JP2000295148A - Adaptive equalizer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an adaptive equalizer for digital communication that is simply configured and where equalization result is fast converged. SOLUTION: The adaptive equalizer of a decision feedback type consists of a transversal filter acting like a feedforward section that receives a received base band signal as an input signal, a transversal filter 2 acting like a feedback section, an adder 3 and a discrimination device 4. A tap coefficient update section 8 uses the root mean square(LMS) algorithm to update tap coefficients of the transversal filters 1, 2 and a step size generating section 9 gives a step size of the LMS algorithm according to a predetermined time function.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an adaptive equalizer mainly used for a digital radio communication receiver with fading.

[0002]

2. Description of the Related Art In recent years, personal handy phones (PH)
Digital wireless communications represented by S) and digital mobile phones are rapidly spreading, and mobile communication systems capable of transmitting a large amount of digital information such as color moving images at high speed are being studied. However, fading with multipath generally occurs in a wireless channel, and radio waves from a plurality of paths having different propagation times arrive at the same time. For this reason, similarly to the (distinct) delay distortion in digital signal transmission by the analog telephone line and the modem, delay distortion (multipath distortion, transmission line distortion) occurs in the received signal, and the transmission quality, that is, the bit error rate characteristic deteriorates. Since the deterioration of the transmission characteristics due to the delay distortion increases as the symbol rate (baud rate) increases, the removal of the delay distortion has become indispensable in future high-speed digital mobile communication devices.

The adaptive equalizer is a typical means for removing the delay distortion, and is widely used in the analog telephone line modem. Conventionally, wireless communication has been installed in fixed communication receivers that perform high-speed transmission. In recent years, however, it has been installed in pan-European mobile telephone (GSM) mobile phones that are digital mobile telephone systems in Europe. It has become When such transmission path equalization is performed, the transmission data is divided into blocks called slots or frames, and a predetermined period called a training signal, that is, a training period in which a transmission signal of a predetermined waveform is inserted at the head of the block. Having. Then, in the receiver, the internal constant of the adaptive equalizer is automatically adjusted so that an error from the default value (reference signal) corresponding to the same waveform or the same data is reduced in the same training period. Then, in a portion other than the training period, the data determination result is often used as a reference signal instead of the specified value.

In general, adaptive equalizers are classified into linear equalizers, decision feedback equalizers, and maximum likelihood sequence estimation equalizers. The maximum likelihood sequence estimation type equalizer has excellent equalization performance. For example, Japanese Patent Application Laid-Open No. 9-294095 discloses a technique for further improving characteristics. However, the maximum likelihood sequence estimation type equalizer is
There is a disadvantage that the amount of calculation or the circuit scale of the signal processing increases with the exponential function of the delay time spread. On the other hand, the linear equalizer and the decision feedback equalizer only increase almost directly in proportion to the spread of the delay time to be equalized. Therefore, when an equalizer is incorporated in a household appliance which is preferably manufactured at a low cost, a linear equalizer or a decision feedback equalizer that is simple in signal processing is suitable.

FIG. 6 is a block diagram of a conventional adaptive equalizer of the decision feedback type. In FIG. 6, reference numeral 1 denotes a transversal filter (comb filter) as a feedforward unit that receives a received signal converted into a baseband signal as an input signal, and has a tap interval τ (τ = T / K, T is a symbol period, K Is a natural number and is mainly 1 or 2),
The number of taps is M (M ≧ 2), and the coefficient of each tap can be set. Reference numeral 2 denotes a transversal filter serving as a feedback unit. The tap interval is T, the number of taps is N (N ≧ 1), and each tap coefficient can be set.

Reference numeral 3 denotes an adder for adding the outputs of the transversal filter 1 and the transversal filter 2, and 4 compares the output signal of the adder 3 with a predetermined threshold value for each symbol period T to transmit the transmission data. A decision unit 5 for making a decision and outputting it as demodulated data is a training signal generation unit for producing a training signal. Reference numeral 6 denotes a switch, which is connected to the training signal generator 5 during the training period and to the demodulated data from the determiner 4 during the other periods.
Reference numeral 7 denotes an error detection unit that outputs a difference between an output of the adder 3 and a training signal or demodulated data as a reference signal obtained from the switch 6 as an error signal e.

Reference numeral 8 denotes a tap coefficient updating unit, which is a value (state quantity) x1, x2, * xM, at each delay tap of the transversal filter 1 and the transversal filter 2.
xM + 1, xM + 2, * xM + N and error detection unit 7
Adaptive control of each tap coefficient according to the error signal e obtained by Note that the input signal and other signals may be a complex signal including two in-phase and quadrature signals obtained by performing quadrature detection (synchronous detection or quasi-synchronous detection) on a signal modulated by a carrier wave. In this case, in the following equations, the in-phase component is represented as a complex number having a real part and the quadrature component is an imaginary part. Further, the above configuration is that of a decision feedback equalizer. However, if the transversal filter 2 is removed, the other parts have the same configuration as the linear filter. Although the above configuration can be configured with digital logic circuits,
It may be configured by a processor such as a digital signal processor (DSP) in which signal processing corresponding to the same configuration is programmed.

The operation of the conventional adaptive equalizer configured as described above will be described below. The input signal is processed by the transversal filter 1, the output of the transversal filter 2 is added by the adder 3, the equalization process is performed, the data is determined by the determiner 4, and the data is output as demodulated data. First, in the training period, the switch 6 is connected to the upper side in FIG. 2, and the training signal is supplied to the error detection unit 7 as the reference signal d. At the same time, each tap coefficient cM + 1, cM + 2, * cM + N is successively updated by the tap coefficient updating unit 8 so that the average power of the error signal e is reduced, that is, the mean square error is minimized.

A successive update algorithm based on the minimum mean square error (MMSE) criterion is described in the literature (Heiken,
Takebe Translation, "Introduction to Adaptive Filters", Hyundai Kogakusha, 1983)
, The least mean square (LMS) algorithm and the recursive least mean square (RLS) algorithm are often used. Among them, the LMS algorithm has a small amount of calculation and can be constituted by a simple circuit or an inexpensive processor.
In this case, the tap coefficient vector w (n + 1) at the time t = (n + 1) τ is sequentially updated by the following formulas (Equation 1) and (Equation 2) by each value at the time t = nτ.

[0010]

(Equation 1)

In equation (1), μ is a constant called a step size constant, and the subscript (n) is the time t = nτ
, And (Equation 2).

[0012]

(Equation 2)

The superscript T indicates matrix transposition, H indicates complex conjugate transposition, and * indicates complex conjugate. The tap coefficients are converged within the training period by sequentially repeating the above operations. In another period during which transmission is continued, the switch 6 is connected to the lower side of FIG. 2, and the ideal signal corresponding to the demodulated data, that is, the value when there is no distortion and no noise when the same data as the demodulated data is transmitted as transmission data. Is supplied to the error detection unit 7 as the reference signal d. Then, the tap coefficients are continuously updated by the above-described successive update formula (Equation 1), and the adaptive operation is performed so that the average error power between the transmission signal and the output of the adder 3 is minimized.

[0014]

However, the conventional adaptive equalizer has a disadvantage that convergence is slow when the above-mentioned LMS algorithm is used. If the step size constant of the algorithm is increased, convergence is accelerated, but there is a problem that error power after convergence increases. Therefore, in order to suppress the error power after convergence and obtain sufficient convergence, the training period must be lengthened, and there is a problem that the data transmission efficiency, that is, the effective transmission speed decreases. Although the above problem can be solved by using the RLS algorithm with fast convergence, as described above, the amount of calculation increases and the circuit scale increases, or the use of a high-speed DSP increases power consumption and manufacturing cost. There was a problem.

SUMMARY OF THE INVENTION It is an object of the present invention to provide an adaptive equalizer which solves the above-mentioned problems and requires a small amount of calculation for signal processing and has a fast convergence of an adaptive operation.

[0016]

SUMMARY OF THE INVENTION The present invention provides a first transversal filter having a received signal converted into a baseband signal as an input signal and a second transversal filter having a determined demodulated data as an input. An adder for adding the outputs of the first transversal filter and the second transversal filter, and comparing the output of the adder with a predetermined threshold for each symbol period of the received signal. A decision feedback type adaptive equalizer comprising: a decision unit that decides transmission data and outputs the demodulated data as demodulated data; and an error detection unit that outputs a difference from a reference signal as an error signal. Depending on the state of the first and second transversal filters, the first mean filter and the second transversal filter are used by the least mean square algorithm. A tap coefficient updating unit for updating a tap coefficient of a transversal filter, and a step size generating unit for giving a step size in the tap coefficient updating unit, wherein the step size generating unit is configured to perform the step size by a function of a predetermined time. Was changed.

According to this configuration, it is possible to provide an adaptive equalizer that requires a small amount of computation for signal processing and has a fast convergence of the adaptive operation.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS According to the first aspect of the present invention, a received signal converted to a baseband signal is used as an input signal.
, A second transversal filter that receives the determined demodulated data as an input, an adder that adds the outputs of the first transversal filter and the second transversal filter, and the adder A determination unit that compares the output of the received signal with a predetermined threshold value for each symbol period of the received signal to determine transmission data and outputs the result as demodulated data; And an adaptive equalizer of a decision feedback type including a least mean square (LMS) algorithm according to the error signal and states of the first transversal filter and the second transversal filter. Tap coefficients for updating tap coefficients of the first transversal filter and the second transversal filter A step size generating unit for providing a step size in the tap coefficient updating unit, wherein the step size generating unit changes the step size by a function of a predetermined time. , The step size is appropriately controlled in time, and the tap coefficients converge quickly in the decision feedback equalizer.

According to a second aspect of the present invention, a transversal filter having a received signal converted into a baseband signal as an input signal, and an output of the transversal filter is predetermined for each symbol period of the received signal. A determination unit that determines transmission data by comparing with a threshold value and outputs the result as demodulated data; and an equalizer including an error detection unit that outputs a difference between the output of the transversal filter and a reference signal as an error signal. A tap coefficient updating unit for updating a tap coefficient of the transversal filter by a least mean square (LMS) algorithm according to the error signal and a state of the transversal filter; and a step in the tap coefficient updating unit. And a step size generator for giving a size. The defined time function which is characterized in that changing the step size, the step size is temporally controlled appropriately, the tap coefficients converge quickly in a linear equalizer.

According to a third aspect of the present invention, in the first or second aspect of the present invention, the step size generator includes register means and bit shift means for bit-shifting the output of the register means. The output of the bit shift means is input to the register means, and the step size is generated by the output of the bit shift means or the register means, so that a circuit or a program constituting a step size generator is simplified.

According to a fourth aspect of the present invention, in the first or second aspect of the invention, the step size generating section includes the register means and a subtraction means for subtracting a constant from an output of the register means, The output of the subtraction means is input to the register means, and the step size is generated by the output of the subtraction means or the register means, so that a circuit or a program constituting the step size generation unit is simplified.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of an adaptive equalizer according to an embodiment of the present invention, FIG. 2 is an explanatory diagram of an operation of the adaptive equalizer, and FIG. 3 shows an example of a time change of a step size of the adaptive equalizer. FIG. 4 is a block diagram showing an example of the configuration of the step size generator of the adaptive equalizer, and FIG. 5 is a block diagram showing another example of the configuration of the step size generator of the adaptive equalizer.

In FIG. 1, reference numeral 1 denotes a transversal filter (comb filter) as a feed-forward section which receives a received signal converted into a baseband signal as an input signal.
And the tap interval is τ (τ = T / K, T is the symbol period,
K is a natural number and is mainly 1 or 2, the number of taps is M (M ≧ 2), and the coefficient of each tap can be set. Reference numeral 2 denotes a transversal filter as a feedback unit. The tap interval is T and the number of taps is N (N ≧
In 1), each tap coefficient can be set.

3 is an adder for adding the outputs of the transversal filters 1 and 2 and 4 is a comparator for comparing the output signal of the adder 3 with a predetermined threshold value for each symbol period T and transmitting the transmission data. A decision unit 5 for making a decision and outputting it as demodulated data is a training signal generation unit for producing a training signal. Reference numeral 6 denotes a switch, which is connected to the training signal generator 5 during the training period and to the demodulated data from the determiner 4 during the other periods.
Reference numeral 7 denotes an error detection unit that outputs a difference between an output of the adder 3 and a training signal or demodulated data as a reference signal obtained from the switch 6 as an error signal e.

Numeral 8 denotes a tap coefficient updating unit, which is a value (state quantity) x1, x2, * xM, at each delay tap of the transversal filter 1 and the transversal filter 2.
xM + 1, xM + 2, * xM + N and error detection unit 7
According to the error signal e obtained by the above, each tap coefficient is adaptively controlled by the LMS algorithm. Reference numeral 9 denotes a step size generating unit which gives a step size that changes according to an appropriate time function as a step size of the LMS algorithm in the tap coefficient updating unit 8. The step size generator 9 may be configured to sequentially read, for example, memories storing the above step sizes. The input signal and other signals are obtained from two signals, in-phase and quadrature, obtained by quadrature detection (synchronous detection or quasi-synchronous detection) of a signal modulated by a carrier in the same manner as in a conventional adaptive equalizer. Complex signal. Although the above configuration can be configured by a digital logic circuit, it can also be configured by a processor such as a digital signal processor (DSP) programmed for signal processing corresponding to the same configuration as the conventional adaptive equalizer.

The operation of the adaptive equalizer configured as described above will be described below. The input signal is processed by the transversal filter 1 in the same manner as the conventional decision feedback equalizer, and then the output of the transversal filter 2 is added by the adder 3 to be equalized. It is determined and output as demodulated data. First, during the training period, the switch 6 is connected to the upper side of FIG.
Supplied to At the same time, the tap coefficient updating unit 8 sets the tap coefficients cM + 1 and cM + so that the average power of the error signal e is reduced, that is, the mean square error is minimized.
2, * cM + N is sequentially updated according to the LMS algorithm. That is, the tap coefficient vector w (n + 1) at the time t = (n + 1) τ is sequentially updated by the following calculation formulas (Equation 3) and (Equation 4) by each value at the time t = nτ.

[0027]

(Equation 3)

In (Equation 3), μ (n) is a step size that changes with time, and is supplied from the step size generator 9. The subscript (n) of each variable is the time t =
It means each value in nτ and is (Equation 4).

[0029]

(Equation 4)

The superscript T indicates matrix transposition, H indicates complex conjugate transposition, and * indicates complex conjugate. μ (n) is set to be large at first to achieve high-speed convergence, and to be gradually reduced to suppress the residual error power. For example, a monotonically decreasing function such as μ (n) = a / n (a is a positive constant), or a function as shown in FIG. 3 in which a large step size such that the tap coefficient does not diverge is maintained for several symbols and then reduced. Good. Μ (n) = b + a / n (a, b
(A positive constant) as in the case of a function that asymptotically approaches a positive value other than 0, it is possible to follow the characteristic fluctuation of the transmission line at a relatively high speed. FIG. 2 shows the relationship between the change in the step size and the transmission data. In the figure, the hatched portion represents a training period. By successively repeating the calculation of (Equation 3) while decreasing the step size from a relatively large value, the tap coefficients can be converged at high speed within the training period. In the data period to be subsequently transmitted, the switch 6 is connected to the lower side of FIG. The reference signal d is supplied to the error detection unit 7. Then, the tap coefficient is continuously updated by the successive update equation (Equation 3), and the adaptive operation is performed so that the average error power between the transmission signal and the output of the adder 3 is minimized.

In the above embodiment, the step size generator 9 is configured to read a value prepared in advance in the memory, but may be configured by a simple logic circuit shown in FIG. In FIG. 4, reference numeral 91 denotes a required bit number D.
The register is a flip-flop circuit and is driven by a clock having a period T synchronized with the symbol of the received signal. 92 is a bit shift circuit. Bold lines indicate wiring based on parallel data. The bit shift circuit is simply a shift of the wiring by m bits. First,
The initial value μ0 is set in the register 91 at the beginning of the slot, that is, before the training period. And time T
Each time, the output of the register 91 becomes (Equation 5) and is input to the register 91 again. Accordingly, since the output of the register 91 or the output of the bit shift circuit 92 decreases in a geometric series, a step size decreasing in an exponential function can be generated.

[0032]

(Equation 5)

Further, the step size generator 9 may be constituted by a logic circuit shown in FIG. In FIG. 5, 9
Reference numeral 1 denotes a register comprising a D flip-flop circuit having the required number of bits, which is driven by a clock having a period T synchronized with the symbol of the received signal. 93 is a subtraction circuit for subtracting Δμ from the output of the register 91, 94 is a match detection circuit for detecting a match between the output of the register 91 and a preset minimum value of the step size, and 95 is a match detected by the match detection circuit 94. A switching circuit outputs the value of the register 91 before output, and outputs the minimum value after a match is detected.

First, an initial value μ0 is set in the register 91 before the beginning of the slot, that is, before the training period. Then, Δμ is subtracted from the output of the register 91 every time T, and the subtraction is input to the register 91 again. Therefore, since the output of the register 91 or the output of the subtraction circuit 93 decreases in the arithmetic series, a step size μ (n) = μ0−nΔμ that decreases by a linear function is generated. After decreasing to the minimum value, the switching circuit 95 outputs a constant value of the minimum value. As described above, when the step size is reduced by the linear function, the initial reduction rate can be kept constant. Therefore, according to the experiments by the inventors, the tap coefficients can be converged more quickly. Further, the configuration and processing of the logic circuit can be realized as software such as a DSP.

Note that the above configuration is for a decision feedback equalizer, but if the transversal filter 2 is removed,
The other parts have the same configuration as that of the linear filter. In this case, the configuration is simpler and exactly the same operation is possible.

[0036]

As described above, according to the present invention, the tap coefficients can be converged at high speed by changing the step size of the LMS algorithm with time. Therefore, it is possible to provide an excellent adaptive equalizer which can be applied to a case where a signal having only a short training period is received, has a small amount of calculation, and can be constituted by a simple circuit or an inexpensive processor.

[Brief description of the drawings]

FIG. 1 is a block diagram of an adaptive equalizer according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating the operation of an adaptive equalizer according to an embodiment of the present invention.

FIG. 3 is a graph showing an example of a temporal change in a step size of an adaptive equalizer according to an embodiment of the present invention;

FIG. 4 is a block diagram showing an example of a configuration of a step size generator of the adaptive equalizer according to one embodiment of the present invention;

FIG. 5 is a block diagram showing another example of the configuration of the step size generator of the adaptive equalizer according to one embodiment of the present invention;

FIG. 6 is a block diagram of a conventional adaptive equalizer using a decision feedback type.

[Explanation of symbols]

 1, 2 transversal filter 3 adder 4 discriminator 5 training signal generator 6 switch 7 error detector 8 tap coefficient updater 9 step size generator

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5J023 DA03 DB05 DC03 DC04 DD07 5K046 AA05 BA01 EE06 EE10 EE32 EE50 EE56 EF05 EF07 EF13 EF21 EF23

Claims (4)

[Claims] 1. A first transversal filter that receives a received signal converted to a baseband signal as an input signal, a second transversal filter that receives determined demodulated data as an input, and the first transversal filter. An adder for adding an output of the filter and the output of the second transversal filter; and comparing the output of the adder with a predetermined threshold value for each symbol period of the received signal to determine transmission data and demodulate data. A decision feedback type adaptive equalizer, comprising: a decision unit that outputs the error signal as an error signal; and an error detection unit that outputs a difference from a reference signal as an error signal, wherein the error signal and the first and second transversal are equalized. The first transversal filter and the second transversal filter by a least mean square algorithm according to the state of the filter. A tap coefficient updating unit that updates a tap coefficient; and a step size generating unit that gives a step size in the tap coefficient updating unit, wherein the step size generating unit changes the step size according to a function of a predetermined time. An adaptive equalizer characterized by the following. 2. A transversal filter having a received signal converted into a baseband signal as an input signal, and comparing an output of the transversal filter with a predetermined threshold value for each symbol period of the received signal. An adaptive equalizer including a determiner that determines transmission data and outputs the demodulated data as demodulated data, and an error detector that outputs a difference between an output of the transversal filter and a reference signal as an error signal, wherein the error signal A tap coefficient updating unit that updates a tap coefficient of the transversal filter by a least mean square algorithm according to the state of the transversal filter and a step size generating unit that gives a step size in the tap coefficient updating unit. The step size generator is provided by a function of a predetermined time. Adaptive equalizer, characterized in that to change the serial step size. 3. The step size generator includes register means and bit shift means for bit-shifting the output of the register means. The output of the bit shift means is input to the register means, and the bit shift means or 3. The adaptive equalizer according to claim 1, wherein said step size is generated by an output of said register means. 4. The step size generating section includes the register means and a subtraction means for subtracting a constant from an output of the register means. An output of the subtraction means is input to the register means, and the subtraction means or the register means 3. The adaptive equalizer according to claim 1, wherein said step size is generated by an output of said means.