JPS5811586B2 - digital beam form - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a digital beamformer for performing phased array beamforming such as used in sonar, particularly when using geometrically symmetrical sensor arrays such as cylindrical play and linear arrays. Concerning effective digital beamformers.

Traditionally, beamformers for sonar have been of an analog type using a tapped analog delay line and a summing amplifier, or of a digital type using a shift register and an adder. Waveform distortion occurs due to analog delay lines, miniaturization is difficult, and the latter requires a large number of output taps of the shift register, so LSI-based shift registers are not used, and a large number of ICs are required. , which had drawbacks such as increased power consumption.

In addition, Japanese Patent Publication No. 43-14911 describes a method for realizing a similar function using an electronic computer, but it uses random access memory (hereinafter abbreviated as RAN) used in ordinary electronic computers as a delay element. For this purpose, the delay time is specified by the memory address, but each time new data is input, the relationship between the delay time and the memory address must be changed, and this change takes a considerable amount of time, which increases processing speed. No.

Also, in order to avoid this address change, it is possible to create a program in advance for all combinations of addresses that need to be changed, but this has disadvantages such as the amount of programming becomes enormous and it is not economical. There is.

In order to solve these drawbacks, the present invention uses a RAM as a delay element, a counter that advances at a cycle equal to the sampling cycle of the input signal, and a read-on memory (hereinafter referred to as ROM) that stores the relationship between the tap number and the sensor number. A desired beam is formed by controlling the RAM and changing the delay time using the address calculated from the output of It can be applied to various soaps with different sensor array sizes and shapes by simply changing the program.
In addition, the present invention realizes a beamformer that can save memory capacity by utilizing the geometric symmetry of the sensor array.

FIG. 1 is a block diagram showing the functions of a beamformer that the present invention is trying to realize, and 1-1.1-2.・
..., 1- has input terminals corresponding to 2 sensors, 2-
1.2-2. ..., 2- is a tapped delay line, 3-1
．． 3-2. ..., 3- is an attenuator, 4 is an adder, and 5 is an output terminal.

Here, a system consisting of an input terminal 1-1, a tapped delay line 2-1, an attenuator 3-1, an adder 4, and an output terminal 5 constitutes a transversal filter for one sensor, and the beam The former can be considered to be the sum of these transversal filters.

If it is necessary to form a large number of beams with different orientations, a large number of the circuits shown in FIG. 1 may be used in parallel, but in this case, the tapped delay lines 2-1, 2-2. ..., 2
− can be shared.

Note that normally, the number of attenuators connected to one tapped delay line is often one per beam output, and there are also cases where all the attenuators have the same gain (no shading).

FIG. 2 is a diagram showing an embodiment of the present invention, and 1-1.1
-2. ..., 1- is an input terminal, 5 is an output terminal, 6 is a multiplexer, 7 is an A/D converter, 8 is an input buffer, 9 is a RAM, 10 is a multiplier, 11 is an adder, 12 is an register , 13 is an output buffer, 14 is a RAM address generator, 15 is an N-ary counter (N is the number of taps in each delay line)
is the zero address counter, 16 is the base counter (K
is the number of sensors), 17 is the table memory, 18 is the table memory counter, 19 is the start address table, 2
0 is a program memory, and 21 is a program memory counter.

Here, input terminals 1-1.1-2. .

On the other hand, the program memory 20 is divided into a 02 section and an LIT section, and as shown in FIG. 3, the 02 section includes an input command W, a beam forming command BF, beam numbers BO, B1 .・
. . , a return command RTN is written, and 0. 1. . . . is written, its address is specified by the program memory counter 21, and its contents are sequentially read out.

The 02 part is sent to the start address table 19,
It specifies the starting address of the table memory 17 where the microprogram and table corresponding to each instruction are stored, and loads it into the table memory counter 18.

Further, the LIT portion is loaded into the binary counter 16.

The table memory 17 is normally a read-only memory (hereinafter abbreviated as ROM), which reads the input buffer 8, writes to the RAM, loads and resets the register 12, writes to the output buffer 13, and clocks the zero address counter 15. , a 0 part that stores microinstructions that control the code and clock of the table memory counter 18, a clock and clear of the program memory counter 21, a 0 part that represents the tap number of the transversal filter, and a 0 part that represents the sensor number; It is divided into a Δ section that stores a clock control signal, and a W section that stores an attenuator weight.

The word in the table memory corresponding to the input command is 0.
All zeros are written to the section Δ, a load signal is written to the first address of the section, a clock signal is written to the other numbers, input buffer read and RAM write signals are written to all addresses of the section c, and this instruction is written from the program memory 20. When called, the constant in Figure 3 is set to 0.
The starting address of the microprogram corresponding to this instruction is loaded from the start-aton staple 19 to the table memory counter 18, and then the decimal counter 16 and the table memory counter 18 are sequentially updated to the 11th stage. only step forward.

At this time, the RAM address generator 14 generates an address for the RAM 9 from the output of the zero address counter 15, the output of the 0 part of the table memory 17, and the output of the advance counter 16, and uses the signal of the c part of the table memory 17 to generate an address for the RAM 9. The signal stored in the input buffer 8 is written to this address in the RAM 9.

Table 1 below shows examples of instructions written to the table memory 17.

Here, the number of sensors was set to 16.

#0 to #15 of table memory 17 (0th address to 15th address)
microinstruction and data corresponding to the input instruction W at address), microinstruction and data corresponding to the beamforming instruction BF and Bo at #16 to #31, beamforming instruction BF to #32 to #47, and beamforming instruction BF to B1. Corresponding microinstructions and data are written, and so on.

The microinstruction in section C reads the input buffer 8 and writes it to the RAM 9 in response to the input instruction W (INP
UT), for the beamforming command BF, read the RAM 9, multiply it by the output of the W part of the table memory 17 by the multiplier 10, add it to the contents of the register 12 by the adder 11 (SUM), and in the last step This means sending the added result to the output buffer 13 and clearing the register 12 (OUTPUT).

Further, CK in part C is a clock signal that increments the zero address counter 15, and is used only at the beginning of the input command W.

CK and L in the delta section are the clock and load signals for the counter 16.

The RAM address generator 14 has the configuration shown in FIG. 4a or 4b.

Here, 14-1 is an input terminal from the zero address counter 15, 14-2 is an input terminal from the output of the n section of the table memory 17, 14-3 is an input terminal from the output of the binary counter, and 22 is the mod N 23 and 24 are multipliers for ×N and ×, respectively; 25 is an adder; 14-4 is an output terminal; RA in which a signal corresponding to the output signal of the nth tap (n is specified by the n section of the table memory 17) of the tapped delay line for the tapped delay line (specified by the n section of the table memory 17) is stored.
Calculate and output the address of M9.

Using FIG. 4a, the N words of signals input from one sensor are stored in a portion of RAM 9 with successive addresses, and using FIG. 4b, they have every other word of address in RAM 9. In either case, the RAM 9 requires a capacity of N words.

In other words, using the example in Figure 4a, if we consider a sensor with a sensor number of 0, its N signals will be 0 to 0.
If the signal is stored at the address N-1 and the count value of the zero address counter 15 is i, this address becomes the latest signal, so the relationship between the address in the RAM 9 and the tap numbers 0 to N in FIG. 1 is as follows. The results are as shown in Table 2 below.

Here, when N or K is an insole of 2, the operation of multiplying by N or 2 is realized by a simple shift, and the adder 25 is not required.

It is a good idea to use Figure 4a when N is in the middle of 2, and Figure 4b when K is in the middle of 2, but in the former case, the sensor number is used as the upper bit to specify the address category, and the zero address counter is used. This is a case where the address order within the address category is specified by using the value obtained by changing the value of 15 by the tap number as the lower bit.The latter is a case where the relationship between the upper bit and lower bit is simply reversed, and both are They perform the same function.

In the case of a beam forming command, the contents of the LIT section of the program memory 20 are loaded into the incanic counter 16 as before, and the start address of the microprogram corresponding to this command is loaded from the start address table 19 into the table memory counter 18. Then, the table memory counter 18 increments sequentially, and the K-ary counter 16 increments to 1 by the clock control signal written in the Δ section of the table memory 17.
Step by step or hold to sequentially generate sensor numbers needed to execute this command.

For example, in the beam forming command BP for beam number B0, 1 is generated in the sensor number corresponding to address 16 of the table memory 17, and address 17.18°...
- Corresponding to 31, 2. is sequentially added to the sensor number. 3. ...to 15. 0 is generated.

The weight of the tap number for the sensor number generated by this forward counter is written in the n part and W part of the table memory 17, respectively, and the register 12 is reset to the last address of this microprogram in the c part. The write signal of the output buffer 13 and the load signal of the register 12 are written to other addresses, and the sensor number k, tap number n, and weight W, which are generated sequentially as the table memory counter 18 advances, are written to other addresses. First, use RA
The M address generator 14 sets the tap number M to this sensor number.
The corresponding R from n and the output of the zero address counter 15
The address of AM9 is generated, the signal stored at this address is read out at the output end of RAM9, the product of this signal and the weight W is calculated by the multiplier 10, and the product is calculated by the adder 11 and the register 12. The output signals corresponding to the designated beams are cumulatively calculated and temporarily sent to the output buffer 13.

This output signal is read out from the output buffer 13 to the output terminal 5 as required.

Note that in part c of the table memory 17, the clock signal of the zero address counter 15 is input to the last address of the input instruction, and the clock signal of the program memory counter 2 is input to the last address of each instruction.
1 clock and the load signal of the table memory counter 18 are written.

In this way, each instruction written to the program memory 20 is executed in the order in which it is written, and several beam outputs are calculated sequentially.

A return instruction is added to the last memory written in the program memory 20, and when this instruction is executed,
A clear signal for the program memory counter 21 is generated,
Return to the first instruction in program memory 20.

By repeating such operations, it is possible to simultaneously form a large number of beams while inputting periodically sampled signals.

Now, the system shown in Figure 2 can be used for general sensor arrays with complex geometries by writing separate microprograms that correspond to beam forming commands for beams in different directions. , in the case of a cylindrical array, its geometrical symmetry can be used to save the capacity of the table memory 17.

Now, assuming that K sensors are arranged at equal intervals on the circumference, in the beam forming calculation, the tap number n and the weight W are not changed, but the sensor number k is relatively shifted by one. This means that the beam direction has been rotated by 360/K (degrees).

This operation can be easily achieved by simply increasing or decreasing the LIT part of the beamforming command by 1, so in order to make the difference in orientation between adjacent beams 360/×L (degrees), L different beamformers are required. All that is required is an instruction, and the capacity of the table memory 17 can be reduced to approximately 1/K compared to the case where it is used as a general array.

In other words, a beam forming command in which the beam direction is simply rotated by an integral multiple of 360/K (degrees) can be realized by simply rewriting the sensor number. For example, the beam forming command BF for beam numbers B0 and B3 in Fig. 3 is Simply 360/K (
If the rotation corresponds to an object rotated by 3 times the angle (degrees), this can be realized by simply replacing the sensor number 1 with the sensor number 4, and replacing the other values accordingly.

That is, in the beam forming command for beam number B0, addresses 16 to 31 of the table memory 17 are used to set the sensor number to 1. 2. ..., 15°K0 was generated sequentially, but for beam number B3, 4 was sent to the sensor number from the program memory 20 to the binary counter 16,
Then, in the table memory 17, specify addresses 16 to 31, which are exactly the same as beam number B0, and sequentially perform 4. to 5
．． ....

K15. 0. 1. 2. By outputting the sensor number 3, the beam forming command BF is executed.

Also, linear array, oval array, etc. have 1 or 2
Even in the case of an array having two planes of symmetry, the forward counter 16 in FIG. 2 can be reversibly changed to a forward counter as shown in FIG. By performing this in part of the LIT section of the formation command, the capacity of the table memory can be saved.

In the case of a linear array, a beam with an orientation between 90 degrees from the broadside direction to the end-fire direction is created in the same manner as in Figure 2, and at this time, the U/D of the KC counter is set to U (forward). , and set the other parts of the LIT part of the instruction to zero.

For a beam in the opposite 90 degree range, you only need to set the LIT part of the same command, the U/D part to D (retreat), and the other parts to -1, which reduces the table memory capacity by almost half. can do.

In addition, in the case of an array with two planes of symmetry such as an elliptical array, only the beam in the direction of the first quadrant is created in the same manner as in Figure 2, and the beam in the third quadrant is created in the direction of the first quadrant. The beam is obtained by shifting the LIT part of its beamforming command by 180 degrees, the beam in the second quadrant is obtained by folding the beam in the first quadrant (corresponding to changing U/D to D), and the beam in the second quadrant is The beam in the 4th quadrant is 18 times the beam in the 2nd quadrant.
Obtained by rotating 0 degrees.

In this way, the capacity of the table memory can be reduced to approximately 1/4.

Here, the sampling frequency fs for the output signal of each sensor
In order to suppress the gain loss due to the roughness of the tap spacing,
Usually, it is selected to be 5 to 6 times the optimal frequency of the input signal, but
Since the output of the beamformer only needs to satisfy the usual sampling theorem, the samples of the output signal of the beamformer can be thinned out to 1/2 or less depending on the signal bandwidth.

Therefore, when thinning out to 1/2, for example, each beam forming process only needs to be performed once for two input commands, as in the program example shown in FIG.

One input command inputs one sample of each sensor's output signal, so if the average time interval at which input commands are executed is shorter than the sampling period 1/fs, the sampling frequency f
Beamforming processing of the sensor output signal sampled at s is possible.

Therefore, if high-speed components are used, a large number of beams can be formed with one beamformer.

Note that in the case of a linear array, the weight W may remain unchanged for each sensor even if the beam direction changes, but in this case, only the W part of the table memory is provided independently;
It is also possible to further save the table memory capacity by taking the address from the output of the binary counter 16.

As described above, since RAM is used as the tapped delay line, connection lines corresponding to each tap output can be omitted, and LSI, R
By using AM, the number of ICs can also be significantly reduced.

Further, by rewriting the contents of the program memory and table memory, it can be used for sensor arrays of different shapes, and it has advantages such as saving memory capacity for arrays with geometric symmetry.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the functions of a beamformer realized by the present invention, FIG. 2 is a block diagram showing an embodiment of the present invention, and FIG. 3 is an example of a program written in the program memory in FIG. Figures 4a and 4b
This figure is a block diagram showing the contents of the RAM address generator of FIG. 2, and FIG. 5 is a diagram showing another example of the binary counter in FIG. 2. DESCRIPTION OF SYMBOLS 1... Input terminal, 5... Output terminal, 6... Multiplexer, 7... A/D converter, 8... Input buffer, 9... RAM, 10... Multiplier, 11... Adder, 12... Register, 13... Output buffer,
14... RAM address generator, 15... Zero address counter, 16... K-ary counter, 17... Table memory, 18... Table memory counter, 1
9... Start address table, 20... Program memory, 21... Program memory counter, 2
2... Subtractor, 23... Multiplier, 24... Multiplier, 25... Adder.

Claims (1)

[Claims] 1.NX where the number of sensors is N and the number of taps is N.
A random access memory 9 capable of storing K words of sensor output, and a microprogram corresponding to an input command written using a number of addresses, each address having a specific tap number and sensor number increment. is stored, and a table memory 17 in which a microprogram corresponding to a beam forming command is written using a plurality of addresses, and each address stores a tap number and an increment of a sensor number; 02 regarding formation instructions
The microprogram corresponding to the input command or the beam forming command is sequentially read from the table memory 17 for each address. a first counter 15 that is incremented in synchronization with the sampling period of the sensor according to an input command of the program memory; A second counter 16 is loaded when the first address of the sensor number is read out, and is incremented according to the increment of the sensor number each time a subsequent address is read out. An address generator that specifies each N word corresponding to each successive address, or specifies each word of the same tap number so as to correspond to each successive address, and uses the value of the first counter 15 as a reference. and a device 14 that generates an address code based on a value changed by the tap number read from the table memory 17 and a value read from the second counter 16, and a plurality of beam directions are common in the table memory 17. A digital beamformer characterized in that it is formed by a microprogram and a different leading sensor number in the program memory.