NO157400B - CONNECTOR FOR MODIFICATION OF THE DYNAMICS AREA OF AN INPUT SIGNAL. - Google Patents

Description

The invention generally relates to switching or circuit devices that change the dynamic range of signals, namely compressors that compress the dynamic range, and expanders that expand the dynamic range. The invention is particularly useful for processing audio or sound signals, but is also applicable to other signals.

Compressors and expanders are normally used together (compander system) to effect noise reduction. The signal is then compressed before transmission or recording and expanded after reception or playback from the transmission channel. Compressors can, however, be used alone to reduce the dynamic range, e.g. to fit the capacity of a transmission channel, without later expansion when the compressed signal is suitable for the end purpose. In addition, compressors alone are used in certain products, especially audio products that are only intended to transmit or record compressed radio programs or pre-recorded signals. Expanders alone are used in certain products, especially audio products that are intended only to receive or play already compressed radio programs or pre-recorded signals.

In certain products, particularly sound recording and sound playback products, a single device is often designed to operate in switchable mode, as a compressor for recording signals and as an expander for playing compressed radio programs or pre-recorded signals.

The degree of compression or expansion can be expressed

in decibels (dB). For example, a compression of 10 dB means that an input dynamic range of N dB is compressed to an output range of (N-10) dB. In a noise reduction system, a compression of 10 dB followed by a complementary expansion of 10 dB is said to provide a noise reduction of 10 dB.

The invention relates in particular to a switching device for modifying the dynamic range of an input signal and which comprises a first circuit with a bilinear characteristic (where "linear" in this context denotes constant gain) which is composed of

1) a linear low-level part up to a threshold,

2) a non-linear (variable gain)

intermediate level part, above the threshold and up to a termination point which provides a predetermined maximum compression ratio or expansion ratio, and

3) a linear high-level part with a gain that

is different from the low-level part gain.

The characteristic is termed a bilinear characteristic because there are two parts with essentially constant gain.

In practice, the threshold and termination point are not always well-defined "points". The two transition regions where the mid-level part transitions into the linear low-level and high-level parts can both vary in shape from a smooth curve to a sharp curve, depending on the control characteristics of the compressor and expander.

It should also be pointed out that switching or circuit devices with bilinear characteristics are distinguished from two other known circuit device classes, namely; (a) A logarithmic or non-linear circuit with either fixed or varying slope and without any linear part, i.e. the gain changes over the entire dynamic range. (b) Circuit devices with a characteristic having two or more parts of which only one part is linear ("uni-linear"). f

A circuit arrangement with a bilinear characteristic has special advantages and is widely used. The threshold can be set above the input noise level or the transmission channel noise level to rule out the possibility that the circuit is controlled by noise. The high-level part with essentially constant gain avoids non-linear processing of high-level signals which would otherwise introduce distortion. In the case of an audio signal, for which the circuit must be syllable-forming or syllabic, the high-level part further provides an area in which to process the overshoots that occur with a syllabic circuit when the signal level increases suddenly. The over-oscillations are suppressed by means of clipping diodes or similar means. Only bilinear characteristics are capable of providing this combination of benefits.

Most of the known circuits with bilinear characteristics currently in use in consumer audio products provide 10 dB of compression and expansion which is sufficient for many purposes. However, this leaves some noise that is audible to some listeners, and for the best possible reproduction accuracy, more compression and expansion is desirable, e.g. 20 dB. It is difficult to provide such a large degree of compression or expansion without encountering problems affecting the character of the signal.

There are known and commercially available circuits which provide a compression or expansion of 20 dB and even more, but these are usually constant-slope logarithmic circuits in which there is a constantly varying gain over the entire dynamic range or nearly the entire dynamic range. Such circuits suffer from higher distortion and greater signal matching problems at very low and very high signal levels than the bilinear circuits in which the gain change is limited to an intermediate part of the characteristic, and overshoot problems are more severe than with circuit devices with a bilinear characteristic. Known constant slope companders use compression ratios in the range of 1.5:2, 2:1 and 3:1, but 2:1 is most common.

The compression ratio is defined as the ratio between the incremental input dynamic range and the incremental output dynamic range. The expansion ratio for a complementary expander is the inverse of the compression ratio. If the compression ratio is 3:1, the expansion ratio is 1:3. It is convenient to use the term with the inverse expansion ratio, which for the example just given is 3:1, i.e. it corresponds to the compression ratio. For simplicity, the present description will be largely limited to the compression ratio, with the understanding that the same considerations apply to the expansion ratio when appropriate changes are made.

A high compression ratio has a disadvantage in that it is difficult to ensure complementarity between the compressor and the expander. In particular, level errors or errors in the transmission or recording medium's frequency response lead to correspondingly increased errors at the expander's output.

It is known (eg, US-PS 2,558,002, US-PS

4,061,874 and Japanese Patent Publication 51-20124) to increase the degree of available compression by connecting a number of compressor stages in cascade. These known circuits (controlled impedance, arrangements, diodes, etc.) multiply the compression ratios of the individual stages so that a high compression ratio results, with the above described disadvantage. For example, one circuit with a compression ratio of 2:1 and a further circuit with a compression ratio of 3:1 will give a total ratio of 6:1. The resulting expansion ratio of 1:6 would place a heavy burden on the uniformity of the transmission channel. Another factor is the demands placed on any circuit that causes the gain change necessary to provide the compressor or expander characteristic. It is relatively easy to cause a circuit to effect precise gain changes over a range of 10 dB, but considerably more difficult to cause the same circuit to effect precise gain changes over a range of 20 dB. It is thus difficult to provide a controlled, reproducible characteristic for use in a compander system. Japanese Patent Publication 51-20124 concludes that a number of series compressors (and expanders) are not suitable as a noise reduction system for high fidelity reproduction systems.

It is also known (US-PS 3,902,131 and US-PS

3 930 208) to cascade a number of compressor stages working in mutually exclusive frequency ranges. Although such arrangements may not result in any increase in compression ratio beyond the compression ratio of a single stage, they do not provide any increase in compression.

From US-PS 3 969 680 an automatic dynamics compander system is known with a compression circuit whose useful signal path comprises a first and a second voltage-controlled amplifier, so that different characteristics can be provided by selecting a clamp on the output of the first and the other amplifier as output. The known device thus contains two voltage-controlled amplifiers in series, but the device is otherwise not relevant to the device according to the present invention, as it does not use circuits that have bilinear characteristics.

From DE-OS 2 038 973 a compander coupling is known in which a summation amplifier is arranged in a compressor and an expander, respectively, to which the analogue signal is supplied both on a linear path, directly in the compressor and inverted in the expander, and also via at least one pair which is formed by an inverting amplitude high-pass circuit and an inverting amplitude low-pass circuit and has equal absolute limit amplitude. The circuitry includes resistors that determine the thresholds for diode limiter and converter devices. The thresholds for several of the aforementioned pairs are preferably separated from each other in order to provide a logarithmic characteristic.

The desired characteristic is achieved by using a combination of circuits in parallel. However, the circuit arrangement does not use bilinear circuits to achieve greater compression or expansion without increasing the compression or expansion ratios.

In light of the foregoing considerations, the purpose is

with the present invention to provide an increased degree of compression or expansion without any undesired large increase in the compression ratio and without placing excessive demands on any circuit involved in performing a gain change.

A further object is to provide an increased degree of audio compression or expansion without producing an undesirably large increase in the overshoots produced during transient signal conditions.

A careful examination of bilinear circuits shows that they have not only the advantages listed above, but also a further advantage, namely a way of solving the problem of high compression ratio and, in the case of audio circuits, also a way of solving the problem of big swings.

It should be noted that the superimposition of the linear areas does not increase the compression ratio in these areas. The compression ratio is increased only in the limited area in which dynamic action takes place. It has therefore been found possible to separate the areas with dynamic effect in such a way that the necessary, total increase in compression is achieved, while at the same time not changing the total, maximum compression or expansion ratio to a significant extent.

A further distinctive feature of this arrangement is that the total result is bilinear with all the accompanying benefits. The effect-shifting possibility with bilinear devices thus represents a further, hitherto unrecognized advantage of this class of device.

The aforementioned purposes are achieved with a switching device for modifying the dynamic range of an input signal, which device comprises a first circuit with a bi-linear characteristic consisting of a low-level part with essentially constant gain up to a threshold level, an intermediate-level part, above the threshold, with variable gain providing a compression ratio or expansion ratio varying between one at the ends of the mid-level section and a maximum value between these ends, and a high-level section of substantially constant gain different from the gain of the low-level section, in that in front of the first circuit, when a compressor for broadcast transmitters, if desired a broadband compressor or broadband limiter can be arranged, and the switching device is characterized according to the invention in that the first circuit is followed by at least one second circuit which also has a bilinear characteristic within a frequency range that is common to the circuits, as the mid-level parts of the circuits' characteristics within a frequency range common to the circuits are mutually offset in such a way that a gain change is achieved over a larger range of intermediate input levels than for any of the circuits individually, and an increased difference between the gains at low and high input levels, but with an overall maximum compression or expansion ratio that, by virtue of the displacement, is essentially no greater than the ratio for a single circuit, and an overall characteristic that is also bilinear.

When it comes to audio circuits, if the circuits have overshoot suppressing (limiting) elements,

is it also possible to shift the circuit thresholds together with the shift of the syllabic thresholds. The overshoots of the low-level circuits, or stages, are reduced accordingly, with minimal overall overshoot of the many stages. This is in contrast to conventional, logarithmic compressors in which large overshoots are produced according to the nature of the matter.

Each of the circuits can introduce a change in the signal's spectral content, for example a low-level treble boost when a compressor is involved. Each successive step can thus be affected by a signal with gradually varying spectral content. When it comes to complex signals, this has the advantage of spectrally spreading the chances of errors in the decoding function. When it comes to a tape recorder with an uneven frequency response characteristic for example, the spectral shift tendency reduces the overall dynamics and frequency response errors in the decoded result.

The degree of displacement that is necessary shall

now taken up for consideration. For simplicity, reference is made to a series connection of two compressor circuits. The compression ratio for each of the first and second circuits will rise from one at the respective threshold to a maximum, and this will be called the rising edge of the compression ratio. The ratio will then fall back to one, and this will be called the falling edge. Strictly speaking, the falling edge can approach one asymptotically, but for all practical purposes it can be considered to have reached one when it is at some value that differs from ~one only by an arbitrarily small amount.

The displacement of the mid-level portions of the first and second circuits results in the falling edge of one circuit overlapping the rising edge of the other circuit. At least as a first approximation, the difference between the two thresholds can be made so that the overlap of the flanks results in a total compression ratio that does not significantly exceed the maximum compression ratio for one or the other circuit itself.

The threshold for the second circuit is preferably lower than the threshold for the first circuit (if more than

two circuits are used, each additional circuit preferably has a progressively lower threshold) when it is a compressor, and vice versa when it is an expander. In principle, the order can be reversed, so that the first compressor circuit has the lowest threshold. When it comes down to it

more than two circuits, the order of threshold levels among these can in principle be reversed and arranged in any order as long as the circuits' mid-level parts are properly offset.

The ideal offset is thus considered to be that which causes the falling edge of one circuit to overlap the rising edge of the other circuit in order to limit, as far as possible, the level range in which dynamic action occurs in the total, series-connected device, while at the same time avoids a significant increase in the maximum compression or expansion ratio beyond the maximum compression or expansion ratio for a single device. If, for example, the maximum compression ratio for each circuit is 2:1, the compression ratio for the total circuit arrangement will rise to 2:1, maintain this value over the overlap and then fall back to one. Ideally, there is thus no increase at all above the ratio of 2:1, in contrast to the previously known arrangements with cascaded compressor stages which multiply the ratio to 4:1.

In practice it may be difficult to achieve an optimum overlap at all frequencies, but it can be realized that the overall maximum compression ratio, provided a reasonable approximation is made to the ideal, can be prevented from rising excessively above 2:1 in the given Example. In a practical circuit arrangement, it could possibly rise to 2.5:1.

A low maximum compression ratio (eg 1.5:1) allows an expander to follow the compressor more easily to provide good complementarity with signal channels that have somewhat unreliable gains and/or frequency responses.

However, a low compression ratio spreads the dynamic effect over a greater range of levels, causing greater sensitivity to noise modulation for a given maximum degree of gain noise reduction ratio at low and high input levels. There is thus a compromise between undesirable effects caused by both high and low compression ratios. The ideal compression ratio will therefore depend on the system environment and system design goals.

The ability to shift bilinear stages provides the designer with an additional means by which an overall circuit can be optimized. When this is done, the shapes of the compression characteristics for individual stages can be designed with displacement specifically in mind. Consideration is also given to the transient characteristics of the circuits, and the opportunity is preferably used to shift the overshoot suppression thresholds in audio compressors and expanders so that minimal overall overshoot results.

A well-known type of circuit called a "sliding band" circuit, which can be used for each of the first and second circuits, produces the special, desirable characteristic for the case of high frequency audio compression or expansion when using of high-frequency boosting (for compression) or high-frequency lowering (for expansion) using a high-pass filter with a variable lower corner frequency. As the signal level in the high-frequency band increases, the filter's corner frequency shifts upwards so that the raised or lowered band is narrowed and the useful signal is excluded from the boost or cut. Examples of such circuits can be found in US-PS

Re 28,426, US-PS 3,757,254, US-PS 4,072,914, US-PS 3,934,190

and Japanese Patent Application No. 55529/71.

Accordingly, each of the first and second circuits may be such a "sliding band" circuit. In principle, the stationary corner frequencies for the two sliding band circuits can be different, and this can be used to provide a degree of compression or expansion that is higher in one part of the treated frequency band than in another. In accordance with an important further development of the invention, however, the corner frequencies are essentially made identical. This leads to the advantage of a sharper distinction or distinction between the frequency range where raising or lowering is used and the area where it is not used, and consequently a sharper distinction between the area where noise reduction no longer takes place, due to the appearance of a significant, useful signal, and the area where noise reduction remains effective.

On the other hand, there are also well-known circles

in which the frequency spectrum is divided into a number of bands by means of corresponding band-pass filters, and the compression or expansion is carried out in each band by means of a gain control device (either an automatically responding diode-type limiter or a controlled limiter) as it turns itself a compressor, with some form of reciprocal or complementary circuit that expands. Examples of such circuits can be found in US-PS 3,846,719. These split-band or multi-band circuits have the advantage of independent action in the different frequency bands, and if this property is required, such circuits can be used as the first, second or further stages of the circuit arrangement according to the invention.

In principle, one of the first and second circuits may be a multi-band circuit and the other may be a sliding-band circuit. This will be of interest in a special situation in which, for example, one wishes to increase the degree of compression or expansion in one part of the total frequency band, as the sliding band circuit and one or more of the band-splitting channels operate over this part of the frequency band.

It is known to construct bilinear compressors and expanders of both sliding band and split band type, using only a single signal path. However, it is usually preferred to construct such devices by providing a main signal circuit which is linear with respect to dynamic range, with a combinational circuit in the main circuit, and an additional circuit which diverts an input signal from the input to or output of the additional circuit and has its output connected to the combination circuit. The additional circuit comprises a limiter (self-acting or controlled), and the limited signal from the additional circuit boosts the main circuit signal in the combination circuit in the case of compression, but bucks the main circuit signal in the case of expansion. The additional path's limited signal is smaller than the main path's signal in the upper part of the input dynamic range. The main circuit and the additional circuit are preferably and most conveniently separately identifiable signal paths.

Such known compressors and expanders are particularly advantageous as they enable the desired type of transfer characteristic to be established accurately without problems of high level distortion. The low-level part with essentially constant amplification is established by giving the additional path a threshold above the noise level. Below this threshold, the further trajectory is linear. The intermediate level part is formed by the area over which the additional path's limiting effect becomes partially effective, and the high level part with essentially constant amplification occurs after the limiter has become fully effective, so that the additional path's signal continues to increase and is negligibly compared to the main path signal . In the highest part of the input dynamic range, the output signal of the circuit is effectively only the signal transmitted by the main linear path, i.e. linear with respect to dynamic range. In dual-path audio circuits, the provision of overshoot suppression is particularly convenient.

An inherent advantage of series-connected stages having mutually offset syllabic (stationary) thresholds, as opposed to a single stage having the same low-level gain or maximum compression effect, is that the overshoot produced by the mutually offset series stages, with the same overshoot suppression threshold in each the shifted series stages and in the single stage, is significantly less than that produced by the single stage.

In the arrangement with the mutually staggered series stages, the amplification is spread among the many stages, and thus each stage must provide less amplification than the individual stage provides. Consequently, the overshoot produced by the first stage in the series stage arrangement is necessarily less than that produced by the single stage, and each stage in the series arrangement must provide much less compression than the single stage. The maximum overshoot output from any stage, whether one of the series stages or the single stage, is essentially independent of the input signal level (the non-linear overshoot suppression mechanism limits the maximum overshoot possible).

The staggered series stage arrangement also has less overshoot than the single stage because the first stage in the staggered series stage arrangement, which has to deal with the largest change in signal level, produces the largest overshoot in the series stages. The subsequent stage(s) must process a signal that has already been subjected to the overshoot suppression mechanism of the previous stage. Also, unlike stationary signals which are subject to the normal gain compounding effect of the shifted series stages, high-level overshoots are essentially independent of the input signal level and add linearly among the series stages. The overshoot effect is not compounded when series-connected stages are used. The total overshoot from the staggered series connection of stages is thus significantly smaller than that produced by a single stage.

A further improvement in overshoot suppression is provided by shifting the overshoot suppression threshold levels in the series-connected stages. The effect is to further reduce the overshoot contribution from the stage or stages after the first stage, so that the total overshoot in the staggered series arrangement is further reduced.

Even a further improvement in overshoot suppression is provided by designing the mutually offset steps as double track steps. The overshoot advantage in a dual path compressor is improved when the side path or side chain contribution is reduced relative to the main path contribution. Splitting the impact into several steps achieves this.

Examples of known dual-path circuits can be found in US-PS 3,846,719, US-PS 3,903,485 and US-PS Re 28,426. Analogous circuits are also known which achieve the same results, but where the additional path has properties that are inverse in relative to limiter characteristics, and the additional path's output signal counteracts the main path's signal for compression and amplifies the main path's signal for expansion (US-PS 3,828,280 and US-PS 3,875,537).

Any of these known bilinear circuits can therefore be used as the first and second circuits in the circuit arrangement according to the invention to achieve the advantages that are a natural part of it, and also to provide a good way to establish the desired degree of displacement. This is done by setting the thresholds and. the dynamic ranges of the two additional paths correctly.

As previously mentioned, it is not essential to form the desired form of bilinear characteristic by means of such "double track" techniques. There are alternatives that work with single lanes, as described in e.g. US-PS 3,757,254, US-PS 3,967,219, US-PS 4,072,914, US-PS 3,909,733 and Japanese Patent Application No. 55529/71. Although these alternative circuits are usually not capable of producing as good results as dual-path circuits,

or may be less convenient and thus less economical, they may produce substantially equivalent results. Consequently, these known circuits can also be used as one or more of the circuits in a switching device according to the invention. If desired, one of the first and second circuits can be a double-track circuit and the other a single-track circuit.

The invention will be described in more detail in the following in connection with design examples with reference to the drawings, where fig. 1 shows an example of a set of curves showing complementary, bilinear compression and expansion characteristics, fig. 2 is a block diagram showing the invention in general terms, fig. 3 shows an example of a graphical presentation of the areas with dynamic effect and how these can be separated in series-connected compressors or expanders, fig. 4 shows a further simplified form of fig. 3, fig. 5 shows a series of idealized bilinear characteristic curves illustrating a general technique for shifting the thresholds of series circuits, FIG. 6 shows a connection core of a previously known sliding belt compressor, fig. 7 shows a connection core of a previously known slide band expander, fig. 8 shows a connection core of a modification of fig. 6 and 7, fig. 9 is a curve recording showing the response below the compression threshold for two series compressors and expanders according to an embodiment of the invention, fig. 10 is a curve recording showing the response below the compression threshold for a previously known compressor and expander according to fig. 6, 7 and 8, fig. 11 shows a curve recording of the input-output response as a function of frequency for a compressor having series devices according to an embodiment of the invention, fig. 12 shows a curve recording of the input-output response as a function of frequency for a previously known compressor having a single device, fig. 13-15 show a series of probe tone couriers illustrating the sliding band effect of an embodiment of the invention and of the circuit of FIG. 6 and 8, fig. 16 shows characteristic curves below the compression threshold for a further embodiment of the invention, fig. 17 shows characteristic curves of a similar nature to the curves in fig. 11, but for a further embodiment of the invention, and fig. 18 shows characteristic curves of a similar nature to the curves in fig. 11 and 17, but show excessive bunching.

Examples of bilinear, complementary compression and expansion transfer characteristics (at a particular frequency) are shown in fig. 1 which (for the compression characteristic) shows the low-level part with essentially constant gain, the threshold, the part where dynamic action occurs, the termination point, and the high-level part with essentially constant gain.

Fig. 2 shows the present invention in general terms: A first bilinear compressor 2 receives the input information and supplies its output signal to a series-connected, second bilinear compressor 4 whose output signal is supplied to a noise-laden, information-transmitting channel N. Two series-connected, bilinear expanders 6 and 8 receive the input signal from channel N at the expander 6 and provides a noise reduction system output at the output of the expander 8. The areas of dynamic effect for the series devices are separated or shifted relative to each other within the frequency range that is common to the devices. Although the figure shows two devices on each side of the information channel N, two or more can be used, and the invention envisages two or more series-connected bilinear compressors or expanders. When designed as a complementary noise reduction system, an equal number of series-connected, bilinear compressors and expanders are provided.

The sequence of stages that have special characteristics in the compressor is reversed in the expander. For example, the last stage of the expander is complementary to the first stage of the compressor in all respects, both with regard to steady state and time-dependent, dynamic response (frequency, phase and transient response under all signal level and dynamics conditions) .

An example of a graphical presentation of the separation or displacement for two bilinear devices is shown in fig. 3 which shows compression ratio as a function of input amplitude level (horizontal axis) for a compressor or expander operating at a particular frequency. For clarity, the curves are shown in idealized form. As a practical matter, the curves are somewhat asymmetric in practical implementations of the applicant's noise reduction systems of type A and type B (US-PS 3 846 719 and US-PS Re 28 426 respectively). Curve 12 refers to the dynamic effect of one compressor or expander (the high level stage). Curve 10 applies to a further compressor or expander (the low-level stage) with a separate area of dynamic action. If the high-level stage is first in the series of compressors (second in the series of expanders), curve 12 represents the variations in compression ratio of the first (compressor) stage as a function of the input level of the first stage, and curve 10 represents the variation of the compression ratio of the second (compressor) stage as a function of the input level of the first stage. The upper curves apply to compressors and the lower curves to expanders. In this example, the regions of action in response to input amplitude level are separated, so that the product of the two curves results in a total characteristic with a compression ratio or expansion ratio not exceeding 2:1 (1:2) between the two maximum compression points 10a and 12a (10b and 12b) for the two devices.

Even with two devices in series, the end operating ranges will thus still remain fixed, as the maximum compression and expansion ratios are not increased. beyond the maximum compression and expansion ratios of single devices, and the advantages of single bilinear devices are retained. Any errors that occur within the area of dynamic effect and that are caused by the series-connected devices should thus not exceed the errors with a single device.

Most bilinear devices determine the fixed end regions with constant gain by means of fixed, preset circuit elements, such as resistors and capacitors, which are inherently stable and cannot introduce dynamic errors, waveform distortions and the like. Any dynamically active parts of the circuits can consequently introduce signal errors only in a transitional operating range between de.linear ranges with constant gain.

It should be noted that the dynamic performance of a conventional, logarithmic compressor or expander,

in the preparation in fig. 3 becomes a straight line. For example, line 11 is the characteristic of a 2:1 compressor, and line 13 is the characteristic of a 1:2 expander. In this analysis it is clear that there is no possibility of separation or displacement of the effects of such devices.

For purposes of analysis and to obtain a first order approximation of required threshold levels to provide optimal displacement in accordance with the invention, it is useful to idealize FIG. 3 even more. It is therefore assumed that each compressor (and expander) immediately achieves its maximum compression ratio at a threshold level and maintains this ratio until it reaches a termination point at a higher level where its dynamic action suddenly stops. A series of compressors and expanders manufactured in the same manner as in fig. 3, then appear as continuous, rectangular curves, as shown in fig. 4. As an example, three bilinear, characteristic compressors and expanders are connected in series. The low level device, which is preferably the third compressor (first expander), has the lowest threshold (T^) shown at -62 dB,

and its termination point (F^) at -46 dB which is the threshold (T2) for the intermediate level step. The intermediate level stage has its termination point (F2) at -30 dB which is the threshold (T^) for the high level stage. The high-level stage has its termination point (F^) at -14 dB. All levels are referenced to the total or general input signal. It is further assumed that each stage has a gain of 8 dB and a maximum compression ratio of 2:1.

Fig. 5 shows idealized characteristic curves (total input signal as a function of output signal) for compression based on the example of fig. 4 (the mirror image expansion curves are omitted for clarity). The figure shows how the dynamic effect of each stage acts alongside the effect of the adjacent stage, resulting in a total compression ratio of 2:1 while achieving a compression of 24 dB.

Based on observations of fig. 4 and 5, a single equation states the relationship between threshold level T, termination point F, maximum compression ratio C and gain G for any particular step:

By using this equation, the threshold levels for each step can be determined as a fairly accurate approximation using an iterative method. For example, if a total termination point (F^) of -14 dB is desired with a step gain of 8 dB and a maximum compression ratio of 2, the equation shows that the high-level threshold (T-^) is -30 dB. This value is then used as the end point (F2) for the intermediate level step to determine that its threshold should be -4 6 dB, and so on. Each step is thus referred back to the result for the previous step in this analysis. The calculated threshold is, however, the total threshold referred to the input of the row. To obtain the threshold for a particular circuit referenced to its own input, account is taken of the cumulative signal gain up to this point. For example, threshing-

len for the step with the lowest level in fig. 5 equal to -46 dB when referring to the input of this stage.

The equation can also be solved with regard to the termination point F, the compression ratio C or the gain G. The designer can thus determine his circuit parameters based on his design goals. Such targets may include requirements that the lowest level threshold should be above the noise minimum limit, that the highest level termination point should be sufficiently low to allow the use of overshoot protection, and that the total maximum compression ratio should not exceed a particular value.

In practical circles, the threshold and termination point are not always well-defined points, as is the case in this analysis. As mentioned in the introduction,

the areas in which the mid-level part of the characteristic transitions into the linear low-level and high-level parts may be smooth or sharp depending on the characteristics of the circuits controlling the dynamic effect. In practice, the threshold area for one circuit will thus overlap the termination point area for another circuit.

Consideration of the above equation and of fig. 5 shows that for the special case with a compression ratio of 2:1, half of the threshold shift is provided by means of the signal amplifications for the steps, and that the other half must be provided by means of a changed bias on the control element and/or a changed control amplifier gain (increased gain for lower threshold). For the compression ratios 1.5:1 and 3:1, 1/3 and 2/3 respectively of the displacement is provided in a similar way by means of the step reinforcements, and 2/3 respectively. 1/3 of the displacement must be provided by the control circuits.

On both fig. 1 and 5, 0 dB is a nominal maximum or reference level. In practice, a free height of approx. 10 - 20 dB above the 0 dB level.

As previously mentioned, it is usually preferred that the high-level stage is the first in a compressor series and that the low-level stage is the last. However, a reversed arrangement is also possible. In the reversed case, the first stage control amplifier needs a high gain to achieve the required low threshold. This low threshold then applies even in the presence of high-level signals, which usually leads to poor noise modulation performance for the overall system when it comes to sliding band systems according to the known technique. In this reversed arrangement, each stage must provide sufficient control amplifier gain to achieve the threshold required for that stage. Furthermore, each threshold is essentially fixed and independent of the operation of the other steps. This is a consequence of the fact that the signal gain for each previous step has essentially dropped to one when the threshold is reached for the corresponding subsequent step. The calculation of the thresholds required for optimal displacement in the reversed case is the same as the calculation in the preferred case. However, the threshold of each step referred to its own input will be the same as the total threshold.

In contrast to the reversed situation, in the preferred arrangement (in which the high-level stage is first in the compressor chain and the low-level stage is last), there is a useful interaction between the stage gains and the thresholds. The thresholds for the downstream stages are determined in part by the signal gains of the preceding stages. In a 2-stage system with a low-level gain of 10 dB per stage, the control amplifier gain requirement for the second stage is thus reduced by 10 dB, by virtue of the low-level signal gain for the first stage. When a high-level signal appears, the 10 dB gain for the first stage is eliminated and the threshold for the low-level stage is effectively raised by 10 dB. With sliding band companders, this improves the noise modulation performance of the noise reduction effect.

In the preferred arrangement, the gains for all preceding stages are fully effective up to the threshold of any particular subsequent stage. In contrast to the previously described system with reversed order, the preferred arrangement thus takes the greatest advantage of the effective signal amplifications for the individual steps. This applies: 1. During signal conditions with a very low level

(sub-threshold level), the control amplifier gain requirement for each stage is reduced by an amount equal to the cumulative signal gains of all previous stages. In the example of fig. 5, the control amplifier gain required for the lowest level stage to achieve a threshold of -62 dB is thus reduced by 16 dB compared to that which would be required if the stage operated independently or in the reversed configuration described in front of. Similarly, the control amplifier gain for the mid-level stage is reduced by 8 dB, thus leading to the most economical circuit.

2. A signal-dependent, variable threshold effect

is achieved, so that the noise modulation effects with sliding band steps are reduced. The effective thresholds for the low-level steps are gradually raised with increasing signal level at a particular frequency. With high signal levels (on the high-level linear portion of the transfer characteristic), the effective threshold of the lowest-level stage is raised by a level equal to all low-level (sub-threshold) stage gains up to that point. In the example of fig. 5, the threshold for the lowest level step, normally -62 dB under low level signal conditions, is thus raised by 16 dB, to -4 6 dB, under high level signal conditions. In a similar way, the threshold for the middle level step is raised to

-38 dB.

In a first practical embodiment of the invention, where series slide-bar devices are used, the compressor 2 and the expander 8 in fig. 2 essentially slide belt devices of the standard B type, as set forth in US-PS Re 28,426, while the compressor 4 and the expander 6 have modified response characteristics. It has been found that, with the noise generated by cassette tapes, a usable result is obtained when the second device (in compressor mode) not only has a shifted input amplitude level response, but also has a cutoff frequency of about two to three octaves lower than the cut-off frequency of a standard B-type device. Specifically, the threshold levels on the second device are lowered, both for the syllabic filters/limiters and for the overshoot suppression limiter, to effect offset, and the corner frequency of the fixed filter is lowered by two to three octaves.

Details of the B-type circuit are shown in fig. 6, 7 and 8 which are the same figures as fig. 4, 5 and 10 in US-PS

Re 28,426, and further details of these circuits, their operation and theory are set out in the said patent document.

The following description of fig. 6, 7 and 8 are taken from US-PS Re 28,426.

The circuit of fig. 6 is specially designed for installation in the recording channel of a consumer tape recorder, since two such circuits are necessary for a stereo recorder. The input signal is supplied at a terminal 10 to an emitter follower stage 12 which provides a low impedance signal. This signal is supplied firstly via a direct through main path which is constituted by a resistor 14, to an output terminal 16, and secondly via a further path whose last element is a resistor 18 which is also connected to the terminal 16. Resistors 14 and 18 add the output signals from the main path and the additional path to provide the required compression law.

The further path consists of a fixed filter 20, a variable cut-off filter 22 comprising a FET transistor 24 (these constituting the filter/limiter unit), and an amplifier 26 whose output is coupled to a dual diode limiter or clipper 28 and to the resistor 18. The non-linear limiter suppresses overshoot of the output signal with suddenly increasing input signals. The amplifier 26 boosts the signal in the further path to such a level that the knee in the characteristic of the limiter or overshoot suppressor 28, which consists of silicon diodes, is effective at the correct signal level during transient conditions. The effective threshold of the overshoot suppressor is slightly above the threshold of the syllabic filter/limiter unit. The resistors 14 and 18 are proportioned so that the necessary, compensatory degree of attenuation is then provided for the signal in the further path.

The output from the amplifier 2 6 is also connected

an amplifier 30 whose output signal is rectified by means of a germanium diode 31 and integrated by means of a smoothing filter 32 to provide the control voltage for the FET transistor 24.

Simple RC filters are used, although equivalent LC or LCR filters could be used. The fixed filter 20 provides a cut-off frequency of 1700 Hz, below which tapering compression takes place. The filter 22 comprises a series capacitor 34 and a shunt resistor 36 followed by a series resistor 38 and the FET transistor 24, with the source-drain path of the transistor connected as a shunt resistor. During rest states with zero signal on the FET transistor's 24 control electrode. the transistor is choked and exhibits essentially infinite impedance. The presence of the resistor 38 can then be ignored. The cutoff frequency of the filter 22 is thus 800 Hz, i.e. a frequency that is significantly below the cutoff frequency of the fixed filter 20.

When the signal on the FET transistor's 24 control electrode increases sufficiently for the transistor's resistance to drop to less than, e.g. 1 k£2, resistor 38 effectively shunts resistor 36, and the cut-off frequency rises so that the filter's passband narrows markedly. The increase in cut-off frequency is of course a gradual effect.

The use of a FET transistor is convenient as such a device, within a suitable, limited range of signal amplitudes, essentially acts as a linear resistance (for signals of one or the other polarity) whose value is determined by the control voltage on the control electrode.

The resistor 3 6 and the FET transistor 2 4 are fed back to an adjustable outlet 4 6 on a voltage divider which includes a temperature-compensating germanium diode 48. The outlet 46 enables the compression threshold of the filter 22 to be regulated.

The amplifier 26 comprises complementary transistors which provide high input impedance and low output impedance.

Since the amplifier drives the diode limiter 28, a limited output impedance is required and is provided by a coupling resistor 50. As already mentioned, the diodes 28 are silicon diodes and have a sharp knee around 1/2 volt.

The signal on the limiter and thus on the resistor

18 can be short-circuited to earth by means of a switch 53. when it is necessary to put the compressor out of action.

The amplifier 30 is an NPN transistor with an emitter time constant network 52 which provides increased amplification at high frequencies. Strong high frequencies (eg a cymbal rattle) will therefore lead to a rapid narrowing of the band in which compression takes place, so that signal distortion is avoided.

The amplifier is connected to the smoothing filter 32

via the rectifying diode 31. The filter comprises a series resistor 54 and a shunt capacitor 56. The resistor 54 is shunted by a silicon diode 58 which allows rapid charging of the capacitor 55 for fast attack, combined with good smoothing under stationary conditions. The voltage on the capacitor 56 is supplied directly to the control electrode of the FET transistor 24.

A complete coupling core of the complementary expander is shown in fig. 7, but a full description is not necessary as essentially the entire circuit is identical to fig. 6, and component values are therefore for the most part not shown in fig. 7.

The differences between fig. 6 and 7 are as follows:

In fig. 7, the further path derives its input signal from the output terminal 16a, the amplifier 26a is inverting, and the signals combined by the resistors 14 and 18 are supplied to the input (base) of the emitter follower 12 whose output (emitter) is connected to the terminal 16a. To ensure low drive impedance, the input terminal 10a is connected to the resistor 14 via an emitter follower 60. Appropriate precautions must be taken to prevent bias getting in the expander.

The amplifier 26a is made inverting by taking the output from the emitter instead of the collector of the second (PNP) transistor. This change involves shifting the 10k£3 resistor 62 (Fig. 6) from the collector to the emitter, which automatically provides a suitable output impedance to drive the limiter. The resistor 50 is therefore left out

fig. 7.

It should be noted that when trimming a complete noise reduction system, it is important to have equal signal levels on the transistors' 12 emitters in both compressor and expander. Measuring terminals M are shown connected to these emitters.

Fig. 8 shows a preferred circuit for replacing the circuit between points A, B and C in fig. 6 and 7. When the FET transistor 2 4 is choked, the second RC network 22 is inactive, and the first RC network 20 then determines the response of the further path. The improved circuit combines the phase advantages of having only a single RC section under quiescent conditions, with the attenuation characteristic of 12 dB per octave for a two-section RC filter under signal conditions.

In the practical circuit, where FET transistors of the MPF 194 type are used, the 39 kQ resistor 36a is necessary to provide a limited source impedance to work into the FET transistor. In this way, the compression ratio at all frequencies and levels is kept at a maximum of approx. 2. The 39 kfi resistor 36a serves the same compression ratio limiting function in the improved circuit as the resistor 36 in the circuit of FIG. 6

or 7. This resistor also provides a low-frequency path for the signal.

Modifications of fig. 6, 7 and 8 must be described

in the following.

In the first practical embodiment of the invention, the compressor 4 and the expander 6 in fig. 2, as mentioned above, arrangements of the one in fig. 6, 7 and 8 showed type with modified characteristics. The changed cut-off frequency and reduced threshold are respectively achieved by modifying the characteristics of the fixed filter (the fixed filter 20 in Fig. 6) and also the gain of the control amplifier by changing its pre-correction characteristics (the emitter-time constant network 52 in the amplifier 30 in Fig. 6 ). The threshold of the overshoot suppressor is lowered by applying suitable DC biases (in the forward direction) to the diodes 28. The impedance of the variable filter (the variable filter 22 of Figs. 6 and 8) is left

substantially unchanged to maintain a suitable fit to the characteristics of available voltage controllable variable circuit elements. Suitable modifications of the B-type sliding band circuit shown in FIG. 6, 7 and 8, is to change the value of the resistance of 3.3 k£2 in the fixed filter 20 to a value of 18 kN to lower its cut-off frequency two to three octaves. To increase the control amplifier gain, the value of the capacitor in the amplifier 30 emitter-time constant network 52 is increased from 0.15 to 0.60 µF (or from 0.1 to 0.4 µF if the implied value of

0.1 uF is used). Preloads of approx. plus and minus 1/4 volt in the forward direction are applied to the silicon diodes 28, so that the overshoot suppression level is reduced by several decibels.

The variable filter 22 has an all-pass frequency characteristic in response to the quiescent control voltage, and thus the total filter cut-off is lowered by two to three octaves. Increasing the capacitor value in the control amplifier's 30 emitter network increases the amplifier's gain at any given frequency. As described above and in US-PS Re 28,426, the variable RC filter 22's cut-off frequency rises as the control voltage (from the amplifier 30, the rectifier 31 and the smoothing filter 32) increases. Thus, with larger capacity values in the network 52, the variable filter reacts by moving upwards in frequency from its resting value in response to signals with a lower level, so that the level response or threshold is shifted in relation to the level response or threshold in the unmodified B-type circuit.

The level response can be shifted in numerous ways in addition to changing the control amplifier's emitter network. Other possibilities include changing the bias voltage of the control element, changing the gain of the control amplifier in some other way, changing the relative signal levels between the filter path and the control signal derivation path, and so on.

Certain details of the circuit in fig. 6, 7 and 8 have evolved over the years and more modern forms of the circuit have been published and are well known in the art. The reference to the special circuit in US-PS Re 28,426 is used for convenience in the preparation. Fig. 9 shows a real curve recording plot of response below the compression threshold for the two series-connected compressors, the first being modified as described above, and the expander response is also shown. Compare this figure with fig. 10 (corresponding to FIG. 12 of US-PS Re 28,426) which shows a real curve recording plot of response below the compression threshold for a single compressor or expander according to FIG. 6, 7 and 8. Fig. 11 shows a chart recording of the input-output response of the series compressors as a function of frequency. Inspection of the response plots shows the two dynamic ranges for the curves indicating the two shifted ranges of action. Although the observability of the dynamic ranges in these curves is useful in showing the offset effect of the devices, in practice it is preferred that the curves be as smooth as possible, without noticeable dynamic ranges or "bumps". Parallel lines A and B are drawn through the threshold areas, line A referring to the standard circuit and line B to the modified circuit. Compare these curves with fig. 12

(corresponding to Fig. 12 of US-PS Re 28,426) showing similar response curves for a single, unmodified B-type scroll compressor. Fig. 11 shows that the compressor which includes series devices provides essentially twice as much compression which is distributed over a larger frequency and level range.

The variable band action of the series devices with staggered action can be seen in fig. 13 and 14 showing a chart recorder probe tone response of the series-connected compressors. Compare these figures with fig. 15 (corresponding to Fig. 15 of US-PS

Re 28,426) which shows a real curve recording obtained from the circuit of FIG. 6 incorporating the circuit of FIG. 8. The variable band effect is shown by recording the compressor frequency response by means of a low-level probe tone (whose level is below the compressor threshold) in the presence of a high-level signal, the probe or sensor tone being detected at the output of the compressor by means of a tracking filter. The high-level signal causes the compressor circuit to operate, as the diagram shows the effect on the filter's turnover frequency.

Fig. 13 shows the response for one probe tone at

-65 dB and 200 Hz signal tones at levels ranging from -28 dB and below to +10 dB. Fig. 14 applies to a 500 Hz signal tone at levels ranging from -34 dB

and below to +10 dB.

In a further practical embodiment of the invention which provides improved performance, both the compressor 2 and the expander 8 in fig. 2 modifications of standard B-type devices. Both series devices have their corner frequencies lowered by two octaves to provide a sharply rising low-level response characteristic. Displacement of the dynamic effect is provided by reducing the thresholds (both syllabic and overshoot suppression) of the other (in compressor mode) device.

A distinctive feature and a useful advantage of the invention is that the frequency responses for the individual circuits are put together or mixed. If the most sharply or suddenly rising noise reduction characteristic is desired, this is achieved by using circuits that have the same low-

in

level (rest) frequency response characteristics.

Consequently, in the improved embodiment, the selection of identical filter characteristics results in approx. 2 octaves below the characteristic of the normal B-type device,

1 a characteristic that rises rapidly above approx. 300 Hz. The system is thus able to achieve a significant noise reduction in the critical range from 300 Hz to 2 kHz, a range in which band noise is noticeable as soon as noise above 2 kHz has been reduced. There is a negligible, audible noise contribution from the band below approx. 300 Hz. By providing only minimal noise reduction effort below 300 Hz, the system avoids the manipulation of fundamental signal frequencies and improves the system's complementarity in practical tape recorders which may for example have frequency response errors due to head unevenness and the like. By avoiding the compression of low frequency signals, system compatibility is further improved due to the fact that the boost of the low frequency signals would

result in annoying rumble and bass boost when encoded tapes are played on systems that do not have complementary expanders.

Referring again to fig. 6 and 8, the resistance in the fixed filter 20 in both of the two series devices in the practical embodiment under discussion is changed from

in

3.3 kfi to 13 kfi, which causes the overall lower cutoff frequency of filters 20 and 22 to shift approx. 2 octaves lower to approx. 375 Hz. In the second arrangement, the capacitor in the emitter network 52 of the control amplifier 30 is increased in value by a factor of approx. 4, as in the embodiment described above. This results in a shift in threshold levels of roughly 10 - 15 dB (depending on signal level and signal frequency). Appropriate biasing is applied to the diode limiter circuit 28 to lower the overshoot suppression level.

In a modification of the last described, practical embodiment, the capacitor 34 in the filter 22 can be increased in value to 0.01 pF to promote the consistency of characteristics from unit to unit and to improve the noise modulation characteristics. As a result of the essentially equal time constants for the fixed filter 20 and the variable filter 22, the arrangement in this case becomes equivalent to a single-pole, variable filter, and the fixed filter can be eliminated. In this case, the resistor 36a (which has a value of 47 kfi in modern forms, for the B-type circuit) is placed in parallel with the source-drain path of the FET transistor 24 to provide a quiescent corner frequency of about

375 Hz. However, it is desirable to retain the fixed filter in the high-level circuit, so that the circuit can be switched to work alone as a standard B-type circuit.

As a practical detail, a consumer product incorporating the practical systems just described would be compatible with existing non-coded and B-type coded "software" (eg tapes and FM radio programs).

The improved systems include a standard B-type device and can therefore be switched to work as a B-type device for full compatibility. On the other hand, when recorded tapes encoded with the improved system become available, existing B-type consumer systems may observe excessive high-frequency information or "clarity", which can be addressed by adjusting the high-frequency tone control in the same way as non-equipped consumer systems currently processing B-type encoded software.

The standard B-type circuit described in US-PS Re 28,426 has a maximum compression ratio of approx. 2:1. This compression ratio has proven to be a good practical choice for consumer cassette tape compander systems.

In the series-connected circuits in the designs described above, each circuit maintains a maximum compression ratio of approx. 2:1, and the maximum compression ratio for the total combination of series circuits is approx. 2:1 at most input signal levels and frequencies. In practical implementations, however, it is difficult to avoid somewhat larger conditions, e.g. 2.5:1, in a small range of levels and frequencies. This can be accepted if the compression ratio is not more than approx. 2.5:1 (or about 1 1/4 times the compression ratio for each circuit), and .if the range of levels and frequencies in which it occurs is not large.

Another particular embodiment of the invention shown generally in FIG. 2, is to design one compressor and expander as a split belt device, as described in US-PS 3,846,719 and US-PS 3,903,485, and the other compressor and expander as a sliding belt device. A suitable split-band or multi-band arrangement is described in "Journal of the Audio Engineering Society", Vol. 15, No. 4, October 1967, pp. 383 - 388. Split-band arrangements conforming to the parameters of this article have become well known as A -type devices.

In a practical embodiment, an A-type compressor receives

in

sor a flat input signal and supplies its output signal to a specially adapted sliding band device. It is particularly advantageous to place the A-type device so that it receives an unprocessed input signal as it is designed to process a flat input signal. Placing the sliding band device first would have the disadvantage of changing the flat input signal to a form less suitable as an input signal for the A-type device. On the reproduction side, the sliding band expander receives the signal from channel N, processes it and supplies it to the A-type expander.

Fig. 16 shows similar curves as fig. 9 for the low-level signal response for: an A-type compressor alone, the sliding band compressor alone!, and the combined compressor response. The expansion response curves are complementary in a similar way as in fig. 9. The A-type device provides a compression of 10 dB up to approx. 5 kHz, above which frequency the increase in level rises steadily to 14 dB

at 15 kHz. This rising response of the A-type characteristic is used to desensitize the sliding band characteristic at high frequencies (see the high frequency portion of the "sliding band" curve in Fig. 16). This is advantageous for reducing the effects of high frequency channel response uncertainties which will be described in more detail below. The combined response curve thus climbs steadily up to 20 dB, where it remains essentially flat until approx. 14 kHz where it falls

of. The slider device is designed to have operating thresholds and resulting areas of dynamic action that are well clear of the thresholds and areas of the A-type circuit.

Fig. 17 shows a series of response curves at different levels for the series-connected A-type and sliding belt compressors. These curves display the same type of information as fig. 11. The shaded area C generally indicates the dynamic ranges resulting from the action of the A-type device, and the shaded area D indicates the action of the sliding band device. This arrangement results in a maximum compression ratio which at any level or frequency does not exceed approx. 2:1

and is therefore relatively free from error amplification effects in practical band tuning channels.

It will be understood that a standard A-type device for example purposes is connected in series with a special sliding belt device. In principle, however, the A-type device can be modified to shift its areas of dynamic action to provide the best fit to the areas of action of the sliding belt device.

The exact degree of displacement required in this and other designs described above will depend on the parameters of the signal processing devices used. The purpose of shifting the areas of dynamic effect is to minimize bunching effects in the response curves. Bunching is an indication of large compression or expansion ratios. See, for example, fig. 18 which shows excessive bunching, i.e. at some frequencies and levels a 10 dB change in input level results in a 2 1/2 dB change in output level, a ratio of 4:1. Optimally, with proper offset, a ratio of 2:1 is never significantly exceeded in a cassette compander system over most of the range of levels and frequencies. In other types of transmission systems, higher compression ratios may be acceptable.

Claims (13)

1. Switching device for modifying the dynamic range of an input signal, comprising a first circuit with a bilinear characteristic consisting of a low-level part with essentially constant gain up to a threshold level, an intermediate level section, above the threshold, with varying gain which provides a compression ratio or expansion ratio varying between one at the ends of the intermediate level section and a maximum value between those ends, <p>g a high level section with substantially constant gain which is different from the low level section gain, whereas in front of the first circuit, when it concerns a compressor for broadcast transmitters, a broadband compressor or broadband limiter can be arranged if desired, characterized in that the first circuit (2, 6) is followed by at least one second circuit (4, 8) which also has a bilinear characteristic within a frequency range that is common to the circuits, the mid-level parts of the circuits' characteristics within a frequency range that is common to circuits, are mutually offset in such a way that a gain change is achieved over a greater range of intermediate input levels than for any of the circuits individually, and an increased difference between the gains at low and high input levels, but with an overall, maximum compression or expansion ratio which, by virtue of the displacement, is in the main case no greater than the ratio for a single circuit, and a total characteristic which is also bilinear. 2. Coupling device according to claim 1, characterized in that each of the circuits is arranged for a maximum compression or expansion ratio of essentially 2:1, and the maximum ratio everywhere does not significantly exceed 2:1. 3. Coupling device according to claim 1 or 2, characterized in that for a compressor the threshold for each of the other circuits (4) is lower than the threshold for the preceding circuit (2, 4), and that for an expander the threshold for each of the other circuits (8) higher than the threshold of the previous circuit (6, 8). 4. Coupling device according to one of claims 1-3, for modifying the dynamic range of audio signals, and where each circuit contains an overshoot suppressor (28) with a threshold level, characterized in that the circuits are arranged with the threshold levels gradually shifted between them, so that a reduction of the overshoots for the overall switching device is achieved. 5. Coupling device according to claims 3 and 4, characterized in that the overshoot suppressors (28) in the individual circuits in the compressor and the expander respectively are arranged with the threshold levels mutually offset in the same way as the threshold levels for the successive circuits. 6. Coupling device according to one of claims 1-5, in which at least one of the circuits consists of a variable filter (22) which provides a boost or cut in a high-frequency or low-frequency region of the signal band, and which responds to signals in this region in such a way that the filter cut-off frequency is caused to slide in a direction which narrows the raised or lowered region, the coupling preferably for audio signals having a rectifying, smoothing and amplifying control circuit (30, 31, 32) which provides a control signal to a controlled impedance device (24) in the filter, to effect the displacement of the filter cutoff frequency, characterized in that each circuit comprises a variable filter (22). 7. Switching device according to claim 6, characterized in that the control circuit (30, 31, 32) in each variable filter (22) has a different gain, in order to set the different thresholds for the circuits in this way. 8. Coupling device according to claim 6 or 7, characterized in that in each of the circuits the rest limit frequencies for the variable filters (22) are essentially the same. 9. Coupling device according to claim 6, 7 or 8, in which the raising or lowering is provided in a high-frequency range, characterized in that the variable filters (22) have resting cut-off frequencies that lie in the range from 300 to 400 Hz. 10. Switching device according to one of claims 1 - 9, in which at least one of the circuits is a double-path circuit comprising a main path (14) which is linear with respect to the dynamic range, a combination circuit in the main path, and a further path whose input is connected to the additional path's input or output and whose output is connected to the combination circuit, the additional path providing a "signal" which, at least in an upper part of the frequency range, enhances or opposes the main path's signal via the combination circuit, but which is so limited that the additional path's signal in the upper part of the input dynamic range is smaller than the main path signal, characterized in that each of the circuits (2, 4; 6, 8) is a dual path circuit. 11. A coupling device according to one of claims 1-10, characterized in that the other or every second circuit (4, 8) is arranged to be affected by a modified signal, e.g. a signal with changed spectral content in the event that an affected circuit (2, 6; 4, 8) changes the signal's spectral content. 12. Coupling device according to claims 1-5 or claim 10 depending on claims 1-5, characterized in that each circuit (2, 4, 6, 8) has a control circuit (30, 31, 32) which provides a control signal to control the circuit's expansion or compression, each control circuit reacting to signals in the respective circuit (2, 4 , 6, 8) in which it is located. 13. A coupling device according to one of claims 6-9 or claim 10 depending on claims 6-9, characterized in that the control circuit or each control circuit (30, 31, 32) reacts to signals in the respective circuit (2, 4, 6, 8 ) in which it is located.