> It's the how it worked I'm trying to remember. The subcarrier feed back to the deck somehow neutralized the time base error of the chroma decode or some such thing. If you can fill in the blanks in my memory maybe I can let this go.
OK:
I'll try. The color-under system works by separating the color information from the luminance information. The process starts by filtering the 3.58MHz color info from the incoming composite video signal during recording. A simple high-pass filter would chop off and lose some of the higher-frequency luminance information, and would also include some luminance in the chroma, thereby corrupting the chroma information. So a better way to separate the two is by using a COMB FILTER, a special type of filter that uses the horizontal line frequency to 'comb out' the chroma from the luma. It's not a perfect separation, but it's better than a high-pass filter.
The luma is recorded on the tape as a frequency-modulated carrier starting at 3.8MHz, going up to about 5.4MHz. The filtered chroma information is down-converted from 3.58MHz to 688KHz and mixed with the luma's 3.8MHz FM carrier. This combined signal is what's recorded onto the tape. It's called 'color-under' because the color information is placed under the luminance information in frequency (Normal composite video contains the color info as quadrature amplitude-modulated sidebands of the 3.58MHz color subcarrier, which means that the color info is above the luminance info).
When you filter, down-convert, and separately modulate the luma and chroma (using two different modulation types), the two components can no longer be in any kind of timing relationship during playback. Normally with composite video, luma and chroma have a close phase relationship since the color subcarrier is in phase with the horizontal timing pulses of the video. Separating the two components destroys that relationship. This is necessary in order to allow color recording on 3/4-inch tape at 3 3/4-inches per second. But without a timing relationship between the two, they must be treated separately during playback if we hope to get quality results in a broadcast environment using a digital Time-Base Corrector or TBC. We want to avoid repeatedly separating and recombining the luma and chroma, since each time it's done, that process damages the integrity of the information. Usually during playback, a typical U-Matic VCR would simply up-convert the 688KHz chroma to an internally-generated 3.58MHz carrier and mix it together with the demodulated luma, producing a (non-coherent) composite video signal. This signal contains all the information, but the luma and chroma no longer have a locked timing relationship. The luma has its timing errors due to tape's mechanical instabilities and the chroma has its own instabilities due to its derivation from an up-converted 3.58MHz reference with no timing relationship to the luma.
If you simply take this non-coherent composite video and feed it into a digital TBC designed for this type of signal, the TBC will once again have to separate the luma and chroma (hopefully using a comb filter) in order to treat the two with separate timing error correction. Upon correction, the two will be combined together AGAIN to make composite video, but this output will be coherent. However, due to the additional processing, there will be still more luma detail loss and some of that detail will be in the chroma, causing ringing and false color edges, and possibly some chroma misalignment with the luma too. But at least the video itself will be a stable coherent signal.
Instead, if you use a BVU U-Matic VCR equipped with the special 'direct' cable connector and couple it with a suitable TBC also equipped with the same connector, the TBC can send a timing correction error derived from the luminance BACK to the VCR as a 'jittery' 3.58MHz reference, thereby counteracting the difference between the luma's and chroma's timing error. This effectively makes the two separate signals coherent again, which allows the TBC to process the two AS ONE composite signal (as it originally was) and avoids additional comb filtering.
How's that?
Dennis Degan, Video Editor-Consultant-Knowledge Bank
NBC Today Show, New York
