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

[TheSuede]

For one, I asked you what exactly you meant by "the APS-C lens has to be better than the FF lens"?

I followed that up with a clarifying question asking if you were talking about two different lenses such as one EF-S lens and one EF lens.

Sorry if I was the cause of any additional confusion...

-Yes, to get two images, one taken with APS and one taken with FF appear equally sharp when viewed at equal presentation size in your presentation format of choice, the lens used on the APS camera needs to be sharper, when sharpness is defined by "lines or line-pairs per mm".

-Yes, at least I am talking about two different lenses in the comparison. Why compare an 85mm on FF with an 85mm on APS? they don't "do" the same thing any more. If the shot taken with the FF camera was at a 10 feet target distance, you'd have to back of to 16 feet to get the same framing on APS. And by backing off to 16 feet, you change the perspective of the shot so much that it wouldn't even be the "same shot" any more.

Taking that into account is where the "FF is sharper than APS" starts to make sense....
Using a 135L 2.0 wide open on a 5D3 gives the same field of view, the same short DoF and the same amount of noise as using the 85L 1.2 on a 60D. But the 135 on 5D3 image will be a LOT sharper and more contrasty.

For the same reason, a bog-standard 85/1.8 on FF gives sharper images with the same short DoF as a 50L 1.2 does on APS. And a 50mm vs 35mm F1.4 on APS and so on.

 

[Rocky]

For one, I asked you what exactly you meant by "the APS-C lens has to be better than the FF lens"?

I followed that up with a clarifying question asking if you were talking about two different lenses such as one EF-S lens and one EF lens.

Sorry if I was the cause of any additional confusion...

-Yes, to get two images, one taken with APS and one taken with FF appear equally sharp when viewed at equal presentation size in your presentation format of choice, the lens used on the APS camera needs to be sharper, when sharpness is defined by "lines or line-pairs per mm".

-Yes, at least I am talking about two different lenses in the comparison. Why compare an 85mm on FF with an 85mm on APS? they don't "do" the same thing any more. If the shot taken with the FF camera was at a 10 feet target distance, you'd have to back of to 16 feet to get the same framing on APS. And by backing off to 16 feet, you change the perspective of the shot so much that it wouldn't even be the "same shot" any more.

Taking that into account is where the "FF is sharper than APS" starts to make sense....
Using a 135L 2.0 wide open on a 5D3 gives the same field of view, the same short DoF and the same amount of noise as using the 85L 1.2 on a 60D. But the 135 on 5D3 image will be a LOT sharper and more contrasty.

For the same reason, a bog-standard 85/1.8 on FF gives sharper images with the same short DoF as a 50L 1.2 does on APS. And a 50mm vs 35mm F1.4 on APS and so on.

Sharpness has got nothing to do with perspective, or DoF. You quoted all the lenses are wide open for the tests. It is not fair either. Every body knows that ALL lenses is not in their best behaviour when they are wide open ( more distortion, less resolution and less contrast ). Are you sure that the lenses can out resolve the APS-C sensor under test??? if not, then the result becomes a test of the lenses, not the sharpness of the sensor.

 

[p666]

Images talk much better than words when you are talking about photography. I decided to test it myself, and took my 60D, and 5D Mark II, and attached the same 35mm prime lens, and took the same shot. The 5D image of 5616 x 3744 I cropped to 1.6, which gave me 3510 x 2340. I then downsampled the 60D image to the same size, and took a 100% crop of both for screen.

So you threw away any detail resolved by the 60D but not by the cropped center of the 5D2. Why would you expect a different result then the one you got?

I'll be the first to say that with current bodies the crop reach advantage only comes into play if you're making large prints from files which are cropped even further than APS-C. It is there. I've got 9-10 MP, 16x20, 7D sports prints that would not have worked with the roughly 3 MP left over had I been shooting a 5D2. But if you're making 13x19's that aren't cropped any further, you'll never see a difference.

But that still doesn't excuse performing your test in a manner which purposely throws away any detail gain in the 60D file. Would you do that in real life? (Same question for the test which started this thread: would you produce prints with no sharpening? Or do you blindly apply equal sharpening to everything regardless of subject / lighting / lens / sensor / final application and view size?)

Ok you do have a point, however, taken the same files, and this time taken the same 3510x2340 crop on the 5D to get the same perspective as the original crop 60D. Now I upsampled the 5D image in Photoshop to the original APS-C size of 5184x3456. Then have taken a 1:1 pixel crop of the centre to compare. Unfortunately there is a slight brightness difference due to changes in ambient light. Quality appears to be very very similar again; there is no disadvantage to simply cropping a FF shot in this instance. Well for 60D vs 5DMk2, although yes, the Mk2 is 3x the price.

5D image:
http://farm9.staticflickr.com/8218/8407024129_52e33f3df5_o.jpg

60D image:
http://farm9.staticflickr.com/8211/8407024335_b175c2d3c9_o.jpg

I'd be interested in seeing a 7D vs 5D Mk2 under the same scenario to see the differences if any. Taking the same lens, mounting them on the other body and isolating the same area as taken by the crop sensor.

2. There are isolated incident that the APS-C can actually sharper than the FF if the lens use for FF have a severe problem in curvature of field or severe unsharp conrners. Since APS-C snesor only using the center part of the FOV of the lens. These problem may not exist. Therefore the APS-C is actually sharper.

I don't see why, because if you were taking the same lens on the full frame, even if it had bad corners, if you took a center 1.6 crop size on the FF to get the same FOV, you would be throwing away the FF bad corners. It would be same as then comparing the same image area. Nearly every lens chart I have ever looked at the lens is sharper in the middle, so unless you were comparing edge of FF to centre of APS-C (which is not really a reasonable comparison) that is not the case.

 

[Sporgon]

@p666, I posted some examples on page 2 of this thread done by my partner in Building Panoramics between 5D mk2 and 650D. I think this is more or less the same sensor as 7D.

However we didn't do it as you describe: it was non scientific, used different lenses to get same focal length. At ISO 100. I can assure you that in this scenario there was just no difference.
Apart from DoF.

Our conclusion with modern APS is that if you're shooting low ISOs, have good lenses and don't require ultra wides there is no difference. Also this thread has highlighted that many people really don't understand what is happening when you enlarge the digital info from you camera chip - either on screen or in print.

If you choose ( new ) full frame and want to keep same focal length and speed there are huge cost implications so the thread has had relevance. For many people their choice of FF will be because of personal and not necessarily technical reasons, and there would be zero differences in their pictures if they had been shot on good APS.

Me? I use FF because I want ultra wide, I want my lenses focal length to be what I understand from 35mm, and I can't stand expensive cameras with pop up flash.

@ dtaylor, really you had the last word in your post about post processing in digital. This is so true.

 

[Rocky]

2. There are isolated incident that the APS-C can actually sharper than the FF if the lens use for FF have a severe problem in curvature of field or severe unsharp conrners. Since APS-C snesor only using the center part of the FOV of the lens. These problem may not exist. Therefore the APS-C is actually sharper.

I don't see why, because if you were taking the same lens on the full frame, even if it had bad corners, if you took a center 1.6 crop size on the FF to get the same FOV, you would be throwing away the FF bad corners. It would be same as then comparing the same image area. Nearly every lens chart I have ever looked at the lens is sharper in the middle, so unless you were comparing edge of FF to centre of APS-C (which is not really a reasonable comparison) that is not the case.

If you read my post carefully, I started by staying that both FFand APS-C are of the same pixel count. Also sharpness has got nothing to do with FOV or DOF. Your disussion talks about cropping the FF to APS-C size. Therefore you have already thrown away more than half of the pixel from FF. Do you expect a 10 Mp picture to be sharper than a 22 MP APS-C picture with an very sharp lens (center area only)??

 

[Sporgon]

@ Rocky : + 1

If nothing else this thread has shown many people don't realise that digital enlargement is all down to the number of pixels on the sensor, and not the physical dimensions of the sensor "format", unlike film format.

 

[Rocky]

2. There are isolated incident that the APS-C can actually sharper than the FF if the lens use for FF have a severe problem in curvature of field or severe unsharp conrners. Since APS-C snesor only using the center part of the FOV of the lens. These problem may not exist. Therefore the APS-C is actually sharper.

well, you have soon 24x36 with the same pixel density then you have both cropping and 24x36 to chose between.
as todays d800 APS crop 15,3Mp. Next will be 54Mp 24x36 if they keep the schedule and the APS crop 24Mp
[/quote}
My discussion is based on both FF and APS-C ARE of the same pixel count. Not pixel density.

 

[tortilla]

Sharpness has got nothing to do with perspective, or DoF.

I disagree. That's not even true when you only shoot ISO test charts. The larger DoF, the smaller the circle of confusion and the higher the resolution/sharpness (disregarding diffraction etc.).

 

[dtaylor]

Some poster throw in "post production", that is not a valid arguement either.

The discussion is merely academic apart from post production. Nobody publishes images without post work, or at the very least non-neutral camera settings.

1. For the same pixel count on both formats, both can be as sharp as each other, assuming that the lens can out resolve both sensor in the imaging area. That will put a lot of stress on the lens for APS-C sensor. The lens need 1.6 time reolving power compare to the FF.

This is false. Resolution is not a single number. It's a graphed MTF curve of contrast achieved at ascending resolutions as recorded on the sensor/film. The higher the resolution, the lower the detail contrast or sharpness.

If you have a FF and APS-C sensor with equal pixel counts, and all other factors are equal (lens, scene, etc), then the APS-C sensor will produce an image that is less sharp than the FF sensor. Within its cropped area, it's recording at a higher resolution than the FF sensor, and the lens will deliver less contrast at that higher resolution.

2. There are isolated incident that the APS-C can actually sharper than the FF if the lens use for FF have a severe problem in curvature of field or severe unsharp conrners. Since APS-C snesor only using the center part of the FOV of the lens. These problem may not exist. Therefore the APS-C is actually sharper.

Yes, there are some lenses where this is the case.

3. FF will be sharper if the lens cannot out resolve the APS-C sensor but it will out resolve the FF.

Sort of. Again, resolution is not a hard number. It would be more accurate to say that FF will be sharper than expected (see point 1) if the lens MTF curve favors lower resolutions, i.e. if it really drops towards the bottom of the chart in the region where the APS-C sensor sits.

It should also be noted that contrary to popular belief the vast majority of lenses can out resolve current DSLR pixel pitches at MTF10. You would almost need a coke bottle to fail to do so. If you want to see lens challenging pixel pitches you have to look at compact cameras.

4. There is no doubt that FF will have better picture quality than thre APS-C, due to less noise, better DR.

Way too broad of a statement. There are actually people who will take this to mean that a 1Ds (original) will have better IQ then a 60D or 7D. It won't. Technology is at least as important as sensor surface area (noise) and pixel size (DR). Further, at this stage of technology, there are ISOs where noise is simply a non issue between the two formats.

6. In print, most of the difference will not be obvious due to the printing process.

I would say it's more due to scaling, but agreed.

7. How many people can tell the difference in music between amplifier with 0.01% distotion and amplifier with 0.005 % distortion??

In a double blind study I would be surprised if anyone could.

 

[rs]

@ Rocky : + 1

If nothing else this thread has shown many people don't realise that digital enlargement is all down to the number of pixels on the sensor, and not the physical dimensions of the sensor "format", unlike film format.

Yes, digital enlargement only takes into account the number of MP. But that argument is conveniently missing out many significant parts of the overall image capture process.

Light, and the scene captured by the lens/camera are analogue. The resulting print (unless its a particularly low res digital print) does its best to look analogue. And similarly speaking, with a good enough display (such as a retina display), its again a good approximation of analogue - and both print and display have light coming off them (reflected or lit) which is analogue, which is how your eyes can see it. With a digital camera, the in between bits of the process are a combination of analogue and digital. It starts off with a lens which is very much analogue. That creates an analogue imaging circle. That in turn is captured by an analogue sensor, which generates analogue electrical signals, fed through analogue amplifiers, before finally getting converted to digital. Only after all that is the RAW file or JPG created and stored. Then its transferred to a computer, possibly had PP done on it, before finally leaving the digital domain when it gets put into some sort of approximation of analogue so we can see it.

By talking about one stage of that process from half way through (the captured MP), and then talking about how magnifying that is the only thing which matters is missing out half the picture.

Comparing two similar sized sensors results in quite similar results, so comparing APS-C to FF is not so clear cut to visualise which is best. So to prove a point, lets compare a much smaller 18MP sensor such as the Panasonic ZS30 to a full frame 18MP sensor such as the 1D X. Printing both images at the same size has exactly the same digital enlargement, but the tiny sensored Panasonic compact analogue image will have been enlarged much more. And I can't imagine anyone trying to pretend that a 1D X, even with the same effective focal length, aperture and ISO will produce images when printed large that could be mistaken for ones taken with the Panasonic. The only way the two could ever look comparable is if their pictures are printed almost postage stamp small - that way the Panasonic analogue image has been enlarged by a more sensible amount, hiding its faults much better.

If we are talking about a purely digital image with no noise, diffraction, lens softness etc. such as CGI, then yes, MP is king. After all, enlarging can only be an enlargement of each pixel, and we don't want to enlarge it to the point where it looks pixelated. But digital photos are not perfect digital files. They are merely digital representation of an analogue capture of the scene. Enlargement of the sensor is the key factor to quality, not enlargement of each subdivision of one stage of the process.

 

[AprilForever]

Full frame is not sharper than APS-C. It has shallower depth of field when framed the same, requiring a higher f-stop to get more in focus. Why did Ansel belong to the f64 club? Because much less than f64, even with tilts, and not much is in focus. Full Frame has the grim problem of using the bad part of lenses too, making corners truly grim on all but the most fine of lenses.

But, mainly, a pixel is a pixel, and it really doesn't matter what size the sensor was. Full Frame can deliver a better image sometimes, but there is a reason Canon's lead Action camera was a crop body (APS-H). Most likely, tachnology has progressed to the point where the 7D MK II will have better IQ at 1.6 crop than the old 1d MK IV had at its 1.3 crop.

APS-C has some serious advantages. Google "7D bird photography". You will see some amazing things...

 

[Rocky]

Sharpness has got nothing to do with perspective, or DoF.

I disagree. That's not even true when you only shoot ISO test charts. The larger DoF, the smaller the circle of confusion and the higher the resolution/sharpness (disregarding diffraction etc.).

As long as your lens can out resolve the senser, the DOF does not matter any more. Also the DOF is governed by the acceptable size of circle of confussion. As long as you have the acceptable size of circle of conffussion, your DOF can either be 1 inch or 10 ft, both will give you the same result on the point of focus.

 

[Sporgon]

Talking of "action cameras" I was really surprised to see how many 7Ds were being used at the London Olympics.

 

[rs]

Full frame is not sharper than APS-C. It has shallower depth of field when framed the same, requiring a higher f-stop to get more in focus. Why did Ansel belong to the f64 club? Because much less than f64, even with tilts, and not much is in focus. Full Frame has the grim problem of using the bad part of lenses too, making corners truly grim on all but the most fine of lenses.

But, mainly, a pixel is a pixel, and it really doesn't matter what size the sensor was. Full Frame can deliver a better image sometimes, but there is a reason Canon's lead Action camera was a crop body (APS-H). Most likely, tachnology has progressed to the point where the 7D MK II will have better IQ at 1.6 crop than the old 1d MK IV had at its 1.3 crop.

APS-C has some serious advantages. Google "7D bird photography". You will see some amazing things...

The reason the Canon 1D line of cameras had a 1.3x crop is when it was introduced in 2001, it was the largest sensor Canon could source - as far as I know, it was the largest in a production camera at the time. FF sensors didn't hit the market until the Contax N Digital was introduced in 2002. That same year, Canon introduced the much more successful FF 1Ds. Due to the limits of the technology it had much slower read out, so it was nowhere near as fast as the 1.3x crop sensor 1D. They continued as two lines of bodies - FF for the highest image quality, 1.3x crop for speed. Nikon managed to combine FF and speed with the D3 (although it was slightly slower than 1D mk III and much lower res than 1Ds mk III). The D4 was their successor, and Canon's rather elaborate answer to it was to combine the two 1 series lines with the 1D X.

Granted, sometime lots in focus is desired. But that's why lenses can be stopped down, and why tilt and shift lenses exist. To get as large a depth of field on FF as on 1.6x, simply use an aperture 1.6x smaller. You will need an ISO just over a stop higher if you want the same shutter speed, but the advantage of FF is still there. And TS-E lenses take things to another level. When you don't want a huge DoF, try a lens like the 70-200 II or the 100L. What's in focus is capable of being so detailed even at the edges of the frame that you'll no doubt be immediately aware of the difference FF makes.

 

[TrumpetPower!]

I see there's still a great deal of misinformation going on in this thread.

The simple version is that, even if the number of megapickles is exactly the same, the larger format will intrinsically produce sharper images.

I'm attaching an image to demonstrate this. It's a simple enough experiment that anybody can perform: set up a still life. Take two images, one with a long lens and the other with a wide lens -- or with a zoom at either end of its range. Crop the wide image to the same field of view as the telephoto image, and scale the two to the same pixel dimensions. You're left with two images with the exact same number of megapickles, and everything else is either constant or can be controlled for. The sensors of the two images is identical, so you've got the same S/N ratio, the same everything.

To do the most fair comparison possible, you'll want to use the same shutter speed for both images, an aperture sufficiently smaller for the telephoto to match the depth of field, and an equivalent higher ISO to keep the exposure the same. That's exactly what I did below.

And, you can clearly see: the image generated with the larger, 135 format sensor is significantly sharper than the image generated with the smaller, P&S-sized sensor. And, again, I stacked the deck against the larger format in every reasonable way I could.

Pixel density is important, yes. But so is format size. For the absolute best results, you'll want the highest pixel density you can get in the largest format you can get. 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.)

Cheers,

b&

 

[dtaylor]

The simple version is that, even if the number of megapickles is exactly the same, the larger format will intrinsically produce sharper images.

All other things being equal, and assuming glass that does not have serious problems at the edges which are trimmed by APS-C...yes.

It's a simple enough experiment that anybody can perform: set up a still life. Take two images, one with a long lens and the other with a wide lens -- or with a zoom at either end of its range. Crop the wide image to the same field of view as the telephoto image, and scale the two to the same pixel dimensions.

While I agree with the principle, this test does not demonstrate it. The differences are way too large to be reliable (different lenses; scaling algorithms).

And, you can clearly see: the image generated with the larger, 135 format sensor is significantly sharper than the image generated with the smaller, P&S-sized sensor. And, again, I stacked the deck against the larger format in every reasonable way I could.

Nah, the differences cut against the WA shot. Again, I agree in principle, but this test isn't the way you demonstrate it.

 

[TrumpetPower!]

Nah, the differences cut against the WA shot.

Seriously?

The wide-angle shot is with the sharpest wide-angle lens ever made for the format, and one of the sharpest lenses ever made, period. And it's at its sweet spot for aperture. Plus, the shot was at ISO 100.

The telephoto shot was with a lens with a built-in soft focus filter (with the filter turned to its least soft setting, of course), fer chrissakes, and it was shot at f/32 (waaaaay past the point of diffraction), and it was shot at ISO 6,400.

I mean, seriously. I compared a Coke bottle with the Hubble, and the Coke bottle still beat the Hubble.

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.

Cheers,

b&

 

[skitron]

7. How many people can tell the difference in music between amplifier with 0.01% distotion and amplifier with 0.005 % distortion??

In a double blind study I would be surprised if anyone could.

Not that it amounts to a hill of beans for this discussion, but it's actually pretty easy if you listen to the cymbals and the speakers/headphones are up to the task.

[/tangent off]

 

[rs]

Nah, the differences cut against the WA shot.

Seriously?

The wide-angle shot is with the sharpest wide-angle lens ever made for the format, and one of the sharpest lenses ever made, period. And it's at its sweet spot for aperture. Plus, the shot was at ISO 100.

The telephoto shot was with a lens with a built-in soft focus filter (with the filter turned to its least soft setting, of course), fer chrissakes, and it was shot at f/32 (waaaaay past the point of diffraction), and it was shot at ISO 6,400.

I mean, seriously. I compared a Coke bottle with the Hubble, and the Coke bottle still beat the Hubble.

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.

Cheers,

b&

You're completely right. With those lenses and those settings, that test couldn't have been more stacked against the full frame setup. Yet full frame easily and conclusively won.

 

[AlanF]

It seems to me that all that "experiment" has shown is that if you use a low quality telephoto lens to take a photo of, say, a small bird sitting in the middle of a tree you get a better image of it than taking a photo of the whole landscape with a wide angle lens and blowing up the teeny weeny central section. Yes, I do know that you downsized the image from the telephoto, but it makes for the same thing in a conparison.

Basically, you have shown the image of the bird that covers the whole sensor - that is makes use of most of the megapixels - is far better than have it cover just a few pixels at the centre of the sensor. That is not a relevant comparison for FF vs crop.