1. Reminder: Please use our affiliate links to get to your favorite stores for holiday shopping!

FF ~ APS-C ~ m4/3 Comparison - UPDATED

Discussion in 'This or That? (MFT only)' started by DHart, Oct 13, 2013.

  1. DHart

    DHart Mu-43 Hall of Famer

    Jan 7, 2010
    Scottsdale, Arizona
    Don
    I am embarking on a project to create landscape images for printing to 40" and larger.

    My available cameras are:

    • 5DMkII w/17-40 f/4 L lens
    • Ricoh GR
    • GX7/EM5/EP5/EPL5 bodies with many m4/3 lenses

    I wanted to do a small (but exceptionally challenging to the gear) test to see just how much difference there would be in capturing fine details between these cameras.

    This is not a test of dynamic range, nor of high ISO noise, nor a comparison of size, nor weight of gear, but simply a test of the ability to clearly render exceptionally fine image detail.

    Of course, it should be expected that the FF sensor would produce finer small detail than the APS-C or m4/3 cameras. But there is much talk by some that the differences are so minor as to be insignificant, these days. I would agree that for many applications, the differences in image detail are insignificant. And for some others, the difference is still significant.

    Certainly, for reproduction at small print sizes such as 11x14 and smaller and for use on the internet... there is little reason to go with a FF sensor if one is simply seeking greater detail. The detail is plenty good enough with m4/3 for these applications.

    But when printing images such as landscapes (or a bar of liquor bottles with all of the fine details on the labels) - which are exceptionally demanding subjects due to the incredible amount of very fine detail (in foliage, for example) - small differences in fine detail can make a nice difference in a large wall print (such as 40" and larger) that will be viewed at relatively close differences, such as in one's home environment.

    Of course, if you back away far enough from a wall print, such additional detail is somewhat less noticeable, however in one's home, walking up to a print is something easy to do and many people will do it to see the finer details in an image (if such details are there to see!).

    To give myself a better idea of the differences between these formats, I selected the 5DMkII, GR, EP5, and my newly acquired GX7 to do a little test.

    All cameras were used on a tripod that didn't move. All captures were RAW and converted in Lightroom to achieve the most pleasing results. The lenses were all set to closely match the 28mm equivalent focal length of the Ricoh GR. The m4/3 lens apertures were set to f/5.6, the GR to f/8, and the 5DMkII to f/11.

    I had no agenda with this exercise other than to observe what the visible differences would be.

    Here is the full image (as created with the 5DMkII). I've changed my hosting, so I can present images better than previously with Photobucket.


    And here are 100% crops of the lower right quadrant of the images:

    GX7 w/14mm f/2.5 lens @ f/5.6 (base ISO of 200)


    GX7 w/12-35mm f/2.8 lens @ f/5.6 (base ISO of 200)


    EP5 w/12-35mm f/2.8 lens @ f/5.6 (base ISO of 200)


    Ricoh GR @ f/8 (base ISO of 100)


    Canon 5DMkII w/ 17-40 f/4 L lens @ f/11 (base ISO of 100)


    Differences are most noticeable when looking at the word "Harmonie" on the Cointreau label, the fine print on the Chateauneufdupape label, the fine print on the round label on the Six Grapes bottle, fine print on the Tequilla Corralejo label, as well as some of the other labels.

    Some folks don't like to view or compare images at 100% and I've never understood that, as 100% view enables one to most accurately evaluate sharpness and IQ - if that is what one would like to evaluate. Some are just not very interested in evaluating fine points of sharpness and IQ, and that's certainly fine for their needs. I like doing so, as it helps me evaluate how well an image is likely to hold up when printed at large print sizes.

    My take away from this exercise (when viewed 100% in LR) essentially confirms what most would expect to find. And that is that the fine details are most crisply and clearly captured with the FF sensor. The APS-C camera came in a close second. Followed by the EP5 and very close to that, the GX7 entries. It would come as no surprise, of course, that the larger format camera would render finer details more distinctly. But it's clear that m4/3, as a format, definitely punches above its weight, given that it's sensor is only about 1/4 the size of the FF sensor!

    None of the example cameras would be a bad choice for typical/smaller print sizes as the results are quite close except at 100% examination in LR.

    As has so often been said in automotive circles, "there's no replacement for displacement." Not to imply, however, that everyone needs "large displacement" for their everyday needs in driving nor in image making.

    Each level of camera used here has good suitability to particular applications and none is perfect for all.

    The bottom line here is that it's nice to have different options for different applications, when one wants them. :2thumbs:
     
    • Like Like x 8
  2. arad85

    arad85 Mu-43 Veteran

    477
    Aug 16, 2012
    Yes, that is a surprise. Unless the APS-C lens was specifically designed to give extra resolution (and the fundamental problem with the GR is you can't separate the lens from the sensor so you can't extrapolate the results), it should appear lower res than the :43: camera (again assuming you used a decent :43: lens).

    Also, if the resolution is that much better, photobucket compression would still show the differences, but they all look pretty similar to me.

    If you'd like to host full sized images, with untouched 100% crops (i.e. you create the JPEG and I'll host)... PM me and I'll give you an e-mail address to send them to...
     
  3. DHart

    DHart Mu-43 Hall of Famer

    Jan 7, 2010
    Scottsdale, Arizona
    Don
    The differences are quite noticeable when looking at the files I uploaded on my computer. But once they got to Photobucket, it appears they were down sampled, which resulted in them all looking almost indistinguishable from each other.

    I will PM you, thanks!
     
  4. cs2012

    cs2012 Mu-43 Rookie

    20
    Jun 27, 2012
    Since the difference in not quite visible in the 100% crops you have posted - and I see very little difference as you yourself realize - it might help if you can make the raw files available and let those who are interested find out for themselves.
     
  5. drd1135

    drd1135 Zen Snapshooter

    Mar 17, 2011
    Southwest Virginia
    Steve
    I have access to a nice printer at the University that will do 48" by whatever prints. About $30 but how often will I do this? I've always been tempted to make a really big print for the wall in my office but to be honest I just can't decide on an image.
     
    • Like Like x 1
  6. DHart

    DHart Mu-43 Hall of Famer

    Jan 7, 2010
    Scottsdale, Arizona
    Don
    Arad and I are trying to come up with a better way to host the samples.

    In the meantime, I'm working on another set of crops, sized so they are less likely to be down sampled in translation.
     
  7. I found that to stop photobucket downsampling images you shouldn't upload files greater than 1024 pixels on the longest side. If they are uploaded at less than 1024 then they should be displaying as is.
     
    • Like Like x 1
  8. DHart

    DHart Mu-43 Hall of Famer

    Jan 7, 2010
    Scottsdale, Arizona
    Don
    Thanks, Nic... the files I uploaded earlier were a good bit larger than 1541 on the longest side. No wonder they got messed up in translation. The ones I just posted were 822 pixels on the long side.
     
  9. arad85

    arad85 Mu-43 Veteran

    477
    Aug 16, 2012
    OK. So here are the untouched files from Don:

    GX7 & 14/1.8 @ f5.6: http://www.arad85.co.uk/hosted/u43/GX7_14-1.8.jpg
    GX7 & 12-35 @ f5.6: http://www.arad85.co.uk/hosted/u43/GX7_12-35.jpg
    Ricoh GR @ f/8: http://www.arad85.co.uk/hosted/u43/GR.jpg
    5DII w/ 17-40 f/4 L lens @ f/11: http://www.arad85.co.uk/hosted/u43/5DMkII_17-40.jpg

    BTW: the 5DII is "cheating" as it is the same subject over a much bigger area so will tend to look sharper ;) The GR does better than I'd expect, but that may be due to the lens. It would be interesting to see the 17-40 on an APS-C sensor....
     
    • Like Like x 1
  10. You can definitely tell that the GR has no AA-filter when you look at the aliased (jagged) edges throughout the image (on the Cointreau label for example).
     
    • Like Like x 1
  11. DHart

    DHart Mu-43 Hall of Famer

    Jan 7, 2010
    Scottsdale, Arizona
    Don
    Andy... thanks for doing that! Photobucket really messed up the original set of files, didn't it!

    The GR sure does a fine job of competing with the 5DMkII... being a little shirt pocket camera... as long as 28mm FOV meets the need.
     
  12. hookgrip

    hookgrip Mu-43 Regular

    150
    May 21, 2013
    Wow, didn't expect to see that much of a difference at ISO 200! The 5D2 file definitely looks the best out of the 3 cameras.
     
  13. arad85

    arad85 Mu-43 Veteran

    477
    Aug 16, 2012
    Or as images (Don asled me to)
    GX7 & 14/1.8 @ f5.6:
    GX7_14-1.8.
    GX7 & 12-35 @ f5.6:
    GX7_12-35.
    Ricoh GR @ f/8:
    GR.
    5DII w/ 17-40 f/4 L lens @ f/11:
    5DMkII_17-40.
     
    • Like Like x 2
  14. The aliasing on the GR makes it appear to me to be over-sharpened compared to the GX7 and the 5DII. The 5DII is showing the additional magnification that an extra 5 or 6 megapixels gives. I know I see a small increase in absolute resolution when comparing my 20MP APS-C Samsung NX200 to my 16MP Olympus E-M5, taking into account pixel-level detail as well as the greater number of pixels overall.
     
    • Like Like x 1
  15. DHart

    DHart Mu-43 Hall of Famer

    Jan 7, 2010
    Scottsdale, Arizona
    Don
    Nic... I agree. Also, perhaps less sharpening (or different sharpening parameters) should be applied to the GR image than what the others received.
     
  16. peter124

    peter124 Mu-43 Regular

    Should the 14mm 1.8 labels actually be 14mm F2.5? Or is there a lens I don't know about?
     
    • Like Like x 1
  17. nickthetasmaniac

    nickthetasmaniac Mu-43 All-Pro

    Jan 11, 2011
    On my screen (3yo MacBook Pro) the GR shows by far the most detail, but as people have mentioned above it also displays the most artefacts and looks over-sharpened.

    I wonder how the various files would hold up to a bit of post sharpening in Lightroom?
     
    • Like Like x 1
  18. LovinTheEP2

    LovinTheEP2 Mu-43 Top Veteran

    619
    Feb 15, 2011
    Toronto
    BTW, the Canon has obvious camera shake; did you use a tripod??
     
    • Like Like x 1
  19. DHart

    DHart Mu-43 Hall of Famer

    Jan 7, 2010
    Scottsdale, Arizona
    Don
    Ooops... yes. Mislabeled. Wishful thinking on my part, perhaps! :smile: Corrected in first post.
     
  20. DHart

    DHart Mu-43 Hall of Famer

    Jan 7, 2010
    Scottsdale, Arizona
    Don
    Did I use a tripod? I sure hope so with a 3.2 sec. exposure!

    BUT...good catch. I did not use a remote release. I held the shutter button down for the long 3.2 sec. exposure which, undoubtedly, contributed some to the shake you see in the image. The jagged-edged paper attachment on the right side of the Correlejo Tequila bottle really shows the effect of the camera shake. The 5DMkII most certainly would have turned in a much crisper image, had I used a remote release. That w a s a l o n g exposure! :thumbup: Alas, my test is most certainly flawed. Nonetheless, when I am creating landscape images that I intend to print to large sizes (40'+) I'm going to do my best to see that I use the 5DMkII with 17-40 L lens for the job... just to get that extra edge of sharp detail when reproducing LARGE.