DE2718631B2 - Process for the digital transmission of high quality audio signals - Google Patents

Description

y = (0.75)

Ax

1/1/1

0 <Λ <

y = (0.75)

I f In Ax
I + InA

1
A.

ί.ϊίΙ with A = 21.76

The invention relates to a method for the digital transmission of high quality audio signals according to the principle of dynamic limitation on the transmitting side and dynamic expansion on the receiving side. Such a method can save transmission time, transmission bit rate or storage capacity can be used. It generates a synchronous output bit rate with constant code word length and can therefore be used in conjunction with digital transmission channels without buffer storage.

For the economical transmission of high quality audio signals with bandwidths of 40 Hz Various types of PCM methods with digital companding are known today up to 15 kHz. A basic block diagram is shown in FIG. 1 specified. The analog audio signal first passes through a Pre-emphasis is band-limited by a low-pass filter in an analog / digital converter with one frequency of 32 kHz sampled and linearly quantized and with 16 384 steps, corresponding to 14 bits per attack value The digitized audio signal is coded in a compressor from 14 to 10 bits per sample compressed and then transmitted over the transmission channel.

In the receiver, the 10-bit signal is in one expand digital expander to 14 bits per sample «. After the reconversion in the digital / analog converter and subsequent low-pass filtering the analog audio signal is de-emphasized and can then be played back. To compand the digital, audio signal is used in the transmission method according to (1) Hessenmüller, H. “The Transmission of Broadcasting Programs in a Digital Integrated Network ", proceedings of the" 1972 Zurich Seminar " Instantaneous value compander is used, which works according to a logarithmic 13-segment A characteristic. Two further processes according to (2) CCI.R. 1974 - 78 Doc. CMTT / 42-E, and (3) Osborne, D. W. and Croll, M. G. Digital sound signals; bitrate reduction using an experimental digital compander BBC RD 1973/41, work on the principle of linear block companding.

The Compenseur Presque Instantane according to (2) collects 32 consecutive 14 bits on the transmit side linearly quantized samples together into a block. According to one of 5 linear 10-bit characteristics, whose quantization levels differ in size by a factor of 2 all sampling values of a block are quantized and uniformly depending on the largest amplitude occurring therein coded. A scale factor of 3 bits also informs the receiver at the beginning of each block which characteristic curve has been switched on at the transmission end. The scale factor and other information are stored in the coder at the Place of the least significant bit (I SB) is added to every second 10-bit code word to be transmitted. The near instantaneous compander (3) also works according to the principle described above, except for deviations listed below. At the transmitter end, the sampled values are stored in the analog / digital converter linearly quantized only with 13 bit, so that you have to switch between 4 linear 10-bit characteristics. In a multiplexer is preceded by a 2-bit scale factor for each code word block to be transmitted. For marking of the higher quantization background noise, an analog »dither« - Signal added.

The transmission system with instantaneous value companding according to (1) is characterized by a high Signal-to-noise ratio at low levels and by a constant code word length. The transmission of the digital

lized audio signal only requires word synchronization.

Compared to the method according to (1), the transmission systems with block comparison achieve (2) and (3) a larger signal-to-noise ratio at high levels. The need to have a factor of scale to transmit leads either to a loss of the signal-to-noise ratio (2) that can in principle be achieved, or to a non-constant code word length (3). Iniolge of the block formation is used for the transmission of the digitized audio signal, in addition to word synchronization, block synchronization is also required.

The task leading to the invention was to provide a method for digital transmission or storage of sound signals achieved at least a dynamic range of 79 dB, with analog / digital and Digital / analog converters can be used whose resolution is no longer in the case of linear quantization than 12 bits per sample

With a constant code word length, the method should have a synchronous output bit rate of no more than Provide 10 bits per sample and only one for the transmission or storage of the digitized audio signal Word synchronization does not require block synchronization, however. The procedure should make it possible To transmit signals recorded in analog compression directly digitally or digitally without expansion save.

The object is achieved as described in claim 1. Appropriate arrangements are in the Described subclaims.

This hybrid compander according to the invention uses the advantages of instantaneous value companding and brings out the favorable properties of an analog compander. So it works across the board the solutions described in (1), (2) and (3), a favorable course of the signal-to-noise ratio over the to achieve the entire dynamic range. Further advantages are that there is no additional switching information must be transmitted and a block synchronization for the transmission is not necessary that with the implementation of inexpensive analog / digital and digital / analog converters can be used, whose linear resolution only needs to be 12 bits per sample.

The difference to the suggestion made in (1) is that in (1) the total companding gain by a digital instantaneous value companding is achieved with a 13-segment characteristic according to the logarithmic Α law, while in the In the invention described here, the companding gain is composed of two components. Of the The first part is generated in an analog Dolby A compander. The second part is through one digital instantaneous value compander with a 9-segment characteristic obtained according to the logarithmic law of A.

In contrast to the proposals made in (2) and (2), the invention is not a block companding.

The basic interaction of the components of the hybrid compander is shown in FIG. 2. At the transmitter end, the analog audio signal first passes through a Dolby A compressor. The analog compressed signal is band-limited by a low-pass filter, sampled in the analog / digital converter, linearly quantized and encoded with 12 bits per sample. The sampled values of the digitized audio signal are compressed from \ 2 to 10 bits in an instantaneous value compressor MK and then transmitted. At the receiving end, the 10-bii code words are expanded to 12 bits in an instantaneous value expander ME. After reconversion in the digital / analog converter and subsequent low-pass filtering, the analog audio signal passes through a Dolby A expander and can then be played back. The digital instantaneous value companding works according to a 9-segment characteristic. The characteristic part for positive code words is shown in FIG. The negative part of the characteristic is in the 3rd quadrant of the coordinate system. It emerges from the part of the characteristic curve shown by rotating it 180 ° around the origin of the coordinates.

ι Ί As using an example in F i g. 4 are shown so5i, results in the interaction of the Dolby A compander with the 9-segment instantaneous value compander a favorable curve of the signal-to-noise ratio over the entire modulation range. in the

The diagram of Fig.4 is for comparison of the signal-to-noise ratio of the method according to (1) plotted against the relative signal level. It denotes 0 dBr the maximum permissible level of a sinusoidal signal at the input of the analog / digital converter before use

Vi the limit.

Compared to the 13-segment characteristic used in method (1), the 9-segment characteristic has an initial steepness that is 4 times lower. For the digital instantaneous value used in the invention

iii companding results in a 12 dB lower gain at low levels. The associated loss of signal-to-noise ratio will be compensated by the gain of the analog Dolby A compander to a rest of 2 dB.

r> If the modulation is high, the 9-segment characteristic runs steeper than the! 3-segment characteristic curve, so that compared to method (1) the instantaneous value companding used in the invention at high Levels a signal-to-noise ratio that is 2.5 dB higher

-in achieved to win the Dolby A compander is further improved. In the higher level range, the Dolby gain depends on the signal statistics. What explained by how the Dolby A system works.

■ 4 "> The entire frequency band is used in the Dolby A processor split into four sub-bands. A separate level control works in each sub-band, so that the Dolby A system can be viewed as a controlled filter, the characteristics of which correspond to the spectrum of the

in signal and a white spectrum for the Tried to set the dolby signal. The resulting filler gain is made possible by the invention exploited, on the one hand, to achieve an advantageously large signal-to-noise ratio and, on the other hand, that

Vi to influence quantization noise subjectively favorably, that is created by the digital instantaneous value companding.

The filter gain of the Dolby A compander at levels above about -4S dBr is for sound pro-

wi gram signals are significantly higher than what is shown in 4, in which the Dolby reference level was chosen to be -6 dBr. The course shown there of the Dolby gain, taken from (4) K. Bertram, Dynamic enhancement with the Dolby Stretcher, remote

tvi seh- und Kino-Technik 24/1970, issue 4 only the gain that occurs when the same level occurs in every part of the band.

The signal-to-noise ratio of the entire hybrid

Compander, measured with the aid of a computer one of the "violin" test signals proposed by the European Broadcasting Union (EBU) is in F i g. 4 plotted against the signal level. Because of his With a high peak factor, the violin signal is no longer free of limiting noise at levels above -15dBr. Measured values determined at these levels are shown in dash-dotted lines in the diagram of FIG.

6 shows a compander characteristic of the Dolby A compander, as described in "TV and Cinema Technology", special print from volume 24 (1970), issue 4, page 123 to 126 is described.

F i g. 7 shows a compander characteristic of the Telcom C 4 compander, as it is in the "Funkschau" 1975, issue 18, p. 571, described by Jürgen Werrnuth, with the even better results can be achieved than with the Dolby A compander. The basic principle for Generation of such a compression characteristic with a 33% gradient over such a large area explained as follows:

Three identical, controllable amplifiers are connected in series, which become control inputs operated in parallel. A regulation ensures that at the output of the last amplifier the Level remains constant, regardless of the size of the input voltage of the first amplifier.

This method has the advantage that it is in the work area of interest - as a result of a high Negative coupling of the control system - works exactly. Also, the control range of each of the three is Amplifier reduced to a third of the input dynamic, which results in favorable demands for the Realize this circuit can be derived. The recovery of the original dynamics in the expander works in a complementary way. Since the Telefunken compander Telcam C4 with four of each other independent control systems works - each system is assigned a certain frequency band - is suppresses the breathing tape noise and the modulation noise of the tape itself.

The explained instantaneous value companding with 9-segment characteristic curve for the hybrid compander according to FIG. 2 can be set on the transmitter side by a logical Switching network for coding (coder) and on the receiver side by a logic switching network for decoding (Decoder).

As in !' i g. 3, the coder arranges each code word

at the output of the analog / digital converter each em dead word

/ U. that is transmitted. The decoder orders everyone received code word

.1 IO.l'D.IK.l 7.1b.l Ϊ.Ι4.Γ J.l > _ «I.

a code word each

II 1> II Ii ll'i Ii UD II «11; Il _h 1Γ-, H 4 Il | ΙΓ.ΐΙΓ | H n

/ ti. which is the digital / analog converter / ugcluhri

2i) The bit U ₁₁ does not carry any information: it is only used to represent the representational value in the Sign & Magniüidc code.

Tables 2 and 3 give a possible coding or decoding rule for realizing

r> tion of the 9-segment instantaneous value companding, if analog / digital and digital / analog converter with Sign & Magnitude code can be used. The coding takes place in the individual segments of the Characteristic different.

In order to distinguish the segments are as in FIG. 3 numbered. The second half segment of (5) and the segments (6) to (9) - not shown in Fig. 3 - form the negative part of the characteristic. The segmentation given in Table 1 uses the zero point symmetry

j-i drove out the 9-segment characteristic. The segments (1) and (9), (2) and (8), (3) and (7), (4) and (6) are combined in pairs. For each pair there is only one coding and Deoodieningsvorsrhrift rrforHprjich, which from Table 2 or 3 in the one specified by the segment

4 (i line can be read. Segments that form a pair are to be distinguished from one another by means of the sign bit * i2. Digits of the code word X] 2XuXu) X <^ XχX7XbX =, XiX ^ X2Xι at the input of the coder, their information at the Coding is lost

4i highlighted in table 1 with hatching.

Here / u 7 sheets / eieliiHinucn

Claims (4)

Patent claims: 1. Process for the digital transmission of high-quality audio signals according to the Principle of dynamic compression on the transmission side and the dynamic expansion on the receiving side, characterized in that on the transmitting side the analog signals with a analog compressor are compressed in such a way that in the for a given signal dynamics to be transmitted required resolution of the analog / digital converter is reduced that the digitized Signals additionally to a minimum transmission bit rate with a digital instantaneous value compressor be compressed so that on the receiving side with a corresponding digital instantaneous value expander, Digital / analog converter and analog expander return the dynamic range to its own original value is stretched and that by the analog expansion of the signals on the On the receiving side, an attenuation of the quantization noise is achieved, which reduces the distance between Signal power and quantization noise power of the digital instantaneous value companding ver r> sizes. 2. The method according to claim 1, characterized in that an analog compander is used is, whose linear gain characteristic is deformed in the range of small levels so that in the jo Compressor small levels are raised and complementarily lowered in the expander and that towards the range of medium levels the deformation decreases and high levels remain unchanged (Fig. 6). π 3. The method according to claim 1, characterized in that an analog compander is used is, whose linear gain characteristic runs independently of the modulation so that im Compressor increased the signals by 33% and in the 4Ί1 Expander can be lowered by 33% complementary to this (Fig. 7). 4. The method according to claim 2 or 3, characterized in that the digital instantaneous value compander works according to a 9 segment characteristic with 4> the following function sequence: