JP2011015159A - Correlation calculation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the number of product-sum computing units and downsize each product-sum computing unit by outputting a correlation value by repeating product-sum calculation using N×N product-sum computing units.SOLUTION: A correlation calculation device includes N×N product-sum computing units (E00-E11) for repeating a process of outputting a cumulative addition value by cumulatively adding respective combinations each comprising any one of N input signals (N is an integer ≥2) in input signals and any one of N replica signals in replica signals by product-sum calculation, and outputting, in the next cycle, a cumulative addition value by cumulatively adding respective combinations each comprising any one of the next N input signals in the input signals and any one of the next N replica signals in the replica signals by product-sum calculation; and an adder for outputting a correlation value of the input signal and the replica signal for each shift amount of the input signals with respect to the replica signals by adding the cumulative addition values in the same cycle or a different cycle which are output by the product-sum computing units in which the amounts for shifting the input signals from the replica signals are equal to one another.

Description

  The present invention relates to a correlation calculation device.

  Correlation processing is performed as processing for detecting the start position of a received signal in digital communication. On the transmission side, a predetermined signal sequence is embedded in a specific portion of the transmission signal. On the receiving side, by performing correlation processing between the received signal and the signal sequence, the position of the signal sequence can be specified and the start position of the received signal can be known.

  FIG. 1 is a diagram illustrating timing detection of a received signal. In a process called cell search in the wireless communication process, a signal start position is detected by searching for a signal string (replica) known to both transmitting and receiving apparatuses. The received signal (input signal) changes due to noise on the propagation path. Therefore, a known signal sequence (replica) cannot be detected by simple match comparison. In general, correlation calculation is performed while shifting the position between the input signal and the replica, and it is determined that there is a known signal sequence at timing t1 of the position with the highest correlation value.

Correlation value P (t) is obtained by complex multiplication of input signal D (u + t) and replica R (u) as in the following equation.
P (t) = ΣD (u + t) · R (u)

  Since the input signal and the replica are complex signals, the correlation value increases as the complex multiplication value of each vector of the input signal and the replica increases. The position of t which becomes the maximum among the correlation values becomes the position of a known signal sequence (replica).

  FIG. 2 is a schematic diagram of the timing detection apparatus. The timing detection apparatus includes a memory 202, a correlation calculator 206, and a maximum value selector 207. The memory 202 receives the input signal 201 and stores the input signal 203 and the replica 204. The correlation calculator 206 performs a correlation calculation of the input signal 203 and the replica 204 in the memory 202 and records a correlation calculation result (correlation value) 205 in the memory 202. The maximum value selector 207 selects and outputs the maximum value timing information among the correlation calculation results 205 in the memory 202.

  FIG. 4 is a diagram illustrating a configuration example of a correlation calculator (matched filter), and FIG. 3 is a diagram illustrating an operation of the matched filter in FIG. The matched filter includes a flip-flop 401 for storing an input signal, a flip-flop 402 for storing a replica, a multiplier 403, and an adder 404. First, a replica is set in the flip-flop 402. Next, the input signal is input to the pipeline of the flip-flop 401 one by one. Correlation between the input signal and the replica can be performed at different timings for each cycle. For example, if the replica has a length of 256, 256 flip-flops 402 and 256 multipliers 403 are required, resulting in a very large circuit. This matched filter has a high speed but has a problem of a large circuit area.

  FIG. 6 is a diagram illustrating a configuration example of the divided matched filter, and FIG. 5 is a diagram illustrating an operation of the divided matched filter of FIG. The divided matched filter includes a flip-flop 601 for storing an input signal, a flip-flop 602 for storing a replica, a multiplier 603, and an adder 604. In order to reduce the area at the expense of speed to some extent, the divided matched filter divides the length of the matched filter and executes the correlation operation a plurality of times. In this method, since the replica is divided to perform the correlation operation, it is necessary to flow the input signal through the pipeline of the flip-flop 601 by the divided number.

  Also, a spread signal generator having a phase shift function, a multiplier for multiplying the output of the spread signal generator and the input signal, and a correlation result for a predetermined number of times for performing in-phase addition over a plurality of signals. There is known a CDMA cell search circuit including a ring buffer for storing the signal and an adder (see, for example, Japanese Patent Application Laid-Open No. 10-164012).

  In addition, it has a searcher that uses a matched filter that captures synchronization and captures synchronization from the upper one or a plurality of minority bits of the input signal, and a sliding correlator that demodulates the input signal by despreading when synchronization is captured. A despreading circuit characterized by this is known (for example, see Japanese Patent Application Laid-Open No. 11-298375).

  In addition, one of the received data caused by each of the plurality of code-multiplexed radio frequency signals and one of the same copy data as one of the plurality of received data are selected and the selected data is held and output. There is known a CDMA system receiving apparatus having a plurality of unit matched filters that multiply the selected data and a replica code and add and output the multiplication result (see, for example, Japanese Patent Application Laid-Open No. 2002-204184).

  There is also known an initial synchronization method in an asynchronous cellular system between DS-CDMA base stations that uses a spread code sequence consisting of a long code unique to each cell and a short code corresponding to each communication channel (for example, Japanese Patent Laid-Open No. 10-101). -126380).

JP-A-10-164012 JP 11-298375 A JP 2002-204184 A JP-A-10-126380

  When a correlation calculator is realized by reducing the circuit scale as in the case of a divided matched filter, a memory for storing intermediate data is required. For example, when the matched filter is divided, it is necessary to flow N input signals by the divided number. Since it is necessary to integrate the divided data, 2N intermediate data memories are required together with at least the N integrated values up to the previous time and the current N correlation values.

  An object of the present invention is to provide a small correlation calculation device.

  The correlation calculation device calculates a combination of any one of N (N is an integer of 2 or more) input signals in the input signal and any one of N replica signals in the replica signal. Cumulative addition is performed by product-sum operation, and the accumulated addition value is output. In the next cycle, any one of the next N input signals in the input signal and the next N replicas in the replica signal are output. N × N product-sum calculators that repeat the process of accumulating and adding a combination of any one of the signals with a product-sum operation and outputting the accumulated addition value, and the input signal for the replica signal By adding the cumulative addition values in the same cycle or different cycles output by the product-sum calculators with the same amount of shift, the input signal for each shift amount of the input signal with respect to the replica signal and the previous And having an adder for outputting a correlation value of the replica signal.

  Since the correlation value can be output by repeating the product-sum operation using N × N product-sum operation units, the number of product-sum operation units can be reduced and the size can be reduced. In addition, since the product-sum operation unit performs cumulative addition, the size of the memory for storing the intermediate data can be reduced.

It is a figure which shows the timing detection of a received signal. It is the schematic of a timing detection apparatus. It is a figure which shows operation | movement of the matched filter of FIG. It is a figure which shows the structural example of a correlation calculator (matched filter). It is a figure which shows operation | movement of the division | segmentation type | mold matched filter of FIG. It is a figure which shows the structural example of a division | segmentation type | mold matched filter. It is a figure which shows the structural example of the product-sum operation unit group in the correlation calculating device by the 1st Embodiment of this invention. It is a figure which shows the structural example of the product-sum operation unit group in the correlation calculating device by the 2nd Embodiment of this invention. It is a figure which shows the product-sum operation unit group which rearranged the product-sum operation unit group of FIG. 7 in the two-dimensional matrix. It is a figure which shows the product-sum operation unit group which rearranged the product-sum operation unit group of FIG. 8 in the two-dimensional matrix. FIG. 10 is a diagram illustrating a correlation calculation method using the product-sum calculation unit group in FIG. 9. It is a figure which shows the product-sum calculation method by the product-sum operation unit group of FIG. It is a figure which shows the process for every cycle of the product-sum operation unit group of FIG. FIG. 10 is a diagram illustrating a correlation calculation method using the product-sum calculation unit group in FIG. 9. It is a figure which shows the correlation calculation method using the product-sum operation unit group of FIG. It is a figure which shows the product-sum operation method by the product-sum operation unit group of FIG. It is a figure which shows the process for every cycle of the product-sum operation unit group of FIG. It is a figure which shows the correlation calculation method using the product-sum operation unit group of FIG. It is a flowchart which shows the correlation calculation method of a correlation calculation apparatus. It is a figure which shows the example of an input signal and a replica. It is a figure which shows the structural example of a product-sum operation unit group. It is a figure which shows an intermediate data memory. It is a flowchart which shows the correlation calculation method of a correlation calculation apparatus. It is a figure which shows the structural example of a product-sum operation unit group. It is a figure which shows an intermediate data memory. It is a figure which shows the structural example of a product-sum calculator. It is a figure which shows the structural example of a correlation calculating apparatus.

(First embodiment)
FIG. 7 is a diagram illustrating a configuration example of a product-sum operation unit group in the correlation operation device according to the first embodiment of the present invention. The product-sum operation unit group includes N × N (for example, four) product-sum operation units (MAC) E00, E01, E10, and E11. Here, N is an integer of 2 or more. The input signal (received signal) d0 is the first complex input signal, and the input signal d1 is the second complex input signal. The replica (signal) r0 is the first complex replica, and the replica r1 is the second complex replica. The product-sum operation unit E00 performs a product-sum operation on the input signal d0 and the replica r0 and outputs the operation result. The product-sum operation unit E01 performs a product-sum operation on the input signal d0 and the replica r1, and outputs the operation result. The product-sum operation unit E10 performs a product-sum operation on the input signal d1 and the replica r1 and outputs the operation result. The product-sum operation unit E11 performs a product-sum operation on the input signal d1 and the replica r0 and outputs the operation result. The same input signal d0 is input to the product-sum calculators E00 and E01, and the same input signal d1 is input to the product-sum calculators E10 and E11. The same replica r0 is input to the product-sum calculators E00 and E11, and the same replica r1 is input to the product-sum calculators E01 and E10.

  FIG. 9 is a diagram illustrating a product-sum operation unit group in which the product-sum operation unit group in FIG. 7 is rearranged into a two-dimensional matrix, and has the same configuration as the product-sum operation unit group in FIG. When the product-sum calculators E00 to E11 are arranged on a square of a two-dimensional matrix, the wiring relationship between the input signals d0 and d1 and the replicas r0 and r1 is relatively easy to understand.

  FIG. 11 is a diagram showing a correlation calculation method using the product-sum calculator group shown in FIG. The correlation calculation device performs correlation calculation between N input signals and K replicas. At the timing of “0”, the amount of replica shift with respect to the input signal is set to 0, the K product-sum operation values are obtained by the product-sum operation of the input signal and the replica, and the K product-sum operation values are added. To obtain the correlation value. At the timing “−1”, the amount of replica shift with respect to the input signal is set to −1, and K product-sum operation values are obtained by product-sum operation of the input signal and the replica. At the timing of “+1”, the amount of replica shift with respect to the input signal is set to +1, K product-sum operation values are obtained by product-sum operation of the input signal and replica, and K product-sum operation values are obtained. To obtain a correlation value. As described above, the correlation between the input signal and the replica is calculated while shifting the timing of the input signal and the replica. The correlation value at each timing is obtained, and the timing at which the maximum correlation value is obtained is the position where the replica in the input signal exists, and the head position of the input signal can be recognized.

  In the wireless communication system of WCDMA (Wideband Code Division Multiple Access), the number N of input signal data is 640, and the number of replica data K is 256. In the LTE (Long Term Evolution) wireless communication system, the number N of input signal data is 2048 × 90, and the number of replica data K is 2048.

  FIG. 12 is a diagram showing a product-sum operation method by the product-sum operation unit group of FIG. Four product-sum calculators A1 to A4 correspond to the four product-sum calculators E00 to E11 in FIG. The product-sum operation unit A1 performs the product-sum operation of the input signal d0 and the replica r0 corresponding to the operation of the position of the replica r0 at the timing “0” in FIG. The product-sum operation unit A2 performs the product-sum operation of the input signal d1 and the replica r1 in response to the operation of the position of the replica r1 at the timing “0” in FIG. The product-sum operation unit A3 performs the product-sum operation of the input signal d0 and the replica r1 in response to the operation of the position of the replica r1 at the timing “−1” in FIG. The product-sum operation unit A4 performs the product-sum operation of the input signal d1 and the replica r0 in response to the operation of the position of the replica r0 at the timing “+1” in FIG. The four product-sum calculators A1 to A4 forming the parallelogram correspond to those extracted from the timings “−1”, “0”, and “+1” in FIG.

  FIG. 13 is a diagram illustrating processing for each cycle of the product-sum operation unit group in FIG. 9. In the first cycle, as in FIG. 12, the product-sum operation unit A1 performs a complex multiplication of the input signal d0 and the replica r0, and the product-sum operation unit A2 performs a complex multiplication of the input signal d1 and the replica r1, and performs a product-sum operation. The unit A3 performs complex multiplication of the input signal d0 and the replica r1, and the product-sum operation unit A4 performs complex multiplication of the input signal d1 and the replica r0.

  Next, in the second cycle, the product-sum operation unit A1 performs complex multiplication of the input signal d2 and the replica r2, the product-sum operation unit A2 performs complex multiplication of the input signal d3 and the replica r3, and the product-sum operation unit A3. Performs complex multiplication of the input signal d2 and the replica r3, and the product-sum operation unit A4 performs complex multiplication of the input signal d3 and the replica r2. Then, the product-sum arithmetic units A1 to A4 respectively perform cumulative complex addition of the complex multiplication value of the first cycle and the complex multiplication value of the second cycle.

  Next, in the third cycle, the product-sum operation unit A1 performs complex multiplication of the input signal d4 and the replica r4, the product-sum operation unit A2 performs complex multiplication of the input signal d5 and the replica r5, and the product-sum operation unit A3. Performs complex multiplication of the input signal d4 and the replica r5, and the product-sum operation unit A4 performs complex multiplication of the input signal d5 and the replica r4. Then, the product-sum calculators A1 to A4 perform cumulative complex addition on the complex multiplication values of the first cycle to the third cycle, respectively.

  Next, in the fourth cycle, the product-sum operation unit A1 performs complex multiplication of the input signal d6 and the replica r6, the product-sum operation unit A2 performs complex multiplication of the input signal d7 and the replica r7, and the product-sum operation unit A3. Performs complex multiplication of the input signal d6 and the replica r7, and the product-sum operation unit A4 performs complex multiplication of the input signal d7 and the replica r6. Then, the product-sum calculators A1 to A4 perform cumulative complex addition on the complex multiplication values of the first cycle to the fourth cycle, respectively.

  Thereafter, similarly, the product-sum calculators A1 to A4 sequentially calculate two input signals and two replicas every cycle. When the input signal and the replica are continuously input to the product-sum calculators A1 to A4, the calculation result is a result obtained by integrating a part of the timings of -1, 0, and +1 in FIG. This value is stored in the accumulator 2613 (FIG. 26) in the product-sum calculators A1 to A4.

  FIG. 26 is a diagram illustrating a configuration example of the product-sum arithmetic units A1 to A4. The product-sum calculators A1 to A4 include a complex multiplier 2611, a complex adder 2612, and an accumulator 2613. The complex multiplier 2611 includes multipliers 2601 to 2604, a subtracter 2605, and an adder 2606. The complex adder 2612 includes adders 2607 and 2608. The accumulator 2613 includes accumulators 2609 and 2610. The product-sum calculators A1 to A4 receive the signals A and B and output the signal acc. For example, signal A is an input signal and signal B is a replica. Signal A has real part signal Re (A) and imaginary part signal Im (A), and signal B has real part signal Re (B) and imaginary part signal Im (B).

  The multiplier 2601 multiplies the real part signal Re (A) of the signal A and the real part signal Re (B) of the signal B, and outputs a multiplication value. The multiplier 2602 multiplies the imaginary part signal Im (A) of the signal A and the imaginary part signal Im (B) of the signal B, and outputs a multiplication value. The multiplier 2603 multiplies the real part signal Re (A) of the signal A and the imaginary part signal Im (B) of the signal B, and outputs a multiplication value. The multiplier 2604 multiplies the imaginary part signal Im (A) of the signal A and the real part signal Re (B) of the signal B, and outputs a multiplication value.

  A subtracter 2605 subtracts the multiplication value of the multiplier 2601 and the multiplication value of the multiplier 2602 and outputs a subtraction value. An adder 2606 adds the multiplication value of the multiplier 2603 and the multiplication value of the multiplier 2604, and outputs the addition value.

  An adder 2607 adds the subtracted value of the subtracter 2605 and the stored value Re (acc) of the accumulator 2609, and performs cumulative addition. The accumulator 2609 stores the addition value Re (acc) of the adder 2607 and outputs it to the adder 2607. An adder 2608 adds the addition value of the adder 2606 and the stored value Im (acc) of the accumulator 2610, and performs cumulative addition. The accumulator 2610 stores the added value Im (acc) of the adder 2608 and outputs it to the adder 2608. The output signal acc includes the real part signal Re (acc) stored in the accumulator 2609 and the imaginary part signal Im (acc) stored in the accumulator 2610, and is output.

  As described above, the product-sum calculator includes the complex multiplier 2611 that performs complex multiplication on the input signal and the replica, and the complex adder 2612 that performs cumulative complex addition on the signal that is complex-multiplied by the complex multiplier 2611.

  FIG. 14 is a diagram showing a correlation calculation method using the product-sum calculator group shown in FIG. In the first scan, as shown in FIG. 13, input signals (0, 1), (2, 3),..., (K-2, K-1), replicas (0, 1), (2, 3),..., (K-2, K-1) are sequentially input, the calculation result area 1401 is stored in the accumulator 2613. The value of the accumulator 2613 is read and the accumulator 2613 is cleared.

  Next, in the second scan, the input signals (2, 3), (4, 5),..., (K, K + 1), replicas (0, 1), (2, 3),. , K−1) in order, the calculation result area 1402 is stored in the accumulator 2613. In this way, by inputting the input signals by shifting two by two, it is possible to obtain the entire value to be subjected to correlation calculation. Since the value of the accumulator 2613 is a value obtained by integrating the odd and even numbers of the respective timings, the correlation value can be obtained by adding all the odd and even results at each timing.

  The correlation value at the timing of “0” is obtained by adding the cumulative addition value of the accumulator 2613 of the product-sum calculator A1 and the cumulative addition value of the accumulator 2613 of the product-sum calculator A2 obtained in the first scan. .

  The correlation value at the timing “+2” is obtained by adding the cumulative addition value of the accumulator 2613 of the product-sum calculator A1 and the cumulative addition value of the accumulator 2613 of the product-sum calculator A2 obtained in the second scan. .

  Further, the correlation value at the timing of “+1” is the cumulative addition value of the accumulator 2613 of the product-sum operation unit A4 obtained in the first scan and the accumulator 2613 of the product-sum operation unit A3 obtained in the second scan. It is obtained by adding the cumulative addition value.

  When obtaining the correlation value at each timing, the correlation values at the timings “0” and “+2” are obtained by one scan, and the correlation value at the timing “+1” is obtained by two scans. Thus, since it is sufficient if there is a memory for two scans as the intermediate data, it is sufficient if there is an intermediate memory for 2 × 4 = 8 data. For example, the first intermediate data memory 2706 (FIG. 27) stores four accumulated addition values of the accumulators 2613 of the product-sum calculators A1 to A4 by the first scan as intermediate data, and the second intermediate data memory 2707. (FIG. 27) stores the four accumulated addition values of the accumulators 2613 of the product-sum arithmetic units A1 to A4 by the second scan as intermediate data. The correlation value at the timing of “+1” is obtained by adding the intermediate data of the first intermediate data memory 2706 and the second intermediate data memory 2707 as described above.

  FIG. 27 is a diagram illustrating a configuration example of the correlation calculation device according to the present embodiment. The correlation calculation device includes memories 2702 and 2703, a product-sum calculator group 2704, a selector 2705, intermediate data memories 2706 and 2707, an adder 2708, a square norm calculator 2709 and a memory 2710. The memory 2702 receives and stores a complex input signal 2701. The memory 2703 stores a complex number replica. The product-sum operation unit group 2704 corresponds to the product-sum operation unit group of FIG. 9, has N × N product-sum operation units, and outputs N × N cumulative addition values. The product-sum operation unit group in FIG. 9 is an example in the case where N is 2. The product-sum operation unit group includes four product-sum operation units and outputs four accumulated addition values. As described above, the product-sum operation unit group 2704 receives two input signals and two replicas from the memories 2702 and 2703 and performs a product-sum operation. The selector 2705 stores the four accumulated addition values of the product-sum operation unit group 2704 in the first intermediate data memory 2706 or the second intermediate data memory 2707. For example, the first intermediate data memory 2706 stores four cumulative addition values of the first scan, and the second intermediate data memory 2707 stores four cumulative addition values of the second scan. . The adder 2708 adds the accumulated addition values of the intermediate data memories 2706 and / or 2707 as described above. The square norm calculator 2709 calculates the square norm of the added value of the adder 2708 and stores it in the memory 2710 as a correlation value. Thereafter, as in FIG. 2, the maximum value selector 207 selects and outputs the timing information of the maximum value among the correlation values in the memory 2710. The timing is the position of the replica in the input signal and indicates the head position of the input signal.

  FIG. 20 is a diagram illustrating an example of an input signal and a replica. The memory 2702 in FIG. 27 stores N input signals, and the memory 2703 in FIG. 27 stores K replicas.

  FIG. 21 is a diagram illustrating a configuration example of the product-sum operation unit group, and has the same configuration as the product-sum operation unit group of FIG. The product-sum operation unit E00 performs a product-sum operation on the input signal d0 and the replica r0 and outputs a cumulative addition value acc_01. The product-sum operation unit E10 performs a product-sum operation on the input signal d1 and the replica r1, and outputs a cumulative addition value acc_10. The product-sum operation unit E01 performs a product-sum operation on the input signal d0 and the replica r1, and outputs a cumulative addition value acc_00. The product-sum operation unit E11 performs a product-sum operation on the input signal d1 and the replica r0 and outputs a cumulative addition value acc_11.

  FIG. 22 is a diagram illustrating the intermediate data memory. The first intermediate data memory 2706 and the second intermediate data memory 2707 correspond to the first intermediate data memory 2706 and the second intermediate data memory 2707 in FIG. 27, and each of the four accumulated addition values acc_00 to acc_11. It has a memory area.

  FIG. 19 is a flowchart showing a correlation calculation method of the correlation calculation device. The loop 1912 has a start end 1901 and an end 1908, and is a loop that increases the variable j by 2 from 0 to N. The loop 1911 has a start end 1902 and an end end 1905, and increases the variable i by 2 from 0 to K.

  In step 1903, the product-sum calculators E00 to E11 receive the input signals d0 and d1 and the replicas r0 and r1. The input signal d0 is the j + i-th input signal. The input signal d1 is the j + i + 1th input signal. The replica r0 is the i-th replica. The replica r1 is the (i + 1) th replica.

  Next, in step 1904, the product-sum calculator E01 performs complex multiplication of the input signal d0 and the replica r1, and cumulatively adds the complex multiplication value to the cumulative addition value acc_00. The product-sum operation unit E00 performs complex multiplication of the input signal d0 and the replica r0, and cumulatively adds the complex multiplication value to the cumulative addition value acc_01. The product-sum operation unit E10 performs complex multiplication of the input signal d1 and the replica r1, and cumulatively adds the complex multiplication value to the cumulative addition value acc_10. The product-sum calculator E11 performs complex multiplication of the input signal d1 and the replica r0, and cumulatively adds the complex multiplication value to the cumulative addition value acc_11. The above processing is repeated as a loop 1911.

  Next, in step 1906, the accumulated addition value of the accumulator is read and stored in the intermediate data memory. The cumulative addition value acc_00 is stored in the intermediate data memory m [1] [j-1]. The cumulative addition value acc_01 is stored in the intermediate data memory m [0] [j]. The cumulative addition value acc_10 is stored in the intermediate data memory m [1] [j]. The cumulative addition value acc_11 is stored in the intermediate data memory m [0] [j + 1].

  Next, in step 1907, the intermediate data in the intermediate data memory is read and added. The adder 2708 adds the intermediate data in the intermediate data memories m [0] [j−1] and m [1] [j−1], and outputs a correlation value COR [j−1]. In this addition, for example, the cumulative addition value of the product-sum calculator A4 in the calculation result area 1401 of the first cycle in FIG. 14 and the cumulative addition value of the product-sum calculator A3 in the calculation result area 1402 of the second cycle are obtained. Addition is performed to obtain a correlation value at the timing of “+1”. The adder 2708 adds the intermediate data in the intermediate data memories m [0] [j] and m [1] [j] and outputs a correlation value COR [j]. In this addition, for example, the cumulative addition values of the product-sum calculators A1 and A2 in the calculation result area 1401 of the first cycle in FIG. 14 are added to obtain the correlation value at the timing “0”. The above processing is repeated as a loop 1912.

(Second Embodiment)
FIG. 8 corresponds to FIG. 7 and is a diagram illustrating a configuration example of a product-sum operation unit group in the correlation operation device according to the second embodiment of the present invention. In the first embodiment, the case of the product-sum operation unit group of four product-sum operation units E00 to E11 has been described. However, in this embodiment, the product-sum operation unit group of 16 product-sum operation units E00 to E33 is described. Explain the case. Hereinafter, the points of the present embodiment different from the first embodiment will be described.

  The product-sum operation unit group includes N × N (for example, 16) product-sum operation units E00 to E33. The input signal d0 is a first complex number input signal, the input signal d1 is a second complex number input signal, the input signal d2 is a third complex number input signal, and the input signal d3 is a fourth complex number input signal. Replica r0 is a first complex replica, replica r1 is a second complex replica, replica r2 is a third complex replica, and replica r3 is a fourth complex replica. The product-sum operation unit E00 performs a product-sum operation on the input signal d0 and the replica r0. The product-sum operation unit E01 performs a product-sum operation on the input signal d0 and the replica r3. The product-sum operation unit E02 performs a product-sum operation on the input signal d0 and the replica r2. The product-sum operation unit E03 performs a product-sum operation on the input signal d0 and the replica r1. The product-sum operation unit E10 performs a product-sum operation on the input signal d1 and the replica r1. The product-sum operation unit E11 performs a product-sum operation on the input signal d1 and the replica r0. The product-sum operation unit E12 performs a product-sum operation on the input signal d1 and the replica r3. The product-sum operation unit E13 performs a product-sum operation on the input signal d1 and the replica r2. The product-sum operation unit E20 performs a product-sum operation on the input signal d2 and the replica r2. The product-sum operation unit E21 performs a product-sum operation on the input signal d2 and the replica r1. The product-sum operation unit E22 performs a product-sum operation on the input signal d2 and the replica r0. The product-sum operation unit E23 performs a product-sum operation on the input signal d2 and the replica r3. The product-sum operation unit E30 performs a product-sum operation on the input signal d3 and the replica r3. The product-sum operation unit E31 performs a product-sum operation on the input signal d3 and the replica r2. The product-sum operation unit E32 performs a product-sum operation on the input signal d3 and the replica r1. The product-sum operation unit E33 performs a product-sum operation on the input signal d3 and the replica r0.

  FIG. 10 is a diagram corresponding to FIG. 9 and showing a product-sum operation unit group in which the product-sum operation unit group of FIG. 8 is rearranged into a two-dimensional matrix, and has the same configuration as the product-sum operation unit group of FIG. . When the product-sum calculators E00 to E33 are arranged on a square of a two-dimensional matrix, the wiring relationship between the input signals d0 to d3 and the replicas r0 to r3 is relatively easy to understand.

  FIG. 15 is a diagram corresponding to FIG. 11 and illustrating a correlation calculation method using the product-sum calculator group of FIG. The correlation calculation device performs correlation calculation between N input signals and K replicas. Similar to the first embodiment, the correlation between the input signal and the replica is calculated while shifting the timing of the input signal and the replica.

  FIG. 16 corresponds to FIG. 12 and shows a product-sum operation method by the product-sum operation unit group of FIG. The 16 product-sum calculators B1 to B16 correspond to the 16 product-sum calculators E00 to E33 in FIG. The 16 product-sum calculators B1 to B16 forming the parallelogram correspond to those extracted from the timings “−3” to “+3” in FIG.

  FIG. 17 corresponds to FIG. 13 and shows processing for each cycle of the product-sum operation unit group of FIG. In the first cycle, as in FIG. 16, the product-sum calculators B1 to B16 perform complex multiplication of the input signals d0 to d3 and the replicas r0 to r3. Next, in the second cycle, the product-sum calculators B1 to B16 perform complex multiplication of the input signals d4 to d7 and the replicas r4 to r7, and the complex multiplication value of the first cycle and the complex multiplication value of the second cycle. And cumulative complex addition. Next, in the third cycle, the product-sum calculators B1 to B16 perform complex multiplication of the input signals d8 to d11 and the replicas r8 to r11, and perform cumulative complex addition on the complex multiplication values of the first to third cycles. Next, in the fourth cycle, the product-sum calculators B1 to B16 perform complex multiplication of the input signals d12 to d15 and the replicas r12 to r15, and perform cumulative complex addition of the complex multiplication values of the first to fourth cycles. Thereafter, similarly, the product-sum calculators B1 to B16 sequentially calculate four input signals and replicas every cycle.

  FIG. 18 is a diagram corresponding to FIG. 14 and illustrating a correlation calculation method using the product-sum calculator group of FIG. In the first scan, as shown in FIG. 17, the input signals (0, 1, 2, 3), (4, 5, 6, 7), ..., (K-4, K-3, K-2). , K-1), replica (0, 1, 2, 3), (4, 5, 6, 7), ..., (K-4, K-3, K-2, K-1) As a result, the calculation result area 1801 is stored in the accumulator 2613. The value of the accumulator 2613 is read and the accumulator 2613 is cleared.

  Next, in the second scan, the input signals (4, 5, 6, 7), (8, 9, 10, 11), ..., (K, K + 1, K + 2, K + 3), replica (0, 1, 2, , 3), (4, 5, 6, 7),..., (K-4, K-3, K-2, K-1) are sequentially input, the operation result area 1802 is stored in the accumulator 2613. Is done. In this way, by shifting the input signals by four each, it is possible to obtain the whole value to be correlated. A correlation value can be obtained by adding all the calculation results at each timing.

  The correlation value at the timing of “0” is obtained by adding the cumulative addition values of the accumulators 2613 of the product-sum calculators B1 to B4 obtained in the first scan. Further, the correlation value at the timing of “+1” is the accumulator of the accumulator 2613 of the product-sum calculators B6 to B8 obtained in the first scan and the accumulator of the product-sum calculator B5 obtained in the second scan. It is obtained by adding the cumulative addition value of 2613. Further, the correlation value at the timing of “+2” is obtained by using the cumulative addition value of the accumulator 2613 of the product-sum calculators B11 and B12 obtained in the first scan and the product-sum calculators B9 and B10 obtained in the second scan. It is obtained by adding the cumulative addition value of the accumulator 2613. Further, the correlation value at the timing of “+3” is the accumulated value of the accumulator 2613 of the product-sum calculator B16 obtained in the first scan and the accumulator of the product-sum calculators B13 to B15 obtained in the second scan. It is obtained by adding the cumulative addition value of 2613.

  When obtaining the correlation value of each timing, the correlation value of the timing “0” is obtained by one scan, and the correlation value of the timings “+1” to “+3” is obtained by two scans. Thus, since it is sufficient if there is a memory for two scans as the intermediate data, it is sufficient that there is an intermediate memory for 2 × 16 = 32 data. For example, the first intermediate data memory 2706 (FIG. 25, FIG. 27) stores the 16 accumulated addition values of the accumulators 2613 of the product-sum calculators B1 to B16 by the first scan as intermediate data. The data memory 2707 (FIGS. 25 and 27) stores the 16 accumulated addition values of the accumulators 2613 of the product-sum calculators B1 to B16 by the second scan as intermediate data. As described above, the correlation value of the timings “+1” to “+3” is obtained by adding the intermediate data in the first intermediate data memory 2706 and the second intermediate data memory 2707.

  FIG. 24 corresponds to FIG. 21 and shows a configuration example of the product-sum operation unit group, and has the same configuration as the product-sum operation unit group of FIG. The product-sum operation unit E00 performs a product-sum operation on the input signal d0 and the replica r0, and outputs a cumulative addition value acc_00. The product-sum operation unit E10 performs a product-sum operation on the input signal d1 and the replica r1, and outputs a cumulative addition value acc_01. The product-sum operation unit E20 performs a product-sum operation on the input signal d2 and the replica r2, and outputs a cumulative addition value acc_02. The product-sum operation unit E30 performs a product-sum operation on the input signal d3 and the replica r3, and outputs a cumulative addition value acc_03.

  The product-sum operation unit E01 performs a product-sum operation on the input signal d0 and the replica r3, and outputs a cumulative addition value acc_10. The product-sum operation unit E11 performs a product-sum operation on the input signal d1 and the replica r0 and outputs a cumulative addition value acc_11. The product-sum operation unit E21 performs a product-sum operation on the input signal d2 and the replica r1, and outputs a cumulative addition value acc_12. The product-sum operation unit E31 performs a product-sum operation on the input signal d3 and the replica r2, and outputs a cumulative addition value acc_13.

  The product-sum operation unit E02 performs a product-sum operation on the input signal d0 and the replica r2, and outputs a cumulative addition value acc_20. The product-sum operation unit E12 performs a product-sum operation on the input signal d1 and the replica r3, and outputs a cumulative addition value acc_21. The product-sum operation unit E22 performs a product-sum operation on the input signal d2 and the replica r0 and outputs a cumulative addition value acc_22. The product-sum operation unit E32 performs a product-sum operation on the input signal d3 and the replica r1, and outputs a cumulative addition value acc_23.

  The product-sum operation unit E03 performs a product-sum operation on the input signal d0 and the replica r1, and outputs a cumulative addition value acc_30. The product-sum operation unit E13 performs a product-sum operation on the input signal d1 and the replica r2, and outputs a cumulative addition value acc_31. The product-sum operation unit E23 performs a product-sum operation on the input signal d2 and the replica r3, and outputs a cumulative addition value acc_32. The product-sum operation unit E33 performs a product-sum operation on the input signal d3 and the replica r0, and outputs a cumulative addition value acc_33.

  FIG. 25 corresponds to FIG. 22 and shows an intermediate data memory. The first intermediate data memory 2706 and the second intermediate data memory 2707 correspond to the first intermediate data memory 2706 and the second intermediate data memory 2707 in FIG. 27, and each of the 16 cumulative addition values acc_00 to acc_33 is stored. It has a memory area.

  FIG. 23 is a flowchart corresponding to FIG. 19 and showing a correlation calculation method of the correlation calculation device. The loop 2312 has a start end 2301 and an end end 2308, and is a loop that increases the variable j by 4 from 0 to N. The loop 2311 has a start end 2302 and an end end 2305, and increases the variable i by 4 from 0 to K.

  In step 2303, the product-sum calculators E00 to E33 receive the input signals d0 to d3 and the replicas r0 to r3. The input signal d0 is the j + i-th input signal. The input signal d1 is the j + i + 1th input signal. The input signal d2 is a j + i + 2nd input signal. The input signal d3 is the j + i + 3rd input signal. The replica r0 is the i-th replica. The replica r1 is the (i + 1) th replica. The replica r2 is the i + 2nd replica. The replica r3 is the i + 3rd replica.

  Next, in step 2304, the product-sum operation unit E00 performs complex multiplication of the input signal d0 and the replica r0, and cumulatively adds the complex multiplication value to the cumulative addition value acc_00. The product-sum operation unit E10 performs complex multiplication of the input signal d1 and the replica r1, and cumulatively adds the complex multiplication value to the cumulative addition value acc_01. The product-sum operation unit E20 performs complex multiplication of the input signal d2 and the replica r2, and cumulatively adds the complex multiplication value to the cumulative addition value acc_02. The product-sum operation unit E30 performs complex multiplication of the input signal d3 and the replica r3, and cumulatively adds the complex multiplication value to the cumulative addition value acc_03.

  The product-sum calculator E01 performs complex multiplication of the input signal d0 and the replica r3, and cumulatively adds the complex multiplication value to the cumulative addition value acc_10. The product-sum calculator E11 performs complex multiplication of the input signal d1 and the replica r0, and cumulatively adds the complex multiplication value to the cumulative addition value acc_11. The product-sum operation unit E21 performs complex multiplication of the input signal d2 and the replica r1, and cumulatively adds the complex multiplication value to the cumulative addition value acc_12. The product-sum calculator E31, the input signal d3, and the replica r2 are subjected to complex multiplication, and the complex multiplication value is cumulatively added to the cumulative addition value acc_13.

  The product-sum operation unit E02 performs complex multiplication of the input signal d0 and the replica r2, and cumulatively adds the complex multiplication value to the cumulative addition value acc_20. The product-sum operation unit E12 performs complex multiplication of the input signal d1 and the replica r3, and cumulatively adds the complex multiplication value to the cumulative addition value acc_21. The product-sum calculator E22 performs complex multiplication of the input signal d2 and the replica r0, and cumulatively adds the complex multiplication value to the cumulative addition value acc_22. The product-sum calculator E32 performs complex multiplication of the input signal d3 and the replica r1, and cumulatively adds the complex multiplication value to the cumulative addition value acc_23.

  The product-sum operation unit E03 performs complex multiplication of the input signal d0 and the replica r1, and cumulatively adds the complex multiplication value to the cumulative addition value acc_30. The product-sum operation unit E13 performs complex multiplication of the input signal d1 and the replica r2, and cumulatively adds the complex multiplication value to the cumulative addition value acc_31. The product-sum calculator E23 performs complex multiplication of the input signal d2 and the replica r3, and cumulatively adds the complex multiplication value to the cumulative addition value acc_32. The product-sum operation unit E33 performs complex multiplication of the input signal d3 and the replica r0, and cumulatively adds the complex multiplication value to the cumulative addition value acc_33.

  The above processing is repeated as a loop 2311.

  Next, in step 2306, the accumulated addition value of the accumulator is read and stored in the intermediate data memory. The cumulative addition value acc_00 is stored in the intermediate data memory m [0] [j]. The cumulative addition value acc_01 is stored in the intermediate data memory m [1] [j]. The cumulative addition value acc_02 is stored in the intermediate data memory m [2] [j]. The cumulative addition value acc_03 is stored in the intermediate data memory m [3] [j]. The cumulative addition value acc_10 is stored in the intermediate data memory m [3] [j-3]. The cumulative addition value acc_11 is stored in the intermediate data memory m [0] [j + 1]. The cumulative addition value acc_12 is stored in the intermediate data memory m [1] [j + 1]. The cumulative addition value acc_13 is stored in the intermediate data memory m [2] [j + 1]. The cumulative addition value acc_20 is stored in the intermediate data memory m [2] [j-2]. The cumulative addition value acc_21 is stored in the intermediate data memory m [3] [j-2]. The cumulative addition value acc_22 is stored in the intermediate data memory m [0] [j + 2]. The cumulative addition value acc_23 is stored in the intermediate data memory m [1] [j + 2]. The cumulative addition value acc_30 is stored in the intermediate data memory m [1] [j-1]. The cumulative addition value acc_31 is stored in the intermediate data memory m [2] [j-1]. The cumulative addition value acc_32 is stored in the intermediate data memory m [3] [j-1]. The cumulative addition value acc_33 is stored in the intermediate data memory m [0] [j + 3].

  Next, in step 2307, the intermediate data in the intermediate data memory is read and added. The adder 2708 calculates the values of the intermediate data memories m [0] [j-3], m [1] [j-3], m [2] [j-3], and m [3] [j-3]. Add the values and output the correlation value COR [j-3]. This addition is, for example, a correlation value calculation at the timing of “+1” in FIG. The adder 2708 also stores the intermediate data memories m [0] [j-2], m [1] [j-2], m [2] [j-2], and m [3] [j-2]. The values are added and the correlation value COR [j-2] is output. This addition is, for example, a correlation value calculation at the timing “+2” in FIG. The adder 2708 also stores the intermediate data memories m [0] [j−1], m [1] [j−1], m [2] [j−1], and m [3] [j−1]. The values are added and the correlation value COR [j−1] is output. This addition is, for example, a correlation value calculation at the timing of “+3” in FIG. Further, the adder 2708 adds the values of the intermediate data memories m [0] [j], m [1] [j], m [2] [j], and m [3] [j] to obtain a correlation value COR. [J] is output. This addition is, for example, a correlation value calculation at a timing of “0” in FIG. The above processing is repeated as a loop 2312.

  As described above, in the first and second embodiments, the N × N product-sum calculators E00 to E11 and E00 to E33 include N (N is an integer of 2 or more) input signals in the input signal. A combination of any one of the above and any one of the N replica signals in the replica signal is cumulatively added by a product-sum operation, and a cumulative addition value is output and input in the next cycle A combination of any one of the next N input signals in the signal and any one of the next N replica signals in the replica signal is accumulated and accumulated by a product-sum operation. The process of outputting the added value is repeated.

  The adder 2708 adds the accumulated addition values in the same cycle or different cycles output from the product-sum calculators E00 to E11 and E00 to E33 having the same amount of shifting the input signal with respect to the replica signal. The correlation value of the input signal and the replica signal for each input signal shift amount with respect to the signal is output.

  The product-sum calculators E00 to E11 and E00 to E33 first repeat the calculation for each of the N input signals and the N replica signals by setting the shift amount of the input signal with respect to the replica signal to 0 in the first scan. Next, in the second scan, the shift amount of the input signal with respect to the replica signal is set to N, and the calculation for each of the N input signals and the N replica signals is repeated.

  In the first embodiment, N is 2, the four product-sum calculators are the first to fourth product-sum calculators, and the first four input signals of the input signal are the first to fourth. The first four replica signals are the first to fourth replica signals.

  In the first cycle of FIG. 13, the first product-sum calculator performs a product-sum operation on the first input signal and the first replica signal. The second product-sum operation unit performs a product-sum operation on the first input signal and the second replica signal. The third product-sum operation unit performs a product-sum operation on the second input signal and the first replica signal. The fourth product-sum operation unit performs a product-sum operation on the second input signal and the second replica signal.

  In the second cycle of FIG. 13, the first product-sum operation unit performs a product-sum operation on the third input signal and the third replica signal. The second product-sum operation unit performs a product-sum operation on the third input signal and the fourth replica signal. The third product-sum operation unit performs a product-sum operation on the fourth input signal and the third replica signal. The fourth product-sum operation unit performs a product-sum operation on the fourth input signal and the fourth replica signal.

  In the second embodiment, N is 4, the sixteen product-sum calculators are the first to sixteenth product-sum calculators, and the first eight input signals of the input signal are the first to the eighth. The first eight replica signals of the replica signal are the first to eighth replica signals.

  In the first cycle of FIG. 17, the first product-sum calculator performs a product-sum operation on the first input signal and the first replica signal. The second product-sum operation unit performs a product-sum operation on the first input signal and the second replica signal. The third product-sum operation unit performs a product-sum operation on the first input signal and the third replica signal. The fourth product-sum operation unit performs a product-sum operation on the first input signal and the fourth replica signal. The fifth product-sum operation unit performs a product-sum operation on the second input signal and the first replica signal. The sixth product-sum operation unit performs a product-sum operation on the second input signal and the second replica signal. The seventh product-sum operation unit performs a product-sum operation on the second input signal and the third replica signal. The eighth product-sum operation unit performs a product-sum operation on the second input signal and the fourth replica signal. The ninth product-sum operation unit performs a product-sum operation on the third input signal and the first replica signal. The tenth product-sum calculator performs a product-sum operation on the third input signal and the second replica signal. The eleventh product-sum calculator performs a product-sum operation on the third input signal and the third replica signal. The twelfth product-sum operation unit performs a product-sum operation on the third input signal and the fourth replica signal. The thirteenth product-sum calculator performs a product-sum operation on the fourth input signal and the first replica signal. The fourteenth product-sum operation unit performs a product-sum operation on the fourth input signal and the second replica signal. The fifteenth product-sum operation unit performs a product-sum operation on the fourth input signal and the third replica signal. The sixteenth product-sum operation unit performs a product-sum operation on the fourth input signal and the fourth replica signal.

  In the second cycle of FIG. 17, the first product-sum calculator performs a product-sum operation on the fifth input signal and the fifth replica signal. The second product-sum operation unit performs a product-sum operation on the fifth input signal and the sixth replica signal. The third product-sum operation unit performs a product-sum operation on the fifth input signal and the seventh replica signal. The fourth product-sum operation unit performs product-sum operation on the fifth input signal and the eighth replica signal. The fifth product-sum operation unit performs a product-sum operation on the sixth input signal and the fifth replica signal. The sixth product-sum operation unit performs a product-sum operation on the sixth input signal and the sixth replica signal. The seventh product-sum operation unit performs a product-sum operation on the sixth input signal and the seventh replica signal. The eighth product-sum operation unit performs a product-sum operation on the sixth input signal and the eighth replica signal. The ninth product-sum operation unit performs a product-sum operation on the seventh input signal and the fifth replica signal. The tenth product-sum operation unit performs a product-sum operation on the seventh input signal and the sixth replica signal. The eleventh product-sum operation unit performs a product-sum operation on the seventh input signal and the seventh replica signal. The twelfth product-sum operation unit performs a product-sum operation on the seventh input signal and the eighth replica signal. The thirteenth product-sum calculator performs a product-sum operation on the eighth input signal and the fifth replica signal. The fourteenth product-sum operation unit performs a product-sum operation on the eighth input signal and the sixth replica signal. The fifteenth product-sum operation unit performs a product-sum operation on the eighth input signal and the seventh replica signal. The sixteenth product-sum operation unit performs a product-sum operation on the eighth input signal and the eighth replica signal.

  As described above, in the first and second embodiments, the correlation value can be output by repeating the product-sum operation using N × N product-sum operation units. And the size can be reduced. In addition, since the product-sum operation unit performs cumulative addition, the size of the memory for storing the intermediate data can be reduced. Specifically, there may be a first intermediate data memory 2706 for storing N × N intermediate data and a second intermediate data memory 2707 for storing N × N intermediate data.

  The above-described embodiments are merely examples of implementation in carrying out the present invention, and the technical scope of the present invention should not be construed in a limited manner. That is, the present invention can be implemented in various forms without departing from the technical idea or the main features thereof.

E00-E33 Multiply-add calculator 2701 Input signal 2702, 2703 Memory 2704 Multiply-add calculator 2705 Selector 2706 First intermediate data memory 2707 Second intermediate data memory 2708 Adder 2709 Square norm calculator 2710 Memory

Claims (5)

A combination of any one of N (N is an integer of 2 or more) input signals in the input signal and any one of N replica signals in the replica signal is performed by multiply-add operation. Cumulative addition is performed to output a cumulative addition value, and any one of the next N input signals in the input signal and any of the next N replica signals in the replica signal in the next cycle. Or N × N product-sum calculators that repeat the process of accumulating the combinations of the two and multiplying them by a product-sum operation and outputting the accumulated addition value,
The amount of shift of the input signal with respect to the replica signal by adding the cumulative addition values in the same cycle or different cycles output by the product-sum calculators with the same amount of shifting the input signal with respect to the replica signal And an adder for outputting a correlation value of each input signal and replica signal.   In the first scan, the product-sum calculator first repeats the calculation for each of the N input signals and the N replica signals by setting the shift amount of the input signal with respect to the replica signal to 0. 2. The correlation according to claim 1, wherein in the second scan, the amount of deviation of the input signal with respect to the replica signal is set to N, and the calculation is repeated for each of the N input signals and the N replica signals. Arithmetic unit. The input signal and the replica signal are complex signals;
The product-sum calculator is
A complex multiplier that performs complex multiplication of the input signal and the replica signal;
The correlation calculation apparatus according to claim 1, further comprising: a complex adder that performs cumulative complex addition on the signals that are complex-multiplied by the complex multiplier. N is 2;
The four product-sum operation units are first to fourth product-sum operation units,
The first four input signals of the input signals are first to fourth input signals,
The first four replica signals of the replica signal are first to fourth replica signals,
In the first cycle, the first product-sum operation unit performs a product-sum operation on the first input signal and the first replica signal, and the second product-sum operation unit performs the first input signal. And the third product-sum operation unit performs a product-sum operation on the second input signal and the first replica signal, and performs the fourth product-sum operation. A product sum operation of the second input signal and the second replica signal,
In the second cycle, the first product-sum operation unit performs a product-sum operation on the third input signal and the third replica signal, and the second product-sum operation unit performs the third input signal. And the fourth product-sum operation is performed. The third product-sum operation unit performs a product-sum operation on the fourth input signal and the third replica signal. 4. The correlation calculation device according to claim 1, wherein the counter performs a product-sum operation on the fourth input signal and the fourth replica signal. 5. N is 4;
The 16 product-sum operation units are first to 16th product-sum operation units,
The first eight input signals of the input signals are first to eighth input signals,
The first eight replica signals of the replica signal are first to eighth replica signals,
In the first cycle, the first product-sum operation unit performs a product-sum operation on the first input signal and the first replica signal, and the second product-sum operation unit performs the first input signal. The third product-sum operation unit performs a product-sum operation on the first input signal and the third replica signal, and performs the fourth product-sum operation. A product sum operation of the first input signal and the fourth replica signal, and a fifth product sum operator performs a product sum operation of the second input signal and the first replica signal. The sixth product-sum operation unit performs a product-sum operation on the second input signal and the second replica signal, and the seventh product-sum operation unit performs the second input signal and the third input signal. The product-sum operation of the replica signal is performed, and the eighth product-sum operation unit is configured to output the second input signal and the fourth replication signal. The ninth product-sum operation unit performs a product-sum operation on the third input signal and the first replica signal, and the tenth product-sum operation unit performs the third product-sum operation. The eleventh product-sum calculator performs a product-sum operation on the third input signal and the third replica signal, and the twelfth input signal and the second replica signal. The product-sum operation unit performs a product-sum operation on the third input signal and the fourth replica signal, and the thirteenth product-sum operation unit performs a product-sum operation on the fourth input signal and the first replica signal. The fourteenth product-sum operation unit performs a product-sum operation on the fourth input signal and the second replica signal, and the fifteenth product-sum operation unit performs the fourth input signal and the second replica signal. A product-sum operation is performed on the third replica signal, and the sixteenth product-sum operation unit performs the fourth input signal. And it performs a product-sum operation of the fourth replica signal,
In the second cycle, the first product-sum operation unit performs a product-sum operation on the fifth input signal and the fifth replica signal, and the second product-sum operation unit performs the fifth input signal. And the sixth product-sum operation, the third product-sum operation unit performs the product-sum operation of the fifth input signal and the seventh replica signal, and the fourth product-sum operation. A product sum operation of the fifth input signal and the eighth replica signal, and a fifth product sum operator performs a product sum operation of the sixth input signal and the fifth replica signal. The sixth product-sum calculator performs a product-sum operation on the sixth input signal and the sixth replica signal, and the seventh product-sum calculator calculates the sixth input signal and the seventh replica. The product-sum operation of the replica signal is performed, and the eighth product-sum operation unit is configured to output the sixth input signal and the eighth replica. The ninth product-sum operation unit performs a product-sum operation on the seventh input signal and the fifth replica signal, and the tenth product-sum operation unit performs the seventh product-sum operation. The eleventh product-sum operation unit performs a product-sum operation on the seventh input signal and the seventh replica signal, and the twelfth replica signal. The product-sum operation unit performs a product-sum operation on the seventh input signal and the eighth replica signal, and the thirteenth product-sum operation unit calculates the product sum of the eighth input signal and the fifth replica signal. The fourteenth product-sum operation unit performs a product-sum operation on the eighth input signal and the sixth replica signal, and the fifteenth product-sum operation unit performs the eighth input signal and the sixth replica signal. A product-sum operation is performed on the seventh replica signal, and the sixteenth product-sum operation unit performs the eighth input signal. And the correlation calculation apparatus according to any one of claims 1 to 3, characterized in that the product-sum operation of the eighth replica signal.

Images