DK145209B - PROCEDURE AND APPARATUS FOR REDUCING THE QUANTIZATION NOISE IN IMPULSE CODE MODULATION SYSTEM - Google Patents

Description

(19) DENMARK ^ Ra /

| J | (12) PUBLICATION MANUAL OR 145209 B

DIRECTORATE OF

THE PATENT AND TRADEMARKET SYSTEM

(21) Application No. 1358/76 (51) int.d, 3 ty Ø3 χ in 3/0 ^ (22) Filing date 26 Mar 1976 (24) Race day 26 Mar 1976 (41) Aim. available Nov. 7 1976 (44) Presented 4 Oct. 1982 (86) International application no.

(86) International Filing Day (85) Continuation Day ~ (62) Stock Application No. "

(30) Priority 6 May 1975 * 5807/75, CH 6 May 1975, 5808/75, CH

(71) Applicant SIEMENS-ALBIS AKTIENGESELLSCHAFT, 8048 Zuerich, CH.

(72) Inventor Josef Matthias Fluhr, CH: Jean-Maurice Lutz, CH: Mik = los Istvan Molnar, CH.

(74) Associate Engineer Giersing & Stellinger.

(54) The method and apparatus for reducing quantum ice is the noise = noise in pulse code modulation = systems.

The present invention relates to a method for reducing the quantization noise in pulse coding modulation systems, where coding and decoding is associated with a compression or expansion with a piecewise linear characteristic, and a coupling in an analog / digital converter to carry out the method.

By translating continuous analog signals into the discontinuous digital form, a quantization noise is known, which is attributed to the fact that

Q

an infinite number of possible analog signal amplitude values must be determined by a finite number of quantization steps. Therefore, a reduction of the quantization noise can only be achieved by increasing the number of t quantization steps. Since, on the other hand, each result of a quantization must be able to be described by an individual password, the number of bits required for each password depends directly on the number of quantization steps, i.e. the greater the number of quantization steps, the greater the number of bits required in each password. The requirement for low background noise in channels where there is no speech (voice pause), and good intelligibility even at low speech level, especially in the low amplitude range, requires a sufficient number of quantization steps. In order to provide this with the least possible number of bits in each password, quantization methods with non-linear quantization characteristics have already been proposed, where increasing quantization steps are used for increasing analog signal amplitude values. In this regard, reference can be made to an article "AD- und DA-Wandler, Verfahren und ihre Anwendung" in the journal "ifeue Technik" no. 3/1972 pp. 80 - 84 and no.

4/1972 pp. 103 - 111.

Thus, a nonlinear quantization can be obtained in a very simple manner in the form of a piecewise linear quantization characteristic. As an example of a non-linear quantization characteristic, the so-called A characteristic (CCITT recommendation G711) can be mentioned, where the ratio of analog signal amplitude and quantization noise within a large part of the dynamic range remains approximately constant. Such a quantization characteristic can e.g. consist of 13 segments, where within each segment a linear, and thus easily achievable quantization characteristic is made possible. FIG. 1 shows such a quantization characteristic where the analog signal instant values are plotted out of the X axis and the characteristic sections plotted out of the Y axis. Analog / Digital conversion by non-linear quantization can be equated to a moment value compression of the present analog signal instantaneous value, and the digital signals thus generated must be expanded by conversion back to analog signals corresponding to the instantaneous value of the compression. It is clear that further signal distortions can only be avoided when the quantization characteristic of the analogue / digital converter, hereinafter abbreviated as A / D converter, just as a mirror image corresponds to the feedback characteristic of the digital / analogue converter. hereinafter referred to as D / A Converter.

In the following, instead of the term "segment", the term "characteristic section" is used. The central characteristic section, which extends partly into the positive and partly into the negative characteristic range, for a e.g. of 13 characteristic sections comprising characteristics are divided into two characteristic sections, one positive and one negative, which together form a straight line. Therefore, a characteristic with 13 characteristic sections consists of 2 x 7 = 14 characteristic sections. The location where two characteristic sections touch each other is denoted by characteristic section boundary and a corresponding amplitude value with characteristic section boundary value. Furthermore, it is noted that in the PCM transmission technique, the analogue / digital converter and digital / analogue technique are often also referred to as coding unit and decoding unit.

FIG. 2 shows the almost ideal course between the quantization steps of the A / D converter and the corresponding amplitude stage of the D / A converter, where the D / A converter stage is plotted along the X axis and the A / D converter stage is plotted out along the Y axis. The smallest deviations are evident when all the mean values of the quantization steps are on a right reference line passing through the coordinate system a. This illustration also shows that either in the A / D converter or in the D / A converter ml the contribution from the the zero first positive and first negative analogue steps correspond to the half contribution of a subsequent amplitude step of the same characteristic section.

As further shown in FIG. 2, additional corrections of the same kind are required at each transition from one characteristic section to another characteristic section. Without these corrections, systematic errors occur, which are attributed to as shown in FIG. 3, that the mean values of the individual amplitude steps, depending on the analog signal instantaneous value, as well as the polarity of the characteristic characteristic section, deviate very much in the positive and negative direction from the right reference line a. These systematic errors cause an additional quantization noise.

To improve this method and thus to reduce the quantization noise in impulse code modulation systems, so far only the proposal for a correction of the D / A transducer placed on the receiver side is known, since within each characteristic section it makes a correction with half a quantization step. A correction is preferably easily provided in the D / A converter. This is due to the fact that the A / D converters used in impulse code modulation systems usually for each supplied analog signal instantaneous value continuously output a digital value to the D / A converter corresponding to the contribution next smaller amplitude value which can be represented by an integer quantization.

From DE Publication 2 245 935 a PCM method with an asymmetric expansion characteristic is known, in which identical weights are used in the encoder and in the decoder. This, however, results in an unchanged use of passwords and threshold values and in the non-linear (segmental linear) quantization, a magnitude-dependent error of magnitude with half an amplitude step of the amplitude step used in that segment. A further error is due to the fact that an unstable location (see Fig. 4) at the transition from the positive region to the negative region of the quantization characteristic, that is, at the innermost amplitude step adjacent to the zero line, is precisely interfering with small amplitudes. To remedy only this latter error, it is proposed to divide the expansion characteristic of an inner segment into at least one segment's distance from the zero line into two differently large regions, one of the two possible threshold values in one region and in the other region being given as reproduced amplitude values.

145209 4

Thus, the aforementioned zero-line unstable location is displaced outward (see Fig. 5), whereby the noise source interfering with small amplitudes is displaced into an area with larger amplitudes.

Furthermore, from DE Publication 2 245 225, a method for correcting the quantization characteristic of the receiver's digital / analog converter is known by means of an additional weight which is only half the smallest amplitude step used. Such a correction on the receiver side can be easily accomplished, already because part of the received password already defines the segment of the quantization characteristic in question, so that a further correction with such a half quantization step must be made only during the fine coding. From DE-publication specification 2,245,225 there is known a further method for improving the signal-to-noise ratio, where no extra weight is required, and thus the same weights can be used both by transmitter and receiver. With this method, at the receiver, at small amplitudes (in the first segment), either the lower or upper limit thresholds of the amplitude interval of a password are reproduced as the reproduced amplitude value. There is thus no correction of the quantization characteristic at all in the innermost segment, which determines the smallest amplitudes, so that no noise reduction measures can be mentioned in this area. Only in the remaining amplitude ranges is a correction made in that the decoded amplitude value in each of these segments is corrected by the size of a quantization step in the neighboring segment concerned, which has a finer step division.

However, it is now possible to conceive PCM sisters, where a single transmitter must supply a large number of receivers, e.g. in the radio network. In such cases it stands. which has a large number of D / A converters placed on the receiving side opposite a single A / D converter. Furthermore, the use of coding methods is conceivable, where from the A / D converter placed on each transmitter analog signal instantaneous value, a digital value is output to the D / A converter corresponding to the contribution next higher amplitude value which can be represented by a whole number of quantization steps. In such cases, advantageously only in the A / D transducer located on the transmitter side, a correction to reduce quantization noise is made.

The object of the present invention is to provide a method for analog / digital converters, according to which a reduction of the quantization noise in PCM systems where encoding or decoding is connected can be achieved with compression or expansion with piecewise linear characteristics.

145209 5

For analogue / digital conversion, an analog / digital converter stores the present analog signal instantaneous value in a capacitor and compares the capacitor's charge with reference values (weighing method) and / or at least a linear discharge of the capacitor is determined where the duration of this discharge is determined. by counting (the counting method) in predetermined element time steps for each characteristic section.

In segmental linear quantization, it is assumed that for analog / digital conversion of an analog signal instant value, its affinity is initially determined by a characteristic section of the compression characteristic by comparisons with the reference section limit values corresponding to the reference values and the subsequent determination of it. The present characteristic section for the said analog signal instant value corresponding to the number of quantization steps takes place second by comparison with for each characteristic panel individual reference values, each of which is composed of a base value characteristic of the present characteristic section and one for each quantization step within a characteristic section individual value value. Of course, the polarity of each analog signal instant value is determined in the case of login signals where both polarities occur, adding the result to the password e.g. in the form of a so-called polarity bit which corresponds to the absolute value of the present analog signal instantaneous value. The determination of the number of quantization steps corresponding to an analog signal instant value within a predetermined characteristic section is also referred to as fine coding.

The method of the present invention is characterized in that each characteristic limit value in the analogue / digital converter, apart from the characteristic limit values coincident with the zero quantization characteristic, is in relation to its theoretical with the receiver digital / analogue converter, the characteristic limit value of the half limit value of a quantization step of the grinding characteristic section in the direction of opposite polarity to the characteristic limit value, and that as a base value, depending on the fine coding method, a value corresponding to the relevant characteristic section limit value of the characteristic section corresponding to the digital / analog converter is determined, or a value that deviates from said characteristic section boundary value by half the size of a quantization step for that characteristic section.

A coupling for carrying out the method according to the invention in an analog / digital converter, in which each analog signal torque value is stored in a capacitor whose charging voltage in an output side connected with a control logic U5209 6 can be compared with the values of the control logic individually. - and disconnectable voltage references, and / or at least approximately linear discharge of the capacitor can be effected by means of the control logic individually switch-on and switch-off power sources, the duration of which discharge being determined by counting in each element source step determined by each power source, that when the smallest quantization step is set equal to an amplitude difference ΔΑ, a

M

a positive and one negative polarity source having the value and / or a device for shortening each first element time step.

Embodiments for illustrating the method according to the invention are described in more detail below with reference to the drawing, in which: FIG. 1 shows the positive area of a so-called A-characteristic, fig. Figure 2 shows the ideal course of a quantization characteristic with piecewise different quantization step heights; Figure 3 shows the two possible uncorrected quantization characteristics in A / D converters; 4 shows an example of a quantization characteristic corrected according to the invention with 14 characteristic sections and binary step arrangement; FIG. Figure 5 shows a coupling device using the iteration method for characteristic section determination and fine coding; 6 and 7 each show a variant of a further design of the coupling device shown in FIG. 5, and FIG. Figure 8 shows a coupling device using the iteration method for characteristic section determination and using the counting method for fine coding.

The FIG. 1, 2 and 3 have already been described. For example, in FIG. 4 - for the sake of clarity - only three character sections S1, S2 and S3 are shown of a positive, first quadrant characteristic region and part of a negative, third quadrant characteristic region of a quantization characteristic. The amplitude steps corresponding to the A / D converter's quantization steps are plotted along the Y axis and the amplitude steps corresponding to the decoded digital values of the D / A converter are plotted along the X axis. The dotted line shown represents an uncorrected quantization characteristic, and the fully drawn line shows a quantized characteristic according to the invention associated with an A / D converter positioned on the transmitter side. The individual quantization steps Q, with the exception of the two characteristic sections S1 and -SI adjacent to the zero point of the quantization characteristic, are each assumed to be twice as large as the quantization steps Q of the neighboring characteristic section adjacent to the polarity. A quantization characteristic with quantization step, soo changes by a factor of 2n from characteristic section to characteristic section, where n is a positive integer, is termed a binary divided quantization characteristic. Of course, however, the method of the invention is equally applicable to non-binary quantization characteristics.

The significance of the two in FIG. The characteristics shown in paragraph 4 must be elucidated using a number example. The amplitude step of the D / A converter is indicated along the X axis and the A / D converter amplitude step along the Y axis. For the D / A converter on the receiver side to give the amplitude value expressed in amplitude element step "12", an amplitude value greater than or equal to the value is added to the A / D converter on the sending side in the case of the uncorrected quantization characteristic. of twelve amplitude element arthres and is less than the value of fourteen amplitude element arthres, one amplitude element arthre corresponding to the contribution of the least of the quantization steps Q occurring in the overall characteristic. In the case of a quantization characteristic corrected in accordance with the method of the invention, on the other hand, the A / D converter has to be transmitted on the transmit side an amplitude value greater than or equal to the value of eleven amplitude element steps and less than the value of thirteen araplitude. elementartrin.

The following are the measures to be taken which are necessary to realize an ideal quantization characteristic in the analogue / digital converter on the transmitter side. In a first step, an analog signal instantaneous value offered to the A / D converter is determined by its relation to a characteristic section S of the quantization characteristic. To this end, using the iteration method (weighing method), the present analog signal peak value is often compared in predetermined order with one of the first order individual reference values URS of each characteristic section S until the first order reference values URS defining the present analog signal peak value are determined.

The sizes designated hereinafter with theoretical characteristic boundary values USG correspond to the actual characteristic boundary values of the U5209 8 receiver-side transducer limit values and, with the exception of the characteristic section boundary values, which do not constitute the characteristic zero point sets, characteristic section boundary values used on the transmitter side.

The reference values of the first order URS of the characteristic sections SI and -SI which touch the zero of the quantization characteristic correspond to the amplitude value zero. Each of the first order reference values URS of the first order to the other characteristic sections -Sn ...- S2, + S2 -... + Sn is determined by the theoretical characteristic threshold value USG closest to the opposite polarity, which is reduced by the value of a half-quantization step Q of the characteristic section S. in the direction of opposite polarity. As shown in FIG. 4 is e.g. the theoretical characteristic section S3 limited by the amplitude values "16" and "32". Thus, the theoretical characteristic section boundary value USG of the characteristic section S3 which is closest to the opposite polarity lies at the aplitude value "16". The value of one-half quantization step Q of the opposite characteristic section S2 in the direction of opposite polarity is an amplitude element step. The reference value URS3 of the first order of characteristic section S3 is shown by the contribution reduction of the amplitude value corresponding to the theoretical characteristic threshold value USG = 16 "16" with the value of an amplitude element step, whereby a reference value URS3 of the amplitude value "15" is obtained. Correspondingly, for the characteristic section S2, an amplitude_ value "7.5" corresponding to the reference value URS2 of the first order appears.

In determining the characteristic section S corresponding to the offered analog signal value S, in another step, the number of quantization step Q corresponding to the specific characteristic section Q is determined in another step. In comparison, using the iteration method (weighing method), the corresponding analog signal frequency is often compared with each number of quantization steps individual reference value UQ of the second order until the second order reference values containing the analog signal instantaneous value are determined. Each of the second order reference numerals UQ of the second order of a quantization step within a characteristic section S is fixed by a second base characteristic URB characteristic of the present characteristic section and an individual step value for each quantization step number, a step value only the whole multiplier of the value of a quantization step for that characteristic section.

145209 9

Each base value URB of a characteristic section S is determined when the iteration method is used, by decrement and, when the counting method is used, by increasing or decreasing a value corresponding to that, towards the opposite polarity nearest theoretical characteristic limit value USG, with a different characteristic individual correction value obtained by halving the size of a quantization step Q for the characteristic section Sx thus determined.

Thus, e.g. in FIG. 4, the theoretical characteristic section S3 is limited by the amplitude values "16" and "32", and a limit value next to the theoretical characteristic section USG lies in the amplitude value "16" in the direction of opposite polarity. The base value URB31 or URB32 for characteristic section S3 is obtained by the contribution reduction or increase of the amplitude value corresponding to USG theoretical value section "16" with the value of half a quantization step Q of the characteristic section S3, ie U with value " or URB32 shows an amplitude value of "14" or pi "18".

Correspondingly, for the characteristic section S2 a base value URB21 or URB22 corresponding to the amplitude value "7" or for the characteristic section S1 is obtained for the amplitude value "-0.5" or "+0.5" corresponding to the base value URBI characteristic range being characteristic section -SI one corresponding to the amplitude value "+0.5" or "-0.5" corresponding to the basic value -URBI.

If the counting method is used for fine coding instead of the iteration method, the sizes of the basic values URB corresponding to the execution of the first counting step must be determined. If the first counter steps are not counted, the same basic values URBI, URB21, URB31, URB41 ... etc. are required as in the iteration method. On the other hand, if the first counting steps are included, the boundary values of the corresponding theoretical characteristic sections USB are required to contribute larger basic values URB22, URB32, URB42. When the duration of the first counter step is shortened, the corresponding characteristic section boundary values USG require identical basic values URB. For example, a shortening of the first counter step may. The first signal step rate signal is first taken from the regular clock signal from the elemantar time transducer after a delay time, and only then begins to compare the voltage of the already discharged capacitor with the reference values.

FIG. Figure 5 shows the principle of a coupling device using the iteration method (weighing method) for characteristic section determination and fine coding. The present PAM analog signal instant value is independently supplied U5209 by its polarity to the one input of a control logic unit SL connected to the output side V. Comparison step V's second input is associated with a plurality of individually switchable and switchable sources with the control logic unit SL G with positive and negative polarity. The number of these sources G, as well as their magnitude and polarity, are based on the number of characteristic sections S, which are subject to a quantization characteristic for a particular dynamic range, and their limit values, and on the number and size of the quantization steps Q present in the characteristic sections S.

Thus, either: a) a source G is required for each characteristic section -Sn ...- S2, + S2 ... + Sn, which is not adjacent to the zero of the quantization characteristic, with a size representing the limit value of the present theoretical characteristic section USG, and for each of the characteristic sections, a source G of a size representing half the value of a quantization step Q belonging to the present characteristic section S.

Or required: (b) a source G of positive or negative polarity each of a size corresponding to half the value for the least quantization step occurring in the quantization characteristic, and for each characteristic section -Sn ...- S2, + S2 ... + Sn, which does not match the quantization characteristic zero, a source G of a size representing the present reference value of the first order URS, and a source G of a size representing the difference between the first order value of the present reference URS and the present basic value URB.

or required: (c) a source G of positive or negative polarity each of a size corresponding to half the value of the least quantization step occurring in the quantization characteristic, and for each characteristic section -Sn ...- S2, + S2 ... + Sn, which does not match the quantization characteristic zero, a source G of a size representing the present reference value of the first order URS, and a source G of a size representing the present base value URB.

In addition to the sources G mentioned in (a) to (c), additional sources G for fine coding are required, the number of quantization steps being determined corresponding to the PAM analog signal peak value found within the preceding characteristic section S of the analog signal peak value.

Thus, e.g. for one according to FIG. 4 binary quantization characteristic with seven characteristic sections + S1 ... + S7 for the positive and seven characteristic sections -S1 ...- S7 for the negative amplitude range in the following sources G:

Eleven sources G for the positive amplitude range with the values

-0.5 / + 1 / + 2 / + 4 / + 7 / + 14 / + 28 / + 56 / + 112 / + 224 / + 448 and eleven sources G

for the negative amplitude range + 0.5 / -1 / -2 / -4 / -7 / -14 / -28 / -56 / -112 / -224 / -448.

With the exception of the source G required for the positive characteristic range of magnitude -0.5, all positive reference values can only be obtained with sources G with positive polarity. Also, with the exception of the negative characteristic range G of magnitude +0.5, all negative reference values can only be generated with sources G of negative polarity. The first-order reference values URS can be represented by these sizes as follows:

For the positive characteristic range: URSI = 0; URS2 = 7 + 1 - 0.5 = 7.5; URS3 = 14 + 1 = 15; URS4 = 28 + 2 = 30; URS5 = 56 + 4 = 60; etc.

For the negative characteristic range: URSI = 0; URS2 = -7-1 + 0.5 = -7.5; URS3 = -14 - 1 = -15; etc.

Of course, for all specified sizes of sources G, the reference values of the first order URS and the basic values URB are purely exemplary. However, these show that the number of necessary sources can be minimized by an appropriate choice and combination of the individual sizes. However, it should only be noted that the sources for fine-coding required for the generation of a basic value URB within the characteristic section belonging to this basic value are no longer available.

The base values URB can be represented using the iteration method for fine coding as follows from the sizes given in the above example. For the positive characteristic range: URBI = -0.5; URB21 = 7; URB31 = 14; URB41 = 28; etc.

145209 12

For the negative characteristic range: URB1 = 0.5; URB21 = -7; URB31 = -14; URB41 = -28; etc.

The second order reference values required for the fine coding according to the iteration method are representative of the positive characteristic range by means of the sizes given in the example given above as follows: - characteristic section SI: the basic value URBI of -0.5 and the combinations of the sizes + 1 / + 2 / + 4 gives as reference values of the second order: - + 0.5 / + 1.5 / + 2.5 / + 3.5 / + 4.5 / + 5.5 / + 6.5 - characteristic section S2: the base value URB21 of +7 and the combinations of sizes + 2 / + 4 / give as reference values of the second order: - + 9 / + 11 / + 13 - characteristic section S3: the basic value URB31 of +14 and the combinations of sizes + 1 / + 4 / +7 gives as reference values of the second order: + 18 / + 22 / + 26 etc.

If the counting method for fine coding is used instead of the iteration method (the weighing method), the sizes of the base values URB must be determined accordingly when completing the first counting steps. If the first counter step is not counted, so that correspondingly the beginning of the assessment of the discharge of the capacitor C is shifted, the same basic values URBI, URB21, URB31, URB41 ____ etc. are required as the iteration method. On the other hand, the first count step is counted, for the corresponding theoretical characteristic cutoff limit values USG are required to have substantially larger base values URB22, URB32, URB42 ____, etc. In the event that the first counter step is counted, e.g. for the quantization characteristic of FIG. 4 the following basic values for the positive characteristic range: URBI = +0.5; URB22 = 9; URB32 = 18; URB42 = 36; etc.

and for the negative characteristic range: URBI = -0.5; URB22 = -9; URB32 = -18; URB42 = -36; etc.

145209 13

FIG. 6 shows a first further design of the coupling device according to the invention using the iteration method for characteristic section determination and fine coding. A present PAM analog signal instant value is also applied here independently of its polarity to a first input on an output side connected to a control logic unit SL connected to a control logic V. A plurality of individually switchable and disconnectable sources G with positive and negative polarity are connected with one input at one of all sources G common summation amplifier SV and via a resistor RZ with the output of summation amplifier SV and with the other input at the comparison step V. Each source G contains a value of its determining ohmic resistance R and one in series with this switched switch K. Each of these by means of the control logic unit SL individually switch-on and disconnect switch K, in the case of a source G with positive polarity, connects said one input to the summation amplifier SV via its resistor R with the positive pol p in one for all sources G with positive polarity common first voltage source UA and in the case of a source G with negative po larity via its resistance R with the negative pol p in one of all sources G with negative polarity common second voltage source UB. The negative pole at the first voltage source UA and the positive pole at the second voltage source UB are firmly associated with the second input of the summation amplifier SV and with a common potential identical to the analogue signal moment value PAM's reference potential.

FIG. 7 shows another embodiment of the coupling device according to the invention using the iteration method for characteristic section determination and fine coding. Here again, the present PAM analog signal instant value, regardless of its polarity, is supplied to a first input on the output side with a comparison logic V connected to a control logic unit SL, and a plurality of individually switchable and disconnectable sources G with positive and negative means G polarity is connected to the second input of the comparison stage V. Each source G contains a value of its ohmic resistance R which on one side is connected to the emitter on a transistor Ta or Tb belonging to each source G and via one to each source G associated locking diode D with control logic unit SL. The collectors of the transistors Ta, Tb of all the sources G are connected via a common collector resistor RK to the reference potential of the PAM analog signal momentum and partly to the second input p in the comparison stage V. At the sources G with negative polarity, their resistors R p on the other side are fixed. connected to the negative pole at a thus common first voltage source Uj, and its positive pole is connected to the bases of the transistors Ta as well as to the negative pole of a second voltage source common to the sources G with negative polarity U2, whose positive pole is associated with the PAM analog signal moment value reference potential. At the sources G of positive polarity, their resistors R p on the other hand are firmly connected to the positive pole at a common third voltage source U3, and its negative pole is connected to the bases of the transistors Tb and to the positive pole of one of the sources G with positive polarity common fourth voltage source U4, and its negative pole with the PAM analog signal moment value reference potential. There can be e.g. for the sources G with positive polarity, transistors Tb of type PNP are used and for sources G with negative polarity transistors Ta of type NPN. In this case, the blocking diodes D are inserted in such a way between the sources G and the control logic unit SI, that positive polarity control pulses from the control logic unit SI can arrive at the emitters of the PNP transistors Tb and the negative polarity control pulses from the control logic unit SI to the emitters. NPN Transistors Ta. The FIG. 7 of a source G is also referred to as a constant current source.

FIG. Figure 8 shows a further coupling device in which the iteration method (the weighing method) and the fine-coding counting method are used for characteristic section determination. The present PAM analog signal instant value is applied independently of its polarity to a first input on an output side connected to a control logic unit SL with comparison step V. This first input is connected to a plurality of individually switchable and disconnectable power sources Gi with positive and negative polarity as well as with a capacitor C. The second input of the comparator stage V is connected to a plurality of individually switch-on and disconnect sources G with positive and negative polarity by means of the control logic unit SL. With the sources G, for determining the characteristic section S, the first reference value for each characteristic section S is generated and for fine-coding the individual basic value URB for each characteristic section S.

Each current source Gi comprises a value determining its ohmic resistance Re, which, on the one hand, is connected to the emitter of a transistor Ta or Tb connected to each current source G1 and to the control logic unit SL via a blocking diode D. associated with each current source G1. , Tb to all the power sources Gi is connected partly to the first input of the comparator stage V and partly to the capacitor C.

On the other hand, at the current sources G1 of negative polarity, their resistors Re are firmly connected to the negative pole of one of these common first voltage sources U *, the positive pole of which is firmly connected to the bases of the transistors Ta and to the negative pole of another voltage source. U2 which is common to the negative polarity current sources G1 and whose positive pole is firmly associated with the PAM analog signal instantaneous reference potential. On the other hand, at the current sources G1 with positive polarity, their resistors Re are firmly connected to the positive pole p1 for these common third voltage sources U3, the negative pole being firmly connected to the bases of the transistors Tb and to the positive pole p a fourth voltage source U4. , which is common to the positive poles power sources Gi and whose negative pole is firmly associated with the PAM analog signal instantaneous reference potential. For example, For the current sources G1 with positive polarity, transistors Tb of the PNP type are used and for the current sources G1 with negative polarity transistors Ta of the NPN type. In this case, the blocking diodes D are arranged on a side between the power sources G1 and the control logic unit SI, so that positive pulse control pulses can arrive from the control logic unit SI to the emitters of the PNP transistors Tb and negative polarity control pulses from the control logic unit SL to the emitters on NPN transistors Yes. With each of these power sources Gi a temporally different, at least approximately linear discharge of the capacitor C is possible, whereby the time elapsed until complete discharge is counted in elementary time steps.

Each source G also contains here its value determining ohmic resistance R and one in series with this switched switch K. Each of these by means of the control logic unit SL individually switchable and disconnect switch K in the case of a source G with positive polarity the other input to the comparator stage V via its resistor R with the positive pole on a fifth voltage source UA common to all sources G of positive polarity, and in the case of a source G of negative polarity via its resistor R with the negative pole of a sixth voltage source UB, which is common to all sources G of negative polarity. The negative pole of the fifth voltage source UA and the positive pole of the sixth voltage source UB are firmly associated with the reference potential of the PAM analog signal instantaneous value. Furthermore, the second input of the comparator stage V via an ohmic resistor RK and the second terminal of capacitor C is firmly connected to the reference potential of the PAM analog signal instantaneous value.

For one of FIG. 4 correspondingly corrected quantization characteristic to reduce quantization noise is required in the coupling devices of FIG. 5, 6 and 7, where fine coding takes place by means of the iteration method (weighing method), in addition to the sources required for analogue / digital conversion for both polarities, only for each additional source with a size corresponding to half of the the contribution of the least in the quantization characteristic contained quantization steps. In the coupling device shown in FIG. 8, where fine coding takes place by the counting method, in addition to the sources of both polarities, which are nevertheless required by analog / digital conversion, an additional source with a size corresponding to each polarity is required. half of the contribution of at least the quantization characteristic contained quantization steps, and / or a device for shortening the first elemental time step. A shortening of the first count step can e.g. The first signal step rate signal is first taken from the elemental timer's regular rate signal after a delay time, and only then begins to compare the voltage of the capacitor already discharged with the reference values.

Claims (4)

145209 17 A method for reducing quantization noise in pulse code modulation systems, where coding and decoding are associated with compression and expansion with piecewise linear characteristics, respectively, which determine in the transmitter's analogue / digital converter a coarse coding of a submitted analog signal torque value characteristic and for fine coding, the corresponding number of quantization steps within the pre-determined characteristic section is determined with respect to a characteristic section-individual basic value, characterized in that each characteristic limit value (URS) in the analogue / digital converter, apart from the characteristic quantization steps, is determined characteristic limit values (URS), relative to its theoretical, with the receiver's digital / analogue converter, the characteristic limit value (USG) is reduced by half the size of a quantization step (Q) of it towards the opposite polarity to that person. the characteristic boundary value (USG) bounding characteristic sections and that as a basic value (URB), depending on the fine-coding method, a value corresponding to the relevant characteristic section limit value (USG) of the characteristic section corresponding to the digital / analog converter is determined. , or a value that deviates from said characteristic section limit value (USG) by half the size of a quantization step (Q) for that characteristic section. Method according to claim 1, in which the fine coding is performed by the iteration method, characterized in that for each characteristic section (S), a base value (URB) reduced in relation to the characteristic section limit value (USG) corresponding to the digital / analogue converter. The method according to claim 1, wherein the fine coding is performed by a counting method, characterized in that the duration of each first counting step is shortened relative to the subsequent ones and that for each characteristic section (S), a threshold with the relevant characteristic section limit value is determined. (USG) of the characteristic section (S) corresponding to the basic value (URB) corresponding to the digital / analog converter. Method according to claim 1, wherein the fine coding is performed by a counting method, wherein each first counting step is included or not counted, characterized in that a characteristic sectional threshold value is defined for each characteristic section (S). (USG) decreased respectively increased base value (URB).