How Much Better?
July 18 (commentary)--"Dear Thom: if the upcoming D400 really has twice the pixels of my D90, how much better would my pictures be if I upgrade? Signed PixelChaser."
Dear PixelChaser: if you have to ask that question the answer is that your photographs will either be about the same or possibly even degrade in quality. Yours, Thom.
It seems like forever that I've been answering this question. I, too, at one time thought more pixels always meant more better. Of course, my first digital camera was .3mp (that's point 3, not 3). I don't mean to be mean in my answer, but in practice, it's proven to be mostly accurate. If you don't know what it is you're chasing and why, you won't achieve it.
Serious shooters basically fall into one or both of two categories: (a) They are pushing every last pixel they've got into very large print work; or (b) they are looking for low-level pixel integrity because they realize that gives them more and better choices in post processing. If you're in one or both of those categories you know exactly what more pixels should provide you.
Category (a) wants pixels of at least similar integrity to what they've got, just more of them so they can print bigger. If they think their current camera is maxed out at prints of 24", then they want the same abilities, but with enough extra pixels to print 36". In the end, many of the people in this category end up doing the same thing they did with film: they go up size. If they were shooting DX at 12mp, they go to FX at 24mp. If they were shooting FX at 24mp, the move to MF at 40+mp. The reasons to do this basically are the same as they were with film: the bigger capture area comes with a built-in advantage when you're basically chasing maximum print size: it decreases the magnification (all else equal). We've had a good run with digital sensors, where today's 16mp DX is arguably giving us similar or better pixel integrity to the old 6mp DX sensors, but it's often quicker, easier, and more productive to just buy a bigger camera.
Category (b) wants their images, usually of some (relatively) fixed size--perhaps they shoot for a magazine--to look better. This is a trickier situation than (a) because "better" can come in a lot of different forms. For instance, if you were shooting inside NBA arenas, low light performance improvements were more interesting than more pixel improvements, which is why the D3 and later D3s won so many converts. If you're shooting at a fixed size (13x19" from your desktop printer, magazine page for your photo editor, 8x10" print from a lab, etc.), more pixels on their own doesn't always give you more better.
To understand that last remark, you need to understand all the meanings we usually cram into the word "resolution." True resolution is measured in line pairs per millimeter, which is a linear measurement. (True resolution is also a chain of resolutions--sensor, lens, etc.--but that's out of the scope of this article.) Thus, the linear change in pixel count is important:
3mp: 2000 pixels on long axis
I've been a little loose with those numbers because it makes it easier to see the thing I want to relate. When we went from our 3mp D1 with 2000 pixels on the horizontal axis to our 12mp D3 with 4000 pixels, we got a probable doubling of resolution (all else equal). We don't get that when we go from our 12mp D3 with 4000 pixels on the horizontal axis to a D3x with 6000 pixels. Instead, we get a 50% increase. Now consider the 16mp D7000 to 24mp D400 leap: 20% increase. Different sources come up with different figures, but I generally use 15% as being the minimum necessary for most people to see any difference in resolution, so the D7000 to D400 is just barely above that bar.
Meanwhile, we've got other factors potentially fighting us. One of the reasons why some of those old 3mp and 4mp images look pretty good these days is that we weren't really recording diffraction impacts. Diffraction wasn't hitting far enough away from an individual photosite to get well recorded into adjacent resulting pixels. It's another area of debate (there are many in this discussion), but some of us use 2x the diagonal of the photosite for the "diffraction impact" mark. Below that, diffraction doesn't significantly lower resolution test numbers. Above that, it does. So as photosites have gotten smaller, diffraction impacts have gotten more visible.
I used the term "pixel integrity" earlier. What are the components of that? A pretty long list, actually, of which here are just some:
- diffraction impacts
- microlens spillover (another light adjacency issue)
- Bayer consistency
- AA filtration level
- consistency/linearity of ADC
- electron migration
- on-board noise leveling (neighbor pair assessments, etc.)
The list goes on and on.
I have little doubt that the camera makers will continue to push forward for both the (a) and (b) shooter. We'll get higher pixel counts, and we'll get a continued devotion to better pixel integrity. The reason is simple: without those things, it gets pretty tough to sell a new DSLR at all. Let's face it, from the D5100 on up Nikon's DSLRs have a pretty long list of features that'll let you do most anything you need to, plus these cameras run from pretty good performance at all things they do to excellent. So without sensor improvement, it would take a big change (did I hear someone say CPM?* ;~) to sell any of us a new camera.
Nevertheless, even though we're getting strong gains in (a) and (b) in sensors, the return is getting lower with each generation. We're very near the point where it should become obvious to everyone that the real choice for tangible gain is to go up size.
*For those with short memories, CPM is my short-hand for communicating, programmable, modular, which is what I've been suggesting for several years now that our cameras really need to be.
Очередное мини-эссе от Хогана в отношении того, что реально даёт увеличение мегапиксельности камеры, и о каких факторах стоит думать, пытаясь ответить на вопрос: "Насколько лучше будут мои фотографии, если я приобрету камеру с новым сенсором" . Много английских букв, посему упрятал под спойлер для интересующихся.
P.S. Товарищ Том закончил писать ещё 3 раздела для этой серии и положил их все тут:
Сколько часов мне нужно посвятить фотографии, чтобы начать снимать как хороший профессионал?