GB2253122A

GB2253122A - Decoding method and apparatus

Info

Publication number: GB2253122A
Application number: GB9102276A
Authority: GB
Inventors: Martin Robert Evans
Original assignee: Cognito Group Ltd
Current assignee: Cognito Group Ltd
Priority date: 1991-02-02
Filing date: 1991-02-02
Publication date: 1992-08-26
Anticipated expiration: 2011-02-02
Also published as: AU1186192A; WO1992014323A1; GB2253122B; GB9102276D0; NZ241489A

Abstract

Data is extracted from a demodulated signal by sampling the signal in successive phases over successive cycles; determining whether each sample value corresponds to a zero cross-over point; recording the occurrence of each cross-over point against the phase in which it occurs in a cycle so as to build up a histogram of cross-over points; selecting a sampling point P from the histogram as the phase of a maximum number of cross-over points; recording the occurrence of successive sample values over successive cycles of the signal in an array of samples values against the phase of a cycle so as to build up the data equivalent to an eye-pattern; determining selection thresholds T both sides of the zero value as minima in the distribution of the recorded occurrences of sample values at the sampling point; and determining whether the sample value at the sampling point P in successive cycles has a value within or outside the selection thresholds T and allocating a corresponding decoded value to that cycle of the signal.

Description

DECODING METHOD AND APPARATUS This invention relates to a method and apparatus for extracting data from a demodulated signal.

When a bit serial data stream is processed for transmission using GMSK or any other partial response signalling method it is subject to line and differential coding and filtering in order to reduce the bandwidth of the transmitted signal.

As a result of these processes, the data is distorted, especially through the interaction between successive bits, known as intersymbol distortion. There are therefore problems in extracting the data from the received signal after demodulation and filtering. However, it is known that the received data signal conforms to certain amplitude and cycle limits which are represented by an "eye pattern" formed by overlaying the oscilloscope traces of successive bit signals.

From this pattern it is possible to identify a zero cross-over point corresponding to a cycle or bit synchronising point, and open areas or "eyes" corresponding to amplitude limits or thresholds between "O" and "1" bit levels. Eye patterns have therefore been used to monitor the performance of cable-based data transmission systems in which line attenuation is a problem that makes it necessary to employ signal compensation techniques.

An object of the present invention is to provide an improved method and apparatus for extracting data from a demodulated signal.

According to the present invention the value of a demodulated input signal is sampled at successive predetermined intervals of time over successive cycles and the data so obtained processed to determine the optimum sampling point for each cycle and the thresholds to be applied at the sampling point to determine a decoded value for that cycle.

In particular, a method according to the present invention consists in the following steps: sampling the value of the input signal at successive predetermined intervals of time (phases) over successive cycles of the input signal; determining for each sample whether or not it corresponds to a zero value of the input signal (a cross-over point); recording the occurrence of each cross-over point against the phase in which it occurs so as to build up a histogram of cross-over points; determining from the histogram the phase of a maximum number of cross-over points (the sampling point); recording the occurrence of successive sample values against phase in a storage array over successive cycles of the input signal so as to build up the data equivalent to an eye-pattern; determining from the distribution of the recorded occurrences of sample at the sampling point, minima (selection thresholds) both sides of the zero value; determining whether the sample at the sampling point in successive cycles has a value within or outside the selection thresholds and allocating a corresponding decoded value to that cycle of the input signal.

This method therefore involves two processes which are preferably conducted simultaneously, one being to generate a histogram of the cross-over points from which a sampling point is determined, and the second being to generate a statistical eye-pattern from which the selection thresholds at the sampling point are determined for use in binary decoding.

The sampling point on the histogram is selected as that phase at which there is a maximum number of cross-over points, although irregularities in the shape of the histogram may lead to the median value being used rather than a simple maximum value.

The invention will now be described by way of example with reference to the accompanying drawings, in which: Figure 1 shows an eye-pattern such as would be produced by a demodulated input signal as displayed on an oscilloscope, Figure 2 shows a histogram of cross-over points such as might correspond to the eye-pattern of Figure 1, Figure 3 shows another histogram of a more irregular shape than that of Figure 2, Figure 4 shows a section through a statistical model of an eye-pattern on the line V-V in Figure 1, and Figure 5 shows a schematic diagram of a system operating according to the invention.

The typical eye-pattern for a GMSK signal is shown in Figure 1, the pattern being produced by overlaying successive cycles of the signal on an oscilloscope. The optimum point for binary decoding of the signal is selected as that point P in the cycle at which there is a maximum number of zero cross-overs. At this point P, there are maximum input values M both sides of zero which can be taken to represent a binary "1" value.

Threshold values T are selected between these maxima M and zero in order to distinguish between binary "0" and binary "1" values, input values within the thresholds T being decoded as binary "0" and input values outside the thresholds T being decoded as binary "1".

In order to determine the zero cross-over maximum P, the signal cycle is divided into sixteen phases and the input signal is analysed for zero cross-overs during successive phases over successive cycles so as to build up a record of the occurrence of zero cross-overs throughout these sixteen phases. Zero cross-overs are detected by setting limits each side of the zero and comparing the value of the signal with these limits. The signal is an oversampled signal so that comparison of sample values occurs once for each phase for each bit period. If a sampled value falls within the limits, it is taken as indicative of a zero cross-over, although more sophisticated tests can be employed for example to determine whether the signal actually passes through both limits.

As shown in Figure 5, successive sampled values of the input signal are held in an input latch 1 and are compared with the two limits in the comparators 2 and 3. A modulo sixteen address counter 6 operates in phase with the input signal and generates a corresponding address to identify each phase of the input cycle. If a cross-over is detected during a particular phase, the phase address generated by counter 6 is used to identify a corresponding cross-over count in a memory 4 which is read into a counter 5. Counter 5 is clocked to add one to the cross-over count, and the count is then written back into the memory 4.

The record of the accumulated cross-over counts for all sixteen phases represents a histogram such as shown in Figure 2. The phase in which the maximum cross-over count occurs can be determined as the simple maximum. However, the histogram may not be regular in shape and may possess more than one peak, as shown in Figure 3. Preferably, therefore the median of the histogram is used, this being the phase at which the area of the histogram is bisected. The median is determined by software using the following algorithm: Let the histogram be H[i] Let the total number of phases be N Find the phase having the minimum number of cross-overs Let this phase be P Find the total area of the graph Let half of this area be A WHILE A = 0 P = (P + 1) mod N A = A - H[P] Optimum phase is P.

At the same time that the input signal is analysed to produce the histogram record, the sampled value of the input signal and its phase are used to progressively build up a statistical model of an eye-pattern. The sampled value passes through a buffer 7 and is combined with the phase address from counter 6 and is used to read a corresponding count out of memory 4 into the counter 5. This count is incremented by one and written back into the memory 4. This process is repeated for successive combinations of sampled value and phase over successive cycles of the input signal.

Once the eye-pattern is complete, the associated data in memory 4 are analysed to determine the thresholds T. The counts of the sampled values at the selected optimum phase P represent a section through the eye-pattern along the line V-V in Figure 1, and are shown in Figure 4. Software is provided to identify the minima between the central maximum at zero and the two outer maxima M. The value of these minima correspond to the thresholds T which are then used for binary decoding of the input data.

The input data are passed through a buffer 8 and written into the memory 4 so that once the histogram has been produced to establish the sampling point P, and the model of the eye-pattern has been produced to establish the thresholds T, these input data can be decoded into binary data.

Although the invention has been described above by reference to a GMSK signal that produces a three level eye-pattern, it could equally well be applied to a system in which the eye-pattern has more than three levels.

Claims

1. A method of extracting data from a demodulated signal comprising sampling the value of the demodulated signal over successive cycles at successive predetermined intervals of time within each cycle to obtain sample values, processing the sample values obtained to determine an optimum sampling point for each cycle and threshold values to be applied at the sampling point to determine a decoded value for that cycle, and comparing the value of the demodulated signal at the optimum sampling point with the threshold values to determine a decoded value thereof.

2. A method as claimed in claim 1 in which the sample values are processed by determining whether or not each sample value corresponds to a zero value of the demodulated signal, recording the occurrence of each zero value against the interval of time in which it occurs in its cycle so as to build up a histogram of zero values, and selecting from the histogram that interval of time for a maximum number of zero values as the sampling point.

3. A method as claimed in claim 1 or 2 in which the sample values are processed by recording the occurrence of successive sample values over successive cycles of the demodulated signal in an array of sample values against the intervals of time in which they occur in their cycles so as to build up a data equivalent to an eye-pattern; and selecting from the distribution of the recorded occurrences of sample values at the sampling point, threshold values corresponding to occurrence minima both sides of the zero value.

4. A method as claimed in claim 3 in which the sample value at the sampling point in successive cycles is compared with the threshold values to determine whether it is within or outside the threshold values, and a corresponding decoded value is allocated to that cycle of the demodulated signal.

5. Apparatus for extracting data from a demodulated signal comprising sampling means to provide sample values of the demodulated signal over successive cycles at successive intervals of time within each cycle, and processing means to determine an optimum sampling point for each cycle and threshold values to be applied at the sampling point to determine a decoded value for that cycle, and comparison means to compare the value of the demodulated signal at the optimum sampling point with the threshold values to determine a decoded value thereof.

6. Apparatus as claimed in claim 5 in which the processing means comprises zero determining means to determine whether or not each sample value corresponds to a zero value of the demodulated signal, zero recording means to record the occurrence of each zero value against the interval of time in which it occurs in its cycle so as to build up a histogram of zero values, and selecting means to select from the histogram that interval of time for a maximum number of zero values as the sampling point.

7. Apparatus as claimed in claim 5 or 6 in which the processing means comprises eye-pattern recording means to record the occurrence of successive sample values over successive cycles of the demodulated signal in an array of sample values against the intervals of time in which they occur in their cycle so as to build up a data equivalent of an eye pattern; and threshold selecting means to select from the distribution of the recorded occurrences of sample values at the sampling point, threshold values corresponding to occurrence minima both sides of the zero value.

8. Apparatus as claimed in claim 7 in which the comparison means compares the sample value at the sampling point in successive cycles with the threshold values to determine whether it is within or outside the threshold values, and allocate accordingly a corresponding decoded value to that cycle of the demodulated signal.

9. A method of extracting data from a demodulated signal substantially as herein described with reference to the accompanying drawings.

10. Apparatus for extracting data from a demodulated signal substantially as herein described with reference to the accompanying drawings.