Home · Register · Join Upload & Sell

Moderated by: Fred Miranda
Username  

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Canon Forum | Join Upload & Sell

1       2       3      
4
       5       6       end
  

Archive 2012 · 5D vs 5DM3 High ISO

  
 
Shutterbug2006
Offline
• • • •
Upload & Sell: Off
p.4 #1 · p.4 #1 · 5D vs 5DM3 High ISO


snapsy wrote:
The extra 2 bits are wasted because they're just oversampling noise (at least on current Canon sensors).


Where do you get that information?



Jan 01, 2013 at 01:15 PM
Shutterbug2006
Offline
• • • •
Upload & Sell: Off
p.4 #2 · p.4 #2 · 5D vs 5DM3 High ISO


Directly from Canon ....

"This extra image information might be overkill if it weren’t for the advanced CMOS image sensors found on both the EOS 50D and EOS 5D Mark II. The pixels on both of these sensors have improved fill factors and lower read noise than those in previous Canon sensors in their class, with the larger-sized pixels on the EOS 5D Mark II providing low-noise performance that’s a notch better than the EOS 50D at every comparable ISO setting. In combination with the camera’s lower levels of background noise and amplifier noise, the EOS 5D Mark II’s 14-bit A/D converter helps to provide smoother tonal gradations, along with the potential to increase detail in bright highlights with the Highlight Tone Priority option. The EOS 5D Mark II also maintains a slightly higher dynamic range at high ISOs than the EOS 50D and the older EOS 5D it replaces.

The expanded dynamic range provides increased detail in deep shadow areas and a superior ability to open up shadows in post processing without banding and excessive noise. Detail is also improved in highlight regions, but the difference is less noticeable as a result of inherently lower noise levels in bright tones and highlight areas. However, having extra bit depth and detail in highlight areas allows a photographer to adjust exposure and contrast in post processing to reduce blown out areas without blocking up shadow areas."

"The potential for smoother tonal transitions, was just mentioned above. Any time you have a scene with a single “color” that moves from light to dark or vice versa, this is a benefit — whether it’s a wide-angle shot of a clear blue sky or a shot of a car with shiny highlights blending into darker tones along its painted body. Skin tones in portraits, especially with “hard” light sources like direct sun or on-camera flash, are another area where 14-bit conversion can mean smoother changes from specular highlights to diffused highlight skin tones.

The increased information in each file also allows cameras like the EOS 5D Mark II and EOS 50D to incorporate Canon’s Highlight Tone Priority option. With much more tonal data, Highlight Tone Priority leverages this information and preserves anywhere from 1/3 to a full stop of additional detail in bright highlights. It can make the difference between washed-out highlights vs. controlled highlights with visible texture and detail. And unlike simply changing exposure to darken an image, Highlight Tone Priority does not change the brightness level of mid-tones or shadows. There’s a chance of a small increase in digital noise in darker areas, but even at high ISOs, it’s a minor difference, and a small price to pay for the potential of added detail in bright highlight areas.

With the expanded tonal information provided by the 14-bit per channel conversion, it’s possible for the camera to make tone curve adjustments on a case-by-case basis. In addition to toning down highlight areas, underexposed areas can be lightened. This is the function of the Auto Lighting Optimizer, a new feature found on the EOS 5D Mark II, EOS 50D, and also on the EOS Rebel XSi and EOS 40D cameras. JPEG images can be fine-tuned as they’re processed and recorded; RAW images from these cameras can have tonal adjustments applied when they’re processed in Canon’s Digital Photo Professional software. Users are free to disable this feature, or set Low, Standard, or Strong levels in the EOS 50D and 5D Mark II.

Another technology that these new cameras utilize is built-in correction for lens vignetting, called Peripheral Illumination Correction. Since tones in sky or other light-colored outer areas of images have a lot more tonal information and much finer steps between them, the camera can automatically apply correction to lighten these areas without risking tonal “skipping”, banding, or noticeable increases in digital noise. Both Highlight Tone Priority and Peripheral Illumination Correction can be applied to JPEG as well as to RAW images."

"It’s a fact that different color dye or light combinations can produce similar visual sensations, which is why we are able to produce accurate colors in print, even though printing inks differ in their density and color, and printing methods range from silver halide to CMYK printing presses. The same holds true for different display technologies, paints, and clothing dyes. Metameric color pairs are those that appear exactly the same under one light source but different under another, proving that the color sensation we see is being formed by color components with different spectral properties.

A 14-bit A/D converter has the ability to map the 10-12 million true colors captured by the camera into trillions of color data combinations. This may provide a benefit down the road in a color-managed system when converting a RAW file or 16-bit TIFF file for viewing on wide variety of color displays or printing on wide gamut color printers. While a 14-bit A/D converter might not be able to produce any more “visual” colors from RAW data than a 8 or 12-bit A/D converter, its increased precision allows for higher color accuracy and smoother gradations between colors, plus additional details in dark, saturated colors that can be attributed to the increased dynamic range.

The increased precision of the 14-bit A/D converter in combination with the computational speed and power of the DIGIC processor also helps images stored in the finest quality JPEG format within the camera, although those improvements are more subtle. When using the RAW+JPEG option, photographers have the best of both worlds. JPEGs can be viewed and e-mailed quickly, while top selections stored in RAW will benefit from the advanced RAW conversion algorithms found Canon’s Digital Photo Professional software.

This is where access to the additional gradations and dynamic range gives serious photographers the ability to adjust shadow and highlight details that would have otherwise been tossed away, or create smooth gradations that might have posterized during the in-camera conversion to an 8-bit JPEG. When satisfied, the higher bit-depth image can either be stored in a non-destructive file format such as a 16-bit TIFF, or output directly to Canon imagePROGRAF ink jet printers using Canon’s advanced 16-bit export modules for Photoshop and Digital Photo Professional software.

As technology progresses, more and more display devices and printers will be compatible with the higher bit-depth image data captured by the EOS 50D, and EOS 5D Mark II, and similar cameras, and use it to automatically fine tune image quality while taking advantage of expanded dynamic range these cameras offer."

I see the difference between an 8-bit, 12-bit and 14-bit file, but perhaps it's because I'm looking for it. It's not a huge, but I believe Canon when they say having the extra data can be beneficial.




Jan 01, 2013 at 01:24 PM
gfiksel
Offline
• • • •
Upload & Sell: Off
p.4 #3 · p.4 #3 · 5D vs 5DM3 High ISO


That piece of Canon information doesn't say anything about noise, which is the main limiting factor in the usefulness of the higher-bit images. When the noise exceeds the quantization, the extra bit do not contain any useful information - the same noise just broken into smaller pieces.


Jan 01, 2013 at 01:40 PM
snapsy
Online
• • • • • •
Upload & Sell: On
p.4 #4 · p.4 #4 · 5D vs 5DM3 High ISO


Shutterbug2006 wrote:
Directly from Canon ....
"The potential for smoother tonal transitions...


The operative word is potential, which is not reached/needed. A good resource on the subject can be found here. And here's image data taken from actual Canon raw with various simulated bit depths. In fact the higher tonal areas are so oversampled that some companies (Sony) actually compress that data in their raws with no perceptual loss of tonality.



Jan 01, 2013 at 02:39 PM
mttran
Offline
• • • • •
Upload & Sell: Off
p.4 #5 · p.4 #5 · 5D vs 5DM3 High ISO


snapsy wrote:
My Canon shadow pushing days are behind me


I know exactly what you mean. Thanks



Jan 01, 2013 at 03:27 PM
Pixel Perfect
Offline
• • • • • • •
Upload & Sell: On
p.4 #6 · p.4 #6 · 5D vs 5DM3 High ISO


While I often take marketing brochures with a dose of salt, IMO the 5D II clearly has better shadow detail and shows far less posterization in areas with fine tonal gradients like the sky when you are post processing, compared to the 5D. I believe the 14 bit data is not a total waste of time. I can push the sahdows further on 5D II and 5D III than I could with 5D.


Jan 01, 2013 at 05:12 PM
skibum5
Offline
• • • • • •
Upload & Sell: Off
p.4 #7 · p.4 #7 · 5D vs 5DM3 High ISO


gfiksel wrote:
That piece of Canon information doesn't say anything about noise, which is the main limiting factor in the usefulness of the higher-bit images. When the noise exceeds the quantization, the extra bit do not contain any useful information - the same noise just broken into smaller pieces.


+1



Jan 01, 2013 at 06:28 PM
Shutterbug2006
Offline
• • • •
Upload & Sell: Off
p.4 #8 · p.4 #8 · 5D vs 5DM3 High ISO


Pixel Perfect wrote:
While I often take marketing brochures with a dose of salt, IMO the 5D II clearly has better shadow detail and shows far less posterization in areas with fine tonal gradients like the sky when you are post processing, compared to the 5D. I believe the 14 bit data is not a total waste of time. I can push the sahdows further on 5D II and 5D III than I could with 5D.


I don't expect Canon spent millions of dollars developing better A/D converters and DIGIC processors to handle the 14-bit files from the newer camera's just to pull the wool over our eyes.

Clearly though there are a few folks here who know better than Canon and should be offered a job.



Jan 01, 2013 at 06:34 PM
Pixel Perfect
Offline
• • • • • • •
Upload & Sell: On
p.4 #9 · p.4 #9 · 5D vs 5DM3 High ISO


Shutterbug2006 wrote:
I don't expect Canon spent millions of dollars developing better A/D converters and DIGIC processors to handle the 14-bit files from the newer camera's just to pull the wool over our eyes.

Clearly though there are a few folks here who know better than Canon and should be offered a job.


Part of the problem is that Canon's so-called 14 bit A/DC's aren't really true 14 bit A/DC's, so they don't offer the full potential of 14 bit data, which is what they mean about the extra 2 bits being wasted. IMO they are not totally wasted.



Jan 01, 2013 at 06:41 PM
snapsy
Online
• • • • • •
Upload & Sell: On
p.4 #10 · p.4 #10 · 5D vs 5DM3 High ISO


Shutterbug2006 wrote:
I don't expect Canon spent millions of dollars developing better A/D converters and DIGIC processors to handle the 14-bit files from the newer camera's just to pull the wool over our eyes.

Clearly though there are a few folks here who know better than Canon and should be offered a job.


Does that mean you read the Emil Martinec article I linked and found errors in his calculations or his understanding of information theory? Or that you performed your own experiments which contradict his findings?



Jan 01, 2013 at 07:17 PM
snapsy
Online
• • • • • •
Upload & Sell: On
p.4 #11 · p.4 #11 · 5D vs 5DM3 High ISO


Pixel Perfect wrote:
Part of the problem is that Canon's so-called 14 bit A/DC's aren't really true 14 bit A/DC's, so they don't offer the full potential of 14 bit data, which is what they mean about the extra 2 bits being wasted. IMO they are not totally wasted.


It's not entirely wasted apparently - some still upgrade their cameras thinking it makes a difference. And in the end that's all that counts from Canikon's perspective.



Jan 01, 2013 at 07:19 PM
corndog
Offline
• • • •
Upload & Sell: Off
p.4 #12 · p.4 #12 · 5D vs 5DM3 High ISO


snapsy wrote:
Does that mean you read the Emil Martinec article I linked and found errors in his calculations or his understanding of information theory? Or that you performed your own experiments which contradict his findings?


One area he points to is the error in rounding off in the ADC, but wouldn't the rounding off be far more accurate in a 14-bit ADC, instead of a 12-bit? If so, wouldn't that be a huge advantage in reducing posterization?

Great article btw!



Jan 01, 2013 at 08:44 PM
snapsy
Online
• • • • • •
Upload & Sell: On
p.4 #13 · p.4 #13 · 5D vs 5DM3 High ISO


corndog wrote:
One area he points to is the error in rounding off in the ADC, but wouldn't the rounding off be far more accurate in a 14-bit ADC, instead of a 12-bit? If so, wouldn't that be a huge advantage in reducing posterization?

Great article btw!


The typical analogy is someone who wants to count the number of cars passing by a point on the highway per hour, using only a random 60 second sample. Lets say a single 60-second sample yields a count of 30 cars. If traffic is following a Poisson distribution, the same as how photons arrive and are counted on an image sensor, the standard deviation of that 30-car sample would be 5.47 cars. That means additional 60-second samples might yield counts anywhere from 24.53 cars to 35.47 cars. Now lets say you need to report those 60-second samples to someone - how much numerical precision would be needed to represent those samples such that no precision would be lost relative to the precision provided by the measurement mechanism itself? Would you need precision to the hundreths of cars, like 28.55? Or tenths of cars, like 28.5? Or single digits, like 28?...when the precision of the measurement mechanism itself is only accurate to +/- 5.47 cars?



Jan 01, 2013 at 09:46 PM
corndog
Offline
• • • •
Upload & Sell: Off
p.4 #14 · p.4 #14 · 5D vs 5DM3 High ISO


Ok, great analogy for getting the point across, but how does one know the resolution of the sampling? Seems like one would need to have intimate knowledge of Canon firmware, as well as getting their hands on the specific model ADC datasheet, no?




Jan 02, 2013 at 12:10 AM
snapsy
Online
• • • • • •
Upload & Sell: On
p.4 #15 · p.4 #15 · 5D vs 5DM3 High ISO


corndog wrote:
Ok, great analogy for getting the point across, but how does one know the resolution of the sampling? Seems like one would need to have intimate knowledge of Canon firmware, as well as getting their hands on the specific model ADC datasheet, no?


Good question. The author of that first article has also written an article describing a procedure by which the standard deviation of noise from two raw images can be calculated and used derive the relationship between ADC units (data values in the raw file) and photon counts, from which other sensor performance data can also be calculated. The technique can be applied to any camera for which raw files can be decoded, plus there are additional shortcuts available depending on the brand. The article is here. These days it's much simpler just to rely on DxoMark's measurements, which have been verified by independent testers.



Jan 02, 2013 at 01:02 AM
vscd
Offline

Upload & Sell: Off
p.4 #16 · p.4 #16 · 5D vs 5DM3 High ISO


I followed the thread with great interest because this was exactly the stuff I wanted to read. I'm currently using the old 5D because I was always pleased with the results and I don't need video or newschool functions. Even liveview is overestimated except of shooting macros, but I got over it with some experiences. The old 5DM2 had (camerawise) no real update because the old AF was build in and the higher resolution was proner to noise.
Now there is a new update available and the same questions arise again. I guess after 7 years and 2 evolutionsteps I would see the difference but I'm still wondering if 12 MP on a fullframe are still to beat. I talked to a professional photograph who's using a Nikon D700 (nearly equal to the 5Dc-Specs) and he doesn't want to upgrade either. Mostly of the bigger RAW of a D800 and the mass of storage and CPU-Power need just to work with the files. There are even just 2 or 3 lenses available to get all the pixels done.

At the moment I think until ISO1600 the 5Dc can still be compared to the new cameras, everything else is overrated by the Marketing which wants everyone to buy new stuff. From ISO 3200 on the newer sensors may rise... but I think this is all based on the old chips from 2004 and later. Canon wants to bring out a new generation this year, mostly because of Dynamic Range issues compared to Nikon. I hope there will be larger steps forward on noise, too.

Except from comparing noise, what about the Diffraction Limited Aperture of the 5DM2/5DM3? You can use apertures up to f13-f15 on the old 5D, getting limited to f11 on the new one (f8 for the D800). Those are problems while shooting macros requiring a bit DOF.

And I'm even still curious about the open letter to Canon and Nikon about the real T-Stops of the chips (LINK). So my 5Dc with 85mm 1.2L II maybe is up again by 1 or 2 steps?

I guess the 5Dc is still limited by the photograph with exceptions of a few of course



Jan 04, 2013 at 11:41 AM
Shutterbug2006
Offline
• • • •
Upload & Sell: Off
p.4 #17 · p.4 #17 · 5D vs 5DM3 High ISO


vscd wrote:
And I'm even still curious about the open letter to Canon and Nikon about the real T-Stops of the chips (LINK). So my 5Dc with 85mm 1.2L II maybe is up again by 1 or 2 steps?


Interesting discussion. Thanks for the link.



Jan 04, 2013 at 03:55 PM
cputeq
Offline
• • • • •
Upload & Sell: Off
p.4 #18 · p.4 #18 · 5D vs 5DM3 High ISO


I'd give it a 1.5 stop advantage, to my eyes.

Impressive for both cameras - the 5Dc 1600 photos are nice for such an age, and 5D3 pulling off similar results at around ISO 5000-ish (I won't do the math) are also impressive - plus you've got more MP to work with if you want to "fake the funk" later on for web or small print use through downscaling



Jan 05, 2013 at 03:28 PM
vscd
Offline

Upload & Sell: Off
p.4 #19 · p.4 #19 · 5D vs 5DM3 High ISO


I just read the review of the 5Dc at dpreview.com again (LINK). Interesting Point (5th) at the list of Conclusion-Pros:

"About a third of a stop more sensitive than indicated".

Could be the effect of t-stops I posted before or at least an interesting fact (LINK). This may give a further small advantage for the classic 5D. You're still amazing my old granny



Jan 06, 2013 at 04:37 PM
snapsy
Online
• • • • • •
Upload & Sell: On
p.4 #20 · p.4 #20 · 5D vs 5DM3 High ISO


vscd wrote:
I just read the review of the 5Dc at dpreview.com again (LINK). Interesting Point (5th) at the list of Conclusion-Pros:

"About a third of a stop more sensitive than indicated".

That was true for the 5D vs 5DM2 but on the 5DM3 Canon increased the actual vs nominal ISO and so it's pretty close to the 5D.

vscd wrote:
Could be the effect of t-stops I posted before or at least an interesting fact (LINK). This may give a further small advantage for the classic 5D. You're still amazing my old granny


The t-stop loss the difference between the 5D and 5DM2 @ f/1.4 is only 0.20EV according to the DxO article. I visually compared the 5D and 5DM3 @ f/1.4 and couldn't see a noise difference, which makes sense because 0.20EV would be hard to visually distinguish.



Jan 06, 2013 at 04:58 PM
1       2       3      
4
       5       6       end




FM Forums | Canon Forum | Join Upload & Sell

1       2       3      
4
       5       6       end
    
 

You are not logged in. Login or Register

Username       Or Reset password



This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.