Thursday, April 25, 2013

DXOMark – What is the practical value of their sensor numbers?

This topic is if interest to me because there is so much partisanship going on betwene “camps” in various forums. The Canon – Nikon wars have raged for years, fueled by their 1:2 positions that actually swap around now and again. DXOMark came along and gave us all “hard” numbers of relative sensor quality to debate over ad nauseum.

I was always a Nikon fan, well before their adoption of the Sony sensors and of course their development of various hybrid versions of these sensors etc. I imagine that the Canon shooters were annoyed when Nikon shooters crowed about their total dominance of the Canon cameras in the DXOMark rankings.

But was there a lot of practical difference in day to day shooting? I really doubt it. Owners of the 500D had a 10 point deficit against the D5000 to deal with, and that is a pretty big number on this scale since the Nikon is only around 72.


So you would wonder then if the Canon fellows all defected to Nikon to get this great new sensor? The answer, obviously, is no. If their shooting was crippled by comparison, then one presumes that they would move in droves. But Canon sold a bucketful of these cameras because it didn’t matter enough to force the migration.

I shot the D70s for years. And that is a further 10 points behind the 500D. Obviously, a terrible camera, right? Well, no. The D70s was magnificent, and I still miss it. I was shooting it right up until I sold it off to finance the D700. The D700 is gone long ago (loved it) and I have moved on, but I remember all of these cameras very fondly. I do not ever remember pining for a camera with better dynamic range because I found it perfectly adequate, despite its being miles and miles behind the newer cameras.


This has all become more interesting to me as I moved into the m4/3 arena. As with the Canon and Nikon war, there is a similar dichotomy in the m4/3 community. Panasonic makes bodies of really varying styles that ignore fierce debates and lots of gnashing of teeth. Olympus cameras have a Sony sensor in them after suffering with older 12mp Panasonic sensors (still great sensors for their time by the way) while Panasonic was shipping the flagship GH2 with the first of the 16mp sensor generation.

That sensor has been surpassed by several new Panasonic sensors and by a Sony sensor, which Olympus sourced to create the wildly popular OM-D E-M5. This camera is widely acknowledged to stomp the Panasonics by the Oly faithful, but when you ask why you get one of two answers:

  • DXOMark score is much higher
  • I can see so much more “depth” in my images, trust me Winking smile

Well, the first point is always true. The numbers are out there … for example, this is how my GX1 and G5 compare against the EM-5 …


Now that is a heck of a beating. The OM-D has matched the D5000 (a higher sensor class, cool) … and the G5 has matched the 500D (also APS-C, cool) … and the GX1 is better by 5 points than the D70s (wow … who’d a thunk?)

So … we have a camera that is as good as an APS-C camera from 3 generations ago and another camera that has the same qualifications, and a third one that is better than an older model from this higher class. That tells us that the m4/3 sensors have come a long way. It also tells us that the same dynamic is at play with Olysonic as was at play with Canikon. Which is to say that people like me go for the best ergonomics (Nikon / Panasonic) and others go for the best sensors (Nikon / Olympus) …

Note that Canon came second for many of us in both those categories. But remember that I bought into Nikon when the sensor wars went to Canon. I did that because the ergonomics easily trumped any perceived advantages of the sensor at the time.

With Panasonic, not only do I prefer the ergonomics (by a country mile), I also prefer the cost of bodies, the video abilities (not even a contest), and in the end I prefer the look. And to be honest, I just cannot put a finger on how much better the sensor is on the E-M5.

The numbers are clear enough … or are they?


Huh? Where did that pounding go? Well, despite the clinginess of the graph lines to one another (i.e. a cursory glance says “nothing to see here,”) in fact the numbers are plotted against “measured ISO”, which for the Oly is much lower than it is for Panasonic. Which basically means that the 200 ISO point fo rthe Oly is way over there on the 100 ISO line.

Getting your head around this is no easy trick. DXO themselves talk about the “fairness” of looking at “measured” ISO for comparison purposes instead of “camera ISO”, which is what determines your ultimate shutter speed. This is very confusing, since camera ISO is how we make the capture, so we really want to know the difference in the quality of the result based on what we need to use to get a certain image (which means shutter speed really matters when you cannot just pop the camera on a tripod and be done with it.)

If we shifted the line for the OM-D to plot it against camera ISO, the line would be well above the Panny bodies, which would accurately describe what the photographer experiences. So why plot it this way? There can be only two answers:

  • This lack of any real difference is meaningful, there is no practical difference
  • Brain fart

The Oly faithful are always spouting these numbers (the raw numbers, not the graphs, obviously) in their defense of the blanket assertion as to the camera’s incredible prowess. And we get statements like that second point way above where the fellow swore up and down that his images were just so obviously better with so much more depth and such. When I challenged him to post some images that illustrate these obvious differences, he dodged repeatedly. The images never appeared. I have never seen someone demonstrate the practical advantages that these numbers give you.

Now, I have seen some truly excellent high ISO images form the Oly, so I know that it has the chops. It also has a really nice (if somewhat over-processed) JPEG output. But no one has shown the head to head differences that really make it shine. When I process images from DPReview and Imaging Resource, I simply cannot find this dominating difference.

So I am at a loss why I cannot find the difference when I process images from these cameras, and why the graph appears to confirm this by plotting against measured ISO and calling that fair. Fair to whom? The Oly’s alleged prowess is not apparent from the graph I clipped above. DXOMark themselves never address this issue, preferring to explain ad nauseum what the number mean in the broad sense. The faithful are sure there is great meaning attached yet never post anything to show us what the difference means in practical use.

So I searched this evening for some answers, after reading the articles discussing the sensor measurement issues on the DXOMark site. And this is what I found …


A year ago someone asked this very question. A day later he posted “anyone?” And that same day the visible post asked if there were any answers. I would think that the lack of answers was a little obvious.But my point is that you just cannot find any useful info to explain the strange break between the reality of the practical use of these cameras and their rather wide gaps in the DXOMark numbers.

How can Canon sell even one camera with Nikon sensors stomping them like that? How can Panasonic sell even one camera when Olympus has three cameras now with that “magical” sensor that stomps all the Panasonic cameras now? The answer is that it matters far less than the faithful (fab boys, whatever) would have you believe.

Anyway … I just thought I would throw this out there. Maybe someone will know of a really good explanation out there somewhere. Or maybe some really good practical samples to show these differences in real world use.