There's nothing contradictory in that. It's just one more stop shifted to highlights. The lowest contrast setting for JPEGs in the camera usually puts the JPEG clipping point just below the RAW clipping point, in the "lead" channel, green. The red and blue channels are scaled for WB, and those are where the most clipping usually occurs. If you set the camera to normal or high contrast, then you start wasting more RAW highlights in the JPEGs.
I don't know how you're gauging this, but the 7D image, taken as a whole, has more DR than either of these cameras, by the normal definition of DR. You are talking about highlight headroom, it seems, and that is often more a matter of metering mode and contrast settings than it is RAW headroom.
This isn't necessarily true. The general rules are that DSLRs have the highest at high ISOs, medium format and compacts the least, and the more pixels, generally, the higher the DR at low ISOs (Screen), but with compacts generally a stop less than a DSLR or MF with the same number of pixels (some compacts beat some DSLRs, though).
The question is how big are you displaying the image. For a small image, you can waste many stops of RAW headroom and still get a usable image on many cameras, at base ISO. The biggest obstacle there is that you may run into banding.
The 7D has way more DR than the 10D, at base ISO, and even more yet, at high ISOs. At ISO 1600, the 7D has about 3 to 4 stops more practical DR than the 10D.
There is no magic in the 40D; must be just a difference in tone curves used.I was wondering what makes the images from my 40D the images so special (other than the fact that I took them). It may be dynamic range. Here is a compilation of dpr's recent DR results for in-camera jpegs from Canon and Nikons. These are prevailing values up to ISO 1600:
Cam... EV
30D... 8.4
40D... 9.1
50D... 8.3
7D... 8.3
5Dii... 8.4
D200... 8.2
D300S... 8.4
The 40D DR is a standout. I wonder why?
Why would you expect the same DR with the same sensor? The downstream electronics and digitization affect DR more than anything. The 20D and 30D happen to have the same sensor and electronics, but the with the 30D Canon started artificial clipping of RAW data below the ADC limits, throwing a wild card into the mix.
I've been wondering about that. How much 'headroom' (to the clipping point) there actually is in RAW. I'm shooting RAW+JPEG, but my computer is old and slow, so I'm not processing the RAWs, just saving them for later, just in case. I'm shooting with Neutral Picture Style, contrast -2, and almost always in daylight with WB set at 5600K. I'm also shooting 'to the right', carefully watching the R/G/B histograms to avoid clipping. Are you saying that with those conditions/settings I can't expect that there's much additional 'headroom' in the RAWs? And does any of the three color channels (usually, under the same conditions) have more 'headroom' than the others?
That would be true if jpeg were recording linear data (ie doubling the illumination doubles the digital number representing it). Most of the time jpeg is used to store data that has been given a "gamma" transformation -- the data is raised to a power, usually 1/2.2, and that compresses the highlights while lifting the shadows so that the 8 bits can store more like a 9.5 stop range of illumination levels.
The 7D might do better at "screen" at low ISOs; some 7Ds have pixel read noise lower than the 5D2 (the IR one does). Having a few more pixels will make the 5D2 a little better in "print".
Clipping blacks does nothing for image quality, except clip away part of the image. That can be done with any RAW, really, if that's what you want.
I know what dynamic range is. I also don't care much about high iso dynamic range. 95% of my shooting is at ISO 100. While the 7D has more dynamic range, no doubt about that, I would much prefer a camera that has the dynamic range of Fuji's S5pro.
----
John
mRAW and sRAW are equivalent to taking a full-size RAW, converting it, and downsampling the result by a factor of 1/2 (sRAW) or about 11/16 (mRAW). The downsampling eliminates all small scale detail, and gets rid of fine scale noise as well. Thus mRAW and sRAW have lower pixel-level noise, because their coarser scale pixels measure the noise at larger image scales, and noise is scale dependent in an image.
That comes at a price; the DR is increased, but the quality at various levels is inferior. For stable subjects, you're better of taking multiple exposures.
DR = headroom + footroom.
With -2 contrast, you should have at least 1/3 stop in the green channe in DPP, maybe another 1/3 in ACR depending on the ISO (1/3 stop groups have different headroom)l, and more in the red and blue. With daylight WB, the red channel has about 1 stop more headroom than the green, and the blue, about a 1/2 stop more than the green. ACR will use the red and blue channels for extended grayscale highlights, IIRC. I don't keep up with the converters, so I can't give up-to-date specifics.
I could (very well) be wrong, but I've got this idea that downsampling affects the noise relatively more than the detail, because the noise isn't affected by the AA-filter (and lens, etc.), but the detail is, meaning that the noise is present at a higher spatial frequency than the detail, and therefore is affected most. (Or maybe one could say that the noise is captured with perfect per-pixel sharpness, and the detail isn't). But of course it doesn't really matter, because I'm also sure it's correct that a good noise filter gives a better result than a simple downsampling.
Finally, dynamic range and noise are two sides of the same coin. DR is the range of tonal levels over which the signal-to-noise is acceptable. Since noise is scale dependent, being larger at finer scales, DR is also scale dependent, being smaller at finer scales. There is nothing bad about this, it's just a fact of the way pixels sample images. The important thing to remember is that finer pixels still have the information about coarser scales in their data, which can be uncovered with proper image processing.
--
emil
--
http://theory.uchicago.edu/~ejm/pix/20d/
Thanks. That's a very valuable/useful piece of information ;-)
But noise doesn't really matter when it is at higher frequencies. Noise is problematic at low frequencies, which dowsampling doesn't remove.
Get some noisy but detailed crops, at 100%, and use the pixelate tool in photoshop interactively; you might see that binned versions don't really improve things much at the same image size. Unfortunately, PS has no way of viewing like this with downsampling insteade of binning, but you can make multiple copies of a crop, and downsize to multiple sizes, and then upsize all to the same size, to compare. I bet you won't find anything better than the noisy original, for any busy, detailed scene.
