End of the megapixel race? Two interesting articles by Ctein

[ErikKaffehr]

Hi,

Many feel that there is an insane megapixel race, and that we have more pixels than ever needed. These two articles by Ctein present a different view:

http://theonlinephotographer.typepad.com/the_online_photographer/2009/02/why-80-megapixels-just-wont-be-enough.html

http://theonlinephotographer.typepad.com/the_online_photographer/2012/02/d800-megapixels.html

The other side of the coin is that Ctein now days shoots an 4/3 camera with 16MP and says it is good enough for A2-size prints.

Ctein is foremost known as a great printer, mostly die transfer, but he is also a photographic artist.

Best regards
Erik

[shadowblade]

The megapixel race is certainly not insane - certainly not if you shoot detailed landscapes or print large.

It may be insane for sports, action and wedding photographers, who don't need great detail or large print and whose voices tend to dominate forums, but the huge success of the D800 and A7r and the continued interest in medium format show that there are plenty of photographers out there who need and want resolution.

Canon appears to have dropped out of the race (possibly due to technical deficiencies) but it seems that Sony/Nikon is pushing ahead regardless. Which I'm very glad of.

[Wayne Fox]

Really old articles.  (I was surprised the one was from 2009, I remember reading it, hard to believe that was 5 years ago).

But it does seem the megapixel “race” is over, as we’ve seen very little movement for several years now.  That doesn’t mean we have enough, just means that most seem to be happy with what they have so there’s not much incentive for makers to invest.  Canon still hasn’t responded with an answer to Nikon’s 36mp, and it hasn’t had much impact on them. 

So the race may be over, but some of us would like to see some movement.  I print larger than A2 for test prints.

[Bart_van_der_Wolf]

Many feel that there is an insane megapixel race, and that we have more pixels than ever needed.

Hi Erik,

It's horses for courses, as usual.

For most uses (probably internet publishing) we have plenty of megapixels. But for some uses, e.g. product/architecture/nature photography where realism (surface texture, tactile quality) is needed (which is not always the case), and especially when larger output is needed, the more the better. Output size and technology is the driving force that determines the requirements, and let's not forget the intended viewing distance.

There is also a development towards smaller photosensitive elements which get combined into output pixels. That allows to extract higher quality from given lenses, and dynamic range expansion of the tiny elements may also be achieved if the technology is applied well. But that's not necessarily the same as a megapixel race, it's more a capture technology.

Cheers,
Bart

[Abe R. Ration]

Hi,

Many feel that there is an insane megapixel race, and that we have more pixels than ever needed.

More pixels are needed for proper sampling. At the moment we either have to accept aliasing even in it's most hideous forms, or a needlessly thick optical stack on top of the image sensor because of the low pass filter. As the world is slowly moving towards mirrorless cameras where wide angle and normal lenses may have exit pupil close to the image sensor, it would be advantageous to have as little material on top of the sensor as possible, at least if one is interested in using lenses not specifically designed for anti alias filter of certain thickness (e.g.Leica M lenses).

[PhotoEcosse]

Horses for courses. (How often have I said that on this chatroom?).

I would love to have 48, 60 or 120 megapixels available.

My reason? Certainly not because I print large. I rarely print larger than A3+ (although sometimes from a small portion of the file).

No, it is because my view of creative photography is that digital photo-processing is essentially data-processing pure and simple - and the more data I have available, the more creative I can be with it.

I certainly find that 36Mp gives me hugely more creative scope than 12 ever did. (Although I happily concede that there was little discernible difference in print quality in an A3+ print from the full image of a 12Mp file and that from the full image of a 36Mp file, all other things being equal - which, of course, they never are).

[mjrichardson]

Morning!

I think it's the same as all races, you either enter or you sit back with a beer watching people getting sweaty, suffering from joggers nipple and getting out of breath!

For me, I don't think about more megapixels because I don't have them, I concentrate on doing what I can with what I have, bigger and "better" only becomes an issue when there is technology available to complement the work I do that has more of whatever it is that is the latest thing, megapixels, dr, ISO, whatever it is. At the moment there isn't anything that does more than I have for what I shoot so I concentrate on the shot.

We all want different things and manufacturers want to make money so there will always be something new around the corner, it's just the nature of things.

Mat

[Telecaster]

Just personally: I no longer care much about printing. I still do it occasionally but it's not the focus of what I do. I am interested in making photos with 4k display in mind, though, and particularly in taking control of the R, G & B sub-pixels of each full-color display element. A nice thing about this interest is that it imposes some hard limits. A 4096x2160 image, if you provide non-interpolated RGB data for each display pixel, requires a Bayer sensor with 8192x4320 photosites (an R, G & B for each display pixel, with the two Bayer Gs averaged in some manner). That's 35.4mp. Other color-creating strategies and/or technologies, such as Foveon, may have different requirements. The apparently upcoming Sony 35mm-format sensor(s) will have the necessary photosite count. Then will come the need for appropriate processing software.

[Edit: fixed math flaw.]

-Dave-

[marcmccalmont]

I hope that someday all sensors are using 2 micron photosites so all lenses at all apertures are diffraction limited (sensors out resolve the best lenses)
no moire and you always get the best out of your lens
Marc

[Jim Kasson]

I hope that someday all sensors are using 2 micron photosites so all lenses at all apertures are diffraction limited (sensors out resolve the best lenses)
no moire and you always get the best out of your lens
Marc

A worthy objective, but 2 um pitch is too coarse to achieve it, thanks to the Bayer CFA.

http://blog.kasson.com/?p=5920

Jim

[dwswager]

Back about the time Nikon introduced the D1 I stated to a buddy that at 6MP digital would replace film.  Check.  Anyone remember Genuine Fractals?  My buddy laughed and said no way.  Reminded me of a co-worker that said his college professor made them use a slide rule because "calculators were just a fad!"

At 12 MP, we basically duplicated the resolution provided by 35mm film.

At 24MP cameras like the D7100, D610 and D750 we blew past anything we were getting before.

I now have a D810 and 36MP works for me.  I actually popped the money for this camera because it was the 1st DSLR I wouldn't mind getting stuck with long term.  In most ways the sensor is 'good enough' and while there might be upgrades, I suspect they will be minimal.  Where this camera will get upgraded in a significant way will be with better signal processing and data throughput opening up additional functionality.

We have to be careful when talking resolution about the camera systems we include.  35mm was smaller, lighter and cheaper than Medium and Large format.  Nikon, Canon, Sony, Samsung, Olympus, etc. still are looking at a mass market product and probably won't ever replace larger digital systems.  I don't know the technical limitation in resolution at 36x24mm sensor size, but I believe you will see roughly 24MP as the new standard and 36-48MP as the specialty cameras.  With that, you will capture about 90% of users.  Then there will be high dollar larger systems for 'extreme' users.  Anyone want to bet Nikon's next flagship camera (D5) will be higher than 24MP?  At best, there might be 3 versions.  The base camera, a high resolution (and slower speed) variant and a maybe a high speed action variant (maybe with less resolution).

I will add that we have hit diminishing returns and the incremental increases will come at ever increasing cost.  Like computers went through a 2 decades where each new model provided increased performance and now computers are commodities.  It's all about the funtionality, features and package.  I've been building my own computers for 20 years (My house has 2 home built desktops, a Home Theater PC and a 9TB Server, in addition to 4 laptops, an iPad, an Android Tablet, 4 Smartphones) and the next one I build will be just marginally better than the i7 I built a 1.5 years ago.

[ErikKaffehr]

Hi,

Nikon and Sony both have 24 MP APS-C, right now. That  would scale to 54 MP. I am a bit surprised that we don't see 54 MP on FF 135 right now.

The major advantage of reducing pixel size is better sampling, ideally the lens should not resolve significantly better than the sensor. But that also means that the resulting image will not be really sharp at actual pixels. Larger image size obviously needs more processing power.

What I would also say is that a 54 MP image properly sharpened and reduced to 24 MP using a decent downscaling algorithm will always be better than an image coming from the same platform using a 24 MP sensor.

Now, some will say that even smaller pixels are needed, and that may indeed be the case, especially with very good lenses using optimal apertures.

Best regards
Erik

Nikon, Canon, Sony, Samsung, Olympus, etc. still are looking at a mass market product and probably won't ever replace larger digital systems.  I don't know the technical limitation in resolution at 36x24mm sensor size, but I believe you will see roughly 24MP as the new standard and 36-48MP as the specialty cameras.  With that, you will capture about 90% of users.  Then there will be high dollar larger systems for 'extreme' users.  Anyone want to bet Nikon's next flagship camera (D5) will be higher than 24MP?  At best, there might be 3 versions.  The base camera, a high resolution (and slower speed) variant and a maybe a high speed action variant (maybe with less resolution).

[Manoli]

The major advantage of reducing pixel size is better sampling, ideally the lens should not resolve significantly better than the sensor. But that also means that the resulting image will not be really sharp at actual pixels.

But does that matter ? It may well be an advantage - you'll get a soft 100% and a file that resolves the optical system (almost) to the full, with no digital artefacts. With the improved sharpening and the new frequency-based clarity controls (Topaz etc) you're able to fine-tune and control apparent sharpness, depending on target output, in post.  If you're going to print at anything over 250 ppi then 'pixel level detail' becomes rather mute.

I can see why this may not appeal to landscapers, but for portrait-ists it seems preferable to the current trend of uber-clinical rendition of some modern asph-lenses.

This reminds me of the Leica v Zeiss debate from the analog days, where Leica were generally considered to have the higher resolving lenses whilst Zeiss were of higher accutance (contrast). B&W photographers generally preferred the softer Leica rendition because you couldn't increase resolution but you you could control and increase the contrast in the print stage.

[ErikKaffehr]

Hi,

Personally I find this kind of softness positive, as it allows for proper rendition. But the image will look bad on screen, at least until proper sharpening is applied and this may cause people rejecting the sensor.

The point Ctein makes it that it takes a lot of pixels for proper rendition of a small detail. Those pixels can be quite soft. Larger pixels will show higher microcontrast but less proper rendition.

My take is that optimal sampling combined with optimal sharpening, correctly downsampled,  will yield a better image than coarse sampling.

Best regards
Erik

But does that matter ? It may well be an advantage - you'll get a soft 100% and a file that resolves the optical system (almost) to the full, with no digital artefacts. With the improved sharpening and the new frequency-based clarity controls (Topaz etc) you're able to fine-tune and control apparent sharpness, depending on target output, in post.  If you're going to print at anything over 250 ppi then 'pixel level detail' becomes rather mute.

I can see why this may not appeal to landscapers, but for portrait-ists it seems preferable to the current trend of uber-clinical rendition of some modern asph-lenses.

This reminds me of the Leica v Zeiss debate from the analog days, where Leica were generally considered to have the higher resolving lenses whilst Zeiss were of higher accutance (contrast). B&W photographers generally preferred the softer Leica rendition because you couldn't increase resolution but you you could control and increase the contrast in the print stage.

[marcmccalmont]

A worthy objective, but 2 um pitch is too coarse to achieve it, thanks to the Bayer CFA.

http://blog.kasson.com/?p=5920

Jim

my thinking is that if the real resolution you want is 36 mpx with 4.8 micron photosites then 130 mpx 2.4 micron photosites would guaranty you 36 mpx resolution w/o artifacts and maximum use of your lens. Since point and shoots are at this pixel density why not?

[ErikKaffehr]

Hi,

My take on the issue, too.

What I see in my experiments is that it takes about f/16 to get rid of aliasing on 6.8 microns, using elderly lenses. Going to 2.4 microns would eliminate most of aliasing at f/5.6 and giving a healthy win in resolution.

It could be discussed if it is better to rely on small pixels or somewhat larger pixels in combination with a properly designed OLP filter. Going from say 4.8 micron pixels to 2.4 micron pixels will loose 1 EV in DR (at SNR = 1) but will not really affect "photographic DR" (at SNR = 10) that much.

Personally, I would say that 3.8 microns, that is what we have on 24 MP APS-C would be OK or at least preferable to 4.8 and 6 microns that we have on 36 and 24 MP full frame right now. That is I don't say 2.4 microns would not be good, just that I would see 3.8 microns as a good step forward.

The way I see it, we have some great lenses coming our way both from Sigma and Zeiss. Hopefully both firms also will sell smaller lenses with smaller maximum apertures. It is an interesting time to live in!

Best regards
Erik  

my thinking is that if the real resolution you want is 36 mpx with 4.8 micron photosites then 130 mpx 2.4 micron photosites would guaranty you 36 mpx resolution w/o artifacts and maximum use of your lens. Since point and shoots are at this pixel density why not?

[MarkL]

Oversampling is common with other areas using digital signal.
With 36MP I can still only make a 24.53 x 16.37" print at 300ppi without interpolating pixels
We have had 39MP digital backs since 2007, then everyone said "omg who really needs that much" and they have since marched on to 80MP

I do think we will see a further divergance of product lines with companies increasingly offering lower MP cameras and higher MP cameras concurrently.

[dwswager]

Oversampling is common with other areas using digital signal.
With 36MP I can still only make a 24.53 x 16.37" print at 300ppi without interpolating pixels
We have had 39MP digital backs since 2007, then everyone said "omg who really needs that much" and they have since marched on to 80MP

I do think we will see a further divergance of product lines with companies increasingly offering lower MP cameras and higher MP cameras concurrently.

People often mistakenly believe that 36MP is 3 times the resolution of a 12 MP camera, when in fact it is less than 2x.  Because of the 2 dimensions of an image, 48MP is double 12 MP.  I personally think that 48MP will be the top end of the 36mm x 24mm sensor MP count.  But, of course, the popular acceptance of the D8x0 cameras might change that.  I would love to see the sales projections versus actual sales for the D800/D800e.

[Geraldo Garcia]

People often mistakenly believe that 36MP is 3 times the resolution of a 12 MP camera, when in fact it is less than 2x.  Because of the 2 dimensions of an image, 48MP is double 12 MP.  I personally think that 48MP will be the top end of the 36mm x 24mm sensor MP count.  But, of course, the popular acceptance of the D8x0 cameras might change that.  I would love to see the sales projections versus actual sales for the D800/D800e.

Well... people who think 36MP is 3 times the resolution (3 times the amount of pixels to be exact) of 12MP are correct, no argument around it.

[Guillermo Luijk]

The megapixel race is certainly not insane - certainly not if you shoot detailed landscapes or print large.

I am not so sure about the not-insanity of the Mpx race when shooting landscapes. This is commonly a large DOF application, so unless you use some large DOF technique to avoid the need of using close apertures (focus stacking, stitching, lens tilting), I wonder how useful can be having tons of Mpx if f/16 on a FF sensor will ruin sharpness on any very high resolution sensor because of diffraction?.

With a highly dense sensor you are not capturing real detail in large DOF applications, just blurred redundant pixels.

This is how diffraction affects a Nikon D800 (36 Mpx) at f/16:


A FF sensor at f/16 becomes diffraction limited (i.e. it begins to loose sharpness because of diffraction) from 12 Mpx:



Diffraction calculator at Cambridge in Colour.

Regards