Megapixels and Print Size
This subject is the perpetual merry-go-round, because there just isn’t a definitive answer. There is just so many factors that contribute to ‘acceptable quality’ that it’s difficult to list them all, but I’m going to try! I’d then like to hear about how you approach this topic yourselves! Please also vote in the poll to your right so I can get an idea of how big you are all printing!
So, why do people give such wildly different opinions when it comes to how many megapixels you need to make a quality print of a given size (and we know this topic nearly often generates a virtual punch-up on photography forums)? Well, there are quite a few reasons…
What is ‘acceptable quality’? We all differ I our perceptions and our eyesight is not all the same either. Some people can genuinely resolve more with their Mk I Eyeball than others. Those who come from a large format film background are likely to have a different perception of quality to someone who spent four decades shooting 35mm TriX developed in Agfa Rodinal!
Normal Viewing Distance: Another layer on top of the above is the concept of Normal Viewing Distance, so not only do we differ in what ‘acceptable quality’ is but we may judge it from different distances! Personally, I think about whether the print I am showing is the kind likely to be ‘sniffed’(!) Print sniffing is not unusual if someone is presenting a two metre wide aerial shot of a city, but it definitely looks a bit strange if it’s a 30cm slow shutter shot of two dancers twirling. In the former case, there may be an expectation (or hope for more detail), but in the latter case as soon as the image is mentally processed its obvious to most people that there is no benefit (nor need) to looking closer: detail is irrelevant to making the shot work. A two metre print of a city skyline, however, has a completely different effect if it is extremely high in resolution across its length compared to one shot on a 6mp camera that only looks ‘sharp’ from the far side of the room. So, our style and presentation affects people’s expectations and, if Mr and Mrs Public go sniffing your grainy Moriyama like image, you can be fairly sure that any adverse reaction to the lack of close-up detail is unimportant: they would not be buying. People who buy prints tend to have a better understanding of these things.
Application/Purpose: Related to the above, there are lots of applications where you want to shed resolution for the final print, not gain it. If you only want a low-resolution crude ‘salt’n pepper’ effect at the end, with huge Moriyama silk screen print type grain, then its probably not advisable to start with a 50MP file! You can, but it will be more work. Lower resolution and ‘bad’ lenses can be very helpful to us.
Uprezzing is another subject and so here we’re only going to look at native files.
Lens Quality & Resolution: This. Makes. A. Big. Difference. Camera manufacturers bundling cheap zooms with $1500 bodies might not want you to realise this (so you end up buying a bunch of pro lenses later, when you realize your new purchase is outperformed by a Sony RX100 compact), but it’s a fact. Great lenses make pixels sing. Pixels that don’t sing are about as much use as choir members who don’t sing. OK, that is a stupid analogy, but sometimes things just come out when you are sitting in Kabul and have limited human contact and gin. While it is true that two lenses of the same quality (even if quite poor) will result in more detail when used on a higher resolution sensor, if you were to put an amazing lens on the lower resolution sensor it may well vastly outperform the higher-res one. A good example is the stunning lens and sensor match found in the Ricoh GR, which I reviewed here. I have also included a snap from Bamyan, Afghanistan at full resolution (jpeg) so you can click on this link and see for yourself (I gave the file a mere 45 points of sharpening with 0.8 radius, but it is otherwise untouched): Ricoh GR: 16 Almost Perfect Megapixels
Aperture: Some people shoot a lot at the optimum apertures for their lens and camera combos. Others tend to stop down a lot for depth of field and suffer the effects of diffraction. If you pack more pixels into a sensor, the ‘pixel pitch’ is higher. To really get a different recording on each of these tiny closely packed pixels, the lens is going to have to be delivering a very detailed image. What this means is that as pixel counts go up, so do the demands on the lenses (they need to be better to make the extra pixels count). Small losses in resolution due to diffraction (as you stop down) can therefore make it harder to lay down the detail necessary for the extra pixels to record more detail. This is why numerous reviews show that the Nikon D800E only holds a resolution advantage around f4-f5.6. At f8 and f11, diffraction the robs the Anti-Alias filter free D800E of any advantage.
Post processing: Someone who shoots straight images, with minimal processing, is able to get more from their pixel count than someone who processes very heavily. A 24MP file printed straight, with a bit of careful sharpening, may look great at a big size. Subject the same 24MP file to a post-processing workout a la Rocky IV and it might not look nearly as good at a given size. Files can only take so much bending before noise and detail start to make an appearance.
Format: Just like with film, a bigger format (sensor) means that for a given pixel count, the pixels are bigger and not as tightly packed than in a smaller sensor. This means lenses don’t need to lay down such high frequency detail, because the same number of pixels are bigger and more spread out. This means that a 30MP medium format back is likely to produce more detailed prints than a 30MP APS-C sensor might, much of the time. I say much of the time, because there are lenses that have astounding resolution and can lay down the very high frequency detail needed by a small sensor with tiny pixels. This is why some people get very excited by the Sony A7S’s measly 12MP: because those pixels punch above their number because they demand less of lenses.
So at the end of all of this, what do I think? Of course it varies, but in very broad general application terms, here goes:
12MP – A3/16”x12”:
20MP – A2/24”x16”
36MP – A1/33”x23″
But none of the above is cast in stone at all! At best, it is a rough guide for those seeking to produce smooth, detailed prints that look like traditional photographs. I have already mentioned how the little Ricoh GR performs more like a 24MP camera due to the perfect match of lens and sensor (and will show the door to a Sony A7R matched with a 28-70 Kit lens in overall terms), but you also have exceptions like the Sigma cameras sporting Foveon sensors that lack a Bayer filter. There is also the Leica Monochrom, which can compete very well with 36MP camera, not just because of it special sensor, but because the lenses are so darned good. While it cannot match the resolution of the Sony A7R on centre, it outperforms it comfortably at the edges and corners with wide angle lenses. The end result is that the overall sense of quality can exceed that of a camera with double the megapixels.
What about devotees of 35mm B&W film who are now working digitally? Well, I am absolutely certain that 6-8MP gives you far more real world detail than 35mm TriX ever did and 16MP is easily competing with 6×7 TriX in many respects, with 36MP being absolutely miles ahead. I’d go so far as to say that, with good lenses, 36MP is right up there with 5×4 Ilford FP4+, having shot that combination for years!
There are heated debates and flame wars on forums because people confuse their own needs with those of others and often very little of the above is considered when throwing absolutes around online. Now that technology is so good, it might be tim to recognise what you actually need, or may in fact be missing when chasing megapixels. Sure, they have their uses, but I remember very well a person who posted an image on line from a 36MP camera, having converted it to B&W, asking how to better achieve a 35mm TriX look. The first step would be shedding 80% of the resolution (and DxO Filmpack’s TriX setting would achieve that very quickly using grain).
***Please share this post so we can collect as many print size votes as possible***