Tuesday, March 27, 2012

Olympus PEN Image Quality Test

Although all the Olympus PEN cameras use the same 12.3 megapixel sensor, incremental improvements to image quality have been made through the models over the last three years. In particular, a weaker anti-aliasing filter was introduced with the third model, the E-PL1, and maintained through subsequent iterations (E-PL2, E-PL3, E-PM1, E-P3). Here I will test the E-PL2 (released January 2011) against the E-P2 (released November 2009) in order to see if there is a discernible improvement in IQ.

I will provide the usual disclaimer that I am no equipment tester; I have no special gear, charts, calibration equipment, etc. Most of the time I simply take "normal" photos and see how they turn out. Every so often I resort to shots under more controlled situations, and that's what I will share with you today. Likely others have performed this same exercise before, but I like to prove things to myself!

I also wish to emphasise that small changes in image quality will not make you (or me!) a better photographer. This test is more or less a matter of curiosity. That said, I will now succumb to the temptation of pixel peeping.

E-PL2 full image at ISO200
Here's the boring shot I took of our DVD shelf, unremarkable unless you have a perverse interest in our viewing habits. I used a tripod and timed release to ensure that camera movement was not a factor. Light was from a window to the left, as you can tell from the highlights. I positioned a black and white photo so I had an area of greater detail near the centre of the image.

The lens is the universally praised Panasonic Lumix G Vario 20mm f/1.7 ASPH, set to f/2.8, its optimal aperture. I used auto-focus, which is very accurate in ample light (though rather slow). I had all special settings on the camera, including noise reduction, etc. set off. The image setting was "natural" and exposure mode Aperture-priority. All this shouldn't matter, since I shot RAW and cooked the files with all enhancements off, using Adobe Camera RAW. I note that real-world shots would look better than these examples, since some sharpening of RAW images is advantageous, and is indeed a part of my standard workflow.

I kept the camera white balance setting on automatic and was intrigued to notice a distinct difference in the colour balance of the results. Since the area of interest is a black and white photo, I reduced the files to monochrome. After all, this is not a test of colour reproduction, which I find to be completely usable and mostly accurate, small discrepancies notwithstanding.

This done, I took a shot at each whole stop ISO setting, doing the same for each camera. The following image shows a test strip of 100% crops. Please click through to Flickr in order to access the full-sized images.

E-P2 versus E-PL2 100% crops

The result? The E-PL2 is indeed sharper, as if a slight haze had been removed from the E-P2 images. And while more detail generally shows up more noise in the bargain, the E-PL2 shots appear to have less obtrusive noise, without any obvious "smoothing". I very much dislike in-camera noise reduction that produces artefacts and which I cannot turn off. That's one reason I like Pentax processing. And neither do I see anything in this Olympus to dislike.

I notice that the E-PL2 seems to be exposing slightly brighter than its counterpart. In this case at least, that seems to be more accurate, and might minimise shadow noise.

Finally, I will go back to my comment about white balance and note that the E-PL2 was also the most accurate in that regard. The E-P2 had a blue tinge.

This seems to be a four-way win for the later camera. Although each factor is small enough, if you are a stickler for getting the utmost from your camera you should scratch the E-P1 and E-P2 off your list. I am pleased to see that incremental improvements do sneak out the door with models that are otherwise largely driven by the need to market a new product or three every retail cycle.

However, back to reality. For most casual shooters these tiny improvements are irrelevant. I would instead recommend purchasing a model based on the criteria in my previous article.


No comments:

Post a Comment