PAL to NTSC?

[Valentin Fuhrer]

Hi all,

I'm working with 25p PAL DV footage shot on a DVX100 Pal cam.

I'm wondering what is the best way to setup my After Effects and Final Cut projects to maintain optimal quality. I'll be staying 100% in the computer until I generate final out for Master Tape (Beta?), and/or ready for encoding for NTSC DVD. Any tape transfer will be handled by a post house, so they'll be working from whatever Quicktime file I give them.

For AE:
What are the appropriate "interpret footage" settings, and composition settings?
What would be the appropriate output render settings (e.g. if I just render to quicktime uncompressed, is that helpful for the post house to receive the highest quality to master from?)

I guess a fundamental question is, with any DV source as input, if I work in a D1 timeline, and render uncompressed, how does that differ from D1 - or is that all there is to it?... (I understand I'm not starting from D1 source, but to output to it, do I just need to get the settings right (uncompressed quicktime, etc) , and click the make it happen button?)

Thanks so much for any help.... just trying to get this all right :-)

[Juan P. Pertierra]

Using After Effects, here's how I do it and I get great results. The best way is to slow the footage down to 24fps. This is a 4% decrease in speed and it is unoticeable. You can then resize the image to NTSC, and burn a DVD in 24p which looks great. The only other hitch is that in slowing the video 4%, you need to stretch the sound channels 4% as well to keep it in synch.

In After Effects, import your 25P clip. Right click on the clip, select interpet as. Make sure the fields box is in NONE, and set conform to rate 24fps.

Now, drop the clip on a NTSC 720x480 24fps timeline. Select the clip in the timeline and do a transform->scale. A little dialog will pop up. Click on the "Include pixel aspect ratio" check box, and select pixels for the units. Now, change the horizontal width to 720, and click OK.
Your video is now NTSC size, while mantaining the aspect ratio correctly. Now, load the audio into a sound program and do a time stretch to 4%, and re-import the audio into the timeline.

Hope this helps,
Juan

[Valentin Fuhrer]

That recipe does help quite a bit... :-)

I'm wondering, when I do the final render in After Effects, what settings should I use for the output quicktime codec?

For best quality, should I just go uncompressed, and let my mpeg encoder do its thing so that I only lose one generation? (i.e. from AE --> uncompressed quicktime --> mpeg encoder to bring into DVD studio pro ... vs. AE --> sorenson codec (or whatever codec) --> mpeg encoder etc.)

Also can anyone tell me, if I work in an AfterEffects NTSC D1 timeline, and render out uncompressed quicktime, do I end up with D1? If not, how do I get from an AE timeline to a true D1 formatted quicktime file?

I realize there are cards that work with D1 and allow alot of real world advantages (i.e. D1 output to video monitor etc. ), but if I only care about outputting a D1 quicktime file properly from AE , is that necessary?

Thanks again.....

[Juan P. Pertierra]

I am a little shaky on this issue but I will tell you what I know.

The main thing here is that you have loaded a 4:2:0 quantized DV stream into the computer, edited, and now you have to be careful what you encode it to. As far as I know, if you encode it in NTSC DV, you should get 4:1:0 because the NTSC DV codec does 4:1:1 quantizing, so you would get really bad color, but I might be wrong.

The best way to go, is to go uncompressed. My favorite is Quicktime for 24fps, because I think AVI has trouble with it...i mean AVI works fine from a DVX100 but that's just because the DVX output is really just 60i with pull down.

IF you have to go DV, I would select one of the 4:2:2 codecs, like for DVCPRO50. Those should preserve the PAL color just fine.

Remember, that you loaded a compressed stream from your camera into your computer for editing(uncompressed), and even if you output it to a DV codec again, you are recompressing and there will be some loss. This is why i'd rather just go uncompressed if you can wait the render times and have the hard disk space. After that, you can go to whatever format you want.

If you already know what formats you want you can get the same quality by just rendering to those formats...however, if you want to keep a file for future use, i'd use uncompressed for best quality.

[Graeme Nattress]

4:2:0 followed by 4:1:1 does not equal 4:1:0. Colour sampling ratios do not concatenate in such a way.

When you convert PAL to NTSC the PAL DV footage (which is 4:2:0) gets uncompressed, and will probably be held internally in a memory buffer as 4:4:4, and this 4:4:4 signal will get recompressed and sampling set to 4:1:1.

The end result may not be quite as good as native 4:1:1, but it will be no where near as bad as 4:1:0!!!

Out of interest, I'm working on an FCP based Standards Conversion plugin that will do PAL <-> NTSC, and Film <-> PAL and Film <-> NTSC.

Graeme

[Juan P. Pertierra]

Graeme,

Thanks! You have cleared up a doubt i've had for a long time.

I have heard so many people say the 4:2:0->4:1:1=4:1:0 that i thought i just didn't understand it...my doubt was that i never understood, if when you load a stream onto the computer and it gets decompressed, all you get is a series of pixels, why would you loose color? I guess my original reasoning is right. I think the problem would appear if somehow you dubbed a 4:2:0 tape to an NTSC tape right? Still not sure how one would go about doing this.

Since you seem to understand this, i have another question...the DV stream that comes from the camera is decimated to 4:1:1 so it only has color info every three pixels right? So then is it up to the computer to approximate the missing color? I would think that different algorithms for filling in this missing data would yield different results..like there are many ways you could do averaging and get different looking color, yet there are no options in any of the capture programs i've tried to control this...is this just internal to the codec? or maybe there's a standarized way to do this that yields the 'best' results....

Juan

[Graeme Nattress]

Well your problem will never occur because you can't do a direct digital copy via firewire from a 4:2:0 source to a 4:1:1 recorder. Bits might be bits, but in that situation it'd be expecting a standards converter of some kind in the middle.

The only way you can do that kind of dub in real life is via SDI from, say PAL DVCPro to PAL DV which would be a 4:1:1 to a 4:2:0 dub, but when the DVCPro player out puts via SDI it gets ucompressed and the colour gets upsampled to 4:2:2. So the DV deck with SDI input will now downsample the 4:2:2 to 4:2:0 and we won't have lost hardly anything. I've seen some really nice dubs done like this (Quad -> D3 -> DigiBeta -> DVCpro -> DV) and they looked awesome.

Yes, when you decompress DV the computer fills in the gaps in the colour, and it's down to the codec how this happens. Apple's DV codec repeats the colour so in 4:1:1 you get 4 identical coloured pixels in a row. This may sound bad, and it can look bad, but you loose least quality with conversions and generations this way. Avid does some funky colour interpolation, and this looks initially better, but is not as robust. The Apple method is better as it allows me to go in and make my own interpolator.

Many DV colour interpolators just blur the chroma or do a strict interpolation. I've written a full up-sampler that also takes the luminance into account to produce visibly superior results to the Apple 4:2:2 filter! You can also read about that as it's part of my Film Effects package.

Remember, DVD is 4:2:0 and you often see them made from 4:1:1 DV without the colour going all over the place and disappearing!!

[Juan P. Pertierra]

Graeme,

Thanks again! You have no idea how long i've been asking about this topic. I didn't understand why my PAL XL1se footage looked great on DVD when all the messages i looked at said that I would loose color in going from 4:2:0 PAL to 4:1:1 NTSC.

Just as a side note, I am almost done with an experiment I started about a month ago. I am extracting full uncompressed 4:4:4 video from the A/D converter of a digital camera. My final goal is to do it with a decent 3CCD cam, but right now I started with a Sony TRV19. It's a bit more complex because i have to do de-mosaquing, but i'm almost done with it. I got the data, right now i'm working on the algorithm and synching the output to the vertical/horizontal signals.

Juan