What is different in a digital videotape machine over an analog? Less than you might expect. You still have a rotary head scanning tape. (All digital VTRs use helical scan, although the first full-frame digital VTR ever built (Code named Annie) was a modified AVR-3.) You still need servos to keep the head drum and capstan running at the right speed. You still need a tape transport to load and move the tape. In fact, the high bandwidth of the digital signal has forced VTR manufacturers back to the precision of the quadruplex days to build a stable enough transport.

The main differences are in the signal system. If analog video is input into the system, it is converted to digital. Four or more channels of analog audio is also converted to digital. In addition, most professional digital formats also have analog longitudinal audio cue and time code tracks to go along with the digital audio. If a format uses video compression, the video is compressed at this time. At this stage, you would think you could just record the data onto the tape, but you can't.

Tiny imperfections in the tape that would produce unnoticeable dropouts in an anal dropouts in an analog system would seriously 'trash' a digital signal. So, the digital video data is encoded using an error correction system, that can replace some lost data completely. This is the Outer Error Correction System. Then the data is Shuffled. Shuffling involves rearranging the data so it is completely out of order. This is done so that a large dropout will cause a data loss that is spread out over a large area of the picture. This way, it will in all likelihood be fixable by the outer error corrector. Finally, the data is encoded by yet another error corrector, the Inner Error Corrector . The inner error corrector will take care of most of the small errors, so that the outer error corrector will only have to deal with large dropouts.

Before recording, the data is encoded into a special code used for digital data recording, and is finally recorded on the tape. Why is this special code needed? As you know, a digital data stream is made up of 1's and 0's. Since direct recording (no bias) is used for digital recording, all that is actually recovered on playback is transitions from one level to the other. If you were to record a 1 followed by ten 0's, there would be no transitions for the entire time the 0's were being transmitted. You would have to have a very stable clock signal to tell the playback electronics where to look for a transition, or lack thereof. Since we know that mechanical jitter is constantly changing the timing of signals being played back from the tape, it is obvious a fixed clock would not work. The clock signal has to be encoded in the data stream, so the playback electronics will know where to look for data transitions.

The playback levels of the signals from off of the tape is very low. A great deal of gain is required in the playback preamps to recover this weak signal. If we wanted to be able to recover long strings of 1's or 0's from the tape, we would have to use DC coupling in the playback preamps to ensure that we knew the exact voltage level at any given moment. This is necessary because AC coupled amplifiers would drift to some voltage during a long string of 1's or 0's and the following electronics would have trouble telling the two logic levels apart. High gain DC coupled amplifiers are very difficult to build and keep in adjustment, whereas high gain AC amplifiers are relatively easy to build. What is needed is a digital code that would not contain a long run of 1's or 0's so it could be amplified by AC amplifiers.

The answer to the above two problems is a family of DC-free codes that can be recovered without reference to a stable external clock. A number of such codes exist, such as Miller-squared, NRZI (Non Return to Zero, Inverted), or EFM(Eight to Fourteen Modulation, the same coding used for compact discs.) All of these coding schemes have been used in digital VTRs.

Because of the wide bandwidth of the digital signal, it is recorded on the tape by several heads at once. The head drum spins much faster in most digital machines than in an analog one, with 6,000 RPM being about average. There are more tracks per scan, and they are narrower than any analog format. The servos have to work very hard in a digital machine to keep data errors down. The data error rate is measured, and a warning generated if it gets too high. Then, the heads are automatically cleaned by a foam roller, and/or the operator is alerted that there is a problem. This usually occurs long before you can see a problem in the picture. Audio data is also recorded by the rotary heads, usually in a special area at the center or ends of the video track. Since audio errors are generally more objectionable than video errors, most digital VTR formats record each audio segment twice. The audio is recorded in separate segments from the video to allow editing of just the video or just the audio, or any combination of both.

On playback, the whole process is reversed. The data is recovered from the tape and run through a digital equalizer. The digital recording code is stripped off. The inner error corrector looks for, and corrects errors. The data is de-shuffled. The outer error corrector corrects any remaining data errors. If there are any errors remaining that are beyond the range of error correction, an error concealment circuit interpolates the missing data from surrounding good data. If necessary, the data is decompressed, and is now once again digital video. It can either be output in digital form, or can be converted back to analog for use by conventional video equipment.

Some formats, such as Digital Betacam, record an analog pilot tone on the tape to help the servos follow the tracks on the tape.

All this digital processing required racks full of circuit boards in the early machines. Due to the rise of consumer digital VTR's, all this signal processing can now be fit on a couple of chips!