Support
You can support this site without any cost or disadvantage at all by clicking this link to Amazon or the one on the left before buying anything – be it underpants, a cupboard, a TV, a pen, a lens or a camera. Amazon is the only shop worldwide, I’m really satisfied with to a 100%, so I have no caveats advertising them. Of course, you can also directly donate a small amount of money, e.g. the amount you would have spent for a magazine, with the button on the left.
UPDATE: Also see "Megapixels - Update: 6 vs. 10 vs. 15"
I added an old EOS 300D (the original Digital Rebel) to the game.
Introduction
In the first days of digital photography, more resolution meant higher quality. Every step in resolution was a big one in image-quality. Of course it was: A resolution of 1024x768 pixels is not enough for bigger than postcard-prints. But when resolution reached the 6MP-area, the differences became less important. And nowadays kind of a complete reversion seems to have taken place in peoples minds: Suddenly, more megapixels are "the devil", mean more noise and worse images. But that's only true for small-sensor point-and-shoots, is it? Or is it not? When I met my pal Marc last week we talked about our very similar cameras and I said how much I love my old, worn and scratched 40D. He said something similar about his nearly identical 50D, but after a while he added in a different context, that 10MP would be so far out-of-date now, that he could not live with it any more in a DSLR, all ergonomic advantages aside. *PUHHH* Of course I got wild and called him a snob and an idiot, a victim to marketing and to not have any knowledge about photography and several other names and things...
But of course we had to resolve this, we had to try, all artistic thoughts aside, purely technical:
The test
This may sound quite time-robbing, but it actually was only a matter of about an hour. Choosing between the test-shots took longest.
First, we both chose the lenses we thought of being sharpest. I chose the 50/1.8, the 85/1.8, the Sigma 18-125 (yes, I'm serious!) and the Samyang fisheye. My buddy chose his 50/1.4, the 100/2, the 135/2L and his 70-200/4L. Then we mounted his 50D on his high-tech tripod, because I don't own one any more, since I have image stabilized lenses ;-) But that's a different story and might be worth another article... for now, I was happy, that he had one. Then we turned the camera to "Neutral" picture-style, sharpening, like all other sliders, to 0 (zero!), the camera to "M"-mode @ 1/2000th second shutter speed, ISO 200 and mounted the first lens, turned it to f/4 and manual focus, turned on live-view, focussed via 10x magnification to the center and took a shot. Then we defocussed the lens completely and repeated the procedure two times. Then we turned the aperture to f/5.6 and the shutter-speed to 1/1000th and repeated it, again three times. Then the same at f/8. From these 72 we chose the sharpest photo on my PC at 200% magnification. We did this WITHOUT knowing, which shot was from which lens, to avoid being biased in any way. When we had identified the sharpest of our shots and had looked up the lens in the EXIF-data, as well as it's sharpest aperture, we mounted it on my 40D, put the 40D on the tripod in the exact same position (the head was left like it was, only the plate on the camera was changed), "Neutral" picture-style, sharpening 0 and so on. We set everything completely identical and took a row of 10 shots, just to be sure. After choosing the sharpest of these 10 again on my PC at 200%, I started post-processing: First, I resized the 40D - pictures to the exact same size as the 50D-images: 4752x3168 pixels. I did this 5 times, using every resampling-method my software knows: Bilinear, Bicubic, Smart size, Resizing of pixels and weighted average. From these 5 pictures I chose the sharpest again. Then I took the crops of the pictures, same size, same position in the frame, everything exactly the same. And at last, unsharp mask was applied to all crops: Radius 0.8, Threshold 0, Amount 200.
The results
So here are the results.
The first astonishing result is, that the sharpest lens of all we had is also the cheapest: The Canon EF 50mm 1:1.8 II. Really. It was a very hard decision between this plastic-toy, the higher-class 50/1.4 USM and the professional Canon EF 135mm 1:2 L USM, but finally we chose the 50mm. Don't get me wrong, all these lenses were exceptional, from the 85/1.8 and 100/2 (which are nearly identical) to the 50/1.4 and yes, even the Sigma and Samyang. but finally we chose the 50/1.8 WITHOUT knowing what lens it was: We looked it up in the EXIF afterwards.
The second surprise was the sharpest aperture, it actually was f/4. As a rule of thumb I normally say, that sharpness theoretically suffers from diffraction on APS-C from f/8 on (f/11 on "fullframe"), but this cheap plastic 50mm is so extraordinarily sharp, that you can really see an all so slight drop in sharpness at f/5.6. Wow. This was not the case with the other lenses, all the others were sharpest at f/5.6 or f/8, sometimes it was impossible to choose between these.
And now, finally, here is the comparison of the 40D to the 50D regarding resolution, or 10MP vs. 15MP:
First EOS 50D, 15 Megapixels, 100% crop from roughly the center, unsharp mask 0.8 pixels and strength 200:
And 40D, 10 Megapixels, resized (bicubic) to 15 Megapixels (exact match 4752x3168 pixels), 100% crop from roughly the center, unsharp mask 0.8 pixels and strength 200:
And two crops from the same picture side-by-side, left 50D (15MP), right 40D (10MP):
And finally, two 200% magnifications. First 50D (15MP):
And 40D (10MP), resized, above settings:
The interpretation
Look at the light areas with less leafes and thin branches to compare fine detail.
These results are practically indistinguishable, you can see the difference for the one and only reason, that the crops are shown exactly besides each other. Yes, the 50D-image seems to show the thinnest branches an all so tiny bit darker, more contrasty, but search and look for a single detail in the 50D-crop that you do NOT find in the 40D-crop at all. And answer yourself: What amount of difference are we talking about here? Look at the pictures! This difference even in a 200%-print would not be worth even only € 10,- price-difference to me. And remember: This is the sharpest lens we had. With any other lens we own, the results will be even closer together, the differences even smaller. Do you own a lens, that's sharper than a EF 135mm 1:2 L USM (and this 50/1.8)? I didn't know that I do before, too, but I doubt that you do, I don't know any.
And what about the 60D, 7D, 18MP? This really will be completely invisible: While the calculatory resolution-difference between 10MP and 15MP at least is about 22% (3.888 vs. 4.752 pixels), the difference between 15MP and 18MP (5.184 pixels) is only 8% even on a purely calculatory basis. If 22% is what is shown above, what will an 8% advance look like? *LOL*
So, to make it short, all this is "IRRELEVANT", nothing more to say. Anyone who says, that he needs a newer body to have more resolution than 10MP in a DSLR is a liar or an idiot or both.
The background
After this test we were left with one question: What may be the reason for this?
One possible explanation could have been, that the lens does only resolve a bit more than 10MP, but not 15MP at it's max. But that would mean, that all of our lenses do have even less resolving power, what would be a) quite unlikely, because we (or at least Marc) had some very high-end-lenses among the tested ones, like the 135/2 or the 70-200/4L, which is among the best zoom-lenses ever tested on any review-site and b) contradict the scientific results at dpreview or photozone or dxomark, which we looked up searching for answers: At least the 50/1.4 was tested there on a 50D and exceeded sensor-resolution at f/4, while the 50/1.8 did so on the 5D Mark II and mine was even a bit sharper than the 1.4. in comparison, the 85/1.8 and the 70-200/4 are both a little bit softer than this, what does exactly match our results, too. So it most likely isn't a matter of the lens.
We searched for answers on wikipedia and scientific sites and articles and while it is quite well known, that more pixels on the same sensor-size lead to more noise in most cases, the reason behind this phenomenon actually may have another impact on pictures. I'll make this as simple as possible, for the scientific explanation and terms, please search the internet or a library. The reason for the increasing noise is, that, if you pack more pixels (meaning photo-diodes), on the same area, the single diodes have to get smaller and / or be closer together. While Canon claims to have achieved the latter, the first has to be the case nonetheless: The photo-diodes got smaller in the 50D. When the diodes get smaller, less light hits every single diode, what leads to more noise, because more "wrong" signals are generated, as the "right" signal is weaker and has to be amplified more - and the amplification generates noise. But the other thing is, that diffraction-effects become more relevant. The smaller the area gets, through which light is collected, the more this light is bended, meaning: The less sharp the picture can be. You see this effect clearly, when you close-down the aperture of your lens - any lens at f/32 does really look soft. An even more simple demonstration for this is to close your eyes nearly completely and make a slit as narrow as possible while still retaining the ability to look through it: The picture you see becomes unsharp and soft. This is the same with the single diodes. When these get smaller, the less sharp is the "picture", that reaches every single diode, so the signal of each diode gets less reliable: Not every pixel, that e.g. has to be completely green, really gets completely green. Translated to the whole picture, this means, that the sharpness on a single-pixel-level gets lower. So, while there still are more pixels in the 15MP-frame, the picture gets less sharp on the lowest level, which doesn't completely compensate for the raised amount, but parts of it. And quite big parts, as we have shown above.
This is the reason, why a theoretical 22% increase in resolution does only result in the increase shown above, of, well... let's say maybe 5%? With a lens exceeding sensor-resolution, of course.
A word about supporting this site
I don’t run this site to earn money. I have a real job to earn my living with, a completely normal job. Since everything I write about here I have bought myself, for myself and with my own money from normal shops or ebay-sellers to actually use it, how much and what I am able to write about , depends on the amount of money that I can save and invest in equipment with good conscience. I share all this, because I want to, not to sell it. But when you find this helpful, maybe even as helpful as buying a magazine or book, of course you can support me, if you want. Your benefit is, that you help me being able to afford things to write about here.
You can use the “Donate” – button on the left to directly send a small amount of money (or a big amount, if you insist). You don’t need a paypal account to do so, every method is possible. If you decide to donate 99 cents, I’m thankful for it, because 10 people being as kind as you, make one new filter tested! The default currency is US $, but it works well with € or nearly any other currency, too.
But even more simple and without any cost or other disadvantage at all it is, to simply use this link to Amazon (or the one on the left) before buying anything there. For you it’s exactly the same as going there directly by typing the web-adress in your browser, you just klick this link first! It doesn’t matter, what you buy or where in the world you buy it, be it underpants, a pen, a cupboard, a lens or a Leica M9, be it in Germany, the USA, the UK or Australia: Amazon’s servers simply realize, that you came there through the link on my page and I get a small percentage of anything you buy FROM THEM. There’s absolutely no cost or other disadvantage for you, Amazon simply pays for my “advertisement” through this. I originally didn’t want to ever do any advertising personally. But then I decided to break this rule for Amazon. I’m a completely satisfied customer and buy everything from them. It’s the only shop in the world I would personally and on my private basis really rate a complete 100% in every regard. They have perfect service, even do call you back, answer emails with real, personal writing, extremely fast delivery even on Christmas-day, always perfect and completely new items, are never considerably more expensive than the very cheapest internet-sellers, have an extremely fast refund-system without being picky or having ever displeased me in any way and sell every good I have ever wanted to buy. They work on a completely different level than any retailer I have ever tried, and deliver it directly to me, without robbing me time and money to drive to the city or mall. I wouldn’t advertise them, if I wasn’t convinced, that it is OK to do so.