JP4036233B2 - Musical sound generating device, musical sound generating method, and storage medium storing a program related to the method - Google Patents

Description

  The present invention relates to a musical sound generating apparatus and a musical sound generating method for generating a musical sound by executing predetermined musical sound generating software on an arithmetic processing unit, and a storage medium storing a program related to the method.

  2. Description of the Related Art Conventionally, a musical sound generator that generates a musical sound by executing predetermined musical sound generation software on a general-purpose arithmetic processing device such as a CPU is known. This is what is called a software sound source or a software sound source. Some soft tone generators also use software (soft effector) to apply an effect (effect) such as a reverberation effect (reverb) to the generated musical sound and output it. In recent years, there is a demand for higher functionality in software sound sources, and there is a demand for imparting various effects.

  In the software sound source, a waveform generation buffer area is provided in order to generate a plurality of samples collectively when generating a musical sound by software. FIG. 3B shows an example of such a waveform generation buffer area. In FIG. 3B, 1, 2,..., 128 are storage areas for 128 sets of waveform samples (which are time-series data arranged in a temporal order). One set of waveform sample storage areas includes DryL, DryR, and Rev. DryL is a waveform sample storage area without stereo left reverb, DryR is a waveform sample storage area without stereo right reverb, and Rev is a waveform sample storage area for input to the reverb. In short, (DryL, DryR, Rev) is used as one unit to have waveform samples in an interleaved format. This is because it is convenient to arrange the output data of each channel in this order in the waveform calculation part.

  In the software sound source, for example, for each frame at a predetermined time interval, waveform samples (128 samples) for one frame of all channels for which musical sounds are to be generated are generated and accumulated in the waveform generation buffer shown in FIG. To calculate the waveform data to be generated. First, 128 samples of the first channel are generated and weighted (by multiplying each value of DryL, DryR, Rev of each sample by a predetermined coefficient) and stored in the waveform generation buffer of FIG. Next, 128 samples of the second channel are generated and weighted and accumulated in the waveform generating buffer of FIG. 3B, and then 128 samples of the third channel are generated and weighted to generate the waveform of FIG. 3B. It accumulates in the generation buffer, and so on. As a result, the accumulated results of the waveform samples of the channel to be sounded are obtained for 128 samples. The generated waveform data is given to, for example, a DMAC (Direct Memory Access Controller) and passed to a sound I / O (an LSI called a codec (CODEC) that executes input / output of musical sound waveform data), and the sound I / O is digital analog (D / A) The sound is converted and emitted by the sound system.

  By the way, there is a demand for realizing various effects even with a soft sound source, but usually the calculation order (so-called soft effector connection relations) when performing multiple effects is not dynamically changeable. There is.

An arithmetic processing unit used for a soft sound source may be one having an internal or external cache memory. However, when the data structure of the waveform generation buffer as shown in FIG. There is a problem that a cache miss tends to occur during calculation processing by a soft effector. For example, in the example of FIG. 3B, the soft effector that performs reverberation calculation takes out 128 samples of Rev stored in an interleaved manner and performs calculation processing, and a cache miss is likely to occur. When the effect to be applied is only reverb, such an overhead does not occur. However, when the effect to be applied increases, a cache miss is particularly likely to occur. For example, when three types of effects (reverb, chorus, variation) are applied and the output is 7 lines, the data structure of FIG. 3B is expanded (DryL, DryR, Rev, ChorusL, ChorusR, VariationL, VariationR). If the buffer for waveform generation arranged for 128 samples as one unit is used, the effector
(1) Variation L and Variation R are collected for 128 samples and variation calculation processing is performed (2) Chorus L and Chorus R are collected for 128 samples and chorus calculation processing is performed (3) Rev is calculated for 128 samples and reverb calculation processing is performed Are calculated in the order of execution, the jumping data area on the waveform generation buffer area must be accessed frequently and frequently, resulting in frequent cache misses and very low efficiency.

  An object of the present invention is to provide a musical sound generating apparatus and a musical sound generating method capable of dynamically changing the calculation order of the effect applying process in a software sound source, and a storage medium storing a program related to the method. .

  The present invention is characterized in that a waveform buffer is prepared for each effect process. The processing executed in the arithmetic processing unit by each effect processing software is to perform predetermined effect processing on a plurality of waveform samples in the corresponding waveform buffer, and write the processed waveform samples to the corresponding waveform buffer. In addition, an add process is provided in which any one waveform sample in the waveform buffer is subjected to a predetermined level control and added to the waveform sample stored in another arbitrary waveform buffer and written again to the same storage position. Regarding the effect processing corresponding to each waveform buffer executed on each waveform buffer and the add processing executed between any two waveform buffers, the processing order of the effect processing and the add processing according to the desired effect algorithm The waveform buffer to be added and the waveform buffer to be added in the add process can be specified.

  Further, the apparatus further includes a specifying unit for the user to specify an algorithm indicating the connection state of the plurality of effects, and the processing order of the plurality of effect processes and the plurality of add processes is changed according to the algorithm set by the specifying unit. To be able to.

  As processing timing, a timing signal is generated at a predetermined time interval, and a predetermined number of samples corresponding to performance information is generated every time the timing signal is generated, and a plurality of buffers respectively corresponding to a plurality of effect processes are generated. By executing the output processing for a plurality of channels, waveform data accumulated in the channels is generated in each of the plurality of buffers. After preparing waveform data corresponding to performance information in a plurality of buffers corresponding to a plurality of effect processes in this way, the above-described effect processing and add processing are executed in a processing order corresponding to a designated effect algorithm, Generate and output waveform data with effect processing. The output waveform data is supplied to the digital-analog converter by one sample every sampling period and converted into an analog waveform.

  As described above, according to the present invention, since the waveform buffer is prepared for each effect process, the connection between a plurality of effects can be easily changed. In addition, since the effect processing is performed on the corresponding buffer, and any buffer can be accumulated in any buffer by the add processing, even if the user arbitrarily designates the soft effector algorithm, Thus, it is only necessary to change the order of the effect processing and the add processing. Therefore, the calculation order of the effect applying process can be dynamically changed according to the user's specification. In addition, waveform sample generation and effect application (effect processing and add processing are performed in a processing order corresponding to a specified effect algorithm) are performed on a buffer prepared for each effect processing. Therefore, the connection between multiple effects can be changed easily.

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

  FIG. 1 is a block diagram of an electronic musical instrument to which a musical tone generator according to the present invention is applied. The electronic musical instrument includes a central processing unit (CPU) 101, a read only memory (ROM) 102, a random access memory (RAM) 103, a drive device 104, a timer 106, a network input / output (I / O) interface 107, a keyboard 108, A display 109, a hard disk 110, a sampling clock (Fs) generator 111, a sound I / O 112, a DMA (direct memory access) controller 114, a sound system 115, and a bus line 116 are provided.

  The CPU 101 controls the operation of the entire electronic musical instrument. The CPU 101 includes a cache memory 117 inside. The cache line (cache block) size of the cache memory 117 is 32 bytes. That is, when the CPU 101 reads data at an appropriate address from the ROM 102, RAM 103, etc., consecutive 32 bytes including the address are copied to a predetermined cache line of the cache memory 117. Thereafter, when any of the 32-byte data is read, the cache line data is supplied instead of reading from the ROM 102 or RAM 103. Since the cache memory 117 is accessed at a very high speed, it can be processed at a very high speed while the data is on the cache memory 117. The cache memory includes a write-through method and a write-back method. The cache memory 117 used here is a write-through method.

  The ROM 102 stores control programs executed by the CPU 101 (software tone generator programs including software effectors) and various constant data. On the ROM 102, waveform data (waveform sample data sampled at a predetermined rate) that is used when the CPU 101 executes a software tone generator program to generate a musical tone is prepared. Note that the control program, various constant data, and waveform data prepared on the RAM 103 can be used instead of the ROM 102. In this case, data such as a control program is supplied from an external storage medium such as a CD-ROM or the network I / O 107 and is expanded on the RAM 103 or stored in the hard disk 110. The RAM 103 is provided with work areas such as various registers, a waveform generation buffer, and a reproduction buffer. The drive device 104 is for inputting / outputting various data to / from an external storage medium 105 such as an FD (flexible disk) or a flash card. The hard disk 110 is a storage device used for recording various data.

  The timer 106 generates a timer clock signal for causing the CPU 101 to generate a timer interrupt at a predetermined interval. The network I / O interface 107 is an interface for exchanging various data via an external public line or a LAN (local area network). The keyboard 108 is a typing keyboard used for inputting various information. The display 109 is a display that displays various types of information. A user can perform various setting operations and instruction operations on the electronic musical instrument using the keyboard 108 and the display 109.

  The Fs generator 111 generates a sampling clock having a frequency Fs to be given to the sound I / O 112. The sound I / O 112 is composed of an LSI called a codec (CODEC). The sound I / O 112 has an analog / digital (A / D) conversion function and a digital / analog (D / A) conversion function. An analog musical tone signal from the external input 113 is input to the A / D input terminal. A sound system 115 is connected to the / A output terminal. The sound I / O 112 includes two FIFO (First In First Out) type stack areas. One is an input FIFO for holding digital waveform data input via the A / D input terminal, and the other is an output FIFO for holding digital waveform data output via the D / A output terminal.

  An analog musical sound signal input from the external input 113 to the A / D input terminal of the sound I / O 112 is A / D converted in accordance with a sampling clock having a frequency Fs, and may be compressed by an ADPCM method or the like as necessary. ) Written to input FIFO. If the input FIFO contains waveform data, the sound I / O 112 issues a request to the DMA controller 114 to process the input waveform data. The DMA controller 114 transfers the data of the input FIFO to a predetermined area on the RAM 103 (a recording buffer area secured in advance) in response to the processing request. Data transfer by the DMA controller 114 is performed by the DMA controller 114 interrupting the CPU 101 (hardware interrupt) for each sampling clock Fs and securing the bus line 116. The CPU 101 is not aware of securing the bus line 116 by the DMA controller 114.

  On the other hand, when waveform data exists in the output FIFO in the sound I / O 112, the waveform data in the output FIFO is D / A converted for each sampling clock Fs and sent to the sound system 115 via the D / A output terminal. And sound is emitted. When the output FIFO waveform data is output, the output FIFO becomes empty, and the sound I / O 112 issues a waveform data acquisition request to the DMA controller 114. The CPU 101 generates waveform data to be output in advance and places it in a reproduction buffer (PB0, PB1 described later) on the RAM 103, and issues a request for reproducing the waveform to the DMA controller 114. The DMA controller 114 interrupts the CPU 101 for each sampling clock Fs to secure the bus line 116 and transfers the waveform data of the reproduction buffer on the RAM 103 to the output FIFO of the sound I / O 112. The transfer of waveform data by the DMA controller 114 is not noticed by the CPU 101. As described above, the waveform data written in the output FIFO is sent to the sound system 115 for each sampling clock Fs and emitted.

  The software tone generator is realized by the CPU 101 executing the tone generation software in the ROM 102. From the viewpoint of using the software tone generator, the tone generator software is registered as a driver, the driver is started, The procedure is such that a MIDI (Musical Instruments digital Interface) event message indicating various performance inputs is output to an API (Application Program Interface) relating to the software sound source, and various processes relating to musical tone generation can be performed on the software sound source. The CPU 101 is a general-purpose arithmetic processing unit, and performs processing different from the software sound source, such as processing for giving performance input to the API, that is, processing for outputting a MIDI event to the API. The process in which the CPU 101 gives a performance input to the API is, for example, a performance input that is generated in real time in response to an operation of a performance operator (not shown) such as a keyboard, or the network I / O 107. A performance input corresponding to a MIDI event input in real time via the interface is output to the API, or a sequence of MIDI events is prepared on the RAM 103 (even if data on the external storage medium 105 or the hard disk 110 is used). (Alternatively, data input via the network I / O 107 may be used.) This is a process of sequentially outputting the data as performance input to the API.

  With reference to FIG. 2, the tone generation principle of the soft sound source will be described. In FIG. 2, each section of S1 to S4 indicates a time frame that is a unit for performing reproduction for a predetermined number of samples (2 × 128 samples). A downward arrow written on the “performance input” line indicates that there was a performance input at that time. The performance input means that various MIDI events such as note-on, note-off, after-touch, and program change are input to the above-described API related to the software sound source. In the example of FIG. 2, there are three performance inputs for the frame S1, two for S2, and one for S3. The software sound source can simultaneously generate a plurality of musical sounds for a plurality of channels, and each musical sound is controlled by a software sound source register for a plurality of channels prepared on the RAM 103. When a note-on event is input as a performance input, the software tone generator assigns sound generation to the software tone generator register and writes various data and note-on for controlling the sound generation of the channel to the software tone generator register corresponding to the assigned channel. . When a note-off event is input as a performance input, note-off is written in the soft tone generator register corresponding to the corresponding channel. Similarly, for performance inputs other than note-on and note-off (for example, after-touch change), data corresponding to the performance input is written in the software tone generator register corresponding to the corresponding channel. Data written to the soft tone generator register in a certain time frame is always used for waveform generation calculation from the next time frame regardless of the type of data.

  The rectangles 201 to 204 for “waveform generation by the CPU” in FIG. 2 indicate sections in which the CPU 101 executes waveform generation calculation including effect provision. In this waveform generation calculation, musical sound waveform generation for a plurality of channels is performed based on data for a plurality of channels set in the software sound source register. The soft tone generator register is rewritten in accordance with the performance input. On the other hand, during the period when there is no performance input, the soft tone generator register holds data written in the past. Therefore, in each waveform generation section 201 to 204, a waveform generation calculation corresponding to the performance input detected in the immediately preceding or previous frame is executed. Since a frame interrupt occurs at the timing when the frame is switched, a waveform generation operation (FIG. 7A described later) in each frame is executed with this frame interrupt as a trigger.

  For example, for the three performance inputs detected in the frame S1, the waveform generation calculation is performed in the section 202 in response to the first frame interrupt of the next frame S2. The CPU 101 generates waveform data (accumulated for a plurality of channels and also provided an effect) in the waveform generation buffer on the RAM 103 by this waveform generation calculation. The generated waveform data is written in the reproduction buffer area on the RAM 103. As the reproduction buffer area, two buffer areas PB0 and PB1 of the same size prepared at consecutive addresses (the two are collectively referred to as a double buffer) are used. Writing to the reproduction buffer area is performed in step 707 of FIG. Further, the buffers PB0 and PB1 are alternately used for each frame. For example, the waveform data generated in the section 201 of the frame S1 is written in the reproduction buffer area PB0 on the RAM 103, the waveform data generated in the section 202 of the frame S2 is written in the reproduction buffer area PB1, and the waveform generated in the section 203 of the frame S3. The data is written in the reproduction buffer area PB0, the waveform data generated in the section 204 of the frame S4 is written in the reproduction buffer area PB1, and the waveform data is alternately written in PB0 and PB1.

  The waveform data written to the reproduction buffers PB0 and PB1 is read and reproduced in the section of the next frame after the waveform generation frame, triggered by a frame interrupt, as shown in “Reading and reproducing” in FIG. That is, the waveform data generated in frame S1 and written to PB0 is the next frame S2, the waveform data generated in frame S2 and written to PB1 is the next frame S3, and the waveform data generated in frame S3 and written to PB0 is In the next frame S4, the waveform data of PB0 and PB1 are alternately read and reproduced as follows. For read reproduction, the DMA controller 114 interrupts the CPU 101 for each sampling clock Fs and transfers the waveform data in the reproduction buffer (PB0 or PB1 designated) on the RAM 103 to the output FIFO of the sound I / O 112. By doing. The frame interrupt is generated when a return occurs when reading the reproduction buffers PB0 and PB1 in a loop (that is, when reproduction of PB1 is finished) and when an intermediate point of loop reading is passed (ie, when reproduction of PB0 is finished). The frame interrupt is a hardware interrupt generated by the sound I / O 112 and indicates the time point when one frame reproduction is completed. That is, the sound I / O 112 counts the number of transfer samples (the number of samples transferred from the PB0 and PB1 to the output FIFO of the sound I / O by the DMAC 114), and is half the size of the reproduction buffer (the sum of PB0 and PB1). Generate a frame interrupt for every number of samples transferred.

  FIG. 3A shows a configuration example of a waveform generation buffer used when the CPU 101 generates a waveform. There are four waveform generation buffers, mixA, mixB, mixC, and mixD. The mix A is a buffer for setting waveform data for dry use, that is, no effect. The mixB is a buffer for setting waveform data for reverberation (rev), that is, input to the reverberation process. The mixC is a buffer for setting waveform data for chorus (cho), that is, input to chorus processing. The mixD is a buffer for setting waveform data for variation (var), that is, input to variation processing. Each of mixA, mixB, mixC, and mixD includes 128 sets of samples (2 in terms of number of samples), with the storage area of the stereo left (L side) waveform sample and the storage area of the stereo right (R side) waveform sample as one set. × 128 = 256 samples). Each of the L-side waveform sample and the R-side waveform sample is a 16-bit (2 byte) sample. Each of mixA, mixB, mixC, and mixD is boundary-adjusted so as to be cached in units of 32 bytes (that is, 16 samples) in order from the top.

  FIG. 4 shows an example of an algorithm from generation of a musical tone by a soft sound source to channel accumulation. The waveform memory 401 is a memory that stores waveform sample data sampled at a predetermined rate. Here, it is assumed that waveform data is prepared on the ROM 102. However, waveform data prepared on the RAM 103 may be used. Waveform data on the RAM 103 is data read from the ROM 102, the external storage medium 105 or the hard disk 110, data input via the network I / O 107, or a waveform obtained by sampling the external input 113 by the sound I / O 112. Data or the like may be used.

  The software sound source first executes a musical sound generation process 402 for the required number of channels (the maximum number of channels is determined according to the CPU capability. Here, the maximum number of channels is 32). Since computation can be performed from any channel, computation processing may be performed with priority on later arrival, or priority may be lowered for a channel whose volume level is reduced. To generate music for one channel, waveform data in the waveform memory is read and interpolated by the waveform reading & interpolation processing 411, filtered by the filter 412, and the output of the filter 412 is divided into 8 series, and multipliers 413-1 to 413 are used. This includes a process of multiplying a predetermined coefficient by -8. The eight series of outputs are the dry L output obtained by multiplying the coefficient for dry L (stereo left) by the multiplier 413-1, and the dry R obtained by multiplying the coefficient for dry R (stereo right) by the multiplier 413-2. Output, reverb L output obtained by multiplying the multiplier 413-3 by the reverb L coefficient, reverb R output obtained by multiplying the reverb R coefficient by the multiplier 413-4, and chorus L coefficient obtained by the multiplier 413-5 L output obtained by multiplying the chorus R coefficient by the multiplier 413-6, the variation L output obtained by multiplying the coefficient for the variation L by the multiplier 413-7, and the multiplier The variation R output obtained by multiplying the variation R coefficient by 413-8. The 8 series outputs obtained for each sound generation channel are mixed (channel accumulation) for each series by mixers 403-1 to 403-8. Then, L and R are interleaved by interleaving processes 404-1 to 404-4, and the waveform generation buffers mixA, mixB, mixC, and mixD shown in FIG. 3A shown in 405-1 to 405-4 are set.

  By inputting an effect edit command using the keyboard 108 and the display 109, the user executes an effect edit processing program (not shown) in step 607 of the main flow in FIG. And parameters can be edited. FIG. 5 is an example of a soft effector algorithm set by user editing. A plurality of effects are applied to the waveform data generated on the waveform generation buffers mixA, mixB, mixC, and mixD by the processing of FIG. It has become.

  In the soft effect algorithm edit, for example, the number of blocks of the soft effector processing (3 blocks in FIG. 5), the processing content of each block (in FIG. 5, reverb, chorus, and variation), connection information between multiple blocks (FIG. 5). In the effect edit processing program, the specified multiple blocks of effects and the processing order of the multiple add processes can be specified in the effect edit processing program. Thus, an effect processing program having the algorithm shown in FIG. 5 is created. The algorithm of FIG. 5 represents processing of the following procedures (1) to (6).

  (1) Read waveform data in the waveform generation buffer mixD 501-4, perform variation processing 507, and overwrite the result on mixD.

(2) add (mixD-> mixA) ... 508 in FIG.
add (mixD-> mixB) 502 in FIG.
add (mixD-> mixC) 504 in FIG.
Execute. Add is a process of multiplying each sample of the buffer described on the front side of-> by a predetermined coefficient and weighting the sample, and adding the weight to each sample of the buffer described on the rear side of->. The add processing uses a common routine, and the weighting coefficient is specified in advance according to which processing result is weighted and added. By these add processes 508, 502, and 504, the results of applying the variation process 507 were respectively weighted and added to the waveform data of the dry buffer mixA, the reverb buffer mixB, and the chorus buffer mixC. It will be.

  (3) Read out waveform data from the waveform generation buffer mixC (data obtained by adding the waveform data prepared by inputting the variation processing 507 in the add processing 504 to the waveform data prepared for input to the chorus processing) and performing the chorus. Processing 506 is performed and the result is overwritten to mixC.

(4) add (mix C-> mix A) 509 in FIG.
add (mix C-> mix B) 503 in FIG.
Execute. By these add processes 509 and 503, the result of performing the chorus process 506 is weighted and added to the waveform data of the dry buffer mixA and the reverb buffer mixB.

  (5) Waveform generation buffer mixB (waveform data prepared for input to the reverberation process is added by adding the weighted waveform data subjected to the variation process 507 in the add process 502, and the chorus process 506 is added in the add process 503. Waveform data obtained by weighting and adding the applied waveform data), reverberation processing 505 is performed, and the result is overwritten on mixB.

(6) add (mixB-> mixA) ... 510 in FIG.
Execute. By this add process 510, the result of applying the reverberation process 505 is weighted and added to the waveform data of the dry buffer mixA. As a result, in mixA, waveform data reflecting variation, chorus, and reverb is set in the waveform data for drying.

  The above-described reverberation process 505, chorus process 506, and variation process 507 are processes for overwriting the mix B, mix C, and mix D with the effect imparting process, and the add processes 502-504, 508-510 are common. It is a routine. Therefore, it is possible to arbitrarily change the order in which a plurality of effects are applied (the connection relationship of the soft effectors) by simply changing these processing orders as appropriate. The common routine add can be used because, as shown in FIG. 3A, the waveform generation buffer is independent for each effect and has the same structure. In the embodiment of the present invention, an algorithm of a soft effector can be arbitrarily designated by designation using a user's keyboard 108 or the like.

  Furthermore, according to the embodiment of the present invention, when waveform generation is executed by the algorithm as shown in FIGS. 4 and 5, processing is performed in units of cache line size, thereby greatly increasing the cache hit rate and generating musical sounds. Speed up calculations. The speeding up of processing by caching will be described in detail with reference to FIGS. 7 (a) and 7 (b).

  Next, the processing procedure of the CPU 101 of the electronic musical instrument described above will be described with reference to the flowcharts of FIGS. 6 (a), 6 (b) and 7 (a), 7 (b).

  FIG. 6A shows the processing routine of the main routine related to the software sound source in the control program of the CPU 101. This main routine is registered in the OS (Operating System) as a driver of the software sound source. When generating a musical sound using a software sound source, first, the driver, that is, the processing of FIG. 6A is activated, and an API related to the software sound source is made valid. In FIG. 6A, various initial settings are made in step 601. With this initial setting, the reproduction buffers PB0 and PB1 are cleared to 0, and the processing to instruct the sound I / O 112 and the DMAC 114 to alternately read and reproduce the reproduction buffers PB0 and PB1 described in FIGS. 1 and 2 is started. Keep it. Next, in step 602, it is checked whether there is any activation factor. If there is an activation factor in step 603, the process proceeds to step 604. If there is no activation factor, the process returns to 602 again. Acceptance of the activation factor in steps 602 to 604 corresponds to acceptance of an instruction for the API related to the software sound source.

  In step 604, the type of the activation factor that has occurred is determined, and the process branches to each process. If the activation factor is the input of a MIDI event, the MIDI processing in step 605 is performed, and then the process returns to step 602 again. The MIDI event input and the MIDI processing in step 605 correspond to the reception of the performance input in FIG. If the activation factor is frame interruption (one frame reproduction complete) in step 604, the waveform generation processing in step 606 is performed, and then the process returns to step 602 again. The frame interrupt is a hardware interrupt that occurs every time the sound I / O 112 completes playback of one frame. The waveform generation processing in step 606 is processing for performing the waveform generation calculation shown in 201 to 204 in FIG. In this waveform generation process, waveform data for one frame (a number of waveform samples corresponding to half of the combined size of the reproduction buffers PB0 and PB1) is generated at a time and written alternately to the reproduction buffer PB0 or PB1. If the activation factor is another request in step 604, processing corresponding to the activation factor is performed in step 607, and then the process returns to step 602 again. In particular, an instruction to sample the external input 113 by the sound I / O 112, an instruction to change the setting of the soft effector algorithm as shown in FIG. 5, and the coefficients of the multipliers 413-1 to 413-8 in FIG. When a request for setting a weighting coefficient that defines a level) or a setting instruction for setting a weighting coefficient for an add process used by the soft effector is made, a corresponding process is performed in the other process of step 607. If it is determined in step 604 that the activation factor is a soft sound source termination request, the termination process is performed in step 608 and the process is terminated.

  FIG. 6B shows a procedure of note-on event processing executed when a note-on event is input in the MIDI processing in step 605 of FIG. 6A. First, in step 611, the MIDI channel, note number, and velocity of the input note-on event are set in the registers MC, NN, and VE, respectively. Next, in step 612, sound generation channels (0th to 31st channels) are assigned. In step 613, information (note number NN, velocity VE, etc.) necessary for sound generation is set in the soft tone generator register of the assigned sound generation channel. In step 614, note-on is written in the soft tone generator register of the assigned tone generation channel, a tone generation start instruction is issued, and the process is terminated.

  Although not shown, MIDI event processing such as note-off and other performance input is the same. That is, if note-off is set, note-off is set in the software tone generator register corresponding to the corresponding tone generation channel, and data corresponding to the performance input is written in the software tone generator register corresponding to the tone generation channel for other performance inputs.

  FIG. 7A shows the procedure of the waveform generation process in step 606 of FIG. By this waveform generation processing, musical tone generation is executed with an algorithm as shown in FIGS. First, in step 701, calculation preparation is performed. This is a process of recognizing a channel on which a waveform generation operation is to be performed with reference to the soft tone generator register, and recognizes whether the waveform for one frame generated by the current waveform generation operation should be set in the reproduction buffer PB0 or PB1. Processing preparations such as processing and clearing all areas of the waveform generation buffers mixA, mixB, mixC, and mixD to zero. Next, in step 702, waveform generation calculation is performed for 16 samples (2 × 16 = 32 samples when each sample of stereo L and R is counted as a unit) for all channels that need to be calculated. This is a process of executing the processing of the algorithm of FIG. 4 for 16 samples and generating 2 × 16 sample waveforms in each of the waveform generation buffers mixA, mixB, mixC, and mixD.

  In step 703, it is determined whether or not waveform samples for one frame have been generated, that is, whether or not 2 × 128 samples have been generated on the waveform generation buffers mixA, mixB, mixC, and mixD in FIG. To do. If one frame has not been completed, the process returns to step 702 to generate the next 2 × 16 samples. By repeating step 702, waveform samples are created 2 × 16 samples from the beginning for each of the waveform generation buffers mixA, mixB, mixC, and mixD in FIG.

  If the waveform for one frame can be generated in the waveform generation buffers mixA, mixB, mixC, and mixD in FIG. 3A in step 703, the process proceeds to step 704. In steps 704, 705, and 706, effect imparting processing such as variation, chorus, and reverb is performed. As described with reference to FIG. 5, this is an algorithm in which the user has designated variation processing, chorus processing, reverb processing, and add processing (instruction for changing the setting of the soft effector algorithm is performed in other processing in step 607). This process is executed in the order shown in FIG. Although not shown, the soft effector processes in steps 704, 705, and 706 are also executed every 2 × 16 samples as in steps 702 and 703. That is, the algorithm processing of FIG. 5 is executed for 2 × 16 samples from the top of the waveform generation buffers mixA, mixB, mixC, and mixD of FIG. 3A, and the processing of FIG. 5 is performed for the next 2 × 16 samples. Then, the process proceeds to obtain 2 × 128 samples having an effect on the waveform generation buffer mixA. The variation process, chorus process, reverb process, and add process described in FIG. 5 are processes for 2 × 16 samples.

  Although the add process described in FIG. 5 does not appear in FIG. 7A, it is assumed that this is included in the effect block process in steps 704 to 706. In other words, the variation processing at Step 704 is the processing of the procedures (1) and (2), the chorus processing at Step 705 is the processing of the procedures (3) and (4), and the reverberation processing is Step 706. Includes the processing of the procedures (5) and (6).

  It should be noted that the effect processes in steps 704, 705, and 706 may be executed all at once, instead of 2 × 16 samples. That is, each of the variation processing, chorus processing, reverberation processing, and add processing described in FIG. 5 may be processing for one frame at a time. Even in this case, the cache works effectively every 16 samples during processing of one frame. In this way, it is considered that the hit rate may decrease in processing between the buffers, but the hit rate increases for registers and buffers used in each effect processing. Compared to the sound channel processing, the effect processing takes a long time to obtain a result, so it is more efficient to process one frame at a time. Moreover, even if it is said for one frame at a time, it is efficient if processing is performed locally in units of 16 samples. In short, in order to make it difficult for a cache miss to occur, it is only necessary to process continuous data in a short time.

  After the soft effector processing, in step 707, the generated waveform for one frame (that is, 2 × 128 samples on mixA) is reserved for reproduction. This is a process of copying the mixA waveform sample to one of the playback buffers PB0 and PB1 (which is not currently used for playback). Since the process of alternately reading and reproducing the reproduction buffers PB0 and PB1 has already started in step 601, the waveform is generated in the next frame simply by copying the waveform that has not been used for reproduction. Executed.

  In this example, the waveform generation buffer mixA and the reproduction buffers PB0 and PB1 are provided separately. However, it is also possible to prepare two faces of mixA, one for PB0 and the other for PB1. In this case, since the waveform generation process is performed directly on the reproduction buffer, the process of copying the generated waveform in step 707 is unnecessary, and the processing speed is further improved.

  FIG. 7B shows a waveform generation processing procedure for 16 samples (2 × 16 = 32 samples if the stereo L and R samples are counted in units) in step 702 of FIG. 7A. First, in step 711, calculation preparation is performed for the first channel. Since the channel requiring the waveform calculation process and the priority between the channels are obtained by the calculation preparation in step 701 in FIG. 7A, the channel with the highest priority is set as the first channel. Next, in steps 712 to 718, waveforms of 16 samples are generated for the channel.

  First, in step 712, an envelope value used in later processing is obtained. The envelope value may be generated by, for example, an operation-type envelope generation process that outputs an ADSR (attack, decay, sustain, release) type envelope waveform. The envelope value generated in step 712 is used in common for the 16 samples currently being generated. As the envelope value generated here, one generated value is commonly used for 16 samples. That is, one envelope value is generated for every 16 samples of the waveform. Next, in step 713, address generation, waveform readout, and interpolation processing (411 in FIG. 4) are performed for 16 samples, and in step 714, filtering processing (412 in FIG. 4) is performed on these samples. Each sample is monaural, not yet divided by stereo LR at this stage.

  Next, in step 715, 2 × 16 samples of mixA are calculated. In FIG. 4, the following processing is performed on 16 samples (monaural) output from the filter processing 412. First, the weight coefficient for dryL (coefficient defining the transmission level of dryL) and the envelope value obtained in step 712 are added. Since both the coefficient and the envelope value are values on the dB (decibel) scale, addition corresponds to multiplication on the linear scale. Thereafter, the multiplier 413-1 multiplies each waveform sample value output from the filter processing 412 by exponential transformation with the addition result of the coefficient and the envelope value. As a result, 16 L-side waveform samples for dry are obtained. The R-side waveform sample for dry is also obtained in the same manner with the coefficient set as the coefficient for dryR. The waveform of 2 × 16 = 32 samples of the dry LR obtained as described above is accumulated in mixA.

  In the same manner as in step 715, in step 716, the waveform of 2 × 16 = 32 samples of reverb LR is accumulated to mixB, and in step 717, the waveform of 2 × 16 = 32 samples of chorus LR is accumulated to mixC. In step 718, the waveform of 2 × 16 = 32 samples of the LR for variation is accumulated in mixD. Of course, different weighting coefficients are used for dry, reverb, chorus, and variation.

  Next, in step 719, it is determined whether there are remaining channels to be calculated. If there is, preparation for calculation of the next channel is performed in step 720 (calculation is performed first on the channel with higher priority), and then the process returns to step 712. By repeating this, waveform generation for 16 samples is performed for each of the stereo L and R for all channels to be calculated and accumulated in the waveform generation buffers mixA, mixB, mixC, and mixD. If there is no remaining channel to be calculated in step 719, the process returns.

  In the waveform generation calculation in FIGS. 7A and 7B described above, the waveform generation including the effect application (the processes in FIGS. 4 and 5) is executed in units of 2 × 16 samples. The 16 samples are 32 bytes, and the waveform generation buffers mixA, mixB, mixC, and mixD in FIG. 3A are boundary-adjusted so that they are cached in units of 32 bytes in order from the top. Therefore, for example, when the first sample of mixA is accessed, 16 samples including the first sample are cached in the cache memory 117, and during the processing for the 16 samples, the waveform generation processing is executed on the cache memory 117. Therefore, the waveform can be generated very quickly. Since it is 2 × 16 samples, the unit is 64 bytes. However, since every 16 adjacent samples are cached in another cache line, two cache lines are used here.

  In particular, in this example, the user can arbitrarily specify multiple types of effect application algorithms, so multiple types of effect application processing are performed in various orders, but waveform generation is independent for each type of effect to be applied. Therefore, when one effect applying process is executed, the process is performed on the corresponding buffer. Since the buffer consists only of waveform samples used for applying the effect (that is, there is no waveform sample unnecessary for the effect processing in between), the cache hit efficiency can be drastically increased and caching can be performed. Is highly effective.

  Next, processing of the DMA controller 114 at the time of reproduction will be described with reference to the flowchart in FIG. At the time of reproduction, a sample request interrupt (hardware interrupt) is issued from the sound I / O 112 every sampling period. In response to this, the DMA controller 114 performs the processing of FIG. First, in step 731, one sample of the reproduction buffers PB 0 and PB 1 is sent to the output FIFO of the sound I / O 112. As described with reference to FIG. 1, the waveform data written in the output FIFO is D / A converted at each sampling period, sent to the sound system 115, and emitted. Note that “DMAB” in step 731 refers to the reproduction buffers PB0 and PB1. The reproduction buffers PB0 and PB1 are described as DMAB because they can be regarded as DMA buffers. Next, in step 732, the pointer p is incremented and the process is terminated. p is a pointer for reading one sample from the reproduction buffers PB0 and PB1. As described above, one sample is transferred from the reproduction buffers PB0 and PB1 to the sound I / O 112 every sampling period while the pointer p is advanced.

  The pointer p is a pointer for sequentially reading from the first sample of PB0 to the last sample of PB1 by incrementing by one, but after reading the last sample of PB1, the first sample of PB0 The pointer value needs to be updated to point to. This processing is set so that the DMA controller 114 automatically performs this processing.

  In the embodiment of the invention described above, as can be seen from steps 702 and 703 in FIG. 7A and FIG. 7B, loop processing for performing waveform generation processing for 16 samples over all channels requiring computation is performed. The waveform generation process for 16 samples is executed in the inner loop, and the waveform for one frame is generated by executing the waveform generation process for one frame until the completion of one frame. However, these loop processes may be interchanged. That is, the waveform generation for 16 samples for one channel is repeated in the inner loop to generate a waveform for one frame, and this is executed for each channel in the outer loop to generate the final waveform for one frame. It may be. In the above embodiment of the present invention, 16 samples across all channels are created, so the cache hit rate is high. However, if the CPU processing power is low, waveform generation for one frame is completed within one frame time. There is a possibility not to. On the other hand, if the above loop processing is replaced, a channel with a high priority is created for one frame at the beginning, so even if waveform generation for all channels is not completed within one frame time, the priority is Higher channels will be pronounced. Note that these methods may be mixed so that the predetermined number of channels is the former method and the remaining channels are the latter method.

  In the embodiment of the present invention, the cache memory is a write-through method, but it may be a write-back method. Since the waveform sample can be updated on the cache memory by using the write-back method, higher-speed waveform generation processing is realized.

  Note that the user may be able to specify not only the connection state between a plurality of effects but also the number and contents of the plurality of effects. Further, since the number of samples serving as a unit of caching differs for each CPU, the processing unit of waveform generation may be changed accordingly.

  The number of waveform generation buffers is four from mixA to mixD in the embodiment of the present invention. This corresponds to the fact that the number of blocks in the subsequent stage is 3, It is changed according to the number of blocks. Since an input buffer and a dry buffer are required for each effect, the number of buffers is the number of effect blocks + 1.

  It should be noted that various related information and operation programs for realizing the above-described musical tone generation method are stored in a storage device such as a hard disk, and the CPU reads them into the RAM before using them, and further uses a CD-ROM ( Compact disk-read-only memory), data stored in portable storage media such as flexible disks and magneto-optical disks can be transferred to a storage device such as a hard disk. Useful for addition (installation etc.) and update (version upgrade etc.). Of course, data may be directly transferred from the portable storage medium to the RAM.

  Furthermore, it is also possible to download related information and operation programs on a storage device such as a hard disk from the communication network side via a communication interface (network I / O) instead of via a portable storage medium. The communication interface is connected to a communication network such as a LAN (local area network), the Internet, or a telephone line, and is connected to a server computer via the communication network. When the related information or operation program is not stored in the storage device (hard disk or the like) of the own device, the present apparatus as a client is connected to the server computer via the communication interface and communication network. Send a command requesting Upon receiving this command, the server computer distributes the requested relationship information, operation program, etc. to this apparatus via the communication network. The device receives the distributed relationship information, operation program, and the like via the communication interface and stores them in the storage device, thereby completing the download.

  Further, the present apparatus may be realized by a commercially available personal computer installed with related information, an operation program, and the like. Of course, in this case as well, data such as related information and operation programs can be distributed in advance in a non-volatile memory such as a ROM, stored in a portable storage medium, and distributed. A method of distributing via an interface is applicable.

Block configuration diagram of an electronic musical instrument to which a musical sound generator according to the present invention is applied Diagram for explaining the principle of musical sound generation The figure which shows the structural example of the waveform generation buffer which concerns on this invention, and the structural example of the conventional waveform generation buffer The figure which shows the example of the algorithm from the musical tone generation with the soft sound source to the channel accumulation The figure which shows the example of the algorithm of the soft effector which gives plural effects to the waveform data Flowchart diagram of main routine and note-on event routine Flowchart diagram of waveform generation processing routine and waveform generation processing routine for 16 samples

Explanation of symbols

  DESCRIPTION OF SYMBOLS 101 ... Central processing unit (CPU), 102 ... Read-only memory (ROM), 103 ... Random access memory (RAM), 104 ... Drive apparatus, 106 ... Timer, 107 ... Network input / output (I / O) interface, 108 ... Keyboard 109, display 110, hard disk 111 sampling clock (Fs) generator 112 sound I / O 114 DMA (direct memory access) controller 115 sound system 116 bus line 117 cache Memory, mixA, mixB, mixC, mixD ... Waveform generation buffer.

Claims (4)

An arithmetic processing unit capable of executing various types of software;
A waveform buffer for storing the generated multiple waveform samples, wherein the waveform buffer is provided in correspondence with a plurality of effects;
A storage device storing a plurality of effect processing software, wherein the processing executed by the arithmetic processing device by each effect processing software performs predetermined effect processing on a plurality of waveform samples of a corresponding waveform buffer, and after processing To write waveform samples to the corresponding waveform buffer,
The arithmetic processing unit executes an add process in which an arbitrary waveform sample in the waveform buffer is subjected to a predetermined level control and added to a waveform sample stored in another arbitrary waveform buffer and then written to the same storage position again. A storage device storing add processing software for
With respect to the effect processing corresponding to each waveform buffer executed on each waveform buffer and the add processing executed between any two waveform buffers, processing of effect processing and add processing is performed according to a desired effect algorithm. A musical sound generating apparatus comprising: a waveform buffer that designates an order and designates a waveform buffer to be added in the add process and a waveform buffer to be added. A specifying means for a user to specify an algorithm indicating a connection state of the plurality of effects;
The musical sound generating device according to claim 1, wherein the means for designating the waveform buffer changes the processing order of the multiple effect processing and the multiple add processing according to the algorithm set by the designating means. A musical sound generation method that is executed by a processing device and is capable of generating musical sounds for a plurality of channels,
Supplying performance information;
Generating timing signals at predetermined time intervals;
Each time the timing signal is generated, a predetermined number of samples corresponding to the performance information are generated and output to a plurality of buffers corresponding to a plurality of effect processes, and executed for a plurality of channels. A generating step for generating channel accumulated waveform data,
After applying effects to multiple waveform samples in a buffer, write the processed waveform samples to the buffer. Multiple effect processing corresponding to each buffer, and waveform samples in any one buffer The effect processing is performed by executing the add processing to be added to the waveform sample stored in another arbitrary buffer and written to the storage position again in the processing order corresponding to the designated effect algorithm. An effect processing step for generating and outputting the applied waveform data;
A musical sound generation method comprising: supplying waveform data output from the effect processing step to a digital-to-analog converter sample by sample for each sampling period and converting the sample into an analog waveform. A storage device readable by a processing device, storing a program related to a musical sound generation method that is executed by the processing device and can generate musical sounds for a plurality of channels,
A program relating to the musical sound generating method is stored in a processing device.
Supplying performance information;
Generating timing signals at predetermined time intervals;
Each time the timing signal is generated, a predetermined number of samples corresponding to the performance information are generated and output to a plurality of buffers corresponding to a plurality of effect processes, and executed for a plurality of channels. A generating step for generating channel accumulated waveform data,
After applying effects to multiple waveform samples in a buffer, write the processed waveform samples to the buffer. Multiple effect processing corresponding to each buffer, and waveform samples in any one buffer The effect processing is performed by executing the add processing to be added to the waveform sample stored in another arbitrary buffer and written to the storage position again in the processing order corresponding to the designated effect algorithm. An effect processing step for generating and outputting the applied waveform data;
A storage medium readable by a processing apparatus, storing a program for executing the step of supplying waveform data output from the effect processing step to the digital-analog converter by one sample at every sampling period and converting it to an analog waveform .

Images