I'm surprised at how much lighter the Canon is than the Sigma and others. Naturally, for an overloaded octo that's important, but be under no illusion: its lighter 'cos it has more plastic and less glass!
People obsess a lot about lenses- it's a favourite topic around the bar in "The Photographer's Arms". Back in the days of film it mattered much more: once the shutter clicked there was little you could do to correct any optical deficiencies. But digital has changed all that - the perceived quality of a lens (when looking at a real image instead of a test target) is much more to do with how the image has been post-processed. Put it this way: with good processing, I'd challenge anyone to tell which lens took which image.
If you really want that wide angle (and at 10mm on a crop sensor it is VERY wide), then simply go with your priority: lowest weight or lowest cost. However I would strongly recommend that you rent one first - not to assess the quality but to find out if it gives you the images you want. The extreme wide angle creates compositional problems/challenges and (on an MR) is likely to get the arms in the shot unless it's pointed severely downwards.
With either the Canon or Sigma lens you will be aware of optical imperfections - barrel distortion, chromatic abberation, vignetting, edge softness, etc. So to get the best out of either lens you'll want to look at DxO. Sorry, Emowillcox, but just because Adobe is a big company does not make LR the best product (as Microsoft also demonstrates). DxO may be a small company but they make the hardware/software that is used (by manufacturers, magazines, etc.) to measure the performance of lenses and sensors, so they know a thing or two about the subject! The core attraction of DxO is that it corrects all the known shortcomings of the particular lens and sensor you are using. So take the same scene with the Canon and the Sigma, process them both through DxO and I guarantee you'll not be able to tell them apart. And you cannot say that about LR (or any of the other raw processors). Check some of the images on my website - they are taken with Nikon D200, D300, D700 and D3s with various Nikon and Sigma lenses, as well as Canon 1DsMkII and 1DsMkIII cameras with various Canon lenses (including 17mm and 24mm shift lenses) - but you simply cannot tell which is which. They are all processed through DxO, of course.
Just to touch on a few other points raised in this thread:
Barrel distortion (the rendering of straight lines as curved lines) is more noticeable with subjects closer to the camera and near the left/right edges, but it's there all the time. DxO eliminates it. Period.
So-called "wide angle distortion" is different and is actually two things. First, there's optical stretching (technically called anamorphism): objects near the edge of the frame appear elongated. Then there's keystoning: the unwanted convergence of
parallel lines (the "leaning building" effect). The first has no real solution (it's just physics), the second can be avoided by keeping the camera dead-level or corrected in post-processing. Neither are anything to do with lens quality.
A tilt-shift lens does not magically make any of these problems go away, but gives the photographer various mechanisms to avoid or at least minimise/relocate the issues. Flying a tilt-shift lens would be somewhat impractical (since you can't twiddle it's little knobs up there) and unlikely to produce superior results.
People obsess a lot about lenses- it's a favourite topic around the bar in "The Photographer's Arms". Back in the days of film it mattered much more: once the shutter clicked there was little you could do to correct any optical deficiencies. But digital has changed all that - the perceived quality of a lens (when looking at a real image instead of a test target) is much more to do with how the image has been post-processed. Put it this way: with good processing, I'd challenge anyone to tell which lens took which image.
If you really want that wide angle (and at 10mm on a crop sensor it is VERY wide), then simply go with your priority: lowest weight or lowest cost. However I would strongly recommend that you rent one first - not to assess the quality but to find out if it gives you the images you want. The extreme wide angle creates compositional problems/challenges and (on an MR) is likely to get the arms in the shot unless it's pointed severely downwards.
With either the Canon or Sigma lens you will be aware of optical imperfections - barrel distortion, chromatic abberation, vignetting, edge softness, etc. So to get the best out of either lens you'll want to look at DxO. Sorry, Emowillcox, but just because Adobe is a big company does not make LR the best product (as Microsoft also demonstrates). DxO may be a small company but they make the hardware/software that is used (by manufacturers, magazines, etc.) to measure the performance of lenses and sensors, so they know a thing or two about the subject! The core attraction of DxO is that it corrects all the known shortcomings of the particular lens and sensor you are using. So take the same scene with the Canon and the Sigma, process them both through DxO and I guarantee you'll not be able to tell them apart. And you cannot say that about LR (or any of the other raw processors). Check some of the images on my website - they are taken with Nikon D200, D300, D700 and D3s with various Nikon and Sigma lenses, as well as Canon 1DsMkII and 1DsMkIII cameras with various Canon lenses (including 17mm and 24mm shift lenses) - but you simply cannot tell which is which. They are all processed through DxO, of course.
Just to touch on a few other points raised in this thread:
Barrel distortion (the rendering of straight lines as curved lines) is more noticeable with subjects closer to the camera and near the left/right edges, but it's there all the time. DxO eliminates it. Period.
So-called "wide angle distortion" is different and is actually two things. First, there's optical stretching (technically called anamorphism): objects near the edge of the frame appear elongated. Then there's keystoning: the unwanted convergence of
parallel lines (the "leaning building" effect). The first has no real solution (it's just physics), the second can be avoided by keeping the camera dead-level or corrected in post-processing. Neither are anything to do with lens quality.
A tilt-shift lens does not magically make any of these problems go away, but gives the photographer various mechanisms to avoid or at least minimise/relocate the issues. Flying a tilt-shift lens would be somewhat impractical (since you can't twiddle it's little knobs up there) and unlikely to produce superior results.
Last edited by a moderator: