GB2139458A - Error correction in data transmission or processing - Google Patents

Abstract

In a data transmission system, 14-bit samples (e.g. of an audio signal) of which one bit is a sign bit are grouped for transmission into groups of 32 samples. Error protection is applied to 7 bits only of each sample. The maximum value sample in each group is determined and securely transmitted, and error protection is applied to the sign bit, and to the bits corresponding to the six most significant bits in the maximum value sample, for all the samples in the block. Error protection is thus improved during quiet passages in the signal.

Description

SPECIFICATION Error correction in data transmission or processing This invention relates to the protection or correction of data in a transmission or processing system, which will hereinafter simply be referred to as a transmission system.

It has been proposed, in particular for use in transmitting broadcast sound signals such as accompany a television signal, to sample the sound signal at a 32kHz sample rate which is linearly quantised to 14bit accuracy. It has also been proposed in such as system to provide a degree of error protection by forming the signal into blocks of for example 32 samples. The amplitude of the largest of the samples in each block is then determined, and an indication transmitted safely in a secure fashion as to which digit is the most significant digit in that maximum sample of the block. At a receiver each sample is then checked against the thus-transmitted maximum value. If any sample has an apparent value exceeding the maximum value, then it is clear that that sample is in error and an adjustment can be made to reduce it to below the maximum value.

This "range bracketing" technique is extremely advantageous in avoiding inadvertent noise during quiet sound passages.

In addition, it is well known to provide error protection to protect only about the seven most significant bits of a 14-bit sample, using one of a number of possible protection algorithsm. Only the most significant bits are protected because errors in the bits of lower significance are less likely to be noticed, and the more bits that are involved in the error protection process, then the more likely it is to fail at a given error rate.

The present invention is defined in the appended claims, to which reference should now be made.

In accordance with this invention we propose a system in which samples in a digital signal are grouped into blocks, and an indication is transmitted of the most significant bit in the maximum sample in that block. Error protection can then be applied to a predetermined number of bits which exclude amplitude bits of greater significance than the said most significant bit of the largest sample in the block.

For example, 32 samples may be grouped into a block each sample comprising 14 bits. Of these bit number 0 is a sign bit, bit number 1 the most significant bit, and bit number 13 the least significant bit, the bits being numbered sequentially between these two. In a conventional system error protection would be applied to bits 0 to 6 inclusive, and a 3-bit check word transmitted.

In accordance with this invention the block would first be tested to locate the sample having the largest amplitude. If in this sample bits 1 and 2 are 0, and bit 3 is the most significant bit, then protection will be applied in accordance with this invention to bits 0 and 3 to 8 inclusive on all the samples in that block.

At the decoder an inconsistency in the check word can be used either to correct a single-bit error or to invoke an error concealment operation.

Thus, the previously proposed range bracketing technique is applied in such a way that the error protection mechanism is scaled for each block in accordance with the range factor. In this way the error performance is improved during quiet passages in the signal. When the signal is below peak levels, as it usually is, the benefit of error protection is extended to bits of lower significance.

1. A data transmitter, in which: multi-bit data samples are grouped for transmission into groups of several samples each, error protection is applied to selected bits of each sample, and an indication is transmitted of the maximum value sample in each group, and in which the bits to be error protected are selected in dependence upon the value of the maximum value sample.

2. A data receiver for use with the transmission of claim 1, in which multi-bit data samples are received in groups of several samples each together with an indication of the maximum value sample in each group, and in which error protecton is applied to selected bits of each sample, the bits to be error protected being selected in dependence upon the value of the maximum value sample.

3. Apparatus according to claim 1 or 2 in which one bit of each sample is a sign bit, and the error protection is applied to the sign bit and to the bit which is the most significant bit of the maximum value sample, as well as a predetermined number of bits of successively lower significance.

**WARNING** end of DESC field may overlap start of CLMS **.

Claims (3)

**WARNING** start of CLMS field may overlap end of DESC **. SPECIFICATION Error correction in data transmission or processing This invention relates to the protection or correction of data in a transmission or processing system, which will hereinafter simply be referred to as a transmission system. It has been proposed, in particular for use in transmitting broadcast sound signals such as accompany a television signal, to sample the sound signal at a 32kHz sample rate which is linearly quantised to 14bit accuracy. It has also been proposed in such as system to provide a degree of error protection by forming the signal into blocks of for example 32 samples. The amplitude of the largest of the samples in each block is then determined, and an indication transmitted safely in a secure fashion as to which digit is the most significant digit in that maximum sample of the block. At a receiver each sample is then checked against the thus-transmitted maximum value. If any sample has an apparent value exceeding the maximum value, then it is clear that that sample is in error and an adjustment can be made to reduce it to below the maximum value. This "range bracketing" technique is extremely advantageous in avoiding inadvertent noise during quiet sound passages. In addition, it is well known to provide error protection to protect only about the seven most significant bits of a 14-bit sample, using one of a number of possible protection algorithsm. Only the most significant bits are protected because errors in the bits of lower significance are less likely to be noticed, and the more bits that are involved in the error protection process, then the more likely it is to fail at a given error rate. The present invention is defined in the appended claims, to which reference should now be made. In accordance with this invention we propose a system in which samples in a digital signal are grouped into blocks, and an indication is transmitted of the most significant bit in the maximum sample in that block. Error protection can then be applied to a predetermined number of bits which exclude amplitude bits of greater significance than the said most significant bit of the largest sample in the block. For example, 32 samples may be grouped into a block each sample comprising 14 bits. Of these bit number 0 is a sign bit, bit number 1 the most significant bit, and bit number 13 the least significant bit, the bits being numbered sequentially between these two. In a conventional system error protection would be applied to bits 0 to 6 inclusive, and a 3-bit check word transmitted. In accordance with this invention the block would first be tested to locate the sample having the largest amplitude. If in this sample bits 1 and 2 are 0, and bit 3 is the most significant bit, then protection will be applied in accordance with this invention to bits 0 and 3 to 8 inclusive on all the samples in that block. At the decoder an inconsistency in the check word can be used either to correct a single-bit error or to invoke an error concealment operation. Thus, the previously proposed range bracketing technique is applied in such a way that the error protection mechanism is scaled for each block in accordance with the range factor. In this way the error performance is improved during quiet passages in the signal. When the signal is below peak levels, as it usually is, the benefit of error protection is extended to bits of lower significance. CLAIMS

1. A data transmitter, in which: multi-bit data samples are grouped for transmission into groups of several samples each, error protection is applied to selected bits of each sample, and an indication is transmitted of the maximum value sample in each group, and in which the bits to be error protected are selected in dependence upon the value of the maximum value sample.

2. A data receiver for use with the transmission of claim 1, in which multi-bit data samples are received in groups of several samples each together with an indication of the maximum value sample in each group, and in which error protecton is applied to selected bits of each sample, the bits to be error protected being selected in dependence upon the value of the maximum value sample.

3. Apparatus according to claim 1 or 2 in which one bit of each sample is a sign bit, and the error protection is applied to the sign bit and to the bit which is the most significant bit of the maximum value sample, as well as a predetermined number of bits of successively lower significance.