A problem with DxoMark dynamic range measurements is that they do not show "usable DR". 14 stops measured DR just means that 14 stops down signal is equal to noise, that is how "engineering DR" is measured. But in a real photograph, such noisy information must be black, it is unusable. In fact, those parts of the image that are bright enough to show detail must have pretty good signal-to-noise ratio to look good. So perhaps only 7-8 out of those 14 is usable.
Again, I direct everyone to the following example:
Based on what you've said, all the "black" in a photograph is unusable because its SNR is too low, and therefor the information is unusable. I think the videos above prove differently. Unusable "blacks" is largely true with Canon sensors, not true in the least with Sony sensors. I think the videos speak for themselves...modern sensors, including canons, offer far more DYNAMIC range than 7-8 stops. Keep in mind, DR and contrast are not the same thing. A final photo may exhibit 7-8 stops of CONTRAST, the measurement of the darkest parts of the scene vs. the lightest parts...but the range of contrast exists within the full dynamic range a sensor is capable of. The range of contrast may start lower and end higher, or start higher and end lower. You may have low-key images that only use the darker range of tones offered by a sensor, or high-key images that only use the lighter range of tones offered by a sensor. You may have something that uses mostly the middle tones. Its called dynamic range for a reason, and its not the same as contrast.
Sensors do indeed provide 12-14 stops of dynamic range, and that full range is usable...either to recover a botched exposure, deal with available light issues, etc. The END RESULT of mucking around with a RAW images exposure in post, shifting highlights and shadows...is a photo that probably has around 7-8 stops of tonal range. Print it, and your probably down to 5-6 stops. The end result does not change the nature of the sensor, though...you still have that original 12-14 stops of DR to work with when taking the shot and working it in post.
Time for me to jump in on this topic. I generally watch from the wings of non-registered users on this forum but this DR topic is my pet peeve with recent Canon bodies too!
I'm an electronics professional by day, love reading all the tech details from various sites but when it comes down to it, let the images speak for themselves.
I don't completely agree with the usability of DXO's sensors tests; they are a good guideline but they seem to not account for the horrendous pattern noise of Canon's recent sensors at low ISO on some bodies.
Also, if you look at the DR curve vs ISO on Canons sensors from way back until now you'll see they're essentially flat before they drop off into the expected slope at higher ISO. The ISO that this occurs has been increasing with newer generations of their sensors, let's call that "improvement."
I was stunned when I saw the curves for Nikon's D7000/5100. THAT, to me, is how that curve should look; you trade sensitivity for DR in a very linear fashion. Canon's curves are not like that, roughly indicating there's a noise floor problem somewhere in their system. That Canon's raw files exhibit more pattern noise at low ISO than at high ISO is baffling but I'll leave analysis of this to sites like clarkvision and others who can delve into details I don't have available
On the practical side...
All the fuss about HDR as a technique should be nearly moot for most scenes if we have a truly decent 14 bit capture. My first Canon body was the 40D. I deliberately held out from going into DSLR until there was more than 12 bits worth of data to work with. The live-view was a massive bonus.
I've done some shots with the 40D where I expose for the highlites and push the shadows in post for an effectively HDR, but realistic-looking shot, based on one properly exposed raw file. They can look great!
With all the hoopla over the 5D2, I was downright unimpressed when I got mine. I had low ISO banding right in the midtones!
In fact, I think my 5D2 and 7D are both turkeys when it comes to all the banding noise in the shadows at low ISO, something my older bodies do not exhibit.
Latest firmware and upgrades to DPP seem to have somewhat mitigated the severity of the problem but when it comes down to it, it seems like I can push the shadows from my 40D a bit more than I can with the 5D2 or 7D.
In fact, I was just thinking about doing a comparison of sorts amongst all the Canon bodies I have access to for exactly this sort of test. I often like to boost shadows in post. The 5D2 and 7D raw files don't often leave me as much to work with as I'd hoped.
I plan to take some shots of "dark" at low ISO and moderate shutter speeds, the sort I'd use in landscape work. I'll then see how far I can push these files in post before the banding noise becomes apparent.
I have access to the following bodies for this test:
350D, 40D, 60D, 7D, 5D2, G11, G12.
It'll take me a bit of time to get this done, got a few projects in the way at the moment but I'll post the results when I do.