All of the updates coming out of Magic Lantern’s camp recently have been RAW video related. And while there has been news enough on that front to keep us excited, we were happy to see something new coming from the ML team today.
Dubbed Dual ISO, Magic Lantern’s Alex (a1ex) has unlocked three full stops of dynamic range that the Canon 5D Mark III and 7D sensors couldn’t previously take advantage of. That brings total dynamic range to about 14 stops.
The details are technical — if you want to get into the nitty gritty, a1ex has prepared an in-depth PDF for you here — but the gist is that your sensor can, in fact, sample half of the lines at ISO 100 and the other half at ISO 1600 (or higher) at the same time.
The photo above shows the difference this makes. On the bottom left you see the photo taken at ISO 100, while the top right shows same shot taken using the ISO 100/1600 combination method (high-res version here).
Here’s another sample shot, this one taken entirely with the ISO 100/1600 configuration (pixel peepers can check out a high-res version here):
The reason this is only possible on the two cameras mentioned is hardware related. The 5D Mark III and 7D both have two ISO amplifiers. The chip that samples the image from the sensor is set via firmware to sample the same ISO from all the lines. The hack allows it to sample half of them at one ISO and the other half at another.
For the 7D, this function only extends to taking photos. The 5D Mark III, on the other hand, can also take advantage of Dual ISO in video mode, assuming you have yours equipped with the ML ability to take RAW video.
Of course, the improvement doesn’t come without pitfalls. You’ll get half-resolution in highlights and shadows in addition to some aliasing and moire in the same areas. You also lose the ability to critical focus when zooming in.
But if none of that seems like too big a price to pay, you can learn more about the tech by checking out the aforementioned PDF or visiting the original forum post where a1ex has kindly provided the code.
Image credits: Photographs by Luke Neumann.