JPH0424902B2 - - Google Patents

Description

[Detailed description of the invention]

(Technical Field to which the Invention Pertains) The present invention relates to a receiving device for narrowband digital data transmission, and in particular, to a narrowband reception device that is used to transmit narrowband digital data to the extent that inter-symbol interference occurs between symbol waveforms representing temporally discrete information. The present invention relates to improvements in demodulating devices for converted data signals. (Prior art) It is well known that data transmission that tolerates inter-symbol interference makes it possible to narrow the transmission path and greatly contributes to the effective use of frequencies, especially in wireless networks. Inter-symbol interference reduces the decision margin on the receiving side with respect to line noise, resulting in lower transmission reliability than transmission without inter-symbol interference. To solve this problem, conventional methods estimate the expected range (value range) of the next input symbol from the decision value of the current input symbol and apply the decision threshold to this, which is called the decision feedback method. It is used in various forms. However, this decision feedback method is generally prone to decision error propagation, and the amount of inter-symbol interference is corrected only for symbols that arrive after the already decided input symbol, which limits the scope of application. There is. A method to solve this drawback is to extract multiple input symbols including the one input symbol to be determined from the input sequence, and then combine them with the expected sequence (hereinafter referred to as the reference sequence) in the extracted interval when there is no noise. A promising method is so-called maximum likelihood determination, in which a reference sequence with the maximum likelihood function is selected and the demodulated digit is determined from a transmitted digit sequence that provides that reference sequence. However, this method requires high-speed addition, subtraction, multiplication, and division operations to calculate the likelihood function, which increases the circuit size and power consumption, which poses problems in terms of miniaturization, low power consumption, and cost reduction. Ta. (Object of the Invention) An object of the present invention is to perform demodulation of a narrowband data signal based on the maximum likelihood judgment method using high-speed processing, and to
An object of the present invention is to provide a narrowband data signal demodulation device that achieves miniaturization, low power consumption, and low cost. (Structure and operation of the invention) The present invention has been made to achieve the above object, and has an input symbol timing with a period T synchronized with the symbol rate 1/T and a period τ which is T or an integer fraction of T. a timing extraction circuit that extracts a sampling timing from an input signal; and an A/D converter that sequentially converts the input signal into an input series of digital values in synchronization with the sampling timing.
a D converter, an N-stage parallel information shift register that is cascade-connected to the A/D converter and shifts the input sequence according to the sampling timing; Information obtained from each stage of the shift register (N+
1) a table ROM that receives one input series as part of an address and another part of the address as an extension address, and the table ROM includes the (N+1) input series of the memory blocks indicated by the extension address. The value of the extended address for the (N+1) input series is stored in the memory location indicated by the input series of the transmission digit series for which the likelihood value calculated based on the likelihood function specified by the parameter is the maximum. It is characterized in that the digit of the sequence position to be demodulated, that is, the demodulation digit, is written in advance. The present invention will be explained in detail below with reference to the drawings. FIG. 1 is a block diagram showing an example of the configuration of a narrowband data signal demodulator 10 according to the present invention.
In the figure, a is a narrowband data signal input with a symbol rate of 1/T, and 1 is a timing extraction circuit that inputs the narrowband data signal input a and extracts timing synchronized with the symbol rate.
This is realized using PLL (Phase Locked Loop), etc. RT is one of the outputs of the timing extraction circuit 1 and is an input symbol timing signal with a period T synchronized with the symbol rate. This is a sampling timing signal having a period τ, and supplies operation timing to the entire demodulator. 2 is an A/D converter that samples the input signal a and successively converts it into a digital signal in accordance with the sampling timing signal CLK.
3-1, ..., 3-M, ..., 3-N are A/D
N which is cascade-connected to the output of converter 2 and shifts every cycle τ of the sampling timing signal DLK.
This is a parallel information shift register with a stage configuration. A/D
(N+1) outputs A _i+M , A _i+M-1 ,... from the converter 2 and each stage of the N-stage parallel information shift register.
..., A _i , ..., A _iL are each time (i+M)
τ, (i+M-1)τ, ..., iτ, ..., (i-
L) will show the input samples from A/D converter 2 at τ. Among these (N+1) outputs, let L be the number of samples that are temporally earlier with one demodulation target symbol in between, and M be the number of samples that are later. Therefore, the relationship between L, M, and N is N
=L+M≧0. 4 is a table for reading out the demodulated digit output x _i of the data signal stored in advance in the storage area addressed by the output of the parallel information shift register at time iτ.
It is ROM (Read Only Memory). That is, the input samples A _i+M , A _i+M-1 , ..., A _i1 ,
..., A _iL series (hereinafter referred to as input series) is input as part of the address of table ROM4,
The information written in advance in the memory location indicated by these is outputted to the outside as demodulated digit output x _i at time iτ. In addition, B is the table ROM from outside.
This is an extended address input that can specify 4 memory blocks. Here, it is assumed that the information to be written into the table ROM 4 is configured based on the following basic procedure. First, a sequence of (N+1) samples of all possible waveforms in a time interval of (N+1)τ of the narrowband data signal input a in the absence of noise, X _i+M-1 ,
..., X _i , ... X _iL (hereinafter referred to as the reference series) are clarified, and then
1) Input sequences A _i+M , A _i+M-1 , ..., A _i , ...
..., A _iL (i.e., one address in ROM), select one reference series that gives the maximum likelihood function value, and set the i-th series value X _i of that reference series to Let the corresponding original digit value x _i be
Write to the address indicated by the above input series. Repeat the above operations for all input series A _i+M , A _i+M-1 ,...
..., A _i , ..., A _iL , it is possible to write the digit value x _i corresponding to each address into all the memory blocks specified by the extended address input B mentioned above. can. Here, as the likelihood function of the reference sequence for the input sequence, it is necessary to use a function that matches the characteristics of the transmission path. For example, the negative squared error function given by the following equation -
E ² or correlation function R, etc. are effective. −E ² = − _i+M 〓 ^j=iL (A _j −A _j ) ² …(1)

[Table] However, , indicates the DC average value of the entire value range of A _j and X _j , respectively. Note that the extended address input B specifies, as a parameter, a subset of all combinations of the reference series, which is one of the independent variables of the likelihood function, or
Alternatively, it may be used for purposes such as specifying the dynamic range of each series value of the reference series. Next, the operation and effects of the demodulator of the present invention based on the configuration example shown in FIG. 1 will be explained in detail using a specific example. First, as an example of the narrowband data signal a, consider the case where the binary data value is passed through a low-pass filter with a Gaussian characteristic having a 3 dB cutoff frequency that is 1/4 of its modulation rate 1/T. . Using this signal to perform FM modulation with a maximum modulation index of 0.5 results in GMSK, a type of narrowband digital FM modulation.
(Gaussian Filtered Minimum Shift Keying)
It is well known that. The upper part of FIG. 2 shows the eye pattern of the narrowband data signal a in this example, and is a waveform diagram in which the time waveform of the signal corresponding to the random sequence of transmitted digits is superimposed on its modulation interval T. . The lower part of Figure 2 shows the input symbol timing signal
In the RT waveform, the rising point of RT is located at the point where the level values of the waveform corresponding to the binary states "1" and "0" of the original transmitted digit are maximally separated (hereinafter referred to as the eye opening point). shall be configured to be located. At this time, the opening point of the eye shows approximately 6-value states indicated by the numbers ~, of which ~ and ~ correspond to the binary states "0" (low level) and "1" (high level), respectively. ing. Therefore, demodulation can be performed for the time being by setting a binary decision threshold between and, but the level difference between and is about 1/5 of the difference between and. It is clear that this method cannot be put into practical use because the transmission reliability is significantly impaired. The reason why the aperture points of the GMSK eye pattern shown in FIG. 2 have six values will be explained. Table 1 corresponds to the eye opening point level of ~
The input patterns of X _i-1 , X _i , and X _i+1 and the characteristics of the eye pattern corresponding to X _i are shown.

[Table] Further, FIG. 6 shows waveform examples (broken lines) after narrow band limitation for all 2 ³ =8 combinations (thick solid lines) of X _i-1 , X _i , and X _i+1 . In this example, it can be seen that the level of the eye opening point of X _i has six values of ~. Although this example deals with an example of 6 values of ~, the number of multi-value levels at the eye opening point, where the sections of symbols that interfere with each other are generally determined by the cutoff frequency of the narrowband limiting filter, also differs. . Now, the number of samples L in the configuration of Fig. 1,
Assuming that the values of M and N are L=M=1, N=2, and the period τ of the sampling timing signal CLK is the same as the period T of the received symbol timing RT, N+1=
Three input series A _i+1 , A _i , A _i-1 are stored in table ROM
Consider an example in which address number 4 is input. At this time, all sets of N+1=3 reference sequences X _i+1 , X _i , X _i-1 are obtained using each sample value of the eye opening point in FIG. It will look like Figure 3. Figure 3 A, B, C, and D show time (i-2) τ,
The set (x _i-2 , x _{i-1 ) of the transmitted digit values x i-2 and x i-1 at ( i-1)τ are (“1”, “1} ”) and (“1
”,
8 sets of reference sequences X _i+1 , X _i , X _i-1 corresponding to the cases where The cases where the digit value X _i at time iτ is "1" and "0" are respectively represented by a solid line and a ● mark, and a broken line and a Γ mark.
In the figure, ●○ indicates overlap between Γ and ●. Therefore, the total 32 shown in Figure 3 A, B, C, and D
A set of reference sequences and any one input sequence A _i+1 , A _i ,
The reference series X _{i+1 ,} X _i , X _i-1 , and if the i-th series value X _i belongs to one of , , then x _i
= “0”, , if it belongs to any of x _i
="1" and write this value to the corresponding address in the table ROM4. With the above configuration, each time a set of input sequences A _i+1 , A _i , A _i-1 is obtained, A _i
The estimated value x _i of the digit value corresponding to It is easy to see that by extracting this using , the demodulated digit output x _i for each input symbol can be obtained. The basis for dividing into the four cases shown in FIG. 3 will be explained. Figure 3 shows the level values at the eye opening points of each of X _{i- 1} , X _i , and X _i+1 for all combinations (eight ways) of the reference series X i-1 , X _i , and X _i+1. FIG. 2 is an explanatory diagram expressing the series in a dendritic form. The number of branches at X _i+1 is X _i-1
The value of X _i-2 immediately before one symbol and the value immediately after one symbol (say X _i+2 ) from X _i+ 1 (2 of “0” and “1”)
Due to the influence of X _i-2 , X _i-1 ,
All combinations of 5 bits of X _i+1 and X _i+2 (2 ⁵ =
32 ways). Figure 3 shows the set of X _i-2 and X _i-1 (X _i-2 ,
X _i-1 ) = (“0”, “1”), (“1”, “1”), (
“1”,
(“0”), (“0”, “0”), and this grouping is also used to explain one usage example of extended address input B described below. are doing. In other words, when the values of the past two bits of X _i-2 and X _i-1 are determined, the set of values of the eye opening point of X _i is as shown above.
Figure 3 A, B, specified by the set of X _i-2 , X _i-1 ,
It can be seen that this is a set in any one of the four groups C and D. Next, the operation of an example in which extended address input B is applied will be explained. FIG. 4 shows the demodulator 10 of FIG. 1 according to the present invention.
This is an application example in which the maximum likelihood judgment method and the decision attribution method are used in combination. 11-1~ in Figure 4
11-l is l one-digit shift registers having a function of shifting the demodulated digit output x _i from the demodulator 10 by one symbol time T in accordance with the input symbol timing signal RT;
The l outputs x _i-1 , x _i-2 , x _il of ~11-l are l digits demodulated before x _i and are fed back to the extended address input B of the demodulator 10. be done. If this configuration is applied to the data transmission path having the eye pattern shown in Figure 2, for example, if l = 2, the extended address input B is configured with two digits (x _i-2 , x _i-1 ). , then the set of reference sequences responsive to the value of one extended address input B is
Any one of the four groups A, B, C, D in Figure 3
It can be seen that the target is narrowed down to a group, that is, eight series. From this, the demodulated digit output _i is the table specified by extended address input B
Input series of memory block of ROM4 A _i+1 , A _i ,
It is clear that if the configuration is configured to write to the memory location specified by A _i-1 , the demodulated digit output that combines the maximum likelihood judgment method and the feedback judgment method can be output to the outside using the input symbol timing signal RT. It is. Figure 5 is a diagram showing the characteristics of the average bit error rate when applied to a GMSK mobile communication line under general Rayleaf aging, where the horizontal axis is the logarithm of the signal-to-noise power ratio (dB), The vertical axis is the average bit error rate. The solid line shows the measured value in the case of demodulation based on the configuration of the present invention shown in FIG. 4, and the broken line shows the measured value in the case of demodulation by 2-bit integration, which is a conventionally known method. Here, the conditions in the embodiment of the present invention are based on the eye pattern shown in FIG. 2 as a baseband signal, and the structural parameters T, τ, L, M, N and l for each T=
By setting τ, L=1, M=0, N=1, and l=2, the maximum likelihood judgment method with N+1=2 input sequences and the decision feedback method with l=2 digits can be combined. This is a combined demodulation method. As is clear from FIG. 5, the maximum likelihood determination method using two input sequences using the demodulator according to the present invention,
It can be seen that the demodulation method in combination with the 2-digit decision feedback method achieves almost the same performance as the conventional demodulation method using 2-bit integration. Another use of the extended address input B is to specify the dynamic range of the reference series as a parameter using the extended address input B, and write to the table ROM4 based on the likelihood function in the specified dynamic range. There is a way to do it. In this case, the demodulator measures the dynamic range of the input signal using another means and reflects the result in the value of extended address input B, thereby adapting to changes in the dynamic range of the input signal. It becomes possible to configure. The above-mentioned uses will be specifically explained. Eye opening point level shown in Figures 2 and 3~
Dynamic cleansing is originally fixed, but temperature changes and changes over time may occur due to malfunctions in the hardware that implements it. An example of this is shown in FIG. As shown in this figure, when a change occurs from dynamic range a to dynamic range b, it is clear that correct demodulation operation cannot be expected. In this case, as shown in Fig. 8, the dynamic range is measured from the eye pattern sample value _Ai using a dynamic range measurement circuit installed separately, and the result (for example, the center level and width of the dynamic range, or Maximum value and minimum value) as extended address B in the table
Supply to ROM4. In this case table ROM4
It is assumed that a demodulation digit previously calculated based on a likelihood function corresponding to the dynamic range specified by the extension address B as a parameter is written in . With this configuration, it is possible to configure a demodulator that adapts to changes in the dynamic range of the input signal. (Effects of the Invention) As explained in detail above, according to the present invention,
Since a maximum likelihood determination method using prior knowledge of inter-symbol interference is used, there are fewer restrictions on inter-symbol interference characteristics. Furthermore, since the calculation results of the likelihood function for the maximum likelihood judgment method are stored in the ROM table in advance, real-time calculation processing is not necessary, and high-speed processing is possible. Furthermore, in realizing this, considering the recent increase in the degree of integration of semiconductor memories, the effects of miniaturization and reduction in power consumption are extremely large. Furthermore, its application has advantages such as excellent versatility, such as expanded application to demodulators that combine the decision feedback method and maximum likelihood method, and adaptation to changes in the dynamic range of the input signal. be.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a configuration example of a demodulating device for a narrowband data signal according to the present invention, FIG. 2 is a diagram showing an example of the eye pattern of an example of narrowband data signal input and the waveform of the input symbol timing signal RT, and FIG. The figure is an explanatory diagram using a dendritic representation of a reference sequence for an example of inputting a narrowband data signal, and FIG. 4 shows an example of a configuration in which the maximum likelihood judgment method and the feedback judgment method are combined using the configuration shown in FIG. 1. The block diagram, FIG. 5, shows the parameters τ=T, L=1, M= of the configuration of FIGS. 1 and 4.
0, N=1, and l=2 are actually measured characteristics of the average bit error rate. Figure 6 is a waveform diagram showing waveform examples (dashed lines) after narrow band limitation for all combination patterns of X _i-1 , X _i , and X _i+1 (thick solid lines), and Figure 7 is a waveform diagram showing changes in dynamic range. FIG. 8 is an explanatory diagram of the dynamic range measuring circuit. 1... Timing extraction circuit, 2... A/D converter, 3-1, 3-M, 3-N... Parallel information shift register, 4... Table ROM, a... Narrowband data signal input, RT ...Input symbol timing signal, CLK...Sampling timing signal,
A _i+M , A _i+M-1 , A _i , A _iL ... Input series, B ... Extended address input, x _i ... Demodulated digit output, 1
0... Demodulator of FIG. 1, 11-1 to 11-l...
...1 digit shift register.

Claims (1)

[Claims] 1. Input symbol timing with period T synchronized to symbol rate 1/T and T or an integer fraction of T
a timing extraction circuit that extracts a sampling timing having a period τ from an input signal; an A/D converter that sequentially converts the input signal into an input series of digital values in synchronization with the sampling timing; an N-stage parallel information shift register that is cascade-connected to the converter and shifts the input sequence according to the sampling timing; and information obtained from each stage of the A/D converter and the N-stage parallel information shift register. (N+1)
A table from which pre-stored demodulation digits are read by receiving input series as addresses.
The table ROM contains, in the memory locations indicated by the (N+1) input sequences, the information expected for the (N+1) input sequences when there is no noise between the input sequence and the input sequence. narrowband data signal input (N+
1) Select (N+1) sample sequences that have the maximum likelihood value calculated based on a likelihood function consisting of a squared error function or a correlation function between all combinations of sample sequences. A demodulating device for a narrowband data signal, characterized in that a digit at a sequence position to be demodulated in a transmission digit sequence corresponding to a sample sequence is written in advance as the demodulation digit to be read out. 2. When configuring the table ROM of claim 1, the past demodulated digit series is supplied to the table ROM as an extension address, and the memory block on the table ROM specified by the extension address is stored with the input series. , from the squared error function or correlation function between the set of (N+1) sample sequences of the narrowband data signal expected in the absence of noise, whose past sequence matches the past demodulated digit sequence. The likelihood value calculated based on the likelihood function becomes the maximum (N+1)
2. The narrowband data according to claim 1, wherein a digit at a sequence position to be demodulated for a transmitted digit sequence corresponding to the selected sample sequence is written in advance as the demodulated digit to be read out. Signal demodulator. 3. When configuring the table ROM according to claim 1, information on the dynamic range of the input series obtained by observing a plurality of sample values of the input series is used as an extended address to configure the table ROM.
and the table specified by the extended address.
The memory block on the ROM stores the input sequence (N+
1) Select (N+1) sample sequences that have the maximum likelihood value calculated based on a likelihood function consisting of a squared error function or a correlation function between all combinations of sample sequences. 2. The narrowband data signal demodulation apparatus according to claim 1, wherein a digit at a sequence position to be demodulated in a transmission digit sequence corresponding to a sample sequence is written in advance as the demodulation digit to be read out.