Friday, November 30, 2012

Olympus OM-D E-M5 versus Panasonic GH3 versus Panasonic G5 at 3200 ISO – High ISO Shootout (test, review, whatever) – **Updated Again on 18 July 2013**

Hmmm … should this be called "The Defender versus the Contender versus the Pretender?” Smile

The OM-D is the darling of the m4/3 market, with a huge following of almost rabid fans. In fact, it is a gorgeous camera with a small yet very capable body. The image quality is stunning and the video is quite good (for an Olympus) … all in all, what is not to love? Well, for one thing the price … because the body is a bit hard to hold securely until you buy the grip and suddenly the body is costing you $1300 …

The GH3, on the other hand, comes with a wonderful dSLR like styling and built in grip and charges that price right out of the box. It also comes with world class video that harkens back to the golden GH2 era (which is still going on :-)

The G5 … what to make of this camera? It uses a variation of the GH2 sensor with the multi-aspect ration disabled. It is the older Panasonic 16mp sensor and thus well behind these two in development. Going by the number of people who toss off comment like “it is just better, I see it every day” and such, the OM-D should crush the older G5. The GH3 might actually be able to stay in the game because it is rumored to use the Sony sensor. I’m not sure that is true, but at the least it uses someone’s flagship sensor.

image I have borrowed the RAW images that Focus Numerique provide for testing and evaluation purposes.

The original images are copyright © Focus Numérique.

I am using them for educational purposes and obviously I therefore am making use of the fair use doctrine for copyright.

Click on the logo at left to go to their site

So … on with it. I’ve used the 3200 ISO original RAW images that were shot with the Olympus 12-50 kit lens on the OM-D, the Panasonic 14-140 all in one kit lens on the GH3 and the Panasonic 12-35 2.8 lens on the G5. All were shot at f/5.6, which should more or less equalize the clarity and sharpness we see. At this focal length, all the lenses are stopped down at least a bit.

I find, though, that the testers were not all that scrupulous about getting identical framing and magnification. Slight differences can make unsubtle differences in perceived sharpness, but this is one of the better test images I have seen, so off we go.

I have processed all three images in Lightroom 4.3 Beta because this version supports the GH3. I have sent the processed TIFFs to CS5 for cropping and crafting of the 100% crop juxtapositions for 5 portions of the scene. I did my best to equalize tone, saturation, sharpness and noise. With these fairly close, we can compare details, edges, dimensionality … all those indefinable things that go to make an image really pop.


But first l think it instructive to see how different they are at typical 1000px web sizes. This is how the vast majority of images are present every day, and it is instructive to see how much difference we would see on the web or on typically small prints, say up to 8x10. This represents the vast majority of usage.

Remember to click on these images to see the full 1000px presentation.

Panasonic GH3 (contender)

Olympus OM-D E-M5 (Defender)

Panasonic G5 (Pretender)

If you are using an IPS panel that is hardware calibrated then you will likely notice that these are so similar at web sizes as to be virtually indistinguishable. I’ve been saying this about the 16mp sensors for some time now. In RAW, the differences are marginal at best. Your lens and your technique will absolutely swamp the sensor differences at high ISO.

It has been asserted to me many times (without supporting backup) that there is a 1 stop+ difference from the G5 to the OM-D at higher ISOs, but for the life of me I cannot find it in the test images here, nor could I find it when I processed test images from imaging-resource or dpreview.com … I tend to get the same excellent results from all the 16mp sensors at these higher ISOs. So I am forced to take all such assertions with a grain of salt. (Truckload would be more accurate.)


And now, the meat of the comparison. We’ll compare 100% crops to see how far apart they really are. This comparison mimics looking at portions of 46” prints from about 20 inches. That is a little insane, yet it is what we pixel peepers do every day, and it is not a bad indicator of what to expect at poster sizes.

Crop 1

The first crop is the upper left corner of the image, specifically the map and a little of the overlapping fan of paint chips plus some of the feather. Now, at 3200 ISO, it should be a bit of a struggle to resolve names from that distance and feather details should be hard to come by.

My analysis says that the GH3 is the detail winner hands down. The lens is obviously very sharp in the upper left corner, while the other two are not as good. In fact, the GH3 has excellent feather detail as well. I suppose that we will have to chalk this up as a lens difference.

What about tonality? Well, try as I might, the OM-D resisted my attempts to soften its contrast and maintain nice edges and noise, so it has stronger shadows that are readily apparent in the paint chips at the right. The Panasonic images have very similar tones, which is strange since the G5 sensor is really old.

I think this is a win for the GH3 hands down, with the G5 coming in an easy second. Again, this might be purely down to the lenses. The OM-D’s kit lens is pretty embarrassing, especially when compared with the 14-140, which sports a 10x range. I have this lens and it is delicious :-)

Crop 2

This crop is just about dead center in this image, so there are no “it’s the lens” excuses here.

Hmmm … my analysis says that these could have come from the same camera and lens. No surprise to me but I will bet a few people are wondering what trickery I am up to. Well, I have been sure for a while that these sensors are very similar in RAW and they all compete quite easily with low end dSLRs.

So, it is a 3-way tie.

Crop 3

Top middle of the image. We’re well away from the corners, so the lenses should not be much of a factor. On the other hand, we have really rich colors, so there is an opportunity for a real difference to show up.

Hmmm … the colors seem very similar on the GH3 and OM-D … makes me wonder of they really do share a sensor. The G5 coloring is clearly slightly different, which makes some sense if the sensor is two years older.

ALERT! I was called onto the carpet by a pedant with certifiable perfect color vision on the previous paragraph. Apparently, my statement is utterly false because the hues are so different between the OMD and the GH3. Apparently, this is offensive to some (presumably the well organized union of pedants with perfect color vision.) The “1000” on the right is blue on the OMD image and some unspecified completely different hues on the two Panasonic images. So I examined them at the pixel level in Photoshop and I agree. The red and green channels on the OMD are in good balance, leaving the numbers looking basically blue. The red channel is stronger on the GH3 and even stronger on the G5 with a weaker blue channel, causing each to lean into the world of magenta, with the G5 going a bit further. This also seems to make the yellow look brighter in the OMD image, although that may be an overly bold statement that will draw further retribution. Anyway, keep that in mind when you process these opinions, in case it matters to you too. For those with great pride in their perfect color vision, please accept my apologies.

Yet I prefer the G5 color and tone, at least on the head gear at right. The Olympus continues to show small but visible edge artifacts. I notice them on most images, but here they make a difference in the left portion of the image. Not sure if this is a demosaicing issue or not.

I think I give this one to the G5 by a hair, with the GH3 next and the OM-D bringing up the rear. ALERT! This is a personal judgment and yours may very well be ordered oppositely.

crop 4

Bottom of the image in shadows. This is where the OM-D should really assert itself.

Well, the OM-D does not disappoint. Its lens is clearly the sharpest of the three at the bottom of the frame. Its tendency to stronger contrast helps it a lot here, although I am also noticing some posterization as the transition to black is a little harsh and sudden. The others are of course somewhat exposed to noise pollution from more open shadows, yet I don’t really see extra noise.

Interestingly, the little flower illustration at left shows up best and cleanest on the G5 again. It really handles edges beautifully. But no matter …

The OM-D wins for its rendering of the car. But it is definitely equivocal. The GH3 and G5 are in a tie for second. This is a contrasty scene that looks much better on this OM-D rendering if you ignore the deep shadow blocking a bit. Cranking the contrast a bit on the others would help them out here, and that can be done without blocking the shadows. This one is pretty complex to judge.

crop 5

Just to the left of crop 4 is an opportunity to show the white bear against the black wall and some writing on the right of the crop.

The G5 shows the writing best, the OM-D shows slightly more detail on the bear’s fur, and the GH3 is more or less similar to both. The OM-D again tries to block the shadows a bit. I would be processing the OM-D on a fairly soft tone curve at all times to prevent this.

There is not much to choose here overall, so I give the G5 and OM-D first place with the GH3 a hair behind.


Conclusion

I had to really stretch to find the differences in some of those I mages. I would say that all three are close enough that you should not be freaking out over the differences at this ISO. You should spend far more time improving your technique and processing skills rather than fretting over perceived differences between these sensors.

In fact, I believe in choosing based on how the camera shoots and what it feels like in the hand. Thus, I am a Nikon and Panasonic fan. But YMMV of course.