What I've taken to doing is rendering a little higher quality proxies with Gearshift: basically the same as the default proxy except with "best" checked. That way I can render a pretty an SD DVD from the proxies instead of the original HDV M2T files in regular SD rendering time. I lose a little color dept from the DV colorspace being lessor, but it is still a little better than if I shot in DV to begin with. That way, aside from rendering the proxies with Gearshift at the beginning of a project, the workflow all the way to DVD is the same as if I'd shot in SD. It saves an incredible amount of time and I still have the option of doing a more time-consuming render later. Nobody other than me wants the HD version yet anyway!

Does the HDV to DVD look significantly better than a high quality DV cam to DVD ?

Does the HDV to DVD look significantly better than a high quality DV cam to DVD ?

Yes.

Do a search using Spot's username (DSE) and you will find he has given some excellent explanations as to why. The two most important are:

1. HDV has better colorspace which matches better the colorspace of DVD, so the colors look better.

2. You will always get better quality when you start with more pixels because all subsequent operations, until the final render when you go down to SD DVD resolution, are done with those extra pixels and are therefore more accurate.

It's an easy test to perform. Set up a controlled shoot that includes repetitive movement. Shoot a minute in HDV, and then a minute in DV. Capture both to your computer. Then, render crop the HDV so it is the same aspect as your DV. Render each to a DVDRW and watch the DVD on an NTSC monitor. You should easily be able to see the differences.

You can also just download this vid, and see the same shot that was captured as both SD (cam downconvert) and as HDV, edited as HDV, and output as SD, both in the same stream. HUGE difference.

Just to add one more point to John's summary:

3. HDV uses 4:2:0 color sampling, as does the MPEG-2 used for DVDs. DV uses 4:1:1 in the U.S. When converting 4:2:0 to a 4:1:1 color sampling format, half the color information is discarded, so you now have 4:1:0. Outputting this to the final 4:2:0 format at the same resolution doesn't add any color info, so it's 4:1:0 which means about 88% of the original color information has been filtered out. No wonder it looks sucky.

In practice, the end result CAN be a little bit better due to chroma smoothing in some codecs.

see the same shot that was captured as both SD (cam downconvert) and as HDV, edited as HDV, and output as SD, both in the same stream. HUGE difference

And if you actually shoot in SD (rather than shoot in HD and then downconvert in the camera) you may get even further degradation.

I was just looking at linked duck video. I see a huge differerence once the digital zoom kicks in, but not before. I have frame captures of both in front of me right now and I'm scrutinizing them pretty closely. They look the same to me. I almost hate to say it because I know there is a colorspace difference that I should be able to see. Anyway, for now, I'll continue to render SD DVDs from the "best quality" proxies knowing that price for the huge difference in rendering time is something that is hard for me to see.