Cabnon 5DS-R pixel size, lens resolution and the law of diminishing returns

Brachytron

Brachytron

Messages
1,823
Name
Mark
Edit My Images
Yes
I've been looking forward to DXO Mark publishing their perceptual megapixel (PMP) ratings for the 5DS-R to do a comparison with the 7D2, because they have very similar pixel sizes. Using the great Canon 24-70mm f2.8 on the 5DS-R gives "only" 32PMP, a drop of 36% from the full sensor resolution. On the 7D2, the drop is 40%, to 12PMP - I guess the difference is due to the anti-aliasing cancellation filter on the 5DS-R. So, as I expected the ratio of PMP to MP is very similar. This being the case, viewing images from both cameras at the same size (up to the limits of the 7D2) should give equivalent perceptual resolution and the only benefits of the high MP sensor at print size and cropping. The Nikon D810 with Nikon 24-70mm f2.8 scores 21PMP out of 36 (a 42% drop).

Taking this a stage further, I am also keen to understand more about the link between lens resolution and pixel size to examine the question "Does the lens out resolve the sensor?". Looking through DXO Mark at a range of cameras with the same Canon 24-70mm f2.8, shows that since the 5D classic, the PMP has dropped as a % of the sensor resolution. The first figure is DXO Mark's PMP, then camera MP, then PMP as % of sensor resolution:

5D 12/12 100%
5D III 18/22 82%
5DS-R 32/50 64%

This all assumes that we believe in the DXO Mark PMP scores!

The laws of MTFs mean that increasing sensor resolution will always improve total system resolution, but I wonder how long it will be and at what MP rating that further increases just aren't worth it in practice? I would be interested to here your views on this. Thanks.
 
5DR2 50/100 50%

-- looks in line with your previous figures, and looks like a worthwhile increase, as worthwhile as any of he previous increases.
 
Last edited:
I looked at the 3 pics which Canon offered before the release of the 5D3 ( a hippo, a still life studio shot and a view taken from the top of a building) and although the resolution was spectacular it seemed to me that the technology was approaching its limit.

Even at 400 ISO there was evidence of noise in the sky in one of them and the studio picture seemed to be hard pressed to produce decent dynamic range resulting in a rather "forced" look.

Obviously some of this could be compensated for in PP but I rather doubt that we will be seeing much increase beyond that pixel density in the near future.
.
 
I just did a quick check, and with a 54mm internal diameter and a flange to sensor distance of 44mm, the EF mount has an effective NA of about 0.5. This means that if you placed the final optical element here (and ignored all requirements to mount it effectively), the the theoretical maximum resolution given purely by the geometry of the mount it about 0.5microns. So to sample this optical resolution, you'd ideally have pixels smaller than 0.25 microns. On a full frame sensor that would be a 13 GIGApixel resolution :) Of course, that's purely an upper limit that purely looks at what the light can do and ignores all engineering requirements. But my guess is that there might still be a little bit of headway possible between the current state of the art and future limits!
 
I'd be interested in the limitations imposed by shutter shock and the granularity of AF fine tune.
 
:)

Well, for absolute acuity we generally use a tripod with mirror lock-up, but even then we still have the tiniest of movement from the shutter action; whilst this affects the whole camera as one unit, you'd think that the blur would start to be perceptible at a pixel level at some threshold even on the most locked down tripod.

And regarding AF fine tune - most lens / DSLR body combinations need some AF fine tuning for optimal acuity. This is normally on a -10/+10 scale (if memory serves me correctly), I just wonder what one increment is in this case and whether the granularity of this is enough to get the best acuity (i.e. does the actual optimum setting fall between eg. +3 and +4 on a given combo). Of course you could just shoot in live view and not worry too much about that (which you would be anyway in mirror lock-up).
 
Some people spend to long thinking about gear and not enough time taking pictures! Is there any chance of dxo going bust?
 
These things might be of interest to people who only ever create ultra sharp images of test charts, news print and brick walls. For those of us silly enough to want to take photographs of real things, I think the cameras and lenses available far exceed our abilities to use them.


Steve.
 
DXO bashing is all well and good but the fact is that not all gear is created equal and therefore bench test measurements are at least something to add to the list of things to think about.

If for example a company, lets call them Bazooka, and another company called Fony both make cameras which are capable of taking pictures it might be nice to have some performance data to read so that we can add it to the mix and an informed choice. We may decide that although the Fony is technically better for other reasons the Bazooka is right for us and that's a perfectly valid way to proceed but at least we've been armed with information to aid our choice and that has to be a good thing.
 
frod said:
:)

Well, for absolute acuity we generally use a tripod with mirror lock-up, but even then we still have the tiniest of movement from the shutter action; whilst this affects the whole camera as one unit, you'd think that the blur would start to be perceptible at a pixel level at some threshold even on the most locked down tripod.
Click to expand...

I first noticed shutter shock on a 14MP APS-C DSLR, the Sony A350, at shutter speeds of around 1/250th sec or longer, on a 500mm lens. It caused ghosting and smearing of sharp fine detail over a few pixels, at first vertically, then as exposure lengthened, all around. No amount of secure support of the camera and lens helped, so it was clearly internal vibration inside the camera. At first I thought it was mirror shock, but later experiments with a 24MP A77, which doesn't move its pellicle mirror, and which is switchable between mechanical shutter and electronic first curtain, showed exactly the same effect at the same speeds, which completely disappeared when I switched out the mechanical shutter. It was a revolution in terms of long lens sharpness. I could get really crisp sharp detail with the 500mm lens on a tripod right down to the longest exposures the light let me try, large fractions of a second. I'd never been able to do that before, and had thought the problems at the lower shutter speeds were just an inferior tripod and so on. The shutter shock blurring was noticeable at 200mm as well. I hadn't been able to see that on the 14MP camera, partly because of the reduced resolution, and partly because at that time I had an inferior long zoom which was going soft at 200mm.

And regarding AF fine tune - most lens / DSLR body combinations need some AF fine tuning for optimal acuity. This is normally on a -10/+10 scale (if memory serves me correctly), I just wonder what one increment is in this case and whether the granularity of this is enough to get the best acuity (i.e. does the actual optimum setting fall between eg. +3 and +4 on a given combo). Of course you could just shoot in live view and not worry too much about that (which you would be anyway in mirror lock-up).
Click to expand...

In my experience of AF fine tune the adjustments are more than fine enough. A much more important problem is that I've yet to find a lens with really shallow DoF where the amount of mechanical play between the two directions of movement doesn't exceed the DoF. So even with AF lock on, manual focus off, I can twitch the lens in and out of focus just using the mechanical play. Also it appears to be the case that the smallest possible increment of the AF stepping motor is about the smallest DoF. That's what I'd expect from an engineer told to squeeze the maximum AF speed and lowest cost without compromising the fineness of tune needed for the shallowest DoF -- not quite as fine as the most critical eye would prefer. I also note with all of my lenses that I've bothered to check, the position at which the AF thinks it's got it varies a little between repetitions, and varies quite a lot depending on which end you approach focus from, infinity or close, enough in that case for one to in sharp focus and the other quite definitely out.

In other words, there are definite limits to the classic DSLR phase-based AF which current high resolution sensors and sharp shallow DoF lenses are showing up.

No, I don't just photograph test charts and brick walls. I'm interested in the real life problem of seeing the feather detail on a small bird perched in the shade of a tree or bush which is far enough away that I'll have to crop down a lot even with a 500mm lens, and need all the sharpness and detail resolution I can coax out of my camera and long lenses. If there was a 50MP crop sensor camera with a 300-900mm zoom I'd be interested.
 
You'd do much better by ignoring DXO and kit altogether and start reading up on Airy discs and diffraction limited systems...
 
woof woof said:
DXO bashing is all well and good but the fact is that not all gear is created equal and therefore bench test measurements are at least something to add to the list of things to think about.

If for example a company, lets call them Bazooka, and another company called Fony both make cameras which are capable of taking pictures it might be nice to have some performance data to read so that we can add it to the mix and an informed choice. We may decide that although the Fony is technically better for other reasons the Bazooka is right for us and that's a perfectly valid way to proceed but at least we've been armed with information to aid our choice and that has to be a good thing.
Click to expand...
I agree, look at the sensors in Nikon compared with Canon. Nikon are better, yet Canon sell more cameras. I don't like dxo sensor scores, for example the 5DS-R has a better ISO score than the 7D2 because it's dynamic range is better.
 
Steve Smith said:
These things might be of interest to people who only ever create ultra sharp images of test charts, news print and brick walls. For those of us silly enough to want to take photographs of real things, I think the cameras and lenses available far exceed our abilities to use them.


Steve.
Click to expand...
Totally agree, but that doesn't stop the science being interesting.
 
You must log in or register to reply here.
Back
Top