Compressed Digital Data		Uncompressed Digital Data	Uncompressed Analog Data
Path	FireWire		HD-SDI	Analog Component
Process				A/D Converter
Form	IBP-frame & MP2		Uncompressed Digital Data
Process		Decode
Audio		Digital 48kHz
Video		Digital YUV
Process		4:2:0 (only MPEG-2) or 4:2:0 to 4:2:2
No Codec		Uncompressed
MPEG-2		I-frame
MPEG-2		IBP-frame (ready for DVD)
DVC		DVCPRO HD (downsamples resolution)
JPEG		PhotoJPEG (draft only)
CineForm		Wavelet
Canopus		HQ HD
Avid		DNxHD
Bits	8-bit	8-bit or 10-bits (not MPEG-2)
Sampling	4:2:0	4:2:0 (only MPEG-2) or 4:2:2:
Type	Native	Intermediate
Storage	Disk File

Note 1: No matter the connection path, recorded HDV is 8-bit, 4:2:0 sampled data.

Note 2: Green indicates highest quality; Blue indicates high quality; Orange indicates moderate quality; Red indicates low quality. Yellow indicates a direct path.

As the chart indicates, there are many intermediate codecs available. None of them increase the quality of HDV — in fact, with the exception of uncompressed video, all require an additional compression step after HDV has been decoded. Moreover, intermediate codecs (with the exception of IBP-frame MPEG-2) are less storage-efficient than is HDV.

While some consider interpolating 4:2:0 sampling to 4:2:2 sampling an increase in quality, this conversion can also be performed by an NLE when HDV special effects are rendered. Likewise, placing 8-bit data within a 10-bit word does not increase image quality.

Using an intermediate codec, however, does offer four advantages:

Ø During the transcode, it is possible for bad frames and broken timecode to be repaired.
Ø When HDV, an interframe codec, is transcoded to I-frame MPEG-2, the intraframe format can be decoded efficiently, thereby improving performance. Apple’s AIC codec is an example of an I-frame MPEG-2 format.
Ø All intraframe, intermediate codecs can be decompressed with great efficiency. This enables more streams of video to be mixed together. However, because each frame requires greater data, total disk throughput will be much larger than what an equal number of HDV streams would require.
Ø Many projects require video be processed sequentially by several applications. With the exception of PhotoJPEG, intraframe, intermediate formats are more robust than HDV. Obviously, uncompressed video is the least fragile of these codecs. However, Avid, Canopus, and CineForm codecs are all ideal for tasks that require multiple compression cycles.

When a frame of video is needed by an NLE’s display or effects engine, the appropriate codec is automatically employed to decompress an intermediate format or decode MPEG-2 to uncompressed YUV data. (If the source is already uncompressed, it is used directly.) There is a little-mentioned process involved in generating the YUV digital video. HDV’s 4:2:0 sampling is interpolated to 4:2:2 or 4:4:4. For an NLE that uses the YUV color space, this process occurs after the decompression or decode. For an NLE that uses the RGB color space, this process occurs during the conversion from YUV to RGB. Because HDV employs symmetric color sampling, interpolation is straightforward.

For display on a computer monitor, YUV is converted to RGB. However, if your computer has a board that outputs analog component or HD-SDI, the data are not converted, thus remaining in YUV colorspace.

When multiple streams of video are combined (mixed) by a special effects engine, the digital video streams, either RGB or YUV, are processed as either 4:2:2 or 4:4:4 video. The output is, of course, also 4:2:2 or 4:4:4 video. This video, either RGB or YUV, is then processed appropriately for internal and/or external display.

Depending on the number of streams, the nature of the codec(s) employed, the number and complexity of the effects applied, and both the computing power and disk throughput of your computer, the output of the display engine or effects engine may be previewed:

Ø only after some, or all, of the effects have been rendered.

Ø at less than full frame rate
Ø at near full frame rate
Ø at full frame rate

These performance differences must considered when:

A) The editor plays back the timeline
B) The timeline is exported to tape

When exporting HDV, realtime export is not possible. However, using today’s powerful computers, realtime timeline playback is sometimes achievable.

Preview render files are the key to full-frame-rate playback when realtime previews are either not possible or when preview playback performance is poor. Preview files can be generated three ways:

1. Rendered results are compressed and stored in a file. This is the optimal codec option because DVC-based codecs require have moderate computation and disk requirements.

2. Rendered results are simply stored in a file. Obviously, this option requires no computation and so is very fast. Unfortunately, with HD, both disk throughput and storage requirements are huge.

3. Rendered results are encoded and stored in a file. Native MPEG-2 encoding is compute-intensive, although both disk throughput and storage requirements are minimal.

When you specify a Sequence codec, you are specifying the codec that will be used when a preview file is generated. This codec is typically also the default export codec.

Tip 7: Liquid supports the options 2 and 3 for preview files: previews using the “2vuy” uncompressed, 4:2:2 codec and previews using the MP@H-14 HDV codec.

So does it make a difference whether a native or an intermediate codec is employed? The answer is “it can, but need not.” To understand this answer, we must consider is how an NLE utilizes preview files. Here again, there are three options.

1. A preview file is decompressed/decoded and played back at full frame rate.

2. A preview file is decompressed/decoded and processed by additional special effects.

3. A preview file is decompressed/decoded and mixed, using special effects, with additional layers of video.

NLEs such as Liquid and FCP never utilize the latter two options. Preview files are never used for anything except playback.

Native HDV preview files — having gone through two cycles of encoding and decoding — will suffer more than will high-quality intermediate codecs. An uncompressed preview file — assuming your system can handle it — offers little loss in quality as there has been only one MPEG-2 encode and decode cycle. The other high-quality intermediate codecs have undergone one MPEG-2 encode/decode cycle and one compression/decompression cycle.

Obviously, those NLEs that (to save render time) use preview files for further computation (options 2 and 3) definitely are sensitive to whether preview files are native or not. The use of a native codec can result in a severe loss in quality. However, by deleting preview files whenever you need to view a Sequence at maximum quality, the NLE will generate new frames from the source(s).

Exactly the same logic applies to exporting a Sequence. FCP and Liquid never use preview files during export. Both NLEs generate every new frame from scratch. Moreover, those NLEs that default to using preview files during export can be forced to not do so.

With FCP and Liquid, when one uses realtime previews, there is no quality difference between native and intermediate editing. However, when you need to render previews, the use of an intermediate codec does offer slightly greater preview quality.

With other NLEs, performance is the primary advantage of editing using an intermediate codec. Performance is increased in two areas. First, given adequate disk throughput and an effects engine engineered for the intermediate codec, more streams can be handled in realtime. Second, preview files, which took time to render, can be reused. And, when these files are reused, quality is minimally impacted.

Now that we understand how multiple SD and HD formats are processed within a Timeline, in the next installment I’ll show you how Liquid can automatically handle different aspect-ratio video in the same Sequence.