Something I learned about "sharpness" on the 7D

elboss wrote in post #9186538
Great discussion ..

But isn't the 18mp sensor denser therefore compensate for the larger output and should exhibit a comparable definition characteristics @100% with a smaller sensor ?

I'm not sure what you're saying about "compensate for the larger output," but the second part of your statement goes off track because at 100% you are looking at the higher resolving sensor at a greater enlargement of the format.

Looking at the format under greater enlargement certainly does mean that any image faults caused by other factors (lens aberrations, subject motion, camera motion, et cetera) will become more visible. That is not, however, because the sensor is resolving more, it's because you're enlarging the format more.

It's not the sensor, it's the enlargement factor. If you didn't enlarge it any more than before, you would not need a higher shutter speed or a better lens. It's not the sensor, it's the enlargement factor.

If your desire is to make larger prints at high image quality, then a higher resolving sensor is one of the factors you must include--and then you see that you have to improve your camera holding (tripod or higher shutter speed) and then the quality of the lens. But remember that the reason you must go to those measures is that you want more enlargement, not because the sensor is making you do it.

What people have misunderstood is that they haven't noticed that when they compare sensors of the same size but with different resolutions both at 100%, they are actually comparing different sized enlargements of the same format. So any non-sensor-caused fault that is governed by the size of the fault projected by the lens is being magnified at a different rate.

When I shot film, I could make a "contact sheet" of my 35mm shots, and they all looked perfectly sharp. A "contact sheet" is made with the film resting directly on the printing paper, so the print is the same size as the negative. But when I put the negative in the enlarger and raised the head to, say, 16x20...whoo, boy, where did all that blur come from? I didn't see it on the contact sheet!

Now, if I'd used Kodak Recording Film which was, as I recall, ISO 1600 (base), I would never have attempted to enlarge it to 16x20 because I already knew KRF was too grainy and too low in resolution...so I never would have seen that my cheap lens was not corrected well enough for a 16x20 print.

But now I decide to use Panatomic-X, which is very sharp and has very fine grain, so I am ecouraged to enlarge it to 16x20...and finally I can see that my lens is cheap.

The thing is that the lens had always been too poorly corrected for a 16x20 print. I had just never tried such a great enlargement with the less-sharp film.

But the thing to remember is that it's not the "fault" of the film (or in our case now, the sensor), it's the fact that I'm looking at a greater enlargement, and it's the greater enlargement that reveals the faults of everything else in the system.

Part of that misunderstanding we see here is probably because they don't realize that "100%" means the monitor is scaling the image to a 1:1 map of image pixels to monitor pixels, so the presentation of an image with more image pixels is going to be larger on the monitor than the presentation of the image with fewer pixels. But if they notice the rulers and see the size of the image, they will see that they're comparing two different enlargements.

pwm2 wrote in post #9187792
Just a tiny note here.

18MP is three times the number of MP as 6MP. But not three times the magnification when looking at a 100% crop. Three times the magnification from 6MP would require 6MP*3^2 = 54MP. In this case, the extra magnification is sqrt(3) or 1.73.

So camera shake with a 18MP body will smear the same distance on the sensor, and give the same amount of smear on the print. But the smear will reach 1.73 times longer on the sensor, counted in pixels. It is then up to the user to decide if the shutter time should be decreased to take advantages of the extra pixels.

You're right. My example needs better resolution.

amfoto1 wrote in post #9184703
Using a Spyder3 on your eyeball?

Of course not. He is using an i1 (formerly known as Eye-One).

Tim

It really is a common problem that anyone considers their 1.6x cameras to give more reach, and that they consider a higher MP camera to allow more cropability.

But a good camera vendor do not build an expensive camera with a huge amount of pixels to allow the image to be cropped more.

Whenever digitally sampling data, the input data must not contain details at a higher resolution than the sensor can sample, or the end result will be severe aliasing effects. A soundcard has a lowpass filter just as there is a lowpass filter in front of the CMOS sensor in the camera. But it is hard to create good lowpass filters that can give very high damping of information above the resolution of the sensor, while not giving artefacts within the sampled region.

In audio, you may use a AD converter capable 192k samples/second, which means that it can handle audio frequencies of 96kHz - way more than we can hear. The reason is that there is no need to have a lowpass filter with a very sharp cutoff. The lowpass filter can start cutting somewhere above 20kHz but it is enough that the damping is really strong when reaching 96kHz and higher. Such a filter will not only be cheap to build. It will also be almost invisible within the 0-20kHz range - the important part that we can hear.

Same with images. Having a camera with 50MP or 100MP means that the sensor will outresolve the lenses, so we will not gain any extra details in our photos. But when the sensor outresolves the lens, it is possible to use a lowpass filter with a quite flat slope. We don't get more details, but we get a picture with less alias problems and with less color artifacts for high-contrast subjects.

But to get the advantages from the higher sensor resolutions, we have to start by accepting that the extra pixels wasn't added to allow us to crop tighter. When all we think about is how tight we can crop, and what the picture looks like at 100%, we will be unhappy with the high-resolution cameras and totally miss their advantages.

pwm2 wrote in post #9188433
But to get the advantages from the higher sensor resolutions, we have to start by accepting that the extra pixels wasn't added to allow us to crop tighter. When all we think about is how tight we can crop, and what the picture looks like at 100%, we will be unhappy with the high-resolution cameras and totally miss their advantages.

Well, it may not have been added specifically to allow us to crop tighter--the higher resolution may decrease aliasing. But it does--because of decreasing aliasing and other benefits--allow us to crop tighter. That can't be seen at 100%.

One of the major IQ benefits I've discovered in the 5D2 over the 5D1 is no more ugly hair moire. If I compare pixel per pixel between my 5D1 and my 5D2, I cannot see that the 5D2 has no hair moire. It can only be seen in the final output.

JefferyG, RDKirk & other participants .... Thanks for the informative replies.

I have been reading posts which states almost the same as following:

I must then ask, why are manufacturer producing camera with very high resolution?

5D_USER wrote in post #9192587
I have been reading posts which states almost the same as following:


I must then ask, why are manufacturer producing camera with very high resolution?

Besides the obvious - many customers buys a camera based on high MP - the extra MP is advantageous at large apertures where the lens isn't diffraction limited. And the extra MP is advantageous when shooting high-contrast subjects because it reduces problems with low-pass filters and aliasing.

You do the audio industry use 96 or 192kHz sampling rates for sound? Why do the CD players use 96 or 192kHz over-sampling when playing sound? It is a fact of life that it helps to sample at a frequency significantly higher than the bandwidth of the input signal. In this regard, the DSLR manufacturers will still have a long way to go until they have sensors that have four, sixteen or maybe 64 times as many pixels in relation to the feature sizes they want to capture.

To some part, people then complain about noise and dynamic range. But a lot of these complaints comes from people looking at 100% crops, instead of looking at actual prints. 9 (3x3) noisy pixels will look noisy on the monitor. But when printed, they will not look noisier than a single large pixel printed to the same area. And the lack of dynamic range of a tiny pixel is offset by the virtual dynamic range of the 3x3 (but noise-modulated) pixel group compared to one larger pixel.

As long as people put their concentration on 100% crops, they will be unhappy with high-MP cameras. When starting to focus on their prints instead, they will notice that more MP isn't a bad thing only invented to fool people into buying new cameras.

Alas, it is almost impossible to get a pixel-peeper to start focusing on the prints, since they are so convinced that their eyes can't be fooled. What they see on a 100% crop just has to be the truth...

5D_USER wrote in post #9192587
I have been reading posts which states almost the same as following:

I must then ask, why are manufacturer producing camera with very high resolution?

Ignoring the issue of how diminishing the returns are or not at high spatial frequencies, the higher pixel density means the MTF of of the sensor and anti-alias filter rolls off at higher spacial frequencies giving better micro contrast at frequencies that are important for image quality.

Interesting thread. Below is a comment made by Canon which seems relevant to the original post. It pertains to Cameras from a few years ago but I guess would still apply today.

'images from more recent EOS professional digital cameras can still look softer than those from earlier models. This is because recent models have more pixels, which means that pixel size is smaller (11.5µm on the EOS-1D; 7.2µm on the EOS-1Ds Mark II). Smaller pixels are more sensitive to camera shake, as a smaller movement will cause the image to move across more pixels. You need to hold the camera steadier - ideally on a tripod.

For the same reason, sports photographers also need to re-think their shutter speeds, as blur from subject movement will be more apparent on cameras with more pixels. Where possible, consider increasing the shutter speed, even if this requires an increase in ISO speed'

A very interesting post which makes a lot of sense. I used to be one to obsess over the 100% screen sharpness of my photos but have gradually realized that as long as the print looks good and a resized web version looks good then it doesn't really matter too much.

Since getting my 7D though i havn't found any difference in 100% sharpness between that and my 5D2 or even my 400d.

Roy C wrote in post #9193398
Interesting thread. Below is a comment made by Canon which seems relevant to the original post. It pertains to Cameras from a few years ago but I guess would still apply today.

Sure does, Roy. Those are almost the exact words used by Mike Owen - European Product Planning Manager for Canon - in his seminar 'Introduction to the EOS-1D Mark IV' I attended back in October.