(This is actually an older post, from about a wek or so ago, but it was left languishing in “Draft” status. But rather than delete it, here it is, out-of-sequence, for posterity)
Nowadays for video editing I mainly use Adobe CS6. However I have still some old projects edited with Sony Vegas (10) which now have new clients. One such project was shot as HDV on a Z1, giving 1440×1080 interlaced, at 50 fields/second, which I call 50i (it doesn’t really make sense to think of it as 25 fps). The required new deliverable from this is a PAL-SD DVD, 720×5786 50i. In addition, I want to deliver high-quality progressive HD (not V) 1920×1080 progressive.
The PAL-SD frame size of 720×576 has exactly half the width of the HDV source and just over half its height. My naive initial thought was that the simple/cheap way to convert from the HDV source to the SD deliverable would be to merely allow each of the HDV fields to be downscaled to the equivalent SD field. This could be performed in Sony Vegas itself, to produce an SD intermediate file as media asset to Encore to produce a DVD.
Some potential complications (or paranoia) that come to mind in this approach are:
- Levels-changes, through processes associated with the intermediate file. For example it might accidentally be written as 16-235 range and read at 0-255 range. In general, uncertainty can arise over the different conventions of different NLEs and also the different settings/options that can be set for some codecs, sometimes independently for write and for read.
- HD (Rec 709) to SD (Rec 601) conversion: I think Vegas operates only in terms of RGB levels, the 601/709 issue is only relevant to the codec stage, where codec metadata defines how a given data should be encoded/decoded. The codec I intend to use is GoPro-Cineform, with consistent write/encode and read/decode settings. Provided Vegas and Encore respect those, there should be no issue. But there is the worry that either of these applications might impose their own “rules of thumb”, e.g. that small frames (like 720×576) should be interpreted as 601, overriding the codec’s other settings.
- Interlace field order. HDV is UFF, whereas SD 50i (PAL) is LFF. Attention is needed to ensure the field order does not get swapped, as this would give an impression of juddery motion.
So I did some experiments…
- Vegas (1) Project Settings:
- Frame Size: 720×576
- Field Order: LFF
- PAR: 1.4568
- Render Settings:
- Frame Size: (as Project)
- Field order: LFF (I think the default might have been something else)
- PAR: 1.4568
- Video Format: Cineform Codec
What Worked:
- Sony Vegas (v.10) project for PAL-SD Wide, video levels adjusted to full-range (0-255) via Vegas’s Levels FX, then encoded to GoPro-Cineform.
- Just as a test, this was initially read into an Adobe Premiere project, set for PAL-SD-Wide. There, Premiere’s Reference Monitor’s YC Waveform revealed the levels range as 0.3 to 1 volts, which corresponds to NTSC’s 0-100% IRE on the 16-235 scale. No levels-clipping was observed.
- So using the 0-255 levels in Vegas was the right thing to do in this instance.
- The Configure Cineform Codec panel in Sony Vegas (v.10) was quite simple, offering no distinction between encode and decode, allowing only for various Quality levels and for the Encoded Format to be YUV or RGB. The latter was found to have no effect on the levels seen by Premiere, it only affected the file-size, YUV being half the size of RGB. Very simple – I like that!
- In Premiere, stepping forwards by frame manually, the movements looked smooth.
In Adobe Encore (DVD-Maker) CS6:
- Imported the intermediate file as an Asset and appended it to the existing main timeline.
- Encore by default assumed it was square-pixels. Fixed that as follows:
- [theClip >RtClk> Interpret Footage] to selrct the nearest equivalent to what I wanted: [Conform to SD PAL Widescreen (1.4587)].
- Why does Encore’s [1.4587] differ from Vegas’s [1.4568] ? Any consequence on my result?
- [theClip >RtClk> Interpret Footage] to selrct the nearest equivalent to what I wanted: [Conform to SD PAL Widescreen (1.4587)].
- Generated a “Virtual DVD” to a folder.
- Played that “Virtual DVD” using Corel WinDVD
- In a previous experiment, involving a badly-produced DVD having swapped field-order, I found this (unlike WMP or VLC) reproduced the juddering effect I had seen on a proper TV-attached DVD player. So WinDVD is a good test.
- Made a physical DVD via Encore.
- The physical DVD played correctly on TV (no judder).
An alternative would be to deinterlace the original 50i to produce an intermediate file at 50p, ideally using best-quality motion/pixel based methods to estimate the “missing” lines in each of the original fields. But would the difference from this more sophisticated approach be noticeable?
There also exists an AviSynth script for HD to SD conversion (and maybe HDV to SD also?).
- It is called HD2SD, and I report my use of it elsewhere in this blog. I found it not to be useful, producing a blurry result in comparison to that of Sony Vegas ‘s scaling (bicubic).