Burst-Error Correcting System

1. Field of the Invention

The invention relates to a digital signal transmission system, and in particular is directed to a system for transmitting digital information signals through a transmission medium, such as magnetic tape and microwave line, which is subjected to disturbances causing burst errors, and for enabling correction of the errors which occur in the received digital signal.

2. Description of the Prior Art

Recently, there is proposed to record a digital information signal, such as audio PCM signal, on a magnetic tape by using video tape recorder. However, it is well known that drop-outs causing a burst error frequently occur in the digital signal reproduced from the tape. There are provided several types of transmission system to correct such a burst error as appearing in the reproduced PCM signal. One of the systems is shown in U.S. Pat. No. 3,409,875, in which the PCM signal is transmitted through two lines, one of which includes a predetermined delay. In the receiving side, the outputs of the lines are applied to an output device through a switch controlled by an error responsive device, which is responsive to unequal signals at the outputs of the lines, so that upon detection of the unequal signals the switch is connected to the line having a delay for a predetermined time. Thus, the system can correct the burst errors appearing in the line. However, it is necessary in the above system to provide the two lines for transmission of the same signal. In other words, the capacity of the transmission medium is needed double the usual one.

An object of the invention is to provide an improved digital signal transmission system in which burst errors contained in the transmitted signal can be corrected at the receiving side.

Another object of the invention is to provide a new system for transmitting digital signals together with parity signals which are composed from the former signals, and for correcting burst errors in the transmitted signal by utilizing the parity signals.

In the system of the invention, a plurality of words of digital information signals are added bit by bit in a modulo 2 adder and a first parity signal is generated for the predetermined number of words of the digital information signals. Each word of the information digital signals and the parity signal are respectively delayed so as to have different delay times to each other. The information signals and first parity signal thus delayed are added bit by bit in a modulo 2 adder to generate a second parity signal for the information and parity signals, and then the predetermined number words of digital information signal and the first and second parity signals are serially transmitted through a transmission line.

The other objects, features and advantages of this invention will be apparent from the following description taken in conjunction with the accompanying drawings.

FIG. 1 is a block diagram of an encoder of a system according to this invention;

FIGS. 2A to 2G show portions of digital information signals appearing at the encoder of FIG. 1 and parity signals which are formed from the digital information signals;

FIG. 3 is a block diagram of a decoder of the system according to this invention;

FIG. 4 is a flow chart showing algorithm of error correcting by the system of FIG. 3;

FIG. 5 is a table which is used for explanation of the error correcting algorithm of FIG. 4;

FIG. 6 is a block diagram of an audio PCM recorder, in which the system of the invention is utilized;

FIG. 7 is another block diagram of the encoder of the system according to this invention;

FIG. 8 shows a portion of the digital information signals appearing at the encoder of FIG. 7 and the parity signals which are formed from the digital information signals;

FIG. 9 is a format of the signals generated from the encoder of FIG. 7;

FIG. 10 is another block diagram of the recorder which is used for the digital information signals processed by the encoder of FIG. 7; and

FIG. 11 is a characteristic curve showing error correcting ability of the system according to this invention.

A description will hereinafter be given on one embodiment of this invention with reference to the drawings. FIG. 1 shows an encoder provided at the transmitting side. An information bit sequence with its one word being formed by one of sampled outputs of an analog signal such as an audio signal is supplied through an input terminal 1 to a demultiplexer circuit 2 where the input information bit sequence is converted into a group of its odd-numbered words and a group of its even-numbered words. Thus, a first information bit sequence H.sub.1 consisting of odd-numbered words and a second information bit sequence H.sub.2 consisting of even-numbered words as shown in FIG. 2A are derived from the demultiplexer circuit 2. These bit sequences H.sub.1 and H.sub.2 are fed to an adder 3 to derive therefrom a first error correcting bit sequence H.sub.3 consisting of parity bits (even parity) as shown in FIG. 2B. The adder 3 and other adders to be mentioned below are all adapted to perform an operation or calculation according to the modulo 2 summation method, and practically formed of exclusive - OR gates. The parity bit sequence H.sub.3 is formed from each one word of the two parallel-transmitted bit sequences H.sub.1 and H.sub.2 shown with the same timing in FIG. 2A, such as P.sub.1 =A.sub.1 .sym.A.sub.2, where a symbol .sym. indicates modulo 2 addition. The bit sequences H.sub.2 and H.sub.3 are then supplied to delay circuits D.sub.1 and D.sub.2 where they are respectively delayed by two words and four words. The delay circuits D.sub.1 and D.sub.2 are usually formed of shift registers. Accordingly, bit sequences H.sub.4 and H.sub.5 shown in FIGS. 2C and 2D are obtained at the output sides of the delay circuits D.sub.1 and D.sub.2. These bit sequences H.sub.4 and H.sub.5 are further applied to an adder 4 where they are added bit by bit to produce a second error correcting bit sequence H.sub.6 consisting of parity bits (even parity) as shown in FIG. 2E, each word thereof being formed from each word of the bit sequences H.sub.4 and H.sub.5 at the same timing. The addition performed in the adder 4 is Q.sub.1 =A.sub.-2 .sym.P.sub.-7, by way of example. These four bit sequences H.sub.1, H.sub.4, H.sub.5 and H.sub.6 are applied to a multiplexer circuit 5. These bit sequences constitute a code formation of a matrix type of four rows by plural columns, and respective words are sequentially seriated at every column of this code formation to obtain a serial data in an order of, for example, A.sub.1, A.sub.-2, P.sub.-7, Q.sub.1, A.sub.3. A.sub.0, P.sub.-5, Q.sub.3,--at an output terminal 6. This serial data is modulated and amplified, if necessary, before being transmitted.

FIG. 3 shows a decoder provided at the receiving side. A received serial data, which is demodulated and amplified as occasion demands, is supplied through an input terminal 7 to a demultiplexer circuit 8, where the received serial data is converted into four parallel-transmitted bit sequences H.sub.1, H.sub.4, H.sub.5 and H.sub.6. The bit sequences H.sub.1 and H.sub.4 are respectively fed to delay circuits D.sub.3 and D.sub.4 to be delayed by four words and two words, respectively. Bit sequences appearing at the output sides of the delay circuits D.sub.3 and D.sub.4 become respectively H.sub.1 and H.sub.2. That is, in the decoder, four bit sequences H.sub.1, H.sub.4, H.sub.5 and H.sub.6 in the same time relation as in the encoder are obtained, and then three bit sequences H.sub.1, H.sub.2 and H.sub.5 in the same time relation as in the encoder are obtained. In order to restore the time relation to the original state as described above, a data synchronizing signal is added, for example, at every four words of the serial data to be transmitted.

The bit sequences H.sub.1, H.sub.4, H.sub.5 and H.sub.6 are supplied to an adder 9 by every one word, and the bit sequences H.sub.1, H.sub.2 and H.sub.5 are supplied to an adder 10 by every one word. The adders 9 and 10 are used for forming syndromes. Since convolutional codes are used in this invention, syndromes from the adders 9 and 10 are respectively supplied to an error correcting logical circuit 11 through a series circuit for four 1-word delay circuits D.sub.5 to D.sub.8 and a series circuit of four 1-word delay circuits D.sub.15 to D.sub.18. Thus, the syndrome derived from the adders 9 and 10, syndromes derived from the respective last stages of the series circuits, and syndromes derived from respective predetermined interstages thereof are fed to the error correcting logical circuit 11. Meanwhile, the bit sequences H.sub.1 and H.sub.2 from the delay circuits D.sub.3 and D.sub.4 are respectively applied through 1-word delay circuits D.sub.9 and D.sub.10 to error correcting adders a.sub.1 and a.sub.2, the outputs of which are supplied to 2-word delay circuit D.sub.11 and D.sub.12, and an output of the delay circuit D.sub.12 applied to an error correcting adder a.sub.3. Outputs of delay circuit D.sub.11 and the adder a.sub.3 are fed to a multiplexer circuit 12 to obtain an error-corrected serial data at its output terminal 13. The delay circuits D.sub.9 and D.sub.10 are provided to maintain a time required for a logic operation of the error correcting logical circuit 11, and the delay circuits D.sub.11 and D.sub.12 are provided to correct an error before two words and also to keep a synchronizing relation of data. Though not shown, the serial data from the output terminal 13 is PCM-demodulated so that the analog signal can be obtained.

An error correcting operation of the decoder will now be described. If an erroneous word contained in received one word is taken as e, and a subscript of word number is attached to e in order to show its corresponding relation with each word of the information bit sequence and parity bit sequence, the syndrome formed by the adder 9 is successively expressed as follows:

Also, the syndrome formed by the adder 10 is successively expressed as follows:

If there is no error, all the bits of these syndromes will become "0". Probability of obtaining a relation of e.sub.i +e.sub.j =0 (e.sub.i .noteq.0 and e.sub.j .noteq.0) is 2.sup.-n, where n (bit) is the word length. If the word length n is selected sufficiently long, a probability that two erroneous words e.sub.i and e.sub.j become equal by chance is negligibly lessened. Further, timings of producing the syndromes by the adders 9 and 10 will become those shown in FIGS. 2F and 2G corresponding to the information bit sequences.

FIG. 4 is a flow chart showing an error correcting logical operation of the error correcting logical circuit 11. In FIG. 4, the side of a decision block attached with a circle mark (O) means "yes", and the side thereof attached with no circle mark means "no". Further, Z.sub.c implies a clear operation such that all the bits of a syndrome held at a corresponding delay circuit are made "0". FIG. 5 shows a mutual relation of syndromes. The syndromes from the adder 9 are formed by respective erroneous words in the horizontal direction of FIG. 5, and the syndromes from the adder 10 are formed by respective erroneous words in the vertical direction of FIG. 5.

At the timings where the syndromes (y.sub.1, y.sub.5, y.sub.9) and (x.sub.-7, x.sub.-3, x.sub.1) are applied to the error correcting logical circuit 11, erroneous words e.sub.1, e.sub.2 and e.sub.-2 respectively contained in respective information words A.sub.1, A.sub.2, and A.sub.-2 can be corrected, and predetermined syndromes are added to the adders a.sub.1, a.sub.2 and a.sub.3 to correct these errors.

For the sake of brevity, the flow chart of FIG. 4 will be partially described with reference to FIG. 5. At first, if x.sub.1 =0 is established, it means that no error exists relating to A.sub.1, A.sub.2 and P.sub.1, so that it is requested to proceed to the next step. If x.sub.1 .noteq.0 and y.sub.1 =0 are satisfied, it means that at least, either of e.sub.2 and e.sub.p1 exists, so that it is requested to examine whether x.sub.1 =y.sub.5 is established or not for the above decision. If x.sub.1 =y.sub.5 is satisfied, it implies that the erroneous word e.sub.2 relating to A.sub.2 exists and that a received data is A.sub.2 +e.sub.2. Accordingly, since x.sub.1 =e.sub.2 is satisfied, an operation of (A.sub.2 +e.sub.2 +x.sub.1) is performed at the adder a.sub.2 so that the correct word A.sub.2 can be obtained. Then the delay circuits D.sub.15 and D.sub.7 are cleared and it is requested to move to the next step. When moved to the next step, x.sub.1 =y.sub.5 =0 is obtained. This clear operation is made for preventing an error correcting operation from being uselessly carried out again despite that the erroneous word e.sub.2 has already been corrected as mentioned above, and also for preventing a miscorrection from occurring in that case. The clear operation is similarly required in other cases.

If x.sub.1 .noteq.y.sub.5 is obtained, it is further requested to decide whether x.sub.1 =y.sub.9 is satisfied or not. And, if x.sub.1 =y.sub.9 is satisfied, it means that the erroneous word e.sub.p1 relating to P.sub.1 exists, so that the delay circuit D.sub.5 is cleared and it is requested to proceed to the next step. When x.sub.1 .noteq.y.sub.9 is obtained, it is also requested to proceed to the next step.

When y.sub.1 .noteq.0 and x.sub.1 =y.sub.1 are established, it means that the erroneous word e.sub.1 relating to A.sub.1 exists, so that an operation of (A.sub.1 +e.sub.1)+x.sub.1 is performed at the adder a.sub.1 to correct the error and the delay circuit D.sub.15 is cleared in order to obtain x.sub.1 =0 at the next step.

When (y.sub.1 =x.sub.1 +x.sub.-3) is established at x.sub.1 .noteq.0, y.sub.1 .noteq.0 and x.sub.1 .noteq.y.sub.1, it implies the existence of the erroneous words e.sub.1 and e.sub.-2 relating to A.sub.1 and A.sub.-2. Accordingly, the syndromes x.sub.1 and x.sub.-3 are respectively supplied to the adders a.sub.1 and a.sub.3 to correct the errors. In this case, the delay circuits D.sub.15 and D.sub.17 are cleared so as to obtain x.sub.1 =0 and x.sub.-3 =0 at the next step. Thus, the error correcting logical operation is similarly carried out at the error correcting logical circuit 11 according to the flow chart of FIG. 4.

FIG. 6 shows an embodiment of a PCM signal recording and reproducing apparatus using a VTR to which the above described invention is applied. In FIG. 6, 14 represents a helical-scan type VTR, which is applied at its input terminal 15i with a PCM signal having the same mode as a television signal. This PCM signal is recorded on a magnetic tape through a recording system of the VTR 14, and a reproduced output from the magnetic tape is delivered through its reproducing system to its output terminal 15o.

Right- and left-channel signals of a stereo audio signal are respectively supplied from terminals 16R and 16L through low pass filters 17R and 17L to sample-and-hold circuits 18R and 18L. Thus sampled signals from the circuits 18R and 18L are fed to A-D converters 19R and 19L to derive therefrom digital code signals, which are then supplied to an encoder 20 to be described later. In the encoder 20, a parity bit is added, time base compression is carried out, and so on to obtain a serial code, which is applied to a synchronizing signal mixing circuit 21. In order to perform the above process, a reference clock oscillator 22 is provided to produce a reference clock which is fed to a pulse generating circuit 23 to produce a sampling pulse, a clock pulse for A-D conversion, a composite synchronizing signal, a control signal for the encoder, and the like. An output of the synchronizing signal mixing circuit 21 is supplied to the input terminal 15i of the VTR 14.

A reproduced PCM signal from the VTR 14 is supplied through its video output terminal 15o to a synchronizing signal separating circuit 24. A composite synchronizing signal separated at the synchronizing signal separating circuit 24 is fed to a pulse generating circuit 25, while the PCM signal from the circuit 24 is fed to a decoder 26 to be described later. After being subjected to processes such as time base expansion, error detection, and error correction in the decoder 26, the PCM signal is supplied to D-A converters 27R and 27L to derive therefrom analog outputs, which are respectively delivered through low pass filters 28R and 28L to output terminals 29R and 29L. A control signal for the decoder 26, a clock pulse for the D-A converters 27R and 27L, a timing pulse for synchronizing signal separation, and the like are produced by a pulse generating circuit 25. A time base in this case is the reproduced composite synchronizing signal.

The encoder 20 is constructed as shown in FIG. 7. A PCM signal S.sub.R relating to the right channel and a PCM signal S.sub.L relating to the left channel are respectively supplied from the A-D converters 19R and 19L of FIG. 6 through terminals 30R and 30L to 1-word delay circuits D.sub.19R and D.sub.19L. Outputs of these 1-word delay circuits D.sub.19R and D.sub.19L are supplied further through 1-word delay circuits D.sub.20R and D.sub.20L to input ends of switch circuits 31R and 31L. The switch circuits 31R and 31L are in synchronism to each other, and each input end thereof is sequentially connected to output ends of the corresponding switch circuit at every 1-word time. Each one word of the PCM signals S.sub.R and S.sub.L, each one word which is delayed from the former by one word, and each one word which is delayed from the first signal by two words, that is, total six words are fed to an adder 32 to be added bit by bit.

A bit sequence H.sub.11 appearing at an output end of the switch circuit 31L is supplied to a multiplexer circuit 33, while bit sequences H.sub.13 and H.sub.15 appearing at the other output ends of the switch circuit 31L are applied through delay circuits D.sub.22 and D.sub.24 to the multiplexer circuit 33. Meanwhile, bit sequences H.sub.12, H.sub.14 and H.sub.16 appearing at respective output ends of the other switch circuit 31R are delivered through delay circuits D.sub.21, D.sub.23 and D.sub.25 to the multiplexer circuit 33. Further, a bit sequence H.sub.17 produced by the adder 32 is supplied through a delay circuit D.sub.26 to the multiplexer circuit 33. If a delay amount of the delay circuit D.sub.21 is taken as d words, delay amounts of the delay circuits D.sub.22, D.sub.23, D.sub.24, D.sub.25 and D.sub.26 are respectively selected as 2d words, 3d words, 4d words, 5d words and 6d words. In this example, d is selected as sixteen words, so that the delay amounts of the respective delay circuits become 16 words, 32 words, 48 words, 64 words, 80 words and 96 words. These seven bit sequences H.sub.11 and H.sub.18 to H.sub.23 are also supplied to an adder 34 to be added bit by bit to produce a bit sequency H.sub.24 consisting of a parity bit sequence Q. This bit sequence H.sub.24 is also fed to the multiplexer circuit 33. Every one word is extracted from these eight bit sequences fed to the multiplexer circuit 33 to form a serial data, which is obtained at an output terminal 35. This serial data is applied to a time base compressing circuit (not shown) in the encoder 20 to form data-lacking intervals corresponding to a horizontal blanking period and a vertical blanking period.

An operation of the encoder 20 will now be described with reference to FIG. 8 and FIG. 9. The adder 32 produces the bit sequence H.sub.17 consisting of parity bits from six words consisting of each word of the PCM signals S.sub.R and S.sub.L, each word thereof before one word from the former, and each word thereof before two words from the first word. For example, a calculation of (L.sub.1 .sym.R.sub.1 .sym.L.sub.2 .sym.R.sub.2 .sym.L.sub.3 .sym.R.sub.3) is performed to form a parity bit sequence P.sub.1 of one word. The six bit sequences H.sub.11 to H.sub.16 appearing at respective output ends of the switch circuits 31L and 31R and the bit sequence H.sub.17 are shown in FIG. 8. Of these bit sequences H.sub.11 through H.sub.17, the bit sequences H.sub.12 to H.sub.17 except H.sub.11 are respectively delayed by the delay circuits D.sub.21 to D.sub.26 to obtain the bit sequences H.sub.18 to H.sub.23. Every one word from seven bit sequences consisting of the bit sequences H.sub.18 to H.sub.23 and the undelayed bit sequence H.sub.11 is supplied to the adder 34 to be added bit by bit to form the bit sequence H.sub.24. By way of example, a calculation of (L.sub.1 .sym.R.sub.-47 .sym.L.sub.-94 .sym.R.sub.-142 .sym.L.sub.-189 .sym.R.sub.-237 .sym.P.sub.-287) is performed to obtain the parity bit sequence Q.sub.1 of one word.

The multiplexer circuit 33 acts to form a serial data from every eight words located at the same timing in FIG. 8. FIG. 9 shows one example of a signal to be supplied to the VTR 14 during one horizontal period between adjacent horizontal synchronizing signals HD. If the word length is selected as sixteen bits, 8.times.16=128 bits will be inserted within one horizontal period.

FIG. 10 shows one example of a circuit arrangement of the decoder 26. The decoder 26 is provided with a time base expanding circuit, though not shown, to produce a serial data with data-lacking intervals being eliminated. This serial data is supplied from an input terminal 37 to a demultiplexer circuit 38 where it is converted into the eight bit sequences H.sub.11 and H.sub.18 to H.sub.24 respectively in a time relation shown in FIG. 8. Each one word of these bit sequences is supplied to an adder 39 where they are added bit by bit to form a syndrome. Meanwhile, delay circuits D.sub.27 to D.sub.32 are provided so as to cancel the differences of delay times between the respective bit sequences in the encoder and those in the decoder. Thus, the bit sequences H.sub.11 and H.sub.18 to H.sub.23 are converted into the seven bit sequences H.sub.11 to H.sub.17 respectively in a time relation shown in FIG. 8 by passing through these delay circuits D.sub.27 to D.sub.32 and each one word of the above bit sequences H.sub.11 to H.sub.17 is fed to an adder 40 where they are added bit by bit to form a syndrome. Further, the bit sequences H.sub.11 to H.sub.16 consisting of information bit sequences are fed through a 1-word delay circuit D.sub.33 to an error correcting adder group a.sub.11 consisting of six adders. Similarly, 16-word delay circuits D.sub.34, D.sub.35, D.sub.36, D.sub.37 and D.sub.38 and error correcting adder groups a.sub.12, a.sub.13, a.sub.14, a.sub.15 and a.sub.16 are respectively provided in turn. Thus corrected information bit sequences are fed to a switch circuit 42 to be converted into right and left PCM signals, which are respectively obtained at output terminals 43R and 43L.

The adders 39 and 40 are each connected at its output side with a series circuit consisting of six 1-word delay circuits and six 15-word delay circuits which are arranged in alternate manner. Syndromes are derived from the last stages of these series circuits and from predetermined interstate positions thereof, and thus derived syndromes are applied to an error correcting logical circuit 41.

The above embodiment of this invention has a conception based upon the enlargement of the embodiment as described previously. Though the detailed description of an error correcting operation of the decoder is omitted, when syndromes y.sub.289 and x.sub.1 are respectively generated from the adders 39 and 40, syndromes y.sub.241, y.sub.193, y.sub.145, y.sub.97, y.sub.49, y.sub.1 and x.sub.-47, x.sub.-95, x.sub.-143, x.sub.-191, x.sub.-239, x.sub.-287 appear as shown in FIG. 10 and these syndromes are applied to the error correcting logical circuit 41.

According to this invention mentioned above, the digital signal transmitting method effective in the correction of burst errors can be achieved. As another type of convolutional code, there is considered a code formation using an error-detecting code, for example, a CRC (cyclic redundancy check) code in place of the bit sequence consisting of the parity bit sequence Q. In this invention, however, the error correcting ability can be enhanced as compared with the above system using such a code formation. For the explanation of comparison in error correcting ability, a graph is shown in FIG. 11 in which ordinate represents number of correction - compensation missing times (number of times per hour) and abscissa represents bit correlation coefficient. According as the bit correlation coefficient approaches 0.999, burst error increases, while according as it approaches 0.900, random error increases. A curve shown in FIG. 11 by a solid line represents a case of using the CRC code instead of the parity bit sequence Q. According to this invention, as shown by a dotted line, the number of correction - compensation missing times can be more reduced and also random error can be more corrected or compensated for.

In the first embodiment mentioned previously, the parity bit sequence Q is added at every three words, while in the second embodiment, the parity bit sequence Q is added at every seven words. However, it is also possible to add the parity bit sequence Q at every given number of words other than the above values.

It will be apparent that many modifications and variations may be effected without departing from the scope of the novel concepts of this invention.