DxOMark, a "comprehensive RAW-based image quality measurement database and a set of scores to evaluate and compare digital cameras and lenses," recently evaluated RED's new 6K DRAGON sensor since it has the ability to pull extremely high-quality still images from video. The result, which probably comes as no surprise to RED themselves, measured the DRAGON as the best overall sensor they've ever tested, breaking 100 points on their scale. Considering that DxOMark has tested all sorts of cameras (some even more expensive than DRAGON like the medium format Phase One IQ180), the results are indeed impressive.
Here is the DRAGON compared to their previous champ, the D800:
It's interesting that dynamic range looks to be around 15 stops, which is pretty much what I thought their dynamic range chart looked like when it was posted last year (no HDRx was used for their testing). I would assume they were using a camera without the new OLPF in front of the sensor, but I'm not sure how much of a difference this would make based on the way they test cameras. The tricky thing about measuring dynamic range is that it can be subjective about what parts of the image are still usable, not just visible. A clipped highlight is a clipped highlight, but each of these cameras can have details in shadows that may or may not be usable depending on how the noise looks.
Speaking of noise, they also had some interesting things to say about that in their detailed analysis:
As the sensor and image processor can deliver very high frame rates, the Epic Dragon is certainly adopting multiple sampling techniques to reduce noise levels (also known as temporal noise reduction). Without such processing, such high SNR would only be possible from a sensor with an exceptional Full Well Capacity. Performances like this seem, to us, above the current technical capabilities of CMOS sensors.
As a side note, it’s interesting to speculate whether rivals such as Nikon, Canon or Sony are already adopting such techniques during video capture or in jpeg. But, this is the first time we can assess this type of performance on still RAW.
And here is some information about the actual testing procedure (emphasis mine):
For the tests, the camera was set to full 6k resolution at 23.98 fps, 16 bit and RC 5:1 compression.
A shutter angle (exposure time) equivalent to 1/50th sec was used. As the camera was supplied with an EF mount, we adopted aCanon EF 24-70mm f2.8 L USM during the tests (though the procedures are not reliant on the lens used). It’s also worth noting that the Dragon’s native sensitivity is ISO 250 and measured at 104, while all other sensitivities were obtained by digital gain.
The camera records REDCODE RAW that must be processed though REDCINE-X, a free application available on the RED website. Currently, RED does not give direct access to the bayer pattern data, so we cannot clamp this sensor analysis and score in our official DxO Mark ranking.
First, with regards to the native ISO, this is something that was stated a while ago when RED visited the ASC. The fact that it can be pushed to such high ISOs is related to how clean the sensor is. 250 may be the base ISO, but the ideal balance between noise and dynamic range, just like with the MX sensor, is quite a bit higher than that. In regards to the sensor analysis, while the results are impressive, since RED doesn't allow you to see the sensor data directly, this result is not actually going to appear in their official rankings. It's unclear if the results would be any different if they could directly access the Bayer data.
It's important to understand that the evaluation is strictly about still images, not necessarily how a camera reacts over an entire scene -- which means you can get away with slightly higher ISOs. They also haven't tested cameras like the Sony F65, F55, Canon C500, and ARRI ALEXA, which are all capable of shooting RAW, and except for the ALEXA, can all shoot 4K. It would be interesting to see the results of the DRAGON compared to these other cameras, especially since 4K is enough to get good-looking still images.
It's also worth mentioning that specs alone don't tell you everything, and final image quality is dependent on lots of different factors.
For more details on the entire test, head on over to the DxOMark website.