differences between QT movie export and QT conversion

We use FCP 5.1. It is DV-PAL project but with lots of effects and titles so we were interested in trying a 10-bit uncompressed export. We did it two ways:
#1: the project was rendered with 10-bit uncompressed chosen in sequence>settings. Then export>QT movie. Made a 123 GB .mov file.
#2: export>using QT conversion. Select 10-bit uncompressed. Made a 123 GB (extensionless) file.
Played both files with QT player using HT1100 projector as reference monitor.
There were obvious differences in color and tone. Faces in #1 showed better modeling. Highlights were stronger. Some saturations were greater. #2 had nothing to recommend it except: some colored titles exported correctly in #2 but became banded in #1.
We'll find a workaround for the banding with #1, but the highlight compression with #2 is fatal.

Postscript: I don't know if the Digibeta tape made from #1 preserved the color and tonality of #1. Anyhow the festival transformed the Digibeta to BetaSP to suit the theater's player. The projector was dim and of course this audience could just as well have been seeing #2 or a mini-DV export.

Dennis Couzin
Berlin, Germany

I don't have a real answer, but can give you some general wisdom to troubleshoot.

The conversion to something else than DV is ONLY worthwhile if you also render your effects to that same codec. Otherwise, you'll just spent bits on something that is already altered by the DV codec. So trash all your renderfiles first if they are in DV.

Next, there 'should' be NO difference in any uncompressed codec. Otherwise, it's not 'uncompressed'. Colors should NOT alter (although you clearly see it happening...)

I don't know how the YUV-RGB handling is done, perhaps you should do another test with different settings. This can be the cause of the difference you're seeing.

Another gotcha (if you work with DV in other applications), the DV codec has a 'high quality' flag. It defaults to 'low', so you have fast playback of clips using QT player.
Some apps don't override this flag, so you end up with horrible quality. You can set it to high in qt Pro and save. (But FCP is smart enough not to make this mistake)

Hth

Bouke

Exporting as a QT movie export without the 1st box (recompress all frames) checked, exports everything "per se" as your sequence settings, and rendered segments rendered to your sequence settings. QT conversion forces a render of everything on your timeline, causing a double compression. However, since you're working in uncompressed, technically there shouldn't be a problem- is your original footage in uncompressed?

It isn't a recommended practice to convert codecs in FCP. Compressor is designed for that, so the better way to do it is to export the clips via compressor- change sequence settings to the format you're exporting in, and export via compressor to the codec you want to be in.

I'm still trying to figure out where the YUV/RGB handling is done, or if the QT uncompressed codec renders colors in YUV... Any word on this?

I did first trash all the render files and re-render with system settings for 10-bit uncompressed 4:2:2.
4:2:2 is already a kind of compression; it's not 4:4:4. If you think about it, the 4:2:2 must be constructed from the 4:2:0 provided by DV PAL scratch file. I think this construction is unambiguous.
But I think that undoing the other aspects of the DV compression require more creativity and thus lead to ambiguity. I think an algorithm applied to the DV compressed image tries figure out what an uncompressing 10 bit (4:2:2) camera would have seen and recorded. Thus two differently made 10-bit uncompressed 4:2:2 images from the same DV compressed image can be different, which is what I saw. So different that side-by-side comparison was unnecessary.
With bum algorithms, it is possible for a 10-bit uncompressed export to look worse than the 8-bit DV compressed original, but I'm not sure my #2 was that bad.
I should also point out that due to a glitch in FCP 5.1 (and FCP 6 too) 10-bit uncompressed PAL exports from DV PAL always get the field dominance wrong.

Dennis Couzin
Berlin, Germany