By Adam Wilt, April 17, 2007
Akira Kurosawa's Rashomon (1950) tells the story of a rape and a murder from the viewpoints of several different witnesses, none of them in agreement. It's a classic film about the subjectivity of perception and the difficulty of discerning truth. Indeed, psychologists refer to a "Rashomon Effect" when discussing this phenomenon. (There's also a famous story about six blind men and an elephant, but as this is an article about video equipment, discussing the viewpoints of sight-challenged observers might set the wrong tone.)
I recently held a three-camera shootout with motion graphics artist Marco Solorio and filmmaker Marshall Spight. We compared the Sony HVR-V1, the Canon XH A1, and the Panasonic HVX200, all in 1080p24 and 1080i60 modes. Marco was interested in the V1 or HVX200 as a 24p project camera. Marshall was on a quest to find a successor to his Sony DSR-500 for future storytelling. I wanted to see how the A1 and V1 compared side-by-side.
The cameras performed just as they have in other tests (see the four-camera tests at www.adamwilt.com/HD/). On a Stouffer grayscale test, all three cameras showed about 8.3 stops of latitude, with perhaps a stop more if we stretched the blacks, squinted, and used our imaginations (all cameras were set up with their knees set to low to maximize dynamic range). There were differences in tonal scale rendering and colorimetry among them, but not to the extent that one camera outclassed the others.
Click To Enlarge
Four cameras lined up, connected to two monitors via an RGB switchbox. We swapped my HP 23-inch LCD for Marco's 24-inch Dell during the test.
|
The A1 and V1 both did well on the test charts in terms of resolution, handily exceeding the detail level available from the HVX (and from the Sony HVR-Z1 I threw in as a control), but we wanted to see how they looked on a real-world scene, so we focused the A1 and V1 on my garden. The scene contained plenty of fine detail at odd angles as well as areas of extreme contrast. The Canon and Sony were set to their default sharpness-Marco disliked the flatness of their unsharpened images-while the HVX was set at Cinelike V gamma, sharpness +2 (see "Exploring the HVX200," Nov. '06 DV, for my rationale).
Marco quickly but reluctantly ruled out the variable-frame-rate HVX for its comparative softness-why shoot HD unless you're getting all the detail you can? We focused on the A1 and V1.
In interlaced mode, both cameras rendered images with the same perceptual sharpness-remarkably, the two were so close that on a static image there was little to tell them apart. But adding slight motion, such as foliage moving in a breeze, gave an edge to the Canon. Its "traditional "1440 x 1080 CCD arrays rendered clean, pellucid images while the Sony's diagonally arrayed 960 x 1080 CMOS sensors added a slight granularity or grittiness to the image, probably due to its residual 540-line aliasing. While the slight motion of fine detail didn't pixellate as obviously as on the HVX200 or the Z1, the V1's pictures looked almost as if they were shot through a fine-ground glass. There was a faintly textured quality to its images, compared to the utterly limpid spatial rendering of the Canon.
Mind you, we came to this conclusion from staring long and hard at the pictures displayed on a 24-inch 1920 x 1200 Dell LCD that Marco brought, as well as my 17-inch 1280 x 768 Panasonic LCD. The difference I'm describing is miniscule, and was only apparent because we were doing a side-by-side test. Were we viewing the V1's pictures in isolation, we would not have seen anything amiss.
Once we put the cameras into progressive, the picture reversed: while the V1's images became somewhat coarser and more aliased, the Canon's diagonals were the deciding factor. The Canon uses vertical pixel shift in 24F mode instead of true progressive CCDs, but the predominance of the green channel's signal (which, at 71.52 percent of the luma signal in HD's ITU-R BT.709 colorimetry, counts for over twice as much as red and blue combined), together with the default level of sharpening, outweighed the smoothing effect of vertical pixel shift. Low-contrast edges looked perfectly fine, but bright, contrasty edges showed half-resolution jaggies.
Again, we were looking at very fine detail in a side-by-side comparison; had we been looking only at the Canon in 24F mode we would have found its pictures perfectly acceptable.
As Marco was solely interested in 24p performance, he declared the V1 the winner. I declared a draw: although the V1 did slightly better in progressive, the A1 was a bit better in interlaced. And while the V1 is lighter and easier to use handheld, the A1 is better in low light, more tweakable, and less expensive. I couldn't pick a winner between the two. Too much depends on what you want to do with a camera, and how you rate all the tangible and intangible points of comparison.
As for the HVX200, don't be too quick to dismiss it. When I went to pull stills for this article, I noticed that while both the Sony and Canon showed characteristic digital-camera purple fringing along overexposed edges at the sides of the picture, the HVX200's images were completely free of this artifact. The HVX200 may render a softer image, but in this case it was more naturalistic than the two sharpies. [Disclaimer: I own an HVX200.]
Marshall's reaction? When he first arrived, I was playing a 720p24 time-lapse HVX200 clip full-screen on my 1440 x 900 MacBook Pro. He gazed at the footage for about 5 seconds, and said, "That's it, I'm done. We don't need to look any further. That's more than enough detail to tell stories with."
One test, three cameras, and three participants: three different viewpoints. True, nobody got raped and nobody got murdered (though the HVX200 felt a bit underappreciated), yet the Rashomon Effect still prevailed. What you see depends on what you're looking for, and you can't count on another person to see the things that are important to you.
That's the problem with tests and reviews in general. Cameras are complex things, and very hard to sum up in a single article. Invariably, information is left out (my camera reviews routinely exceed their assigned word counts, yet there's only so much detail I can cram in about cameras with manuals that routinely exceed 150 pages), and what gets left out depends on who's doing the writing. Give a news shooter a digital cinema camera, and he'll complain about its ungainliness for single-person run-'n'-gun operation and ignore its ability to shoot 24p. Give a cinematographer an ENG camera, and you'll hear about its excessive onboard image processing and its unsuitability for working with film-style accessories. Swap the cameras between the two reviewers, however, and entirely different pictures will emerge.
The problem is even more acute when a camera is rated by a single number, as DV requires (and DV is far from alone in this practice). It's argued that a single number is what readers want, that it allows them to size up a camera at a single glance and compare it to others-but is that really possible?
I've given every HD camera I've tested a 4.5 rating. Three different Sonys, a Panasonic, a JVC, and two Canons-spanning the range from $4,000 to $10,000-as different from one another as it's possible to get. Yet they've all wound up with a 4.5. I'd prefer not to supply a numerical rating for cameras, but the magazine has insisted, so I have to choose something. My 4.5 ratings aren't intended to say, "Buy this one now" so much as, "This one does something amazing for the price and breaks new ground. See if it meets your needs."
These are the early days of affordable HD, and each camera when it first came out has done something better than anything else available. Upon release, Sony's HDR-FX1 and HVR-Z1 were the first affordable 1080i camcorders, and were well worth their 4.5 ratings. Today, they might only earn a 3 or 3.5. The new kids on the block-Sony's HVR-V1 and the Canon XH series-capture sharper images and offer more flexibility and control, for less money than the Z1 first sold for. Things change quickly when a new market opens up, but that means that comparing a Z1 with an A1 or a V1 based on their numerical ratings is misleading. There has been progress in the past two years.
So what's a reader to do?
Read any review with an eye on the fact it's going to be both incomplete and opinionated, simply because time and space are limited and the reviewer can't help but have a viewpoint. If you can discern where the reviewer is coming from, you can better judge how the reviewer's viewpoint relates to your own concerns.
Read as many reviews as possible, to diminish the effect of individual bias. But don't fall victim to analysis paralysis, wherein you're too busy reading reviews to ever pick up a camera and shoot with it.
Finally, as Marshall noted, the HVX200 had more than enough detail for his purposes. He didn't need to look any further. He could focus his attention on writing and telling better stories. In the end, it's the story you tell, not the camera you use to tell it, that counts.
At least, that's my viewpoint.
|