DE2211681C3 - Method and arrangement for reducing the probability of errors when transmitting code words with different weights of the bit positions of a code word by means of a multilevel code - Google Patents

Description

40

45

55 implemented and transmitted that on the receiving side only if the signal-to-noise ratio is too low expanded amplitude steps or the higher-value bit positions of the recovered from them Few-bit code words processed further and only one of the originally several four-bit code words and the slow sequence of these Multiple code words an undisturbed analog signal with a lower bandwidth is obtained (Fig. 5),

7. Application of the method according to claim 6 for the multilevel transmission of the previously pulse code modulated on accompanying tones of a television transmission in the vertical blanking gaps of the television signal.

The invention relates to a method for reducing the error probability as a result of interference when transmitting code words with different values of the bit positions of a code word by a multilevel code.

It also relates to an arrangement for carrying out such a method.

Multi-level transmission methods are known per se. To get a high bit rate per unit of To achieve bandwidth, several binary strings of the same clock and sampling length are common transmitted in parallel as a multilevel signal. On the Senüeseite, those in the various Binary character strings detected at the same cycle time are converted into one of a series of amplitude values and transmitted. On the receiving side, the received amplitude signal is returned to the original, separated binary strings are converted back.

For the simultaneous transmission of π binary strings, at least 2 "-1 amplitude levels must be provided.

In the figure! is a well-known multilevel coding on the example! of n = 3 to be transmitted bit strings with 7 amplitude levels shown schematically.

If in the known method several, z. If, for example, π binary character sequences are transmitted simultaneously as a multi-level signal with linear amplitude steps over a disturbed transmission path, then these binary character sequences are evenly and at the same time endangered in their reception information if the interfering signals exceed half the distance between the amplitude levels in the multi-level signal.

The largest possible amplitude Am of the multi-level signal is known and is usually transmitted as a synchronous signal and reference variable at regular intervals. The critical value of the interference signals 5, which must not be exceeded with regard to error-free transmission, then results as

At the

I)

d. This means that the signal-to-noise ratio on the line must be more than 2 (2 "-1) for the multi-level signal to be transmitted with uniform amplitude steps (see Fig-1).

Various attempts have been made to meet the high requirements of the multilevel transmission method to reduce the signal-to-noise ratio, e.g. B. by coding or decoding in several stages and

22 ii

the application 'of phase modulation techniques (PT-AS 14 37 534) or by using to keep symmetrically structured multi-level signals small (DT-ASt9 09 4l2),

The inventive method has the same task, already recognizable from the heading, but goes a different way, which results in the known effect of a bit error from the position of the disturbed bit within a multi-digit code word corresponding to a binary number. In the case of ^{| 0} a pulse-code-modulated speech signal whose, for example, 127 amplitude levels are each expressed by a 7-digit Codev / oii, a corruption of a bit to which the highest code value, i.e. 64 amplitude levels in the analog signal, is assigned '5 (most significant), compared to the corruption of a bit to which, for example, only 2 of the possible 127 amplitude levels are assigned (low significant), has a significantly higher influence on the transmission quality of the binary converted signal. ^zo

According to the Anglo-Saxon notation (most significant bit), bit positions of higher significance should be understood to mean those bit positions that generally require a higher level of immunity to interference compared to the other bit positions, or ¹ S especially bit positions that characterize a large amplitude of an originally analog signal.

The Anglo-Saxon names are also used as abbreviations MSB and LSB in Figs. 2 ... 5.

The quality of a multiple parallel digital transmission is according to the previous consideration especially of the transmission security of the most significant bits dependent. While in a serial digital transmission from this known consideration none Measures to reduce the interference can graze, this changes with a Parallel transmission in a multilevel code in which the bits to be transmitted are sorted according to their Cooewertness is possible.

The prerequisite for the multilevel coding method according to the invention is the ord ^r Jng of the bits according to their different valency, ie the structure of the code words to be transmitted as multi-digit binary digits. Code words with a different structure can be converted into the required form by code converters known per se.

The invention solves the problem of reducing the probability of errors when transmitting code words with different weights of the bit positions of a code word by a? n multilevel code in that the amplitude levels of the multilevel signal correspond to the valence of the different Measure bit positions of different sizes, and the bit positions of higher significance compared to those of lower value Valence through a large amplitude level can be distinguished.

It becomes a special kind of companding of the multi-level signal, which differs from the known companding of analog signals in the Execution, from companding when converting ana'og to digital signals (PCM companding) Another difference compared to the mentioned and well-known differs in the setting of objectives However, in cases where companding is used, larger analog signals or samples are pressed on the transmitting side and stretched on the receiving side, but vice versa higher value when companding the Mefir level signal Binary characters are stretched on the sending side and pressed on the receiving side.

The ones to be provided for carrying out the procedure Facilities, digital / analogue converter on the transmission side, Analog / digital converter on the receiving side and possibly serial / parallel converter on the Transmitting and parallel / serial converters on the receiving side are known and are not the subject of the Invention.

The increase in the amplitude levels for higher-order bit parts can be achieved, taking into account the It is necessary to have the control limit of the system grounded, at the same time the amplitude levels for the low-order ones To reduce bit positions.

The following explanations and considerations include the symbols used and their Preceding explanations:

At the maximum amplitude with linear gradation,
Ak maximum amplitude with companded gradation,
a step height with linear steps,
b smallest step height with companded steps,
η number of bit positions in a code word,
m Number of more significant bit positions of a code word.,

ρ increase in importance,

5 Limit value of the permissible interference voltages with linear support

c _ ^α
2 '

Sk Gience value of the permissible interference voltages in non-stretched areas with compensated gradation,

Sk = \,

5k ' limit value of the permissible interference voltages in the stretched range with companded gradation,

SA '= (I + y) Sk = (I + y) ~,

y superelevation; (1 + y) = superelevation factor,
V loss factor for the limit value of the interference voltages in the unstretched area of the companding characteristic compared to a linear characteristic, K improvement factor for the limit value of the interference voltages in the stretched area of the companding characteristic compared to a linear characteristic.

FIG. 2 shows an example of a characteristic curve formed according to the method described and its assignment to 2 " / j-bit code codes and is also intended to clarify the terms and symbols used more clearly than in the following simplified examples. This characteristic curve is not considered any further It should be noted, however, that the degree of elevation y does not necessarily have to be uniformly the same, as shown in the figure, but can also be staggered.

Considered in detail and compared with the case of a linear gradation according to FIG. 1, simple ones become Examples of the dimensioning of the elevation of the amplitude steps, initially for the general case

of cant dimensions that can be freely selected within limits. These considerations and formulas also apply to the Case of exaggeration measures y limited to odd numbers corresponding to even-numbered exaggeration factors.

If, based on the previous consideration of the linear gradation, one assumes the case, which is clear for a quantitative comparison, that the step exaggeration for m of π bit positions takes place by an exaggeration amount y , then in the area of the exaggerated amplitudes disturbances by the factor (y + 1 ) must be larger in order to be harmful in the same way as with the linear gradation.

Since with the same number of stages 2 "-l the

Ak ϊ <

Ratio - ^ when the characteristic curve is stretched compared to the Ratio ^{Am increased} with linear characteristic,

the signal-to-noise ratio for the unstretched stages is reduced by a loss factor V. This results from the ratio of the signal-to-noise ratios, which in turn are given as the ratio of the maximum amplitudes to the limit value of the interference voltages, which is accordingly

V =

AkISk AmIS

In this, Am / S = 2 (2 "- I) from f 1) and

AkISk = 2 (2 "- I -fy (2 ^m

V = - ■

2 "- I + y (2 ^m - I) 2" - I

(2)

35

For the stretched steps, the permissible interference voltages are increased by an improvement factor

^K AkISk 'The improvement factor K is then the same

(5)

K =

1 + y ν

(y + I) (2 "- I)

40

45

(6)

A numerical example with / 7 = 3, m = * 1 and y = 3 shows that for a single higher-order bit position (m = \) out of a total of 3 bit positions to be transmitted, the limit value of the permissible interference voltages by the factor K = 2.8 (approx. 9 db) can be increased, while it is only reduced by the loss factor V = 1.43 (approx. 3 db) for the remaining 2 low-order bit positions.

This by the method described in his Relationships that can be achieved in a simple example are shown as a characteristic curve in the general form Fig. 3 shown.

The arrangements used to carry out the method described are basically based on FIG Sending side digital / analog converter that converts an entered code word into a corresponding analog value and on the receiving side analog / digital converter, the a received analog value into a code word implement. However, commercially available and therefore inexpensive converters have a characteristic curve loss * either with a linear step or with a non-linear one Gradation, in which the step height decreases with higher values. They are therefore for what has been described so far general procedures cannot be used. Conceivable, but expensive, would be converters specially designed for the required degree of elevation.

In contrast, the method described can be used with commercially available and inexpensive converters linear amplitude grading can be carried out if certain concessions are made be able. These are mentioned and explained later

Two are basically the same, only in the Advantages slightly different embodiments possible. These are only given below for the Sending side described; the opposite way on the receiving side is for the expert Of course. Common to both embodiments is the case-by-case increase in the individual to transmitting code words representing binary numbers depending on a certain value of the more significant Bitctelien.

In the first embodiment, each binary number representing a multi-digit code word becomes one then added to the cant / corresponding binary number if at least one of the more significant Bit positions has the specific value, in the examples the value "1". A code converter, which is connected upstream of the digital / analog converter and constructed from commercially available logic circuits, is used for this purpose can. The result of such a code conversion is exemplified in the tables in FIG. 3 to be seen. Accordingly In the example with y = 3, the binary number 11 is added to the original code word. This creates A code word with a larger number of digits for at least some of the possible code words.

For the later considerations it should be noted that for an odd y the exaggeration factor is an even number.

In the second embodiment, the amplitude levels are increased by increasing the number of digits in the binary numbers representing the code words in relation to the lower-order bit positions by inserting a further bit position between the higher-order and lower-order bit positions. Instead of each n-bit code word, an f / j + / 7> bit code word is created which is lengthened by the number of increases in value ρ. Each insertion of an additional bit segment is equivalent to an excess amount

y = 2-

for the more significant bit positions.

In Fig. 4 are in the series of the example assumed 3-bit code words the additional bit positions (ZB) between the more significant (MSB-) and the lower-order (LSB) bit positions inserted between brackets.

In this case, the maximum amplitude results as

Ak - 2 "- 1 + yp = 2" - 1 + 2 "~ ^m p.

Furthermore, it follows from [7] that y and, since ρ can only be an integer, the product yp is always an even

and the exaggeration factor (\ + y) is always an odd number.

One to be provided in the receiving transducer and shown in FIGS. 3 and 4 indicated test threshold PS with a level in the case of, for example, only one excessive level the same

in the middle of the elevated step, with an even-numbered y, as always with the second embodiment, optionally possible with the first embodiment, cannot be realized with normal commercial converters, since the threshold lies in the middle between two normal steps.

A test threshold that can be implemented with commercially available converters lies outside the middle of the excessive level. The value for the degree of exaggeration y for the more significant bit positions can therefore only be included in the comparison calculation

(8th)

can be used. The one for the Pig.4 is an example assumed values n-3, m-1, p-1 and ^ = 2 "- '= 4 from [7] inserted into [4] then result in one Loss factor:

V =

2 "- 1 The improvement factor

kr>' ^{+ 1} with (8) yields

2 "' OT - 1)

i /

2 "

2 "- I '

With the crazy. Example! The further assumed value η = 3 is the improvement factor K = 2.55 (approx. 8 db) for the higher-order bit positions and the loss factor V = 1.57 (approx. 4 db) for the lower-order bit positions for the case of using commercially available converters.

With the same assumed numerical values, including the even-numbered y = 4, but without taking the test threshold into account, the first embodiment according to [7] results in an improvement factor / i «* 3.18 (approx. 10 db) with the same loss factor V" 1, 57 (about 4 db).

A comparison of the two embodiments that can be implemented with commercially available converters shows:

For both embodiments, digital / analog or analog / digital converters are provided with at least one Connection required on the digital side over the number of bit positions of the code words to be transmitted

The first embodiment requires a simple code converter each on the sending and receiving side. An odd-numbered excess amount y is possible, as a result of which a test interval can be formed in a simple manner in the middle of the elevated step and a somewhat more favorable improvement factor is achieved.

In the second embodiment, the code words to be transmitted are also recoded, this However, it happens without a special code converter in a particularly simple manner in that at least one connection on the digital side is not assigned characters in the OLK or A / D converters, rather it is at a fixed potential The test threshold, which is formed in a simple manner, is not in the middle of the excessive level, and the improvement factor that can be achieved is somewhat smaller.

ίο The procedure described so far and explained using simple examples leads to the transferring party Multi-bit code words with very high maximum amplitudes and / or very low, difficult to distinguish Stages.

According to a further development of the invention, the bit positions of a plurality of successive ones are in series incoming multi-bit code words rearranged before the method is carried out in such a way that all bit positions each one of the multi-bit code words in the same place one A plurality of little-bit code words lie, with the Bit positions of the little bit code words are again assigned different values and the little bit code words one after the other according to the one described Process converted into amplitude values and transmitted be genes. The multi-bit code words are given as All of them have different values, and the few-bit code words obtained result in smaller maximum amplitudes and / or larger ones according to the method Stages.

FIG. 1 shows an example of this. 5 the rearrangement of three successive 12-bit code words which are supposed to have arisen from the analog sampled values A 1, A 2 and A 3 shown in the first line 5.1 by an analog / digital conversion. In the from Converters of serially transmitted 12-bit code words CW \, CW2 and CWZ in the second line 5.2 are the twelve bit positions without a value indicated by different symbols, an oblique cross, a horizontal cross and an empty field, as a specific one of the three 12-bit code words correspondingly identified The values of the 3 x 12 = 36 bit positions of the three code words CWl, For example, CW2 and CW3 are written in series into a memory and from the filled memory successively the values of three out of different ones Bit positions originating from code words are read out in parallel. These three bit positions each form a new 3-bit code word CWa ... CWl, as shown in line 53 each the 3-bit code words contain one bit position from each of the 12-bit code words. In each of the 3-bit code words only the bit position originating from the first 12-bit code word CWl is considered to be of higher value in the sense of described procedure viewed.

In general terms, rearranging the bit positions from η / V-bit code words Λ / Λ-bit-Co- words.

The in the in F i g. 5, assumed 12-bit code words are suitable for the signals a 12 kHz wide frequency band and thus an audio frequency channel with broadcast quality or the Play the accompanying sound of a television broadcast. The samples in line 5.1 have a Repetition frequency of 24 kHz. The described procedure in connection with the also described BU position rearrangement allows in the latter case the

Inserting accompanying sound in gaps in the transmission of the image signals using the so-called "sound-in-vision" technology.

On the receiving side, at regular time intervals, e.g. during the duration of an im

703608/176

f # € i £ 5 & em

Connection to the "zero signal" transmitted to the reference variable, the interference voltage is measured and, if this has one Exceeds the minimum value equal to half of the first amplitude level, the evaluation of the amplitude levels corresponding to the lower-order bit positions ver be prevented.

In the case of multi-level coding with previous rearrangement of the bit positions of multi-bit code words, this means that when interference occurs, only every it " multi-bit code word is evaluated on the receiving side and the bandwidth of the recovered analog signals is reduced to the n« ^en part in the receiving end / N-bit) digital

• ' '' y

10

The amplitude-modulated pulses obtained from the tal / analog converter are then fed to the downstream low-pass filter only with a repetition frequency equal to the if ^tn part of the transmission-side sampling frequency, and the low-pass filter can be switched to different cut-off frequencies depending on the detection of interference.

Interferences are therefore no longer audible as such, rather, only recognizable as a reduction in quality due to the reduced bandwidth. Against this is no or little disturbed connection on the Receiving side surrenders the full, original bandwidth.

For this purpose 3 sheets of drawings

L. "r

-J-

Claims (6)

22 Π claims ;. 1. Method for reducing the probability of errors in the transmission of code words with different values of the bit positions of a code word through a multilevel code, characterized in that the amplitude stages of the multi-level signal correspond to the Values of the various bit positions are measured differently and the bit positions are higher Valence compared to those of lower valency distinguished by a higher amplitude level will. 2. The method according to claim 1, characterized in that, on the receiving side, the signal-to-noise ratio with respect to the low one at periodic intervals Amplitude values are determined and the result is stored over the duration of the period and that at If the signal-to-noise ratio is too low, only the expanded amplitude steps or the higher-value bit positions recovered from them are released for further processing. 3. The method according to claim 2, characterized in that to determine the signal-to-noise ratio in periodic intervals, preferably following the reference signal, an amplitude signal with the Value 0 is omitted and transmitted. 4. Arrangement for performing the method according to claim 2 using conventional linear digital / analog or analog / digital converter, characterized in that the transmitting-side digital / analog converter has a binary number corresponding to the incoming code word depending on the value of the more significant bits is preceded by a Uberhöhungsmaß (y) increasing or non-increasing codeword converter, and that the reception-side analog / digital converter, a the recovered binary number is followed depending on their value to the Überhöhungsmaß (y) degrading or nichterniedrigender codeword converter (F i g. 3 ). 5. Arrangement for performing the method according to claim 1, using conventional linear digital / analog or analog / digital converter, characterized in that the higher-order bit positions of each code word to be regarded as a multi-digit binary number to be transmitted compared to the lower-order code word is increased by inserting at least one further bit position (ZB) between the higher and lower-order bit positions, and that for this purpose on the digital side of the converter between the connections for the lower-order and those for the higher-order bit positions, at least one connection is assigned a fixed value ( Fig. 4). 6. The method according to claims 1.2 or 3 for Conversion of multi-bit code words derived from a broadband analog signal into multi-level signals, characterized in that the bit position len several successive multi-bit code words before performing the method in such a way be rearranged so that all bit positions of one of the multi-bit code words are in the same position of a plurality of few-bit code words, the 6 $ Bit positions of the little bit code words are again assigned different values and the little bit code words one after the other according to the method 35