Ok, I finally got a chance to test the E-M5 against my GH2 in dynamic range. I chose a high DR area, and then decided to see how they did when overexposed 3 stops and pulled back, and underexposed 3 stops and pushed back. I metered until the clouds just barely started to clip in JPEG. The clouds were slightly different between the two cameras which led to the same shutter speeds working despite the 1/3 stop difference in ISO, but final images were right on in exposure (identical mid tone values for the 0EV EC shot). I then increased exposure 3 stops and then decreased exposure three stops. Did the same for each camera. In Lightroom 4.1 RC2, I then pulled back the highlights until no more detail could be recovered, and with Shadows, I increased black point until it didn't clip and used +100 on the shadows slider for each file. This is the overall scene (this is the E-M5 mid-shot): Here's is what the GH2 could recover in the highlights (three stop overexposure, pulled in Lightroom 4.1): And here's what the E-M5 could recover in the highlights. The clouds are still too far gone to recover, but they were well over 2.5 stops blown, so it's not surprising...but the E-M5 kept all the detail in the buildings as well as the water, while the GH2 lost all of that. Looking at 100% crops of the building area, you can see that not only did the GH2 ultimately lose the detail, but that it also lost all color data for some of the detail it could recover. The E-M5, on the otherhand, perfectly recovered everything but the clouds. GH2: E-M5: Next, I took a look at the shadows. There wasn't a lot of extreme shadowing in the frame, but the bridge above, of course was the place to look. This much shadow recovery is equivalent to a 3 stop push. There is a lot of noise for both cameras in the pushed shadows, that's for sure, but the E-M5 again retains more detail, has better noise control, and retains the rust color of the bridge steel. The GH2 loses the color information and a lot of the darkest detail. Upper Left Crops: GH2: E-M5: So, this definitely confirms what I've been seeing while I shoot regularly...the E-M5 has significantly better dynamic range than the GH2 (or any other m4/3 cam I've ever used) in both the shadows and the highlights (though the difference in everyday use is bigger in the highlights). In my experience with the E-M5 so far, it's been very difficult to clip highlights at all...you really have to almost try to do it.
But the lens on the GH2 has better corners (looking at the rivets, the seem sharper and less elongated on the GH2 shot than the OMD. Just curious -- the same lens on both, or which different lenses?) Thanks for the test. I'm an Oly user, but it'll be fun to see how the GH3 improves!
Same lens on both (Oly 12), but I focused about halfway into the shot to maximize DOF...the differences here are almost certainly due to a slight difference in focus point (which would affect the bridge area the most, since it was by far the closest thing).
It's very tough to do DR testing right. This is one of the best that I've seen . The only technical issue I can see that could have skewed results are that changing lighting conditions could have affected the DR of the scene, which is why I always try to do this sort of test without a cloud in the sky. Nevertheless, it seems a very solid test. Each of these was shot at the camera's respective nominal base ISO (ISO 160 vs ISO 200, correct? I'm very surprised at how much better the E-M5 fared.
Maybe I'm slow, but just to confirm this is shot in jpeg? Would it make any difference to see the RAW files pushed?
These were shot at base ISO, and they were shot in RAW and developed in Lightroom 4.1 RC2...you can't recover detail in JPEG.
Thank you. Yes, I'd have preferred more consistent lighting, but the lighting did not change much at all...the cloud cover was relatively consistent, and there wasn't much time between shots. As I said, the final exposures were identical (within 1% value in Lightroom for the same spot)
Nice test and write up Jordan. Thank you! But all these tests you keep writing about in relation to the EM5 are just killing me!!!
Cool! Thanks for the real-world tests. I love my E-M5, but I'm really looking forward to what the GH3 will bring to the table.
Very nice test I was thinking that GH2 and E-M5 are more comparable but this shows that Oly is truly better. Curious if the matrix in OMD is the same as in G3 but better programmed and used by Olympus core.
After using the GX1 for the past several months (which is the same sensor as the G3), and the OM-D, I can't imagine that they're the same sensor. They have very similar characteristics in some sense, but in others, they are very different. The files out of the E-M5 are so much richer than the GX1. I'm not saying the GX1 is a bad camera, or that it's incapable of quality output...it most certainly makes some very good images. But the E-M5 has a depth to the images that the other m4/3 have been lacking. The DR is so improved, not just in recoverable highlights. The files feel to me much like I felt when I first got my Canon 1Ds Mark II. There's just a depth to the images I haven't seen on a smaller sensor. It really does seem to compare very well against current APS-C sensors, and that's saying something with the recent APS-C releases.
Cool stuff. That seems like a simple and unambiguous test, which is also easy to relate to for anyone who's ever grabbed a Lightroom slider. As a G3 shooter, I'm also jealous!... but I'll live. Mostly, I'm heartened that Oly significantly pushed the state of the art in 4/3 sensors... I love it! Your move, Panasonic!
I gotta agree with you on this. Feels like my 40D raws, there is so much more headroom than with my GF1.
I think it is still most definitely possible that they are the same sensor. It reminds me of the Nikon d7000 vs my Pentax K-5 - most definitely the same sensor, and yet Pentax does a noticeably better job of bringing the most from it (dynamic range, high iso, noise rendering). Considering Oly's performance with jpeg files and Panasonic's lack thereof, it wouldn't be surprising to me to see Olympus get that much better performance from the same hardware. Also, an excellent comparison, Dynamic range is such an important sensor aspect, and to see the e-m5 perform this well is great.
Thanks for writing up your test, Jordan! Your findings mirror what I've seen in my very limited shooting with my EM5 thus far. Not only does the EM5 have a good amount more DR than other M43 cameras I've used, but it also does a much better job on highlight recovery with LR. The thing that always bothered me about all the 4/3 and M43 cameras I've used in the past is that you had to be REALLY careful not to blow the highlights, because you got all kinds of weird color shifts, if not outright clipping in one channel before the others with even half a stop highlight recovery. This is in contrast to my Canon DSLRs where I could often pull back a full stop with no ill effects, and even more if I worked at it a bit. A great part of your image to highlight what I'm on about is the short building that's obscured by the trees. In the GH2 shot you can see that some of the windows have a color shift. This is very unnatural-looking. The EM5 renders a much more natural, flat look in the windows. You can see this in the other buildings, too. The tones are more natural and filmic. I've had this problem with all my 4/3 and M43 cameras until the EM5. I've worked around it by underexposing slightly, but this tends to increase shadow noise when you're shooting mid ISOs. In my brief time with the EM5 thus far I've been really happy to see that I can shoot the EM5 without worry of color shifts in the near-clipped highlights, much like with my 5D2 and 7D. I think everyone who has shot with 4/3 and M43 in recent years realizes that their sensors don't quite measure up to a benchmark like the 5D1 or 1Ds2, but someday would. As a long-time 5D1 shooter I think the EM5 finally matches it. Base ISO noise is still higher than the 5D1 (Canon is the king of smooth low ISO, and the 5D1 is the poster child for this), but DR, high ISO noise, and color fidelity are a match IMHO. For me, that's as good as I need. Any future development is gravy. The next M43 camera I buy won't be for sensor developments, but for usability developments (faster C-AF, shorter viewfinder blackout, better buffering algorithms).
It may not be a huge difference, but it's a difference in a very important area, and an area where 43/M43 has historically lagged the competition. Every little bit of water is precious to a man lost in the desert, ya know?
