Canon 5DS and 5DSR Cameras -- Too Much?

[cooltouch]

Well, with the release of these two new 50.6mp FF DSLRs, Canon has regained top bragging rights in the megapixel wars. Earlier today, I watched a rather lengthy presentation given by Canon USA in which it was emphasized repeatedly how these new cameras would provide unsurpassed image resolution, how they would rival medium format digital cameras or digital backs, while providing a more convenient form factor and a substantial savings.

Well, this all sounds really great -- until one begins to think a bit more critically. Like, for instance, are there any lenses available that can resolve the detail that this huge new sensor can store? Back when Modern Photography magazine still existed, their lens tests always listed resolution in line pairs per millimeter (lppmm). Any lens that managed lppmm numbers in the high 60s always scored "Excellent." Only a very tiny few lenses tested had resolution numbers which exceeded 70 lppmm. And if you crunch the numbers, with a full frame sensor, 70 lppmm translates into 16.9 megapixels.

So, is Nikon building optics that are way sharper now that can handle the 800-series cameras 36mp? Or now that Canon has introduced a 50mp sensor, can we assume that Canon is now producing lenses that will resolve the amount of detail a 50.6mp sensor can hold?

Honestly, I doubt it very much. A 50.6mp sensor would require lenses with over 120 lppmm resolution in order for the sensor's output to be fully realized. That's almost twice the resolution that the old Modern Photography magazine considered to be "excellent." Chances are, the only lenses on Earth that will get close to these resolution numbers are highly specialized process lenses that are good for only a very narrow range of tasks.

You know what this reminds me of? These huge new sensors remind me of modern flatbed scanners with their very high claimed resolution numbers -- like 4800 ppi (as with the Epson 4870 and 4990) or 9600 ppi (as with the 700-series Epsons), when even the best of them test out at barely a tad over 2000 ppi. All the leftover resolution just ends up as file bloat. Which makes me wonder if the same thing is happening to digital image files internal to the camera.

You know, Nikon was smarter by half when they chose a 16.2mp sensor for their classy, retro DF FF DSLR. A 16.2mp sensor is "matched" to the resolution of some of the best lenses around. Nikon realized that there was little point in using a bigger sensor, since the extra resolution would be wasted. So I guess I gotta ask the folks at Canon -- are we really gonna be able to see a clear difference with photos taken with these new ultra high-output sensors?

[calvin83]

With a 50MP Bayer filter, you will need to pay more for an high resolution lens. I would rather they made a 16MP full-color RGB sensor rather than a 50MP Bayer filter.

[MartinCrabtree]

I pine for Kodachrome 25.

[OPAL]

Most people viewing their images just on their PC monitors! Many people's Hobby is to watch the sharpness of their glasses, increasing the images up and over 100% on the monitor.

With this new 50MP Canon DSRL, these people will be crying thick tears, if they're notice and seeing the optical qualities of their lenses!

[cooltouch]

I dropped in at sonyalpharumors.com and was browsing through their listings to see if I could find out anything more about the new mirrorless FF models that Sony is supposed to be releasing very soon, and I came across an article quoting Tamron as stating that they had no plans to make lenses for the A7 cameras. Here's part of the quote:

"The Spanish site DSLRmagazine asked Tamron if they plan to make lenses for the Sony A7 system. But the official answer was: 'Currently we have no intention to cover this line of business, although we have a very good relationship with that particular brand.'

He also confirmed the new 15-30mm Sp lens can resolve the new 50 megapixel FF sensors. This fits with a rumor I got back many months ago. That source gave me the lens specs as first and also told me Sony would make a close to 50MP FF camera too. Question now is only WHEN that camera will be announced to fight back the Canon 5ds."

I would sure like to find out more about Tamron's definition of "resolving" a 50mp sensor. The math is straightforward. A full frame sensor will be 36mm x 24mm, if it follows the same size conventions of 35mm film cameras. If we use the Canon value of 50.6 megapixels, that's 50.6M = 24mm * x pixels/mm * 36mm * x pixels/mm. Simplifying things, we get 50,600,000 = 864 * x^2 or 50,600,000/864 = x^2. Take the square root of both sides and we get x = 242. This is the number of pixels per millimeter with a 50.6 mp FF sensor. But we want to put this in terms we're used to seeing, namely Modern Photography's test chart results, which are in line pairs per millimeter. So we just divide 242 by 2 to get pairs. Which gives is 121.

Now, I don't know about you, but I have never seen a lens test in which a lens had resolution numbers that were even close to 121 lppmm. If Tamron is now producing lenses that hav 121 lppmm resolution, that all by itself is sensational, headline material, IMHO,

Doing a bit more googling, I ran across an article at photographylife.com that has this paragraph:

Lenses could be a source of problems too. If you want edge to edge sharpness, forget about using lenses older than 5 years. Lenses designed for film cameras and early digital cameras might do well in the center of the frame, but will surely suffer everywhere else. Thats because many lenses are not optimized for high resolution digital camera sensors and older film lenses are not designed to perform on a flat sensor. This all means that to take full advantage of the 50+ MP sensor, one has to use the latest generation lenses that are specifically designed to yield maximum sharpness and contrast on imaging sensors. If your lenses are unable to resolve 50 MP, there will be little advantage to owning such a high resolution camera. Images will also suffer from poor technique and handling of such equipment in the field. Hence, while there are clear benefits to high resolution sensors, there are a few variables to keep in mind and it all has to come together to truly be advantageous.

I have some problems with that paragraph, especially where he writes "older film lenses are not designed to perform on a flat sensor." That right there tells me he doesn't know what he's talking about. Yes, I'm aware of the concept of plane curvature, but that is due do the spherical nature of optics and is concerned more with the curvature of the "plane" of the subject rather than the film plane. A film =plane= is just as flat as any sensor is. But still, it is undisputable that lens designers are having to resort to a different bag of tricks nowadays with digital cameras than they used to have to deal with with film. But it seems to me that resolution isn't just some sort of magical value that can be greatly increased just because they wish it to be so, which is the impression the author is giving me.

[sichko]

I'm not sure that 50 MP is enough. A Nikon 24 MP APS-C D7100 has a pixel density equivalent to a ~ 55 MP on FF. And it shows moire with a kit lens!

[calvin83]

Some superior lenses reach more than 120lp/mm at 10% contrast in the centre part if they stopped down to F4.5 to F5.6(Table 2 from http://vitaleartconservation.com/PDF/affects_of_lens_quality_on_image_resolution_v3a_draft.pdf).

and from Zeiss http://www.zeiss.com/content/dam/Photography/new/pdf/en/cln_archiv/cln13_en_web.pdf

[cooltouch]

Thanks for the links, Calvin. That first one is especially informative, with tons of data.

I found it interesting to note that, despite its stating that very high resolution lenses exist, in table 2 on page 8, it in fact mentions only one: the Schneider Symar 150 at 80 lppmm. It alludes to a few others, but their data is buried in tables in which it isn't as easy to extract and understand.

I have one roll of Kodak Panatomic X left (170 lppmm). I think I'll save it for a special occasion.

[tromboads]

Well that's it, this 50Mp FF sensor shares a similar pixel density to current APS bodies. And people arn't running out in droves buying $$$$ glass because their images look sh1t.

This size will be fine. It's a current model APS with a wider FOV

Now are current APS sensors far too dense? Probably. And to answer the questions about old glass, does old glass lens look crap in the center on current APS bodies?

[Snodge]

I must admit that as a Canon 5D3 owner/user, I'm a little perplexed at these 2 cameras they've released. I'm sure that there are some photographers that can make use of the extra size of the images, however it seems to me that the low pass filter bypass is just an easy way (or easier way than removing it completely) to respond to the Nikon and Sony equivalent cameras, and that the built in intervalometer - again a nice to have - is to address one of the main reasons people are using Magic Lantern on their cameras.

I suppose they have to do something to keep up with other manufacturers, but I would have expected that there would be a much better frames per second, and improved ISO performance over previous models.

While I am quite content with my 5D3 at the moment, if Canon continue as they are at the moment, by the time I need a new camera body I would almost certainly be switching to A.N. Other manufacturer.

[lm4187]

Tromboads and sichko are correct. The resolving power of the 50MP FF sensors are the same as the current 24MP APS-C sensors (within 5% or so). The pixel pitch of the 50MP is about 4.1 microns, of the 24MP about 3.9 microns. In fact, the FF sensor has slightly less resolution.

The Bayer filter lessens the resolution because of the interpolation required to fill in the 'blank' pixels in each color plane. And that is for the X-Y distances. Look at the diagonal interpolations and the resolution goes down even further.


Cheers,
Lou

[cooltouch]

[sichko]

[calvin83]

I would suggest reading the following article before continuing the discussion.
http://luminous-landscape.com/do-sensors-aeoeoutresolveae%C2%9D-lenses/

[sichko]

[cooltouch]

That's a very informative article. It combines much of what camera optics experts have written with much of what telescope optics experts have written. I just skimmed through it. It will take a good deal of study before I can assimilate it all. It does remind me of when I used to be a much more active astronomer than I am now, and when I actually conducted some resolution tests that are discussed in the article. When I began this thread, though, I specifically decided not to include things like airy disks and diffraction limits in this discussion.

Sichko asks "Can you really resolve one line pair in the image using only one line pair on the sensor ? For reasons suggested by Lou (lm4187) and for other reasons, some suggest that rather more are needed - as many as six sensor lines per image line."

Look, I was just trying to keep things simple without getting complicated or picky. When I'm talking about line pairs, I mean that by definition, one line pair is equivalent to two pixels. When referring to the AF Test Charts that are still one of the most widely used methods for determining resolution, when counting the smallest distinguishable pair of lines, this must be the equivalent to a pair of pixels. And that is the Rosetta Stone that I've used to translate between more traditional measurements of resolution and pixel-based resolution.

[visualopsins]

Using these cameras will we be discussing (and choosing) mf lenses by how much zoom from 100% pixels to make a sharp image? Like, "with this lens we need to zoom 3X from 100% pixels to see any sharp detail" and show the 3X100% pixels sample crop. Great cameras for lens testing, yes?

As if it matters, the digital audio guys are wanting 1MHz sample rate -- having sensors "out resolve" lenses gives some wiggle room, yes?

[cooltouch]

[visualopsins]

[sichko]

[cooltouch]

Aye, and methinks your explanation has the ring of truth to it, me lad. I can relate what you've written to the concept of "significant figures" I was first taught in my Chem 1a class many years ago. If you reach the point where the significant figure is +/- a certain value and you've reached that certain value, attempting to achieve a more precise value through measurement results in basically a random result. I can equate that to "blur". So with a sensor that has a higher mp count, the critical +/- sigfig willl be proportionally less than that of a sensor with a lower mp count. Is this a valid comparison, do you think?

[visualopsins]

How to calculate sensor spatial resolution?

[sichko]

[calvin83]

[cooltouch]