confused by JVC website re downsampling

brianluce

Carbonite Member
JVC says:

"When HD originals are intended for release on SD media, especially DVD, JVC recommends finishing in original native high definition, thereby preserving maximum image quality. Once completed, an HD project is easily output in SD for final release.The final DVDs will be of significantly better quality than if the project is posted in DV from an in-camera conversion.

Here's why: The HDV format records significantly higher resolution and the chroma is subsampled at 4:2:0. If a DV dub is made from an HDV recording, the extra resolution is lost, and the chroma subsampling effectively becomes 4:1:0. This is less than optimal for DV which records at 4:1:1 (in the U.S.), and less than optimal for DVD which utilizes MPEG2 at 4:2:0. Modern non-linear editing systems can edit the HDV recording in HD, and do any processing (extra enhancement, effects, etc.) in the higher resolution HD domain, and with the full HDV chroma resolution. The final DVD can be made from the NLE's downconverted output, but with the full chroma resolution and 24p cadence of the original MPEG2 recording. "
------------------------------------------------------

I don't get this. first they say HD will look better than footage shot natively in Standard def DV. but in the second paragraph it says that the extra res and chroma is lost and you end up with 4.1.0 subsampling which is "less than optimal for dv". isn't this contradictory? how can downsampled hd burnt to dvd look better if it's not even 4.1.1? I'm also confused that they seem to use the terms "HD" and "HDV" interchangebly.
 
I think they mean, don't convert to DV and edit and encode from that. If you do you're getting effectively 4:1:1 colour. They're saying to keep it in HDV (Or, I assume, some other 4:2:0 codec)
 
HDV is 420. If you go to DV that's 411 (natively) then your HDV downconvert will be 410.

Going directly to DVD is going from HDV 420 to DVD 420 which is far superior than a costly transcode to DV. If you must work in SD then I frame 50Mb 422 or DVCPro50 would be the best solution to work in.

good luck
 
stephenlnoe said:
Going directly to DVD is going from HDV 420 to DVD 420 which is far superior than a costly transcode to DV. If you must work in SD then I frame 50Mb 422 or DVCPro50 would be the best solution to work in.

good luck

What is I frame 50mb? is that an option in component out? same for dvcpro50?
 
I don't know if it is a problem in PAL land. My take on it is this:

1 - If you need/want HD, obviously shoot in HD
2 - If you need/want SD, for highest quality shoot in HD, then downsize in your NLE to SD for the DVD/output format. Possible formats that I've toyed with are AIC, uncompressed SD, and DV50. DV50 is probably good enough and is the easiest/smallest to work with. DO NOT output to DV, since there will be significant loss of detail which is inherent to the DV format. JVC is saying that due to color resolution differences in HDV and DV, you will be BELOW DV quality.
3 - If you need DV format, just shoot in DV to begin with.

I'll try to shoot a couple frames tonight showing the difference between HD->SD, HD->DV, and straight DV.
 
brianluce said:
JVC says:
" If a DV dub is made from an HDV recording, the extra resolution is lost, and the chroma subsampling effectively becomes 4:1:0. This is less than optimal for DV which records at 4:1:1 (in the U.S.), and less than optimal for DVD which utilizes MPEG2 at 4:2:0.

And I think JVC have a case of sour grapes here. The SOny HDV to DV downconvert does not produce 4:1:0, but proper DV 4:1:1 (in NTSC, 4:2:0 in PAL). When you downsample video, you don't begin by throwing away your chroma resolution like JVC suggests.

However, you will get best DVD results by downconverting HDV to uncompressed in your NLE, no doubt.

Graeme
 
I think they meant that by going dv you go from 4:2:0 HD to 4:1:1 dv and when you encode your DVD you end up trying to go back from 4:1:1 SD to 4:2:0 DVD but really can only get 4:1:0. I think they are saying NTSC DV in general is really bad for DVD and it is best to stay either 4:2:0 or 4:2:2 if you can.
 
kidtexas said:
I don't know if it is a problem in PAL land. My take on it is this:

1 - If you need/want HD, obviously shoot in HD
2 - If you need/want SD, for highest quality shoot in HD, then downsize in your NLE to SD for the DVD/output format. Possible formats that I've toyed with are AIC, uncompressed SD, and DV50. DV50 is probably good enough and is the easiest/smallest to work with. DO NOT output to DV, since there will be significant loss of detail which is inherent to the DV format. JVC is saying that due to color resolution differences in HDV and DV, you will be BELOW DV quality.
3 - If you need DV format, just shoot in DV to begin with.

I'll try to shoot a couple frames tonight showing the difference between HD->SD, HD->DV, and straight DV.

this post really confuses me. i thought the point was NOT to convert to SD in the nle but to stay hdv all the way until you encode to dvd.
the other thing is how can you convert a a 4.2.0 hdv signal into 4.2.2 dvcpro50?

hope you get the frames up kid.

btw, my ultimate goal is delivering the good quality SD.
 
brianluce said:
this post really confuses me. i thought the point was NOT to convert to SD in the nle but to stay hdv all the way until you encode to dvd.
the other thing is how can you convert a a 4.2.0 hdv signal into 4.2.2 dvcpro50?

Exactly. When you encode to DVD, you are converting to SD... either in your NLE or DVD program. My frame of reference is FCP, so you edit in HD and then convert on the way to the DVD studio pro.

It might be more manageable to edit in DVCPRO50 or DVCPRO HD, then send that to your dvd program. In FCP, you can achieve this by simply making a new sequence and setting it's encoding to the desired codec. Drag your original footage into the sequence and FCP goes to town. I'm sure there is a quality loss from going HDV -> DVCPRO50/HD -> DVD, but when it comes down to it, for smaller projects destined for DVD, is it noticeable.

Mind you, there is nothing from keeping you from just changing the sequence setting to another codec after the edit. I believe this setting only affects rendered video - FCP works 4:4:4 internally.

My plan on implementing this (until FCP can handle 720p24 more appropriately) is to convert the HDV footage to AIC for the edit. At the end of editing, export to via Compressor to DVD and whatever HD codec I feel is appropriate.

The take home point is that if you are starting out with HD and ending up with SD, don't pass by DV in between.
 
Thomas Smet said:
I think they meant that by going dv you go from 4:2:0 HD to 4:1:1 dv and when you encode your DVD you end up trying to go back from 4:1:1 SD to 4:2:0 DVD but really can only get 4:1:0. I think they are saying NTSC DV in general is really bad for DVD and it is best to stay either 4:2:0 or 4:2:2 if you can.

That could well be what they mean, but that's not what they actually say, at least the way that I read it. But properly smoothed or upsampled 4:1:1 DV chroma converts to 4:2:0 extremely well....

Graeme
 
All conversions were done in FCP. Sorry for the light difference between the HD originated shots and the DV shot. The cam is more sensitive in DV and I didn't adjust the iris right. I think you still get the idea though.

I see I didn't stretch them to compensate for anamorphic. Oh well.

Ok, I probably screwed this up, but here are the images:

HD -> SD uncompressed - 8bit 4:2:2
uncomp.png


HD -> SD Apple Intermediate Codec
aic.png


HD -> DV
hd2dv.png


DV
dv.png
 
Last edited:
That doesn't look like a color sampling error, that just looks like bad de-interlacing. You should be able to get far better downconversion than that, even going HD->DV in-camera on an FX1 or HVX would deliver results better than that.
 
Looks more like FCP "preview quality" DV. Put it in quicktime player and turn on HQ playback, make sure de-interlacing is of. It should NOT look like that at all.

Graeme
 
Cool. I thought something was wrong since it looked even worse than most DV I taken captures from. The above pics have been updated.

Sorry for the confusion.
 
Last edited:
Graeme_Nattress said:
That could well be what they mean, but that's not what they actually say, at least the way that I read it.
Graeme

thanks Graeme, i read that a bunch of times and found the choice of words confusing.
 
For those with Photoshop, grab one of the first 2 images and one of the 2nd 2 images. Bring them both in, convert them to LAB mode, and then look at the AB channels (turn off the L channel).

You can clearly see the difference in color resolution there.
 
kidtexas said:
Cool. I thought something was wrong since it looked even worse than most DV I taken captures from. The above pics have been updated.

I didn't read this far and was looking at the pics above wondering what the hell Graeme and Barry were talking about:Drogar-BigGrin(DBG)
 
Back
Top