When the megapixel wars were in full swing, it wasn’t uncommon to find photographers poo-pooing the race to add more pixels. While there certainly is some truth to that point of view, it’s not entirely representative of the pixels situation.
The counter point to the fewer better pixels argument is that more pixels really do capture more detail. Moreover, even if at the pixel level they are noisier, in the resulting images the noise becomes finer grained and as a result less distracting.
The reality is, that so long as the increase in pixels isn’t causing a problem for other needs — for example, you aren’t limited to 3 FPS when you need 10 FPS to do you job — having more pixels makes better images overall.
However, there are a number of places where this does fall apart to some degree. And ultimately this is where the real balance has to be struck between more pixels, better pixels, and marketing numbers that sell the next generation of cameras.
To start with the most obvious place where more pixels is problematic is that they demand better quality lenses; at least if you want to maintain pixel level sharpness or quality.
Actually this is one of the places that I think is most problematic in terms of photography, and how we’re moving forward. The way I envision it, and the way I get people to try and picture it, is that there are two regimes.
One is where individual pixels matter. This is how we’ve done digital photography from the start. One side example of this approach is how you have to zoom in to 1:1 magnification in Lightroom to see the affects of noise reduction and sharpening. Even sharpening tools, like photoshop’s unsharp mask, have their parameters specified in pixels.
The other side of the coin is that pixel stop mattering. In effect they become just part of the continuous color and detail in the image. The ultimate extension of this is the premise of sub-diffraction sensors, where the pixels are so small that they behave more like the grains of silver in film than what we’re traditionally use to in digital imaging. You have lots of pixels, pixels far smaller than the details you’re capturing, but you don’t care about the pixels per say, and you don’t work with the pixels per say, you work with the details in units like microns or millimeters.
Between these two regimes, there is also a point where pixel counts are ridiculously high, but not high enough that they can be completely ignored. I don’t have a good name for this region, but it’s where I’m increasingly believing we are in the digital photography world. With 30, 36, 42, and 50 megapixel “full frame” sensors, the pixels are increasingly becoming unimportant individually, but there is still enough limited resolution that you can’t move to the next paradigm.
The combination of increasingly higher resolution sensors, and the processing approach that’s build around pixel level manipulations, creates interesting side effects that propagate out to things like lenses and rules of thumb.
For example, higher resolution sensors combined demand higher resolution lenses to drive pixel level sharpness. That drives up the cost and complexity of lenses, and the systems as a whole.
The other side effect is boils down to rules of thumb. Many of our rules of thumb were empirically derived from film. While some people may hanker for the good old days of film and how awesome it was perceived to be, the reality is that resolution wise it was quite limited. Fine grained film stocks on small format (36×24 mm) cameras, really only could hack 16–18 MP worth of digital information. With cameras like the 5D mark IV and the D810, we’re easily at 2x that resolving power, and the 5DS is resolving 3 times what 35 mm film could.
It shouldn’t be unsurprising then that if you’re holding to the pixel sharpness as your objective, you’re going to have problems with rules of thumb that were designed to be sharp on cameras that only effectively resolved half of what a modern DSLR does.
In fact, as an aside, one very interesting take away from Mr. Clark’s calculations is that modern DSLRs at reasonable film speeds of even ISO 100 easily and handily beat many, if not most, medium format cameras shooting film in resolving power. And at higher ISOs digital completely decimates film’s ability to resolve detail.
One example of the problem with film era rules of thumb can be seen in the 500 rule for astrophotography. For those that aren’t familiar with the 500 rule, if you divide 500 by the focal length of your lens, the resulting numbers is approximately the longest exposure you can shoot in seconds before starts start rendering as short trails and not point sources.
This may have worked for 16ish MP equivalent film, but on a 30 MP digital sensor, using the 500 rule defined exposure time does result in multi-pixel long trails not point sources. Admittedly as long as you’re not making huge prints, the trails aren’t a serious problem. However, if you are, or you’re trying to keep stars from trailing at 100% magnification, you can’t roll with the old rule of thumb.
There’s a further consequence to needing faster shutter speeds for trail free stars. You also need a faster lens.
My reasoning for this is pretty simple. When I was shooting in dark sky conditions with 20 second exposures on an f/2.8 lens, I needed to shoot at ISO 12,800 to get good detail in the stars, and especially the Milky Way. If I have to cut the exposure time in half to keep the starts from trailing, then I have to push the ISO up another stop to keep the exposure levels the same.
Or you need a faster lens, something like a 24mm f/1.4. With two stops more aperture, I could either shoot at f/2 and keep the ISO the same while cutting down the trails. Or shoot at f/1.4 and cut the ISO by a stop to better control noise.
Now I will certainly admit that this is a bit of a niche point. After all not everyone is shooting stary skies in dark places without wanting star trails. However, it is indicative of some of the problems with simply translating a rule of thumb derived for film to our current high resolution digital bodies.
Truthfully I can say the same thing about the hand holding rule of thumb. Since moving to my 5D mark IV at 30 MP, many images that I’ve shot handheld using the 1/focal length for a shutter speed, have had camera motion issues. Practically speaking, this was not something I saw consistently with my 22 MP 5D mark III.
My point here isn’t that we should eschew increases in resolution. However, while we continue to move towards higher and higher resolution cameras, it’s important for us as photographers to be aware of how that change is impacting many of rules and guides that we become accustom to.
- http://www.clarkvision.com/articles/film.vs.digital.1/ ↩