AU550201B2 - Waveform clipping circuits - Google Patents

Description

2r

WAVEFORM CLIPPING CIRCUITS

10

. This invention relates to speech signal processing and in particular to soft clipping circuits especially for an electronic circuit network in a telephone subscriber's instrument. 15 Conventional electromechanical telephone networks _are inherently slow and, even under overload conditions, generate only minor harmonic outputs. In contrast the new electronic networks are fast and, when overload and consequent speech waveform clipping occurs, 20 they produce an excessive proportion of^'harmonics that are outside-.- he voice band (typically 300 Hz to^' 3.4 kHz). These harmonics are transmitte 'with little attenuation through the hybrid balance network, which is tuned to reject signals in the voice band, to the 25 instrument receiver thus resulting in poor side-tone performance. Furthermore, in order for the hybrid to remain balanced and the return loss to remain greater than -14db, as required by most telephone administrations, the impedance of the instrument as .30 'seen' by the line should be 600 ohms (900 ohms in USA)

&- irrespective of the signal amplitude. When electronic subsets at present in use are overloaded causing clipping the output transistor stage can be ^*driven into saturation. When this occurs the impedance 'seen' by the 35 line falls to that of a saturated transistor, i.e. only a few ohms. Attempts have been made to prevent output transistor saturation by the use of clipping diodes.

S ^U*"BSTITUTE SHEET ' However the impedance of the diodes then shunts the 600 ohm output impedance causing the aforementioned problems of impedance changes, hybrid imbalance and the generation of out of band harmonics. The problem of harmonic generation may be overcome by the use of circuits which soft clip the speech waveform in contrast e.g. to diode circuits which produce hard clipping. A soft clipping circuit does not have a sharp cut-off level but progressively reduces the overall voice channel gain beyond a predetermined threshold amplitude. This provides a smooth transition from high to low gain thus preventing 'splatter'. When soft clipping is applied to a signal substantially all the power is concentrated in the fundamental and the lower order harmonics, the latter being in or close to the voice frequency band. Thus rejection to the receiver path by the hybrid network is still obtained.

Soft clipping alone however does not overcome the major problems of changes in the output impedance to the line and in particular the reduction of impedance caused by output stage saturation. The latter effect can be overcome by ^"suitable choice of the .circuit operating parameters which ensure that the waveform is always below the amplitude at which saturation occurs. The problem of impedance changes related to gain transformations has however proved intractable with conventional telephone circuits which employ negative feedback for gain regulation and thus suffer from an impedance reduction as the gain is reduced. According to one. aspect of the present invention there is provided a speech processing circuit e.g. for an electronic telephone subset, said circuit including a variable gain element whereby soft clipping of a speech waveform may be effected, and negative feedback means, coupled between the circuit output and the variable gain element, whereby the gain of said element is controlled such that the circuit output

impedance is maintained substantially independent of an input speech signal amplitude.

According to another aspect of the invention there is provided a speech processing circuit e.g. for an electronic telephone subset, said circuit including a continuously variable gain input stage, a signal amplifier coupled to the input stage, an output stage coupled to the amplifier, and rectifying feedback means coupled between the output stage and the input stage, the circuit being such that, at predetermined positive and negative output signal threshold levels within the saturation levels of the output stage, the feedback means generates a feedback signal whereby the gain of said input stage is progressively reduced. According to a further, aspect of the invention there is_^.provided a speech processing circuit e.g. for an electronic telephone subset, said circuit including first and second long tailed transistor pairs whereby speech signals are soft clipped, an output amplifier stage coupled to said long tailed pairs, and negative feedback means coupled between long tailed pairs, wherein the transistors of said pairs are so constructed that gain maxima of said pairs are symmetrically disposed about the speech signal zero level and spaced from that level by a voltage corresponding to the onset of soft clipping, and wherein said feedback means is arranged to feed the tail circuits of said long tailed pairs with respective first and second currents whereby the combined output impedance of the long tailed pairs is maintained substantially constant.

The arrangement operates by progressively attenuating the waveform as a predetermined threshold level is approached. This not only substantially prevents the generation of out of band harmonics but also provides a much higher sound quality than that produced by conventional hard clipping. It has in fact been found that the human ear responds primarily to the rate of the zero crossings of a waveform. Thus soft clipping of an audio signal gives very little subjective impression of reduction of loudness. Furthermore, soft clipping of a signal reduces the amplifier power necessary to provide audibility e.g. in hearing aid applications.

The arrangement described herein overcomes the disadvantages of previous telephone circuits in that means are provided whereby the circuit output impedance, i.e. the impedance 'seen' by the line, is maintained substantially constant for all input signal levels.

Embodiments of the invention will now be described with reference to the accompanying drawings in which:- Fig. 1 is a schematic circuit diagram of a soft clipping circuit embodying the techniques described herein;

Fig. 2- illustrates the characteristic of a feedback amplifier employed in the circuit of Fig. 1;

Fig. _,3 shows one form of soft clipping circuit arrangement; -

Fig. 4. shows a modification of the circuit of Fig. 3; Fig. 5 is a circuit diagram of an alternative soft clipping circuit;

Figs. 6a_ and 6a_ illustrate the gain characteristic of the variable gain- element of the circuit of Fig. 5; ' and Fig. 7 illustrates the signal handling characteristic of the circuit of Fig. 5.

Referring to Fig. 1, the schematic circuit shown illustrates the soft clipping technique employed herein. The circuit, which is intended primarily for use in a telephone subset, includes a speech signal input stage I/P having a controllable gain, typically less than unity. The output of this input stage is fed to one input amplifier AMP 1 the output of which drives the base of an output transistor TR1. The.other terminal of the amplifier AMP 1 is coupled to the emitter -circuit of the transistor e.g. via resistor network Rl, R2 and R3. Output to the line is taken from the collector of TR1 and, to ensure correct matching to the line impedance, an impedance Zl (typically 600 Ω) shunts the collector - to the circuit ground.

In order to provide impedance matching to the line under all input signal level conditions it is essential that the input drive to the base of TRl is maintained below the level at which that transistor is driven into saturation with consequent shunting of impedance Zl. This is achieved via a rectifying amplifier AMP 2 coupled to the circuit output, i.e. TRl collector. The characteristic of this amplifier is shown in Fig. 2. For input voltages below a positive or above a negative threshold value the amplifier output is zero. If however these thresholds are exceeded a rapidly increasing positive output signal is produced. This signal is fed .to a control input of the controllable gain stage thereby reducing the gain o'f^" that stage. The rectifying amplifier input threshold is set at a level below that at which saturation of transistor TRl occurs thus ensuring that the drive to TRl base is suitably limited.

It is of course necessary that the gain of the input stage should transform smoothly from a higher gain to a lower gain state and vice versa so as to prevent hard clipping of an input 'signal.

It will of course be understood that the term gain as employed herein is understood to include an amplification of less than unity and, in some instances, a zero amplification. A circuit that achieves soft clipping and prevents output stage saturation is shown in Fig.^' 3. In this circuit the controllable gain input stage comprises first and second transistor long tailed pairs, TR31,TR32 and TR33,TR34 fed from a current mirror CM31. Constant current sources CC31 and CC32 generating currents 131 and 132 respectively are coupled one in each common emitter circuit.

Input speech signals are fed to the base of transistor TR31 of the first pair whilst the base of the input transistor TR33 of the second pair is grounded, i.e. the second pair receives a zero signal. The outputs of the two pairs, i.e. the collectors of TR32 and TR34, are commoned and are fed via an amplifier AMP 31 to the base of an output transistor TR5 the collector of which is, in use, coupled to the line. Impedance matching to the line is provided by shunt impedance Z31. It should be noted that impedance Z31 may. in some applications include a reactive element.

The circuit of Fig. 3 is arranged such that, for input signals below the threshold level at which soft clipping is intended to^' commence, the major proportion of the signal generated at the output common collectors of _.the two long tailed pairs derives from the first pair, i.e. the zero signal secon'd^'pair has only a small effect on the signal output and the stage is thus in its higher gain condition. When the output signal to the line exceeds the threshold of the rectifying amplifier AMP 32 that amplifier generates an output current whose magnitude increases in correspondence with the signal level above the threshold. This current is fed into the tail circuit of the first long tailed pair thus reducing the current through that pair. At the same time a substantially equal and opposite current generated by inverter INV 1 flows from the tail circuit of the second long tailed pair. This increases the current through the transistors of the second pair. The net effect is to reduce the proportion of the stage output due to the first long tailed pair and to increase the output due to the second pair thus producing an effective reduction in stage gain. The sloping characteristic of the amplifier AMP 32 ensures that this reduction in gain is effected-.smoothly thus providing effective soft clipping of a speech waveform.

Fig. 4 shows a modification of the soft clipping circuit of Fig. 2. In this arrangement the - variable gain stage, as before, comprises first and second long tailed transistor pairs TR41.TR44; TR42,TR43, the first of which receives the input speech signal. The tail circuit of this first pair includes a constant current generator CC41 whereby the signal transfer characteristic of that pair is defined. The input of the second pair, TR42,TR43 is grounded so as to produce a zero output. The stage output to amplifier AMP 41 is thus, as before, a weighted combination of the output of each pair, the weighting being determined by the selective tail currents. Gain control and hence soft clipping is provided by a feedback current from a rectifying amplifier AMP 42, as described with reference to Figs. 2 -and 3, which provides tail current for the second pair via inverter INV 41 whenever the output signal at the colle'ctor of transistors TR45 exceeds a predetermined threshold. This increases the weighting of the second pair component, i.e. the zero signal, in the combined output to amplifier AMP 41 thus providing a smooth soft clipping action. Again the threshold of amplifier AMP 42 is such that the output transistor TR45 is never driven into saturation. It will be apparent to those skilled in the art that the gain characteristic of a transistor long tailed pair varies with signal amplitude and that, in some applications, it may be desirable to provide linearisation of the gain characteristic-. A circuit which embodies this linearisation is shown in Fig. 5. Referring now to Fig. 5, the circuit shown in Fig. 5 includes a variable gain input stage via which

SUBSTITUTE SHEET speech signals are fed to an output amplifier, and a feedback amplifier whereby the gain of the input stage is controlled. The variable gain stage comprises first and second transistor long tailed pairs TR51,TR54 and TR52,TR53 fed from a current mirror CM and each of which is provided with a constant current source S,, S₂, coupled in the tail of the pair.

The transistors TR51,TR54 and TR53,TR52 of each pair are so designed that they have an emitter area mismatch in the ratio x : 1. As can be seen from Fig. 6a_ this mismatch separates the signal/gain characteristics of the two pairs by a distance V = 2 jt_^ log x

_. q where k is Boltzmann's constant, t is the absolute temperature and q is the electronic charge. Fig. 6b_ shows the resultant gain characteristic of the two long tailed pairs. This characteristic has a substantially constant gain portion extending %V either side of the signal zero crossing. Beyond this constant gain region the characteristic falls smoothly away to zero. The stage is thus .linear in operation for input signals below the soft clipping threshold.

The output of the variable gain stage is fed to an output stage which, typically, comprises an amplifier AMP 51 driving an output transistor TR55 the collector load of which Z51 presents the required impedance to the line. Typically this impedance is 600 ohms, or 900 ohms in USA. . .

A proportion of the output_^ signal is fed to one input of a two input amplifier AMP 52 having first and second current outputs. The other input of amplifier AMP 52 is connected to a reference potential, e.g. the circuit ground. The outputs of the amplifier, i.e. currents la and Ilo are applied respectively to the tail circuits of the long tailed pairs TR51,TR54 and

TR52,TR53. It should be noted that the sum of the two currents la and lb remains substantially constant. Thus,

SUBSTITUTE SHEET whatever current is added (subtracted) at the tail circuit of one pair is subtracted (added) at the tail circuit of the other pair and the total current -passed by the two pairs in combination remains constant. In this way the output impedance to the line of the whole circuit including the feedback loop is maintained at a substantially constant value.

The soft clip voltage threshold where the major gain regulation occurs is arranged substantially equal to the voltage level (positive or negative) where the peak of the gain (g ) characteristic (Fig.- 6a_) occurs and where the corresponding long tailed pair is in balance. When this occurs any feedback in the loop is "notched out" as show by the negative feedback gain curves in Fig. 7. Hence the cur.rent in the corresponding long-tailed pair can be reduced thus reducing the forward gain without causing a change in the output impedance.

The design of the circuit is determined from the specification requirements of the end user, e.g. the telephone administration, wherein the output signal level for the onset of soft clipping is' defined. From that signal level the emitter area ratios of the long tailed pair transistors can be defined. The feedback characteristic is -also defined so that the feedback notch coincides with this soft clip level as indicated in Fig. 7. The circuit parameter calculations involved will be apparent to those skilled in the art.

Advantageously the arrangements described herein may comprise or form a part of an integrated circuit. In particular such arrangements may be used in a telephone subscriber's instrument to provide clipping of input speech without the generation of harmonics. The arrangement may also be employed as part of a circuit for a hearing aid, as a soft clipped waveform provides a high degree of audio perception together with a relatively low power input.

SUBS I TE SHEE

Claims (9)

CLAIMS : -

1. A speech processing circuit e.g. for an electronic telephone subset, said circuit including a variable gain element whereby soft clipping of a speech waveform may be effected, and negative feedback means, coupled between the circuit output and the variable gain element, whereby the gain of said element is controlled such that the circuit output impedance is maintained substantially independent of an input speech signal amplitude.

2. A speech processing circuit e.g. for an electronic telephone subset, said circuit including a continuously variable gain input stage, a signal amplifier coupled to the input stage, an output stage coupled to the amplifier, and rectifying feedback means coupled between the output stage and the input stage, the circuit being such that, at predetermined positive and negative output signal threshold levels within the saturation levels of the output stage, the feedback means generates a feedback signal whereby the gain of said input stage is progressively reduced.

3. A speech processing circuit^' e.g. for an electronic telephone subset, said circuit including first and second long tailed transistor pairs whereby ^' speech signals are soft clipped, an output amplifier stage coupled to said long tailed pairs, and negative feedback means coupled between the output stage and the long tailed pairs, wherein the transistors of said pairs are so constructed that gain maxima.of said pairs are symmetrically disposed about the speech signal zero level and spaced from that level by a voltage corresponding to the onset of soft clipping, and wherein said feedback means is arranged to feed the tail circuits of said long tailed pairs with respective first and second currents whereby the combined output - impedance of the long tailed pairs is maintained substantially constant.

SUBSTITUTE SHEET

4. A circuit as claimed in claim 3, wherein said negative feedback means includes a rectifying amplifier adapted to generate an output only for signals above a predetermined threshold level. • 5. A circuit as claimed in claim 3 or 4, and including feedback means provided with a notch characteristic, the location of the notches of the characteristic corresponding to the soft clipping level of the circuit.

6. A circuit as claimed in any one of claims 1 to 5 and in the form of an integrated circuit.

7. A speech processing circuit substantially as described herein with reference to the accompanying drawings.

8. A telephone subset provided with a speech processing circuit as claimed in any one of claims 1 to 6.

9. A method of .speech signal processing substantially as described herein with reference to the accompanying drawings.