Petavoxel recently bemoaned the fact that the majority of sensors in recent micro four thirds EVIL cameras (or MILCs, if you prefer) perform poorly compared to contemporary APS-C sized sensors in digital SLRs. The only exception was the Panasonic GH1, which put up quite a respectable showing compared to its μ4/3 stablemates.
Yesterday dpreview posted their review of the E-PL1, and they were very happy with its high ISO performance. I quote:
Overall, the E-PL1′s images are the most natural and convincing here – avoiding the D3000′s overly contrasty, noisier images … Most impressive is the E-PL1′s ability to produce results comparable with the EOS 500D and Pentax K-x, despite its smaller sensor.
But what does DXO Labs have to say? They disagree, showing that the big three leave the E-PL1 gasping for photons with a lowly 487 points in the low-light ISO stakes.
Hence, the paradox.
How can this be? Is either DXO Labs or dpreview writing nonsense? The keen observer will notice that there are two significant differences in the way these two respected websites measure image quality:
- DXO Labs is very mathematical in their analyses, and eventually reduce all results to numbers, where dpreview show the images taken of their test scene, and then draw subjective conclusions.
- DXO Labs look only at the raw sensor performance – this means using raw file output and standardised conversion software. Dpreview looks only at JPEGs when analysing high ISO performance.
So what does this mean? Did Olympus stumble upon a remarkable JPEG engine which extracts superior image quality from an inherently inferior sensor? Or do numbers only tell part of the tale? And which camera then takes “better” photographs?
As you saw above, dpreview is quite confident that the Olympus outperfoms the ageing baby Nikon. Looking at JPEG performance only, the E-PL1 blows the D3000 out of the water, but concerning RAW performance it’s a very close call. Yet, even in their RAW comparison, DPReview goes so far as to say “the Olympus is clearly still capturing the better representation of the scene.” I trust the objective, repeatable and well-defined measurement protocols of DXO Labs, but I certainly don’t see the 15% advantage the D3000 is supposed to have over the E-PL1. Looking at these samples, you see how difficult it can be to see these differences in the “real world”.
Which camera will delight you with the best pictures depends far more on the quality of your lenses, the speed and accuracy of its auto-focus, and whether the image stabilisation (if available) works well. Is it important that your $500 large-sensor camera is small? Get the E-PL1. Still not small enough? Get the new Sony NEX-3. Should it be ergonomic? Then go for the Nikon. Or do you want the best low light capability and have a bit more to spend? Then ditch both and buy a Canon 550D. They can all take great photographs. And there is no such thing as the “best” camera – it all depends on what you need.
Read Ken Rockwell’s (hypocritical) view on the “Equipment Measurbator” for a good laugh about getting blinded by the numbers. In our case the number’s don’t lie, but they can distract you from focussing on the important stuff, which almost always boils down to usability for your purposes.