Nikon

DSLR Movies, Pros and Cons

In my new tradition of rambling on about a subject only to post again a few days later with a more succinct summary of my thoughts, here’s a quick rundown of why you should be excited about shooting video with your DSLR, and why you should reserve some modicum of wait-and-see caution.

First, the reasons to get excited, ranging from the obvious to the more obvious: 

  • In the case of the D90, what a sweet deal. $1,000 for the body. I paid nearly that for my HV20. The real win here is that a resolution that is low to medium for a DSLR is positively overkill for HD. Nikon’s “bargain” SLR is overqualified by a mile for 720p video.
  • Use your DSLR lenses. If the 5D MkII (or whatever) shoots video too, then both Nikon and Canon peeps can rejoice about this one. A big drawback of the RED One is the expense of the lenses (which need to be top-notch), and a perceived drawback of next year’s Scarlet is the fixed lens.
  • A big-ass sensor. The D90’s sensor is roughly the size of the RED One’s. The rumored 5D MarkII’s is way bigger. That means more control of depth-of-field and more predictable results from your stable of lenses. It also means that a $1,000 camera is now making images that, at web resolutions, look an awful lot like those from much more expensive kit.
  • 24p. I’m still reeling from this one, but somehow the D90’s video wound up being 24 fps. Hallelujah. It so easily could have been anything else.

And now the reasons to reserve judgement:

  • 720p only, at least in the case of the D90. Personally I love 720p, but it’s quite a subset of what your DSLR can do.
  • 24p. And that’s all, on the D90 anyway. No overcranking or undercranking. And who knows what frame rates other DSLRs will offer. A camera on which video is an afterthought is not likely to offer a wealth of options here.
  • CMOS. CMOS roll. Roll MOS roll. Are DSLR chips, which have never had to fear shearing, skewing and wobbling from rolling shutters (mechanical shutters negate this), going to fare well when recording video? Or will they be jello-cams?
  • Limited running time. Aparently there’s some red-tape reason why the D90 is restructed to five minutes of video. Those five minutes will still cost you about 600MB.
  • No external audio input. But the D90 does have a mic. You’ll be dual-systeming it. Fortunately there’s cool software out there for syncing audio based on waveform matching.
  • Manual control. Although things look good for the D90, the thing about shooting video is that you need the same kind of quick access to key manual exposure controls that SLR stills shooters require. But will the video options be ghettoized in a deep menu?
  • Manual focus. Not in and of itself a problem, since that’s how pro video and film gets shot, but that LCD screen won’t be reliable for critical focus at HD res. HD cameras have handy focus-assist features like edge enhancement and LCD zoom.
  • You’ll be at the mercy of a codec. The D90 scores well here on paper, but you are still dealing with heavy (and inefficient) compression of a baked-in color palette that may not respond well to agressive grading, as tends to be the case with perceptual compression. This is a tough thing to swallow when you’re holding in your hands a camera that lives and breaths raw formats for stills. Maybe someday someone will make a DSLR that lays down CinemaDNG sequences. Maybe? How about Please God Yes.
  • But maybe worse than baked-in color and codecs, you’re at the mercy of whatever the camera can muster in realtime debayering. Most video from compact cameras looks bad not because of compression but because of the hasty techniques used to rapidly build an image from a tiny subset of the sensor’s photosites. I offer up my beloved LX2 as an example—its videos may be 848x480 on disk, but what’s actually present in the images is far less resolution than that. This hard-to-quantify factor could be the real pitfall, although none of the (heavily recompresed) D90 sample vids I’ve seen have exhibited egregious demosaicing artifacts.

I said I’d stop hypothesizing, but this is an interesting enough subject that I felt it was worth clearing the air about both why this thrilling new trend in DSLRs is so great, and why it’s not as great as one might hope.

UPDATE: Mike commented below with a link to dpreview’s sample D-Movies. If you download the original AVI files you can see some sizzle and color sparkle that is symptomatic of the expedient debayering I describe above.

More D90

From robgalbraith.com:

The D90 allows you to select the aperture (from wide open to f/8) prior to commencing, then it handles the adjusting of ISO and shutter speed automatically as needed to maintain video brightness as lighting conditions change during recording. To disable automatic exposure adjustment, it’s possible to lock exposure prior to beginning the recording.

Sounds a bit like the HV20/30 dance to me, where you’re always aiming the thing at a light trying to lock it into 1/48th shutter. Stock up on ND if you want that filmic shutter and shallow DOF at the same time in daylight! Still, if this is true, you can lock down your auto settings and, perhaps somewhat clunkily, explicitly set a shutter speed. 

Of course, unlike Rob, I view the lack of automatic white balance and as a good thing. I’m also baffled by his desire for a digital zoom. I guess what stills folks want from a video camera is different than what film folks want from a stills camera that, uhm, shoots video.

The boards are positively buzzing about this (dvxuser, scarletuser, dvinfo, hv20.com, Rebel Café). There’s some clarification, some idolizing, some bashing, and enough hypothetical confusion that I’m happy to refrain from any further speculation of my own.
We’ll see real sample movies and detailed hands-on reviews soon enough.

Still more on this.