Originally Posted by hazydave
And that's kind of a shame. The XF100 is actually more of a compromise than the consumer 1-chip Canons.
Why not. I started into video with 1-chip cameras... my last serious one was the Sony HVR-A1. In those days, that was the best low-cost HDV camcorder you could buy (only $2600). I started to think about upgrades when the 1-chip Sanyo FH-1 ($300) delivered better images, at least in some situations.
There are two problems at work here: color and sensitivity. Sensors themselves, CCD or CMOS, don't know color. They have to be filtered. My Sony has the usual filter, invented by Dr. Bayer at Kodak, which puts arrays of microfilters. For the 2Mpixel HDV image, you have 1M green, 500K blue, and 500K red sensors. Given the filter, each sensor is receiving only 1/3 of the light entering the lens. For each pixel, you interpolate across the neighboring pixels... for example, if I'm on a green pixel, I have red and blue pixels in either direction. There's a pretty good chance that the red value for that particular green pixel would have been the average of the pixels before and after this current pixel. Same with red and green.
The problem is, that doesn't always work. So along borders, there are hard discontinuities in color. So you get weird color fringing. This was a pretty big deal in SD, less so in HD, but still an issue. And kind of a big one if you're chromakeying. My HVR-A1 did this kind of interpolation. So does the Canon XF100... one chip, 2/3 of the light tossed out, a 2Mpixel area, color fringing.
Most modern consumer single chip cameras have larger sensors. Part of that's because consumer think a video camera should also be a still camera, so they want megapixels. Bad idea.. but there's some help in video. When you have an 8Mpixel sensor for a 2Mpixel frame, as you would in HD, you can do "pixel bucketing" with a single-chip sensor. This is why Canon and JVC single-chippers have >8Mpixel sensors... Sony's doing something that's a different compromise.
The good thing is that you now have one red, one blue, and two green (or possibly, one green and one "white") sensor-level pixels, which get combined into a single video pixel. So there's no color interpolation failure. You're still tossing out 2/3 of the light, and worse yet, for the same sized sensor, your sensor sites (eg, hardware pixels) are now 1/4 the size of what they were... so there's less actual light making it to the sensors.
A 3-chipper is very different... it uses a diachroic prism to split essentially all of the light entering the lens into red, blue, and green components, each with their own sensors. Panasonic is the only company doing this in the consumer market; most pro cameras use 3 chips, at least until you get to huge sensors in DSLRs or digital cinema models, like the Red Camera. You get basically all of the light captured. Naturally, there's the temptation to use smaller sensors, but, as in the Panasonic case, they don't have to have huge megapixel counts... an HD frame only need 2Mpixels. The overall effect is that a 3-chip Panasonic with 1/4" sensors is getting slightly more light per video pixel than a single-chip 1/2.5" camcorder with 8Mpixel+ Bayer pattern.
Sony uses a kind of intermediate interpolation. They only use about 4M pixels in the video frame, but turn the sensor 45 degrees off the horizontal. This is the same technique FujiFilm used nearly a decade ago. They don't have any actual pixel values recorded... everything in the video image is interpolated. But based on the tilted sensor, the distance for interpolating a real 4Mpixel into a virtual 8Mpixel is less per interpolation. And the don't screw the color up as much along horizontal and vertical lines, which our brains are hard-wired to detect.
So yeah, Pannys screens suck, relatively. But they're delivering the best video. One other advantage of the Panasonic approach... targeting a smaller sensor, they can build a better lens in the same space. The TM700 lens is f1.5, versus f1.8 or f2.0 on most other consumer models (JVC's actually at f2.8.. they have a fairly huge 1/2.3" sensor, and can't deliver a higher aperture without growing the camera size still more).
Sony pushed ahead with "hybrid" image stabilization... if Panasonic doesn't have that in the new models, you'd kind of wonder what they're for. Sony's system uses the normal optical means for X and Y stabilization. But optical can't deal with Z-axis stabilization (eg, anti-rotation). But it's the perfect application of digital stabilization.