Exposing the RED MX
EXPOSING THE RED MX -- AN UPGRADED AND AN UPDATE OF BEST EXPOSURE PRACTICES
It's been a while since this initial article was written. Indeed, this article covered strategies when RED had released Build 16, then a breakthrough build for the startup camera company that burst onto the scene, and have now delivered over 7000 cameras to customers worldwide.
We've seen, build 17, 18, 19, 20, 21, and most recently a new breakthrough with build 30, which works on the original RED One and an "upgraded" version with the new Mysterium-X sensor -- which demonstrates a lower noise floor and more dynamic range.
As I read through this article, I realized that it was due for a much needed update, so I wanted to share some new strategies for exposing the RED camera. First, you have to understand a bit about the RED philosophy. This is a company that continuously works to improve its products and has offered these upgrades to its customers through a series of free software upgrades.
The other thing to understand is the nature of capturing RAW data, the core of the RED system. RED has chosen to store its sensor data in a compressed file format known as .R3D. By storing data this way, it means that footage you shoot today can benefit from future tools. This means that footage captured on RED can -- and does -- look better a year, or two, or five down the road.
I caught up with Ted Schilowitz, (RED's leader of the rebellion) who probably racks up more air miles than George Clooney's "Up in the Air" character. Ted took the time to walk me through some exposure strategies after the recent improvements to RED's software decoding tools and the new sensor hardware hit the marketplace.
The first thing to understand is that RED is doing a pretty interesting thing that we really have not seen with camera companies before -- they are upgrading the imager in the existing camera body. While other companies, such as Panansonic gave us the DVX-100, 100a, 100b and many flavours of Varicam, RED has chosen to allow its owners to "upgrade hardware within the camera, rather than buy an entire new camera.
The hardware upgrade is allowing users to put the next generation sensor into their existing camera for about a third of the original price of the camera body. This extends the life of your investment and also gives you the same sensor as the newly announced EPIC camera (working prototypes were shown at NAB this year).
Much of the previous article dealt strategies on how to handle blue channel noise that is inherent with all silicon based sensors. The silicon generally makes all of the cameras see light in the blue spectrum, in RED's case, the sensor's native balance is about 5000K.
The first thing you notice with am MX sensor in a RED one camera is that there is far less noise when "pushing" the image in low light scenes. In RED's first generation M sensor there was a lot of visible noise in low light situations and you could not rate the camera past about 500ASA without seeing some very unpleasant artifacts under tungsten lighting. RED's RAW philosophy also means they do not do any kind of "autocorrection" in the camera, including sharpening and noise reduction, so when you push things you get what the sensor is seeing and you take care of it in post production.
These kinds of images are virtually eliminated with the M-X sensor. Not only are the images virtually noise free, but there is a far wider dynamic range. The new sensor (which will record 5K images to REDCODE RAW in the EPIC camera) has been measured at 13.5 stops of dynamic range. This is the kind of range which starts to rival what we are used to seeing in 35mm film stocks.
RED has also introduced a new and improved color science called REDCOLOR and a gamma setting called REDGAMMA. They have also introduced a new technology called FLUT. The way FLUT works is that it moves increases the usable range of the image without ever clipping the highlights. This makes for wider histograms and allows much more usable image area to be used. Just a reminder that the post technology improves all images captured in an .r3d file to date. When I loaded tungsten images captured two years ago on previous camera builds, the shots were greatly improved by the new algorithm decoding the RAW data.
With the new sensor the change is even more dramatic because the sensor has a significantly lower noise floor. 2000 ISO can be a very acceptable ISO for capturing low light images now because the sensor has very good tolerance in the blacks. When you boost a shot, there is not any noise to boost because the sensor is capturing such a clean image.
While the sensor is not any "faster" because it is still rated at 320 ISO, you can feel more comfortable going to 500, 640 or 800 because the image is noise free.
According to Ted, RED scrutinizes the image pretty closely. As part of their low light testing on the new sensor, they put the images up on a 4K projector in a theater to pick out the deficiencies. Projecting the images in this resolution allows them to see pretty quickly if the image is breaking down.
RED has also streamlined its offering of free software decoding tools. Previously there was REDALERT, REDCINE, and REDRUSHES. They have blending these tools into one application called REDCINE-X and they have also added a GPU R3D realtime playback and render acceleration card called the REDROCKET (at USD $4750.00 it is not inexpensive, but it does rip through R3D decoding).
Listen to full interview with Red's Ted Schilowitz:
Here is a preview of what some of the RED menus look like
SAMPLE when camera is in IRE VIEW MODE. Click on image for larger view.
SAMPLE when camera is in REDCOLOR View. Click on image for larger view.
FALSECOLOR View. Click on image for larger view.
SAMPLE when camera is in RAW View. Click on image for larger view.