Is Full Frame sharper than APS-C? My answer here

Samuel H

Samuel H

Major Contributor
I have to decide between APS-C and full frame, and one of the things to look at is sharpness. With APS-C, as opposed to full frame, you're only using the central portion of the image circle, the sweet spot: you're not using those blurry corners, and this should lead to sharper pictures. On the other hand, because the sensor is smaller but final image resolution is the same, you're blowing up the analog image that the lens projects on the sensor; this is bad, because every defect will be magnified. Which force is bigger?


I ran some tests trying to find out, shooting stills with a 5D2.


My conclusion: as a general rule, a lens will always be sharper on full frame than on APS-C, both in the center and in the corners.


You can see the tests and read a bit more about them here: http://www.similaar.com/foto/lenstestsff/lenstestsff.html
 
Probably true, as long as video is 2K or lower res. But the difference will probably appear again in 4K video.
 
Well, it's not exactly the same. Here it is "Does a central crop from a 36mp D800 resolve more than a 22mp 5DmarkII?".

Of course, if you put a horrible lens on the full frame camera and an awesome one on the APS-C, you're giving the smaller sensor a better chance to shine. But personally my question was "will my lenses deliver sharper images on full frame or on APS-C?", so for me the fair comparison is "my very nice 35mm set at f/2.8 on APS-C" vs "my very nice 50mm set at f/4.5 on full frame", and the latter wins by a wide margin... unless your final image is moderately low res, then they are about the same.
 
It is simple physics really.
System MTF is a product of lens MTF multiplied by sensor MTF. With the same pixel count the APS-C sensor will be much denser. It will need much better optics to reach the MTF of a FF system.

Here is an example: a fullHD full frame (36mm wide sensor) image has a resolution of around 26-27 lp/mm. This means it is feeded by spatial frequencies up to around 13 lp/mm (anything above this will only cause aliasing). So you are really interested in the lens' MTF result in spatial frequencies up to this value.
A fullHD APS-C image (say, 24mm wide) has a resolution of 40 lp/mm. It is feeded by spatial frequencies up to 20 lp/mm.

So, to reach an equivalent result with the APS-C system, you will need a lens with MTF at 20 lp/mm roughly equalling the MTF of the FF lens at 13 lp/mm. But as MTF always drops with spatial frequencies increasing, you will need a really stellar lens for your APS-C system to compete. That's why larger sensors generally achieve better delineation and tonality even with average quality lenses. And this is why large format and medium format matter to high-end photographers, not the shallow DOF factor.

Now with video there are other things to consider, as mentioned above. Compression and motion blur largely render the difference irrelevant.
 
Indeed, once I've seen the empirical tests, everything seems obvious. But before tests, the "you're only using the sweet spot of the lens" argument seemed to have some merit too; not enough, it seems.
 
Samuel H said:
My conclusion: as a general rule, a lens will always be sharper on full frame than on APS-C, both in the center and in the corners.
Click to expand...

I don't think this rule will hold up for video once you add Super35 ( slightly larger than APS-C ) sensors into the mix.

Compare the footage from a Canon C100/C300 camera against any Full Frame video camera and the Full Frame camera will lose when it comes to sharpness and detail.

As we move up to 4K and 8K acquisition formats, then dedicated Super35 / APS-C lenses will be required to record the same ( or better ) detail as Full Frame lenses.
 
TheDingo said:
I don't think this rule will hold up for video once you add Super35 ( slightly larger than APS-C ) sensors into the mix.

Compare the footage from a Canon C100/C300 camera against any Full Frame video camera and the Full Frame camera will lose when it comes to sharpness and detail.

As we move up to 4K and 8K acquisition formats, then dedicated Super35 / APS-C lenses will be required to record the same ( or better ) detail as Full Frame lenses.
Click to expand...

The C100 and C300 are special in that they take the QFHD sensor image, and then interpolate it down using an intelligent pixel binning scheme to generate a 444 1080p image, before compression. So it will obviously be a better image than what a 4k camera, with a 4k sensor, will produce, because it is starting with a superior image, because of a new way to read and process sensor data.

If you could find a full frame camera that let you record raw mosaiced images at 3840x2160, than you could probably get an image that is comparable if you post-processed it similar to how the Cxxx cameras process their images.
 
TheDingo said:
I don't think this rule will hold up for video once you add Super35 ( slightly larger than APS-C ) sensors into the mix.

Compare the footage from a Canon C100/C300 camera against any Full Frame video camera and the Full Frame camera will lose when it comes to sharpness and detail.
Click to expand...
Only because all existing full frame video cameras "cheat". Line skipping, etc. When there is a full-frame camera that uses the full sensor as efficiently as an S35 video camera uses its sensor, then the advantage will once again be with the larger sensor.

As an example, "full frame" VistaVision film is substantially sharper than S35 film.

As we move up to 4K and 8K acquisition formats, then dedicated Super35 / APS-C lenses will be required to record the same ( or better ) detail as Full Frame lenses.
Click to expand...
That's the case for any format though. The smaller the sensor, the more demanding the requirements of the lens, all other things being equal. Of course, all other things are rarely equal, and a S35 chip with 1920x1080 resolution is far less demanding of its lens, than a full frame camera with 30 megapixels would be. In the end, it's all about resolving information down to the pixel level. Smaller pixels require sharper lenses. Bigger pixels can get away with softer lenses.
 
An S35 sensor with a Metabones Speed Booster could produce a sharper image vs. the same lens on a full frame sensor, due to increased MTF performance. Right now I'm not aware of any digital camera that can capture full frame video with full 1920x1080p resolution (1000+ TV Lines measured (horizontal) resolution (true 1920h pixels)). As we later discovered, the 5D3 appears to be scaling ~1620x910 into 1920x1080 (which matches closely to our chart measurements of around 810 TV Lines (this is post first artifacts, but is actual, usable resolution when post-sharpened to remove what appears to be Gaussian-like blur)). The 1DC looks similar in full frame 1080p mode (1DX has strange anisotropic aliasing/sharpening artifacts). I will test the FS700 with Speed Booster against the 5D3 shortly (same lenses).
 
J.F.R. said:
So explain to me how the GH2 Hack is able to produce such sharp images?
Thanks
Click to expand...

Current full frame cameras have hardware/software/business limitations which reduce effective resolution. My 'ancient' Panasonic TM700 camcorder ($800 when new) also provides higher resolution vs. current full frame cameras (when given sufficient light).
 
Why not? At the same pixel size, a larger sensor will have more pixels, and therefore more resolution.

Assuming all other things are equal, having more surface area results in more resolution.
 
My tests were:
* for a given mpix count (at least not very different)
* with a similarly strong optical low-pass filter (in relation to their pixel size)
* with complete sensor readout in both cases
* with similar image processing

Why? Because the choice I have to make keeps all those things constant. Other people have to make different choices, and so these tests may be useless to them. In particular, these tests are useless if you want to compare a GH2 with a 5D2 (actually, that should be obvious to anyone who has seen a side-by-side comparison).

(I'm trying to get a hint of what will be true in 2 or 3 years, not to see what was true in yesterday's cameras)



edit: you can also look at it as follows: I wanted to know if "you're only using the sweet spot of the lens so your images will be sharper" had any merit; it turns out it's not enough to counter the additional magnification that happens when you use a smaller sensor
 
Last edited:
You must log in or register to reply here.
Back
Top