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

[dtaylor]

Nah, the differences cut against the WA shot.

Seriously?

There is no way to quantify and compensate for the many variables which are different. Your test, however much you think it proves the point, is simply invalid.

As for your argument about the lenses, it's pointless because the crop-n-scale trumps everything else. Even the f-stop selection in this case. You didn't have any where near as many pixels on target, and then you let a computer algorithm (which one? what settings?) fill in the missing data. This bears no analogy or relevance to having an actual APS-C sensor, of the same MP resolution, take the shot with the same lens and aperture positioned for the same FoV. (Side note: the soft filter argument was silly. That does not come into play on that lens unless you select it.)

Again, I agree in principle because the principles are well established and understood. But the test you posted was horribly flawed.

If you don't think that this test demonstrated the importance of format size over pixel dimensions...well...all sorts of unflattering adjectives come to mind, with "hopeless" being the one best suited for a family-friendly forum.

Nobody can tell what your test demonstrated because of all the confounding variables.

I've compared FF and APS-C a number of times...all other factors being equal...and I've always found the same thing: FF is sharper out of camera, but the range is small enough to be closed by bumping up sharpening in camera or post. When the sensors are otherwise comparable of course.

[dtaylor]

OK - here's a simple question. Has anyone got a Pentax Q (12MP, small sensor) and a 5Dc (12MP, FF sensor) - if we look at a comparison of the two, it should solve this issue once and for all, even though the 5Dc is using much older tech.

Why don't you just head over to DPReview or Imaging Resource and download studio files for the cameras you wish to compare? I would be careful with compact cameras because they will probably have sharpening even at 'neutral' settings.

As to crop vs FF: You will find FF a bit sharper. Until you run a sharpening pass on crop. The difference is there, but it doesn't amount to all that much at low to mid ISO. At high ISO crop cameras start to apply more NR, causing more blur, and post sharpening to compensate starts emphasizing noise.

The point of TrumpetPower's test is both of them are adjusted to a very low number of MP, therefore they're both on an equal footing (downsampling a high MP image to low MP is very similar to starting with low MP in the first place).

That's not even close to an equal footing. As good as our scaling algorithms are, scaling up is destructive while scaling down can actually yield more final detail then direct sampling at the lower resolution.

[tortilla]

But, if you have a choice between two cameras of different formats and different megapickles, go with the bigger format. (With, of course, the usual caveats that sufficient technology age can skew the results.)

Except 1D MK III vs 7D.

Oh wait, the 1D MK III ist two years older, I can't compare them ... 

Just joking... as you said, due to technology developments (and other reasons, as discussed) it doesn't make sense to compare sensors only in regard of size. (And likewise the whole thread doesn't make much sense IMO).

[TrumpetPower!]

So, dtaylor, we have a test in which we have two images with the exact same number of megapickles, the exact same shutter speed, the exact same field of view, the exact same depth of field, the exact same sensor technology, better optics on the smaller sensor...and your point is that this is somehow an invalid comparison of two sensors with the same number of megapickles but different areas because of some magic resampling faery dust that made the image with the larger sensor become dramatically better than the image with the smaller sensor.

Just out of idle curiosity, had I performed the test with film -- you know, that plastic stuff with the smelly chemicals? -- would you still be maintaining that the test was somehow invalid because the enlarger has some sort of magic faery dust rendering the comparison simply invalid and horribly flawed?

Or, I have a better idea.

You're such an expert at this. How would you go about performing a comparison between two sensors with the same number of megapickles and the same sensor technology but different format sizes?

That is, after all, the question under discussion: does format size matter? So, how, exactly, would you hold all the variables constant except for format size?

Tell us your ideal experimental protocol.

b&

[dtaylor]

Alan, I was being gracious in ignoring your complaint, as it indicates that you really don't understand even the simplest fact about what a crop sensor camera actually is. Specifically:

Alan's complaint was spot on. He very clearly understands what's going on.

[TrumpetPower!]

But, if you have a choice between two cameras of different formats and different megapickles, go with the bigger format. (With, of course, the usual caveats that sufficient technology age can skew the results.)

Except 1D MK III vs 7D.

Oh wait, the 1D MK III ist two years older, I can't compare them ... 

Just joking... as you said, due to technology developments (and other reasons, as discussed) it doesn't make sense to compare sensors only in regard of size. (And likewise the whole thread doesn't make much sense IMO).

Eh, not exactly.

Until relatively recently, comparing formats was commonplace and trivial. The exact same film you loaded in a cartridge into your SLR was being used in large format backs. And you really had to be clueless in those days to insist that you could get the exact same quality out of the smaller format.

We've been in a rapid growth phase in the digital world where, in about a decade, we're seeing quality improvements comparable to what used to only be achieved by going with a larger format. The latest-and-greatest APS-C sensors have basically caught up with the earliest full-frame sensors. Those first full-frame sensors were going toe-to-toe with medium format film; today's full-frame kits outperform the view camera that Ansel Adams used -- and today's high-end medium format kits surpass anything anybody's ever had access to, perhaps with the exception of some of the insane custom setups.

However, there's very good reason to suspect that this age of rapid growth is just a phase, and that it won't be all that long (historically speaking) before it'll come to an end. Eventually, we'll be back to marginal improvements, and all camera sensors will have roughly the same pixel densities and noise characteristics and the like. When that day comes, once again, the defining characteristic will be format size, not sensor technology.

And we even see that to some extent today, within a manufacturer's lineup. Canon's currently-produced full-frame cameras outperform their currently-produced APS-C cameras by about a stop or so (what you'd expect), and the same is true with Nikon's full-frame v APS-C, and Sony's, and Leica's, and so on. Between manufacturers the lines aren't always so clear...but, then again, there were always wars between Kodak and Fuji films.

Cheers,

b&

[Sporgon]

@ TrumpetPower, on page 2 I posted four comparison pictures of 18MP APS and 21MP FF, shot at 100 ISO, subject close.

There is just no difference in sharpness, colour depth, tonal grad, shadow, - nothing.

You commented that you weren't surprised............

[TrumpetPower!]

Alan, I was being gracious in ignoring your complaint, as it indicates that you really don't understand even the simplest fact about what a crop sensor camera actually is. Specifically:

Alan's complaint was spot on. He very clearly understands what's going on.

He very clearly does not.

The question was whether two sensors with the same megapickles but different sizes would produce images of comparable sharpness.

There is only one way to test such a question: make two images using sensors of different sizes and the same number of megapickles. Which is exactly what I did.

You know, all y'all who think that sensor size has nothing to do with sharpness an only to do with noise or tonality or whatever are more than welcome to demonstrate otherwise. Complaining that magic resampling faery dust somehow introduced more sharpness into a test that was significantly stacked in favor of the loser isn't exactly all that impressive.

Cheers,

b&

[dtaylor]

So, dtaylor, we have a test in which we have two images with the exact same number of megapickles,

False. From your test post:
Crop the wide image to the same field of view as the telephoto image, and scale the two to the same pixel dimensions.

You had two files with very different dimensions and then let a computer scale them. Computer generated pixels are not equal to directly sampled pixels. If they were, we could all shoot 1 MP sensors and scale to any print size we wanted.

I'm tired of debating your test. Your variables were all over the place, and it was basically a test of pixels on target and computer scaling, not of comparable FF/crop sensors. Yet you defend it like it was your child. Catching up on this thread I find you are actually being insulting to those who point out the test flaws. Grow up and try listening to them instead.

Just out of idle curiosity, had I performed the test with film -- you know, that plastic stuff with the smelly chemicals? -- would you still be maintaining that the test was somehow invalid because the enlarger has some sort of magic faery dust rendering the comparison simply invalid and horribly flawed?

It would be invalid for the same reason it's invalid now: fewer sampling points on target. It would just be grains instead of pixels.

You're such an expert at this. How would you go about performing a comparison between two sensors with the same number of megapickles and the same sensor technology but different format sizes?

Gee...I don't know...how about getting the actual cameras and putting the same lens on each; changing position to achieve the same FoV; and making sure that all other factors (scene; lighting; aperture; shutter; settings) are equal, thereby isolating the difference you want to measure (sharpness)? 

On second thought...nah...that's just crazy talk 

[dtaylor]

There is only one way to test such a question: make two images using sensors of different sizes and the same number of megapickles. Which is exactly what I did.

Last I checked you used one sensor, two lenses, cropped one image resulting in fewer MP, then scaled up thinking that interpolated pixels = true MP.

[TrumpetPower!]

@ TrumpetPower, on page 2 I posted four comparison pictures of 18MP APS and 21MP FF, shot at 100 ISO, subject close.

There is just no difference in sharpness, colour depth, tonal grad, shadow, - nothing.

You commented that you weren't surprised............

Yes, I wasn't surprised. You were comparing APS-C and 135 formats, which are only a stop apart, and latest-generation tech with first-generation tech. You found they were within a stop of each other. Of course! There's only a stop between the formats, and close to a stop of improvements the past decade.

b&

[TrumpetPower!]

If they were, we could all shoot 1 MP sensors and scale to any print size we wanted.

Ah. I see.

You really are that ignorant of the subject, no?

Scaling down is throwing away information. It's as if the information never existed, and the end result is the same as if you had never started with it in the first place.

Scaling up is inventing information. That information must come from somewhere. It's because there's no good place for that information that you can't upscale wantonly.

Gee...I don't know...how about getting the actual cameras and putting the same lens on each; changing position to achieve the same FoV; and making sure that all other factors (scene; lighting; aperture; shutter; settings) are equal, thereby isolating the difference you want to measure (sharpness)? 

On second thought...nah...that's just crazy talk 

Yes, that's crazy. Spectacular fail crazy.

Use the same lens on different formats and you get a different field of view. Change position and you get a different perspective in addition to a different field of view. Keep the aperture the same between different formats and you get different depth of field.

I was very specific in what I did and didn't change, and I did it for a reason. My test really was a true apples-to-apples comparison. The shooting position was the same, so the perspective was the same. The shutter speed was the same so the motion blur (not that there was any) was the same. The aperture was different, yes, but it resulted in the same depth of field. The ISO setting was different, yes, but it resulted in the same exposure -- and, unsurprisingly if you know the basics of photography, it resulted in basically the same amount of noise.

The only remaining variables were format size and sharpness. And, as is utterly unsurprising to anybody who knows as much about photography as you'd learn from an introductory textbook, the larger format captures more information per fraction of the frame, and thus a sharper image.

b&

[TrumpetPower!]

There is only one way to test such a question: make two images using sensors of different sizes and the same number of megapickles. Which is exactly what I did.

Last I checked you used one sensor, two lenses, cropped one image resulting in fewer MP, then scaled up thinking that interpolated pixels = true MP.

You checked worng.

The full-frame image was scaled down, not up. The small format image was a 100% crop.

And the small format image had a significantly superior lens.

b&

[Sporgon]

@ TrumpetPower, yes indeed, but that test was at close distance with large and easily defined subjects.

We took that camera on a shoot with 5D mk1, ( not the mk2 with which we did the test ) and out of interest shot the same frames on each camera. ( We stitch ). Although the 18MP 650 clearly had more resolution than the 13MP 5D, in the final picture, in the hills and woods behind our main subject, about quarter mile away, the ones from the 5D had more definition and a very subtle improvement in total graduation - what I refer to as "brio" in a picture. It is only a very very subtle difference but it is there.  I presume this is something to do with the fact that when dealing with far away subjects the actual image projected onto the FF sensor will be larger, and this, coupled with the larger pixels collecting more light give the benefit that isn't seem when making "test" shots of close subjects.

BTW, I wouldn't call a 5D Mk 2 first generation technology.

[AlanF]

It is perfectly obvious that a large format with a long focal length lens beats out a small format with the same number of megapixels with a short focal length lens, all things being equal, if they have the same field of view. But your "experiment" did not do that. Your set up had the long focal length lens with a narrow field of view and compared it with a short focal length with a wide field of view and then lost resolution from the small format by cropping.

Alan, are you serious?



Scenario a: two cameras, one with a 4.27mm x 6.4mm sensor and a 24mm lens, the other with a 24mm x 36mm sensor and a 135mm lens. Both sensors have 6.25µ pixels.

Scenario b: one camera with a 24mm x 36mm sensor with 6.25µ pixels. One image is made with a 24mm lens and cropped to only use the 4.27mm x 6.4mm central portion of the sensor; the other image is made with a 135mm lens and the full sensor.

Educate us: what, exactly, is the difference between those two scenarios?

Or are you somehow under the misimpression that lens focal length is a function of format size? Is that it? Do you think that a 24mm lens on a 4.27mm x 6.4mm sensor is still a wide-angle lens? If so, you need to brush up on your introductory optics. That small sensor would need a 4.27mm lens to produce the same field of view as a 24mm lens does on a 24mm x 36mm "full frame" sensor. That's why a 50mm lens is still a 50mm lens whether it's on a Rebel or a 5D, but it's a normal lens on the 5D and a telephoto on the Rebel. The lens hasn't changed, and it's still the exact same focal length as always, regardless of what camera you attach it to.

b&

Yes, I am serious. Although you claim in your customary gracious manner that I am clueless about sensors and optics, I do, in fact, know a little. The first is that crop sensors generally have much smaller pixels than larger sensors. For example, the Canon PowerShot G9 with its 7.6x5.7mm sensor has a pixel size of 2.5 µm compared with the 6.25 µm pixel 5D III. So your scenario with both large and small sensors having 6.25 µm is both false and irrelevant for the arguments in this forum because resolution depends, among other factors on pixel density or size.  The smaller size of the crop pixels goes to compensate for other factors and is crucial in any analysis.

Consider the following, which highlights the importance of pixel size and applies to a real scenario.  The height of an image I from object of height H a distance v away from the lens of focal length f for v>>f is I = H*f/v.  According to the Nyquist-Shannon sampling theorem and the Nyquist limit, I has to span ~2 pixels in order to be resolved. The 7D has a pixel size of 4.3 µM, so I has to be greater than 8.6 µm to be resolved. For the 5D III, I has to be >12.5 µm. For the same lens on a 7D and 5D III, f/v is the same for both at the same distance away and so I is the same. For the 5D III and a 100mm lens, two lines 3 mm apart cannot be resolved at distances greater than 24 metres. For the 7D and its crop sensor, the distance is 34.9 metres. So, between 24 and 34.9 metres, the 7D can resolve the two lines whereas it is optically impossible for the 5D III so to do.

It is your ignoring of the effects of pixel size (density) that invalidates your arguments.