aCuria wrote:
Yes you should look at the raw files. I linked to them together with the crops.
If you start with a 16-bit file and then reduce the bit depth, it’s kind of like supersampling. You’ve got a lot more tonal information to begin with, so when you downsample to 8-bit it still looks really good. That’s why 8-bit exports from high bit-depth sources often hold up better than people expect.
Because of this, tests where you just take a high-bit-depth file and reduce the bit depth aren’t that meaningful. You’re still benefiting from all the extra precision upstream. If you really want a fair comparison, you need to look at sensors that are natively 8, 10, 12, 14, or 16-bit and compare those directly.
Where lower bit depth really starts to show is when you begin editing. That’s why white balance adjustments look much better on 10-bit video than on 8-bit, and why a 14-bit RAW photo holds up even better when you push it. You’ve got more headroom before things start to break apart.
If the original 16-bit image has smooth gradation, especially in areas like skin tones, then exporting to 8-bit for the web will still look great. The tonal values get mapped through the gamma curve, and you can retain visually smooth transitions even in 8-bit.
But if the original file already has poor gradation, an 8-bit export will just faithfully preserve those limitations, and it’ll look flat or banded as a result. Garbage in garbage out. ...Show more →
Point number 3 pretty much sums up what I was trying to get at in my earlier comments that 'other factors' likely have a much greater impact on the differences in image quality between the various sample images.
Point number 3 pretty much sums up what I was trying to get at in my earlier comments that 'other factors' likely have a much greater impact on the differences in image quality between the various sample images.
Blog post Point number 3:
photographic sensors do not generate color information with 16-bit fidelity. The signal is already quantized, and color differences at the bottom 2 bits of a 16-bit file are buried in noise.
I agree!
aCuria wrote:
Personally I doubt the Canon is capable of clean 14 bit, and I doubt the x2d is capable of clean 16 bit. However, the x2d is likely legitimately 2 bits better than the canon, that's probably what we are seeing here.
In fact I said pretty much the same thing in this thread! (see above quote)
Blog post Point number 1:
A 16-bit file can, in theory, encode 96 dB of dynamic range versus 84 dB for 14-bit.
I am not very happy at all about how he presents this. In practice it cannot be done. 84db of analog signal cannot be saved into a 14 bit file without an "ideal" ADC.
Since our ADCs are not "ideal", headroom has to be built into the pipeline.
The blog post claims that the MF sensor is capable of true 14 bit signal. Using a 16 bit ADC and pipeline makes sense for this kind of sensor.
I recently upgraded (about two months ago) from the R6 Mark II to the R5 Mark II. The main reason for the upgrade is that I shoot a lot of birds and often need to crop heavily. I absolutely love the 45 MP sensor—it gives me plenty of headroom to crop smaller frames without losing image quality.
Pre-capture is a blessing when used properly. I don’t keep it turned on all the time; I use it only when needed. I do wish Canon would update the firmware to allow assigning pre-capture to a custom button so it can be toggled on/off easily.
I definitely see a noticeable improvement in autofocus compared to my R6 Mark II. I believe the R6 Mark III is probably quite similar in this regard. As for frame rates, anything beyond 30 FPS becomes overwhelming to sort 30 FPS is already more than enough.
The only area where I’m a bit disappointed is noise performance. Canon still has a long way to go compared to Sony. In low light, images beyond ISO 1600 aren’t great in my experience. People say you can shoot up to ISO 4000–6000, but I really wonder how practical that is.
@Vksrik - What is your high ISO noise reduction approach with the R5II?
I frequently use it at ISO 3200 and higher, but I run everything through Adobe Denoise. I downloaded the DP Review studio scene comparison RAW files at ISO 6400 from the R5II, Sony a1, a7RV and Nikon Z9 and ran them all through default Denoise (50% strength). In respect to noise characteristics, it's splitting hairs IMO between Canon and Sony. The Nikon appears a lot smoother with the least fine detail retention. The a7R has somewhat of a resolution advantage and the a1 appears slightly sharper at the pixel level compared to the R5II, perhaps due to the R5II's antialiasing filter?
IMO there are other reasons to pick one camera/system over the other.
If you're referring to low ISO noise performance with shadow recovery, then yes, Sony has an advantage. A 'problem' with the a7R series is it's currently a very poor fast action e-shutter option and the a1 series is in a higher price category than the R5II, if buying new, though I believe you can get the original a1 at used prices comparable to the R5II.
I run everything through Denoise now, even ISO 100 files. It slightly 'tightens up' fine details and is effective at controlling Canon low ISO shadow noise when pushing files, within reason. At high ISO it's a great solution (IMO) for generally cleaning up files but one has to be realistic that it will not recover fine details originally lost to high ISO noise. You can see this in the fuzzy green artificial foliage in the test scene with all four cameras, to varying degrees. It appears to work best when you can give it as much detail/information as possible. IOW, as sharp a file as possible to start with will result in the best AI noise reduction. Otherwise it will kind of improvise, though it's probably still better than non-AI NR options.
Here is the DPR studio scene page where you can make camera and ISO selections.
Thanks for the info and links! I use Photoshop to edit my bird images. I work with layers, use Neat Image to remove noise from the background layer, and mask the subject to protect fine details.
A friend of mine shoots with a Sony A1, and he confidently shoots at ISO 6400. In my opinion, with Canon, the images are usable at that ISO, but they’re not as clean as Sony’s.
I also use Adobe Camera Raw or Lightroom Classic noise reduction for landscape photography.
Vksrik wrote:
A friend of mine shoots with a Sony A1, and he confidently shoots at ISO 6400. In my opinion, with Canon, the images are usable at that ISO, but they’re not as clean as Sony’s.
It would be interesting to see R5II and a1 side by side images processed through the same NR processes to see what differences remain. Just from the DPR studio scene, the differences between the two seem minimal.
Vksrik wrote:
I also use Adobe Camera Raw or Lightroom Classic noise reduction for landscape photography.
FYI, there is a pretty significant difference between Adobe's modern AI Denoise process and the traditional manual denoise controls in ACR and Lightroom. I'm not sure if you are referring to the former or the later here, but I think if you have access to the latest AI Denoise feature, you will find that your ISO 6400 images clean up very well.
It's also worth mentioning that any comparisons of noise between cameras should be performed at common output sizes. Higher resolution bodies will generally show more noise at the pixel level than lower resolution bodies, but when output size is normalized to that of the lower resolution camera, these differences often disappear or are greatly reduced.
(And you may already know all of this, of course.)
wondering the same. Also, I'm quite curious to hear from someone who owns both how the build quality differs, and how the experience of the difference of the EVF is
