The pixels on the sensors are oriented on the diagonal: think of the holes in a chain-link fence, and you'll understand their layout (see the diagrams at http://bssc.sel.sony.com/BroadcastandBusiness/minisites/HDV1080/HVR-V1U/). The active area consists of 1080 staggered rows of 960 photosites each. It's as if the even rows were pixel-shifted compared to the odd rows. Each row, however, is sampled 1920 times: each photosite is sampled by itself and a sample is taken in between photosites, combining information from the two photosites on either side and the ones immediately above and below. In essence, each scan line comprises two sets of interleaved sample types: 960 pin-sharp samples and 960 more diffuse samples averaging the four surrounding photosites. Sony claims that this gives the V1 full 1920 x 1080 resolution. What I saw was impressive, but don't throw away your CineAltas just yet.

The CMOS imager also allows true progressive scan, at both 24p and 30p rates. The 24p is recorded and output as 60i using the standard 2:3 pulldown. The 24p is a true 24 fps mode that evenly samples 24 frames per second, not the pseudo-24p CineFrame mode from earlier Sony HDV cameras. I had no problem capturing the V1's 24p as 60i using Final Cut Pro 5.0.4, converting the clip to DVCPRO HD, and then extracting the original 24 frames per second using Cinema Tools. (Cinema Tools refuses to work on raw HDV clips, hence the conversion to DVCPRO HD. I also used Apple Intermediate Codec and uncompressed formats successfully.) The recovered 24p clip showed perfectly even timing between frames with no temporal syncopation: it's really 24p. The shutter speed defaults to 1/48 second, 1/24, and 1/40 (but not 1/32), and higher and lower speeds can be dialed in.

Sony is working with NLE vendors to allow real-time 2:3 pulldown extraction during capture. Vegas may already support it.

The CMOS sensor allows scanning from a reduced area at higher than normal speeds. The V1 exploits this with Smooth Slow Recording: the camera captures 3 seconds of 640 x 360 pixel video (or 6 or 12 seconds of 512 x 320 pixel video) to an internal memory buffer at 240 fps, for a 4x slow-mo. It then up-converts these small pix to DV(CAM) or HDV depending on format settings, and records them to tape. The resulting clips are a bit soft and show some compression artifacts. Although they won't intercut cleanly with full-res material, they're perfectly suited for motion analysis work-the system's symbology for Smooth Slow Recording shows a stylized golfer, the target market for the many low-res, high-speed shooting schemes available on various camcorders.

Like other CMOS camcorders, the V1's chips use a rolling shutter, sampling scan lines sequentially instead of capturing the entire frame at once. Still frames taken from fast pans show tilted vertical lines-just like stills grabbed from a tube camera's clips or photographs taken with still cameras using vertical focal-plane shutters. The 1x playback of such pans on CRTs look better than CCD camera pans do, because CRTs are also sequentially scanned, but all-at-once displays may make the V1's pix look distorted. (Of the two LCDs I use, the Panasonic BT-LH1700W scans like a CRT: the Z1 CCD pans look tilted and the V1 CMOS pans look normal. The HP L2335 buffers the image and displays it all at once: the Z1 pans look normal and the V1 pans show tilt.) Fast-moving objects can look distorted in extracted stills, though I have yet to see anything objectionable in moving video. If you're considering a V1 for film-out, you'll want to test this aspect of the camera's imaging and see if it suits your needs.

Performance

The V1 hasn't shipped yet (at least, not as of the time I wrote this), so it's too early to generate hard data on how it performs. Still, what I saw may be of interest, with the caveat that the released version of the camera may or may not behave like the prototype I worked with.

Click To Enlarge DVX100, HVR-V1, and PD-150: the V1's forward handgrip provides the best balance.

The zoom ring is a free-spinning ring like the one on the PD170, instead of the indexed, hard-stopped ring on the Z1. While this means the V1 gives up focal length markings and the repeatability of the Z1's and HVX200's zooms, it does allow the zoom ring and zoom rocker to be active at the same time. The prototype's zoom ring lacked the instant responsiveness of the PD150 I use-there was a slight lag between turning the ring and seeing movement, a delay I put down to pre-release firmware (I hope).

Sensitivity on the V1 is rated at 4 lux, compared to 3 lux on the Z1. I found the prototype to be roughly one and a half stops slower than the Z1, two stops slower than the Panasonic HVX200, and three stops slower than the standard-definition PD150 and Panasonic DVX100.

Dynamic range and noise on the prototype looked roughly comparable to those of the Z1, but I'm withholding detailed judgments until a shipping version of the camera is available for test.

Resolution looked surprisingly good, both on test charts and on real-world material. One telling factor is that sharpness can be dialed down on the V1 without the picture going to mush, as it will on the Z1.

In interlaced mode, when dual-row readout is involved, each output scan line is the sum of two adjacent scan lines on the sensors (see http://www.adamwilt.com/TechDiffs/FieldsAndFrames.html). Recall that the even lines are pixel-shifted with respect to the odd lines, so in effect you're getting 1920-pixel sampling (plus 1920 softer, four-photosite in-between samples). Images were noticeably more detailed than similar images from the Z1. Equally importantly, the V1 shows much less aliasing than the Z1 or the HVX200.

Click To Enlarge An informal comparison of detail rendering: V1 on top, Z1 below. Proper test-chart shots will be withheld until I can test a production model.

In progressive mode, it's more of a mixed bag. Dual-row readout isn't used, so each output scan line is the sum of 960 sharp samples and the interposed 960 soft samples. Sharpness is still high, but aliasing is more evident; test charts and side-by-side tests with a Z1 and an HVX200 (both with 960 samples per scan line) show aliasing atop the sharper signal. The aliasing is at the same frequency as on the Z1 and HVX, but much less prominent, since the in-between samples fill in some of the missing information, albeit somewhat less crisply. The result is a sharp 960 x 1080 image mixed with a softer, offset 960 x 1080 image. It may not be possible to pick a single resolution figure and declare that the camera's limit.

Vertically, this same alternation of sharp and soft samples leads to an image that looks like a 1080-line image at times, a 540-line image at other times, and something in-between the rest of the time, especially when gentle vertical motion is present. It's hard to describe, really. Test charts show a clean 800+ lines vertically on static pictures, but near-horizontal lines show more vertical aliasing than might be expected, although it's a subtle effect.

The zone plates on a DSC CamAlign chart indicate that the V1's predominant aliasing is in the diagonal direction. The V1, with its 45-degree tilted pixels, is not the camera to choose for detailed shots of chain-link fences.

Overall, the raw resolution of the prototype V1 surpasses that of either the Z1 or the Panasonic HVX200, and may even challenge Canon's crisp XL H1. I'm eager to test the final, shipping version.

Conclusion

Sony squeezed out some features in the quest to put HDV into a PD170-sized package: analog-in recording, 50 Hz/60 Hz switchability, low-light capability, and standard video connectors on the camera body. But the result is a compact, lightweight Handycam that shoots sharp, clean HDV while offering the best laid-out controls found on a small Sony, and true progressive capture to boot. Rolling shutter will be an issue for some, but overall, the V1 looks like a worthy addition to the choices available to HDV shooters.