English - Espaņol
Sharpness - Bokeh
Foto tutorial (English)
Foto tutorial (Espaņol)
Equipment recommendations US-ES
Flaat for Canon
Flaat for Nikon
Flaat for the BMC
Flaat for NEX-5N
Old Picture Style Tests
550D video lineskip
APS-C vs Full Frame
Badly assembled lenses and image quality
Lens mount compatibility chart
ISO on different cameras
High ISO on the 5D3
DIY: DR test chart
RGBWK Bayer sensors
Notes on DoF-FoV
Notes on crop-DoF-FoV
Custom Cropmarks for Magic Lantern on the Canon 550D
How many megapixels do I want?
How many megapixels can I see?
Quick Monitor Calibration Chart
After the results for sharpness, now it is time for some dynamic range tests. As before, cameras considered:
* Sony a7s (XAVC-S)
* Panasonic GH4 (4K internal)
* Blackmagic Cinema Camera (RAW)
* Blackmagic Pocket Cinema Camera (Prores and RAW)
* Canon 5D Mark III (H.264 and RAW)
* Sony NEX-5N (H.264)
I tried to use similar lenses on all of them:
* Leitz Elmarit-R 90mm f/2.8 on the a7s, the 5D3, and the 5N+speedbooster
* Leitz Summilux-R 50mm f/1.4 on the pocket with 0.58x speedbooster
* Leitz Elmarit-R 35mm f/2.8 on the BMC
* Panasonic 12-35mm f/2.8 on the GH4 (didn't have a m43-to-leicaR adapter)
I moved the tripod a bit to match the framing in all cameras.
The test target (which you can see above) was the same as the one in my sharpness test, but this time I moved the tungsten fresnel a lot closer, to create a very hot spot on the left side, as you can see above.
All cameras were tested at their optimal ISO: 100 on the 5D3 and the 5N, 200 on the GH4, 800 on the BlackMagic cameras, and 3200 on the a7s (3200). Aperture was set to match exposure in all cases, which means it ranged from f/2.8 in the 5D3 and f/4 on the GH4 to f/11 on the a7s (I always overexpose by one stop on this one) and f/13 on the pocket.
We used Flaat settings on the a7s, the 5N and the blackmagics, CineLike D with contrast=-5 and everything else tuned down on the GH4. No idea which settings we used on the 5D3+H.264, I forgot to check that one (I was more interested in the RAW results, for which I knew standard Adobe settings with no sharpness and contrast dialed down would work great).
I shot that target with all available shutter speeds from 1/25s to 1/200s (from there onwards the fresnel started to create flicker). All footage was brought into AE (using Flaat or standard settings and 2010 process for the RAW clips), then I looked at the last shutter speed that met the following criteria:
* Useful shadows: even in the shadow area, I can see the separation between the black Prada box and the brown table it's on, and the shadow area of that black Prada box doesn't show too much noise (for me).
* Useful highlights: detail in the white rolled paper on the left and color in all the skin tone patches.
In previous tests I also required that shadows should show fine textures. Not this time. I intended to do it, but the box was too far from the camera this time and several cameras showed no texture even on the well-lit areas. Yes, that's how smart I am.
Regarding the 5D3+RAW, it does have about a stop more headroom than what I show in that chart, but it creates banding in some areas, plus that extra stop is completely color-less (look at the color chart above: the top 2 patches in the skin-tone column have gone grey; same for the top 3 patches of the pastel column on the left). Resolve may not have this problem, but I don't know how to use it, and this is the best I could get on AE.
As expected, the GH4 is lagging in terms of DR... unless you are ready to go overboard with noise reduction in post. As is, the shadows are extremely noisy. It's a very fine noise, without the huge color splotches of older DSLRs, but I wouldn't consider it usable. But, since the 4K image from the GH4 is so incredibly sharp, you can go crazy with the NR and still have a very sharp image. For this test, I applied MB Denoiser II at maximum strength (300%). In any case, beware that even with Cinelike -5 the image is pretty contrasty compared with what you can get in other cameras. Also, NR at 4K is VERY slow, it probably means you want to do transcode+NR overnight to ProRes or DNxHD.
Finally, my a7s: it is close to the top of the chart here, and that's after I substracted a third of a stop from the measurement in the highlights, because color gets funky when it is close to clipping, something that my test chart wouldn't show because it doesn't have the appropriate bright color gradients. But I tested this on the side looking at how bright areas in skin go from looking nice to clipping in an ugly fashion. My conclusion: as long as I'm 1/3 stop below clipping, everything is fine. The downside here is that, whereas rolloff in the a7s is nice for desaturated areas, it is pretty ugly for anything that has color, from skin to sky to basically everything out there. So, lots of DR, but make sure you don't clip anything that is not white.
If you want to see what these differences mean, click on the links below to see a frame grab from each camera. They were all taken with exactly the same exposure (1/50s in all cases, ISO at its optimal setting, and aperture set accordingly as explained above). The only processing in these images is that I have adjusted levels to match 16-235 on all cameras, and I have applied a ton of noise reduction to one of the GH4 samples.
5D3 - 5D3+RAW - 5n - a7s - BMC+RAW - GH4 - GH4+NR - Pocket+ProRes - Pocket+RAW