The HDV downsampling thing...

Astronuts wrote on 11/30/2004, 3:24 PM
Hi,

sorry for my ignorance - I just took a look at the bali images on the VASST site and was struck at the quality of the 'downsampled' image. There's nothing freaky going on here is there? I mean with 'downsampling' you are simply setting up the 1440x1080 timeline to render out a new file at SD resolution yes? Is this done from the m2t file or is it from the cineform avi? I shoot with an FX-1E - 'PAL' version so I would just take my (1440 x 1080/1.333/25fps/up-field first) timeline/region and set the target file to be (720x576/1.333/up-field first/25fps), yes? And then just let vegas do its thang? Sorry, just this constant referral to 'downsampling' instead re-rendering or some other more commonly used term has me concerned that I was misinterpreting what was going on.

Thanks.

David.

Comments

Spot|DSE wrote on 11/30/2004, 11:23 PM
You are correct. All that is done to downsample is Pan/crop applied to shift aspect, and then conformed to 720 x 480.
farss wrote on 12/1/2004, 12:21 AM
If you're shooting with the PAL version I guess you might want to covert to SD PAL in whcih case 720x576.
But here's a question, as HDV is UFF shouldn't you convert to LFF for both PAL and NTSC. Now that's simple enough but doesn't that reverse the field cadence and cause problems with motion?
That's one of the few things that confuses me with video, well OK, the other one is how much to charge.
Bob.
vectorskink wrote on 12/1/2004, 3:37 AM
Hi guys

I am confused as well. I downloaded some .m2t files from the FX1 off the net and downsampled to SD on a DVD for a test and thought 'Hmmm What does DVD use? UFF or LFF?' I used progressive scan in the end. I then thought whatabout if you wanted to PTT the downsampled SD footage back to DV, which is LFF???? Is PAL LFF??

Cheers
Tim
farss wrote on 12/1/2004, 3:54 AM
PAL and NTSC DV are both LFF.
Bob.
Spot|DSE wrote on 12/1/2004, 6:51 AM
LFF=Lower Field First
UFF=Upper Field First
Coursedesign wrote on 12/1/2004, 9:39 AM
It might be worth noting that NTSC uses 4:1:1 color sampling, while HDV uses 4:2:0.

PAL uses 4:2:0 and DVDs use 4:2:0 (although they don't use the same 4:2:0 sampling algorithm, it should be a miniscule loss compared to rendering an NTSC 4:1:1 to DVD 4:2:0 which cuts the color resolution to 4:1:0).

It seems that when downconverting HDV footage to NTSC SD, it would be better to use at least for example the free Matrox DV50 4:2:2 codec as this would eliminate the color sampling loss. Anybody care to try?
Spot|DSE wrote on 12/1/2004, 9:41 AM
I've not done this with the Matrox, but have done this with the Sony 4:2:2 codec, and frankly, didn't seen any discernable difference.
Coursedesign wrote on 12/1/2004, 11:05 AM
You didn't see any difference on the DVD? The Vegas timeline would look the same of course.

If so, it could be that the MC MPEG-encoder is successful in its chroma smoothing. Would be interesting to know the limitations of that.

I'm a huge fan of intelligent interpolation after seeing what it can in do in practice for uprezzing video, but sometimes there just isn't enough information.