I’m in a dealer’s showroom, stuffed with equipment used by Hollywood
DPs—ARRI cameras, DigiPrime lenses, etc. The owner shows me a prototype
HD camera so new it’s still in an unopened box. He hands me a cassette
that looks like a shrunken 3/4-inch cartridge or maybe an 8-track tape.
The camera looks like a small ENG camera without a VTR and doesn’t seem
special in any way. But Michael, the proprietor, excitedly explains
that the camera records a 4:2 image.
I figure he means 4:3 or 4:2:2 and I didn’t hear him correctly. I ask.
He repeats. 4:2. I mull this statement, then a gloomy light bulb
appears above my head (metaphorically...but then, everything here is a
metaphor). There is no third number because all the color is crammed
into a single channel, as with S-VHS. This new camera is HD Y/C. I wake
up in a sweat.
I made up that last line, but regrettably the rest is what I actually
dreamed. Let me make two points here: one, most of my dreams aren’t
nearly this pathetic and two, the low-cost HD formats and cameras we
already have are much better than they could be, and much better than
many currently consider them.
General wisdom of every new technology progresses through four phases:
Disbelief (there’s no way it could be any good), Hype (it’s the most
amazing thing ever), Backlash (it’s over-hyped junk) and finally
Acceptance or Rejection (it’s another tool with advantages and
disadvantages that will/won’t work for me). HDV has entered the
backlash stage. The first-blush of new love has faded from our cheeks
and loins, now we have to decide if we want to commit or move on.
HDV Image Quality
There’s no denying that HDV images can look great. Thanks to luma
sampling that equals HDCAM and exceeds DVCPRO HD, interviews, scenics
and many other shots have a level of detail that’s stunning. But
motion, especially fast-moving and detailed motion as in sports
footage, presents quite a challenge for the real-time MPEG encoders in
standard HDV cameras. It’s a challenge HDV can often meet, but not
always. (Go to www.studiomonthly.com to view examples of
both a good and bad HDV clip and you’ll see what I’m talking about.)
This is the downside of HDV’s impressive MPEG compression efficiency. A
brief review: Unlike DV, JPEG and similar intraframe compression
schemes, HDV’s MPEG compression searches for both spatial redundancies
within single frames and temporal redundancies across multiple frames.
An HDV camcorder will record a complete image of one frame (an
I-frame), but several subsequent frames hold just information that
differs from that complete image. The MPEG compression engine will
predict where to place the pixels in these subsequent frames.
Most HDV cameras record one I-frame followed by 14 predicted (B or P)
frames. This forms the 15-frame group of pictures (GOP) that makes HDV
compression so efficient. However, with high-motion footage, the
further a predicted frame is from an I-frame, the worse the prediction.
You can choose one of two paths to address this shortcoming. Shoot
carefully. Or use a shorter GOP.
Shooting carefully is always a good idea. For HDV, follow the
recommendations in the Hands-On HDV guide
(www.studiomonthly.com/5451.html),
minimize abrupt changes in camera motion with a tripod, dolly or
Steadicam-style camera stabilizer and keep light levels higher than you
would with DV. The pixels on those high-definition CCDs are really
small and need plenty of light to expose a good image, and changes in
color challenge MPEG compressors as much as changes in position (i.e.,
motion).
Once the camera is set and you have controlled camera movement and
light levels, you may still encounter content with motion that produces
images you don’t like. At that point, you can decide that you, the
client, and the audience can live with those images or that they won’t
notice a problem (and in my experience, most clients and audiences
won’t), that you like the long-GOP, high-motion aesthetic (some do) or
that you want a shorter GOP.
Even though this frame was shot handheld from a car and the bike cranks are
spinning around 100rpm, the image looks nice, with just expected motion
artifacts. This is probably an MPEG I-frame.
Several frames later, the MPEG compressor is trying to guess where all those
pixels from the previous I-frame should be placed. It isn’t doing a great
job in this predicted (B or P) frame. However, on playback most general
viewers think this footage looks fine even after the flaws are pointed out
to them.
Camera manufacturers have developed two responses to the challenge
motion presents a 15-frame GOP. JVC says use a shorter HDV GOP.
Panasonic says use a much shorter GOP, and don’t use HDV. And on a
related note, Canon decided to keep the 15-frame GOP, but maximize CCD
resolution and camera image control. While I don’t yet have a lot of
hands-on experience with all the low-cost HD cameras—at this writing
they aren’t all shipping—the different approaches are intriguing.
JVC’s GY-HD100U camcorder records images to a 6-frame GOP, minimizing
motion estimation needs. But according to the JVC Web site, a feature
of the camera’s CCDs, High Speed Twin Readout, can cause "a small
difference in the shading or color...between the left and right
portions of the screen." I’ve heard a couple of scary stories about
strong split black levels, but have only seen very minor differences
when I’ve looked for it myself; differences that are basically
unnoticeable. And the pro.jvc.com Web site explains the issue and offers workarounds.
The Panasonic AG-HVX200 records HD images as DVCPRO HD, a format with a
1-frame GOP. That takes care of predicted frames. But for HD, the
camera’s P2 solid-state media is limited to relatively short record
times, doesn’t record HD to tape (just to P2 cards or third-party hard
disk recorders) and has lower luma sampling than HDV. For 720p, DVCPRO
HD samples at 960 x 720 pixels versus the 1280 x 720 for 720p HDV; for
1080i, DVCPRO HD samples at 1280 x 1080 pixels compared to HDV’s 1440 x
1080. However, the HVX200 records at a variety of frame rates, offers
some 1080p recording and uses a 100 Mbps HD signal (and also supports
50 Mbps SD recording to DVCPRO50—a great format). Overall, Panasonic
says the HVX200 offers 81 different image-format and frame-rate
options. And the P2 cards’ recording time, about 20 minutes of 720/24p
on an 8 GB card, won’t be an issue for many users.
Canon’s XL H1 takes a different tack—it offers CCDs with much higher
resolution than some other HDV cameras, time code in/out, an HD-SDI
port and a completely adjustable color matrix. But the camera is front
heavy, and Canon’s not yet clear about how its 24f and 30f modes work.
However, the flexibility in image rendering is unprecedented at this
price level.
The perfect low-cost HD camera has yet to be announced. And, of course,
there’s more to a camera and a format than the size of its GOP. But
compare this to the pre-DV days when those on a budget debated the
virtues of Hi8 and S-VHS. With HD we can already choose from several
absolutely useful cameras. In fact, all of these HD cameras are better
than my dream camera.
Post your comments below
