WO2011102291A1 - 高速フーリエ変換回路 - Google Patents
高速フーリエ変換回路 Download PDFInfo
- Publication number
- WO2011102291A1 WO2011102291A1 PCT/JP2011/052844 JP2011052844W WO2011102291A1 WO 2011102291 A1 WO2011102291 A1 WO 2011102291A1 JP 2011052844 W JP2011052844 W JP 2011052844W WO 2011102291 A1 WO2011102291 A1 WO 2011102291A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- data
- memory
- fourier transform
- fast fourier
- stage
- Prior art date
Links
- 230000015654 memory Effects 0.000 claims abstract description 167
- 238000004364 calculation method Methods 0.000 claims abstract description 76
- 239000000470 constituent Substances 0.000 claims description 4
- 238000000034 method Methods 0.000 description 53
- 238000012545 processing Methods 0.000 description 50
- 238000010586 diagram Methods 0.000 description 30
- 238000003775 Density Functional Theory Methods 0.000 description 22
- 238000004891 communication Methods 0.000 description 10
- 238000007792 addition Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 description 3
- 230000001174 ascending effect Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000007796 conventional method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
- G06F17/141—Discrete Fourier transforms
- G06F17/142—Fast Fourier transforms, e.g. using a Cooley-Tukey type algorithm
Definitions
- the present invention relates to a fast Fourier transform circuit that processes a Fourier transform at high speed.
- Cooley-Tukey-type fast Fourier transform In the Cooley-Tukey-type fast Fourier transform (hereinafter simply referred to as “fast Fourier transform”), generally, a calculation is performed using a two-point DFT (Digital Fourier Transform Discrete Fourier Transform) as a constituent element (cited documents 1 to 3). reference).
- a two-point DFT Digital Fourier Transform Discrete Fourier Transform
- radix-4 fast Fourier transform an operation using a 4-point DFT as a constituent element
- the data number N needs to be a power of 2
- the data number N needs to be a power of 4.
- An object of the present invention is to solve such problems and to provide a fast Fourier transform circuit capable of reading and writing data processed at individual stages of fast Fourier transform at high speed without dividing a memory.
- the fast Fourier transform circuit of the present invention includes a computation unit that performs a fast Fourier computation including a plurality of digital Fourier transforms, a memory that stores input / output data of the computation unit, and a computation unit for processing target data. And a means for controlling the writing of the calculation result to the memory by the calculation unit so that the reading order of the data from the memory is the same in each stage for a plurality of stages of calculations to be performed.
- the present invention it is possible to read and write data processed by individual stages of fast Fourier transform at high speed without dividing the memory, and to increase the speed while suppressing an increase in LSI area.
- FIG. 1 is a block configuration diagram showing a fast Fourier transform circuit according to a first embodiment of the present invention. It is a signal flow figure explaining operation
- FIG. 5 is a diagram illustrating a flow of processing for writing data from a write buffer to a memory. It is a figure explaining the flow of the process of the data by the fast Fourier transform circuit shown in FIG.
- FIG. 5 is a diagram illustrating a flow of processing for writing data from a write buffer to a memory. It is a figure explaining the flow of the process of the data by the fast Fourier transform circuit shown in FIG.
- FIG. 5 is a diagram illustrating a flow of processing for writing data from a write buffer to a memory. It is a figure explaining the flow of the process of the data by the fast Fourier-transform circuit shown in FIG.
- FIG. 5 is a diagram illustrating a flow of processing for writing data from a write buffer to a memory. It is a figure explaining the flow of the process of the data by the fast Fourier transform circuit shown in FIG.
- FIG. 5 is a diagram illustrating a flow of processing for writing data from a write buffer to a memory. It is a figure explaining the flow of the process of the data by the fast Fourier transform circuit shown in FIG. 1, and is a figure which shows the flow of the process of writing the data from the write buffer to the memory when the process of the first stage is completed.
- FIG. 20 is a diagram illustrating a calculation configuration example of a 2-point / 4-point DFT calculation unit in the fast Fourier transform circuit illustrated in FIG. 19.
- FIG. 19 performs the first stage operation, data is read from the memory, stored in the read buffer, and 2-point / 4-point DFT operation section. It is a figure which shows the flow of a process of the calculation by, and the storage of the calculation result to the write buffer.
- FIG. 20 is a diagram illustrating a flow of a process of writing data from the write buffer to the memory when the 2-point / 4-point DFT operation unit in the fast Fourier transform circuit illustrated in FIG. 19 performs the first stage operation.
- the 2-point / 4-point DFT operation section in the fast Fourier transform circuit shown in FIG. 19 performs the second stage operation, data is read from the memory, stored in the read buffer, and 2-point / 4-point DFT operation section.
- FIG. 20 is a diagram showing a flow of processing of writing data from the write buffer to the memory when the 2-point / 4-point DFT operation unit in the fast Fourier transform circuit shown in FIG. 19 performs the second stage operation.
- the 2-point / 4-point DFT computation unit in the fast Fourier transform circuit shown in FIG. 19 performs the third stage computation
- the computation result of the 2-point / 4-point DFT computation unit is stored in the write buffer, and the write buffer
- the write buffer It is a figure which shows the flow of a process of writing the data from memory to memory. It is a figure which shows the example of whole structure of the communication system using an OFDM system, and is a figure which shows only the structure regarding communication of one direction.
- the fast Fourier transform circuit includes a four-point DFT operation unit 1, memories 2A and 2B, a read buffer 3, a write buffer 4, a selector 5, a rotator generation unit 6, and a control unit 7.
- the read buffer 3 has a configuration of B complex data ⁇ (N / B) ⁇ 2 banks, stores data read from the memories 2A and 2B, and stores 4-point DFT operation unit 1 for each radix B data. Output to. Since the read buffer 3 has a 2-bank configuration for N complex data, the data is output to the 4-point DFT operation unit 1 in the other bank while data is read from the memories 2A and 2B in one bank. Can do.
- the write buffer 4 has a B complex data ⁇ (N / B) ⁇ 2 bank configuration, stores the calculation results of each stage by the 4-point DFT calculation unit 1, and writes them in the memories 2A and 2B. Since the write buffer 4 is composed of two banks for N complex data, the calculation result from the four-point DFT operation unit 1 is received in the other bank while writing to the memories 2A and 2B in one bank. Can do.
- the selector 5 selects the read / write destination of the memories 2A and 2B. That is, control is performed so that one of the memories 2A and 2B is a reading side and the other is a writing side. Reading and writing to the same memory 2A or 2B do not occur at the same time.
- the rotor generation unit 6 generates a rotor that multiplies the output of the 4DFT calculation unit 1.
- the rotor is a well-known matter in the field of fast Fourier transform, and the description thereof is omitted here.
- the multiplication of the rotor, which is the output of the rotor generation unit 16, and the output of the 4-point DFT calculation unit 15 is performed before writing to the write buffer 4. It may be done when writing.
- the control unit 7 performs reading of the memories 2A and 2B, generation of a write address, generation of addresses of the read buffer 3 and the write buffer 4, and a plurality of stages of calculations performed by the calculation unit on the processing target data.
- the calculation unit writes data to the memory so that the reading order of data from the memory is the same in each stage.
- the control unit 7 also performs the generation control of the rotor in the rotor generation unit 6, and the management and control of the number of times and the processing stage of the 4-point DFT calculation unit.
- an input memory and an output memory can be provided separately.
- the selector 12 is configured to be able to access these memories.
- One DFT operation processing group is called a stage, and is designated as a first stage, a second stage,..., A P stage from the input side.
- the number of stages P 2
- four 4-point DFT operations are performed at each stage.
- Cooley-Tukey-type fast Fourier transform can take two types of configurations, a time-thinning type and a frequency-thinning type, depending on how it is decomposed into two-point or four-point DFT.
- a fast Fourier transform which is a radix-4 frequency decimation type configuration, will be described as an example.
- FIG. 3 is a diagram for explaining the configuration of a four-point DFT in a radix-4 fast Fourier transform.
- the output complex data X ′ (m) is expressed as follows.
- j is an imaginary unit.
- FIG. 4 is a diagram for explaining a general example of storing complex data in a memory.
- Data is stored in one physical address of the memory (the contents of a plurality of addresses can be read continuously by designating one address).
- four complex data A i , A i + 1 , A i + 2 , and A i + 3 from the right are stored in one physical address B (i / 4). If it does not become complicated, the storage place is not particularly limited.
- description will be made assuming that the data is stored as shown in FIG.
- the following write address WA (S, k, m) and read address RA (S, k, m) will be described as addresses where individual complex data are stored.
- the write address WA (S, k, m) is calculated by the following formula.
- a mod b indicates that a remainder is obtained when a is divided by b.
- the converted data is rearranged in ascending order by using the address calculated in this way and taking out the address stored in the memory after the end of the final stage (P stage) by bit reverse.
- FIG. 6 is a diagram for explaining a method of calculating a write address to the memories 2A and 2B in the fast Fourier transform circuit shown in FIG.
- control unit 1 sets the write address and the read address of the memories 2A and 2B, and the 4-point DFT operation unit 1 performs the data number N to the power of the radix B.
- writing to the memories 2A and 2B is performed in the order of addresses that can be read from the memory at the same address.
- data for N / 4 addresses of the memories 2A and 2B can be stored in one bank.
- these banks are referred to as bank # 0 and bank # 1.
- the 4-point DFT processing is performed by the 4-point DFT calculation unit 1 using the data stored in the read buffer 3 in this way.
- the read buffer 3 holds data for N / 4 times of four-point DFT processing.
- the next four-point DFT processing N / 4 times of data are read from the memories 2A and 2B and stored in the other bank of the read buffer 3. Since four complex data are stored in one address of the memories 2A and 2B, data corresponding to one 4-point DFT process can be read from the memories 2A and 2B in one cycle. Since data for one 4-point DFT process can be read from the memories 2A and 2B simultaneously with the DFT process, the 4-point DFT process can be executed in one cycle.
- the address of the write buffer 4 is wn
- the data for N / 4 times of the 4-point DFT processing in the 4-point DFT operation unit 1 needs to be stored in the read buffer 3. For this reason, in the four cycles immediately after the start of each stage, the processing by the 4-point DFT operation unit 1 is not performed, and data necessary for the processing is read from the memories 2A and 2B and stored in the read buffer 3. After reading 4 ⁇ (N / 4) complex data, the bank of the read buffer 3 is switched, and data is read from the memories 2A and 2B and stored in the read buffer 3. At the same time, data is read from the other bank of the read buffer 3 and processing by the 4-point DFT operation unit 1 is executed.
- the write buffer 4 does not write to the memories 2A and 2B until the results of four times of the 4-point DFT processing are stored. After the start of processing, when the results for 4 points DFT processing N / 4 times are stored, the write buffer 4 switches the bank and stores the DFT processing result following the bank after switching. In parallel with this, the write buffer 4 writes the DFT processing result stored in the other bank into the memories 2A and 2B.
- the minimum execution cycle is N / 4 + 4 ⁇ 2 cycles.
- control unit 7 controls each unit and starts processing from the first stage.
- the selector 5 switches the connection so as to read from the memory 2A in which the input data is stored.
- FIG. 7 is a diagram showing a flow of processing for reading data from the memory 2A to be read and storing it in the read buffer 13.
- data required for calculation in the 4-point DFT calculation unit 1 is not complete until 4 words (16 complex data) are read from the memory 2A. For this reason, the calculation by the 4-point DFT calculation unit 1 is not performed, and only the data is stored in the read buffer 3.
- FIG. 8 to FIG. 11 are diagrams showing the flow of processing of reading data from the memory 2A, storing it in the read buffer 3, calculating by the 4-point DFT calculation unit 1, and storing the calculation result in the write buffer 4, respectively. is there.
- the 4-point DFT operation unit 1 sequentially performs 4-point DFT operations (FIG. 8).
- the calculation results z (1, 0) to z (1, 3) of the 4-point DFT operation are stored in the write buffer 4, and at the same time, the data necessary for the next 4-point DFT operation is read from the memory 2A word by word, and the read buffer 3 is stored. It is assumed that at least one word (4 complex data) is read from the memory 2A and stored in the read buffer 3 for one 4-point DFT operation.
- the read buffer 3 has a memory. Data x (4) to x (7) for at least one word from 2A is stored.
- FIG. 16 is a diagram showing a flow of a process of writing data from the write buffer 4 to the memory 2B when the first stage process is completed.
- N the number of 4-point DFT operations
- data z (1,48) to z (1,60) not yet written in the memory 2B is stored in the write buffer 4 for 4 words (16 complex). Data) is left. This data is sequentially written into the memory, and the first stage processing is completed (FIG. 16).
- FIG. 17 is a diagram for explaining the writing of data to the memory 2A in the second stage.
- FIG. 18 is a diagram for explaining the writing of data to the memory 2B in the third stage.
- the selector 5 switches the connection for each stage, reads data from the memory 2B or 2A in which the calculation result is stored in the previous stage, performs the calculation, and stores the result in the other memory 2A or 2B. Store. Therefore, in this case, data is written in the memory 2A in the second stage, and data is written in the memory 3A in the third stage.
- the method of writing the calculation result of the 4-point DFT to the write buffer 4 is different from the first stage.
- the data write address to the memory 2A or 2B is N / 4 PS + 1 interval according to the equation (3).
- the address (address for storing one complex data) interval is 4 in the second stage and 16 in the third stage.
- the read buffer 3 and the write buffer 3 and the write buffer can be obtained by devising a method for arranging complex data in the memories 2A and 2B and a method for writing the output of the 4-point DFT operation unit 1 to the memories 2A and 2B.
- a buffer 4 is introduced. This eliminates the need to unnecessarily divide the memories 2A and 2B even when performing a four-point DFT in one cycle.
- a decoding circuit and a test circuit added to the memories 2A and 2B are not required, and the circuit scale can be reduced. For example, when each of the memories 2A and 2B is divided into four, a decoding circuit and a test circuit for six memories are required, but according to the present embodiment, such a circuit can be reduced.
- the number of memories can be reduced to about 1/4 when performing physical placement at the time of LSI manufacture, and the layout becomes easy. Thereby, area saving (reduction of chip cost) can be realized.
- FIG. 19 is a block diagram showing a fast Fourier transform circuit according to the second embodiment of the present invention.
- the fast Fourier transform circuit shown in FIG. 19 includes memories 2A and 2B, a read buffer 3, a write buffer 4, a selector 5, a rotator generation unit 6, a 2-point / 4-point DFT calculation unit 11, and a control unit 12.
- the memories 2A and 2B, the read buffer 3, the write buffer 4, the selector 5, and the rotator generation unit 6 are the same as those shown in FIG.
- the 2-point / 4-point DFT calculation unit 11 can execute one 4-point DFT calculation or two 2-point DFT calculations in one cycle. The control is performed by the control unit 12.
- FIG. 20 is a diagram illustrating a calculation configuration example of the 2-point / 4-point DFT calculation unit 11 in the fast Fourier transform circuit illustrated in FIG.
- the first stage is composed of two two-point DFTs, and the second and subsequent stages are composed of four-point DFT operations.
- FIG. 21 is a diagram showing a configuration diagram of a two-point DFT used in the first stage processing.
- the outputs X ′ (m) of the two two-point DFTs are expressed as follows.
- X ′ (m) and x ′ (m) are complex numbers, and j is an imaginary unit.
- the value obtained by dividing the number of data N by the value P ⁇ S + 1 to the power of m, and the value obtained by dividing the value of B by the power of S ⁇ 1 and dividing the value twice the order k by the power of B Is set to a value obtained by multiplying the value obtained by multiplying the half of the value of B by the value obtained by dividing the value of k in order by the half of the value obtained by multiplying B to the S-1 power.
- the control unit 12 is configured to multiply the read address RA (S, k, m) from the memories 2A and 2B in each of all P stages by the value obtained by dividing the number of data N by the value of B and m, A value obtained by adding the value of the order k is set.
- the write address is as follows, unlike Equation (3).
- the read address is the same as in equation (4).
- the storage address of the data read from the memories 2A and 2B in the read buffer 3 is the same as that in Expression (6).
- the write address to the write buffer 4 varies depending on the processing stage as follows. a) Write address to the write buffer 4 in the first stage b) Write address to the write buffer in the second stage c) Write address to the write buffer in the third stage and thereafter The same as the equation (8).
- FIG. 26 is a diagram showing a flow of processing for writing data from the write buffer 4 to the memory 2B in the first stage.
- the word stored in the write buffer 4 is stored in the memory 2B in one word in parallel with the writing of the results of the fifth to eighth 2-point DFT operations by the 2-point / 4-point DFT operation unit 11 into the write buffer 4.
- Write (4 complex data) one by one. After the eight 2-point DFT operations are completed, the data stored in the write buffer is written into the memory B. At this time, the write address interval to the memory 2B is 1.
- FIGS. 22 to 26 are diagrams equivalent to FIGS. 22 to 26 for explaining the operation in the second stage, respectively. Since the operation of the second stage is the same as the operation of the first stage described with reference to FIGS. 22 to 26 except that the write address to the write buffer 4 is different, the description is omitted.
- FIG. 32 is a diagram showing a flow of processing of storing the calculation result of the 2-point / 4-point DFT calculation unit 11 in the write buffer 4 and writing data from the write buffer 4 to the memory 2B in the third stage.
- the operation in this case is the same as the operation of the first embodiment described with reference to FIGS. 8 to 16 except that the interval of the write address to the memory 2B is expressed by the following equation. Is omitted.
- frequency decimation type fast Fourier transform has been described as an example, but the present invention can also be implemented by time decimation type fast Fourier transform.
- orthogonal frequency division multiplexing which is a radio access method with high frequency utilization efficiency, has been used to improve communication speed.
- wireless LAN Local Area Network
- mobile communication LTE (Long Term Evolution), which is being standardized with a new communication method in 3GPP (Third Generation Partnership Project), and OFDM. Is adopted.
- FIG. 33 is a diagram illustrating an example of the overall configuration of a communication system using the OFDM method. Here, for the sake of simplicity, only the configuration related to one-way communication is shown.
- This communication system includes a base station 101.
- the base station 101 includes an encoder 102, a modulator 103, an OFDM signal generator, a D / A converter 105, and a plurality of antennas 106 (only one is shown in the figure).
- the communication system also includes a mobile station 111 such as a user terminal that communicates with the base station 101.
- the mobile station 111 includes a decoding unit 112, a demodulation unit 113, an OFDM signal demodulation unit 114, an A / D conversion unit 115, and a plurality of antennas 116 (only one is shown in the figure).
- a CPU (not shown) of the base station 101 inputs data to be transmitted to the encoding unit 102 as information bits.
- the encoding unit 102 performs CRC (Cyclic Redundancy Check) addition and convolutional encoding on the input information bits.
- the modulation unit 103 modulates the input code data.
- the OFDM signal generation unit 104 maps the modulated data on the frequency axis, and converts the data on the frequency axis into data on the time axis by inverse digital Fourier transform.
- the converted data is output to the D / A converter 105.
- the D / A conversion unit 105 converts the digital signal output from the OFDM signal generation unit 104 into an analog signal. Then, the modulated data converted into the analog signal is transmitted via the plurality of antennas 106.
- the mobile station 111 on the receiving side receives data transmitted from the antenna 106 on the transmitting side 101 via the plurality of antennas 116.
- the data received by the antenna 116 is affected by noise when propagating through space after being output from the antenna 106.
- Data received by the antenna 116 is input to the A / D converter 115.
- the A / D converter 115 converts input data from an analog signal to a digital signal.
- a / D conversion section 115 outputs the converted digital signal to OFDM signal demodulation section 114.
- the OFDM signal demodulator 114 converts the digital signal on the time axis output from the A / D converter 115 into data on the frequency axis by digital Fourier transform and maps it on the IQ plane.
- the demodulator 113 demodulates the data output from the OFDM signal demodulator 114.
- Demodulation section 113 outputs the demodulated data obtained by demodulation to decoding section 112.
- the decoding unit 112 performs error correction decoding on the input demodulated data. Using the decoded data obtained as a result, a processing circuit such as a CPU at a subsequent stage performs a predetermined process.
- DFT digital Fourier transform
- LTE the number of subcarriers in the 20 MHz band is 1200.
- DFT it is necessary to perform 1,199 complex multiplications and complex additions 1200 times. Moreover, this must be done within 66.67 microseconds of one OFDM symbol period.
- Cooley-Tukey type fast Fourier transform As an algorithm for reducing the amount of calculation, for example, the above-described Cooley-Tukey type fast Fourier transform is used.
- the amount of computation in the case of 1200 subcarriers is 2048 complex multiplications and complex additions + (3 complex additions and 1 complex multiplication 2048 times) ⁇ 5 times It becomes.
- the number of data needs to be a power of 2. Therefore, 2048 which is larger than 1200 and the smallest power of 2 is the number of data.
- the rotator that is subject to complex multiplication may be represented by only the real part or only the imaginary part, so the number of multiplications may be smaller. However, there is still no change in the amount of computation.
- the amount of calculation can be reduced by performing a fast Fourier transform with a 4-point DFT as a component.
- input / output data is generally stored in a memory.
- the four-point DFT it is necessary to take out four complex data serving as inputs from the memory.
- the memory is a single port, four cycles are required.
- the former increases the area of the memory cell itself, and the latter leads to an increase in area due to the test circuit or decoder circuit associated with the memory macro.
- dividing the memory increases the number of wirings by four times, increasing the difficulty of layout design at the time of manufacturing the LSI, and may increase the LSI area so that wiring is possible.
- the present invention can be used for a measuring apparatus that performs Fourier transform, such as a spectrum analyzer.
Landscapes
- Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Discrete Mathematics (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Complex Calculations (AREA)
Abstract
Description
図1は本発明の第1の実施の形態に係る高速フーリエ変換回路を示すブロック構成図である。ここでは、高速フーリエ変換の基数B=4の場合の構成例を示す。
図2は、図1に示す高速フーリエ変換回路中の4点DFT演算部1の動作を説明するシグナルフロー図である。ここでは、基数B=4、データ数N=16の場合を示す。
図4は、複素データのメモリへの一般的な格納例を説明する図である。1複素データを1アドレスに格納した場合のアドレスをAi(i=0,1,…,(N-1)、Nはデータ数(FFTポイント数))とするとき、連続する4個の複素データをメモリの1物理アドレス(ひとつのアドレスを指定することで、複数のアドレスの内容を連続的に読み出すことができる)に格納する。図4に示す例では、右からAi,Ai+1,Ai+2,Ai+3の4複素データを1物理アドレスB(i/4)に格納しているが、回路構成が複雑にならなければ、格納する場所は特に限定されるものではない。以下では、簡単のため、図4に示すとおりに格納されるものとして説明する。
図5は、メモリへの書き込みアドレスの従来の計算方法を説明する図である。ここでは、基数B=4に対してデータ数N=16の場合を示す。
リードバッファ3およびライトバッファ4は、それぞれともに2バンク構成で、1バンク当たり、N個の複素データを格納できる容量を有し、基数B=4で4複素データをメモリ2A、2Bの1アドレスに格納する場合に、1バンクにメモリ2A、2BのN/4アドレス分のデータを格納できる構成となっている。説明のため、これらのバンクをバンク#0、バンク#1とする
(a)S=1のとき
図7から図18は、図1に示す高速フーリエ変換回路によるデータの処理の流れを説明する図である。これらの図を参照して、高速フーリエ変換回路の動作を詳細に説明する。ここでは、N=64の場合を例に説明する。以下の動作は、制御部7が各部を制御することにより実行される。処理の対象となる入力データは、メモリ2Aに格納されているものとする。
以上説明した実施の形態においては、複素データのメモリ2A、2Bへの配置方法と、4点DFT演算部1の出力のメモリ2A、2Bへの書き込み方法を工夫することで、リードバッファ3およびライトバッファ4を導入している。これによって、1サイクルで4点DFTを行う場合においても、メモリ2A、2Bを不必要に分割する必要がなくなる。メモリ2A、2Bに付加するデコード回路やテスト回路が不要となり、回路規模削減が可能となる。たとえば、メモリ2A、2Bをそれぞれ4分割した場合、メモリ6個分のデコード回路とテスト回路が必要となるが、本実施の形態によれば、そのような回路を削減することが可能となる。
図19は、本発明の第2の実施の形態に係る高速フーリエ変換回路を示すブロック構成図である。この実施の形態の基本的な構成は、第1の実施の形態と同等である。ただし、B/2=2のべき乗ではあるが、B=4のべき乗ではないデータ数N(log2Nが奇数)の高速フーリエ変換処理を実行できる構成となっていることが異なる。
a)第1ステージにおけるライトバッファ4への書き込みアドレス
式(8)と同じ。
これらのアドレスを用いた動作例を図22から図32を参照して説明する。以下の説明では、データ数N=32とする。
2048回の複素乗算と複素加算+(3回の複素加算と1回の複素乗算を2048回)×5回
となる。ここで、クーリー・チューキー型の高速フーリエ変換ではデータ数が2のべき乗であることが必要であるため、1200より大きく、かつ最も小さな2のべき乗である2048がデータ数となる。
2A、2B メモリ
3 リードバッファ
4 ライトバッファ
5 セレクタ
6 回転子生成部
7、12 制御部
11 2点/4点DFT演算部
Claims (8)
- 複数のデジタルフーリエ変換を構成要素とする高速フーリエ演算を行う演算部と、
上記演算部の入出力データを格納するメモリと、
処理対象のデータに対して上記演算部が行う複数ステージの演算に対して、上記メモリからのデータの読み出し順が各ステージで同じとなるように、上記演算部による演算結果の上記メモリへの書き込みを制御する手段と
を有することを特徴とする高速フーリエ変換回路。 - 請求項1記載の高速フーリエ変換回路において、
前記演算部は、基数B、ただしBは2以上の整数、のデジタルフーリエ変換を構成要素とし、
前記制御する手段は、データ数Nが上記基数Bのべき数のデータに対して前記演算部が行うP=logBNステージの演算に対して、次のステージでの演算対象となるデータが前記高速フーリエ演算の順番k=0,1,・・,N/B-1に対してデータ数Nを上記基数Bで割った値ごとに同じアドレスでメモリから読み出すことのできるアドレス順に、前記メモリへの書き込みを行う
ことを特徴とする高速フーリエ変換回路。 - 請求項2記載の高速フーリエ変換回路において、
前記制御する手段は、第S=1,2,・・,Pステージのそれぞれにおける前記メモリへの書き込みアドレスWA(S,k,m)、ただしm=0,1,・・,B-1、を、前記基数Bを残りステージ数に1を加えた値P-S+1乗した値で前記データ数Nを割った値にmを乗じた値と、前記基数BをS-1乗した値で前記順番kの値を割った値に前記基数BのS乗を乗じた値と、前記基数BをS-1乗した値で前記順番kの値を割っときの余りの値とを加算した値に設定し、
第S=1,2,・・,Pステージのそれぞれにおける前記メモリからの読み出しアドレスRA(S,k,m)を、前記基数Bで前記データ数Nを割った値にmを乗じた値と、前記順番kの値とを加算した値に設定する
ことを特徴とする高速フーリエ変換回路。 - 請求項1記載の高速フーリエ変換回路において、
前記演算部は、4以上の複数B点またはB/2点のいずれかのデジタルフーリエ変換を1サイクルで実行する構成であり、
データ数Nが上記B/2のべき乗であるが上記Bのべき乗ではないデータに対して、前記演算部による演算が、B/2点のデジタルフーリエ変換を2組実行するひとつのステージと、B点のデジタルフーリエ変換を実行するlogB(N/2)ステージとに分解され、
前記制御する手段は、上記ひとつのステージおよび上記logB(N/2)ステージのそれぞれにおいて、次のステージでの演算対象となるデータが前記高速フーリエ演算の順番k=0,1,・・,N/B-1に対してデータ数Nを上記Bの値で割った値ごとに同じアドレスでメモリから読み出すことのできるアドレス順に、前記メモリへの書き込みを行う
ことを特徴とする高速フーリエ変換回路。 - 請求項4記載の高速フーリエ変換回路において、
前記制御する手段は、
前記ひとつのステージにおける前記メモリへの書き込みアドレスWA(S,k,m)、ただしm=0,1,・・,B-1、を、m=0,1,..,B/2-1のときにはkの2倍にmを加算した値、m=B/2,..,B-1のときにはN/2とkの2倍とm-B/2とを加算した値に設定し、
前記logB(N/2)ステージのそれぞれにおける前記メモリへの書き込みアドレスWA(S,k,m)を、前記Bの値を残りステージ数に1を加えた値P-S+1乗した値で前記データ数Nを割った値にmを乗じた値と、前記Bの値をS-1乗した値で前記順番kの2倍の値を割った値に前記基数BのS乗の半分の値を乗じた値と、前記基数BをS-1乗した値の半分の値で前記順番kの値を割ったときの余りの値とを加算した値に設定し、
前記ひとつのステージおよびlogB(N/2)ステージのそれぞれにおける前記メモリからの読み出しアドレスRA(S,k,m)を、前記Bの値で前記データ数Nを割った値にmを乗じた値と、前記順番kの値とを加算した値に設定する
ことを特徴とする高速フーリエ変換回路。 - 請求項1から5のいずれか1項に記載の高速フーリエ変換回路において、
前記メモリから読み出されたデータを格納し、前記演算部に出力するリードバッファと、
前記演算部による演算結果を格納し、前記メモリに書き込むライトバッファと
をさらに有する
ことを特徴とする高速フーリエ変換回路。 - 請求項6記載の高速フーリエ変換回路において、
前記演算部の構成要素である前記複数のデジタルフーリエ変換を、前記リードバッファおよび前記ライトバッファの1サイクルのデータ転送で実行する
ことを特徴とする高速フーリエ変換回路。 - 請求項2から5のいずれか1項に記載の高速フーリエ変換回路において、前記Bの値は4であることを特徴とする高速フーリエ変換回路。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US13/579,184 US20130046806A1 (en) | 2010-02-16 | 2011-02-10 | Fast fourier transform circuit |
EP11744578A EP2538345A1 (en) | 2010-02-16 | 2011-02-10 | Fast fourier transform circuit |
JP2012500571A JPWO2011102291A1 (ja) | 2010-02-16 | 2011-02-10 | 高速フーリエ変換回路 |
CN2011800097694A CN102763101A (zh) | 2010-02-16 | 2011-02-10 | 快速傅里叶变换电路 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2010030796 | 2010-02-16 | ||
JP2010-030796 | 2010-02-16 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2011102291A1 true WO2011102291A1 (ja) | 2011-08-25 |
Family
ID=44482878
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2011/052844 WO2011102291A1 (ja) | 2010-02-16 | 2011-02-10 | 高速フーリエ変換回路 |
Country Status (6)
Country | Link |
---|---|
US (1) | US20130046806A1 (ja) |
EP (1) | EP2538345A1 (ja) |
JP (1) | JPWO2011102291A1 (ja) |
CN (1) | CN102763101A (ja) |
TW (1) | TW201209602A (ja) |
WO (1) | WO2011102291A1 (ja) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013097236A1 (zh) * | 2011-12-31 | 2013-07-04 | 中国科学院自动化研究所 | 多粒度并行fft计算装置 |
WO2017053986A1 (en) | 2015-09-25 | 2017-03-30 | Microsoft Technology Licensing, Llc | Method and system for approximate quantum circuit synthesis using quaternion algebra |
WO2016164325A1 (en) * | 2015-04-10 | 2016-10-13 | Microsoft Technology Licensing, Llc | Method and system for quantum circuit synthesis using quaternion algebra |
CN113569189B (zh) * | 2021-07-02 | 2024-03-15 | 星思连接(上海)半导体有限公司 | 一种快速傅立叶变换计算方法及装置 |
TWI863789B (zh) * | 2024-01-02 | 2024-11-21 | 財團法人工業技術研究院 | 數論轉換運算電路 |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS62273440A (ja) * | 1986-05-21 | 1987-11-27 | Hitachi Ltd | 核磁気共鳴装置 |
JPS63292267A (ja) * | 1987-05-25 | 1988-11-29 | Nippon Telegr & Teleph Corp <Ntt> | 高速フ−リエ変換用アドレス生成回路 |
JPH11110370A (ja) * | 1997-01-22 | 1999-04-23 | Matsushita Electric Ind Co Ltd | 高速フーリエ変換装置および方法、可変ビットリバース回路、逆高速フーリエ変換装置および方法、並びにofdm受信および送信装置 |
JP2004166083A (ja) * | 2002-11-14 | 2004-06-10 | Matsushita Electric Ind Co Ltd | 符号化装置及び方法 |
JP2006155487A (ja) | 2004-12-01 | 2006-06-15 | Sony Corp | 高速フーリエ変換処理装置および演算処理用制御装置 |
JP2008186396A (ja) * | 2007-01-31 | 2008-08-14 | Mitsubishi Electric Corp | 高速フーリエ変換装置 |
JP2008533873A (ja) * | 2005-03-11 | 2008-08-21 | クゥアルコム・インコーポレイテッド | 高速フーリエ変換トゥイドル乗算 |
JP2008217359A (ja) * | 2007-03-02 | 2008-09-18 | Fujitsu Ltd | 高速フーリエ変換装置及び高速フーリエ変換処理方法 |
JP2009095008A (ja) * | 2007-09-20 | 2009-04-30 | Mitsubishi Electric Corp | ターボ符号復号装置、ターボ符号復号方法及び通信システム |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6081821A (en) * | 1993-08-05 | 2000-06-27 | The Mitre Corporation | Pipelined, high-precision fast fourier transform processor |
WO2002091221A2 (en) * | 2001-05-07 | 2002-11-14 | Jaber Associates, L.L.C. | Address generator for fast fourier transform processor |
TWI298448B (en) * | 2005-05-05 | 2008-07-01 | Ind Tech Res Inst | Memory-based fast fourier transformer (fft) |
-
2011
- 2011-02-10 US US13/579,184 patent/US20130046806A1/en not_active Abandoned
- 2011-02-10 EP EP11744578A patent/EP2538345A1/en not_active Withdrawn
- 2011-02-10 CN CN2011800097694A patent/CN102763101A/zh active Pending
- 2011-02-10 JP JP2012500571A patent/JPWO2011102291A1/ja not_active Withdrawn
- 2011-02-10 WO PCT/JP2011/052844 patent/WO2011102291A1/ja active Application Filing
- 2011-02-16 TW TW100105116A patent/TW201209602A/zh unknown
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS62273440A (ja) * | 1986-05-21 | 1987-11-27 | Hitachi Ltd | 核磁気共鳴装置 |
JPS63292267A (ja) * | 1987-05-25 | 1988-11-29 | Nippon Telegr & Teleph Corp <Ntt> | 高速フ−リエ変換用アドレス生成回路 |
JPH11110370A (ja) * | 1997-01-22 | 1999-04-23 | Matsushita Electric Ind Co Ltd | 高速フーリエ変換装置および方法、可変ビットリバース回路、逆高速フーリエ変換装置および方法、並びにofdm受信および送信装置 |
JP2004166083A (ja) * | 2002-11-14 | 2004-06-10 | Matsushita Electric Ind Co Ltd | 符号化装置及び方法 |
JP2006155487A (ja) | 2004-12-01 | 2006-06-15 | Sony Corp | 高速フーリエ変換処理装置および演算処理用制御装置 |
JP2008533873A (ja) * | 2005-03-11 | 2008-08-21 | クゥアルコム・インコーポレイテッド | 高速フーリエ変換トゥイドル乗算 |
JP2008186396A (ja) * | 2007-01-31 | 2008-08-14 | Mitsubishi Electric Corp | 高速フーリエ変換装置 |
JP2008217359A (ja) * | 2007-03-02 | 2008-09-18 | Fujitsu Ltd | 高速フーリエ変換装置及び高速フーリエ変換処理方法 |
JP2009095008A (ja) * | 2007-09-20 | 2009-04-30 | Mitsubishi Electric Corp | ターボ符号復号装置、ターボ符号復号方法及び通信システム |
Also Published As
Publication number | Publication date |
---|---|
CN102763101A (zh) | 2012-10-31 |
JPWO2011102291A1 (ja) | 2013-06-17 |
EP2538345A1 (en) | 2012-12-26 |
TW201209602A (en) | 2012-03-01 |
US20130046806A1 (en) | 2013-02-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Lin et al. | A dynamic scaling FFT processor for DVB-T applications | |
EP0855657B1 (en) | Fast fourier transforming apparatus and method | |
JP2009535678A (ja) | パイプラインfftのアーキテクチャおよび方法 | |
CN112100568B (zh) | 定点傅里叶变换fft处理器及处理方法 | |
CN112800386B (zh) | 傅里叶变换处理方法和处理器、终端、芯片及存储介质 | |
WO2011102291A1 (ja) | 高速フーリエ変換回路 | |
US9250996B2 (en) | Multicore type error correction processing system and error correction processing apparatus | |
CN111737638A (zh) | 基于傅里叶变换的数据处理方法及相关装置 | |
CN101937423B (zh) | 一种流水式fft/ifft的处理系统 | |
KR100989797B1 (ko) | Fft/ifft 연산코어 | |
US9727531B2 (en) | Fast fourier transform circuit, fast fourier transform processing method, and program recording medium | |
US8838661B2 (en) | Radix-8 fixed-point FFT logic circuit characterized by preservation of square root-i operation | |
CN115544438A (zh) | 数字通信系统中的旋转因子生成方法、装置和计算机设备 | |
US20140365547A1 (en) | Mixed-radix pipelined fft processor and fft processing method using the same | |
EP2144173A1 (en) | Hardware architecture to compute different sizes of DFT | |
CN101938442B (zh) | Dft处理器的预检测基运算方法、混合基运算方法和系统 | |
US20140219374A1 (en) | Efficient multiply-accumulate processor for software defined radio | |
JP3065979B2 (ja) | 高速フーリエ変換装置および方法、可変ビットリバース回路、逆高速フーリエ変換装置および方法、並びにofdm受信および送信装置 | |
Lin et al. | Low-cost FFT processor for DVB-T2 applications | |
Yang et al. | A novel design of pipeline MDC-FFT processor based on various memory access mechanism | |
CN113094639B (zh) | 一种dft并行处理方法、装置、设备及存储介质 | |
KR20120101807A (ko) | 메모리 기반 고속 푸리에 변환 프로세서 | |
Vishwanath | Efficient Hardware Architecture for Ultra-High Sampling Rate FFT Analysis of Acoustic Emission Signals | |
Locharla et al. | Implementation of input data buffering and scheduling methodology for 8 parallel MDC FFT | |
Jung et al. | Low-complexity multi-mode memory-based FFT processor for DVB-T2 applications |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 201180009769.4 Country of ref document: CN |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 11744578 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2012500571 Country of ref document: JP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 2011744578 Country of ref document: EP |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
WWE | Wipo information: entry into national phase |
Ref document number: 7936/CHENP/2012 Country of ref document: IN |
|
WWE | Wipo information: entry into national phase |
Ref document number: 13579184 Country of ref document: US |