According to eos-numerique.com (in French), the long awaited 5D replacement will be presented on Sept 17th. This is a serious Canon enthusiasts' site run from France and moderated on both sides of the Atlantic (i.e. Québec). Here is what they have to say:

Capteur CMOS 24X36 24 MP
Nouvelle batterie, nouveau grip
Processeur DIGIC IV
Boîtier en alliage de magnésium
Live view
Mode vidéo
Système de nettoyage intégré

Translation: full frame 24 mpx; new grip & battery; new digic 4 processor (like the 50D); magnesium alloy; live view + video; dust removal

Nothing revolutionary: the Nikon D700/D90 have opened the road... but it looks like a serious upgrade nevertheless. Unfortunately, no news on the hottest topic: ISO range and follow-up to the 5D's legendary low light performance.

Canon won't outdo their flagship EOS-1Ds Mark III that "only" has 21.1 MP.

Why not? They have to react to the market.

Exactly right! I can't be the only one who is considering switching systems. When the 5D replacement is announced, there'll be a lot of frenetic comparisons between it and the A900. The longer the time interval between upgrades, the bigger the improvement expected.

What road did the D700 and D90 open?

And, btw, how on earth do these two come in this comparison? What does the D90, an entry level DSLR (though with very good noise characteristic) have to do with a full-frame, 24 Mpix camera?

Neither the D3 nor the D700 opened any road, except inside Nikon.

There will also be comparisons between the 5D successor and the D700. The question is: Which camera is the 5D successor competing against -- the D700 or the A900? Surely it's not both; barring some unforeseen breakthrough that would be highly improbable. Think of it this way: Reasonable price, solid build, lots of features, performance, high ISO performance or high resolution -- choose 4 (and there is currently no way to reconcile high ISO performance and high resolution at any price) .

In this day and time I would be careful about a declaration that there is "no way" that a technology (combination) under discussion is possible. The rumored "way" is a new-technology variable ISO technique previously mentioned on this and other forums:

http://www.apical-imaging.com/UserFiles/ad-iso.pdf

IMO, the logical fallacy behind the oft-expressed, overly-simplistic idea that a manufacturer will not preempt an existing high-end model is that it fails to take into account the temporal nature of product cycles. The argument would make sense if the new-model starting point and ending point of all models within a product line's hierarchy were synchronized. But of course they are not, for reasons related to a manufacturer's development resources and market conditions, including competitive pressures such as those now seen by Canon from Nikon.

So, the preemption question must be analyzed in light of a particular point in time relative to the current product cycle of the two models under consideration. The 5D was announced on August 22, 2005. As such, it has passed through about 150% of its logical product life cycle. (The only reason that it lasted so long was because [a] it was a classic in terms of image quality; and [b] there was no competition.)

The IDsMarkIII was announced on August 22, 2007. As an example, for the sake of argument, if its replacement is announced at the beginning of 2009, the IDsMarkIII will have been two-thirds of the way through its product life-cycle right now. Overlap of some key features between models (and thus preemption and cannibalization of sales) may necessarily occur at points in time of the end of the product cycle of the higher-end model and the start of the product cycle of the lower-end model.

Indeed, such an occurrence may be seen as a sign of the end-of-life of the higher-end model (the model's "Winter") and anticipation of the birth of a replacement higher-end model (the model's "Spring"). In its admittedly late review of the 1Ds Mark III last month, DPR said "although there's no doubt that the Mark III sets the standard for digital SLR resolution and significantly narrows the gap between '35mm format' and 'medium format' digital cameras, it is also undoubtedly a niche product that will, by now, have been bought (or earmarked for purchase) by pretty much everyone who is ever likely to hand over the best part of $8000 for one."

I hope that my fellow photogs who have recently purchased the 1Ds will not shoot the messenger. None of this means that a 1Ds purchase as late as today was a mistake. You will probably have recognized the value of your dollars spent before the 5D replacement even ships, and certainly before the 1Ds Mark III replacement is announced and ships. And, as many others have commented, at some point it is probably a good idea to call one's acquisition of camera body technology "good enough."

Best,
Bruce

I'm not interested in opening PDF files (they slow down my computer and hide until I reboot), so post relevant quotes here if you consider it worth my time and yours. Frankly, it appears you want to pile on speculation about what the resolution, price, etcetera of the coming 5D successor will be with speculation about some technological breakthrough which I have already written is "highly improbable."

I am careful about declaring there is "no way", because that is in the context of what Canon will offer in a 5D successor; next time include the word I wrote just before the part you selectively clipped -- "currently". It would be ludicrous for Canon to leapfrog their flagship DSLR technology with a DSLR that would be competing with those that cost less than half as much.

Web forums are riddled with speculative threads espousing unreasonable expectations of what will or should be available tomorrow or next week, if half this stuff had come to pass we would already have 135 format DSLRs with 14 stops of DR that shoot 30 fps for a $1000. Someday we will, but not tomorrow or next week or even next year.

It's not clear to me what this Apical system is doing. Having looked at their website, it seems that they've developed a system to analize and process the data in a particular way after A/D conversion and after the signal has been read.

Anyone who's taken the trouble to compare equal exposures at ISO 100 and ISO 1600, from their 5D or 20D, should be very impressed with the vastly superior shadows that the ISO 1600 image contain. The difference is like chalk and cheese.

Whether you compare a correctly exposed ISO 100 shot with the overexposed ISO 1600 shot, or a correctly exposed ISO 1600 shot with the necessarily underexposed ISO 100 shot, the difference in the shadows noise and general quality of the image in the shadows, including resolution, is huge.

I understand that Canon achieve this low noise result at high ISO by applying an anlog boost to the voltage from each pixel prior to A/D conversion. The ISO setting determines the amount of preamplification applied.

Apical appear to claim they are applying a variable ISO value to different parts of the scene, after software analysis of the scene. This seems to me like a variable fake ISO, in a similar manner to Canon's ISO 3200 which in most of their models is the same as ISO 1600 underexposed one stop with digital values later multiplied, together with all noise introduced by earlier processes in the chain.

The following crops of shadows are from equal exposures; same lens, same F stop, same shutter speed, but different ISOs. The ISO 100 shot was close to a correct ETTR, maybe 1/3rd stop underexposed. The ISO 1600 shot was of course overexposed.

Click to view attachment

I got the same general impression from reading the Apical white paper, Ray. That is, ISO amplification in the digital domain on a pixel-by-pixel basis.

As I understand it, analog amplification of the base sensor signals is the traditional means of increasing ISO. That amplification process introduces noise; the greater the gain the more noise introduced. Once a sensel value is A/D converted it can be manipulated digitally without the further introduction of noise.

I must admit that I neither doubt nor understand the reason for your clean-signal, high-ISO results. (Probably something simple that I am missing here.)

I find little that is simple in these circuit designs . But my simple understanding of the broad concept, is that it's better to amplify a low level signal before reading and converting it to the digital domain. The reading and the D/A conversion have their own noise. When that further noise is added to a low level signal, the signal-to-noise ratio suffers greatly. I think this is what my comparison crops above are demonstrating.

Although this news has now become a reality, I'd just like to add, for the record, some points which are relevant to my above comparison crops.

Quoted from the Dpreview preview of the 5D MkII, it appears to be confirmed that Canon have a pre-amplifier at each photosite. This, it would seem to me, is largely responsible for Canon's low noise at high ISO.

Looks like my mates at EOS-numerique.com were right!

(give or take a couple of minor details)