Color resolution when downconverting to SD
 

I have read several places that HDV color resolution, when downconverted to SD in the right codec, can hold 4:2:2 colors.

Can anyone verify this, and if so explain exactly how it works ?

1080i HDV (1440x1080 pixels, 4:2:0) contains 720x540 chroma values (Cr and Cb). Enough for 4:4:4 SD (NTSC). Your NLE shouldn't lose chroma values when downsizing. It will in all likeliness decompress the HDV material to uncompressed 4:4:4 or 4:2:2 before the downscale to SD.

Now that's a great reason to shoot in HD.

When you are downsampling resolution, as long as it is into a codec that will not restrict colour(chroma) info, the downsample will not stay locked at the origional 4:2:0, but will condense the chroma info that it has available. If that is SD 4:4:4, it will take advantage of that space. You aren't gaining anything, you are just condensing what was already there.

For best results I have found using a codec that does intelligent upsampling to 4:2:2, such as AspectHD. Then perform FX work or do CC in 16bit rendered out uncompressed or other non-restrictive codec. When all is done downsample to SD uncompressed and then render to DVD. I see the point of having the prestine 4:4:4 SD downsample, is it allows super quality DVD's. Very pro looking. I see the pre-downsampled HD as the best to do FX manipulation on, as the final downsample "cleans" everything up. Thats my preferance.

Agreed with Ken on all accounts.

-Steve

Well said Ken.... When SD is the final intended output, the "oversampling" afforded by the source HD signal produces very professional looking SD by increasing the effective chroma fidelity of the SD output compared to the 4:2:0 HD source.

My point exactly :)

Thanks guys for the valuable information. Now the follow-up question:

Will the condensed color (now 4:2:2 or 4:4:4) be an advantage (compared to SD 4:2:0) when pulling keys and doing blue-screen work?

You bet.... But SD is not usually 4:2:0 unless it's been converted into or captured (less common) in MPEG. Most forms of DV are 4:1:1, but the answer of "you bet" still applies for 4:1:1 DV source.

That's true in 60Hz countries, but the previous poster is from Norway. For PAL, SD normally is 4:2:0 in the DV variants, at least as far as DV and DVCAM are concerned, the exception here being DVCPRO which is 4:1:1. (And that is far less common than the other two.) An obvious advantage is when a DVD is the end product - using DV or DVCAM means the same colour space (4:2:0)exists throughout the production process.

Almost... except the 4:2:0 sampling on a DVD is different in structure to the 4:2:0 sampling used in PAL DV...

Could you elaborate? I'm aware that a DVD uses MPEG2, whereas DV is intra frame compression only, but my quote referred to colour space which I thought was the same in each case. Or am I missing something?

Basically the matrix of colour samples are a different shape.

PAL DV uses co-sited 4:2:0. Cb and Cr are sampled with every second Luminance samples on alternating lines. It goes:

Line A: Y+Cb, Y, Y+Cb, Y
Line B: Y+Cr, Y, Y+Cr, Y

MPEG-2 uses inter-sited 4:2:0. Cb and Cr are still sampled on alternating lines, but they're offset from the Luminance samples (which is a little hard to depict without a graphic). It (sort of) goes:

Line A: Y, Cb, Y, Cb
Line B: Y, Cr, Y, Cr.

So basically the colour samples for MPEG don't directly align with the colour samples for DV (either PAL or NTSC).