Question regarding ultrawide lenses on crop and full frame cameras

[insanitybeard]

I'm a hobbyist photographer who mainly shoots landscapes, and have been using a 7D for the last 3 years. My ultrawide landscape lens is the Canon EF-S 10-22, which does yield good results, but coupled with the pixel dense 7D does struggle with fine detail towards the corners when the files are viewed near to 100%. Whilst I always expected that going full frame would give benefits especially regarding landscape photography, and is something I intend to do in the medium term when i can afford it, I was quite amazed at the difference in sharpness comparing the sample crops on The Digital Picture: http://www.the-digital-picture.com/Reviews/ISO-12233-Sample-Crops.aspx?Lens=271&Camera=736&Sample=0&FLI=0&API=3&LensComp=100&CameraComp=453&SampleComp=0&FLIComp=0&APIComp=3
Looking at this comparison, with both lenses set to f8 and at their widest setting to provide a similar FOV in their respective formats, the EF-S 10-22 used on a 60D crop body @10mm appears very unsharp and with more visible CA than even the 17-40L ( itself not regarded as a stellar performer at the wide end on FF until considerably stopped down) used on a 1Ds mk. III. Is this showing the limitations of ultrawide glass on pixel dense crop sensors or could it be partly down to a poor copy of the 10-22 being tested? This difference indicates that I would see a marked improvement in IQ by switching to a FF body, even using the 17-40, which I already own. Any thoughts/ input appreciated!

 

[qwerty]

Full frame lenses have an advantage not because of pixel density per se, but because when you view the final image at a given output size the FF image does not need to be magnified as much. A 40-something MP full frame camera (i.e. with the same pixel size as the 7d) with the 17-40 would show the same qualitative results compared to the 10-22 on the 7d (and might actually have slightly better resolution due to increased pixel density, though this effect would be small near the corners as the blur from the lens dominates). This is basically due to the difference between lp/mm (line pair per millimeter at a given contrast) and lp/ph (line pairs per picture height) - lp/mm is a measure of the 'sharpness' of the lens, while lp/ph tells you the 'sharpness' of the resulting image. I am too lazy to check right now, but I would expect that the 10-22 has a better lp/mm (because it is a newer design and because of the higher pixel density of the 7d), but not enough to overcome the size advantage of the FF sensor.

If you go to photozone.de, you can see a comparison of the image resolution with the 17-40 tested on a 5dii, 50d, and a 350d. (You need to compare the center points between the different format sizes; the APS-C 'edge' and 'corner' is at a different place than that for full-frame.) You can see that the image resolution (lp/ph) at the center is best for the full frame, next best for the 15MP (38MP FF equivalent density) APS-C and worst for the 8MP (20MP FF equivalent density) APS-C.

 

[Don Haines]

As a 60D shooter, also with a 10-20mm ultrawide lens, I agree with the lack of image quality, particularly in the corners. I found that with static landscapes that I get much better results with the 17-55 and stitching multiple images together to get a decent ultra-wide shot with minimal distortion.

If things are moving, that solution is terrible.... FF would be so much better for wide angle photography.

 

[Mt Spokane Photography]

Its very difficult to design a ultra wide angle lens and get sharp undistorted edges, and still sell it at a popular price. Most users of rebels would not consider a high price for a truly excellent lens.
The new15mm Zeiss is easily the best.
http://www.bhphotovideo.com/c/product/850101-REG/Zeiss_1964831_Distagon_T_15mm_f_2_8.html

 

[Artifex]

Its very difficult to design a ultra wide angle lens and get sharp undistorted edges, and still sell it at a popular price. Most users of rebels would not consider a high price for a truly excellent lens.
The new15mm Zeiss is easily the best.
http://www.bhphotovideo.com/c/product/850101-REG/Zeiss_1964831_Distagon_T_15mm_f_2_8.html

I totally agree with you.
However, if you are looking for a very sharp and even cheaper ultra wide angle lens, you can check for the Samyang 14mm f/2.8. The biggest down sides? Very heavy "mustache" distortion, no possibility to put filters and the fact that it is a manual lens, though at 14mm, autofocus is not very useful, especially if you use hyoerfocals. If you can live with those, you can have a 14mm almost as sharp as the 2360$ Canon 14mm, but for about 400$.

 

[traveller]

Full frame lenses have an advantage not because of pixel density per se, but because when you view the final image at a given output size the FF image does not need to be magnified as much.

I think that you're getting your concepts a bit mixed up with film; how exactly does a pixel get "blown up" -it is a dimentionless unit?

A 40-something MP full frame camera (i.e. with the same pixel size as the 7d) with the 17-40 would show the same qualitative results compared to the 10-22 on the 7d (and might actually have slightly better resolution due to increased pixel density, though this effect would be small near the corners as the blur from the lens dominates).

And then you start talking about pixel density, which you stated was not the issue!

 

[insanitybeard]

Its very difficult to design a ultra wide angle lens and get sharp undistorted edges, and still sell it at a popular price. Most users of rebels would not consider a high price for a truly excellent lens.
The new15mm Zeiss is easily the best.

I quite agree that wide angles are one of the hardest lens groups to design, even more so for crop sensors, but what astounds me with The Digital Picture's sample crops is not just the difference in corner sharpness, it's everywhere- centre frame included! Could there be other factors influencing this?

 

[yablonsky]

Hi,

I tested the 10-22 on my 50D. The result was poor compared to my 17-40. I sold it after 2 weeks. Later I moved to FF with the 5D mark II and sold the 50D. My 17-40 was better on the 5D mark II than on the 50D. The size of the sensor is the clue. You get much more detail in 100% magnification. All lenses are performing better than on the 50D.

 

[qwerty]

Full frame lenses have an advantage not because of pixel density per se, but because when you view the final image at a given output size the FF image does not need to be magnified as much.

I think that you're getting your concepts a bit mixed up with film; how exactly does a pixel get "blown up" -it is a dimentionless unit?

A 40-something MP full frame camera (i.e. with the same pixel size as the 7d) with the 17-40 would show the same qualitative results compared to the 10-22 on the 7d (and might actually have slightly better resolution due to increased pixel density, though this effect would be small near the corners as the blur from the lens dominates).

And then you start talking about pixel density, which you stated was not the issue!

The image resolution (in lp/ph) is equal to the lens resolution (in lp/mm) times the sensor size (about 24mm for a FF sensor, about 16mm for a APS-C sensor). Alternatively and equivalently, to get a 4x6 print, the FF image as projected by the lens onto the sensor needs to be enlarged by some 4.2x while the APS-C image needs to be enlarged by about 6.3x.

Increasing pixel density while keeping the same sensor size will generally increase resolution (both lp/mm and, as a result, lp/ph) (at least to the point where the pixel pitch is such that the lens "outresolves" the lens by a large margin, as is probably the case in the FF corners of the 17-40 at apertures larger than F/8). This is only tangentially related to the change in overall image resolution due to sensor size, and discussed ad nauseum elsewhere in these forums. It also produces a change in the opposite direction than the OP was asking about (e.g. asking if higher pixel density was causing lower image resolution).

 

[Mt Spokane Photography]

Its very difficult to design a ultra wide angle lens and get sharp undistorted edges, and still sell it at a popular price. Most users of rebels would not consider a high price for a truly excellent lens.
The new15mm Zeiss is easily the best.
http://www.bhphotovideo.com/c/product/850101-REG/Zeiss_1964831_Distagon_T_15mm_f_2_8.html

I totally agree with you.
However, if you are looking for a very sharp and even cheaper ultra wide angle lens, you can check for the Samyang 14mm f/2.8. The biggest down sides? Very heavy "mustache" distortion, no possibility to put filters and the fact that it is a manual lens, though at 14mm, autofocus is not very useful, especially if you use hyoerfocals. If you can live with those, you can have a 14mm almost as sharp as the 2360$ Canon 14mm, but for about 400$.

Not mine: Its easily the worst lens I've ever owned, and I've had some really bad ones. My Canon 15mm FE is much sharper.

 

[Artifex]

Its very difficult to design a ultra wide angle lens and get sharp undistorted edges, and still sell it at a popular price. Most users of rebels would not consider a high price for a truly excellent lens.
The new15mm Zeiss is easily the best.
http://www.bhphotovideo.com/c/product/850101-REG/Zeiss_1964831_Distagon_T_15mm_f_2_8.html

I totally agree with you.
However, if you are looking for a very sharp and even cheaper ultra wide angle lens, you can check for the Samyang 14mm f/2.8. The biggest down sides? Very heavy "mustache" distortion, no possibility to put filters and the fact that it is a manual lens, though at 14mm, autofocus is not very useful, especially if you use hyoerfocals. If you can live with those, you can have a 14mm almost as sharp as the 2360$ Canon 14mm, but for about 400$.

Not mine: Its easily the worst lens I've ever owned, and I've had some really bad ones. My Canon 15mm FE is much sharper.

It probably was a bad copy, as Samyang's quality control is probably the worst I have ever seen. First copy I got couldn't focus farther than 1,5m. I returned it and got a great one, very sharp straight from f/2.8 and stunning at f/5.6.

 

[insanitybeard]

Further to this topic, I've checked some more of The Digital Picture's sample crops, this time for the 70-200 f4 L IS (a lens that I own): http://www.the-digital-picture.com/Reviews/ISO-12233-Sample-Crops.aspx?Lens=404&Camera=736&Sample=0&FLI=3&API=2&LensComp=404&CameraComp=453&SampleComp=0&FLIComp=4&APIComp=2 compared on a 60D crop body and 1Ds mk III full frame body- the situation is the same, the full frame crop is sharper everywhere, and indeed, this seems to be a general thing- the same lens used on a FF body appears sharper than the same lens used on a crop body. In the comparison, the aperture is set at f5.6, so diffraction should not be having any effect on the crop sensor result. I understand that the full frame sensor will deliver more total resolution, and that the 1Ds III has 21.1 megapixels compared to the 60D's 18 but can somebody explain why the full frame crop is sharper everywhere? Surely it must be possible to achieve sharp results at a lower total resolution? If pixels are a dimensionless unit, then surely it cannot be because of the crop sensor image requiring more magnification to appear the same size as the full frame image. I apologise is this has been discussed before, but I'm really trying to understand the reasons for these differences and how much difference going full frame would make to me.

 

[qwerty]

Further to this topic, I've checked some more of The Digital Picture's sample crops, this time for the 70-200 f4 L IS (a lens that I own): http://www.the-digital-picture.com/Reviews/ISO-12233-Sample-Crops.aspx?Lens=404&Camera=736&Sample=0&FLI=3&API=2&LensComp=404&CameraComp=453&SampleComp=0&FLIComp=4&APIComp=2 compared on a 60D crop body and 1Ds mk III full frame body- the situation is the same, the full frame crop is sharper everywhere, and indeed, this seems to be a general thing- the same lens used on a FF body appears sharper than the same lens used on a crop body. In the comparison, the aperture is set at f5.6, so diffraction should not be having any effect on the crop sensor result. I understand that the full frame sensor will deliver more total resolution, and that the 1Ds III has 21.1 megapixels compared to the 60D's 18 but can somebody explain why the full frame crop is sharper everywhere? Surely it must be possible to achieve sharp results at a lower total resolution? If pixels are a dimensionless unit, then surely it cannot be because of the crop sensor image requiring more magnification to appear the same size as the full frame image. I apologise is this has been discussed before, but I'm really trying to understand the reasons for these differences and how much difference going full frame would make to me.

I think I explained the reason for this pretty clearly above; however, some people seem to be getting confused by details that do not get to the crux of the reason why FF cameras tend to have better image sharpness (image sharpness being distinct from, though related to, lens sharpness and sensor resolution).

Imagine a world where sensors were perfect, and the only degradation to the image came from the lens (1). If you put a FF lens on an FF and a APS-C sensor with the same subject distance, the image formed on the APS-C sensor would be a crop from the FF sensor; it would be exactly the same as you would get by taking the FF image and cropping it (now you see why APS-C sensors are called crop sensors). If you blew both images up to the same size, you would magnify the lens aberrations to a greater extent for the APS-C sensor than the FF sensor, so the APS-C image would look worse (just like occurs if you blow up a crop of an image). If you magnified the images by the same ratio, the APS-C image would look like a crop of the FF image; details in the image would look the same but the FF image would give a wider FOV.

The same thing happens if you take the photos with the same FOV (either by moving farther away with the APS-C camera, or by using a shorter focal length), with the obvious caveats for perspective or differing lens performance at differing focal lengths

For a concrete example, imagine that you have a lens that gives a uniform resolution of 100 lp/mm at MTF-50. In the above example, the FF image would then be able to show 2400 line pairs vertically in an image, while the APS-C would only be able to show about 1600 line pairs with the same contrast. So, of course the FF image would look sharper.

If the above isn't crystal clear, don't worry about the other issues raised in this thread.


And with respect to the 'pixels are a dimensionless unit so surely that can't be the reason...' stuff, it hurts my head too much trying to figure out enough about what you do not understand to try to explain to you what's going on (3). Specifying a number of pixels and a format also gives you the pixel pitch, which has the dimension of length (cf. the dimensions for lens resolution.) But on an even more fundamental level, just because something is dimensionless doesn't mean that changes in that number can't explain some phenomenon (2).


(1) Somehow I can only read that sentence in Don LaFontain's voice...
(2) Just search for "Reynold's number", which is part of the introduction to dimensional analysis in a fluids course
(3) Charles Babbage could have phrased this more elegantly

 

[insanitybeard]

Qwerty, thanks for taking the time and effort to respond! I will take a good long look at what you have written to try and understand better. I had previously assumed that the increased pixel density of the crop sensor camera was 'magnifying' any flaws of the lens, which is one of the points you make.

 

[qwerty]

Qwerty, thanks for taking the time and effort to respond! I will take a good long look at what you have written to try and understand better. I had previously assumed that the increased pixel density of the crop sensor camera was 'magnifying' any flaws of the lens, which is one of the points you make.

Its not the increased pixel density that is magnifying the flaws with APS-C, its the fact the the image sensor (and thus the relevant portion of the image projected by the lens) is smaller. Higher pixel density (for a fixed image size) will actually lead to sharper images in general (1).

If you made a FF camera with the same pixel density of the 7d, images from it would be much sharper than those from the 7d (and slightly to significantly sharper than a 20-something MP FF camera, depending on lens quality). But pixel density is one of those factors you should ignore for now.

(1) I am not even going to get into noise considerations here. Also note that once your pixel density is high enough, additional increases will not make a substantial change in image quality.

 

[insanitybeard]

Qwerty, if you're out there I have re-read your posts and the key points make sense to me, namely that the larger sensor will of course offer more resolution because of it's larger size and therefore greater line widths per picture height, and also because of the cropped sensor image requiring greater relative enlargement to appear at the same size of the full frame image- not to be confused with the total number of pixels, which, for example, comparing the 1DX and 7D might be similar, but the 7D's are crammed into a much smaller sensor. Of course I always knew the FF sensor would provide better IQ, due to it's larger surface area and consequently greater light gathering ability, it was just the difference in sharpness particularly apparent in TDP's sample crops that astounded me.

The other thing I think was confusing me was the units for resolution- lp/mm, which is independent of sensor size, and lp/ph, which is not.

Anyway, it makes more sense to me now, thanks again for taking the time.