera you'll ever need.

Worst camera naming ever.  

Except maybe for "*ist" from Pentax.

But it's close.

So here's my short review: "the 5D3 is more and better camera than the name suggests. If you shoot Canon, buy one."

- N.

[mattbr]

As I wrote, the 5D MKIII sensor would have been fine if it didn't have the D800 to contend with. But, this is a marketplace issue, not one related to the ability of the camera and sensor to produce great IQ, which the MKIII certainly can.

Michael, this honestly sounds a little bit disingenuous (or very diplomatic) coming from someone as knowledgeable as you are. Even cursory examination of the new sensor yields the conclusion that the improvements over the previous version are somewhere in the realm of the minimal, and definitely extremely underwhelming given the time it took to release the camera (or at the very least way below what could be expected in terms of progress). That the D800 came out essentially only makes these shortcomings painfully apparent. While the quasi-nonexistent evolution in resolution might, to a degree, be compensated by lens designs in some cases (specifically, the hallowed day the 24-70 II is finally released, and replaces the rattling, out of alignment engineering disaster that was v1's zoom column...), the rest of the design likely simply will not. Had the D800 not been released, my conjecture would be that logic would have called for a/b's against the mkii... And that the conclusion most testers would have reached would have been "whoopedee-doo, Canon solved some of the banding problem !". I, for one, would be curious to see a double-blind experiment in differentiating, say, the dpreview test shot between the two, taken from the raw files. I'm quasi certain that, assuming no boosting of the shadows is allowed, the results would come through inconclusive.

Now, this all could of course have been partially compensated by with major improvements to other areas in the camera. Which have, thankfully happened. The AF is (complexity put aside...) really good, the camera feels much more solid than the mkii, and is much more comfortable to use. Someone else said that the mkiii is what the mkii should have been, and i can't agree more. If there is something in it that could not have been released when the 7d was released, i'd be extremely curious to know what it is (oh, and before someone rips me a new one for insensitivity to environmental catastrophes : D800). It is, in my mind, and having shot about 10k frames through it by now, at best, 18 months too late and 30% overpriced - in other words, an epic failure, maybe unprecedented for the brand.

The article i'd be curious to see, from Luminous Landscape, or anyone else, isn't a review - a review of it, is, as you've clearly understood, a waste of time. It's an investigative one : how could this happen ?

[Hans Kruse]

I think that a 5D mkIII review and contrasting it with the D800 is well worth the effort. The 5D mkIII has lots of improvements over 5D mkII and in many ways a more general purpose camera than the D800.

What I would like to see in such a review is what Michael does best is to have both feet planted on the ground and tell his opinion about the two camera systems. Examples based on real photography and not just test shots (as we also see on the forums) where one of the cameras has an advantage over the other. This could be in dynamic range. When is the 5D mkIII really falling short and for resolution. How big would you need to print in order to see the difference and here a (double) blind test would certainly be more convincing than a trust me statement. The 5D mkIII has a different implementation of various things like AF, fps, live view so how does it fare against the main competitor in real photography. Wild life is an area where I see the 5D mkIII as a stronger camera than the D800 due to shooting speed.

That the D800 came around at 36MP should not really have been a big surprise since we already saw what such a sensor could bring from the D7000 sensor and other cameras using the same sensor. The next step up for Nikon (and Sony)  for full frame would logically be a FF sensor based on the 24MP APS-C sensor giving a 54MP sensor.

What is more surprising is that Canon chose to stay at 22MP and not base a new FF sensor on the 18MP APS-C sensor which would give a 46MP sensor.

Thoughts about these choices and why they were made and where we are heading for the future (and as mentioned what is relevant for real photography of various types) might be more interesting than how much you can lift shadows from the D800

[dreed]

And that the conclusion most testers would have reached would have been "whoopedee-doo, Canon solved some of the banding problem !"

The operative word here is some as images that have been uploaded thus far reveal that the problem still exists in the 5DIII. You've got to wonder just what Canon have been doing...

[Hans Kruse]

I don't think it is fair to say that there is no improvements in the sensor of the 5D mkIII. DxO says the following on the pixel level.
The 5D mkIII is at the pixel level the best at ISO's at ISO 1600 and above. DR suffers at lower ISO's as we know. I have yet to see a good review with actual photographical examples of shadow lifting as you would do in real photographs where the 5D mkIII falls short. Of course they exist, but artificial lifting of shadows does not really prove that much IMHO. The reason I mention this is that even the Canon 1Ds mkIII does quite well even in high DR scenes using Lightroom 4 editing capabilities (I mean in real photos). I'm certainly in no way against more DR, but I'd like to have to brought into perspective of we do as photographers.

[BJL]

artificial lifting of shadows does not really prove that much IMHO.

By artificial lifting of shadows, you seem to be referring to what we used to do with low contrast printing and dodging and burning in order to print with scenes of high subject brightness range. Most photographs do not need it, but some do, and the dynamic range limitations of digital cameras have long been a sore point with some photographers, so I cannot agree that this is either "artificial" or irrelevant to assessment of DSLRs. Especially when looking at the new top of the line models from what has for some time been the leading brand.

[MatthewCromer]

DR is a REAL photographic requirement in some situations.

For example:

http://lightskyland.com/images/2402_1200.jpg

That image required a significant highlight recovery as well as pushing the shadows 3-4 stops.

[BJL]

I don't think it is fair to say that there is no improvements in the sensor of the 5D mkIII. DxO says the following on the pixel level.

True: a roughly 1-2dB improvement, and it looks slightly better in the fairer "print" normalised comparison: maybe 2dB. Not a lot of progress for 3 1/2 years, but about the same as from D3X to D800 in "print" normalisation, and I speculate that the progress is limited by the fact that the noise in these tests is dominated by photon shot noise, and so is close to the physical limits under the constraint of using a Bayer CFA with adequate color discrimination. (Only in far darker parts of the image does dark noise become relevant, or at very high ISO where the Canon sensors avoid their DR limitation). This is suggested by the fact that when you compare the 5D3 and D800 with "print" normalisation, the SNR 18% curves are almost identical. That amazing closeness over the whole ISO speed range from quite different sensor technologies makes me think that the cause is a common fundamental physical limit, i.e. photon shot noise, not noise generate in the cameras.


P. S. Given that extreme closeness, I am not sure where Ray gets his claim of better high ISO noise performance for the 5D3, let alone his apparent sentiment that this high ISO noise advantage plus 6fps vs 4fps was enough reason for Canon to choose a sensor design with substantially inferior resolution and dynamic range. Perhaps some comparison involving a choice of lenses that eliminates or reverses the pixel count advantage of the D800, making the DxO "screen" comparison more relevant.

[Bernard ODonovan]

Got it.     

Hi Yakim

I recently read through this review:

http://www.kenrockwell.com/canon/5d-mk-iii.htm

It's not all roses but it seems overall for Ken it's the more practical camera, mainly due to the ''Canon 5D Mk 3 shooting mode dial: C1, C2 and C3''.

AF and WB are a couple of the many items he has issue with, so worth a good read to ensure all the issues will not affect your likely use...