WO1986004201A1

WO1986004201A1 - Broadcast stereo companding system and apparatus

Info

Publication number: WO1986004201A1
Application number: PCT/US1985/002479
Authority: WO
Inventors: Aldo G. Cugnini; Daniel W. Gravereaux; David W. Stebbings
Original assignee: Cbs Inc.
Priority date: 1985-01-04
Filing date: 1985-12-13
Publication date: 1986-07-17
Also published as: NO169268B; DK164681B; EP0207964B1; KR870700188A; DK234086A; KR900006261B1; AU5306986A; EP0207964A1; NO862139L; BR8507150A; DK164681C; ATE56114T1; EP0207964A4; NO169268C; AU578792B2; DK234086D0; DE3579460D1

Abstract

A stereophonic broadcasting system incorporating companding of the difference signal in which both the usual difference signal(s) and a compressed version of te difference signal(s') are transmitted to one or more remote receivers (70). At the receiver (70) the usual difference signal(s) and the compressed difference signal(s') are combined (84) and the resulting signal then expanded (86) to obtain a noise-reduced difference signal for dematrixing (74) with the sum signal (M) to obtain the original left (L) and right (R) signals. Utilizing both difference signals increases the effective level of the received difference signal by about 6dB, whereas the noise is increased only about 3dB, for a net improvement of about 3dB in signal-to-noise ratio, and also permits compression of the difference signal according to any desired compression law and the use of the unchanged difference signal(s) as a reference signal for controlling the expansion (146) of the compressed difference signal(s') so as to cause the amplitude of the companded difference signal to equal the level of the usual difference signal(s). That is, the expander is adaptive to any compression characteristic that might be employed at the transmitter (24). Further the availability of the usual uncompressed difference signal(s) at the receiver enables the adaptive decoding of dynamic parameters of the received signal, such as frequency response and attack and recovery times, so that all of the parameters of the original signal can be restored automatically.

Description

BROADCAST STEREO COMPANDING SYSTEM AND APPARATUS

BACKGROUND OF THE INVENTION

This invention relates to FM broadcasting systems and, more particularly, to transmission techniques for in¬ creasing the channel capacity and coverage area of FM stereophonic broadcasting systems for both radio and tele- vision applications while maintaining compatibility with existing monophonic and stereophonic radio receivers and with the transmission standards for stereo television re¬ cently adopted by the Electronic Industries Association (EIA). The potential of FM sound broadcasting has long been recognized, primarily due to its relative immunity to electromagnetic interference and its ability to provide full audio bandwidth with low noise. Although FM stereo adds a new acoustical dimension to radio reception, it does so only at the expense of serious degradation of signal-to-noise ratio. The noise penalty in stereophonic broadcasting is well known, there being several factors which contribute to the higher noise levels and coverage losses resulting from multi-channel sound transmissions. When a broadcast station converts to biphonic service, monophonic coverage is reduced because signal power must be divided among the various components of the more com¬ plex baseband signal. (The term "biphonic" will be used hereinafter to clearly differentiate two-channel broad- casting from other forms of stereophony such as triphonic and quadraphonic broadcasting.) The biphonic signal-to- noise ratio because of the wide bandwidth of the composite signal containing the monophonic sum signal M, the pilot signal p, and the stereophonic difference signal S. With a baseband spectrum extending to 53kHz for biphonic trans- missions, the noise level is particularly high because of the rising spectral characteristic due to frequency modu¬ lation. The so-called "triangular" noise spectrum in¬ creases 6dB per octave with- increasing frequency of the composite signal, and although audio de-emphasis counter- acts this somewhat, the noise problem is still severe. After demodulation, the noise components of the difference channel sub-carrier are added, statistically independent, to the noise already present in the monophonic signal dur¬ ing audio dematrixing. Instead of here describing the factors that must be taken into account in estimating the theoretical loss of signal-to-noise ratio, attention is directed to United States Pat. No. 4,485,483 dated November 27, 1984 of Emil L. Torick and Thomas B. Keller for a summary of a number of studies that have been made of signal-to-noise degrada¬ tion. In general, these studies conclude that there is a 26dB penalty for stereophonic programming with wide audio separation, whereas for monophonic receivers noise degra¬ dation is in the range from 1dB to 7dB. Such losses of signal-to-noise ratio result in a reduction in the effec¬ tive area of coverage of a broadcast station, studies hav¬ ing shown that with biphonic transmission the two-channel reception range is only approximately one-half the mono¬ phonic reception range. The above-mentioned patent describes a biphonic FM radio broadcasting system that is fully compatible with existing receivers and which through improvement of signal-to-noise significantly extends the area of coverage of an FM broadcasting station. In the disclosed system (hereinafter sometimes referred to as the "Torick/Keller system") the usual left and right signals are convention- ally atrixed to obtain conventional sum (M) and differ¬ ence (S) signals. The difference signal is used to amplitude-modulate a first sub-carrier signal and at the same time is applied to a compressor which compresses its dynamic range to produce a compressed difference signal S¹. The compressed difference signal S' is used to amplitude-modulate a second sub-carrier signal of the same frequency but in quadrature phase relationship with the first. Suppressed-carrier, double-sideband modulation of each sub-carrier is employed, with the frequency of the sub-carrier signal being sufficiently high to assure a frequency gap between the lower sidebands of the modulated sub-carrier signals and the M signal. A conventional low- level phase reference pilot signal, lying within the aforementioned frequency gap, is employed for detection purposes at the receiver. The M signal,. the two modulated sub-carrier signals, and the pilot signal are frequency modulated onto a high frequency carrier for transmission purposes. The receiver includes a demodulator for deriv¬ ing the M signal, the normal difference signal S and the compressed difference signal S', and an expander for com- plementarily expanding the derived compressed difference signal. The expanded noise-reduced version of the differ¬ ence signal is combined with the derived sum signal M to obtain the original left (L) and right (R) signals. The receiver also includes switch means for applying the nor¬ mal difference signal, instead of the expanded version of the derived difference signal, to the combining means to enable the receiver for reproduction of conventional stereophonic signals.

In effect, then, the Torick/Keller system embodies the concept of companding a channel that is additional to 5 that normally used for stereo and to code it in a way so as to not increase the bandwidth requirements for trans- . mission. By companding the difference (L-R) signal, which gives 22dB to 26dB signal-to-noise improvement in the transmission chain, the stereo listener theoretically en-

10 joys the same signal-to-noise ratio as does the conven¬ tional monophonic listener. This amount of reduction of received noise greatly increases the effective stereo ser¬ vice area; in general, the radius from the transmitter to the point at which an acceptable signal is no longer re-

15 ceived is at least doubled as compared to that for conven¬ tional stereo, which means that four times the number of potential listeners will receive an acceptable signal.

In the application of stereophonic sound to tele¬ vision according to the system recently adopted by the

20 Electronic Industries Association (EIA), the difference signal (L-R) is compressed by a dbx Inc. compressor to give the signal-to-noise improvement necessary to overcome the penalty in signal-to-noise as compared to monophonic transmission. The compressed difference signal is used to

25 amplitude-modulate a sub-carrier and the amplitude- modulated sub-carrier and the usual sum signal M, and a pilot signal, are frequency modulated onto a high fre¬ quency carrier for transmission purposes. Only the com¬ pressed difference signal is transmitted; that is, there

30. is no transmission of an uncompressed difference signal. Thus, the signal can be compatibly received by existing monophonic television receivers, but cannot be received by existing FM stereo radio receivers; but since there had not previously been a standard for stereophonic tele¬ vision, compatibility was not a problem in the adoption of the EIA stereo television system. A primary object of the present invention is to provide an FM stereophonic broadcasting system which ex¬ hibits a greater signal-to-noise improvement than that ob¬ tainable with the Torick/Keller system.

A corollary object is to provide an FM stereo broadcast system which better utilizes the greater channel capacity of the Torick/Keller system while still realizing its improved signal-to-noise advantage.

Still another object of the invention is to compat- ibily improve the stereo television system recently agreed to by the EIA, wherein only a compresed audio difference signal of reduced maximum amplitude is transmitted, by al¬ so transmitting an unchanged difference signal in quadra¬ ture.

Another object of the invention is to provide an adaptive expander for use in the receiver having the capa¬ bility of adapting to any compressor characteristic that may be employed at the transmitter.

Yet another object is to provide an expander for use in an FM stereophonic system that is capable of adap- tive frequency response decoding and adaptive attack and recovery time decoding so that all parameters of the orig¬ inal signal can be restored automatically regardless of the compression law used in transmission. SUMMARY OF THE INVENTION

The FM stereophonic broadcasting system according to the invention, whether for radio or television, is sim¬ ilar to the Torick/Keller system in that the usual differ¬ ence signal S and a compressed difference signal S¹ are both transmitted. However, unlike the Torick/Keller radio system, in which only the expanded version of the received compressed difference signal is matrixed with the sum sig¬ nal M to obtain the original L and R signals, the receiver according to the invention combines the usual difference signal S and the compressed difference signal S', expands the resulting signal to obtain a noise-reduced difference signal and matrixes this noise-reduced difference signal with the derived sum signal M to obtain the original L and R signals. The signal content of S and S' being the same, combining them by summing increases the effective level of the received difference signal by 6dB, whereas the noise is increased only 3dB by virtue of the fact that the noise from the S channel and the noise from the quadrature S¹ channel can be considered as coming from two statistically independent noise sources whose outputs add as the square root of the sum of the squares; this means that the re¬ sultant noise in the sum signal (S + S¹) would be only 3dB, resulting in a net improvement of 3dB in signal-to- noise ratio. Alternatively, the process of decoding and adding the S and S' signals can be accomplished in a sin¬ gle operation by arranging the decoding axis to be at 45° with respect to the axes of transmission of the S and S¹ signals. In this case, the signal level from the decoder is increased by the square root of the sum of the squares of the S and S' signals, resulting in a received signal level increase of 3dB, yet with the same noise contribu¬ tion as for a one-axis decoder. There is, therefore, a 3dB improvement in received signal-to-noise ratio, the same result as when both axes are decoded independently and summed after decoding.

The transmission of the unchanged normal difference signal S, coupled with the utilization at the receiver of both the normal and the compressed difference signal, allows the transmission of the compressed signal according to any desired companding law. For example, instead of being limited to the 2:1 slope typical of the "CX" com¬ pander described in commonly assigned United States Pat. No. 4,376,916, or other known companding systems, -the in¬ vention makes possible the use of a compander system hav- ing an infinity-to-one (∞:1) or similar companding charac¬ teristic which provides a subjective 10dB to 12dB improve¬ ment against noise pumping for a given amount, of gain re¬ duction over what is realizable with such prior art com¬ panders. Additionally, because it is transmitted completely unchanged, the normal difference signal can be used at the receiver as a reference signal for controlling the level of the noise-reduced signal delivered by the expander so as to cause the amplitude of the sum signal (S+S¹) to equal the level of the normal difference signal and thus insure proper dematrixing with the received sum signal M. This feature offers the important advantage of the expand¬ er being adaptive to any compressor characteristic that might be employed at the transmitter, from which it fol- lows that should improved companding characteristics be developed it would not be necessary to build new receiver decoders to exploit them; it would be necessary only to alter the compressor (at the transmitter) and control the output level of the "new" expander with the normal differ¬ ence signal. Since the received unmodified signal S con¬ tains the complete information about the original (L-R) biphonic difference signal, the difference in the dynamic characteristics of the received S and S' signals gives a clue as to the values of these time constants; a "smart" decoder can be made sensitive to these values and be arranged to apply appropriate corrections. Also, since the signal S is unmodified, the relationship between the signal amplitudes of the various frequency bands of such companders as Dolby A and Dolby B or dBx for stereo tele¬ vision is also unchanged in this signal. It cari, there¬ fore, be used as a reference to re-establish the correct dynamic frequency response of any companding system.

The above-outlined improvements are also obtainable in the agreed EIA/dbx stereophonic television application by additionally transmitting an unchanged difference sig¬ nal in quadrature with the compressed difference signal, which can be done with no loss in service area and without affecting compatibility for existing stereo television re¬ ceivers. By exploiting both signals at the receiver in accordance with the principles outlined above, it is pos¬ sible to design an improved receiver that would (a) re- ceive an improved signal-to-noise ratio signal with less modulation of the noise; (b) enable the use of adaptive decoders; (c) require no more spectrum space; and (d) have improved high signal level signal-o-noise ratio. BRIEF DESCRIPTION OF THE DRAWINGS

Other objects, features, and advantage of the in¬ vention, and a better understanding of its construction and operation, will be had from the following detailed description when considered in conjunction with the ac- compan ing drawings, in which:

FIG. 1 is a frequency diagram of the composite baseband signal developed in accordance with the princi¬ ples of the present invention;

FIG. 2 is a simplified block diagram of a transmit- ting terminal for generating and transmitting the compos¬ ite signal of FIG. 1;

FIG. 3 is a simplified block diagram of a feedback type of compressor useful in the system of FIG. 2

FIG. 4 is a simplified block diagram of a feed for- ward type of compressor useful in the system of FIG. 2;

FIG. 5 shows the steady state compression and ex¬ pansion characteristics of one companding law useful in the FM broadcasting system according to the invention;

FIG. 6 shows the steady state compression and ex- pansion characteristics of another companding law useful in the system;

FIG. 7 is a simplified block diagram of one embodi¬ ment of a receiving terminal constructed in accordance with the invention; FIG. 8 is a simplified block diagram of an expander for use in the receiver system of FIG. 7;

FIG. 9 is a simplified block diagram of an alterna¬ tive receiving terminal embodying the invention; FIG. 10 are curves showing the level of the sum of the output signals of the two demodulators of the receiver shown in FIG. 7;

FIG. 11 are curves showing the level of the output signal from the single demodulator of the receiver shown in FIG. 9;

FIG. 12 are curves showing signal-to-noise ratio as a function of input level of the stereo difference signal for the two-axes demodulator of FIG. 7; FIG. 12A is a curve showing the improvement in signal-to-noise ratio of the two-axes demodulator of FIG. 7 with an infinity-to-one compressor over that of a 3:1 compressor;

FIG. 13 are curves showing signal-to-noise ratio as a function of input level of the stereo difference signal for the single axis 45° demodulator of FIG. 9;

FIG. 13A is a curve showing the improvement in signal-to-noise ratio of the single axis demodulator of FIG. 9 with an infinity-to-one compressor over that of a 3:1 compressor;

FIG. 14 is a frequency diagram of the composite baseband signal developed in accordance with the invention for FM stereo television transmission.

FIG. 15 is a simplified block diagram of a receiv- ing terminal including the adaptive expander of the inven¬ tion;

FIG. 16 is a simplified block diagram of an alter¬ native form of adaptive expander embodying the invention; FIG. 17 is a simplified block diagram of a modified version of the adaptive expander shown in FIG. 16; FIG. 18 is a block diagram of a feed-forward type of adaptive expander embodying the principles of the in¬ vention;

FIG. 19 is a simplified block diagram of a parallel-type frequency correcting adaptive expander;

FIG. 20 is a simplified block diagram of a series- type frequency correcting adaptive compander; and

FIG. 21 is a simplified block diagram of an adap¬ tive expander having automatic adaptive time constants.

DETAILED DESCRIPTION

Inasmuch as the FM stereophonic radio broadcasting system according to the invention is designed to be com¬ patible with the existing two-channel stereo system approved by the Federal Communications Commission, it will be useful background for an understanding of the present invention to briefly review the basic principles of the existing system. In the current system, stereophonically related signals L and R are added together to form a sum signal M having frequencies up to about 15,000 Hz. To this is added a double-sideband suppressed 38kHz sub- carrier signal Ssinωt, where S = (L-R), and a 19kHz pilot signal for receiver synchronization purposes. The compos¬ ite modulation signal can be written: e_m = M + psinωt + Ssinωt Eq. ( 1 )

2 where ω = 2 ^*π^*f_s = 38kHz, and p is the amplitude of the 19kHz pilot. It is seen from the baseband spectrum illus¬ trated in FIG. 1 that it comprises a monophonic channel M from about 50Hz to 15kHz, a 19kHz pilot, and a stereo¬ phonic difference channel Ssinωt from 23 to 53kHz. In common with the Torick/Keller system, in the transmitter of the present system a compressed difference signal designated S' is added to the conventional compos¬ ite FM signal. The compressed difference signal is con¬ veniently transmitted as a double-sideband suppressed 38kHz quadrature sub-carrier signal S'cosωt. Thus, the composite baseband signal may be represented by the fol¬ lowing equation: e_m = M + + Ssinωt + S'cosωt Eq. (2)

where p represents the amplitude of the pilot signal and _ω = 2τr •38kHz. The quadrature sub-carrier requires no additional spectrum space and, as has been noted earlier, imposes only a small penalty in modulation potential, which is largely overcome by the present invention.

A transmitter for generating this composite signal is illustrated in FIG. 2 which, in the interest of sim¬ plicity, omits some of the more conventional transmitter circuits. The two audio frequency signals L and R, de- rived from separate sources (not shown), are applied via conventional 75 μ sec. pre-e phasis networks 6 and 8, re¬ spectively, to the inputs of a conventional matrix network 10 consisting, for example, of a network of summing am¬ plifiers arranged to produce at the output terminals of the matrix the two audio signals M = (L+R) and S = (L-R). The monophonic sum signal M is applied via a first delay device 11 to one input to an adder 12,. and the stereophon¬ ic difference signal S is applied via a second delay de¬ vice 13 to the input of a first modulator 14, and also to the input of a compressor 16 of a companding system; the compressor may be one of several types which will be des¬ cribed later. The compressed difference signal from the compressor, designated S', is applied to the input of a second modulator 18, the output of which is delivered to adder 12 where it is linearly combined with the monophonic sum signal M and the signal from modulator 14. The delays introduced by delay devices 11 and 13 are such as to in¬ sure that the M signal and the two modulated signals arrive at the adder simultaneously. The sub-carrier and pilot signals are derived from a carrier generator 20 which provides a sine wave signal having a frequency of 38kHz which is applied to modulator 14 and also to a phase shift network 22 of known construc¬ tion for providing a 90° phase displacement between .the sub-carrier signal applied to modulator 18 and the sub- carrier applied^' to modulator 14. The modulators 14 and 18 comprise suppressed-carrier amplitude modulators of known construction which serve to amplitude-modulate the two sub-carriers with respective audio frequency signals so as to produce the two double-sideband, suppressed-carrier, amplitude-modulated. sub-carrier signals Ssin ωt and S'cosωt. These two signals are then combined in adder 12 with the sum signal M and a 19kHz sine wave pilot signal, also derived from carrier generator 20. The composite signal produced at the output of adder 12 set forth in Eq. (2), having the amplitude coefficients shown in FIG. 2, is then applied to the FM exciter of a transmitter 22 and frequency modulated onto a high frequency carrier for transmission to one or more remote receivers.^* It will be noted that the amplitude coefficients of the signal com¬ ponents in common with conventional FM stereo are the same as for conventional FM stereo, and that the amplitude co¬ efficient of the quadrature signal is also 0.45; that is, it is unnecessary to reduce the amplitude coefficients of signals M and S in order to avoid over-modulation of the transmitter. Modulator 14, designated the "S" modulator, receives the in-phase carrier, and the second modulator, designated modulator "S"¹, receives the quadrature sub- carrier.

The compressor 16 may take the form of any of sev¬ eral known arrangements, including the feedback type il- lustrated in FIG. 3 and the feed-forward type shown in FIG. 4. Referring to FIG. 3, the difference signal (L-R) is applied to the compressor 16 via a pre-emphasis network 25, if required, and enters the input of a variable gain element 26 which is controlled by a direct voltage derived from its output signal, designated S', and which is a function of the difference signal (L-R). This output sig¬ nal is fed to a control voltage generator circuit which includes a full-wave rectifier 28 for producing a direct voltage signal having a level proportional to the ampli¬ tude of the applied alternating signal S'. The control loop further includes a settable voltage reference device 30, which may be incorporated as part of the rectifier or, alternatively, may be inserted at some other point in the loop, for determining the knee of the transfer character¬ istic of the compressor. The direct voltage output from rectifier 28 is proportional to the amplitude of signal S' when the latter exceeds the threshold level V_ref_#, and is equal to the threshold voltage when the amplitude of signal S^! is less than the threshold, so that input sig¬ nals having amplitudes above the threshold are compressed and signals having amplitudes below the threshold are not compressed. The direct voltage produced by rectifier 28 and threshold device 30 is applied to an integrator 32 which may be of the type described in commonly assigned Pat. No. 4,376,916 which includes a network of at least three signal paths having differing time constants, the network having a common input terminal for receiving the rectified signal and a common output terminal at which the control signal for the variable gain element is devel¬ oped. All except one of the signal paths each includes a diode for causing the respective path to be conductive on¬ ly in response to the rectified signal exhibiting a change in amplitude sufficient to exceed a predetermined voltage. and the said one signal path conducts in response to any differential in voltage between the input and output ter¬ minals The signal delivered by integrator 32 is amplified by a suitable amplifier 34 and applied as a control signal to variable gain element 26. The control signal delivered by the described network causes the output signal de¬ livered by element 26 to not noticeably modify the infor¬ mation content of the source signal. The control signal generating loop has a high loop gain which enables genera- tion of a very linear transfer characteristic, the nature of which will be discussed following description of the feed-forward type of compressor illustrated in FIG. 4.

Referring now to FIG. 4, the feed-forward type of compressor has the same basic elements as the just des- cribed feedback type but differs in that the control sig¬ nal -is derived from the input to a variable gain element 40 instead of from the output. Briefly, the difference signal S, with or without pre-emphasis as necessary, is applied to the input of variable gain element 40 and also to the input of a full-wave rectifier 42, which may have a settable voltage reference device 44 embodied therein for setting the knee of the transfer characteristic. The di¬ rect voltage from the rectifier and voltage reference de¬ vice 44 is applied to an integrator 46 having the configu- ration described above in connection with FIG. 3, the out¬ put of which is amplified by a controllable amplifier 48 and then applied as a control signal to variable gain ele¬ ment 40. The compressed output signal S', the amplitude of which is a function of (L-R) , is applied to the input of modulator 18. As is known, the feed-forward type of compressor is subject to two restrictions which do not ap¬ ply to the feedback type: (1) in order for the input sig- nal versus output signal to follow a given law the vari¬ able gain element itself must have a calibrated and pre¬ scribed transfer characteristic; and (2) the control sig¬ nal must be developed with appropriate gain and offsets to match that law. In other words, the gains and offsets of the control signal generating loop and the variable gain element itself must all be chosen specifically to give the desired characteristic.

FIG. 5 is a conventional signal level diagram which depicts the steady state characteristics of an infinity- to-one companding system of which compressor 16 (of either the feedback or feed-forward type) forms a part. The in¬ put signals to be processed for compression or expansion are represented along the abscissa between -60dB and a standard operating level of OdB. The absolute value of the input voltage rises along the abscissa from left to right up to the rated level. The output levels for com¬ pression or expansion are represented along the ordinate between -60dB and OdB. The absolute value of the output voltage rises along the ordinate, from the bottom to the top, up to the rated level.

Curve 50 represents the relation between the output and input levels of the normal difference signal S which, since it is not subjected to compression, has a gain slope of one, and is included in the diagram primarily for ref¬ erence purposes.

• Curve 52 represents the relation between the output and input signal levels of the compressor 16. For input signal levels in the range between -60dB or lower and about -32.7dB, the characteristic 52 has a compression slope of one, so that input signal levels lower than -32.7dB relative to the rated level are not compressed. For input signal levels higher than about -32.7dB rela¬ tive to the rate level, the characteristic has a compres¬ sion slope of infinity; that is, beginning at the knee 54 of the characteristic, in this example set at -32.7dB, the output level is the same regardless of the level of the input signal, which in this example is at a level of -6.7dB relative to the rated level. Although this results in the received signal level being ambiguous in that it may correspond to many levels, by combining the S and S' signals at the receiver the infinity-to-one companding law for the S' signal can be accurately decoded; more par¬ ticularly, the S and S' signals are combined at the re¬ ceiver in such a way that the combination of S and S' gives a final slope to the characteristic into the expand- er which differs from infinity-to-one. Thus, unlike cur¬ rently conventional companding systems which require a prescribed curve for the input-output characteristic in order to build a complementary expander that will track the input signal without the need for a separate pilot tone or other means, the expander characteristic in the present system differs from the compressor characteristic by reason of the combination of the S and S' signals at the receiver which, in turn, enables the use of the infinity-to-one compressor characteristic in the first place. It can be seen that the dynamic range of input signals between -60dB and -32.7dB has been compressed to a range between -34dB and -6.7dB for the output signals, and that between the break point 54, which occurs when the am¬ plitude of the input signals begin to exceed the threshold ^vref./ the output level remains at -6.7dB. While the breakpoint 54 can be set at input levels other than -32.7dB, depending on the nature of the noise likely to be encountered in the transmission channel, it is important to the performance of the system that the slope of the characteristic be exactly one for input signal levels be¬ low the break point. The vertically oriented arrow 56 ex- tending between the characteristic 50 for the S signal and the compressor characteristic 52 represents the approxi¬ mately 26 dB gain increase in the S¹ difference signal as compared to the difference signal S required to compensate for the noise penalty for stereophonic programming rela- tive to monophonic.

An important advantage of using the infinity-to-one compression law for the S' difference signal can be seen from a comparison of its characteristic with the 2:1 com¬ pression characteristic typically employed in the "CX" companding system described in Pat. No. 4,376,916, shown in the dash-dot line curve 58 and a 3:1 compression char¬ acteristic shown by the dashed line curve 59. Compared to the 3:1 compression characteristic, as indicated by the vertical arrows, the infinity-to-one compression charac- teristic 52 gives a 10dB to 12dB improvement against noise modulation or "pumping," often an anomoly of companding systems. Stated another way, the infinity-to-one compres¬ sion law reduces the audibility of noise pumping yet still gives improved signal-to-noise ratio for input sig- nals throughout the range from about -44dB to about OdB. In particular, at the lower end of this range of levels, the ear is very sensitive to noise pumping caused by lack of sufficient program level to mask the noise so that slight variations in level of the program enables the ear to hear variations in the level of the noise. It is seen from comparison of characteristic 52 with the 2:1 charac¬ teristic 58 that the improvement is even greater; for ex- ample, at an input signal level of -30dB there is an im¬ provement of about 15dB of output signal level relative to a compressor having the 2:1 characteristic.

A further advantage of the infinity-to-one compres- sion law over the 3:1 compression characteristic is that the knee 54 occurs at an input level of -32.7dB whereas the 3:1 curve of a "CX" expander has to have the knee set at -49dB in order to get the same total amount of compres¬ sion for an acceptable signal level. Thus, in the case of a "CX" expander having a 3:1 compression law, by defini¬ tion, gain variations are still occurring at input signal levels of -49dB, causing the operating signal-to-noise ra¬ tio level at that point to be much lower than in the case of the infinity-to-one characteristic, thus making the noise pumping all the more audible.

While the described infinity-to-one companding characteristic provides the listed advantages, it is poss¬ ible, by virtue of utilizing both the S and S¹ difference signals at the receiver, to use a greater than infinity- to-one characteristic for the S' difference signal and thereby achieve additional advantages. More specifically, as shown in the signal level diagram of FIG. 6, it is pos¬ sible to use a compressor having a characteristic 60 for the S' difference signal which extends with a slope of one up to within a dB or two of 100% modulation and then grad¬ ually drops back from that level as the level of the S difference signal rises, such that addition of the S and S¹ signals gives an infinity-to-one characteristic (de¬ picted by dotted line characteristic 62) that is always maintained at a level such that the square root of the sum of the squares of signals S and S' does not exceed 100% modulation. The curve 60 has a re-entrant characteristic in the sense that it starts to come down, from a break point at an input level of about -28dB, as the level of the S difference signal increases, and intersects the S difference signal curve at the -3dB point. This somewhat unusual characteristic can be quite easily derived by sub¬ tracting the difference signal S from the output of a reg¬ ular infinity-to-one compressor having the characteristic illustrated in FIG. 5. This re-entrant characteristic not only permits maximum possible modulation of the 38kHz transmission signal, it also insures that the 38kHz sub-, carrier has substantially constant amplitude over a wide dynamic range of program level; that is, this characteris¬ tic makes it possible to transmit at about 100% modulation all of the time, which is the best that can be done in any communication system.

A first embodiment of a receiver according to the invention is shown in the block diagram of FIG. 7 and, again, in the interest of simplicity, some of the more conventional FM receiver circuits (e.g., RF and IF stages and discriminator networks) have not been shown and will be only briefly mentioned as necessary. A received FM signal is amplified in the RF and IF stages (not shown) of a receiver/FM demodulator 70, and demodulated in any of the known FM detection circuits (not shown) to derive the audio signals contained in the received signal, namely, the signals M, S, S¹ and the pilot, each with the indicat¬ ed amplitude coefficient. The monophonic sum signal M is separated from the higher frequency components of the com¬ posite signal by a low-pass filter 72 and applied as one input to a de-matrixer circuit 74 of conventional design. The remaining components of the composite signal are se- lected by a bandpass filter 76 designed to pass frequen¬ cies in the band from 19kHz to 53 kHz and to reject fre¬ quencies below this band, and then applied to an S demodu¬ lator 78 and to an S' demodulator 80. The pilot signal is derived by means (not shown) and applied to a carrier gen¬ erator 82 which regenerates quadrature versions thereof, sin t and cosωt, which are applied to demodulators 78 and 80, respectively. So as to-derive the unchanged differ¬ ence signal S and the compressed difference signal S', which has the function of (L-R) established by the com¬ pressor at the transmitter, the S and S' signals are summed in an adder 84 and the sum signal is applied to an expander 86 wherein the signal S' is complementarily ex¬ panded (in a manner to be described) to obtain the differ- ence signal S at its output for delivery via an optional de-emphasis circuit 87 (if required) to a second input to de-matrixer 74 when a switch 88 is in the position shown. The de-matrixer 74, which may be of the conventional con¬ struction illustrated in the aforementioned Torick/Keller patent including an adder and a subtractor, combines the M and noise-reduced S signals to produce as outputs the sig¬ nals 2L and 2R, the amplitude of which is then reduced by one-half to obtain signals L and R for application to the left and right loudspeakers, respectively (not shown), all typical of the mode of operation of a conventional two- channel FM receiver.

In addition to operating as just described, the re¬ ceiver is fully compatible with conventional monophonic and two-channel (biphonic) stereophonic broadcasts. When a monaural broadcast is being received, the output of the receiver/FM demodulator 70 comprises only the monaural signal M consisting of (L+R) . This signal is selected by low pass filter 72 and applied to de-matrixer 74, and since no signal is applied to the second input of the de- matrixer, only the signal^' M appears at each output of the de-matrixer for application to the left and right loud- speakers, respectively.

For enabling the receiver to reproduce a received conventional two-channel stereo signal, the switch 88 is actuated, preferably automatically, from the position shown to the dotted line position so as to connect the output of the S demodulator 78 to the second input of de- matrixer 74. Such automatic switching can be achieved, for example, by any of several known techniques, such as modulating the pilot tone, or adding a separate identifi¬ cation signal, when a compressed difference signal is transmitted; a detector in the receiver (not shown) re¬ sponsive to the identification signal produces a signal for actuating the switch 88 from the solid line position to the position shown by dotted line in FIG. 7. Thus, when a conventional two-channel stereo signal is received, the M signal, as before, is applied to one input of de- matrixer 74, and the S signal, derived from demodulator 78, are combined in the de-matrixer to produce output sig¬ nals 2L and 2R, the amplitude of each of which is reduced by one-half prior to application to the left and right loudspeakers, respectively.

FIG. 8 is a block diagram of the expander 86 of FIG. 7 for complementarily expanding a signal compressed at the transmitter with the feedback type compressor shown in FIG. 3. Repeating a portion of the block diagram of FIG. 7, the S and S* signals from demodulators 78 and 80, respectively, are summed in adder 84 and applied to the expander 86 which essentially comprises a compressor of the type shown in FIG. 3 and a subtractor circuit 90. The subtractor 90 is arranged to subtract from the sum signal (S+S') a noise-reduced signal S¹ derived by feeding the output signal from subtractor 90 through a variable gain element 26' to the subtractor and controlling the gain of variable gain element 26' with a control signal derived from its output signal, which is the same as the signal applied to subtractor 90. The control signal generating loop is the full equivalent of the comparable control loop of the compressor shown in FIG. 3 and accordingly its com¬ ponents are identified by corresponding, but primed, ref¬ erence numerals. The subtractor 90 may be a simple com¬ parator amplifier having high stability, and if the other components are identical to corresponding components of the compressor the expander will be complementary with the compressor. The requirement that the signal S appear at the output of subtractor 90 while the signals at its two inputs are (S+S¹) and S', respectively, is met by the var¬ iable gain element 26' having a very low phase delay at the highest audio frequency encountered, which is about 15 to 16kHz. In other words, the delay through the variable gain element must be less than a few degrees at 16kHz, or a few microseconds. In practice, the subtractor 90 pref¬ erably comprises an operational amplifier having a transit delay of the order of 100 nanoseconds, and similarly for variable gain element 26', thereby to enable subtraction of two. signals which seemingly haven't yet arrived. The resultant output signal S, then, is the sum of the S sig¬ nal and the noise-reduced S' signal- which, by virtue of the noise reduction in the S' signal when the S and S' signal levels are equal, has a signal-to-noise ratio 3dB better than if only the usual difference signal (L-R) were used. Further details of the received improvement in signal-to-noise ratio as a function of input signal level will be described later. The output signal from subtract¬ or 90 is applied to the optional de-emphasis circuit 87, if required, and then applied to de-matrixer 74. It will be understood that the de-emphasis circuit would be in¬ serted only if pre-emphasis was employed in the trans¬ mission chain.

Referring now to FIG. 9, there is shown in block diagram an alternative form of receiver in which demodula¬ tion and summing of the unchanged difference signal S and the compressed difference signal S' is accomplished by us¬ ing a single quadrature demodulator having a decoding axis set at 45° to the normal carrier generator axes of sinωt or cosωt. Again, some of the more conventional FM receiv¬ er circuits (e.g., RF and IF stages and discriminator net¬ works) haVe not been shown and will be only briefly men¬ tioned. A received FM signal is amplified in the RF and IF stages (not shown) of a receiver/FM demodulator 100, and demodulated in any of the known FM detection circuits (not shown) to derive the audio signals contained in the received signal, namely, the signals M, S and S' and the pilot. The monophonic sum signal M is separated from the higher frequency components of the composite signal by a low-pass filter 102 and applied as one input to a de- matrixer circuit 104 of conventional design. The remain¬ ing components of the composite signal are selected by a bandpass filter 106 designed to pass frequencies in the band from 19 Khz to 53Khz and to reject frequencies below this band. The S and S' signals are applied to a quadra¬ ture demodulator 108 having a demodulation axis set at 45°, and the pilot signal is derived (by conventional means not shown) and applied to a carrier generator 110 which regenerates responsively thereto a signal sin (at + 45°) (or, alternatively, cos (ωt - 45°))which is applied to demodulator 108. By using the single quadrature demodula- tor, the demodulation and summing of the S and S' signals take place simultaneously, in one shot, and, as indicated, the output signal from the demodulator will be the square root of the sum of the squares of the S and S' signals, and a function of the^' (L-R) difference signal applied to the transmitter. This signal is applied to the expander 86, which may have a configuration similar to the expander shown in FIG. 8, wherein the signal is expanded to obtain a noise-reduced output signal corresponding to the origin¬ al difference signal S. This signal is applied to an op- tional de-emphasis network 112 (if required) and then applied to a second input to de-matrixer 104 for combina¬ tion with the sum signal M to produce as outputs the sig¬ nals 2^'L and 2R, the amplitude of which is then reduced by 1/2 to obtain signals L and R for application to the left and right loudspeakers, respectively (not shown), all typ¬ ical of the mode of operation of a conventional two- channel FM receiver, including subjecting the de-matrixer output signals to the usual 75 μ sec de-emphasis in net¬ works 106 and 108. Although a 45° decoder has been described in detail by way of example, it will be understood that demodulation axes other than 45° may be used. The results will be sim¬ ilar except for the introduction of a ratio between the gain coefficients for the S and S' signals in the summing action. Such other axis angle may be preferable over the 45° axis angle under certain conditions. Although the system of FIG. 9 cannot reproduce a conventional two-channel stereo signal, it is fully com¬ patible with conventional monophonic broadcasts. When a monaural broadcast is being received, the output of re- ceiver/FM demodulator 100 comprises only the monaural sig¬ nal M consisting of (L + R) . This signal is selected by low-pass filter 102 and applied to de-matrixer 104, and since no signal is applied to the second input of the de- matrixer, only the signal M appears at each output of the de-matrixer for application to the left and right loud¬ speakers, respectively.

It will have been seen from the foregoing that there are basically two different ways of using the S sig¬ nal in the receiver's decoder: (1) two-axes decoding shown in FIG. 7 using sinωt and cosωt demodulators for re¬ spectively-deriving the S and S' signals, adding S and S" together and expanding the sum, and (2) 45° axis decoding shown in FIG. 9, in which demodulation and summing of the S and S' signals both happen in one process and expanding the resulting signal. The relative merits of these two approaches will be seen from a comparison of FIGS. 10 and 11 which respectively illustrate the level of the output of the adder 84 in the two-axes demodulator of FIG. 7 (i.e., S + S') and the level of the output signal from the single 45° axis decoder of FIG. 9 (i.e., the square root of the sum of S² + S'²), both as a function of input lev¬ el. Both figures include the usual input/output curves, the curve for the unchanged difference signal S having un¬ ity slope and the curve S' for the difference signal com- pressed with the infinity-to-one compressor described ear¬ lier. The knee of the compressor is set at about -32.7dB, above which the output level is -6.7dB regardless of the level of the input signal. Referring specifically to FIG. 10, the dotted line curve labeled (S + S¹) depicts the level of the output signal from adder 84 (FIG. 7) as a function of input signal level and is seen to gradually increase from -6.667dB at an input level of about -32.7dB, to about 3.3dB for OdB maximum input signal level. As has been described previously, separate demodulation of the two axes has the effect of introducing two statistically independent random noise generators which together add 3dB of noise to the transmission link noise but, since the output signal rises by 6dB at the -6.7dB input level, there is a net improvement of 3dB in signal-to-noise ra¬ tio.

Turning now to FIG. 11, it will be noted that the dotted line curve labeled S² + S'², for input S and S' signals identical to those depicted in FIG. 10, also in¬ creases gradually, but not as steeply as the dotted line curve in FIG. 10, to a level of about +0.8dB at an input level of OdB, as compared to the plus 3.3dB in FIG. 10. However, the signal-to-noise ratio is the same as for the other case because the noise contributed by the demodula¬ tor, since it has but a single axis, will be 3dB less than for the two-axes case in which there was an increase in signal level but also more noise. In the single axis de- modulator of FIG. 9, not only does the demodulator con¬ tribute less noise than do the two demodulators of the FIG. 7 system, but there is a lower output signal level because of the fact that the demodulator derives the square root of (S² + S'²). FIGS. 12 and 13 respectively depict signal-to-noise ratio as a function of input signal level, with an arbi¬ trarily chosen -40dB noise level on the transmission medi- um, for the two-axes demodulating system of FIG. 7 and the single-axis modulator of the FIG. 9 receiver. Each of these figures show three curves: (1) a straight line la¬ beled S having unity slope relating the unchanged differ- ence signal to signal-to-noise ratio; (2) a curve labeled S' which depicts signal-to-noise ratio as a function of the level of the difference signal compressed by 3: 1 com¬ pressor; and, (3 a curve labeled (S + S¹) in FIG. 12, and labeledVs² + S*² in FIG. 13, which depict signal-to-noise ratio as a function of input level for the case where the S' signal is compressed according to the infinity-to-one compression law. Considering first the third curve of FIG. 13, which shows the signal-to-noise ratio for the single axis 45° decoder, it is seen that the signal-to- noise ratio is improved at the higher input signal levels, and that at OdB input level the signal-to-noise ratio is slightly better than it is for the unchanged difference signal S; this is the result of taking the square root of the sum of the squares of the unchanged and compressed difference signals.

The same curve in FIG. 12, for the case in which the unchanged and compressed difference signals are de¬ rived from separate demodulators, there is a noise contri¬ bution from two axes, or 3dB more noise than in the single-axis case, and, furthermore, the actual received signal level is now (S + S'), which does not involve squares and square roots; thus, the derivation of the ex¬ act signal-to-noise ratio as a function of signal level involves a different set of calculations than for the cor- responding curve for the single-axis case.

The shape of the curve for the two cases is gener¬ ally the same but they differ by one or two dB here and there, as will best be seen from a comparison of FIGS. 12A and 13A which show the improvement in signal-to-noise ra¬ tio realized from using an infinity-to-one compressor in¬ stead of a 3:1 compressor for the two-axes demodulator case and the single 45° axis demodulator case, respective¬ ly. It will be noted in both figures that at input levels of about -49dB the signal-to-noise improvement of the infinity-to-one compressor goes to zero; this stems from the fact that in order to have a fair comparison the sig- nal level performance was arbitrarily chosen such that in the absence of signal, the background noise level will set the infinity-to-one compressor to have the same noise im¬ provement as the 3:1 compressor, that value being about 26dB of noise improvement. Also the maximum signal levels for both were arbitrarily set at the same value of -6.7dB, as was the output for maximum input signal level. The curves of FIGS. 12A and 13A are also of the same basic shape, but have differences in detail which graphically show the differences between the curves of FIGS. 12 and 13, most notably at the -35dB input level, where the sin¬ gle axis 45° decoder has about a 2dB advantage in the im¬ provement as compared to the two-axes decoding method. The reason for this is the contribution of noise by the two demodulation axes, which causes the slightly lower re- ceived signal-to-noise ratio for the two-axes case. How¬ ever, the 9dB improvement in signal-to-noise ratio as com¬ pared to a "CX" compressor, for example, is still signifi¬ cant.

Besides its signal-to-noise superiority over the two-axes demodulator, the single 45° axis decoder has the further advantage that conventional stereo receivers cur¬ rently utilize a single-axis decoder, albeit one fed by a sinωt signal rather than the sin(ωt + 45°) required for the demodulator of FIG. 9. Furthermore, the circuit is somewhat less expensive to manufacture in that less com¬ ponents are required for demodulating and summing in a single operation.

As has been briefly discussed previously, the prin¬ ciples of the present invention are also applicable to the recently agreed EIA/dBx stereo television system. As shown in FIG. 14, the baseband spectrum of this system comprises a monophonic channel M from about 50Hz to 15kHz, representing the sum of stereophonically related left and right signals, and a compressed stereophonic difference signal S' from 23kHz to 53kHz, which is transmitted as a double-sideband suppressed sub-carrier signal S' sinωt for example (or S' cosωt) . The baseband signal also includes a pilot signal positioned in the gap between the M channel and the difference signal channel. The pilot frequency is the horizontal line frequency of the NTSC television sys¬ tem, namely, 15.734kHz, and the sub-carrier frequency is the second harmonic of the pilot frequency, or 31.468kHz. In accordance with the present invention, the base¬ band signal would additionally include an uncompressed difference signal S transmitted as a double-sideband sup¬ pressed kHz sub-carrier in quadrature with the compressed difference signal sub-carrier (so as not to require any additional spectrum) . This composite signal would be fre¬ quency modulated onto a high frequency carrier for trans¬ mission purposes, using a transmitter substantially as shown in FIG. 2. Because the signal includes the sum sig- nal M and the compressed difference signal S', it is com¬ patible with existing EIA/dbx receivers, and there would be no loss in service area. Unlike the EIA/dbx stereo television receiver, in which only the expanded version of the received compressed difference signal is matrixed with the sum signal M to obtain the original left and right channel signals, the receiver would combine the uncom- pressed difference signal S and the compressed difference signal S', using, for example, the adaptive expander des¬ cribed herein below to achieve the above-described im¬ provement in signal-to-noise ratio in the received signal and with less noise pumping. Such adaptive decoder is controlled by the uncompressed, and thus unmodified, dif¬ ference signal and is, therefore, capable of recovering dynamic level, frequency, and attack and recovery time constant information contained in the original difference signal. It should now be evident that the transmission of the uncompressed difference signal S along with a com¬ pressed ve^'rsion S¹ of the difference signal, and the. util¬ ization of both at the receiver to obtain a noise-reduced difference signal, provides several improvements in FM stereophonic radio and television broadcasting systems in¬ cluding (1) increased channel capacity, (2) improved re¬ ceived signal-to-noise ratio as compared to currently available companding systems and (3) the ability to de¬ code the uncompressed difference signal and the compressed difference signal either separately or together, all of these advantages being realizable while maintaining com¬ patibility with existing FM and biphonic receivers in the case of radio, and with the EIA/dbx system in the case of television. As applied to FM stereo radio, the utiliza- tion of both compressed and uncompressed difference sig¬ nals at the receiver provides the further advantage of en¬ abling the use of any of several companding characteris- tics, with their attendant advantages, including the infinity-to-one and re-entrant characteristics shown in FIGS. 5 and 6, as well as the use of adaptive expanders which will now be described. FIG. 15 is a block diagram of a receiver in associ¬ ation with a preferred form of adaptive expander according to the invention. Again, for purposes of simplicity, some of the more conventional FM receiver circuits (e.g., RF and IF stages and discriminator networks) have not been shown and will be only briefly mentioned as necessary. A received FM signal is amplified in the RF and IF stages (not shown) of a receiver/FM demodulator 120, and demodu¬ lated in any of the known FM detection circuits (not shown) to derive the audio signals contained in the re- ceived signal, namely, the signals M, S, S' and the pilot signal, each with the indicated amplitude co-efficient. The monophonic sum signal M is separated from the higher frequency components of the composite signal by a low-pass filter 122 and applied as one input to a de-matrixer cir- cuit 124 of conventional design. The remaining components of the composite signal are selected by a bandpass filter 126 designed to pass frequencies in the band from 19kHz to 53kHz and to reject frequencies outside this band, and then applied to an "S" demodulator 128 and an "S"¹ demodu- lator 130. The pilot signal is derived by known means (not shown) and applied to a carrier generator 132 which regenerates quadrature versions thereof, namely, sinωt and cosωt, which are applied to demodulators 128 and 130, re¬ spectively, so as to obtain the signals S and S" (which has the function of (L - R) established by the compressor at the transmitter), respectively. The availability at the receiver of both the un¬ changed difference signal and the compressed version thereof enables the use of an expander that adapts to the compression characteristic, whatever its form, this being accomplished by using the received unchanged difference signal S as a reference level for developing a control signal for a variable gain element connected to receive the compressed difference signal S¹ and for producing a noise-reduced difference signal for application to de- matrixer 124.

In the embodiment of the expander shown in FIG. 15 the received unchanged difference signal S is full-wave rectified by a rectifier 134 to produce a direct current signal having a level proportional to the amplitude of the difference signal S. This direct current signal is ap¬ plied to an- integrator 136 which preferably is of the com¬ plex form described in the aforementioned United States Pat. No. 4,376,916. The output signal from integrator. 136 which describes the envelope information of the unchanged difference signal S, is applied as one input to a comparator 138 which, for example, may be an operational amplifier 140, with the signal from integrator 136 applied to its positive input.

The compressed difference signal S' produced at the output of demodulator 130 is applied as one input to an adder 142 in which it may, if desired, but not necessar¬ ily, be added to the unchanged difference signal S; the dotted line connection 144 signifies the optionality of adding the two signals together at this point. Assuming that the unchanged difference signal is not added to the compressed difference signal, the compressed difference signal is simply transmitted to the input of a variable gain element 146, the output of which is applied to a sec¬ ond full-wave rectifier 148 which produces a direct cur¬ rent output signal having a level proportional to the am¬ plitude of the signal appearing at the output of variable gain element 146. This direct current signal is applied to a second integrator 150, which may have the same con¬ struction as integrator 136 but, on the other hand, it need not have the same time constants as those used in in¬ tegrator 136, nor those used in compressor 16 at the transmitter, the reasons for which will shortly become evident. A settable reference voltage level for this loop is established by a device 152 labeled V_ref_#, which ele¬ ment may be embodied in the rectifier circuit 148. The signal developed in the chain including rectifier 148 and integrator 150 is applied as a second input to comparator 138, more specifically, to the negative input of opera¬ tional amplifier 140. Any output signal from comparator 68 representing a difference in the amplitude of the two applied signals is amplified by a suitable amplifier 154 to produce a control signal which is applied to the con¬ trol element of variable gain element 146 to control the gain thereof.

In operation, if the levels of the output signals from integrators 136 and 150 are the same, there will be no change in the magnitude of the control signal applied to variable gain element 146 which, in turn, signifies that the level of the output signal from the variable gain element 146 corresponds to that of the unchanged differ¬ ence signal S. Viewed in another way, the function of comparator 138 is analogous to that of the threshold level of conventional expanders in that no change is made to the voltage level determined by the unchanged difference sig- nal S until the signal level is above the so-called knee of the transfer characteristic. In the present circuit, the threshold, instead of being a preset fixed signal as is the case in conventional expanders, is a signal derived from the unchanged difference signal S, which, of course, varies in amplitude with certain attack and recovery times, but is at the level that the compressed signal S* should be. Thus, instead of -there being a fixed reference voltage, the comparator 138 provides a changing reference level which dynamically varies with the changes in level of the unchanged difference signal S itself. The purpose of the V_ref_# device 152 is to establish a fixed gain difference equal to the fixed gain difference of the compressor employed at the transmitter so as to preclude variable gain effects on signals below a certain level, that is to say, to provide the equivalent of a "knee" to offset errors that could occur if the signal-to-noise ratio of the difference signal S is not sufficiently high. At input signals levels above such "knee", the level of the signal appearing at the output of variable gain element 146 is determined by the amplitude of the control signal developed in comparator 138 which, in turn, is dependent on the amplitude of the unchanged difference signal S. By virtue of controlling the gain of variable gain element 146 with a control signal proportional to the difference between the level of the compressed difference signal and the unchanged difference signal, the_.-output of variable gain element 146 is the signal S,- except that it has been subjected to noise reduction by reason of the ex- pansion of the compressed difference signal that takes place in the control loop. The signal S produced at the output of variable gain element 146 is applied to the in¬ put of an optional de-emphasis circuit 156 (which is used only if preemphasis is used in the transmitter) and the output thereof is applied via a switch 158 to the second input of dematrixer 124.

The level of the signal S appearing at the output of variable gain element 146 is set by the V_ref. device 152 and amplifier 154 to be equal to the level of the re¬ ceived unchanged difference signal S and using that signal as a reference level gives the expander the important ad¬ vantage of being capable of adapting to any of a wide va¬ riety of companding laws. Thus, in the event a new type of compression transfer characteristic were to be devel¬ oped in the future, it would not be necessary to scrap the described expander and replace it with a new receiver ex¬ pander in order to take advantage of the new transfer characteristic. Because its operation depends only on the level of the unchanged difference signal S, any type of compression law currently available, or as may be devel- oped in the future, would be satisfactorily decoded by the adaptive system of FIG. 15.

Unlike most existing companding systems, which re¬ quire a very specific law for compression and a specific complementary, law for expansion to achieve satisfactory performance, this adaptive system is not critically sensi¬ tive to the compression law, and even time constants are not as important, because the receiver always has a refer¬ ence level derived from the unchanged difference signal S. Reverting now to FIG. 15, in addition to operating as previously described, the receiver is fully compatible with conventional monophonic and two-channel (biphonic) stereophonic broadcasts. When a monoaural broadcast is be¬ ing received, the output of receiver/FM demodulator 120 comprises only the monaural signal M consisting of (L + R) . This signal is selected by low pass filter 122 and applied to de-matrixer 124, and since no signal is applied to the second input of the de-matrixer, only the signal M appears at each output of the de-matrixer for application to the left and right loudspeakers, respectively.

For enabling the receiver to reproduce a received conventional two-channel stereo signal, the switch 158 preferably is automatically actuated from the position shown to the dotted line position so as to connect the output of demodulator 128 to the second input of de- matrixer 124. Such automatic switching can be achieved, for example, by any of several known techniques, such as modulating the pilot tone, or adding a separate identifi¬ cation signal, when a compressed difference signal is also being transmitted; a detector in the receiver (not shown) responsive to the identification signal produces a signal for actuating the switch 158 from the solid line position to the dotted line position. Thus, when a conventional two-channel stereo signal is received, the M signal, as before, is applied to one input of de-matrixer 124 and the S signal derived from demodulator 123 are combined in the de-matrixer to produce output signals 2L and 2R, the am¬ plitude of each of which is then reduced by one-half prior to application to the left and right loudspeakers., -

FIG. 16 shows an alternative configuration of the adaptive expander of FIG. 15, the difference being that instead of having two integrators, each following a re¬ spective full-wave rectifier and the outputs of the inte¬ grators applied to the comparator, the outputs of the rec- tifiers are applied to the comparator and its output ap¬ plied to a single integrator. More particularly, the un¬ compressed difference signal S from demodulator 128 is ap¬ plied -to a first full-wave rectifier 160 and is also ap- plied as one input to an adder 162 in which it is prefer¬ ably added to the compressed difference signal S¹ from de¬ modulator 130. The direct current signal from rectifier 160, which has a level proportional to the amplitude of the difference signal S, is applied to one input of a com- parator, such as a difference amplifier 164. A settable reference voltage level for this loop is established by a device 166 labeled V_ref.-|, which element may be embodied in the circuit of rectifier 160, and the purpose of which will be described presently. The signal from adder 162, representing the sum of signals S and S', is applied to the input of a variable gain element 168, the output of which is applied to a second full-wave rectifier 170 which produces a direct current output signal having a level proportional to the amplitude of the output signal from variable gain element 160. This direct current signal is applied to the other input of difference amplifier 164. A second voltage reference device 172, designated V_ref 2, establishes a reference voltage level for this loop. Any output signal from the comparator representing a differ- ence in the amplitude of the two applied signals (after taking into account the reference voltage levels estab¬ lished by elements 166 and 172) is applied to an integra¬ tor 174, preferably of the complex type described above. The output signal from integrator 104 is amplified by a suitable amplifier 176 to produce a control signal which is applied to the control element of variable gain element 168 to control the gain thereof. As in the expander of FIG. 15, the function of the difference amplifier is anal¬ ogous to that of the threshold level of conventional ex¬ panders in that no change is made to the voltage level de¬ termined by the unchanged difference signal S until the signal level is above the knee of the transfer character¬ istic. Voltage reference V_ref_#-| is provided to limit the lower knee of the expander as in conventional expand¬ ers, and V_ref.2 is provided to cut off its associated loop, because otherwise the loop gain would try to make the output of variable gain element 168 to be constant at the level of the uncompressed difference signal S. Other¬ wise, the operation is similar to that of the adaptive ex¬ pander of FIG. 15 and produces a noise-reduced difference signal S at the output of variable gain element 168 for application to de-matrixer 124.

FIG. 17 shows yet another implementation of the same basic adaptive expander which effects a saving in components as compared to either of the previously des¬ cribed implementations in that it requires only one full- wave rectifier and one complex integrator to achieve sub¬ stantially the same results. More particularly, in this case the uncompressed difference signal S is directly ap¬ plied to one input of a difference amplifier 180 and is added to the compressed difference signal S¹ in an adder 182 to produce a sum signal which is applied to the input of a variable gain element 184. The output signal from variable gain element 184 is applied to a gain element such as an amplifier 186, the output of which is applied to the second input of difference amplifier 180. Any re- suiting difference signal is full-wave rectified by a rec¬ tifier 188 and the resulting direct current signal, after level adjustment by a voltage level setting device 190, is applied to an integrator 192, again of complex type, the output of which is amplified by a suitable amplifier 194 to develop a control signal for variable gain element 184. As before, the output of the variable gain element is a noise-reduced signal S for application to the de- matrixer.

The gain element 186 is included to avoid signal ambiguity at the output of full-wave rectifier 188 which would occur unless steps are taken to make sure that the signal from variable gain element 184 is always either larger or smaller than the difference signal S. Other¬ wise, if the two signals applied to difference amplifier 180 should be equal, the output signal from the full-wave rectifier cannot know which signal was the larger. To eliminate this possibility for ambiguity the gain element 186 is connected in the loop following the variable gain element 184 to make sure that the signal in this loop is always dominant over the difference signal S. Because gain element 186 is inside the control loop for variable gain element 184 it will automatically wash out because of the feedback nature of the system, and will not affect the amplitude of the final noise-reduced signal S appearing at the output of variable gain element 184. Alternatively, a suitable gain element can be inserted in the S signal path to the difference amplifier to insure that this signal is always dominant, and inserting in the output line to the dematrixer an attenuator to offset the gain of the element in the S signal path.

FIG. 18 is a block diagram of yet another implemen- tation of an adaptive expander embodying the invention, this one being of the feed-forward type as opposed to the three feedback types just described. Although the feed- forward type is subject to the restrictions, not appli¬ cable to the feedback type, that the variable gain element must have a calibrated and prescribed transfer character¬ istic, and the control signal must be developed with ap- propriate gain and offsets to match that law, it can oper¬ ate satisfactorily if such restrictions are taken into ac¬ count. The feed-forward embodiment has many elements in common with the previously described adaptive expanders but are arranged differently as will now be described. Specifically, the uncompressed difference signal S is full-wave rectified by a rectifier 200, the output of which is applied as one input to a difference amplifier 202, the level of the rectified signal being settable by a voltage level setting device 204 designated V_ref_#-|. The uncompressed difference signal S is also added to the com¬ pressed difference signal S* in an adder 206 and the re¬ sulting sum signal applied to a second full-wave rectifier 138 and also to the input of a variable gain element 210. The direct current signal from rectifier 208, after being subjected to level setting by a voltage level setting de¬ vice 212 designated V_ref_Φ2 and which may be embodied in the rectifier circuitry, is applied to the other input of difference amplifier 202. Any difference signal appearing at the output of amplifier 202 is applied to an integrator 214 of the complex type described earlier, the output of which is applied as one input to a "law" generator 216, to which is also -applied a reference voltage V_ref 3 derived from a suitable voltage source 218. For a given change in level of the signal from integrator 214, the "law" genera- tor 216 produces the correct output voltage or current for application to the control element of variable gain ele¬ ment 210 to insure that gain element 210 gives the correct amount of gain or attenuation as may be necessary at any instant of time. For example, since the variable gain element typically is logarithmic, the law generator 216 would contain a logarithmic circuit element to match the changes in the signal from integrator 214 to decibels of attenuation in the variable gain element 210. Thus, the output of variable gain element 210 is a noise-reduced difference signal S for application to the dematrixer.

The availability at the receiver of the uncom- pressed difference signal also enables the design of a frequency correcting adaptive expander, one form of which is shown in the block diagram of FIG. 19. The principle of the illustrated frequency correcting expander is to utilize the uncompressed difference signal to adaptively expand a compressed signal containing unknown frequency responsive elements. For example, the Dolby A companding system has four companded channels, each with a distinct and separate frequency band which is compressed indepen¬ dently according to its own signal levels. The function of the system of FIG. 19 is to separate such a multi-chan¬ nel signal into corresponding separate channels and using the uncompressed difference signal, which contains all of the signal information of the compressed difference sig¬ nal, to separately adaptively expand the multiple channel signals, and thereafter combining the resulting noise- reduced channel signals to obtain a noise-reduced and fre¬ quency corrected difference signal for application to a dematrixer. To this end, the uncompressed difference sig¬ nal S and the compressed difference signal S' are summed in an adder 220 and the resulting signal applied to an ar¬ ray of filters including a low pass filter 222 designated LPF-j and a multiplicity of band pass filters 224, 226 and 228 respectively designated BPF2, BPF3 and B FN, de¬ signed to pass successively higher frequency bands in the audio frequency spectrum. The uncompressed difference signal S is applied to an equivalent set of filters 230, 232, 234 and 236 respectively designated LPF1, BPF2_/ BPF3 and BPFN ^an^ having the same pass bands as the corre¬ spondingly designated filters in the other set. Each of the filters receiving the sum signal from adder 220 is connected to its own variable gain element 242, 244, 246 and 248, respectively, the outputs of which are applied to respective full-wave rectifiers 252, 254, 256 and 258. The direct current signals from these rectifiers are applied to one input of a respective difference amplifier 260, 262, 264 and 266. Similarly, the frequency bands of the uncompressed difference signal S passed by filters 230, 232, 234 and 236 are full-wave rectified by respec¬ tive rectifiers 237, 238, 239 and 240 and the respective direct current output signals are applied to the other in¬ put terminal of difference amplifiers 260, 262, 264 and 266, respectively. The difference signals delivered by the difference amplifiers are applied to respective com¬ plex integrators 268, 270, 272 and 274 and the output sig¬ nals therefrom after suitable amplification (not shown) are applied as control signals for variable gain elements 242, 244, 246 and 248, respectively. It -will have been observed that the system of FIG. 19. is a group of four adaptive expanders each constructed according to the block diagram of FIG. 16, except that each is surrounded with a low pass filter or a band pass filter. The outputs of the multiple variable gain elements are combined in an adder 276 which combines all of the separately adaptively ex¬ panded signals together to produce at the output the total noise-reduced and frequency corrected difference signal S for application to the dematrixer. Thus, the system of FIG. 19 consists of a multiplicity of adaptive expanders, each of which operates on its own part of the frequency spectrum as defined by the associated filters. In prac¬ tice, the filters need not have steep rolloff characteris¬ tics and may be, for example, of the order of 6dB or 12dB per octave. It will be understood that FIG. 19 illus¬ trates an ultimate general case of a frequency correcting adaptive expander and that it may be modified to have more or less expanders to meet a specific application.

FIG. 20 is a block diagram of a frequency-correct¬ ing adaptive expanding system for providing essentially the same function as that achieved with the system of FIG. 19, except in this case the filtering process is arranged in series rather than in parallel. Also, rather than showing a general case, FIG. 20 shows how the system can be applied to the agreed EIA/dbx Inc. stereo tele¬ vision system in which the audio difference signal is com- pressed by a compressor which has two variable gain ele¬ ments arranged in series and includes de-emphasis and bandpass networks for effectively dividing the audio fre¬ quency spectrum into two bands. Thus, for enabling the adaptive expansion of a dbx/compressed signal, the system of FIG. 20 has two variable gain elements connected in series and includes two sets of filters for dividing the audio frequency spectrum into substantially the same bands as are used in the dbx compressor. More specifically, the uncompressed difference signal S, which again is used as a reference signal, is applied to a first filter 280, and may optionally be added to the compressed difference sig¬ nal S' in an adder 282; the resulting signal is applied to a second filter 284 having the same pass characteristics as filter 280. In practice, both of filters 280 and 284 may be de-emphasis filters having characteristics compar¬ able to those which establish the upper band in the dbx Inc. compressor. As in the arrangement of FIG. 16, the selected band of the signal S is full wave rectified by a rectifier 216 and the selected band of the signal from ad¬ der 212 is applied to a variable gain element 288. This variable gain element operates on a selected high frequen- cy band and functions as a variable de-emphasis element. The overall output of the expander is rectified by a rec¬ tifier 290, and the direct current output signals from rectifiers 286 and 290 are applied to respective inputs of a difference amplifier 292, the output of which is applied to a complex integrator 294 which develops a control sig¬ nal for variable element 288. The remaining portion of the spectrum of difference signal S is applied to a second filter 296 and the frequency-corrected and noise-reduced signal delivered by variable element 288 is applied to a wide band variable gain element 304. The output of ele¬ ment 304, which is the output of the expander, is also applied to a bandpass filter 298 which, again, has the same pass characteristics as filter 296. As before, the selected band of signal S is full-wave rectified by a rec- tifier 300 and the resulting direct current signal is applied as one input to a difference amplifier 302, and the signal passed by filter 298 is applied to a full-wave rectifier 306, the output of which is applied to the other input of the difference amplifier 302. The output signal from the difference amplifier is applied to a complex in¬ tegrator 308, the output of which after suitable amplifi¬ cation (not shown) constitutes the control signal for var- iable gain element 304. Thus, it is seen that the system of FIG. 20 is similar in function to the system of FIG. 19 except that the individual adaptive expanders are connect¬ ed in series with filter or variable de-emphasis elements connected between them, and the full noise-reduced and frequency corrected signal S appears at the output of the last variable gain element (i.e., element 304) in the chain. It is to be understood that in the interest of clarity only two series-connected expanders have been shown, and that additional stages can be provided should the application require.

Since the uncompressed difference signal S contains all of the dynamics of the original signal which may, for example, be music, it can be used in an adaptive expander to determine what attack and recovery times were used in the transmitter's compressor, whatever its type. By way of background, in the design of companding systems there is always a compromise between the attack and recovery times. Considering first the attack time, it can not be so short that the rate of change of gain of the signal would be so large during signal correction as to produce an audible click due to the rapid amplitude modulation of the signal. Another disadvantage of an attack time that is too short is that sharp peaks in the signal will tend to over modulate the transmitter. Accordingly, attack times of 10 to 20 milliseconds are typically used in com¬ panding systems in order to maintain a higher average mod¬ ulation level by, in effect, missing the peaks on the pro¬ gram signal while at the same time not continuously over- loading the transmitter, as would be the case if the attack time is too long. Now briefly considering the attack and recovery times of companding systems, in order to keep the modula¬ tion of the transmitter at a maximum, the attack time of the compressor should be as short as possible so that upon occurrence of a sudden high level transient the compressor is gain adjusted downward in order not to over modulate. If the recovery time were long and the transient followed immediately by a passage of quiet music, the expander would take too long to reduce gain, and the transmission channel noise would be heard fading away. On the other hand, a recovery time that is too short is undesirable be¬ cause there will be gain modulation during each cycle, or cycle-following of low frequencies and consequent increase in distortion. These conflicting requirements of attack and recovery times are usually met by making compromised choices or by using complex integrators, such as the one used in the "CX" companding system described in commonly assigned Pat. No. 4,376,916, which has multiple and com¬ plex attack and recovery times for different signal condi- tions.

The above-outlined characteristics of companding systems are equally present when considering an adaptive expander because it is not possible to make the attack time of the adaptive expander infinitely short, nor its recovery time very long. If this could be done the output dynamics of the signal would be good because if, for in¬ stance, the reference signal S had a transient that went up and then recovered in a certain way, the compressor at the transmitter would alter those dynamic ratios and an adaptive expander would, if not limited by recovery or at¬ tack time integrators, seek to establish the correct sig¬ nal level and put the output right. However, in practice. for a low frequency transmitted signal such an adaptive expander itself would cycle-follow and cause distortion. Consequently, even an adaptive expander must have attack and recovery time elements, preferably of the complex type as shown and described in the embodiments discussed thus far. However, it will be seen that it is possible to adapt the integrator of the adaptive expander to the attack and recovery time constants of the compressor at the transmitter. Referring to FIG. 21, which illustrates in block diagram form a feedback type of adaptive expander corre¬ sponding generally to the arrangement shown in FIG. 16, the uncompressed difference signal S is full-wave recti¬ fied by a full-wave rectifier 310 and the direct current output signal, after being subjected to level setting by a voltage reference element 312 (which may be incorporated in the rectifier circuitry) , is applied as one input to a difference amplifier 314. The uncompressed and compressed difference signals are summed by an adder 316 and the re- suiting signal is applied to a variable gain element 318. As before, the output signal from the variable gain ele¬ ment is full-wave rectified by a rectifier 320 and its dc output signal adjusted in level by a voltage reference de¬ vice 322 and thereafter applied as the second input to difference amplifier 314. The output signal from the dif¬ ference amplifier is applied to a complex integrator 324 which includes a variable attack time element 326 and a variable recovery time element 328. The output of the in¬ tegrator, after suitable amplification by an amplifier 330, becomes the control signal for variable gain element 318. Again, as before, the output of the variable gain noise-reduced signal S and, as will be seen, is also adapted to the dynamics of the original signal.

When the signal at the output of difference ampli¬ fier 314 becomes large on a transient basis, it is known that there is a large error in either the attack or the recovery time constants of integrator 324, and for the reasons discussed above it is desirable that the integra¬ tor be adjusted to always go for the longest possible attack time and the longest possible recovery time. This is accomplished by also applying the output signal from difference amplifier 314 to two sensing amplifiers: an attack time error amplifier 332 and a recovery time error amplifier 334. Each of the sensing amplifiers 332 and 334 has an associated voltage reference element 336 and 338, respectively, which establishes a voltage level against which the output signal from the amplifier is compared to determine the magnitude of the error signal from differ¬ ence amplifier 314. It is possible to distinguish whether the error is an attack time error or a recovery time error because the polarity of the error signal for attack time is the opposite of the polarity of the error signal for recovery time. Thus, the error signal can, in either case, be integrated to develop a control signal for alter¬ ing the attack and recovery time constants of integrator 324. To this end, the error signal from sensing amplifier 332 is applied to an attack error integrator 340, which develops a control signal which is applied to the control element of variable attack time element 326 of integrator 324, and the error signal output of sensing amplifier 334 is applied to recovery error integrator 342, which devel¬ ops a recovery time control signal for adjusting the re¬ covery time of variable recovery time element 328. Vari- able attack time element 326 may take the form of an RC circuit including a series resistor and a capacitor con¬ nected to a ground reference in which the resistive ele¬ ment is controllably variable; for example, the resistor may be either a variable current source or a field effect transistor (FET) arranged to be controlled by the control signal from the attack error integrator. Thus, for exam¬ ple, if the attack time of integrator 324 had initially been set to have a long time constant and a signal is re- ceived having a short attack time, a large attack time er¬ ror would be sensed, and if it exceeded the level estab¬ lished by V_ref_# 336, the control signal would change the resistance of the variable resistance element in a direc¬ tion to shorten the attack time constant. Meanwhile, the attack error integrator 340 would very slowly relax to again exhibit a long attack time constant.

The time constant of the variable recovery time element 328 may be controlled by the parallel combination of a resistor and a capacitor for discharging the main in- tegrating capacitor; again, the resistor may take the form of either a variable current source or a variable resis¬ tor, or similar circuit element, adapted to be controlled by the control signal developed by the recovery error in¬ tegrator 342. As in the case of the variable attack time element, the controllable resistor in element 328 would normally be set to exhibit its longest possible recovery time constant (within reason) and to be shortened in re¬ sponse to the recovery time control signal. It will now be apparent that the system of FIG. 21, which is basically the same as that of .FIG. 16 except that the complex inte¬ grator is replaced by a controllable complex integrator, will automatically adapt to the attack and recovery dynam- ics of the compressed signal, all made possible by using the uncompressed difference signal S as a reference.

It will, of course, be understood that the three general types of adaptive expander described above are not mutually exclusive and, in fact, they can be combined in a variety of permutations and combinations. For example, one or the other of the frequency-correcting arrangements of FIGS. 19 and 20 could be combined with the just des¬ cribed adaptive attack and recovery time expander fea- ture. Also, it will be recognized by ones skilled in the companding art that the described adaptive techniques for expanding compressed signals are equally applicable to compressors should a situation arise requiring them.

It should now be evident that the transmission of the uncompressed difference signal S along with a com¬ pressed version S' of the difference signal, enables util¬ ization of the uncompressed difference signal at the re¬ ceiver as a reference signal to make an expander adaptive to any companding law and which enables decoding of dynam- ic parameters of the received signal, such as frequency response, and attack and recovery time constants, so that all of the parameters of the original signal can be re¬ stored automatically, regardless of the transmission sys¬ tem. Although several specific embodiments of the inven¬ tion have been illustrated and described, they are exemp¬ lary only, and such variations and modifications as will now be suggested by those skilled in the art will be un¬ derstood as forming a part of the present invention inso- far as they fall within the spirit and scope of the appended claims. For example, the invention may be used in other than the described frequency modulation system, including amplitude modulation, phase modulation, delta modulation, pulse and pulse code modulation. The carrier may be any of a variety of forms, including visible or infrared light, and multiplexed satellite distribution systems. The various functions shown in the block diagrams may be implemented in many ways so as to achieve ideal and/or ro¬ bust characteristics; for example, the full-wave rectifier and integrator typically form an envelope detector of which various implementations, including r.m.s. detectors, are available. Also, although the described use of a quadrature channel permits the transmission of the com¬ pressed difference signal without requiring an increase in bandwidth, if additional bandwidth is available, it is then possible to use an additional carrier, sufficiently removed in frequency from the S subcarrier, to contain the compressed difference signal.

Claims

1. A receiver for use in a biphonic FM stereophonic broadcasting system which is adapted to receive a high frequency carrier modulated by a stereo sum signal M and first and second sub-carriers respectively modulated by a stereo difference signal S and a compressed version S' of said stereo difference signal, said receiver comprising: demultiplexing means for deriving said stereo sum signal M, said stereo difference signal S and said co - pressed stereo difference signal S', means responsive to both said stereo difference signal S and said compressed stereo difference signal S' for expanding said compressed stereo difference signal S' to obtain a noise-reduced stereo difference signal having an amplitude level substantially equal to that of the stereo difference signal S, and de-matrix means for combining said noise-reduced stereo difference^* signal with said stereo sum signal M to obtain left and right channel stereophonic signals.

2. A receiver as defined in claim 1, wherein said means for expanding comprises: first and second demodulators for separately de od- ulating said first and second sub-carriers for obtaining said stereo difference signal S and said compressed, ster- eo difference signal S', means connected to receive and for summing said stereo difference signal S and said compressed stereo dif- ference signal S' to obtain the sum signal (S + S'), and expander means for expanding said sum signal (S + S' ) to obtain said noise-reduced stereo difference signal.

3. A receiver as defined in claim 2 for use in a ster- eophonic FM broadcasting system in which said first and second sub-carriers are of the same frequency and dis- placed in phase by 90° and said high frequency carrier is further modulated by a pilot signal having a frequency which is one-half the frequency of the sub-carriers, wherein said demultiplexing means includes means for de- riving said pilot signal, and wherein said first and sec- ond demodulators are connected to receive said pilot sig- nal.

4. A receiver as defined in claim 1 for use in a ster- eophonic FM broadcasting in which said first and second sub-carriers are of the same frequency and displaced in phase by 90° and said high frequency carrier is further modulated by a pilot signal having a frequency which is one-half the frequency of the sub-carriers, wherein said demultiplexing means includes means for deriving said pi- lot signal and wherein said means for expanding comprises: quadrature demodulator means connected to receive and responsive to said pilot signal for quadrature de odu- lating said stereo difference signal S and said compressed stereo difference signal S' and producing an output signal representing the square root of the sum of the squares of the signals S and S' and expander means for expanding said output signal to obtain said noise-reduced difference signal.

5. A receiver as defined in claim 1, wherein said means for expanding is adaptive and comprises: an electronically controllable variable gain device having input, output and control terminals, means for applying said compressed stereo differ- ence signal S' to the input of said variable gain device, and control signal generating means for generating a control signal responsively to both said compressed stereo difference signal S' and said stereo difference signal S and applying the control signal to the control terminal of said variable gain device for causing the gain between the input and the output of the variable gain device to have a value dependent on the value of said control signal.

6. A receiver as defined in claim 5, wherein the con- trol signal generating means of said means for expanding includes means including comparator means for dynamically comparing a signal proportional to the signal produced at the output terminal of said variable gain device against a signal proportional to said stereo difference signal S for producing an error signal proportional to differences be- tween said signals S and S', and means for processing said error signal to produce said control signal.

7. A receiver as defined in claim 6, wherein the means of said expanding means for applying said stereo differ- ence signal S' to said variable gain device includes means for summing said stereo difference signal S and said com- pressed stereo difference signal S' and applying the re- suiting sum signal to the input terminal of said variable gain device, and wherein said control signal generating means in- eludes rectifier means for causing said control signal to be a direct current signal which substantially follows dy- namic variations of the error signal produced by said com- parator means.

8. A receiver as defined in claim 7, wherein the rec- tifier means of said control signal generating means com- prises first rectifier means responsive to the stereo dif- ference signal S for producing a first rectified signal having a level proportional to the amplitude of and which substantially follows dynamic variations of the stereo difference signal S, and means for applying said first rectified signal to said comparator means, and second rectifier means responsive to the signal produced at the output terminal of said variable gain de- vice for producing a second rectified signal having a lev- el proportional to the amplitude of and which substantial- ly follows dynamic variations of the output signal from said variable gain device, and means for applying said second rectified signal to said comparator means.

9. A received as defined in claim 8, wherein the con- trol signal generating means of said expanding means in- eludes complex integrator means consisting of a network of signal paths having differing time constants for modifying said control signal to minimize program modulated noise in spite of rapid changes in level of said signals S and S'.

10. A receiver as defined in claim 9, wherein said com- plex integrator means comprises a single integrator cir- cuit having an input connected to receive the error signal from said comparator means and having an output coupled to the control terminal of said variable gain device.

11. A receiver as defined in claim 9, wherein said com- plex integrator means comprises: a first integrator circuit having an input connected to receive said first rectified signal and having an out- put coupled to said comparator means, and a second integrator circuit having an input con- nected to receive said second rectified signal and having an output coupled to said comparator means.

12. A receiver as defined in claim 7, wherein said stereo difference signal S is directly applied to said comparator means, wherein the rectifier means of said control signal generating means comprises a single rectifier responsive to said error signal for producing a rectified error sig- nal having a level proportional to the amplitude of and which substantially follows dynamic variations of said er- ror signal, and wherein said control signal generating means fur- ther includes complex integrating means consisting of a network of signal paths having differing time constants for modifying said rectified error signal to produce said control signal.

13. A receiver as defined in claim 5, wherein the means in said expanding means for applying said compressed ster- eo difference signal S' to said variable gain device com- prises means for summing said stereo difference signal S and said compressed stereo difference signal S' and apply- ing the resulting sum signal to the input terminal of said variable gain device, and wherein said control signal generating means in- eludes comparator means for dynamically comparing a signal proportional to said sum signal against a signal propor- tional to said stereo difference signal S for producing an error signal proportional to differences between said sig- nals S and S', and means for processing said error signal to produce said control signal.

14. A receiver as defined in claim 13, wherein the con- trol signal generating means of said expanding means fur- ther includes first rectifier means responsive to the stereo dif- ference signal S for producing and applying to said com- parator means a first rectified signal having a level pro- portional to the amplitude of and which substantially fol- lows dynamic variations of the stereo difference signal S, second rectifier means responsive to said sum sig- nal for producing and applying to said comparator means a second rectified signal having a level proportional to the amplitude of and which substantially follows dynamic vari- ations of said sum signal, and wherein said means for processing said error signal includes integrator means consisting of a network of sig- nal paths having differing time constants.

15. A receiver as defined in claim 14, wherein said er- ror signal processing means further includes circuit means for modifying the output signal from said integrator means to produce a control signal which is adapted to the gain characteristics of said variable gain element.

16. A receiver as defined in claim 9, wherein the means for applying the stereo difference signal S to said first rectifier means includes first filter means for ap- plying to said first rectifier means a selected frequency band of the difference signal S, and wherein the means for applying the output signal from said variable gain device to said second rectifier means includes second filter means having substantially the same pass characteristics as said first filter means for coupling said selected frequency band of said output signal to said second rectifier means.

17. A receiver as defined in any of the preceding claims, wherein said receiver further comprises: switch means connected to said de-matrix means and operative alternatively to couple the stereo differ- ence signal S, instead of the noise-reduced difference signal, to said de-matrix means for enabling said receiver for reproducing conventional two-channel stereophonic broadcasts.

18. An improved biphonic FM radio transmission system for transmitting stereophonically related left and right channel signals, said system comprising: at the transmitter matrix means for combining the left and right channel stereophonic signals to obtain a sum signal M cor- responding to the sum of the left and right channel sig- nals and a stereo difference signal S corresponding to the difference between the left and right channel signals; compressor means connected to said matrix means for compressing the dynamic range of said difference signal S to produce a compressed stereo difference signal S" which is compressed according to a desired compression 1aw; generating means for generating two sub- carriers of the same frequency and spaced 90° apart in phase and a pilot signal having a frequency which is one- half the frequency of the sub-carriers; modulator means connected to said generating means, said compressor means and said matrix means for am- plitude-modulating each sub-carrier with a respective one of said stereo difference and compressed stereo difference signals S and S' to develop two double-sideband suppresse- d-carrier signals, the frequency of said sub-carriers bei- ng sufficiently high as to insure a frequency gap between the lower sidebands of the modulated sub-carrier signals and the frequency bandwidth of said sum signal M, with the frequency of the pilot signal being located within said gap; and frequency modulating means connected for freq- uency modulating the modulated sub-carrier signals, the pilot signal, and the sum signal M onto a high-frequency carrier for the purpose of transmitting the same to one or more remote receivers; and at the receiver deriving means for receiving said frequency modulated high-frequency carrier and for deriving said sum signal M and said amplitude-modulated sub-carriers respec- tively modulated by said stereo difference signal S and said compressed stereo difference signal S'; means including demodulator means for combin- ing and demodulating the amplitude-modulated sub-carrier signal for producing a combined signal and for expanding the combined signal to obtain a noise-reduced stereo dif- ference signal having an amplitude level substantially equal to that of the stereo difference signal S; and de-matrix means connected to receive and for combining said noise-reduced difference signal with said sum signal M to obtain said left and right channel stereo- phonic signals.

19. A biphonic FM radio transmission system as defined in claim 18, wherein said demodulator means comprises first and second demodulators each for demodulating a dif- ferent one of said amplitude-modulated sub-carriers to de- rive said stereo difference signal S and said compressed stereo difference signal S', wherein said combining means includes means for summing said stereo difference signal S and said com- pressed stereo difference signal S' to produce a sum sig- nal (S + S') , and wherein said means for expanding comprises an expander for expanding said sum signal (S + S') to obtain said noise-reduced stereo difference signal.

20. A biphonic FM radio transmission system as defined in claim 18, wherein the compressor in the transmitter has a compression characteristic which has unity slope for in- put signal levels up to a predetermined input signal level and which has infinity slope for input signal levels above said predetermined level.

21. A biphonic FM radio transmission system as defined in claim 18, wherein the compressor in the transmitter has a compression characteristic which has unity slope for in- put signal levels below a first predetermined level, and a slightly negative slope for input signal levels greater than said first predetermined level and less than a second higher predetermined level.