Reading mode: Light Dark

7D vs 40D image quality - my take after downloading DPP 3.7.2

Go to last post
Steen Bay said:
I hope/expect that Canon a couple of years from now will use BSI (back-side illuminated) sensors with a considerably improved QE.
Click to expand...
The quantum efficiency of BSI sensors is worse if you compare the active area only. The reason Sony and probably other manufacturers in due time will use BSI for their small sensor cameras is that the wiring on those small sensors take up a significant part of the sensor surface area. Large sensor cells such as encountered in DSLR don't benefit as much of the movement of the active area to the back side of the sensor as the wiring takes up relatively less space. With the 50D Canon introduced gap less microlenses which channel all light to the active areas and thus the loss in QE due to less efficient sensor area will translate fully to the performance of the DSLR sensor. So to sum it up: BSI on a DSLR sensor will result in worse performance.
--
regards
Karl Günter Wünsch
 
Karl Gnter Wnsch said:
Steen Bay said:
I hope/expect that Canon a couple of years from now will use BSI (back-side illuminated) sensors with a considerably improved QE.
Click to expand...
The quantum efficiency of BSI sensors is worse if you compare the active area only. The reason Sony and probably other manufacturers in due time will use BSI for their small sensor cameras is that the wiring on those small sensors take up a significant part of the sensor surface area. Large sensor cells such as encountered in DSLR don't benefit as much of the movement of the active area to the back side of the sensor as the wiring takes up relatively less space. With the 50D Canon introduced gap less microlenses which channel all light to the active areas and thus the loss in QE due to less efficient sensor area will translate fully to the performance of the DSLR sensor. So to sum it up: BSI on a DSLR sensor will result in worse performance.
Click to expand...
No! I'm pretty sure that there is room for almost a doubling of the QE, and more than a doubling of the fill-factor. Don't confuse 100% micro-lens coverage with 100% fill-factor (and/or 100% QE). The 'fill-factor' is the size of the photosensitive area relative to the pixel size, and that can be increased by a considerable amount.
 
Steen Bay said:
No! I'm pretty sure that there is room for almost a doubling of the QE, and more than a doubling of the fill-factor. Don't confuse 100% micro-lens coverage with 100% fill-factor (and/or 100% QE). The 'fill-factor' is the size of the photosensitive area relative to the pixel size, and that can be increased by a considerable amount.
Click to expand...
I think you'd really have to revolutionize the method of color separation to get close to maximum photon capture. Currently, the red and blue channels turn a lot of light into heat, instead of charge. If the wavelengths could be actively sorted and guided into their respective photosites, with high efficiency, the sensitivity would increase a big last step.

Higher per-unit-of-area capacity will not help the DR of the lower ISOs it affords, if the current tech which adds lots of noise after gain is still used.

--
John

 
Karl Gnter Wnsch said:
Jeff Kingston said:
I owned a 30D, 40D, 50D, 5DI and now have a 5DII and 7D.
Click to expand...
Except for the FF cameras I have had similar lineups. I still have a 40D, 50D and of course the 7D at my disposal and the results speak for themselves. The 7D is easily the best crop DSLR by quite a margin. Besides the lengthy feature list high ISO performance is astounding as are the levels of detail. Biggest of all though is IMHO the native increase in dynamic range without employing any tricks like ALO or HTP...

regards
Karl Günter Wünsch
Click to expand...
Will you please show your results so that they can speak for themselves?
 
the 100 -400 is a very nice lens for outside. I have the 70-200 f2.8 L and canon 1.4 ext. Very happy with my zoom, best zoom in Canon's line, and I use light collapsing monopod to help stabalize in low light.

For multipurpose photography, i hold to aquiring great glass is more important than the latest body, since the earlier bodies are good also. Glass will make more of a difference for many photography moments.

MAC
Steven Sterne said:
It is taking Adobe (and Apple) a long time for RAW support for the 7D - I don't remember if it took this long for the 40D or not. I hope that when you have a chance to test the 7D for sports, you post back and tell us hat you find to be the real-world differences between the 40D and the 7D.

Like you, I have found that the 40D's focus for sports is best with center point and a quick trigger on the shot - that's how I always use it for soccer. The tracking is not as sophisticated as the 7D's. I'm still getting used to the way I have to pan the lens when I'm tracking, and I've been really surprised at how differently the 100-400 IS behaves under mode 2 (for panning) compared with mode 1.

Steven
Click to expand...
 
ejmartin said:
Not sure I agree here. The measurements I made on BB's first 7D gave an average gain of about 2.1 e-/14-bit raw level at ISO 100.
Click to expand...
I'm starting to wonder how you measured it; I get wildly different results at every tonal level, if I subtract one OOF color checker from another, multiply the noise by .707, subtract the read noise in quadrature, and divide the mean ADU from a single copy by the resulting sigma, and scale for maximum ADU divided by the mean.

I do notice a tremendous decrease in noise when one image is subtracted from another; even before I multiply by .707 the noise is significantly lower in the subtraction. I can actually see vertical banding in the red channel in the white patch with IRIS' whitepoint just above that level.
ejmartin said:
This translates to about 28500 electrons at saturation at ISO 100 (clipping is about 13600 according to Gabor).
Click to expand...
Mine clips at 14,044 with black still at 2048; 1196 adjusted.

--
John

 
John Sheehy said:
ejmartin said:
Not sure I agree here. The measurements I made on BB's first 7D gave an average gain of about 2.1 e-/14-bit raw level at ISO 100.
Click to expand...
I'm starting to wonder how you measured it; I get wildly different results at every tonal level, if I subtract one OOF color checker from another, multiply the noise by .707, subtract the read noise in quadrature, and divide the mean ADU from a single copy by the resulting sigma, and scale for maximum ADU divided by the mean.

I do notice a tremendous decrease in noise when one image is subtracted from another; even before I multiply by .707 the noise is significantly lower in the subtraction. I can actually see vertical banding in the red channel in the white patch with IRIS' whitepoint just above that level.
John Sheehy said:
This translates to about 28500 electrons at saturation at ISO 100 (clipping is about 13600 according to Gabor).
Click to expand...
Mine clips at 14,044 with black still at 2048; 1196 adjusted.
Click to expand...
http://forums.dpreview.com/forums/read.asp?forum=1019&message=33335705
http://forums.dpreview.com/forums/read.asp?forum=1019&message=33345971

I had BB take two OOF GMCC charts, and used the difference to extract the variance as a function of average RAW level in a square. I looked at all squares for the blue channel; the result was so linear I only bothered with the bottom row of grayscale squares thereafter for the other channels (which also gave nice linear results).

Indeed there seems to be an inordinate amount of gain variation between columns in most 7D's I've looked at.

--
emil
--



http://theory.uchicago.edu/~ejm/pix/20d/
 
John Sheehy said:
I do notice a tremendous decrease in noise when one image is subtracted from another; even before I multiply by .707 the noise is significantly lower in the subtraction. I can actually see vertical banding in the red channel in the white patch with IRIS' whitepoint just above that level.
Click to expand...
Is the vertical banding consistent from shot to shot -- is it a gain fluctuation for individual channels, or is it something else?

--
emil
--



http://theory.uchicago.edu/~ejm/pix/20d/
 
ejmartin said:
John Sheehy said:
I do notice a tremendous decrease in noise when one image is subtracted from another; even before I multiply by .707 the noise is significantly lower in the subtraction. I can actually see vertical banding in the red channel in the white patch with IRIS' whitepoint just above that level.
Click to expand...
Is the vertical banding consistent from shot to shot -- is it a gain fluctuation for individual channels, or is it something else?
Click to expand...
I don't see much in the way of artifact after subtracting one image from its twin, and the remaining noise looks like a nice, dense starfield (fixed pattern offset read noises are subtracted along with scalar ones). I'm not sure why I was getting weird results previously, but today I am seeing what would be about 30 - 32KPh with a max adjusted ADU of 14335, but is a little less at 11996 (ISO 100's lower clipping). Of course, the latter is relevant for max usable capacity, but the former for QE.

Now, when are converters going to allow subtraction of a correction file before converting?

--
John

 
John Sheehy said:
ejmartin said:
John Sheehy said:
I do notice a tremendous decrease in noise when one image is subtracted from another; even before I multiply by .707 the noise is significantly lower in the subtraction. I can actually see vertical banding in the red channel in the white patch with IRIS' whitepoint just above that level.
Click to expand...
Is the vertical banding consistent from shot to shot -- is it a gain fluctuation for individual channels, or is it something else?
Click to expand...
I don't see much in the way of artifact after subtracting one image from its twin, and the remaining noise looks like a nice, dense starfield (fixed pattern offset read noises are subtracted along with scalar ones).
Click to expand...
I was asking whether, without taking a difference image, you were seeing vertical banding in a given RGGB subarray that could be attributed to column variation of gain, and if so is it stronger than in previous models. This would be visible for instance upon averaging a bit in the vertical direction.
John Sheehy said:
Now, when are converters going to allow subtraction of a correction file before converting?
Click to expand...
It is my hope that a future release of PerfectRAW will allow such a possibility. There is also word that RawTherapee is going to be GPL'd, so some enterprising coder could add such a feature, since RT and PR are both based on dcraw which has a blackframe subtraction capability.

--
emil
--



http://theory.uchicago.edu/~ejm/pix/20d/
 
ejmartin said:
How much flat field correction does the 7D need?
Click to expand...
The sigmas plummet when you subtract an OOF ColorChecker from its twin; that's even before you multiply by .707.

IIRC, some dropped by about 80%.

The 7D's minor RAW ills are highly correctable - so close, but so far.

--
John

 
GeorgeML said:
Won’t discuss image level noise since I’m not into resizing.
Click to expand...
Every time you print you (or your software) resizes. And if you're not viewing at 100% on screen you are resizing. Any practical use of an images involves image resizing. Per pixel noise is only of interest to people who solely view images 100% on scree.
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top