I did this the other day Graeme. The render times are ridiculous, and the storage requirements are huge, but it looks phenomenal!

Workflow to uncompressed 480p60:
- Capture Premiere Pro 1.5.1 to Cineform intermediate
- Edit
- Render to 1440x1080i uncompressed avi (because I can't import Cineform* to AE yet)
- Import 1080i uncompressed into After Effects*
- 1440x1080, PA 1.333, 59.94 fps timeline, deinterlace upper field first best quality
- color correct (if need be), add effects, etc. in 16-bit colour
- render to 720x480, 59.94 fps, PA 1.2 uncompressed.

*note: If I had AspectHD, this would be a lossless copy-paste operation into After Effects, omitting all intermediate renders and using source files directly. Since I can't do this, I am limited to using an uncompressed intermediate render.

I will be rendering all DVDs I author by this method. The uncompressed 480p60 or 480i looks tremendously better than any DV go-between, as there is no additional compression or colour space reduction. Unless shot on a DV camera, I will never touch DV footage again. All my SD will come from down-conversions.

-Steve

Steven, that really should theoretically produce excellent results, and it's great to hear that it's doing it in reality too.

Graeme

Robert,
The Camera conversion is NOT as good as the Vegas conversion, not by a long shot. The camera conversion is better than a PD 170, but not *tremendously* so.
I'm capturing uncompressed from the cam by going from the cam uncompressed out, to a Decklink HD card, bringing that into Vegas and editing there. Or, I'm bringing it in via the component out into the SD Connect from Convergent and sending that in via firewire.
The BEST image, is going into the Decklink, then converting in Vegas to a 4:2:2 SD stream, if that's where I'll end up. What I'm doing for *most* of my work is using our GearShift tool to convert the raw m2t files to DV proxy, editing the DV proxy, and replacing the proxy with the m2t files when I'm done, and either printing those to SD/MPEG, or back to the HDV deck, which will be then used to transfer to HDCAM at another facility.

a ques for Graeme - will fc5 clean up these problems ? I mean , who wants to undergo these workflows- not me . thanks Kurth

Don't know. I don't know anyone who has FCP5 yet who'd be qualified to look at the supposed new scaling in there to see if it does correct filtering or not. I'll be writing about it, no doubt, as soon as I myself know for certain what is going on.

Graeme

See later message.

I agree and disagree. When you do a lossless downconversion in Vegas or After Effects you get a better color conversion, that's for sure. However, none of the conversions I've tried in those programs looked as good as Z1's HDV->DV downconversion when it comes to smoothness of fields rendering (and yes, I have the fields set up in the correct order). Putting the colors aside, the sharpness of the in-camera and software conversion is pretty much the same, however when you do it in software the're a very noticeable flickering on some highly detailed horizontal lines, well, on just about anything that's got a lot of detail, such as wide shots of trees, grass, fences, etc.. This is especially noticeable during vertical panning. The lines are sort of gliding together with your panning. So I don't think it's a very good idea to discard the in-camera downconversion for all kinds of material.

With the software downconversion, while it looks good, but because of the horizontal lines issues, you know it was downconverted. However, with the in-camera downconversion, it looks like in was natively shot in 576i or 480i scanning mode. So, the choice should be made on whether you'd like better color reproduction or better interlaced fields downconversion.

I think the only hardware that could achieve both would be one of the super expensive Snell & Wilcox converters. Sorry, but the regular software's (Sony, Adobe, Canopus, etc.) up- and downconversion of the interlaced material still leaves much to be desired.

If you're seeing interlace twitter, then incorrect or insufficient filtering has been applied in the downconversion. There's no reason why software cannot do this correctly.

Graeme

You're doing something wrong then. I've seen several hardware conversions, and they can't touch what Vegas is doing.

No I don't think I do it wrong. I export from HDV 60i project (M2T) to a normal DV NTSC preset, while choosing a lossless codec (Sony codec for example) instead of DV. Of course I tried DV codec as well, with similar results. After Effects does the same thing. You pretty much have to look for this artifact, otherwise you might not even notice it. Once you do though, it's quite obvious.


If you'd like to check it out yourself, you can actually use the 60i clip you have on Vasst of the surfer walking on the beach. Try converting this M2T from camera and then do the same thing in Vegas or After Effects: pay attention to water waves, you'll see that the software conversion is not as smooth on those rolling waves, you see some lines (basically the very thing everybody's trying to avoid as much as possible these days). I mean it looks very good, but not as good as the in-camera conversion.

And if you try it, please try to be as objective as possible, because those lines are quite visible. Also, I'm not talking about 24p downconversion (which looks good), but a normal 60i.

Douglas, can I clarify one thing in your comment? Are you talking about the PD170 in-camera Widescreen or standard 4:3 for the comparison?

I am moving from PD150 to Z1 next month (when next shipment of hot cakes arrive) for Widescreen SD DV. I know Z1 SD native WS is far better than PD150 in-camera WS, but did not test the PD150 full 4:3 resolution against the Z1 recording DV/DVCAM.

Thanks.

You cannot just scale down an interlaced image and get a interlaced image. Your fields will get blended together. Certain software may handle this correctly by dealing with the fields. The best way to scale down interlaced and end up with interlaced is to...


1. Seperate fields into 60/50 de-interlaced frames.
2. Scale down your 60/50 de-interlaced frames.
3. Combine your frames back into fields by taking every other line from the scaled down frames.


This will give you a perfect scaled down SD interlaced image.

If you are trying to go from 60/50 interlaced HD down to 30p/25p you should de-interlace first. During the scale down filtering if you keep the fields this could cause odd issues with your SD footage by keeping tiny little pieces of fields left over. Cineframe 30/25 do not have this issue since they are already de-interlaced.

The sad thing about doing your HD to SD down converting in the camera is that it gets turned into 4:1:1 DV when sent out the firewire port. This feature is mainly for people who cannot edit HDV yet to be able to deal with the video they shot.

Well, then wouldn't you think that highly praised Vegas or especially After Effects would do some sort of the above described process in background when you do a simple scale down. If not, then what good are they for in processing of interlaced video? You can always de-artifact 4:1:1 DV with Magic Bullet, but if you have obvious lines on your video, that is very hard to fix.

Again, indeed perhaps this difference in the fields rendering is too subtle to see even by such professionals that grace this board. I saw it and I am reporting it here. I did nothing wrong exporting, and there's no reverse fields issue or any other anomaly. It's just that the horizontal details are rendered by popular software not as nicely as by the in-camera HDV->DV conversion. That's all.

Let's put aside the uncompressed (or CineForm) codecs for now (and 4:1:1 color issues) and simply try to render the HDV material with a lot of [u/]high frequency[/u] information:

1. Through the in-camera downconversion to DV
2. Through Vegas/Adobe/Canopus/etc. downconversion to DV

... and see what looks better and more natural. The choice number 1 would win (... yes I know the colors will be better with number 2). If there's any bandwith on this site I can upload those clips. If there's none, I encourage all the pessimists to try it themselves before coming to any conclusions of me doing something wrong. There were quite a few times I've seen professionals doing obvious mistakes, including messing up pulldown and many other things. So it's not surprising that some of you won't admit that the software conversion may be inferior to the in-camera conversion. I'm not trying to flame anyone, but to just objectively compare the results without biasing one's mind towards a preferred solution. Believe me I'd love to use the software downconversion because of the 4:1:1 issue, but I'm not willing to give up a natural looking image for that.

By the way, what's up with Vegas messing up color space everytime I use any other codec, but the one supplied by Vegas? Even exporting to uncompressed AVI results in shrinked color space. But if I use Vegas DV or Sony YUV codec everything is fine. No such problems in Premiere or After Effects...

Thanks.

Ruslan.

I just want to shoot with an FX or Z1 in 1080i, and then do pulldown's in native HDV in FCP5. Anyone know if this can be done? (and still look good, that is...) Do I need a Nattress filter? ;-)

Michael, I'm getting pretty darn good results here. Why not try the demo and see? I'm now testing software to vastly improve 4:2:0 as well as the 4:1:1 that I do now.

Graeme