So I've been contemplating the current crop of camera's out there these past few and it seems as though the choices have never been tougher. On the one hand
we have the faithful D810 with very good resolution, noise and phenomenal dynamic range. Then we have the A7R II sporting an all new sensor and which seems
to come so very close to the D810 in terms of noise and DR but with the benefit of added resolution. And then we have the 5DS/r. Canon's newest high resolution
wonder. With an all new sensor that promises better noise handling and dynamic range than any other Canon sensor to date.
But the question remains as to how good Canon's sensor really is in real world application as the studio samples seem to indicate that it's showing more noise
than the competition at the same sensitivity. And so I set out to try and figure this out without having to go out and buying these camera's myself.
As a landscape shooter, my main concerns revolve mainly around super clean files with as as much post processing latitude as possible. This usually involves
RAW files a very low noise ceiling, and very good dynamic range. Especially in the shadow areas. Since most of my landscape shots involve shadow recovery
to some degree or another as I compensate for lighting throughout the scene. Which will typically run anywhere between 1 and 3 stops in most cases.
And so to test my theory, I downloaded the RAW files for each system from Imaging Resources online and DPReview. Though I must add that I'm finding
DPReviews Studio samples to be rather unintuitive given that the entire setup rests on a vary shallow focal plane. Which in turn, doesn't leave much room
for spacial definition.
Whatever the case, it was at this stage that I began to see just how impressive the 5DSr RAW files actually are. And while there's no denying that there is
noise present at nominal ISO, the impressive part, is just how well the files clean-up and resist push processing afterward. ie, one aspect of higher resolution
sensors is that the noise patterns follow pixel definition. Which in turn, allows for better noise reduction in cases where the software makes use of
neighboring pixel data for edge and color reconstruction. Such as with Topaz Denoise for example. To which I'd add, "really" seems to like the 5DSr files.
And while I haven't had much time to play with the files beyond that of my own quick and dirty post processing methods, I'd add that these may be the best
high ISO noise reduction files I've worked with short of the 645z. And yes, they work out to be better than the A7R II and D810. though I'd also add that
I haven't gone beyond ISO3200 to date. And so this could change as I enter the upper end of the sensitivity range.
Anyways, in getting down to it, here's an ISO3200 image scaled to match the A7S which I thought was worth sharing. Granted the backdrop textures were
lost due to a heavy handed mid-tone NR. Though that's an easy fix with a little adjusting and so I'm expect this file to get even closer to nominal. Which is
no small feat for a 50mp FF sensor imo.
ISO3200 RAW, Photo Ninja Beta, Resize Magic, Topaz Denoise and horrible color profile conversion. More to come...
I also wanted to add*(forgot) that one interesting aspect of the 5DSr files is that while the shadow areas contain more noise than with the D810, that the
limited recovery I did, seem to contain more information. ie, the D810 being flat and devoid of information, whereas the 5DSr seem to contain more tonal
information. - Make of this what you will, but this is what I've observed to date.
Last edited by JohnBee; 08-17-2015 at 02:00 AM.