(commentary)
Last week I posted an article on sansmirror.com about why 16mp might be enough resolution for most people. A corollary position to that exists, and I’ll discuss it here today on dslrbodies, because it’s a little easier to explain with the past and current Nikon DSLRs.
The question I get asked a lot is this: "if I print at 11x14” (or pick any other smallish size such as 8x10”), would I see a difference between an image from a D700 (12mp) and a D8xx (36mp)?" My answer is yes. But it’s my answer for my trained eyes.
Two problems come up with the “more megapixels are good” corollary argument. The first is what I just alluded to: most people aren’t trained to see differences. This has been proven time and time again in surveys and tests. It’s a form of Least Common Denominator, the same thing that defines sRGB Color Space, actually: what’s the least technology we need so that virtually all people can’t see a difference from a higher level of technology across a wide range of products?
The same thing happened in music. CDs are digitized. We lost something when we moved from the analog world. But we lost even more when we took the digital tracks and created the MP3 world, by applying compression to the music. Most people can’t hear the difference between an MP3 and a CD track, but it’s most certainly present. Amongst other things, MP3 compression works on the premise that when something is loud and complex, you can’t make out the subtleties. Using classical music, for example, the bombastic conclusion of Beethoven’s Ninth Symphony is so loud and complex most people can’t make out the individual instruments/singers. On the other hand, in the quieter, less complex sections of the work you can make out everything. So, to over generalize what’s being done, you heavily compress the loud/complex and don’t or lightly compress the quiet/simple passages.
Eyes work in some ways similar to ears, in some ways different. But as JPEG images have proven, you can clearly reduce large amounts of data into much smaller sets of data, and most people can’t tell the difference.
Some of this has to do with the smallest arc that the eye can see. Stand at a “normal” viewing distance from something and you don’t necessarily see the finest detail. Try it with paintings at a museum, for example. Stand at the distance everyone is viewing something from, then move very close. Now you see the brush structures you didn’t see when standing back. Trained eyes can see them without having to move as close as most people would have to in order to see them, because they know what to look for.
In fact, when I was deep into filmmaking, I was a mean machine when it came to “seeing” things that others didn’t. Hot frames, splices, microphones dipping into the frame briefly, distant items that weren’t consistent with the period, slight changes in positions between cuts, you name it, I saw it, even it was just in one frame out of 24 a second. My students would often contest this with me until I’d take the film and show it one frame at a time to reveal what I saw. But it takes knowing what to look for and how to see film this way.
That said, while it certainly isn’t the majority, some viewers of photography have better training and eyes than others, and can certainly see things other don’t. When I look at an 8x10” print from a 12mp and 36mp camera, the one telltale of the 36mp camera tends to be increased edge acuity. A small difference in definition, to be sure, but sometimes an important one. It’s why MF works so well for landscape photography, for example, while point and shoot cameras get the broad strokes mostly right these days, but somehow fall a bit short. It’s the same with 12mp and 36mp on the same size print, only much closer to one another in impact. I can usually spot this, but 90% of viewers probably can’t.
The Web with its JPEG compression and variable browsers (and this site’s automatic image resizing) isn’t the best place to try to demonstrate this, but I’ll give it a go. The following are two shots taken moments apart. One is a crop simulating a smaller pixel count (the lens was zoomed out), the other uses all the pixels (lens zoomed in), then downsized. It’s not a perfect comparison, as the leopard changed angle on me between the shots and thus slightly different parts of his face are in the focus plane, but still, see if you can figure out which one has twice as many megapixels behind it:
Yes, it’s subtle. And the slight change in focus plane and the extra JPEG compression won’t help you figure it out, but on my big calibrated display there’s a difference between the two images that’s pretty clear. Really smart folk will figure which is which based upon something else, though: noise.
So what’s the other problem with the corollary position of More Pixels are Better? That would be the high ISO thing I also wrote about last week. If we’re mostly shooting at ISO 6400, more pixels probably isn’t as good a choice as better pixels. The above shots were at ISO 3200 on a D800E, by the way. I’ve done a little bit of noise reduction, but not much.
Nikon seems to realize that we need different types of pixels for different purposes, though we’re now got dissimilar DSLRs instead of similar ones: the D4s and D810 are the pro cameras for “better pixels” (low light D4s) and “more pixels” (good light D810). The D4s can’t match the D810’s dynamic range at below ISO 800, while the D810 can’t match the D4s’s dynamic range above ISO 800. Not that either are slouches, only that they are each optimized for different duties.
Where do I stand on more pixels? Depends upon what I’m doing, which is exactly the Nikon philosophy, basically: pick up the right camera for the right job. Most of the time that’s going to be the “more pixels” camera for me. That’s because I rarely work at truly high ISO values and thus value the increase in dynamic range in the low register plus the increased acuity in a same-sized print. Sometimes—for me that tends to happen with indoor or night sports—I go the opposite direction and pick the “better pixels” choice.
It’s at this point I should mention Sony. While I’m sure it wasn’t a slam dunk decision internally on their part to do so, I believe what they did with the A7 series is dead on right. A7s = better pixels, A7r = more pixels, A7 = Goldilocks. (There are other subtle differences, like video and focus capabilities that need more fleshing out, but I’m talking mostly about sensor choice here.) I don’t change batteries, lenses, accessories, or even camera controls when I pick one Sony camera over the other for the job I’m doing today.
That’s exactly what I’ve consistently asked Nikon for: a beefy body with vertical grip that has sensor choice (e.g. D4s and D4x, maybe even a DX version), plus a smaller and lighter body that has sensor choice (e.g. D800h and D800x, and definitely a DX version). Instead, we’ve gotten two different cameras (D4s and D810 currently) with different batteries, different cards, different chargers, and different camera controls, and no DX followup to the D300.
Pity, because that makes me less likely to keep my D4s. I just won’t use it enough compared to the D810. Indeed, I’ll tend to use the D810 at higher ISO values than I normally would, because it’s “good enough.”
Which brings me to my final thought. “Good enough” is mostly a consumer thing. If Nikon wants to create "good enough" cameras with Coolpix, or CX, or DX, great. But professional products should be better than “good enough.” They should be “greatest possible.” Indeed, I’ll be bold here and make this assertion: if the camera body sells for less than US$1000, then it’s okay to be “good enough.” If it sells for more than US$1000, it should be designed to be “greatest possible” given its inherent choices (DX or FX, small body or big body, DSLR or mirrorless, etc.).
As I noted, I like what Sony did with the A7 series. I hope Nikon matches it with the high DX and the FX lines. There’s a large group of us that don’t want to compromise: we want a Stradivarius violin, a Bach trumpet, a Loree oboe, and so on. Products that with practice and care, we can maximize what we get out of them.
Someone recently showed me a recent photo about to be used in an ad campaign for a major company. In a posed, static shot, the pro photographer missed focus by a inch or two while using a D4. How much do you want to bet that if I took the same shot with my D810, nailed the focus, that the acuity in it would just pop in ways that makes the other guy’s photo look weak?
That’s where the corollary argument for more pixels really comes into play: shot discipline and clients that want the best. Getting it done right with the right product is far better than getting it done wrong with the wrong product. It’s how your work starts to shine in ways that others' don’t.
As you can tell, I have sort of a dual position about pixel counts. I want more. More sampling is always better, all else equal. But the general market doesn’t necessarily need more pixels. Far too many other aspects of the camera are now starting to take more importance than pixels for the consumer market: size, weight, cost, low light use, focus performance, controls, ease of learning/use, and so on.
But bring on the 54mp FX cameras, I say. And make the 16-24mp cameras better for the majority of the users, too.
When I write articles like this and the previous one on sansmirror.com, sometimes folk in various Internet fora seem to think I’m contradicting myself. Please read what I wrote carefully. I believe I have held a consistent position for quite some time on pixel counts: I’ll always take more because more sampling is always better (and produces things like the extra acuity and burying the noise as in the example shown). But in trying to serve customers of this Web site, I also have to put that in perspective. Pretty much once we hit 24mp the argument for more pixels for most customers has started to disappear. They don’t print big, they don’t crop that much, and they’re often better served by something else being pushed forward instead of pixel count. Thus, I’d argue that, as an industry, increasing pixel counts don’t serve the majority of the customers any more.
Things were different when we were using 2.5mp D1’s: 6mp did make a difference in what we could do with our photos. Suddenly, two-page magazine spreads started to approach film when done with care. We got another healthy boost that could easily be seen by most people when we went from 6mp up to 12mp. But 12mp, as I noted in the other article, almost gets you to 11x14” prints at native print resolution. At each print size increment (4x6, 8x10, 11x14, 13x19”, etc.), we lose more and more folk with more pixels: they just don’t output that big.
So my position is this: I still do some big work, and I value extra acuity. I’ll take 54mp FX. I might even take 108mp FX, though I’d be deep in the tail of perceptual improvements: very little gain. Most people should be fine with 16-24mp, though (CX, m4/3, DX, or FX). I believe that’s been my position for over five years, and it’s not changing with anything I see on the horizon.
Wait, what about Lytro, and their “megarays” or Sigma and their three-layer approach? Those are different stories for another day. However, in passing, I’ll point out that the Sigma Foveon sensor does have one of the benefits of more megapixels: higher acuity.
