> What does ICP stand for? <
Interlaced Chroma Problem, as examined and reported by Don Munsil and Stacy Spears. It's not a bug, but a problem common to all DVD players and MPEG decoders, that exists whenever the chroma fields from an interlaced static image are upscaled. It can look like CUE, but isn't, and can appear streaky, or as a mottling or plaid effect in static solid color areas.
Vertical low-pass filtering applied after
deinterlacing can completely correct the ICP issue (e.g. 3-tap vertical blur with coefficients of 0.25, 0.50, 0.25). This can be done without loss, because there are only 240 lines of chroma actually on the disc. The problem arises when that is separated into two 120 line fields, upsampled, and interleaved into a 480 line frame. This results in false-color artifacts (chroma aliasing).
Check out the Special Report on the www.hometheaterhifi.com
site for lots more details, and a beautifully written analysis and explanation by Don and Stacy. It's a gem.
The approach that's used in the Sage chip is a side-effect of a cost saving technique, due to their throwing away one field of chroma entirely. It hides the chroma problems somewhat effectively, but can introduce a flickering effect on sharp edges, especially animation. They also use a vertical low-pass filter, intended to remove chroma artifacts common in composite video sources, which is very effective. So it could definitely be improved upon, since it unnecessarily reduces the vertical chroma resolution from 240 to 120 lines.
> Also, wouldn't vertical chroma filtering adversely affect source material that does not have the Chroma Upsampling Error, such as laser disk programs and broadcasts? <
No, not if it's only decimated once, rather than twice, as the Faroudja chip does. These are composite video sources, and the filtering we're discussing can be very effective in minimizing artifacts normally present in those domains, and is frequently applied anyway as a chroma-noise reduction technique.
Re: the lip-sync error, Allan explained:
> With the de-interlacer's four field lookahead, there is no way around it other than delaying the audio to match. If you add a 3d comb filter to the video signal path, such as from a high end laser disk player, you increase the lip sync. error. <
Thanks, Allan. This is very good information to have and consider. I always assumed that any
of the video processing operations that were applied in the temporal domain (which is most effective) would have a (cumulative) impact, regardless of how they were implemented. Yet we hear some folks commenting on certain approaches as being objectionable, and others seem not to be aware of any anomalies.
> I wanted to let you know that the 67ms video delay is very noticeable. <
Thanks, Ken. So the problem is there, and how apparent it is will be dependent on each user's sensitivity, as well as the presence or absence of other A/V processing elements in the signal chain.
I appreciate that you picked out some of the most egregious examples, and gave an unvarnished appraisal of the unit's capabilities.