Recording video with DSLR cameras has been the big new trend for a few months now, and while I have been right in the thick of those developments, I have been silent about that on here until now. I have been putting off writing about my experiences with DSLR based film making for a number of reasons; besides being busy, I was working with beta hardware and software from Canon for a while, so felt obligated not to talk about it. And the projects I was working on weren’t completed, so I wasn’t sure what details I should be avoiding posting online. Most all of those developments are now public in some capacity, so I guess it is time to discuss all of the options and technical details on here.
DSLR cameras with HD video options offer small lightweight formfactor with film like depth of field, using interchangable lenses and solid state recording. The flip side it that many of the options on standard video cameras that ease the management of larger projects are not available on DSLRs. Timecode is a big deal on a large projecct, but could be ignored when making a home video. The lack of real timecode on DSLR videos makes synchronizing external audio and tracking large numbers of assets more challenging. There is also a record length limitation on most current DSLRs. Monitoring is also an issue, with the viewfinder turning off when using the external video output, which would never happen on a regular video camera.
First off was the Nikon D90, that shot 720p24. That is a good frame rate for film makers, but the quality of the MPEG4 files is not worth serious work, especially with a lot of motion in the frame. Canon’s 5D MarkII was the next option to hit the market, with 1080p30. Dealing with the framerate is the most challenging aspect of using this camera. 30p is a fairly unusual framerate, supposedly requested by the AP for reporters to get both pictures and video for the web. Canon’s next 1080p release in the Rebel T1i offered an even more awkward and useless framerate of 20fps. Canon finally got it right with the next release, in the new 7D, which has options for 29.97, 25, and 23.976 fps at 1080p. It also has options for 50 and 59.94 fps at 720p, which makes it a much more versatile film making tool. The new 1D MarkIV offers those options on a slightly larger sensor, and with less rolling shutter artifacting, a phenomenon I will go into more detail on later. Nikon has released a number of other cameras, but with no new features beyond the D90’s 720p24. Panasonic released the Lumix GH1 which offers 1080p24, but the 4/3 lense system is not as universal as the Canon and Nikon offerings. Canon has pushed the technology the furthest, so most of my dialog will be in regards to their offerings and workflow options.
Regardless of which DSLR camera you chose, there are many challenges to be overcome in using them on a large production. First off, as mentioned before, you don’t have real timecode or keycode to track your footage or sync audio with. It is a tapeless format, so you have to have a separate data backup solution. In the case of the 5D there is the framerate issue if you plan to intercut it with film. There are also issues with monitoring and playback, since calibrated HDMI based monitoring solutions are hard to come by. All video capable DSLRs have CMOS sensors and therefore exhibit rolling shutter issues during rapid movement. This is caused by the top of the frame recording a slightly earlier moment in time than the bottom of the frame, with any rapid changes producing inconsistencies in the picture, changing from top to bottom. I am most familiar with the specific issues of the Canon 5D workflow, since I am in the middle of production on a feature film that is being shot primarily on the 5D. I will go into much more detail about how we solved those issues to leverage the enormous potential of that camera in my next article.
Some great info here, as this field is so new and of great interest. Now that I just purchasing a 7D, I’m realizing it is dramatic shift from simple Standard DV acquisition… CF cards, backup solutions in the field, XLR inputs, and then monitoring on systems like a CS3… I’ll be looking forward to your next article!
Thanks for posting!