dxomark lens scores: Adapted Full Frame vs Native

Discussion in 'This or That? (MFT only)' started by Hookem3119, Jan 25, 2016.

  1. Hookem3119

    Hookem3119 New to Mu-43

    5
    Nov 4, 2013
    Houston
    Jeff
    I'm trying to decide whether to put money into the Kipon adapter that supports AF and a Canon EF tele, or the Oly 50-200mm and MC-14 TC.

    Q: Using Dxo's scoring system, the native mft teles generally score very low for sharpness compared to standard full frame lenses. How much of that is due to the sensor on the body the lens was tested on? Or, would a standard USM (non "L") 70-300mm score comparably to the Oly 75-300MM? They have not tested the Oly 50-200mm. Would it score comparably to a Canon 70-200mm L or 70-300mm L if it was tested on a full frame 24 mp body?
     
  2. tkbslc

    tkbslc Mu-43 Hall of Famer

    99% or greater. The maximum resolution score for their test is handicapped on m4/3 due to the smaller sensor and less megapixels.
     
    • Agree Agree x 1
    • Informative Informative x 1
  3. Hookem3119

    Hookem3119 New to Mu-43

    5
    Nov 4, 2013
    Houston
    Jeff
    Thanks. I was thinking that optically, the pro level Olys were roughly equal to the to the equivalent size and speed L series, and that the Canon consumer level 70-300 probably wouldn't resolve more detail than the native mft 75-300. I would just be sacrificing lens weight and focus speed. I do like the idea of all the glass that is available in Canon EF mounts though. But that is for another thread.
     
  4. pdk42

    pdk42 One of the "Eh?" team

    Jan 11, 2013
    Leamington Spa, UK
    DxO's test methodology is seriously biased against u43. Their "perceptual megapixel" measurements are guaranteed to make smaller sensors appear worse and of course that means that lens tests will be skewed too. Believe me, I've had Canon consumer zooms in the past and I can promise you that the 50-200 beats them hands down.

    What I found amusing was that they bent their own rules when they released their own camera so that the results looked good.
     
    • Like Like x 1
  5. Klorenzo

    Klorenzo Mu-43 All-Pro

    Mar 10, 2014
    Lorenzo
    You can verify how much the score is influenced by the sensor going for example here:

    Canon EF 135mm f/2L USM mounted on Canon EOS 20D : Tests and Reviews | DxOMark

    and compare the results with the Canon 350D (8MP) and the 5DS: 14 vs 39.

    BTW that same lens, fourth best Canon lens in their charts, on a 18MP camera (Canon 7D) scores 19 where the Olympus 75 on a 16MP sensor scores 27(!).
     
    • Useful Useful x 1
  6. pellicle

    pellicle Mu-43 All-Pro

    Feb 10, 2010
    Southport, OzTrailEYa
    pellicle
    Hi

    it has everything to do with the "physics facts of life" and the testing simply reveals that. The basic issue is magnification - real magnification.

    First a metaphor: if you sell pizza by the slice count not the diameter then an 8 slice pizza would be more pizza than a 6 slice pizza.

    pizza.


    The fact is that m43 is still a smaller sensor and thus to see the same image or the same feature size requires a larger magnification than if you were using (so called) "Full Frame" (which is actually tiddly compared to the giants in film like 8x10 inches).

    The facts of life are that everything has resolution limits, same reasons we can't see the lunar lander on the moon no matter how big the telescope because there are resolution limits of light. The more you magnify the image the more you'll approach that limit and see mush.

    So the more megapixels into a smaller sensor must meet a point of diminishing returns. Meaning that as you put more pixels onto the sensor you get closer to fully capturing every last possible detail from that lens, but its a diminishing return, in that you'll need a 40MPix sensor to see that the lens actually has a full resolution of perhaps 13PMPix (note the P for perceptual)

    Thus we see why there is a limit to the resolution of effective or "perceptual megapixels" you see when looking at DxO reviews.

    I have a few pages of mine I suggest you read:
    This page is quite old, but perhaps more technical ... it was written before m43 and back then I used to think poorly of 43 compared to APS:
    megapixel madness

    This page has some practical examples of lens testing in the real world and its effect on print (or of course screen) viewing:
    in my view ...: lens tests: looking for better pictures, what's missing?

    So to answer your question: what seems a simple question has a complex answer. In short if you want more perceptual megapixels (not just more mush) then you will need to spread that capture over a bigger area not a smaller one. So the m43 lenses will actually do better (due to being more recent designs with a higher "design bar" than the older lenses. M43 lenses are pretty darn good for what they are.

    Lastly have a peek at this page, it may help you understand "Perceptual Megapixels" better.
    in my view ...: a Perceptual Mega Pixel explained?

    Best Wishes

    PS: when I wrote that MegaPixel madness post back in 2006 I wrote this:
    I'd forgotten that wish , and I'm happy to say that I got it back in 2009 with my G1 and more so now with cameras like GM5 :)

    So it does have a happy ending, and there are cameras out there to suit all your needs (even if its not one camera size fits all).
     
    Last edited: Jan 25, 2016
  7. eteless

    eteless Mu-43 All-Pro

    Jun 20, 2014
    The native glass is generally all of modern design, sharp within reason and often smaller. The main appeal as I see it is the higher end telephoto options available (most longer primes are generally fantastic).

    I guess the gist is that you get what you pay for, high end lenses are probably going to be pretty good (you really have to go hunting for a bad lens these days*).

    *= The exception would be some of the absolutely horrible consumer kit lenses for APS-C, you don't have to hunt far to find a bad one.
     
    • Agree Agree x 1
  8. pdk42

    pdk42 One of the "Eh?" team

    Jan 11, 2013
    Leamington Spa, UK
    Thanks for that. I read your article on perceptual megapixels and found it very interesting. Your comment about how more than 16Mp on u43 is unlikely to give any benefit is intriguing in light of the GX8 and Pen F.
     
    Last edited: Jan 25, 2016
  9. pellicle

    pellicle Mu-43 All-Pro

    Feb 10, 2010
    Southport, OzTrailEYa
    pellicle
    Hi

    well DxO has as yet no lenses listed as being tested on the top m43 cameras

    Panasonic Lumix DMC-GX8 vs Olympus OM-D E-M5 Mark II vs Panasonic Lumix DMC-GH4 | DxOMark

    so it will be interesting to see ... by my expectation is that its like audio, if you go up in sampling rate, after a point you don't get any audible increase in fidelity (after you've avoided Nyquist driven aliasing).

    I guess that in reality I'm not saying you'll see NO improvement, but just no significant one, and certainly not the linear one that the pixels may imply.

    I see that comparing the lens results for the Oly 75mm on the GH3 and the EM-5 (which both have the same pixel counts) shows 11 vs 13 PMPix respectively. Will that translate to anything observable in even a huge print is a moot point (and in my view serves to highlight issues in the testing OR sensor coverings).

    PS what I'd be very interested to see is how well those lenses do on the sensor shift stuff which is used to give the 40Megapixels

    PPS:
    Image comparison: Digital Photography Review
     
    Last edited: Jan 25, 2016
  10. sriracha

    sriracha ballistic photons

    69
    Jan 22, 2011
    DxO's scoring system* provides directional guidance for the controlled results of a specific body + lens generated in a lab environment. the question they do not really answer is "how is this applicable and useful"

    allow me to provide some suggestions... after you score your camera + lens (or just camera, or just lens) use the following series of awesome formulas:
    • why you suck = DxO score 1
    • the cost per point you paid to suck = $body&lens1 / DxO score1
    • where awesome begins = DxO score 2
    • the cost per point you paid for awesomeness = (($body&lens2 - $body&lens1) + $resellbody&lens1) / (DxO score2 - DxO score1)
    joking aside OP... but only partially. what i listed above is the actual impact to a photographer that DxO skirts around. mathematically speaking, because DxO applies a log curve to their scores, at some point the cost increase far outstrips the increase in scores.

    let's bring it down to a business case. what is the value you are creating for $5,000 worth of Leica awesomeness forged in the fires of Mordor that gives you +5 DxO points?
    • a working photog rationale:
      • let's assume $ASP per print = $1,000 cause you're a badass
      • payback = ($5,000 / $1000)... boom! you paid the cost of the lens in 5 shots. but where's the fun in that? you just sacrificed 100 hobits to procure the glass from Mordor! you need to charge more so they did not perish in vain!
      • value added = ($5,000 / ???) here's the hard question, can you actually charge more? because if not, you essentially spent $5,000 procuring some pieces of glass and metal that gave you nothing you did not already have... except DxO points
    • an enthusiasts rationale:
      • F logic. i will sacrifice as many hobits as needed for +5 DxO points! my preciousssssss....
    anyway, i suggest browsing the forums to find a good example of the target shot you are looking for with the glass you are thinking of. then factor in the new body + glass and see if it is worth it to you... because ultimately that is what the score comes down to, personal preference

    *note: their scoring system does not take into impact on not using tripods, light variation throughout the day, post processing, de-noise, sharpening, blurring etc that people do which largely detracts from their scores...
     
    • Like Like x 1
  11. pellicle

    pellicle Mu-43 All-Pro

    Feb 10, 2010
    Southport, OzTrailEYa
    pellicle
    on blurring, focus errors are a factor. For instance I found that the Sigma 30mm f2.8 was unable to 'step' precisely enough when focused closely. Like the other lens testing concepts I assume (although I am not sure if its enumerated) that their method involves distance from the target as a function of focal length (like say 51x focal length). Its possible that the lens may not step to that exact spot (as my observation of the Sigma was) and so will produce a "result" which rates the lens lower.

    I'm also not sure if they average the results of subsequent tests or only test a lens once.

    Either way its a guide: akin to a tape measure printed on rubber ... it may stretch or shrink but its about right generally speaking
     
  12. PakkyT

    PakkyT Mu-43 Top Veteran

    767
    Jun 20, 2015
    New England
    Yes, but if you want to take some kick ass photos of charts, graphs, and checker boards DxO's scores can't be beat. ;)
     
    • Agree Agree x 1
  13. fredlong

    fredlong Just this guy...

    Apr 18, 2011
    Massachusetts USA
    Fred
    If you like the idea of all that Canon glass, then a Canon body would be the best solution.
    Fred
     
    • Agree Agree x 4
  14. Speedliner

    Speedliner Mu-43 All-Pro

    Mar 2, 2015
    Southern NJ, USA
    Rob
    DXO is useful for comparing within similar platforms. Not so useful for cross-platform comparisons.

    However, a poorly rated canon lens isn't going to suddenly excel on MFT. A quality highly ranked one will though. A quality, speed boosted FF lens does offer some advantages. That addition of light is useful, especially in MFT where more light is almost always helpful.
     
    • Agree Agree x 1
  15. LowriderS10

    LowriderS10 Monkey with a camera.

    May 19, 2013
    Canada
    Stick with native mount lenses...they're excellent.

    This is purely anecdotal (though I should have a thread somewhere on this forum with evidence to back it up), but when I switched from Canon (5D Mark II and L lenses) I kept my Canon gear, just in case it had better IQ than the Oly (so I was going to use the Oly for travelling and the Canon for "serious" work).

    Well, I ended up shooting the E-M5 + 9-18 (yes, the lowly 9-18) and the 5D Mark II with the venerable 16-35L II and guess what? The Oly outperformed the Canon more often than the Canon gave me better results (I'd say in about 75% of the time the Oly was better, in 25% the Canon). Then I bought the Oly 75 1.8 and shot it against my 5D Mark II and the 135L...the 135L was my favourite lens of all time...I loved it. But guess what...other than bokeh, the Oly 75 was right up there with the 135L in every way...in fact, I think the 75 was sharper wide open.

    So, I sold all my Canon gear and never looked back. :)
     
  16. Hookem3119

    Hookem3119 New to Mu-43

    5
    Nov 4, 2013
    Houston
    Jeff
    Thanks, this is exactly what I was looking for. I knew that the DXO scores were relative to the sensor, that's why I posted the question. What I did not know, was how to quantify that. I really did think about getting a used APSC sensor camera and the Bigma lens just to use for birds and wildlife, but decided that 50-200 would accomplish the same with the 2x TC, but still give me the option of premium glass for everyday shooting. I have the Oly lens and EC-14 on the way. I'll add the EC-20 when I stumble into a good deal. If anyone is interested, I'll be selling my 70-300 ED when the new one arrives.