Update: DSLRs as motion picture cameras

In the one week or so since I posted about DSLRs, there has actually been several press releases about functional improvements to a few of the problems I listed. These improvements only scratch the surface of the mountain of issues that prevent DSLRs from being professional motion picture tools (the central reason is actually the attitude of the manufacturers that HD video on DSLRs is an added bell-and-whistle for still photo 'shooters', and not a primary function), but regardless, this week simprovements are worth listing:


1. The Foundry Demos Technology to Nuke CMOS Artifacts

This is actually month old news, but The Foundry is developing software that will remove the rolling shutter artifacts from footage, which is currently the #1 problem with the Nikon cameras. Even though this is helpful, I would assume that even on a high end computer, actually executing the motion estimation algorithm would take several seconds per frame, and add an extra step in the post-production process, not to mention extra storage for the export of the corrected files. It's clearly a fix, but an extremely inefficient one. If these cameras are ever going to be viable, the rolling shutter needs to simply not exist in the first place.


2. Firmware update for Canon 5D Mark II

Canon announced that in early June, they will have a firmware update that will allow for manual aperture, shutter, and ISO. This is a significant step that puts them leaps ahead of the current Nikon HD lineup (D90 and D5000), which in addition to automatic controls, also have the rolling shutter issue mentioned above. Canon seems to be taking professional filmmakers at least a little bit more seriously than Nikon.

DSLRs as motion picture cameras

I have a collection of manual Nikon lenses and accessories and want Nikon to make a DSLR that can function as a (at least somewhat) professional motion picture camera so that I can actually shoot something with it. However, their initial offerings (the D90 and D5000) are purely consumer products leave a lot to be desired. In spite of this, a lot of people (I call them "shooters," not cinematographers, sorry for the condescension) have decided to actually try and make films with these products because they are in search of the magical "DoF" that their 1/3" video cameras lack.

Trying to shoot something with DSLR lenses (whether on front of a 35mm adapter or on an actual DSLR) is already a setback because still lenses aren't designed for motion picture photography, particularly because they have a lot of issues such as breathing and focus markings that aren't suitable for a focus puller.

Just to be clear, I don't expect Nikon or any other manufacturer to put out a professional product with a $1000 price tag, but there seem to be some people who think of these products as "almost there" or even acceptable for professional use. However, even some consumer camcorders in the $1000 price range have full manual controls and a higher bitrate than the D90. People are getting hooked in by the "DoF" and seem to be willing to throw everything else away for it.

Below are the issues that I've seen with the camera, based off video samples and literature. Only one of these issues are specific to Nikon HD DSLRs (the rolling shutter), and the rest actually apply to all of the HD DSLRs on the market today (currently there is one offering from Canon and one from Panasonic as well). The only advantages I can see with these cameras are that you could sneak them into places that don't allow motion photography, or have permit situations that are prohibitive.



1. Rolling shutter

Nikon employs CMOS sensors that scan the frame from top to bottom, left to right. Therefore, from the top the the bottom of the frame there will be a 1/48 sec exposure timing difference (I'm actually assuming that this is the duration of exposure for these cameras, but I'm not sure since it's not under manual control... more on that below). Regardless, this discrepancy in exposure causes a "jello" effect that has been very noticeable in 100% of the handheld work that I've seen, a good portion of all moving camera work, and even some shots that were locked off. Until Nikon works this out, I would say that it's the #1 issue with their cameras.


2. No real viewfinder

There's a reason that camera operators look through a high quality viewfinder. It allows them to:

1. Check focus
2. Check the frame at maximum resolution (either optically or electronically) for any issues that need to be taken care of by them (such as lens flares or anomalies), or issues that need to be reported to someone else (such as positioning of set dressing).
3. It allows them to see the image in a totally glare-free environment.

Unfortunately, an SD LCD with a Hoodman attached to a D90 doesn't really cut it. Yes, this is what the operator looks at for steadicam shots, but thats because they don't have a choice.

It's certainly possible to shoot a film without a viewfinder, but it's a complete handicap for the operator, and you really don't want to shoot an entire movie that way.

One workaround is to do it Roger Deakins style and operate a remote head while looking at a full size HD monitor with a larger hoodman that can actually get the glare under control (the D90 does have HDMI out). However, if the production can afford a remote head, they probably aren't using a DSLR to shoot the film.


3. Bitrate

The encoder on the camera employs a variable bit rate, which seems to fall between 4-24 mbps depending on the complexity of the frame. Unfortunately, there is no way to set the bitrate manually, but some people have gone so far as to point the camera at a complex chart before shooting in order to force the camera into a higher bitrate for a limited amount of time.

4-24 mbps isn't high enough to produce an image that is viewable on a full rez HD display or projection without noticeable compression artifacts in at least some parts of the image. As a reference, DV is 25 mbps (albeit a codec that isn't as efficient), and DVCPROHD is 100 mbps.


4. No manual shutter or ISO

Nikon has promised to do a firmware fix for this, but it may take a while. This is a basic function that is needed for any kind of real image control.


5. No over/under cranking

Lack of overcranking is bigger issue than lack of undercranking, but my guess is that 24fps is the approximate maximum data throughput of the camera at this time. Lack of this ability pretty much precludes it from any use on commercials or music videos, and is still a major limitation for narrative filmmaking.

RSS feed

You can view updates as RSS by using this link with your reader:

http://feeds2.feedburner.com/cineblog-bkantor

Issues with the Color application

After color timing a feature film, three commercials, a web series, and a short film on Final Cut Color, I am finally ready to list my complaints about the application. I do consider it an amazing value (it has some features that are typically only available in expensive hardware based systems), however, it hasn't really reached it's full potential due to several flaws that could be easily ironed out.


1. Secondaries are pre-clipped

This is the single most difficult thing about the application. In a professional color grading application, if I make a primary grade that results in blacks or whites being clipped at 0 or 100 IRE, respectively, and then I make a secondary grade that affects the clipped area, I still have access to the information that was clipped. The clipping is not "burnt in" (i.e. the information is not thrown away) until the shot actually leaves the application. Not so in FCP Color. If I make my primary grade and clip anything, It comes in pre-clipped to the secondaries tab. This makes doing basic tasks like using a power window to take down a hot area impossible. No other professional color grading application behaves like this.


2. No support for speed changes or speed ramping

If I grade something in Lustre or Scratch that has source footage with speed ramping done in post (as opposed to in-camera), the timeline needs to be prepped beforehand... either the shot needs to be pre-rendered with the speed change, or the shot needs to be included as an extra shot at the end of the timeline without a speed change, so that it can be brought back into the original editing application or another application later to have the speed ramp applied.

With FCP Color, it's essentially the same. If I have a shot in FCP that has a speed ramp, I can still send it to Color and round trip it back, but when it shows up back in the FCP timeline, the in/out points and speed ramp need to be manually re-entered. Apparently it's not information that can be encapsulated in the XML that FCP uses to communicate to Color. I would expect this much from an application that's supposed to be standalone, but Color is part of a suite of applications, and I would think that Apple could figure out a way to make them communicate better with each other.


3. The render queue is clumsy

If I grade and render out an entire project, and then go in and make spot changes based off notes, the only way to make sure the clips I modify are all re-rendered is to manually add them to the render queue. The app doesn't keep a list of what needs to be rendered. This could be as easy as adding an "Add un-rendered to render queue" button, but it's not there. In FCP, all I need to do is Apple+R to render anything that's un-rendered, but in Color that functionality just doesn't exist. To complicate matters, there is also no shift-select ability to easily add a group of clips to the queue. One has to command-click all of them.


4. The timeline is clumsy

There is no basic scrolling to navigate the timeline. You can scroll the portion of the timeline that is visible to you (you can't zoom out all the way), or you can go immediately to the beginning or the end of the project, or you can even manually enter a timecode, but you can't just scroll or scrub through it in the same way you can in any other color grading application, or any NLE.

"Herpes Boy" Trailer

The "Herpes Boy" trailer is up at http://www.youtube.com/watch?v=SoPLqvdbhQA

This film was shot on the Red camera handheld with a 10-100 zoom. More about my experiences with the Red are forthcoming.