February 27, 2006
HVX-200 CCD InformationPlease note, this blog has been archived and now lives at www.discretecosine.com
Those who are of a geekish bent might enjoy this article from DVXuser.com. It's an official statement from Panasonic on the design and implementation of the imaging block in the HVX-200. The brief summary is that they're using 960x540 CCDs which are intentionally misaligned from each other. This means that the Red CCD sees slightly different bits of light versus the Blue CCD and so on.
It's an interesting approach to dealing with HD on a small CCD. By using a lower resolution chip, they're able to have bigger pixels, and thus (theoretically) better low light performance. It does raise a couple concerns. Follow the jump to read more.
One of the major factors differentiating the HVX-200 from the Canon XL-H1 and Sony Z1U is the use of a progressive scan CCD. This allows them to generate a 24p image without having to do any deinterlacing.
Deinterlacing always (for most cases of "always") costs you some vertical resolution, no matter how clever the scheme. You're filling in information that wasn't captured. With the most basic deinterlacing algorithm (where you double every line) you end up with half the vertical resolution of a true progressive image. Using fancy math and whatnot, most deinterlacing routines are closer to 70% of the full resolution.
However, the chips on the Panasonic are already half the resolution of the chips on a camera like the XL-H1. So, it seems to me that you're losing much of the advantage of a true progressive chip. Yes, the chip offset should recover some of that, but the offset technique does strike me as cheating just a bit. I'm not quite sure why, maybe because I'm associating it with the kind of sneaky marketing that advertises a 3 megapixel still camera as being 8 megapixels after interpolation.
Additionally, one could make the case that the 4:2:2 moniker is a bit inaccurate when referring to the color sampling on this camera. It's perhaps more like 4:1.3:1.3 or some such. You're getting more samples in the codec, but that added sampling accuracy in the data stream doesn't necessarily correspond to sampling accuracy in the original image. I'll withhold judgement on that until I can shoot some tests and look at the individual channels.
Hey Pana, want to send me a loaner?
Posted by at February 27, 2006 9:44 AM | News