Hi-Vision, MUSE, and the Optical Disc

Genesis of Hi-Vision

After the successful colour television broadcast of the 1964 Tokyo Olympiad, NHK [Nippon Hoso Kyokai, the Japanese state broadcaster] began to examine what the next likely development of television would be. Their research suggested three major alterations would increase the value of television to the viewer:

1. A wider aspect ratio.
2. An increase in the total resolution of the picture.
3. High-fidelity multichannel sound.

Other modifications, such as 3-D picture and interactivity, were examined and judged to be of little or no widespread interest. Various other facts came out of the research — notably that only part of the increased resolution would go to making the picture visually sharper, as viewers would tend to reduce their viewing distances [in terms of screen heights] to increase the visual angle subtended by the image.

Accordingly, in 1970 NHK opened discussions with Japanese electronics manufacturers concerning the development of the next-generation television system. Work on what became known as Hi-Vision advanced rapidly enough that in 1979 Matsushita Electric was able to offer a closed-circuit system scanning 30 frames per second, interlaced 2:1, with a resolution of 1125 scan lines [1035 visible] and a video bandwidth of 30 MHz. These specifications remained constant for the Hi-Vision studio format. There were, however, two important differences between this early implementation and the final product. Firstly, the picture aspect ratio was 5:3, a screen shape for motion pictures more common in Japan and Europe than the United States; this was eventually supplanted by the 16:9 ratio proposed by Sarnoff Laboratories, striking a geometric mean between Academy and Cinemascope formats while approaching closely the 11:6 VistaVision matte. Secondly, what is more to our purpose, it employed a full-bandwidth composite format with the colour subcarrier at 24.3 MHz.

For broadcasting purposes, a baseband signal more than seven times the bandwidth of NTSC posed severe problems of channel allocation even before audio transmission was considered. Simply duplicating the existing service coverage might absorb all the spectrum available for television, and the immense investment in conventional equipment militated for the retention of NTSC broadcasts into the forseeable future. Allocating new spectrum for terrestrial broadcasts or supplying every household with a separate CATV service were awkward options at best. About this time, however, the advance of communications techniques opened the possibility of direct-broadcast satellite television service to the home. Studies showed that the whole of the Home Islands could be blanketed by a single transponder, and this avenue was selected for delivery of the new service. In practice, part of the Republic of Korea is covered too, much to the annoyance of the authorities there, who do not allow television broadcasts after 7 pm and have little fondness for the Japanese anyway.

Problems remained. Chiefly, BS series orbiters were limited to 100 watts broadcast power and 27 MHz transmission bandwidth per transponder. Furthermore, the frequency modulation technique used for satellite transmissions was ill-suited to frequency-interleaved composite video signals. In Great Britain at the same time, Doctors Wyndham and Lucas were solving the very same problem for conventional PAL transmissions. Their MAC [Multiplexed Analog Component] system separated the colour subcarrier from the luminance signal, concatenating and time-compressing the two components to fit in the ordinary line scan period. Dispensing with frequency-interleave required, of course, additional bandwidth and extensive processing at transmitter and reciever, but the technique gave good results. NHK engineers adopted this solution to one aspect of their transmission difficulties, but they were not out of the woods just yet. Three 30 MHz picture components would require, of course, 90 MHz of bandwidth. Filtering the two chroma components to half the luminance bandwidth still required 60; vertical subsampling, transmitting the two colour-difference terms in alternate lines, would occupy 45. This was still far too much bandwidth for a transponder which ordinarily carried two NTSC channels at better than broadcast quality. Having already made the commitment to radical signal-processing steps at the head end and the home, the interested parties decided to go for broke.

The MUSE System

The final delivery format is a masterwork of signal processing. Known as MUSE, it is a complete audio-visual delivery format furnishing a creditable high-definition picture and multichannel digital sound within a comparatively narrow bandwidth. Its leading characterstic is the Multiple Sub-Nyquist Encoding which forms the heart of the video processing, and supplies the acronym. As the name will imply to the technically-minded, the video signal is broken up by discrete time sampling into frequency bands which are manipulated separately. In static scenes field- and frame-offset subsampling converts high-frequency information into aliasing components with a four-field structure, and full-frame pans are compensated by encoding motion vectors. Moving objects are low-pass filtered and line-offset subsampled, while colour components are low-pass filtered, vertically alternated, and time compressed. The subsampling pattern constrains the number of active video lines to 1032, the next lower multiple of four to 1035, and to accommodate the processing, the chrominance (C) signal is delayed by one line, and the luminance (Y) signal by five lines. Audio is Differential Pulse Code Modulation encoded and multiplexed into the vertical blanking interval. DANCE [DPCM Audio Nearly instantaneously Compressed and Expanded] A-Mode provides four discrete channels, normally Left, Centre, Right, and Surround, at better than FM radio quality from 15 bit, 32 kHz sampling, along with 128 kbps of independent data; B-Mode supplies Left and Right with a 48 kHz sampling frequency and 16-bit depth, and a 112 kbps data stream. An ingenious set of synchronising signals effectively prevents resampling error at the decoder to ensure the integrity of the reconstructed picture.

The master clock frequency is 48.6 MHz, and the baseband MUSE signal occupies 8.1 MHz. This permits its transmission by wideband FM within the 24 or 27 MHz channels (depending on ITU-R region) used in the 12 GHz satellite broadcast band. The output signal has a bandwidth of approximately 20 MHz luminance for luminance, and 7 MHz chrominance, in static areas of the image ; for moving areas of the picture these figures fall to 14 and 3.5, the loss of resolution being largely masked by blurring in the camera and the eye. These figures are hardware-dependent, and a new encoder and decoder introduced in 1996 provide improved performance without loss of compatibility. Approximately 640 horizontal elements per picture height are available, or a picture element with dimensions roughly 1.5:1, comparable to what is provided by VHS tape or anamorphic 16:9 NTSC LaserDisc. The absence from the signal of temporal aliasing components below 4.05 MHz simplifies the task of motion detection and enables a very simple high-quality NTSC downconversion.

The above statements refer specifically to MUSE-E, the standard first transmitted by satellite in December 1988, which was in operation eight hours per day by November 1991, and was demonstrated in the United States and Italy via terrestrial microwave link and VSB-AM over two adjacent 6 MHz UHF channels. There are additional variants of the system. MUSE-T, for studio purposes, employs only intrafield subsampling and requires a bandwidth of 16.2 MHz, but is said to have "no perceptible impairment" with respect to master-tape picture quality. Several related "Narrow MUSE" systems were proposed for terrestrial broadcast in the United States, using a single 6 MHz channel or a non-adjacent 3 MHz augmentation channel.

Of considerably greater interest is a non-broadcast use, the first delivery format for prerecorded high-definition video programming to be offered to home users.

MUSE Hi-Vision LD

It is possible to play back a signal of almost arbitrary bandwidth from an optical disc, either by increasing the speed at which the pits pass the pickup, or by reducing the size of the pits, which requires improving the sensitivity of the optical system. Both Sony [with the HDL-2000] and Sanyo [HVL-BM2000] produced machines for reproducing baseband high-definition signals, at the cost of a short playing time — say 30 minutes for a double-sided CLV disc. For the consumer market, this was not acceptable. The obvious solution was to record a compressed signal which the home user could reproduce through the same decoder used for Hi-Vision broadcasts.

Recording the MUSE-E signal still required extending the frequency response of the disc-pickup system from that used with NTSC (although the absence of sidebands generated by a chroma subcarrier did ease the requirements somewhat). A linear velocity of 14 m/s was obtained by moving the inner playback radius outward to 76 mm for CAV discs, and by increasing the rotational speed at the inner radius of 55 mm to 2700 RPM for CLV discs. The Modulation Transfer Function of the optical pickup was improved by replacing the infrared laser diode used in NTSC players, wavelength 780 nm, with a red laser operating at 670 nm, and increasing the numerical aperture of the pickup lens to 0.55 from 0.5. This made it possible to reduce the track pitch from 1.8 to 1.1 micron, for a playing time of 30 minutes per side in CAV or 60 in CLV, the same as for NTSC discs. The overall result was to allow modulating the MUSE signal on a carrier frequency of 12.5 MHz, with a deviation of ±1.9 MHz. Incidentally, almost all Hi-Vision LD players can also play back NTSC LaserDiscs, with better performance than a standard player due to the improved optical and electronic characteristics.

As the MUSE synchronization signal is in the same amplitude range as the video signal, it is somewhat difficult to discriminate, so a pilot tone of 2.28 MHz (135/2 times the line frequency) was recorded on the disc. Although satisfactory timebase accuracy was possible using this tone to drive a tangent servo, actual consumer players incorporated digital time-base correctors, with frame memory to permit trick play effects on CLV discs. The resulting analog-digital-analog conversion step, placed ahead of the A-D conversion in the decoder, risks introducing signal distortion, even with the resampling information embedded in the MUSE signal. A proposal to solve this problem by transmitting the signal digitally to the decoder appears not to have been taken up.

The MUSE signal, of course, incorporates audio along with video, and signalling information for the Hi-Vision LD player, such as chapter and frame number, was encoded in line 564, reserved for just such purposes in the MUSE encoding standard. On many discs, however, an additional signal is recorded, in the form of a Red Book PCM digital audio EFM track, as used on NTSC and PAL LD. Although the various refinements applied to standard-definition digital-audio LD, such as LD+G subtitles and dts 5.1 surround sound, could be applied to the high-definition format, this line of development appears not to have been pursued.

While the picture quality is generally high, there are difficulties, particularly with film-sourced material. Many of the film-to-video transfers were made with what were not even the best techniques of the time, for example, the use of a projector-camera "film chain" rather than a flying-spot scanner, and are (by today's standards) mediocre. Also, since conventional 24fps film transfers employ a five-field sequence, abacd, and MUSE is based upon a four-field abcd sequence, disturbances in the encoding may result. It is possible to flag an individual field as containing no motion from the previous field, with obvious advantages for film-source material, but whether this was ever done is not clear. Progress in encoding could be expected to improve the quality of later releases somewhat.

One particular issue which many users have reported is a green tint to the image. As MUSE employs an elaborate system of preemphasis and transmission gamma to improve noise performance, with different values for the luminance and chrominance signals, distortions in the reproducing equipment could cause shifts in relative amplitude, which would affect color reproduction. For example, a uniform attenuation of the Y signal by 5% with respect to the C signal would result in reductions of the decoded R, G, and B values by 2%, 1%, and 5% respectively, based on a sample of 12 representative colours. Such shifts could be caused by resampling jitter, or by nonlinearities in the analog electronics, which agrees with the observation that the tint is equipment-dependent.

Unfortunately for the video enthusiast, MUSE LD was not a commercially successful format. Players and software were introduced at prohibitively high prices, media quality was as noted not necessarily stellar, and releases were sparse. Only about 100 titles, mainly a random selection of American movies, were released, the last in 1998. Market penetration was slight, even counting the bizarre package deal sold as a virtual aquarium. Player production continued for another four years to satisfy demand for the Ultimate LaserDisc Player in the form of Pioneer's immensely sophisticated HLD-X9, but that has ceased now. With the changeover to digital satellite broadcasting in Japan, and the introduction of prerecorded 12cm optical media supporting the 1080×1920 pixel matrix digital high-definition format, it would seem that MUSE in all its incarnations is definitively obsolete.