JP3625205B2 - Adaptive equalizer and receiver - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an adaptive equalizer and a receiving apparatus, and is suitably applied to a digital wireless communication receiver that receives and demodulates a radio signal affected by fading, for example.
[0002]
[Prior art]
Conventionally, a radio channel generally receives radio waves of a plurality of paths having different propagation times due to the influence of multipath fading. For this reason, delay distortion occurs in the received signal, and the bit error rate characteristic deteriorates. Since the delay distortion increases as the symbol rate increases, in future high-speed digital mobile communication devices, the influence of this delay distortion cannot be ignored, and removal of the delay distortion is indispensable.
[0003]
An adaptive equalizer is a typical means for removing this delay distortion. In wireless communication, it is conventionally installed in a receiver that performs high-speed transmission, and more recently, it is also installed in a GSM mobile phone, which is a European digital mobile phone standard.
[0004]
When such transmission path equalization is performed, transmission data is divided into slot units, and has a training period in which predetermined data called a training sequence, that is, a predetermined transmission signal is inserted into a part of the slot. Then, the receiver automatically adjusts the internal tap coefficient of the adaptive equalizer so that the error from the default value (reference signal) corresponding to the default data becomes small during this training period. In the part other than the training period, the data determination result (demodulated data) is often used as a reference signal instead of the predetermined value.
[0005]
FIG. 7 is a block diagram showing a configuration of a conventional decision feedback type adaptive equalizer 10. In FIG. 7, reference numeral 1 denotes a transversal filter of a feedforward unit that receives a received signal converted into a baseband signal. The transversal filter 1 has a tap interval of τ (τ = T / K, T is a symbol period, K is an oversampling rate), M taps (M ≧ 2), and each tap coefficient can be set. Is. Reference numeral 2 denotes a transversal filter of the feedback unit, where the tap interval is T, the number of taps is N (N ≧ 1), and each tap coefficient can be set.
[0006]
The output of the transversal filter 1 and the output of the transversal filter 2 are added by an adder 3. The addition result by the adder 3 is output to the determiner 4 and the error detector 7. The determiner 4 compares the output signal of the adder 3 with a predetermined threshold for each symbol period T, determines the received signal, and outputs it as demodulated data.
[0007]
Reference numeral 5 denotes a training signal generator for generating a training sequence. Reference numeral 6 denotes a switch, which is connected to the training signal generator 5 side during the training period and to the determiner 4 side during other periods. Thereby, the error detection unit 7 obtains the difference between the output of the adder 3 and the training sequence or demodulated data as the reference signal obtained from the switch 6 as the error signal e, and the error signal e is obtained as the tap coefficient update unit 8. To send.
[0008]
The tap coefficient updating unit 8 responds to the error signal e with values (state quantities) x ₁ , x ₂ ,..., X _M , x _{M + 1} , x at each delay tap of the transversal filter 1 and the transversal filter 2. _{M + 2, ........., x M} + N and the tap coefficient values _{c 1, c 2, .........,} c M, c M + 1, c M + 2, ........., adaptively controls the _{c M + N.}
[0009]
Incidentally, FIG. 7 shows the configuration of the decision feedback type adaptive equalizer 10. However, if the transversal filter 2 is removed, the other parts have the same configuration as that of the linear filter.
[0010]
The operation of the conventional adaptive equalizer configured as described above will be described. The input signal is processed by the transversal filter 1, and the signal processed by the transversal filter 1 and the signal processed by the transversal filter 2 are added by the adder 3, whereby equalization processing is performed. Next, the data of the output of the adder 3 is determined by the determiner 4 to obtain demodulated data.
[0011]
In the training period, the switch 6 is connected to the training signal generator 5 side, and the training sequence is supplied to the error detector 7 as the reference signal d. At the same time, the tap coefficient updating unit 8 causes the tap coefficients c ₁ , c ₂ ,..., C _M , c _{M + 1} , c _{M + 2} ,..., C _{M + N} so that the mean square error of the error signal e is minimized. Are updated sequentially.
[0012]
As a sequential update algorithm based on the least mean square error criterion, an LMS (Least Mean Square) algorithm or an RLS (Recursive Last Square) algorithm is often used. In particular, the LMS algorithm has a small amount of calculation and can be realized by a simple circuit or an inexpensive processor. In this case, the tap coefficient vector w (n + 1) at time t = (n + 1) τ is sequentially updated using each value at time t = nτ using the following equation.
[0013]
w (n + 1) = w (n) +. mu.x (n) e ^* (n) (1)
In the equation (1), μ is a constant called a step size constant, and the subscript (n) means each value at time t = nτ, and x (n), w (n), e (n) is a value obtained by the following equation.
[0014]
x (n) = (x ₁ x ₂ ... x _M x _{M + 1} ... x _{M + N} ) ^T
w (n) = (c ₁ c ₂ ... c _M c _{M + 1} ... c _{M + N} ) ^T
e (n) = d−w ^H x (2)
Here, * in the equation (1) represents a complex conjugate, a superscript T in the equation (2) represents a matrix transpose operation, and a superscript H represents a complex conjugate transpose operation. The tap coefficient update unit 8 converges the tap coefficient within the training period by sequentially repeating the above calculation.
[0015]
On the other hand, during a period other than the training period, the switch 6 is connected to the determiner 4 side, and when the ideal signal corresponding to the demodulated data is transmitted, the value when there is no distortion and no noise is supplied to the error detection unit 7 as the reference signal d. Is done. Then, the tap coefficients c ₁ , c ₂ ,..., C _M , c _{M + 1} , c _{M + 2} ,..., C _{M + N} are continuously updated by the above sequential update formula, and the demodulated data and the output of the adder 3 are updated. It operates adaptively so that the average error power is minimized.
[0016]
[Problems to be solved by the invention]
However, the conventional adaptive equalizer has a drawback that convergence is slow when the above LMS algorithm is used. Therefore, if the step size constant of this algorithm is increased, convergence will be accelerated, but the mean square error value after convergence will increase. Therefore, in order to suppress the mean square error after convergence and to obtain sufficient convergence, there is a problem that the training period must be lengthened and the data transmission efficiency is lowered.
[0017]
The present invention has been made in view of this point, and an object of the present invention is to provide an adaptive equalizer and a receiving apparatus that can obtain an equalized signal with a fast convergence of an adaptive operation and a small error.
[0018]
[Means for Solving the Problems]
In order to solve this problem, the present invention adopts the following configuration.
[0025]
An adaptive equalizer according to the present invention includes a first transversal filter that receives a received signal converted into a baseband signal, a second transversal filter that receives the determined demodulated data, An adder for adding the outputs of the transversal filter and 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 the received signal to obtain demodulated data An error signal is generated using a determination unit, an error detection unit that outputs a difference between the reference signal or demodulated data and the output signal of the adder as an error signal, and the error signal and the tap coefficients of the first and second transversal filters. A tap coefficient updating unit that updates the tap coefficients of the first and second transversal filters by performing an adaptive algorithm calculation so that Thereby generating a step size used in the tap coefficient updating unit, an adaptive equalizer comprising a step size generating unit that changes according to the step size to the error signal, the step size generating unit, the error detection unit The absolute value of the ratio of the output from the error storage unit and the error storage unit that temporarily stores the error signal output from the step size storage unit that temporarily stores the output signal from the step size generation unit and the output from the error storage unit And a multiplier for calculating the product of the output from the divider and the output from the step size storage unit, and the output from the error detection unit as an input, and the step size is determined by the output of the multiplier. The structure to generate is taken.
The adaptive equalizer of the present invention compares a transversal filter that receives a received signal converted into a baseband signal and compares the output of the transversal filter with a threshold value that is predetermined for each symbol period of the received signal. A determination unit that determines a received signal and obtains demodulated data, an error detection unit that outputs a difference between the reference signal or demodulated data and the output signal of the transversal filter as an error signal, and tap coefficients of the error signal and the transversal filter. And performing an adaptive algorithm calculation so as to reduce the error signal, thereby generating a tap coefficient updating unit for updating the tap coefficient of the transversal filter, a step size used in the tap coefficient updating unit, and the step size. An adaptive equalizer comprising: a step size generator that changes in accordance with an error signal The step size generation unit includes an error storage unit that temporarily stores an error signal output from the error detection unit, a step size storage unit that temporarily stores an output signal from the step size generation unit, and the error storage unit. A divider for calculating the absolute value of the ratio of the output and the output from the error detection unit; and a multiplier for calculating the product of the output from the divider and the output from the step size storage unit. A configuration is adopted in which the output is used as an input and the step size is generated by the output of the multiplier.
[0026]
According to these configurations, the step size generator can generate the step size according to the error with high accuracy.
[0031]
DETAILED DESCRIPTION OF THE INVENTION
The essence of the present invention is that a transversal filter is calculated using an adaptive algorithm operation in accordance with an error between an output of a transversal filter for filtering an input received signal and demodulated data or a reference signal obtained by determining the output. Is to adaptively change the step size of the tap coefficient updating unit for updating the tap coefficient.
[0032]
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0033]
In FIG. 1, reference numeral 100 generally indicates the configuration of an adaptive equalizer according to an embodiment of the present invention. The adaptive equalizer 100 inputs the received signal converted into the baseband signal to the transversal filter 101 of the feedforward unit. The transversal filter 101 has a tap interval τ (τ = T / K, T is a symbol period, K is an oversampling rate), M taps (M ≧ 2), and each tap coefficient can be set. Is. Reference numeral 102 denotes a transversal filter of the feedback unit, which has a tap interval of T, the number of taps is N (N ≧ 1), and each tap coefficient can be set.
[0034]
The output of the transversal filter 101 and the output of the transversal filter 102 are added by an adder 103. The addition result by the adder 103 is output to the determiner 104 and the error detection unit 107. The determiner 104 determines the received signal by comparing the output signal of the adder 103 with a predetermined threshold for each symbol period T, and outputs the received signal as demodulated data.
[0035]
Reference numeral 105 denotes a training signal generator for generating a training sequence. Reference numeral 106 denotes a switch, which is connected to the training signal generator 105 side during the training period and to the determiner 104 side during other periods. Thereby, the error detection unit 107 obtains a difference between the output of the adder 103 and the training sequence or demodulated data as a reference signal obtained from the switch 106 as an error signal e (n), and the error signal e (n) Is sent to the tap coefficient updating unit 108.
[0036]
The tap coefficient updating unit 108 includes values (state quantities) x ₁ , x ₂ ,..., X _M , x _{M + 1} , x _{M + 2} ,..., X at each delay tap of the transversal filter 101 and the transversal filter 102. _{In addition} to controlling _{M + N} , the tap coefficient values C _{1 to} C _{M + N of the} transversal filter 101 and the transversal filter 102 are set so as to reduce the error value according to the error signal e (n) obtained by the error detection unit 107. Is controlled adaptively.
[0037]
In addition to this configuration, the adaptive equalizer 100 includes a step size generator 109. The step size generation unit 109 generates a step size based on the output from the error detection unit 107, and supplies this to the tap coefficient update unit 108 as the step size of the LMS algorithm in the tap coefficient update unit 108.
[0038]
The step size generator 109 is configured as shown in FIG. The error value | e (n) | from the error detection unit 107 is input to the error storage unit 201 of the step size generation unit 109. The error storage unit 201 is a register composed of a DFF (D flip-flop) circuit having a required number of bits, and is driven by a clock having a period T synchronized with the symbol of the received signal. The error value | e (n−1) one symbol before | Is output.
[0039]
The division unit 202 divides the output | e (n−1) | from the error storage unit 201 and the output | e (n) | from the error detection unit 107, and the division result | e (n) | / | E (n-1) | is output. The step size storage unit 203 is a register composed of a DFF (D flip-flop) circuit having a required number of bits, and is driven by a clock having a period T synchronized with the symbol of the received signal. The step size μ (| e (n− 1) |) is output. The multiplication unit 204 multiplies the output μ (| e (n−1) |) from the step size storage unit 203 and the output | e (n) | / | e (n−1) | from the division unit 202. This results in a new step size μ (| e (n) |).
[0040]
That is, the step size generation unit 109 obtains a new step size μ (| e (n) |) by executing the following expression, and sends this to the tap coefficient update unit 109. ing.
[0041]
μ (| e (n) |) = μ (| e (n−1) |) × (| e (n) | / | e (n−1) |) (3)
As apparent from the expression (3), the new step size μ (| e (n) |) is set large when the absolute value | e (n) | of the error signal is relatively large, and is small. If so, set it small. As a result, the residual error is suppressed and the convergence calculation can be performed at high speed.
[0042]
The operation of the adaptive equalizer 100 of the present embodiment configured as described above will be described. First, prior to the training period, initial values e0 and μ0 are set in the error storage unit 201 and the step size storage unit 203, respectively. At a certain time n, the error storage unit 201 stores | e (n−1) |, and the step size storage unit 203 stores μ (| e (n−1) |).
[0043]
The input signal is processed by the transversal filter 101, and the signal processed by the transversal filter 101 and the signal processed by the transversal filter 102 are added by the adder 103, and equalization processing is performed. Next, the data of the output of the adder 103 is determined by the determiner 104, whereby demodulated data is obtained.
[0044]
In the training period, the switch 106 is connected to the training signal generation unit 105 side, and the training sequence is supplied to the error detection unit 107 as the reference signal d. At the same time, the tap coefficient updating unit 108 uses the tap coefficients c ₁ , c ₂ ,..., C _M , c _{M + 1} , c _{M + 2} , so that the mean square error of the error signal e (n) is minimized. , C _{M + N} are updated sequentially.
[0045]
At this time, the tap coefficient vector w (n + 1) at time t = (n + 1) τ is sequentially updated using the following formulas with each value at time t = nτ.
[0046]
w (n + 1) = w (n) + μ (| e (n) |) x (n) e ^* (n) (4)
In equation (4), μ (| e (n) |) is a step size that changes according to the error signal e (n) that is output from the error detection unit 107, and is supplied from the step size generation unit 109. . The subscript n of each variable means each value at time t = nτ, and x (n), w (n), and e (n) in equation (4) are values obtained by the following equation.
[0047]
x (n) = (x ₁ x ₂ ... x _M x _{M + 1} ... x _{M + N} ) ^T
w (n) = (c ₁ c ₂ ... c _M c _{M + 1} ... c _{M + N} ) ^T
e (n) = d−w ^H x (5)
Here, * in the equation (4) represents a complex conjugate, a superscript T in the equation (5) represents a matrix transpose operation, and a superscript H represents a complex conjugate transpose operation. The tap coefficient updating unit 108 converges the tap coefficient within the training period by sequentially repeating the above calculation while using the step size μ (| e (n) |) generated by the step size generation unit 109.
[0048]
On the other hand, during a period other than the training period, the switch 106 is connected to the determiner 104 side, and when the ideal signal corresponding to the demodulated data is transmitted, the value at the time of no distortion and no noise is supplied to the error detection unit 107 as the reference signal d. Is done. Also at this time, the tap coefficient updating unit 108 continuously updates the tap coefficient by the above sequential update formula while using the step size μ (| e (n) |) generated by the step size generating unit 109, The adaptive operation is performed so that the average error power between the demodulated data and the output of the adder 103 is minimized.
[0049]
As described above, the adaptive equalizer 100 according to the present embodiment includes the step size generation unit 109. When the absolute value | e (n) | of the error signal is relatively large, the step size is increased. By reducing the size, the residual error in the adaptive algorithm is suppressed, and the tap coefficients can be converged in a short time.
[0050]
That is, when the error value is large, a large step size is selected, so that the tap coefficient updating unit 108 executes a tap coefficient calculation calculation that converges the large error to 0 at high speed. On the other hand, when the error value is small, a small step size is selected, so that the tap coefficient updating unit 108 executes a tap coefficient calculation operation that converges the residual error to 0 with high accuracy. As a result, it is possible to achieve both an increase in convergence speed and an effect of suppressing residual error.
[0051]
FIG. 3 shows experimental results when equalization processing is performed using the adaptive equalizer 100 of the embodiment shown in FIG. 1 and the conventional adaptive equalizer 10 shown in FIG. In FIG. 3, the horizontal axis represents the number of received samples, and the vertical axis represents the square error in decibels. In FIG. 3, the transmission path conditions are unchanged among 200 samples, and the initial 50 samples represent the training period. As can be seen from the experimental results, the adaptive equalization process of the present embodiment in which the step size is changed according to the error has a faster convergence speed than the conventional adaptive equalization process in which the step size is fixed. I understand that.
[0052]
According to the above configuration, an adaptive algorithm calculation is used in accordance with an error between the output of the transversal filters 101 and 102 for filtering the input reception signal and the demodulated data or the reference signal obtained by determining the output. By changing the step size of the tap coefficient updating unit 108 that updates the tap coefficients of the transversal filters 101 and 102, an adaptive equalizer that requires a small amount of calculation for signal processing and has a fast convergence of the adaptive operation. 100 can be realized.
[0053]
As a result, for example, it is possible to apply to a case where a signal having only a short training period is received, which can be configured by a simple circuit or an inexpensive processor, and further it is possible to improve transmission efficiency.
[0054]
In the above-described embodiment, the case where the step size generator is configured as shown in FIG. 2 has been described. However, the present invention is not limited to this, and may be configured as shown in FIG. 4, for example. In FIG. 4, the step size generation unit 300 includes an address generation unit 302 and a step size storage unit 301.
[0055]
The step size storage unit 301 has a memory or register configuration, stores a finite number of step size candidate values, and is driven by a clock having a period T synchronized with a symbol of the received signal. The address generation unit 302 generates an address for the step size storage unit 301 based on the output e (n) from the error detection unit 107 (FIG. 1), and the error signal e (n) from the error detection unit 107 is large. The address value storing a value having a large step size in the step size storage unit 301 is generated. As a result, a step size close to the step size according to the above-described equation (3) can be output from the step size storage unit 301.
[0056]
FIG. 5 shows experimental results when equalization processing is performed using the adaptive equalizer to which the configuration shown in FIG. 4 is applied as the step size generation unit 109 in FIG. 1 and the conventional adaptive equalizer shown in FIG. Shown in In FIG. 5, the horizontal axis represents the number of received samples, and the vertical axis represents square error in decibels. Further, in FIG. 5, the condition of the transmission path is unchanged among 200 samples, and the initial 50 samples represent the training period. As can be seen from the experimental results, even when the step size generator 300 as shown in FIG. 4 is used, the convergence speed equivalent to that of the step size generator 109 shown in FIG. 2 can be obtained. Further, if the step size generation unit 300 in FIG. 4 is used as the step size generation unit, the circuit configuration can be simplified as compared with the step size generation unit 109 in FIG.
[0057]
Further, in the above-described embodiment, the case where the present invention is applied to the decision feedback type adaptive equalizer 100 has been described. However, if the transversal filter 102 is removed, the rest of the configuration is a linear type with the same configuration. This is the configuration of the adaptive equalizer. In this case, the configuration becomes simpler and the same effects as those of the above-described embodiment can be obtained. That is, the present invention can be applied not only to the decision feedback type adaptive equalizer but also to a linear type adaptive equalizer.
[0058]
In the above-described embodiment, the case where the LMS (Least Mean Square) algorithm is used as the adaptive algorithm in the tap coefficient updating unit 108 has been described. However, the present invention is not limited to this, and for example, an RLS (Recursive Last Square) algorithm. May be used.
[0059]
Here, FIG. 6 shows an example when the adaptive equalizer 100 of the embodiment is applied to an actual receiving apparatus. The receiving apparatus 401 inputs the received signal converted into the baseband signal to the adaptive equalizer 100. The received signal subjected to equalization processing by the adaptive equalizer 100 is subjected to error correction processing by an error correction unit (for example, Viterbi decoding) 403 having a predetermined error correction function.
[0060]
The reception performance evaluation unit 404 evaluates the reception performance of the adaptive equalizer 100 based on the bit error rate of the signal after error correction. Specifically, the reception performance is compared with a predetermined threshold value, and the comparison result is sent to the control unit 402. The control unit 402 controls whether or not to operate the step size generation unit 109 of the adaptive equalizer 100 according to the reception performance.
[0061]
Specifically, first, the control unit 402 controls off the step size generation unit 109 of the adaptive equalizer 100. At this time, if the reception performance obtained by the reception performance evaluation unit 404 is good, the step size generation unit 109 is kept off. On the other hand, when the reception performance obtained from the reception performance evaluation unit 404 is poor, the step size generation unit 109 is turned on. Thereby, in the receiving apparatus 401, the step size generating unit 109 generates a step size corresponding to the error, and the tap coefficient updating unit 108 has sufficient reception performance without performing the adaptive algorithm calculation using this step size. If it is obtained, power consumption can be suppressed by not operating the step size generator 109.
[0062]
【The invention's effect】
As described above, according to the present invention, the adaptive algorithm calculation is performed according to the error between the output of the transversal filter for filtering the input received signal and the demodulated data or the reference signal obtained by determining this output. By adapting the step size of the tap coefficient updater that updates the tap coefficient of the transversal filter using the adaptive method, it is possible to obtain an equalized signal with a fast convergence of the adaptive operation and a small error An equalizer can be realized.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an adaptive equalizer according to an embodiment of the present invention. FIG. 2 is a block diagram showing a configuration of a step size generator. FIG. 3 shows an adaptive equalizer according to an embodiment. FIG. 4 is a block diagram showing another configuration example of the step size generation unit. FIG. 5 is an adaptive equalizer when the step size generation unit of FIG. 4 is applied. FIG. 6 is a block diagram showing an example when the adaptive equalizer of the embodiment is applied to a receiving apparatus. FIG. 7 is a block diagram showing an example of the conventional adaptive equalizer. Block diagram showing the configuration 【Explanation of symbols】
100 Adaptive equalizer 101, 102 Transversal filter 103 Adder 104 Determinator 105 Training signal generator 107 Error detector 108 Tap coefficient updater 109, 300 Step size generator 201 Error storage unit 202 Dividers 203, 301 Step size Storage unit 204 Multiply unit 302 Address generation unit 401 Receiving device 402 Control unit 403 Error correction unit 404 Reception performance evaluation unit

Claims (2)

A first transversal filter that receives the received signal converted into the baseband signal, a second transversal filter that receives the determined demodulated data, the first transversal filter, and the second transversal filter An adder for adding the output of the transversal filter, a determiner for comparing the output of the adder with a predetermined threshold value for each symbol period of the received signal and determining a received signal to obtain demodulated data, and reference An error detection unit that outputs a signal or a difference between the demodulated data and an output signal of the adder as an error signal, and the error signal and the tap coefficients of the first and second transversal filters, The tap coefficients of the first and second transversal filters are updated by performing an adaptive algorithm calculation so that becomes smaller And-up coefficient updating unit, as well as generating the step size used in the tap coefficient updating unit, an adaptive equalizer comprising a step size generating unit that changes according to the step size to said error signal, and
The step size generation unit includes an error storage unit that temporarily stores an error signal output from the error detection unit, a step size storage unit that temporarily stores an output signal from the step size generation unit, and an error storage unit A divider that calculates an absolute value of a ratio of an output and an output from the error detector; and a multiplier that calculates a product of an output from the divider and an output from the step size storage unit, The adaptive equalizer characterized in that the step size is generated by the output of the multiplier with the output from the detection unit as an input . A transversal filter that receives a received signal converted into a baseband signal, and compares the output of the transversal filter with a predetermined threshold value for each symbol period of the received signal to determine the received signal and to generate demodulated data Using a determination unit that obtains the difference between the reference signal or the demodulated data and the output signal of the transversal filter as an error signal, the error signal and the tap coefficient of the transversal filter, By performing an adaptive algorithm calculation so as to reduce the error signal, a tap coefficient update unit that updates the tap coefficient of the transversal filter and a step size used in the tap coefficient update unit are generated, and the step size is comprising: a step size generating unit that is changed according to the error signal, the A that the adaptive equalizer,
The step size generation unit includes an error storage unit that temporarily stores an error signal output from the error detection unit, a step size storage unit that temporarily stores an output signal from the step size generation unit, and an error storage unit A divider for calculating an absolute value of a ratio of an output and an output from the error detection unit; and a multiplier for calculating a product of an output from the divider and an output from the step size storage unit, The adaptive equalizer characterized in that the step size is generated by the output of the multiplier with the output from the detection unit as an input .

Images