Let's compare the dynamic range of our cameras

Here's one for the 1D3:

HOSTED PHOTO
please log in to view hosted photos in full size.

Finally, the differences look minor. So I don't know how can we interpret this...
Is it really a difference in the dynamic range or is it just a programming artefact ?
I wonder also why the differences appear only on the right side of the histogram ?

Logically the differences should be on the left. The dynamic range is determined by two factors. The top limit is set by the strength of incoming light that causes the sensor to reach saturation, the point at which any further increase in luminance input does not cause an increase in electron output. Over the years this has remained fairly constant at 3.5 stops over medium grey - despite attempts to discover better sensor materials. The bottom limit is the point at which noise overwhelms image data. The less noise the camera produces, the lower the limit.
Of course, Canon is free to calibrate the medium grey point wherever they want and apparently they shift it downward on models with greater DR in order to utilize the extra DR at the top.

www.dxomark.com

Interesting topic and exchange. My two bits (spoken from the arm chair of course since I don't have insiders access to their system).

The low limit in hardware is 0 and a file that has 0 (pure black) may very well have noise in the shadows but to exclude or "soft" limit the low end based on noise should not preclude what the A/D / imaging engine is capable of generating. All I am saying is that my file better have pure black at 0 in it if indeed the scene is exactly that regardless of noise. Perhaps the raw data is conditioned to produce usable image data and scaled accordingly such that while the hardware is capable of 0 the noise floor dictates an artificial bottom that 0 is set to. In other words the raw data is conditioned ((aside from noise reduction) scale and b/w thresholds). Most likely this is the case since they most likely have to deal with non-linearities in the sensor cells.

tzalman wrote in post #11426915
Logically the differences should be on the left. The dynamic range is determined by two factors. The top limit is set by the strength of incoming light that causes the sensor to reach saturation, the point at which any further increase in luminance input does not cause an increase in electron output. Over the years this has remained fairly constant at 3.5 stops over medium grey - despite attempts to discover better sensor materials. The bottom limit is the point at which noise overwhelms image data. The less noise the camera produces, the lower the limit.
Of course, Canon is free to calibrate the medium grey point wherever they want and apparently they shift it downward on models with greater DR in order to utilize the extra DR at the top.

Canon 500D

HOSTED PHOTO
please log in to view hosted photos in full size.

I have counted the number of positions the vertical sliders can take in the RAW histogram for the classic 5D. Here are the results. From left to right starting the counter by 0 :

Position of the slider -> value of the counter

extrem left -> 0
-8 stop -> 1
-6 stop -> 5
-4 stop -> 21
-2 stop -> 48
0 stop -> 80
+2 stop -> 112
extrem right -> 138

What do you think ?

Where are you getting the readings for those values?

This is from my 30D. Seems identical to the 20D.

HOSTED PHOTO
please log in to view hosted photos in full size.

DPP controls generate parameters which feed one or more algorithms for manipulating Raw data. You are discovering the resolution of the controls. Lenses are calibrated in 1/3 stops and my guess is DPP is designed to cover the expected range of control based on well established trends in photography. Once an image is exported the manipulated data is stored in the specified output format at the selected resolution. There is no point in working with a reduced data for one graph and not for another. I suspect all manipulation is at the full working resolution of the sensor.

tonylong wrote in post #11429986
Where are you getting the readings for those values?

I have used a new feature of next version of DPP++ Plugin which allowed me to control the sliders of black and white points of the RAW histogram with maximum precision by using keyboard or buttons clics. Each clic or each key press moves the slider by 1 position. That's how I did the count.

HOSTED PHOTO
please log in to view hosted photos in full size.

yb98 wrote in post #11430164
I have used a new feature of next version of DPP++ Plugin which allows me to control the sliders of the RAW histogram with maximum precision by using keyboard or buttons clics. Each clic or each key press moves the slider by 1 position. That's how I did the count.

OK, I checked out the Brightness slider, which starts out at 0 -- moving it from the far left to the far right went from -2.0 to +2.0 (you could move it in fine increments) but it did the exact same for the 1D3 and the 5D, so no difference from arriving at the far right position, which is slightly farther for the 1D3 than the 5D.

I think the difference (if there is one) is at the moment the sensor captures the data...

Have you noticed that when you move the brightness slider the shape of the raw histogram changes, but when you move the black/white sliders of the raw histogram, the shape doesn't change ?

yb98 wrote in post #11430238
Have you noticed that when you move the brightness slider the shape of the raw histogram changes, but when you move the black/white sliders of the raw histogram, the shape doesn't change ?

Yeah, it's interesting playing around, not just with the brightness, white and black points, but note that the contrast, highlight and shadow controls have less affect on the histogram and more affect on the curve...