JP4553663B2 - Pull-in method and blind adaptive equalizer in blind adaptive equalizer - Google Patents

Description

  The present invention relates to an adaptive equalizer applied to a demodulator in the field of digital communication, and more particularly to a pull-in method and a blind adaptive equalizer in a blind adaptive equalizer that does not require a training signal (reference signal).

  In the digital communication field, an adaptive equalizer is used when restoring an original transmission signal from a deteriorated received signal. Conventionally, this adaptive equalizer has a series circuit configuration of a carrier recovery circuit and a decision feedback equalizer, and a training signal is used in the decision feedback equalizer (see Non-Patent Document 1).

"Study of performance desired for 7GHz band digital FPU and waveform equalization method of QAM FPU", Journal of the Institute of Image Information and Television Engineers, Vol. 51, no. 9, pp. 1550-1559 (1997) (FIGS. 6 and 7)

  However, since the adaptive equalizer according to the above prior art requires a training signal, communication interruption or redundant training signal is necessary, and it takes time to pull in and the configuration is complicated. is there.

  The present invention has been made in consideration of such problems, and an object thereof is to provide a pull-in method and a blind adaptive equalizer in a blind adaptive equalizer that is easy to pull in and has a simple configuration.

  The pull-in method in the blind adaptive equalizer of the present invention includes a multiplier that multiplies a complex input signal and a carrier recovery signal, a decision feedback type that receives an output of the multiplier and obtains an equalization output signal and a determination output signal, etc. In a pull-in method in a blind adaptive equalizer comprising an equalizer, and a carrier recovery circuit that is supplied with the equalization output signal and the determination output signal and outputs the carrier recovery signal, tap coefficients of the determination feedback equalizer Is updated with a CMA algorithm that does not affect the phase rotation of the carrier, and carrier recovery of the carrier recovery circuit is started using phase error detection with a limited determination range, the determination feedback type, etc. After the equalization error signal in the equalizer becomes small and the carrier regeneration in the carrier regeneration circuit converges, the decision feedback type etc. And a tap end coefficient step for switching so that the carrier regeneration of the carrier regeneration circuit is performed using phase error detection that does not limit the determination range. Features.

  According to the present invention, the tap coefficient of the decision feedback equalizer is updated by the CMA (Constant Modulus Algorithm) algorithm that does not affect the phase rotation of the carrier and the determination range of the carrier regeneration circuit is limited in the step of starting pull-in. Carrier recovery is started using the detected phase error detection, and after the equalization error signal in the decision feedback equalizer becomes small and the carrier regeneration in the carrier recovery circuit converges, the decision feedback type The updating of the equalizer tap coefficient is switched to the decision-oriented algorithm, and the carrier regeneration of the carrier regeneration circuit is switched to carrier regeneration using phase error detection that does not limit the determination range.

  Such a configuration eliminates the need for a training signal and makes it easy to pull in even under a fading environment that is practically inferior. As a result, the pull-in time can be shortened.

  Here, a Godard algorithm can be adopted as the CMA algorithm, and an LMS (Least Mean Square) algorithm or an RLS (Recursive Least Square) algorithm can be adopted as the decision-oriented algorithm.

  Note that phase error detection in which the determination range of the carrier reproduction circuit is limited in the pull-in start step is performed from the origin on the IQ coordinate in a multilevel QAM (Quadrature Amplitude Modulation: 16QAM, 64QAM, 256QAM, 1024QAM,. Only the outermost 4 symbols or the neighboring symbols including these 4 symbols, for example, 12 (3 × 4) symbols, or 24 (6 × 4) symbols, etc. are detected. By limiting the determination range, the symbol pull-in range at the start of pull-in can be expanded.

  In this case, it is an input signal (received signal) to the carrier reproduction circuit in order to easily distinguish whether the outermost four symbols or the neighboring symbols including these four symbols are within the determination range. The absolute values of the I and Q values of the digitized output signal are collected in the first quadrant, and further rotated in the negative direction by π / 4 so that they overlap on the I axis or on and near the I axis (actual Will vary in the vicinity of the I axis). The determination range is a rectangular range including the symbol point superimposed on the I axis or the symbol point adjacent to the symbol point. By determining whether or not there is a symbol within the rectangular range, the software and hardware configuration can be simplified.

  A blind adaptive equalizer according to the present invention includes a multiplier that multiplies a complex input signal and a carrier reproduction signal, a decision feedback type that is supplied with an output of the multiplier and obtains an equalization output signal and a determination output signal, and the like. A blind adaptive equalizer comprising: an equalizer; and a carrier recovery circuit that is supplied with the equalization output signal and the determination output signal and outputs the carrier recovery signal, wherein the determination feedback equalizer is tapped with a CMA algorithm A tap first updater for updating a coefficient and a tap second updater for updating a tap coefficient with a decision-oriented algorithm, wherein the carrier recovery circuit limits a determination range and performs a first phase error for carrier recovery A detector and a second phase error detector that performs carrier recovery in the entire range that does not limit the determination range, and at the start of pulling, the tap first updater and the first phase error Play the carrier with out unit, at the time of retraction ends, and performs the reproduction of carrier using the second phase error detector and the tap second updater.

  With this configuration, no training signal is required and the pull-in time can be shortened.

  Also in this case, the first phase error detector detects only the outermost four symbols or the neighboring symbols including these four symbols when viewed from the origin on the IQ coordinate in the multi-level QAM modulation method. It is preferable that the phase error is detected by limiting the determination range within the region.

  According to the present invention, at the time of pull-in in the blind adaptive equalizer that does not use the training signal, first, the pull-in start processing is performed by limiting the determination range of the carrier reproduction circuit and adopting the CMA algorithm, and at the end, carrier regeneration is performed. Since it has been devised to perform the process at the end of pulling in which the determination range of the circuit is the entire range and adopts the decision-oriented algorithm, the pulling time can be shortened. In addition, the configuration of the blind adaptive equalizer can be simplified.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 shows a configuration of a blind adaptive equalizer 10 in which a pull-in method in a blind adaptive equalizer according to an embodiment of the present invention is implemented.

  This blind adaptive equalizer 10 basically includes a carrier recovery circuit 12 for recovering a carrier and a conjugate signal exp (−) of the carrier recovery signal exp (jθn) recovered by the carrier recovery circuit 12. jθn) and the complex input signal r (n) = I + jQ, and the output {r (n) × exp (−jθn)} = u ′ (n) of the multiplier 14 is the input signal u ′ ( a decision feedback equalizer 16 which is supplied as n) and generates an equalized output signal y (n) = yi + jyq and a decision output signal y ^ (n) = I ^ + jQ ^.

  Here, the variable n is a sample time, which is taken as Ts / 2 in this embodiment. Θn is the carrier reproduction phase at time n. j is an imaginary unit. Further, I means an I coordinate value, and Q means a Q coordinate value. Further, “^” means a determination output.

  Here, the carrier reproduction circuit 12 includes a first phase error detector (Phase Detector) 21 that detects a phase error by limiting a determination range from the equalized output signal y (n), and an equalized output signal y (n). Alternatively, the output of the second phase error detector 22 that detects the phase error in the entire range that does not limit the determination range from the determination output signal y ^ (n), and the outputs of the first phase error detector 21 and the second phase error detector 22. According to the output value of the loop filter 26, a loop filter 26 that filters the phase error output from the first and second phase error detectors 21 and 22, and generates a low-frequency signal. And a numerically controlled oscillator 28 that outputs a carrier reproduction signal exp (jθn) having a different frequency.

  Here, the first phase error detector 21 determines the carrier phase error determination range of the detected symbols, as shown in FIG. 2, for example, in the 64 QAM system, the four outermost contours when viewed from the origin on the IQ coordinate. The phase error detection is performed only in the ring-shaped frame region 66 for detecting only the symbol (symbol with the maximum amplitude) 79.

  The second phase error detector 22 is configured to detect a carrier phase error for all 64 symbols.

  On the other hand, the decision feedback equalizer 16 includes a plurality of (in this embodiment, four) 1/2 symbol time (Ts / 2) delay units 52 and a corresponding number (in this embodiment, five) taps. A feed forward filter 30 composed of a coefficient unit 54 and an adder 33 for adding the outputs of the tap coefficient units 54, and a plurality (two in this embodiment) of one symbol time (Ts) delay units 56, a corresponding number (two in this embodiment) of tap coefficient units 58, a feedback filter 32 including an adder 34 that adds the outputs of the tap coefficient units 58, and a feed forward filter 30. And an adder 35 that adds the outputs of the feedback filter 32 and outputs an equalized output signal y (n).

  The decision feedback equalizer 16 is further supplied with the equalized output signal y (n) = yi + jyq to determine the symbol position, and the determination output signal y ^ (n) = I ^ representing the symbol position after the determination. A determination unit 36 that outputs + jQ ^, a tap first updater 41 that updates tap coefficients of the tap coefficient units 54 and 58 based on the equalization output signal y (n), and an equalization output signal y (N) or the tap second updater 42 that updates the tap coefficients of the tap coefficient units 54 and 58, the tap first updater 41, and the tap second updater 42 are switched based on the determination output signal y ^ (n). A tap updater 38 including a switch 44 is provided.

  Here, the tap first updater 41 is configured to update the tap coefficient with a CMA algorithm such as the Godard algorithm that does not affect the carrier phase rotation, and the tap second updater 42 includes the tap coefficient. Is updated with a decision-oriented algorithm such as an LMS algorithm or an RLS algorithm.

  For example, the tap coefficient update formula by the Goddard algorithm is given by the following formula (1) as is well known, and the tap coefficient update formula by the LMS algorithm is given by the following formula (2), as is well known.

  That is, in the equation (1), the current tap coefficient w (n + 1) is equal to the previous tap coefficient w (n), and the equalization error signal e () regarding the input signal u (n) and the equalized output signal y (n). n) plus the term multiplied by the conjugate. In the equation (2), the current tap coefficient w (n + 1) is equal to the previous tap coefficient w (n), the input signal u (n), the equalized output signal y (n), and the determination output signal y ^ (n). Is obtained by adding a term multiplied by the conjugate of the equalization error signal e (n).

w (n + 1) = w (n) + μ1u (n) e ^* (n), e (n) = y (n) (R2−y (n) | ² ) (1)
w (n + 1) = w (n) + μ2u (n) e ^* (n), e (n) = y ^ (n) −y (n) (2)
w (n + 1), w (n): Tap coefficient (vector)
u (n): input signal (vector)
e (n): error signal e ^* (n): conjugate error signal. “ ^* ” Represents a conjugate.
y (n): equalized output signal {y (n) = w ^* (n) u (n) = yi + jyq}
y ^ (n): judgment output signal R2: constant μ1, μ2: step size

  Further, the blind adaptive equalizer 10 performs overall control such as switching of the switches 24 and 44 by transmitting / receiving signals to / from the carrier recovery circuit 12 and the decision feedback equalizer 16 and performing necessary calculations. The control unit 60 which performs automatically is provided.

  The blind adaptive equalizer 10 shown in FIG. 1 uses analog signals I and Q obtained through a low noise amplifier, a BPF (Band Pass Filter), a down converter, and a quadrature detector, respectively, in a radio receiver, for example. It is a digital signal processor (Digital Signal Processor) for processing a complex input signal r (n) = I + jQ converted into a digital signal by an A / D converter.

  Next, the operation of the above-described embodiment will be described based on the operation flow shown in FIG.

  First, the control unit 60 connects the switch 44 of the decision feedback equalizer 16 to the tap first updater 41 side in the pull-in start step after the time t0 (demodulation start point) in the pull-in mode, and makes a decision feedback. The tap coefficient of the type equalizer 16 is updated by a CMA algorithm that does not affect the phase rotation of the carrier, and the switch 24 of the carrier recovery circuit 12 is connected to the first phase error detector 21 side to determine the carrier recovery circuit 12 Carrier reproduction is started using phase error detection with a limited range.

  At this time, the initial value of the tap coefficient is, for example, the tap coefficient F0 of the feedforward filter 30 is set to “1”, the tap coefficient of the tap coefficient unit 54 of the other feedforward filter 30 and the tap coefficient unit 58 of the feedback filter 32. All coefficients are set to “0”.

  As can be seen from the well-known derivation principle, in the CMA algorithm, only the amplitude information of the input signal u (n) is targeted for evaluation, and therefore it does not affect the phase fluctuation, for example, the presence or absence of a carrier frequency offset. For this reason, even if the input signal u (n) in which the frequency offset between transmission and reception is still included in the input signal u (n) of the decision feedback equalizer 16 at the beginning of reception, the tap coefficient is The eye pattern of the equalized output signal y (n) is opened as it is updated.

FIG. 4 shows the DFE average equalization error power Pe (Pe is the square of the absolute value of the input / output of the determiner 36, that is, | y ^ (n) −y (n ) | It is an average value of ² or a moving average value.

  FIG. 5 shows a temporal change in the carrier frequency error fe [Hz] obtained from the loop filter 26.

  As can be seen from FIG. 5, since the eye pattern opens to some extent after the time t0 at the start of the pull-in, the first phase error detector 21 of the carrier recovery circuit 12 operates normally, and the frequency offset Thus, the frequency offset is reduced and the carrier frequency error fe is reduced.

  FIG. 6 is an explanatory diagram showing an operation procedure of the first phase error detector 21. In the first phase error detector 21, only the outermost symbol (four outer symbols) 79 of 64QAM multilevel modulation is used for phase error detection. In FIG. 6, only the first quadrant is shown, but it is confirmed for each symbol whether or not the reception point is in the determination range 70, and if it is within the determination range 70, the ideal point (the four outermost points) And the calculated phase error is supplied to the loop filter 26.

  The determination range 70 is an original determination range, but software and hardware for discriminating whether or not the equalized output signal y (n) = yi + jyq that is a received signal is within the determination range 70 is complicated. Therefore, in this embodiment, the phase error detection is performed using the new determination range 72.

  The newly set determination range 72 will be described with reference to FIG.

  In order to use the new determination range 72, first, in the equalization output signal y (n) that is the reception point, I component = yi, Q component = yq absolute values | yi |, | yq | Collect in one quadrant.

  Second, the equalized output signal y (n), which is the reception point consisting of the outermost symbols 79 collected in the first quadrant, is multiplied by exp (−jπ / 4) and rotated 45 ° to the right to enter the first quadrant. The received points including the collected outermost symbol 79 are collected on the I axis. This value is equalized output signal y ′ (n) = y′i + jy′q.

  Third, it is confirmed whether the equalized output signal y ′ (n) = y′i + jy′q, which is a reception point, is within the rectangular determination range 72. The rectangular determination range 72 is made adjustable. That is, the determination range 72 is determined by setting the equalization output determination amplitude median, the equalization output determination amplitude range half value, and the equalization output determination phase range half value by the control unit 60.

  Fourth, when the equalized output signal y ′ (n) = y′i + jy′q that is the reception point is within the rectangular determination range 72, the equalized output signal that is the received signal with the reference symbol point Rs. -sign (yq) · yi + sign (yi) · yq, which is a phase error of y ′ (n) = y′i + jy′q, is supplied to the loop filter 26. If it is not within the determination range 72, the previous phase error is supplied to the loop filter 26. This phase error detection is less susceptible to errors due to fading because the carrier amplitude (symbol point amplitude) is large. The sign function sign (X) is a function that takes +1 when X is positive and -1 when X is negative.

  During the operation at the setting of the pull-in start time t0, that is, the pull-in state in which the determination range of the carrier regeneration circuit 12 is limited to the four outermost symbols 79 and the tap first updater 41 based on the CMA algorithm is adopted. In this embodiment, when the value of the DFE average equalization error power Pe shown in FIG. 4 is equal to or lower than a predetermined threshold value or after a predetermined time has elapsed since time t0 during the operation at the start step, At the time point t1 in the mode, the switch 44 is switched to the tap second updater 42 side, and the update of the tap coefficient of the decision feedback equalizer 16 is switched to the LMS algorithm. By switching to the LMS algorithm, tap coefficients converge more accurately.

  Next, at the time point t2 in the pull-in mode, in other words, at the time point t2 (see FIG. 5) when the carrier frequency error fe which is a residual component after the compensation of the frequency offset is almost eliminated, the switch 24 detects the first phase error. Switching from the detector 21 side to the second phase error detector 22 side, the determination range is the entire range, the phase error of all symbols is detected, and carrier recovery is performed. At this time, since the tap updater 38 is switched to the tap second updater 42 side that performs the tap update by the LMS algorithm by the switch 44, the equalization error power is further reduced and the frequency of erroneous determination is also reduced. .

  After this time t2, since the equalization error signal in the decision feedback equalizer 16 becomes small and the carrier regeneration in the carrier regeneration circuit 12 has converged, the tap coefficient of the decision feedback equalizer 16 is Updating is performed by the LMS algorithm by the tap second updater 42, and the carrier regeneration of the carrier regeneration circuit 12 is operated in a step at the end of pulling using phase error detection in the entire range that does not limit the determination range.

  Next, as shown in FIG. 3, the demodulated data is determined to be valid and pulled in at time t3 after the elapse of a predetermined time from time t2 or at time t3 when the DFE average equalization error power Pe becomes equal to or less than a predetermined threshold. The complex input, which is the received signal, ends the mode and remains in the operation form at the end of the pull-in step, that is, the carrier regeneration using the second phase error detector 22 and the operation form using the tap second updater 42. A tracking mode is entered in which data is reproduced by the determination output signal y ^ (n) of the signal r (n).

  As described above, in the above-described embodiment, as shown in FIG. 3, a training signal (reference signal) becomes unnecessary by having a pull-in start step and a pull-in end step in the pull-in mode. In practice, the blind adaptive equalizer 10 that can be easily pulled in even under a fading environment can be realized. As a result, the pull-in time (the pull-in mode duration: t3-t0) can be shortened.

  Note that either of the switches 44 and 24 at the time point t1 and the time point t2 between the pull-in start step and the pull-in end step may be switched first or simultaneously. At the time of actual installation, the switching order can be selected so that the pull-in time in the pull-in mode is shortened in the fading environment at the installation point or reception time.

  The present invention is not limited to the above-described embodiment. For example, the decision feedback equalizer 16 is changed to a linear equalizer, or the first phase error detector 21 has the outermost contour of 64QAM multilevel modulation. Various configurations based on the description in this specification, such as the use of symbols (four outer symbols) 79 and inner symbols 80 and 81 in the vicinity thereof (see FIG. 6, a total of eight symbols) for phase error detection, etc. Of course, it can be taken.

  In the first phase error detector 21, when a total of 12 symbols 79 to 81 of the outline are used, a total of 12 symbols is used, the determination range 72 is widened, so the probability that a symbol can be detected increases. There is a possibility that the convergence time will be shorter, but at the same time, the convergence time may become longer due to an increase in the error, which is determined by these two trade-offs. Therefore, although not shown, it is possible to use a total of 24 symbols, that is, six symbols on the outline.

It is a block diagram of the blind adaptive equalizer which concerns on one Embodiment. It is explanatory drawing of the determination range in the step at the time of a pull-in start. It is an operation | movement flowchart of one Embodiment. It is explanatory drawing of DFE average equalization error electric power. It is explanatory drawing of a carrier frequency error. It is explanatory drawing of the phase error detection using only the outermost symbol. It is another explanatory view of phase error detection using only the outermost symbol.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Blind adaptive equalizer 12 ... Carrier recovery circuit 14 ... Multiplier 16 ... Decision feedback equalizer (DFE)
21 ... First phase error detector 22 ... Second phase error detector 24, 44 ... Switch 30 ... Feed forward filter 32 ... Feedback filter 41 ... Tap first updater 42 ... Tap second updater 79, 80, 81 ... symbol

Claims (4)

A multiplier that multiplies the complex input signal and the carrier recovery signal; a decision feedback equalizer that is supplied with an output of the multiplier and obtains an equalization output signal and a decision output signal; the equalization output signal and the decision output In a pull-in method in a blind adaptive equalizer comprising a carrier recovery circuit that is supplied with a signal and outputs the carrier recovery signal,
The coefficient of the tap of the decision feedback equalizer is updated with a CMA algorithm that does not affect the phase rotation of the carrier, and carrier regeneration of the carrier regeneration circuit is started using phase error detection with a limited decision range. A step at the beginning of the withdrawal,
A first switching step for switching the tap coefficient of the decision feedback equalizer to be updated by a decision-directed algorithm when an equalization error signal in the decision feedback equalizer becomes a predetermined threshold value or less ;
After the execution of the first switching step, when the carrier regeneration in the carrier regeneration circuit has converged, the carrier regeneration of the carrier regeneration circuit is switched to be performed using phase error detection that does not limit the determination range . 2 switching steps;
And updating the tap coefficient of the decision feedback equalizer with the decision-directed algorithm, and carrying out carrier regeneration of the carrier regeneration circuit using phase error detection that does not limit the decision range. A pull-in method in a blind adaptive equalizer. The retraction method according to claim 1,
In the multi-level QAM modulation method, the phase error detection that limits the determination range of the carrier recovery circuit in the step of starting pulling is performed by using only the four outermost symbols or the four symbols as viewed from the origin on the IQ coordinate. A pull-in method in a blind adaptive equalizer, which is performed by limiting a determination range within a region in which a nearby symbol including the symbol is detected. A multiplier that multiplies the complex input signal and the carrier recovery signal; a decision feedback equalizer that is supplied with an output of the multiplier and obtains an equalization output signal and a decision output signal; the equalization output signal and the decision output A blind adaptive equalizer comprising a carrier recovery circuit that is supplied with a signal and outputs the carrier recovery signal;
The decision feedback equalizer includes a tap first updater that updates a tap coefficient with a CMA algorithm, and a tap second updater that updates a tap coefficient with a decision-oriented algorithm,
The carrier recovery circuit includes a first phase error detector that performs carrier recovery by limiting a determination range, and a second phase error detector that performs carrier recovery in an entire range that does not limit the determination range;
At the start of pull-in, the carrier is regenerated using the first tap updater and the first phase error detector, and at the end of pull-in, the carrier is regenerated using the second tap updater and the second phase error detector. the stomach line,
A blind adaptive equalizer characterized in that a carrier is regenerated by using the second tap updater and the first phase error detector between the start of pull-in and the end of pull-in . The blind adaptive equalizer according to claim 3,
In the multi-level QAM modulation system, the first phase error detector makes a determination within an area in which only the outermost four symbols as viewed from the origin on the IQ coordinates or a nearby symbol including the four symbols is detected. Blind adaptive equalizer characterized in that phase error is detected in a limited range.

Images