The audio industry has been waiting for it for more than five years. The TV
broadcast industry has wanted it much longer than that. It's been available
in Japan for nearly six years and in West Germany for nearly four. "It" is
stereo TV, and by the time you read this, TV broadcast stations will be getting
ready to go on the air with this new audio service, and TV receiver manufacturers
(not to mention leading audio manufacturers) will be feverishly gearing up
to produce the hardware needed to receive stereo transmissions.
While there's no point in rehashing the reasons that stereo TV has been so long in coming to the United States, it is important to note that we in the U.S., having finally arrived at a decision and selected a system, will more than likely benefit from the long delay. As a result of the long deliberations which led to the selection of a system, we will probably end up with the highest quality of audio possible within the present NTSC broadcast standards. Furthermore, unlike the Japanese or the German TV systems, we will be able to enjoy simultaneous stereo and bilingual or secondary audio programming. Bilingual soundtracks may well prove to be as important as, or even more important than, stereo sound for TV. Many regions of the United States are populated by large ethnic minorities whose first language is not English, and these viewer/listeners will be able to enjoy their favorite TV programs accompanied by soundtracks in their native languages. This will enlarge the stations' audiences, raising their ad revenues (especially if the commercials are also bilingual). Stations expect to make money on this, but not to make much on stereo. Luckily, both stereo and a second language can be transmitted at once; in the Japanese multi-channel TV audio system, the broadcaster must choose either a stereo soundtrack or a second language soundtrack.
Last December 22, various segments of the video industry assembled in Washington to vote for one of three multi-channel audio transmission systems that had been undergoing tests for several years. One of the major decisions to come out of the initial series of tests was that there was an obvious need for noise reduction or "companding." Experiments with "compatible companding" revealed that some companding could be tolerated by people owning older TV sets not equipped with the necessary expander circuits, but that it was difficult to devise an effective companding system without altering sound balance for those owning the older, mono TV sets. So, after a great deal of deliberation, it was decided that companding should be applied only to the stereo difference (L-R) channels and not to the mono (L + R) sum channel, which the mono listener would continue to receive as before.
A separate committee task force was set up to evaluate several companding systems. This task force conducted extensive subjective listening tests, using experienced listeners--ranging from recording and broadcast engineers to musicians and audio critics and journalists-who where asked to make a large number of "forced choices" in blind A-B comparisons between pairs of companding systems.
The tests were conducted using noise and interference levels simulating those which would occur with each of the three transmission systems in close-in (strong signal) and suburban or fringe area (weaker signal) locations. It was agreed that noise reduction would be even more important in the case of the Second Audio Program (S.A.P.) channel, the one used to transmit a second-language audio track or a secondary audio program. (As we will show shortly, this second audio program channel is normally noisier than the stereo channel). The action of the various proposed companding systems was judged by the auditors for the S.A.P. channel as well. In addition to these subjective tests, extensive laboratory measurements were made to evaluate the performance of each of the proposed noise-reduction systems.
Fig. 1--Base-band frequency allocations of Zenith multi-channel TV sound
system; fH is the TV horizontal-line frequency (15,734.26 Hz).
Table I--Signal specifications for multi-channel TV sound.
Notes fH = 15,734.26 Hz; DSB = double sideband; SC = suppressed carrier; FSK = frequency-shift keying.
Fig. 2--Encoder for Zenith multi-channel TV sound system.
Zenith Transmission System
After two days of presentations by all proponents, the vote was finally taken.
The winning transmission-system proponent was Zenith, while the winning companding-system proponent was dbx. Figure 1 shows the spectrum occupancy and modulation standards of the chosen transmission system. The main-channel modulation consists of an L + R audio signal. An L R stereo difference audio signal causes double-sideband, suppressed-carrier amplitude modulation of a subcarrier at twice the horizontal-line frequency.
Audio bandwidth of each signal extends to 15 kHz, and the main channel pre-emphasis remains as it has been in the past, 75 µS. Pre-emphasis of the stereo subchannel is a part of the companding system, which will be described shortly.
The combined peak deviation of the main channel and stereophonic sub channel is always 50 kHz, with the main channel accounting for 25 kHz of this. When the L and R channels are statistically independent (as will usually be the case), the main and subchannel signals interleave, so peak deviation due to the stereo subchannel can also be up to 50 kHz; this helps keep S/N from falling as low in stereo as it otherwise might. When the L and R signals are not statistically independent (which will be most true as the signals approach mono), or when the L + R and L-R do not have matched pre-emphasis characteristics, the relative levels of L + R and L R components assume their respective, natural levels, as dictated by the acoustic scene.
A stereo pilot signal is also transmitted, as a continuous-wave frequency of 15,734.26 Hz (the TV horizontal-line rate) with a main-carrier deviation of 5.0 kHz. The subcarrier for the S.A.P. channel is at five times the horizontal line frequency (or 78.67 kHz). The S.A.P. channel is frequency-modulated to a peak deviation of 10 kHz by a signal that is band-limited to 10 kHz; when not modulated, it is locked to 78.67 kHz. The pre-emphasis on the S.A.P. channel is, again, part of the chosen companding system. Main carrier deviation due to this subcarrier is 15 kHz. The S.A.P. channel is noisier than the main-channel or stereo-subchannel audio, due both to its own low level of modulation and its low deviation of the main carrier.
A third subcarrier, known as the Professional Subchannel and intended for voice or data transmission, is located at approximately 6.5 times the horizontal-line frequency. This last subcarrier causes 3-kHz deviation of the main carrier.
Figure 2 is a simplified block diagram of the encoder required at a transmitter to broadcast the new system, while Fig. 3 shows a basic block diagram of the elements of a decoder circuit for multi-channel-sound TV receivers or tuners. It is expected that appropriate ICs will be available for both the basic decoder and corn pander.
The companding circuits are not shown in these diagrams. In the encoder, a compression circuit would go in the L-R line between the stereo multiplexer and pilot adder, and another would be inserted just before the S.A.P.'s FM modulator. In the decoder, the L-R signal would be expanded in the stereo decoder, and the S.A.P. would be expanded in its decoder.
dbx Companding System
Although the companding system chosen by the industry for noise reduction bears the dbx name and was proposed by dbx, its operation is more sophisticated than that of the familiar dbx noise-reduction system used in consumer tape recording. The corn pander works in two stages. First, It provides wide-band amplitude compansion to reduce dynamic range in the transmission channel at all audio frequencies. This section utilizes a 1:2:1 linear dB compander, similar to dbx's noise reduction for tape recording. In addition, the compander provides variable pre-emphasis/de-emphasis which adapts itself to the spectral distribution of the program material, to take full advantage of the limited channel-headroom available. The spectral compressor is able to boost or reduce high-frequency levels, depending upon the input signal spectrum.
Rms detectors are used to control both the amplitude and spectral companders, thereby providing minimum sensitivity to impulse noise while maintaining appropriate reaction times for music signals. A clipper is provided within the compressor control loop for preventing channel overload without inducing compressor/expander tracking errors. Band-limiting filters are also included in the compressor design.
Compensation for the phase errors caused by band-limiting throughout the system is provided in the form of a complementary filter in the L + R channel. The compressor design is shown in block diagram form in Fig. 4, while the expander block diagram is shown in Fig. 5.
How Good Is the Chosen System?
As anyone who has switched from stereo FM to mono reception of the same FM signal knows, unless you are in a strong signal-reception environment, stereo FM is a lot noisier than mono. The full impact of this signal-to-noise deterioration is especially severe when you are listening to a station whose transmitter is many miles away.
Unfortunately, when stereo FM broadcasting was approved back in 1961, noise-reduction systems such as Dolby, dbx and the like had not been invented. Happily, as we enter the stereo TV era, we have an excellent noise reduction system built into the new system to take care of the signal-to-noise deterioration that would otherwise have occurred as we switch from mono to stereo TV sound (or to the Secondary Audio Program channel, be it bilingual service or an entirely different audio program). The following data was extracted from the many, many pages of data in the report submitted to the FCC by the Electronic Industries Association Multichannel TV Sound Committee to support the industry recommendation. In "City Grade" reception tests, the chosen systems yielded S/N ratios between 65 and 68 dB for stereo reception, with a split-sound type of receiver, while the S.A.P. channel, using the same type of receiver, yielded S/N ratios between 78 and 79 dB. The real advantage of companding showed up more definitively when tests were conducted for "Grade B" signal-reception conditions. Such conditions are represented by a video carrier-to-noise ratio of only 30 dB, as might be expected in outlying areas served by a TV station. Again, using a split-sound receiver, signal-to-noise ratio of the Zenith system, without companding, was just over 50 dB in stereo.
With dbx companding added, the signal-to-noise ratio increased to between 63 and 64 dB. The improvement was far more dramatic in the case of the normally noisier S.A.P. channel. With no companding, S/N measured a very noisy 43 dB. When dbx companding was added, S/N improved to a remarkable 77 or 78 dB! With intercarrier types of receivers (those that do not have separate video and audio i.f. circuits), S.A.P. signal-to-noise without companding in a Grade B signal environment was even poorer, between 36 and 42 dB. With the chosen dbx companding system added, S/N improved to between 62 and 63 dB, still an acceptably low level of background noise.
Fig. 3--Decoder for Zenith multi-channel TV sound system.
Once stereo TV transmissions begin, I expect that we'll see a number of new product categories appearing in both [...]
(adapted from Audio magazine, June 1984)
= = = =