JPH07202636A - Digital filter and nyquist point retrieving method - Google Patents

Abstract

PURPOSE:To miniaturize and make a device light in weight by thinning the prescribed range of the calculation of the evaluation value in a Nyquist point retrieving without performing a tap calculation for the prescribed number of delay elements. CONSTITUTION:By the delay elements 1, 2 and 3,... of serial connection, a multiplier 4 and an adder 5, a non-recursive digital filter is composed. The multiplier 4 is provided for every three delay elements, for instance, thins the tap calculation for a filtering and reduces arithmatic amount per unit time. But, since the data after the filter passes is outputted for every delay time TM per delay element, the sampling period for the calculation of the evaluation value in retrieval of a Nyquist point can be set to be short. Thus, the accuracy of the estimation of a Nyquist point is improved. While an evaluation value Ws is apart from a prescribed value Pj, the execution of the calculation of the evaluation value is also thinned for every 3TM and further, arithmetic amount per unit time is reduced.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital filter and a Nyquist point search method. More specifically, the present invention relates to a digital filter used for matching transmission characteristics in digital communication with Nyquist conditions, and a Nyquist point search method for searching a time at which intersymbol interference does not occur on the receiving side. .

The invention is particularly, but not exclusively, applicable to mobile radio receivers, which are expected to use 16QAM.

[0003]

2. Description of the Related Art In digital communication, it is known that intersymbol interference does not occur if transmission characteristics satisfy the Nyquist condition. Therefore, in order to form such a transmission characteristic, it is common practice to provide a filter in the transmitter and the receiver. In addition, such a filter
In this specification, it is referred to as a Nyquist filter or a matched filter.

According to the Nyquist theory, if the transmission characteristics satisfy the Nyquist condition, the received waveform r (t) of code r ₀ is
As shown in (a) of FIG. 9B, r (t ₀ ) = r ₀ r (t ₀ ± kT) = 0. Here, time t ₀ is the reception time of code r ₀ , and k is k ≧
An integer of 1 and T is a code interval. Note that t ₀ , (t ₀ ± k
In this specification, the time at which intersymbol interference between the codes does not occur, such as T), is referred to as the Nyquist point.

The Nyquist filter is a FIR (Finit
e Impulse Response) filter or a digital filter called an acyclic filter. However, the application of such a digital filter is not limited to the Nyquist filter.

FIG. 9 (a) is a signal flow diagram showing the structure of such an FIR filter. In the figure (a), 20
Reference numeral 0 is a delay unit that delays the input signal by one clock and outputs the delayed signal. The input and output of each delay means 200 are called taps. 201 is a multiplication means. Also, H _{1 to} H
_n is a coefficient by which each tap is multiplied.

The FIR filter connects the delay means 200 and the coefficient H is applied to the taps formed by their inputs and outputs.
_{1 to} H _n are respectively multiplied, and all the multiplication results are added. In addition, such multiplication
In this specification, this is called tap calculation.

In FIG. 1A, it is shown that the addition of the multiplication results is performed by the addition means 202, and the addition result is the filtered signal Y (n).

As described above, in the FIR filter, n (n-1) delay means are connected to n multiplication means and n delay means are connected.
Number of coefficients and addition means for adding n number of multiplication results.

In the received signal passed through the Nyquist filter, if the transmission characteristics are adjusted by the Nyquist filter so as to satisfy the Nyquist condition, there should be a Nyquist point. However, on the receiving side, the Nyquist point is not known even though the code interval T is known.

Therefore, conventionally, there has been proposed a Nyquist point search method in which a known signal is inserted into a transmission signal and the receiving side detects the known signal from the received signal to estimate the Nyquist point.

Such a conventional Nyquist point search method will be described with reference to FIG. 9 (b). In (b) of the figure, (a) is the known signal component of the received signal. However, another code component such as (a) is also superimposed on the received signal. As described above, t ₀ and (t ₀ ± kT) are Nyquist points.

(C) is an evaluation value that indicates the correlation between the received signal and the known signal. The calculation method of such an evaluation value is
It is set based on the properties of the known signal, such as the phase and the amplitude. Further, the calculation method is set so that the evaluation value becomes an extreme value at the reception time t ₀ of the known signal.

The calculation of the evaluation value is performed at a time of t ₁ + k _r U (k _r is an integer in which k _r ≧ 0). Here, U is the calculation interval of the evaluation value. In the conventional Nyquist point search method, the evaluation value calculated at such a time is compared with a predetermined value to search for a time when the evaluation value is closer to the extreme value than the predetermined value.

[0015] In FIG. (B), since the evaluation value Q _A at time (t ₁ + 2U) is closer to the extreme than the predetermined value P _A, such time (t ₁ + 2U) is the Nyquist point It has been estimated to be. In addition,
Obtaining an evaluation value that indicates the correlation between the received signal and the known signal as described above is referred to as known signal detection processing in this specification.

[0016]

In such a conventional Nyquist point search method, the error ΔT generated between the estimated Nyquist point and the true Nyquist point reaches the calculation interval U at the maximum. Since the Nyquist point estimation error ΔT increases the reception error rate, it is desired to shorten the calculation interval U as much as possible.

At this time, the received signal for which the evaluation value is to be calculated needs to pass through the Nyquist filter. Therefore, if the calculation interval U is shortened, the calculation interval also becomes shorter in the Nyquist filter. .

Further, if the Nyquist filter is composed of an FIR filter extending over the same convolution section, if the calculation interval is shortened in this way, the delay means increases and the tap calculation also increases.

As described above, in the receiver for digital communication, in order to reduce the reception error rate, the number of taps of the FIR filter is increased, and the calculation interval of the FIR filter is shortened. Known signal detection processing. Are required to be performed at shorter intervals.

However, when the Nyquist filter to which the FIR filter is applied and the known signal detection process are to be realized by using, for example, a DSP (Digital Signal Processor), the calculation amount per unit time required for them is the above-mentioned. As described above, it becomes necessary to increase the number of DSPs used.

As the number of DSPs used increases, so does the mounting space, power consumption, and cost.
Therefore, particularly in a technical field where small size and light weight are required, such as a communication device mounted on a vehicle, it is desired to reduce the calculation amount per unit time.

The technical problem of the present invention is to pay attention to such a problem, and in the digital filter and the Nyquist point search method, the reception error rate can be suppressed,
It is to reduce the amount of calculation required for them per unit time.

[0023]

FIG. 1 (a) shows a principle diagram of a digital filter according to a first aspect of the present invention. The digital filter according to claim 1 is a non-cyclic digital filter having a delay unit that delays a signal to form a tap, and the delay unit before, or after the delay, like the delay units 1, 2, and 3. The delay means is provided so that tap calculation is not performed for taps.

FIG. 1 (b) shows the principle of the Nyquist point search method according to the second aspect. The Nyquist point search method according to claim 2, wherein an evaluation value indicating a correlation between the received signal and the known signal, which has a characteristic of showing an extreme value with respect to a known signal included in the received signal, such as W _S. It is assumed that the calculation method is set.

On such a premise, the Nyquist point search method according to claim 2 obtains the evaluation value with respect to the sampling value of the received signal, and searches for the Nyquist point from the extreme value of the evaluation value, According to the size of the evaluation value, the calculation interval of the evaluation value is Q ₁ ,
Switching is performed such that the evaluation value becomes shorter as the evaluation value is closer to the extreme value like Q ₅ and Q ₆ with respect to Q ₂ , Q ₃ and Q ₄ .

FIG. 1 (c) shows the principle of the Nyquist point search method of the third aspect. According to a Nyquist point search method of a third aspect, in the Nyquist point search method of the second aspect, the sampling value of the received signal is stored for a predetermined time at an interval shorter than the current calculation interval.

[0027] Then, as Q ₁₁ to Q _12, when instructing that the amount of change in the evaluation value had passed the extreme value, from the sampled value stored, Q ₁₂
On the other hand, as in Q _13, the evaluation value is calculated at shorter time intervals by tracing the passage of time in reverse.

[0028]

In the conventional digital filter, the calculation amount per unit time decreases if the calculation interval is lengthened.
However, in such a configuration, the output interval of the filtered signal becomes long, and when it is applied to the Nyquist filter, the execution interval of the known signal detection processing in the subsequent stage is unavoidable. In other words, if the amount of calculation is reduced in that way, the Nyquist point estimation error increases,
The reception error rate also increases.

Therefore, the digital filter according to claim 1 is
The number of calculations per unit time is reduced by thinning out some tap calculations. If some tap calculations are thinned out, even if the output interval of the filtered signal is shortened by shortening the calculation interval of the digital filter, the calculation amount per unit time decreases.

If the output interval of the signals filtered in this way is shortened, it becomes possible to shorten the execution interval of the known signal detection processing in the subsequent stage when applied to the Nyquist filter. That is, the reception error rate can be suppressed by reducing the Nyquist point estimation error.

In FIG. 3A, tap calculation in the tap formed by the output of the delay means 1 and the tap formed by the output of the delay means 2 is thinned out, and the tap formed by the input of the delay means 1, It is shown that tap calculation of taps formed by the output of the delay means 3 and the delay means 3 is performed in the multiplication means 4, and the multiplication results are added in the addition means 5. However, the drawing does not limit the thinning interval of tap calculation in the present invention.

According to the Nyquist point search method of the second aspect, it is estimated whether the reception time is far from the reception time of the known signal, and the execution interval of the known signal detection process in the time zone far from the reception time of the known signal is set. By increasing the length, the amount of calculation per unit time is reduced.

According to the Nyquist point search method of the second aspect, by shortening the execution interval of the known signal detection process in the time zone close to the reception time of the known signal, the estimation error of the Nyquist point is reduced and the reception error rate is reduced. It is possible to suppress.

In FIG. 5B, the evaluation value is compared with a predetermined value Pj, and if the evaluation value is farther from the extreme value than the predetermined value Pj, the evaluation value Q ₁ is calculated every 3T _M , Q ₂ , Q
₃ and Q ₄ are obtained, and if the evaluation value is closer to the extreme value than the predetermined value Pj, the evaluation values Q ₅ , Q ₆ ... Are calculated for each T _M. However, the figure does not limit the execution interval of the known signal detection processing in the present invention.

At this time, depending on the characteristics of the evaluation value and the execution interval of the known signal detection processing in the time zone away from the reception time of the known signal, the time corresponding to the extreme value of the evaluation value is received. There is a possibility that the execution interval of the known signal detection processing in the time zone close to the time may be exceeded.

Therefore, in the Nyquist point search method according to a third aspect of the present invention, the sampling value of the received signal is stored in advance, and a known signal is obtained by reversing the passage of time based on the stored sampling value of the received signal. By performing the detection process, the Nyquist point estimation error is reduced,
It is possible to suppress the reception error rate.

In FIG. 6 (c), if the evaluation value is farther from the extreme value than the predetermined value P _LAST , the evaluation values Q ₁₀ , Q ₁₁ and Q ₁₂ are calculated every 6T _{M, and} thus calculated. The slope of the line segment connecting the evaluation values is the extreme value R ₁ and the extreme value R ₁ to 6T.
If the slope is gentler than the slope of the line segment connecting the evaluation value R _{2 at} the time of returning to _M, the evaluation value Q _{13 at the} time of going up T _M from the stored sampling value of the received signal.
Has been shown to seek. However, the figure does not limit the execution interval of the known signal detection processing in the present invention.

[0038]

EXAMPLES Next, examples of how the digital filter and the Nyquist point search method according to the present invention are actually embodied will be described.

[Outline of Digital Communication System] First, an outline of the digital communication system will be described with reference to FIG. In the figure, 20 is a transmitter and 21 is a receiver.

In the transmitter 20, the transmission data is input to the matched filter 22. The output of the matched filter 22 is converted into an analog signal in the D / A converter 23. The analog signal is modulated and transmitted by the transmitter 24.

At this time, the calculation in the matched filter 22 and the conversion into the analog signal in the D / A converter 23 are performed in synchronization with the clock generated by the clock signal generator 25.

In the receiving device 21, the transmitting device
Twenty transmission signals are received by the receiving unit 26. The receiver
At 26, demodulation of the received signal is performed. The demodulated signal is converted into a digital signal in the A / D converter 27. The digital signal is input to the matched filter 28.

The output of the matched filter 28 is converted into received data in the data demodulation section 29. The output of the matched filter 28 is also input to the known signal detector 30.
The known signal detector 30 detects a known signal from the output of the matched filter 28 and searches for a Nyquist point.

At this time, the data demodulation section 29 demodulates the received data at each code interval with reference to the Nyquist point of the known signal retrieved by the known signal detection section 30. The conversion into a digital signal in the A / D converter 27, the calculation in the matched filter 28, and the demodulation of the received data in the data demodulator 29 are performed in synchronization with the clock generated by the clock signal generator 31. .

[Structure of Embodiment of the Present Invention] Next, the present invention is applied to the matched filter 28 and the known signal detection unit.
An example of the present invention will be described with reference to FIG. The data output from the A / D converter 27 is input to the port 40a of the DSP 40.

The data input to the port 40a is stored in the register 40b. At this time, an interrupt request is output from the port 40a to the arithmetic unit 40c. The data stored in the register 40b is read by the arithmetic unit 40c. In the arithmetic unit 40c, the control described later is executed.

In the DSP 40, the ROM 42, and the RAM 43, data transfer is mutually performed via the bus 41 to which they are connected. The ROM 42 stores the control procedure of the control executed by the DSP 40. The RAM 43 stores data necessary for the control executed by the DSP 40.

[0048] The ROM42 is tap coefficients h ₁ to h _m shown in (b) are also stored. The tap coefficient h ₁ ~h _m
Is used in the matched filter calculation described later.

FIG. 7C shows an example of data stored in the RAM 43. The number of samples, the current index, and the sampling value are stored in this area. The numbers next to the sampling values are the index numbers of those sampling values. Hereinafter, this area will be referred to as a sampling value storage area.

[Control Procedure of Sampling Value Storage Control] Next, the control procedure of the sampling value storage control executed in the DSP 40 will be described with reference to the flowchart shown in FIG. When the interrupt request is output from the port 40a to the arithmetic unit 40c, the control is executed by temporarily suspending the control being executed.

In step H50, the register 40
The data stored in b is read. The data is temporarily stored in the RAM 43. In the following step H51, the current index in the sampling value storage area is read and stored in the variable INX secured in the RAM 43.

The number of samples and the current index in the sampling value storage area are initialized to 0 by separately executed initialization control (not shown).

In the following step H52, the variable I
1 is added to NX. By the addition, the variable INX
When N exceeds N _LAST , the control shifts to step H54 according to the determination in step H53. Otherwise,
The control proceeds to step H55. In step H54, 1 is assigned to the variable INX, and control is performed in step H54.
Move to 55.

In step H55, the variable INX
Is assigned to the current index of the sampling value storage area. In the following step H56, the data read in step H50 is stored in the sampling value of the sampling value storage area corresponding to the index designated by the current index of the sampling value storage area.

In the subsequent step H57, the number of samples in the sampling value storage area is substituted into the variable CNT secured in the RAM 43. In the following step H58, the variable CNT is compared with N _LAST . If the variable CNT is N _LAST or more, the control ends.

On the other hand, if the variable CNT is smaller than N _LAST , the control shifts to step H59. Step H59
In, the value obtained by adding 1 to the variable CNT is stored in the number of samples in the sampling value storage area. Then, the control ends. When the control ends, the execution of the control suspended by the control is restarted.

[Control Procedure of Nyquist Point Search Control (No. 1)] Next, the control procedure of the Nyquist point search control executed in the DSP 40 will be described with reference to the flowchart shown in FIG.
The control is activated by the host control when it is determined that the known signal reception time is approaching due to the synchronization protection performed by the host control (not shown), or when it is in the asynchronous state.

At step H70, a constant K _N is added to the variable N secured in the RAM 43. The variable N is initialized to 0 by the initialization process. Also, addition and subtraction with respect to the variable N, addition and subtraction result is the further subtracting the greater if N _LAST than N _LAST, are calculated as further addition of N _LAST if addition and subtraction result is 0 or less.

In the following step H71, the sampling value of the sampling value storage area indexed by the index N, (N−L), ..., (N− (m−1) L) is read out, and Variables secured in RAM 43 x (0), x (-L), ..., X (-(m-1)
L). Here, m is the number of tap coefficients. L is a constant. The sampling value storage capacity of the sampling value storage area is ensured so as to satisfy N _LAST > (m-1) L.

In the subsequent step H72, the variables x (0), ..., X (-(m-1) L) are read out, and a matched filter operation described later is performed. The calculation result y () of the matched filter calculation is stored in the RAM 43. Regarding the calculation result y (), the latest predetermined number is stored. They are written as y (0), y (-1), ..., Y (q) in ascending order.

In the subsequent step H73, the calculation results y (0), ..., Y (q) are read out, and known signal detection processing (details not shown) is performed. The evaluation value obtained in this step is the variable ER
Memorized in.

In the following step H74, the variable E is
R and the constant P ₁ are compared. The constant P ₁ has the extreme value of the evaluation value as ER ₀ , and if the evaluation value is convex downward, P ₁ > P ₂ > ER ₀ is satisfied, and if the evaluation value is convex upward, It is selected to satisfy P ₁ <P ₂ <ER ₀ . Here, P ₂ is a constant described later. In the figure, it is assumed that the known signal detection process is set so that the evaluation value is convex downward.

If the variable ER is greater than or equal to the constant P ₁ , control returns to step H70. Otherwise,
The control proceeds to step H75. If the evaluation value is convex upward, the magnitude relationship under these branching conditions is reversed.

In step H75, the variable ER is substituted into the variable E _min . Then, the control is step H76.
Move to.

At step H76, the variable N is set to 1
Is added. In the following step H77, the sampling value in the sampling value storage area is read into the variable x (), as in step H71. In the subsequent step H78, a matched filter operation described later is performed, as in step H72.

In the subsequent step H79, the known signal detection processing is performed, and the evaluation value is stored in the variable ER, as in step H73. Then, the control shifts to step H80.

In step H80, the variable ER and the variable E _min are compared. If the variable ER is greater than or equal to the variable E _min , control transfers to step H82. Otherwise, control transfers to step H81.
If the evaluation value is convex upward, the magnitude relationship under these branching conditions is reversed.

In step H81, the variable E _min
The variable ER is substituted into. Then, the control returns to the step H76.

In step H82, the variable E _min
And the constant P ₂ are compared. If the variable E _min is the constant P ₂ or more, the control ends. Otherwise, control transfers to step H83. If the evaluation value is convex upward, the magnitude relationship under these branching conditions is reversed.

At step H83, 1 is subtracted from the variable N and stored in the RAM 43. Then, the control ends.

The subtraction result is passed to the data demodulation process (not shown). In the data demodulation process, the subtraction result is N ′, and the index N ′ + K _SYM N _SYM (K _SYM = 1, 2, 3, _...
The demodulation of the received data is performed from the sampling value of the sampling value storage area indexed by (...).

Here, N _SYM is the number of data input to the port 40a between adjacent codes in time series.
The addition of N ', like the addition of the variable N, it is assumed that the further N _LAST is reduced when in excess of the N _LAST.

[Control Procedure of Nyquist Point Search Control (Part 2)] Next, another example of the control procedure of the Nyquist point search control executed by the DSP 40 will be described with reference to the flowchart shown in FIG. Note that, in the figure, steps given the same reference numerals as the above-mentioned reference numerals are the same as the corresponding steps corresponding to the same reference numerals, and therefore description thereof is omitted.

In this example, if the variable ER is greater than or equal to the constant P ₁ in step H74, the control shifts to step H91. Otherwise, control transfers to step H90.

In step H91, the variable ER is substituted for the variable TMP. Then, the control shifts to the step H70.

In step H90, the variable E _{min is set.}
The variable ER is substituted into. In the following step H92, the result of subtracting the variable E _min from the variable TMP is substituted into the variable A.

In the following step H93, the variable A is
And the constant α are compared. Is the constant number alpha, the evaluation value from said value of said evaluation value in K _N T _SMP only reversed up time from the time when an extreme value ER _0, is set to a value obtained by subtracting the extreme ER _0. Here, T _SMP is the port 40a.
This is the interval at which data is input to.

If the variable A is larger than the constant α, the control shifts to the step H76. Otherwise, control transfers to step H94. If the evaluation value is convex upward, the magnitude relationship under these branching conditions is reversed.

At step H94, 1 is subtracted from the variable N. In the following step H95, the sampling value in the sampling value storage area is read into the variable x () as in step H71. In the following step H96, a matched filter operation described later is performed, as in step H72.

In the subsequent step H97, the known signal detection processing is performed, and the evaluation value is stored in the variable ER, as in step H73. Then, the control shifts to step H98.

In step H98, the variable ER is compared with the variable E _min . If the variable ER is equal to or more than the variable E _min , the control shifts to step H100. Otherwise, control transfers to step H99.
If the evaluation value is convex upward, the magnitude relationship under these branching conditions is reversed.

In step H99, the variable E _min
The variable ER is substituted into. Then, the control returns to the step H94.

In step H100, the variable E
_min is compared with the constant P ₂ . If the variable E _min is the constant P ₂ or more, the control ends. Otherwise, control transfers to step H101. If the evaluation value is convex upward, the magnitude relationship under these branching conditions is reversed.

At step H101, 1 is added to the variable N and stored in the RAM 43. Then, the control ends.

The addition result is passed to the data demodulation process. In the data demodulation process, the addition result is set as N ′, and the index N ′ + K _SYM N _SYM (K _SYM = 1, 2, 3, _...
The demodulation of the received data is performed from the sampling value of the sampling value storage area indexed by (...).

[Matched Filter Operation] Next,
The matched filter calculation will be described based on the data flow diagram shown in FIG.

The variables x (0), x (-L), ..., X (-(m-1)) stored in the RAM 43 as described above.
L) is multiplied by the tap coefficients h ₁ , h ₂ , ... H _m , respectively. In the figure, the multiplication means 110
In, it is shown that such multiplication is performed.

Then, all the multiplication results are added.
The result of this addition is y (0). In the figure, it is shown that such addition is performed in the addition means 111.

[Description of Operation] Next, the operation of the Nyquist point search control will be described based on the graph shown in FIG. In FIG. 5A, the vertical axis represents the variable ER and the horizontal axis represents the variable N. Also,
(A) shows the evaluation value. In FIG. 3A, the evaluation value is assumed to be convex downward. (A) is the extreme value of the evaluation value. Further, in the same figure (a), the constant K _N =
It is assumed to be 4.

Now, when N = 10, the above step H
70, H71, H72, H73 are executed and the variable ER is changed to Q.
_{Assume 100} . The value Q ₁₀₀ is the constant P ₁
If it is larger than the above, the steps H70, H71, H
72 and H73 are executed. At this time, the variable ER is Q
_It is assumed to be ₁₀₁ .

If the value Q ₁₀₁ is smaller than the constant P ₁ , then this step H75, H76, H77, H78,
H79 is executed. At this time, it is assumed that the variable ER becomes Q ₁₀₂ . If the value Q ₁₀₂ is smaller than the Q ₁₀₁ , the steps H75, H76, H77, H78 and H79 are executed again through the step H81. At this time, it is assumed that the variable ER becomes Q ₁₀₃ .

The value Q ₁₀₃ is larger than the Q ₁₀₂ and the Q ₁₀₂ is the constant P _{2 as} shown in (c) of FIG.
If it is smaller than this, this step H80, H82,
H83 is executed. At this time, N '= 19 is stored.

In FIG. 9B, the vertical axis represents the variable ER and the horizontal axis represents the variable N. Further, (A) shows the evaluation value. It is assumed that the evaluation value is convex downward in FIG. (A) is the extreme value of the evaluation value. Further, in the same figure (b), it is assumed that the constant K _N = 4.

Now, when N = 26, the above step H
70, H71, H72, H73 are executed and the variable ER is changed to Q.
_It is assumed to be ₁₁₁ . The value Q _{111 corresponds} to the constant P ₁
If it is larger than the above, the steps H70, H71, H72 and H73 are executed again through the step H91. At this time, it is assumed that the variable ER becomes Q ₁₁₂ .

The value Q ₁₁₂ is smaller than the constant P ₁ and
If (Q ₁₁₂ -Q ₁₁₁ ) is less than or equal to the constant α, then the steps H94, H95, H96, and H97 are executed. At this time, it is assumed that the variable ER becomes Q ₁₁₃ . If the value Q ₁₁₃ is smaller than the Q ₁₁₂ , the steps H99, H95, H
96 and H97 are executed. At this time, the variable ER is Q
_It is assumed to be ₁₁₄ .

The value Q ₁₁₄ is larger than the Q ₁₁₃ and the Q ₁₁₃ is the constant P _{2 as} shown in (c) of FIG.
If it is smaller than the above, this step H98, H100
, H101 is executed. At this time, N '= 33 is stored.

[0097]

As described above, the digital filter according to the present invention is configured so as to thin out a part of tap calculations, so that the output interval of the filtered signal is kept short unlike the conventional one. Moreover, it is possible to reduce the amount of calculation per unit time while keeping the convolution section long.

Therefore, by applying the digital filter according to claim 1 to a Nyquist filter, it is possible to obtain a reception signal that has passed through the Nyquist filter at short intervals while reducing the amount of calculation per unit time. It has become possible to suppress the reception error rate by reducing the Nyquist point estimation error.

In the digital filter according to the first aspect, since the amount of calculation per unit time is reduced in this way, for example, the number of DSPs used is reduced, and the size, weight and cost of the device can be reduced. .

According to the Nyquist point search method of the second aspect, as described above, the execution interval of the known signal detection process is lengthened in the time zone in which the reception time is away from the reception time of the known signal, which is different from the conventional method. As a result, the amount of calculation per unit time has decreased.

In the Nyquist point search method of the second aspect, since the amount of calculation per unit time is reduced in this way, for example, the number of DSPs used is reduced, and the size, weight and cost of the device can be reduced. became.

In the Nyquist point search method according to the third aspect, the sample value of the received signal is stored as described above, and when the received time has passed the received time of the known signal, it is stored. Since the known signal detection processing is performed by moving up the time based on the sample value of the received signal, it is possible to reduce the estimation error of the Nyquist point and suppress the reception error rate.

[Brief description of drawings]

FIG. 1 is a block diagram showing a basic principle of the present invention and a time chart.

FIG. 2 is a block diagram showing an outline of a digital communication system.

FIG. 3 is a block diagram showing an embodiment of the present invention and a data configuration diagram.

FIG. 4 is a flowchart showing a control procedure of sampling value storage control.

FIG. 5 is a flowchart showing a control procedure of Nyquist point search control.

FIG. 6 is a flowchart showing another example of the control procedure of Nyquist point search control.

FIG. 7 is a data flow diagram showing a calculation method of matched filter calculation.

FIG. 8 is a graph showing an operation of the example.

FIG. 9 is a signal flow diagram for explaining the configuration of a conventional FIR filter and a time chart for explaining a conventional Nyquist point search method.

[Explanation of symbols]

 1, 2 and 3 delay means 4 multiplication means 5 addition means 27 A / D conversion section 40 DSP 40a port 40b register 40c arithmetic section 41 bus 42 ROM 43 RAM

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location // H04B 3/23 4229-5K

Claims (3)

[Claims] 1. A non-recursive digital filter having delay means for delaying a signal to form a tap, wherein delay means for performing tap calculation at a tap before or after a delay is provided. And a digital filter. 2. An evaluation value indicating the correlation between the known signal and the received signal whose calculation method is set so as to have a characteristic of showing an extreme value with respect to the known signal included in the received signal. For the sampling value of, when searching for the Nyquist point from the extreme value of the evaluation value, according to the magnitude of the evaluation value, the calculation interval of the evaluation value, the evaluation value to the extreme value. A Nyquist point search method characterized by switching so that it becomes shorter as it gets closer. 3. The Nyquist point search method according to claim 2, wherein the sampling value of the received signal is stored for a predetermined time at an interval shorter than the current calculation interval, and the amount of change in the evaluation value is stored. When instructing that the extreme value has passed, it is characterized in that the evaluation value is calculated at shorter time intervals by tracing back the elapsed time from the stored sampling value. Nyquist point search method.