JP5316091B2 - Waveform memory sound source - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain musical sound with various performance strength in high fidelity, even when a waveform data of sound of a musical instrument with an attenuation type is stored in a memory with a small capacity. <P>SOLUTION: A whole waveform data in which a connection point is set is stored in a waveform memory, for a waveform data (f) of a maximum strength. As for a waveform data W(mf) of a performance strength (mf) which is weaker than the maximum strength, the waveform data (f) creates an attack section waveform data Wa(mf) composed of waveforms of a connection section which is connected to a connection point or after, after a waveform in an attack feature section by performing attack waveform creation processing A, and it is stored in the waveform memory. As for waveform data W(p) of performance strength (mp) and (p), which are weaker than the waveform (mf), attack section waveform data Wa(mp) and Wa(p) which are created by similar attack waveform creation processing B and C are stored in the waveform memory. Thereby, a storage capacity of the waveform memory is reduced. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to a waveform memory sound source that synthesizes musical sounds from waveform data stored in a waveform memory.

  Conventionally, a waveform memory sound source is known as one of sound sources in a musical sound generation device. The waveform memory sound source stores waveform data obtained by sampling a musical tone when playing an actual musical instrument and converting it into a digital signal in the waveform memory, and reading the waveform data from the waveform memory at a reading speed according to the note number to be generated. Is read out to generate a musical tone. In order to reproduce a musical tone that approximates the tone of a natural musical instrument in such a waveform memory sound source, it is necessary to store waveform data for each pitch or range in the waveform memory for each tone color. The amount of waveform data stored in the waveform memory is enormous. Furthermore, the tone color of the musical tone that is generated differs according to the performance intensity such as the pitch (tone range) of the musical tone and the key depression speed of the keyboard. For example, when a key is pressed with a strong touch, a musical sound with an increased high frequency component is generated, and a musical sound with a decreased high frequency component is generated as the touch of pressing the key becomes weaker. Therefore, in a musical sound generating device equipped with a waveform memory sound source, it is also necessary to store musical sound waveform data corresponding to the pitch (tone range) and performance intensity in the waveform memory. As described above, if waveform data corresponding to the pitch (sound range) or performance intensity is stored in the waveform memory for each tone color, a waveform memory having a huge memory capacity is required.

  By the way, the high-frequency component such as the harmonic component of the fundamental wave is contained in the attack part of the musical sound waveform, but the attenuation of the harmonic proceeds rapidly in the sustaining part, and the remaining component is almost the fundamental component. Therefore, the frequency component included is lower than that of the attack portion. Therefore, as a method of reducing the memory capacity of the waveform memory sound source, instead of storing the waveform data of all waveforms in the waveform memory, the waveform data of the attack part and a part of the waveform data of the continuous part are repeatedly reproduced. It has been proposed to store in a waveform memory as loop waveform data. When generating a musical tone in such a waveform memory sound source, first, the waveform data of the attack portion is read, and then the musical tone waveform is generated by performing a loop reproduction process of repeatedly reproducing the loop waveform data. .

Japanese Patent No. 3666346 Japanese Patent No. 3534012 Japanese Patent No. 3655903

  In the waveform memory sound source, if the waveform data of the attack part and the loop waveform data are stored in the waveform memory in order to reduce the memory capacity, the continuation part that repeatedly reproduces the loop waveform data when generating a musical sound There was a problem that the tone became monotonous. Also, in the conventional waveform memory sound source, when all waveforms from the start to the end of the musical sound waveform are stored in the waveform memory, an enormous memory capacity is required, so normally only waveform data of one performance intensity is stored. There is a problem that musical sounds having various performance intensities cannot be faithfully obtained because they are stored in the waveform memory.

  Accordingly, an object of the present invention is to provide a waveform memory sound source that can faithfully obtain musical tones with various performance intensities even when waveform data of attenuation-type instrument sounds is stored in a memory with a small capacity.

  In order to achieve the above object, the waveform memory sound source of the present invention is a full waveform having the strongest performance intensity, and has a plurality of connection points corresponding to a plurality of performance intensities and a plurality of connection points corresponding to the plurality of performance intensities. A first waveform data for which a read start address is set, and a plurality of fourth waveform data corresponding to a plurality of performance intensities, each having an attack feature in the waveform data of the corresponding performance intensity, followed by A waveform memory storing a plurality of fourth waveform data including a connection portion for connecting to the connection point corresponding to the performance intensity of the first waveform data, and a performance included in the supplied sounding instruction A selection unit that selects any one of the plurality of fourth waveform data according to the performance intensity indicated by the intensity data, and a fourth selected by the selection unit according to the sound generation instruction. Reading out the shape data from the head address, and reading out the first waveform data from the reading start address corresponding to the performance intensity among the plurality of reading start addresses in synchronization with the reading; And connecting the first waveform data read out to the fourth waveform data at a predetermined point after the end of the connecting portion of the fourth waveform data read out by the unit, and based on the connected waveform data The main feature is the provision of a musical tone forming unit for forming musical tone waveform data.

  According to the present invention, a plurality of first waveform data having the highest strength at which a connection point is set, an attack feature, and a plurality of subsequent connections for connecting to the first waveform data after the connection point. Since only the fourth waveform data is stored in the waveform memory, the storage capacity of the waveform memory is reduced. When synthesizing a musical tone, the fourth waveform data is read from the head address, and in synchronization with the reading, the first waveform data is read from the read start address corresponding to the fourth waveform data, and the fourth waveform data is read. Since the musical tone waveform data is formed by connecting the first waveform data after the data connection portion, a plurality of all waveform data corresponding to the performance intensity is almost the same as when stored in the waveform memory as it is. Music can be synthesized with quality. Accordingly, it is possible to faithfully obtain musical tones having various performance intensities.

It is a circuit block diagram which shows the structure of a musical sound production | generation apparatus provided with the waveform memory sound source of the Example of this invention. It is a circuit block diagram which shows the detailed structure of the waveform memory sound source of the Example of this invention. It is a figure which shows the outline | summary of the formation method of the 1st waveform data memorize | stored in the waveform memory in the waveform memory sound source of this invention. It is a figure which shows the outline | summary of the formation method of the 2nd waveform data memorize | stored in the waveform memory in the waveform memory sound source of this invention. It is a figure explaining the attack waveform creation process in the formation method of the waveform data memorize | stored in the waveform memory in the waveform memory sound source of this invention. It is another figure explaining the attack waveform creation process in the formation method of the waveform data memorize | stored in the waveform memory in the waveform memory sound source of this invention. It is a conceptual diagram of the process which produces the waveform data Da (x) of the deterministic component of the attack part waveform data Wa (x) concerning the formation method of waveform data. It is a conceptual diagram of the process which produces the attack part waveform data Wa (x) concerning the formation method of waveform data. It is a conceptual diagram of the phase smoothing process in the smoothing process concerning the formation method of waveform data. It is a conceptual diagram of the amplitude smoothing process in the smoothing process concerning the formation method of waveform data. It is a figure explaining the process which produces attack part waveform data by performing a cross fade in the formation method of waveform data. It is a conceptual diagram of the process which produces attack part waveform data by performing a cross fade in the formation method of waveform data. It is a figure explaining the process which produces | generates a musical tone in the musical tone production | generation apparatus concerning the formation method of waveform data. It is a figure which shows the timing chart in which a musical tone is produced | generated when note-on is detected in the musical tone production | generation apparatus concerning this invention. It is a flowchart of the note-on event process performed in the sound source of the musical tone generator according to the present invention. It is a flowchart of the link point interruption process performed in the sound source of the musical sound production | generation apparatus concerning this invention. It is a flowchart of the fade-out completion | finish interruption process performed in the sound source of the musical tone production | generation apparatus concerning this invention. It is a flowchart of the pitch change event process performed in the sound source of the musical tone generator according to the present invention. It is a part of flowchart of the other note on event process performed in the sound source of the musical sound production | generation apparatus concerning this invention.

FIG. 1 shows a circuit block diagram showing a configuration of a musical sound generating apparatus 1 having a waveform memory sound source of the present invention.
In the musical tone generating apparatus 1 according to the present invention shown in FIG. 1, a CPU (Central Processing Unit) 10 controls the overall operation of the musical tone generating apparatus 1 and forms waveform data to be stored in the waveform memory 23. And musical tone control processing. A RAM (Random Access Memory) 12 is set with a work area for the CPU 10 and a storage area for various data. The flash ROM 11 stores operation software such as waveform data forming processing and musical tone control processing executed by the CPU 10. Since the flash ROM 11 is rewritable, the operation software can be rewritten and the operation software can be easily upgraded. The timer 13 is a timer that functions as a watchdog timer, an interval timer, and a timer interrupt. A hard disk drive (HDD) 14 having a hard disk can record waveform data of all waveforms corresponding to the tone range and performance intensity for each tone color. Waveform data to be stored in the waveform memory 23 is formed by performing waveform data formation processing on the waveform data read from the HDD 14. The HDD 14 can store an operating system (OS) and various application programs and data.

  The MIDI I / O 15 is a MIDI interface for exchanging MIDI events with an external MIDI device, and the musical tone generator 1 can receive a MIDI signal output from the MIDI device. A musical sound signal can be generated based on the received MIDI signal and emitted from the sound system 21. A panel switch (panel SW) 16 is provided in the panel, and an event of a panel operator operated by scanning the panel SW 16 is detected and output to the bus line 25. The panel display 17 includes display means such as liquid crystal, and displays various setting screens on the display means, and operation screens during waveform data formation processing and musical tone control processing. The writing circuit 18 samples the musical sound waveform signal input from the external waveform input terminal 24 and sends it to the HDD 14 for recording, or writes the waveform data subjected to the waveform data forming process to the waveform memory 23. The buffer (BUFF) 19 temporarily stores the waveform data that has been subjected to the waveform data forming process written from the RAM 12 to the waveform memory 23 or the waveform data read from the waveform memory 23, and transfers it to the CPU 10, RAM 12, etc. To do. The sound source 20 is a waveform memory sound source that sequentially reads waveform data from the waveform memory 23 based on each information supplied via the bus line 25 and generates a musical sound signal based on the read waveform data. The sound source 20 has a plurality of tone generation channels, and the tone signal generation processing of each tone generation channel is performed by time-division multiplexing control, and is configured to be able to generate an independent tone signal for each tone generation channel. . In this case, the sound source 20 performs a plurality of time-division channel operations, receives sound source parameter information from the CPU 10 at the time of note-on, and generates a plurality of musical sound signals for a plurality of channels. The sound source parameter information is assigned channels, pitch information, envelope parameters, and the like. In each channel of the sound source 20, the waveform data recorded in the waveform memory 23 is read, and the read waveform data is processed along the sound source parameter information to generate a musical sound signal. The tone signal generated by the sound source 20 is emitted from the sound system 21. The access management unit 22 manages access time slots of the waveform memory 23 so that accesses from the writing circuit 18, the sound source 20, and the buffer 19 to the waveform memory 23 do not collide with each other. Note that each unit of the musical tone generating device 1 exchanges data with each other through a bus line 25 including a control bus, an address bus, and a data bus.

  Next, FIG. 2 shows a circuit block diagram showing a detailed configuration of the sound source 20 according to the embodiment of the present invention. In the sound source 20 shown in FIG. 2, the control register 38 has a CPU 10 that controls the note-on signal, note-off signal of each tone generation channel, parameters for controlling the characteristics of the tone signal generated in each tone generation channel, link information described later, Various parameters such as effect parameters for controlling the sound effect applied by the DSP 36 are written. Furthermore, the CPU 10 can write the plurality of parameters of the same channel or different channels simultaneously to the control register 38 by performing the writing using the plurality of simultaneous writing units 39. Specifically, a plurality of parameters that the CPU 10 wants to write at the same time are passed to the plurality of simultaneous writing units 39 in advance, and the plurality of simultaneous writing units 39 perform the following sampling cycle and the next in response to a trigger from the CPU 10 thereafter. The plurality of parameters are sequentially written in the control register 38 at the timing of the gap between the sampling periods. The plurality of parameters written in this way are reflected at the same timing (the next sampling cycle) on the musical tone signal generated in each tone generation channel. This is called "simultaneous writing". The parameters of each tone generation channel written to the control register 38 include note number NN, pitch data PD, volume difference, start address and end address of waveform data to be read, original pitch of the waveform data, amplitude envelope (amplitude EG) parameters , Crossfade control data and the like are included. The F number generation unit 31 includes an integer part and a decimal part based on the note number NN, pitch data PD, and original pitch of the channel supplied from the control register 38 for each time division channel timing, and A frequency number of the channel (hereinafter referred to as “F number”) for pitch shifting the pitch of the waveform data to the pitch indicated by the note number NN and the pitch data PD is generated and sent to the reading unit 32. The reading unit 32 accumulates the supplied F number for each time-division channel, thereby obtaining an address that proceeds at a speed corresponding to the F number from the start address to the end address of the channel from the control register 38. Generate waveform data samples from the waveform memory 23 based on the integer part of the generated address, and interpolate the read waveform data samples between samples based on the decimal part of the generated address. , The interpolation sample of the channel is generated.

  The amplitude change control unit 33 controls the change in volume from the rising edge to the falling edge of the musical sound at each time division channel timing in accordance with the note-on signal, note-off signal, and volume EG parameter supplied from the control register 38. Generates volume envelope (volume EG) waveform data for the channel. At each time division channel timing, the generated volume EG waveform data is multiplied by the interpolation sample of the waveform data output from the reading unit 32 in the first multiplier M1, and volume control of the waveform data is performed. The crossfade (Xfade) control unit 34 generates a crossfade coefficient of the channel at each time division channel timing in accordance with the crossfade control data supplied from the control register 38. At each time division channel timing, the generated crossfade coefficient is multiplied by the waveform data sample output from the first multiplier M1. In this case, in the channel in which fade-in is instructed by the cross-fade control data, the crossfade gradually increases (fade in) from 0% (0.00) to 100% (1.00) within a predetermined crossfade time. A coefficient is generated, and a crossfade coefficient that gradually decreases (fade out) from 100% (1.00) to 0% (0.00) within a predetermined crossfade time is generated in a channel in which fadeout is instructed. . The fade-in and fade-out here are originally for connecting waveform data having the same shape (or almost the same shape), so there is no restriction on the change curve. In order to reduce the time required for crossfade, it is preferable to use a change curve that changes linearly on a linear scale. In addition, the value of the crossfade coefficient generated for each channel can be directly set to 100% (1.00) to 0% (0.00) by the crossfade control data.

  For each time division channel, the waveform data sample output from the second multiplier M2 is supplied to the mixer 35 and is weighted and mixed for each of a plurality of output destinations. Based on the effect parameter from the control register 38, the DSP 36 gives various acoustic effects such as reverb and chorus to the sample of waveform data supplied from the mixer 35 for each sampling period, and returns it to the mixer 35. Further, the mixer 35 adds the waveform data from the DSP 36 to the sample of waveform data having a DAC (digital-analog converter) 37 as an output destination for each sampling period, and then outputs the sample to the DAC 37. In this way, the mixing output from the mixer 35 to which the effect is applied is converted into an analog signal by the DAC 37 and output to the sound system 21 at every sampling period.

Next, an outline of a first waveform data forming method for forming waveform data to be stored in the waveform memory 23 of the musical tone generating device 1 will be described with reference to FIG.
For example, playing a keyboard instrument with the highest performance strength (f), performance strength (mf) weaker than performance strength (f), performance strength (mp) weaker than performance strength (mf), performance weaker than performance strength (mp) Keys are pressed at four performance intensities (p), and each tone is recorded on the HDD 14 with the same microphone and at the same recording level. The waveform data recorded in the HDD 14 are the waveform data W (f) of the performance intensity (f), the waveform data W (mf) of the performance intensity (mf), the waveform data W (mp) of the performance intensity (mp) in FIG. This is waveform data W (p) of performance intensity (p). Next, waveform data to be stored in the waveform memory 23 is created. All waveform data W (f) having the highest performance intensity (f) is stored in the waveform memory 23 without being subjected to waveform creation processing. Stored as first waveform data. Next, attack waveform creation processing A is performed on the waveform data W (mf), and the attack waveform data Wa (mf) comprising the attack feature of the waveform data W (mf) and the connection portion connected to the waveform data W (f). ). Attack waveform data Wa (mf) having a storage capacity considerably smaller than the total waveform data is stored in the waveform memory 23. In the attack waveform creation process A, the waveform data W (f) is used in the area after the attack feature of the waveform data W (mf) by using the waveform data W (f) which is the maximum intensity as will be described later. The attack waveform data Wa (mf) that can be connected to () is created.

  In addition, the waveform data W (mp) is subjected to the same attack waveform creation processing B as the above processing, thereby creating the attack portion waveform data Wa (mp) having a considerably smaller storage capacity than the entire waveform data. The waveform data W (p) stored in the memory 23 is subjected to attack waveform generation processing C similar to the above-described processing, so that the attack portion waveform data Wa (p) that has a considerably smaller storage capacity than the total waveform data is stored. Created and stored in the waveform memory 23. In this way, a plurality of attack part waveform data Wa (mf) to attack part waveform data Wa (p) corresponding to the performance intensity stored in the waveform memory 23 are set as the fourth waveform data. One waveform set shown in FIG. 3 is a waveform set corresponding to the performance intensity of a certain tone range in a certain tone color. In this one waveform set, only the waveform data W (f) with the performance intensity (f), which is the maximum intensity, is required for all waveform data, and the waveform data with the performance intensity weaker than the maximum intensity is considerably higher than the entire waveform data. Since only the attack portion waveform data Wa (x) having a small storage capacity is required, the capacity of the storage area for the waveform data required for the waveform memory 23 can be greatly reduced.

  Further, when generating a musical tone using the waveform memory 23, the attack portion waveform data Wa (x) of the instructed performance intensity is read, and the waveform data W ( After connecting to f), the connected waveform data W (f) is read and reproduced. Thus, it is possible to synthesize musical tones with a quality that is almost the same as when reproducing all waveform data of the instructed performance intensity. In this case, using the first sound channel and the second sound channel of the sound source, the attack part waveform data Wa (x) is read and sounded in the first sound channel, and in synchronization with that, the second sound channel is read. The waveform data W (f) is read and sounded, and the waveform of the attack part of the first sounding channel is created by crossfading the output of the two sounding channels at the timing when the crossfade waveform data (described later) is read. The data Wa (x) is smoothly shifted to the waveform data W (f) of the second tone generation channel. When the instructed performance intensity is the highest intensity, only one sounding channel is prepared, and all waveform data in the waveform data W (f) of the performance intensity (f) is read and reproduced on one sounding channel. To. Note that the connection from the waveform data Wa (x) to the waveform data W (f) does not necessarily need to be based on a cross fade.

The outline of the second waveform data forming method for forming the waveform data to be stored in the waveform memory 23 of the musical tone generating device 1 will be described with reference to FIG.
Also in the second waveform data forming method, the performance data (f) waveform data W (f) and performance strength (mf) waveform data W (mf) are stored in the HDD 14 in the same manner as the first waveform data forming method. The waveform data W (mp) of the performance intensity (mp) and the waveform data W (p) of the performance intensity (p) are recorded. Next, the entire waveform data of the waveform data W (f) of the performance intensity (f), which is the maximum intensity, is stored in the waveform memory 23 as the first waveform data. Next, attack waveform cut-out creation processing D is performed on the waveform data W (f), and the attack waveform data Wa (f) composed of the attack feature portion and the connection portion of the waveform data W (f) is cut out. Waveform data Wa (f) is stored in the waveform memory 23. Further, the waveform data W (mf) includes the attack portion waveform data Wa (mf), which is created by performing the attack waveform creation processing A similar to the first embodiment and has a considerably smaller storage capacity than the entire waveform data. Store in the waveform memory 23. Further, the attack part waveform data Wa (mp) is created by performing the attack waveform creation process B in the same manner as the first embodiment, and the attack part waveform data Wa (p) is created by performing the attack waveform creation process C. And stored in the waveform memory 23, respectively. A plurality of attack part waveform data Wa (f) to attack part waveform data Wa (p) corresponding to the performance intensity stored in the waveform memory 23 in this way are set as the fourth waveform data.

  FIG. 4 shows one waveform set corresponding to the performance intensity of a certain tone range in a certain tone color stored in the waveform memory 23. In the one waveform set shown in FIG. 4, only the waveform data W (f) of the performance intensity (f), which is the highest intensity, is stored as all waveform data, which is considerably less than the entire waveform data at each performance intensity. Attack portion waveform data Wa (x), which is a storage capacity, is stored. Thus, the capacity of the waveform data storage area required for the waveform memory 23 can be greatly reduced. When a musical tone is generated using the waveform memory 23, the attack portion waveform data Wa (x) of the instructed performance intensity is read out and the waveform data Wa (x ) Is connected to the waveform data W (f) by the connecting portion, and thereafter, the connected waveform data W (f) is read and reproduced. Thus, it is possible to synthesize musical tones with a quality that is almost the same as when reproducing all waveform data of the instructed performance intensity. In this case, a first sounding channel that produces the attack waveform data Wa (x) and a second sounding channel that produces the waveform data W (f) are prepared, and the first sounding channel is created by crossfading at the connecting portion. The waveform data Wa (x) of the attack portion of FIG. 5 is smoothly shifted to the waveform data W (f) of the second sound generation channel. Even if the instructed performance intensity is the highest intensity, the performance intensity is shifted from the waveform data Wa (f) of the first sound channel attack section to the waveform data W (f) of the second sound channel. Therefore, it is possible to eliminate the need to change the tone generation process. Also in the second embodiment, the connection from the waveform data Wa (x) to the waveform data W (f) does not necessarily need to be based on a cross fade.

Next, the attack waveform creation processes A to C will be described with reference to the drawings. In the figure to be referred to, the waveform data of performance intensity (x) weaker than the maximum intensity is shown as W (x).
In the attack waveform creation process, first, as shown in the upper part of FIG. 5, the time axis is set so that the amplitude levels at which the waveform data W (f) and the waveform data W (x) are attenuated substantially coincide with each other. Line up. However, the attenuation waveform shown in FIG. 5 is an image diagram, and the actual attenuation waveform has a faster rise and a slower fall, and the fall curve is an exponential curve. Next, the waveform data W (f) of the performance intensity (f) and the waveform data W (x) of the performance intensity (x) are frequency-analyzed, respectively, and the peak trajectory data (deterministic component) and residual waveform data (residual) Waveform). This extraction method is basically the same as the extraction method described in Japanese Patent No. 3534012 to Japanese Patent No. 3659053. In the lower part of FIG. 5, the deterministic component d (f) and the residual waveform R (f) obtained by separating and extracting the waveform data W (f) of the performance intensity (f) by frequency analysis, and the performance intensity (less than the maximum intensity) A deterministic component d (x) and a residual waveform R (x) obtained by separating and extracting the waveform data W (x) of x) by frequency analysis are shown. In the frequency analysis processing of waveform data, first, the waveform data to be analyzed is subjected to spectrum analysis while moving the analysis section in the time axis direction to create a peak locus of the frequency component. The spectrum of the waveform can be analyzed by performing FFT (Fast Fourier Transform). Then, peak trajectory data (deterministic component) is obtained from the spectrum of all created peak trajectories, and the waveform obtained by inverse Fourier transform of the obtained peak trajectory data (deterministic component) from the original waveform data. The residual waveform data (residual waveform) is obtained by subtracting the data. Furthermore, the spectrum analysis may be performed again on the remaining waveform to obtain more precise peak locus data (deterministic component). Such spectrum analysis on the residual waveform is performed as many times as necessary.

  Based on the deterministic component d (x) in the waveform data W (x) of the performance intensity (x) weaker than the maximum intensity shown in the lower part of FIG. 5 thus obtained, the attack of the waveform data W (x) A connection start point and a connection end point for shifting from the waveform data W (x) to the waveform data W (f) having the highest intensity are set in an area after the feature portion. An example of the connection start point and the connection end point set in the waveform data W (x) is shown in FIG. 6. In FIG. 6, the connection start point is shown as the smoothing start point SS, and the connection end point is the smoothing point. Shown as the end point SE. Further, a waveform end point WE corresponding to the end point of the attack portion waveform data Wa (x) stored in the waveform memory 23 is set. The area after the attack feature of the waveform data W (x) is the characteristic of the waveform data W (x) when compared from the beginning of the waveform data W (f) with the same attenuation state on the time axis. Is a region after the point where the waveform data becomes faint and timbre-like, and a start point SS, an end point SE, and a waveform end point WE are sequentially set in this region. As a method of determining the region, for example, the level balance of the main components of the spectrum is compared between the waveform data W (x) and the waveform data W (f), and the region after the point where they substantially match is set as the region. Can do. Or, from the earliest point that sounds the same after playing back the waveform data W (x) and waveform data W (f) with the same attenuation state from the same point on the time axis based on the sense of hearing It can be. Furthermore, the waveform data W (x) and the waveform data W (f) having the same attenuation state may be compared, and the waveform may be the first and subsequent points that look the same. Note that the start point SS, end point SE, and waveform end point WE are determined for each waveform data of the plurality of attack portion waveform data Wa (x). In this specification, in order to distinguish whether each data is data in the time domain or in the frequency domain, the first letter of the alphabet is capitalized for the waveform in the time domain, and the component in the frequency domain The first letter of the alphabet is lowercase.

  By setting the start point SS, end point SE, and waveform end point WE, the deterministic component d (f) of the waveform data W (f) becomes component (df1), component (df2), component (df3) The residual waveform R (f) is divided into a waveform (Rf1), a waveform (Rf2), a waveform (Rf3), and a waveform (Rf4). The deterministic component d (x) of the waveform data W (x) is divided into a component (dx1), a component (dx2), a component (dx3), and a component (dx4), and the residual waveform R (x) is Waveform (Rx1), Waveform (Rx2)), Waveform (Rx3), and Waveform (Rx4). Therefore, the attack is based on the deterministic component (df2) and component (df3) of the waveform data W (f) with the highest intensity and the deterministic component (dx1) and component (dx2) of the waveform data W (x). The waveform data Da (x) of the deterministic component of the partial waveform data Wa (x) is created.

FIG. 7 shows a conceptual diagram of processing for creating the waveform data Da (x). In FIG. 7, by performing a smoothing process on the deterministic component (dx2) of the waveform data W (x), an intermediate deterministic component (di2) corresponding to the connection portion (deterministic component) of the third waveform data. Form. In the smoothing process, the final phase and amplitude of the resulting intermediate deterministic component (di2) is changed to the first phase and amplitude of the deterministic component (df3) of the connected waveform data W (f). The phase and amplitude of each of the spectral components that can be matched with any of the spectra that make up the deterministic component (df3) out of all the spectra that make up the deterministic component (dx2) The trajectory is adjusted in the time axis direction or level direction. Details thereof will be described later with reference to FIGS. As a method of forming the intermediate deterministic component (di2), for example, the waveform data W (f) indicated by the broken line from the deterministic component (dx2) of the waveform data W (x) is not used for the smoothing process. It may be formed by crossfading the deterministic component (df2).
When the intermediate deterministic component (di2) is formed, the deterministic component (dx1) and the waveform data W (f) of the waveform data W (x) arranged on the time axis as shown in the second stage of FIG. An intermediate deterministic component (di2) is placed between and connected to the deterministic component (df3). Of these three components, the deterministic component (dx1) corresponds to the attack feature, the intermediate deterministic component (di2) corresponds to the connection portion, and the deterministic component (df3) corresponds to the same characteristic portion. The connected state is shown in the third row of FIG. 7. The deterministic component (dx1), the intermediate deterministic component (di2), and the deterministic component (df3) connected and arranged on the time axis. , The waveform data Da (x) of the deterministic component of the attack portion waveform data Wa (x) is created.

  The generated waveform data Da (x), the residual waveform (Rf2) and waveform (Rf3) of the waveform data W (f), and the residual waveform (Rx1) and waveform (Rx2) of the waveform data W (x) By performing weighted synthesis, attack portion waveform data Wa (x) which is the fourth waveform data is created. FIG. 8 shows a conceptual diagram of processing for creating the attack portion waveform data Wa (x). As shown in FIG. 8, from the start point SS to the waveform end point WE, the weight of the waveform data Da (x) is 100% (0 dB), and the weight of the residual waveform (Rf2) of the waveform data W (f) The starting point SS is 0% (−∞ dB), and is increased (fade in) toward the ending point SE. The ending point SE is 100% (0 dB), and the weight of the waveform (Rf3) is ended. 100% (0 dB) from the point SE to the waveform end point WE. The weight of the remaining waveform (Rx1) of the waveform data W (x) is 100% (0 dB) from the beginning to the start point SS, and the weight of the waveform (Rx2) is 100% (0 dB) at the start point SS. Then, it decreases (fades out) toward the end point SE and becomes 0% (−∞ dB) at the end point SE. Here, the waveform obtained by adding and combining the residual waveform (Rx2) faded out from the start point SS to the end point SE and the residual waveform (Rf2) faded in corresponds to the connection portion (residual waveform) of the third waveform data. To do. Also, since the fade-in and fade-out here are for connecting two different waveforms, the weight change curves each change linearly on a decibel scale so that the volume after mixing is kept almost constant. It is good to do. Waveform data Da (x) thus weighted, residual waveform (Rf2) and waveform (Rf3) of waveform data W (f), residual waveform (Rx1) and waveform (Rx2) of waveform data W (x) Are combined to create attack waveform data Wa (x). The waveform data from the end point SE to the waveform end point WE in the attack part waveform data Wa (x) is the waveform obtained by inverse Fourier transforming the deterministic component (df3) of the waveform data W (f), and the waveform data W Since it is the sum of the residual waveform (Rf3) in (f), it should have the same shape as the waveform data from the end point SE to the waveform end point WE in the waveform data W (f). The crossfade waveform data (same characteristic part) in the attack part waveform data Wa (x) is an area from the end point SE to the waveform end point WE.

  Here, the smoothing process will be described with reference to FIGS. FIG. 9 shows a process related to the phase of the y-th component which is one of the spectra constituting the deterministic component d (x) in the smoothing processing of the deterministic component d (x) of the waveform data W (x). The horizontal axis is the time axis, and the vertical axis is the phase. In the phase smoothing process, by shifting the phase from the start point SS to the end point SE among the phases of the y-th component of the deterministic component d (x) in the phase direction or expanding and contracting in the time axis direction, The phase at the end point SE is adjusted to match the first phase of the corresponding y-th component of the deterministic component (df3) to which it is connected. The shift in the phase direction is performed by adding a phase offset that smoothly changes during the period from the start point to the end point SE, or by multiplying the phase coefficient that changes in the same way. In the expansion and contraction in the time axis direction, first, a point where the phase of the y-th component of the deterministic component d (x) matches the first phase in the vicinity of the end point SE is detected, and the point matches the end point SE. Thus, the phase from the start point SS of the y-th component of the deterministic component d (x) to that point is expanded or contracted in the time axis direction. The adjustment may be performed by combining a shift in the phase direction and expansion and contraction in the time axis direction. The phase smoothing process is performed for all components that can be associated with any one of the deterministic components (df3) among all the components of the deterministic component d (x), and the deterministic component d ( The phase of each component of x) coincides with the first phase of the deterministic component (df3) at the end point SE.

FIG. 10 shows the amplitude of the y-th component which is one of the spectra constituting the deterministic component d (x) in the smoothing process of the deterministic component d (x) of the waveform data W (x). The horizontal axis is the time axis and the vertical axis is the amplitude. In the amplitude smoothing process, the amplitude from the start point SS to the end point SE of the amplitude of the y-th component of the deterministic component d (x) is shifted in the amplitude direction or expanded or contracted in the time axis direction. The amplitude at the end point SE is adjusted to match the first amplitude of the corresponding y-th component of the deterministic component (df3) to which it is connected. The shift in the amplitude direction is performed by adding an amplitude offset that smoothly changes during the period from the start point to the end point SE, or by multiplying the amplitude coefficient that changes. In the expansion and contraction in the time axis direction, first, a point where the amplitude of the y-th component of the deterministic component d (x) matches the first amplitude is detected in the vicinity of the end point SE, and the point matches the end point SE. Thus, the amplitude from the start point SS of the y-th component of the deterministic component d (x) to that point is expanded or contracted in the time axis direction. The adjustment may be performed by combining the shift in the amplitude direction and the expansion and contraction in the time axis direction. In the amplitude smoothing process, the above-described process is performed on all components that can be associated with any one of the deterministic components (df3) among all the components of the deterministic component d (x). In addition, for the components that could not be matched among the deterministic components (df3), a process of reducing (fade out) between the start point SS and the end point SE is performed, and among the deterministic components (df3) For components that could not be associated with each other, a process of raising (fade in) between the start point SS and the end point SE is performed. In this way, the amplitude of each component of the deterministic component d (x) becomes completely coincident with the initial amplitude of the deterministic component (df3) at the end point SE.
9 and 10 each show the phase or amplitude in an exaggerated manner. Actually, the phase or amplitude increases as shown between the waveform data W (f) and the waveform data W (x). Will not slip. In addition, instead of performing the smoothing process by the shift process or the control of the time axis as described above, the phase and amplitude of each component that can be associated with the deterministic component (dx2) and the deterministic component (df2) You may make it carry out by cross fade.

By the way, the attack part waveform data is simplified by using the crossfading (smoothing process in the time domain) of the waveform data instead of using the precise method by the smoothing process in the frequency domain using the frequency analysis process as described above. It can also be created in a simple way. The attack part waveform data thus created will be described as Wa ′ (x).
When creating the attack portion waveform data Wa ′ (x) by performing crossfading of the waveform, first, the waveform data W (f) and the waveform data W (x) are attenuated as shown in FIG. Arrange on the time axis so that the amplitude levels that follow are approximately the same, and in the area after the attack feature of the waveform data W (x), the waveform data W (f) with the highest intensity from the waveform data W (x) The crossfade start point SS, end point SE, and waveform end point WE for shifting to are set as shown in FIG. In this case, the setting of the start point SS, the end point SE, and the waveform end point WE can be determined by hearing or by comparing the waveform shape.

  By setting the start point SS, end point SE, and waveform end point WE, the waveform data W (f) becomes the waveform data (Wf1), waveform data (Wf2), waveform data (Wf3), and waveform data (Wf4). The waveform data W (x) is divided into waveform data (Wx1), waveform data (Wx2), waveform data (Wx3), and waveform data (Wx4). The attack part waveform data Wa ′ (x) is created by crossfading the waveforms of these divided areas. FIG. 12 shows a conceptual diagram of processing for creating the attack portion waveform data Wa ′ (x). As shown in FIG. 12, the weight of the waveform data (Wf1) in the waveform data W (f) is 0% (−∞ dB) until the start point SS, and the weight of the waveform data (Wf2) is 0% at the start point SS. Ascending (fade in) toward the end point SE, the end point SE is set to 100% (0 dB), and the weight of the waveform data (Wf3) is set to 100% from the end point SE to the waveform end point WE. The In addition, the weight of the waveform data (Wx1) in the waveform data W (x) is 100% from the beginning to the start point SS, and the weight of the waveform data (Wx2) is 100% at the start point SS toward the end point SE. It decreases (fades out) and is 0% at the end point SE, and the weight of the waveform data (Wx3) is 0% from the end point SE to the waveform end point WE. Since the fade-in and fade-out here are also for connecting two different waveforms, it is preferable to use change curves that change linearly on a decibel scale so that the volume after mixing can be kept substantially constant. The weighted waveform data W (x) and the waveform data W (f) are weight-synthesized to create attack portion waveform data Wa ′ (x) as shown in FIG. The process of creating the attack part waveform data Wa ′ (x) by performing the waveform cross-fading can be performed more easily than the smoothing process, but the sound quality is slightly inferior. Of the formed attack waveform data Wa '(x), the waveform data from the beginning to the start point SS (same shape as Wx1) is faded out to the attack feature from the start point SS to the end point SE ( Waveform data obtained by adding and synthesizing Wx2) and the waveform data (Wf2) to be faded in corresponds to the connection portion, and waveform data from the end point SE to the waveform end point WE (the same shape as Wf3) corresponds to the same characteristic portion.

Next, processing for generating a musical tone of performance intensity (x) in the musical tone generating apparatus 1 according to FIG. 1 will be described. The present invention relates to formation of waveform data to be stored in a waveform memory and generation of musical sounds using the waveform memory. In musical tone generation, two of the four waveform data appearing in the process of waveform data formation. Only waveform data (first waveform data, fourth waveform data) is used. Therefore, it is conceivable to renumber only the waveform data to be used for the musical sound generation, but this is not done in this specification in order to maintain the correspondence with the waveform data formation process.
The waveform memory 23 stores a waveform set shown in FIGS. 3 to 4 for each tone range of each tone color. FIG. 13 shows the waveform data W (f) having the highest intensity selected as the first waveform data included in one of the waveform sets selected to generate the combined waveform data CW (x), The attack portion waveform data Wa (x), which is waveform data, is shown. As shown in the figure, a first point (first p) is set in the attack portion waveform data Wa (x), and a second point (second p) is set in the first waveform data W (f). The first point on the attack waveform data Wa (x) is the crossfade waveform data (same characteristic part) for connecting to the second and subsequent points of the first waveform data W (f). Corresponding to the end point SE described above, the same characteristic portion is waveform data from the end point SE to the waveform end point WE. That is, the waveform characteristics after the first point in the attack portion waveform data Wa (x) are the same as the waveform characteristics after the second point of the first waveform data W (f). The first point and the second point are respectively determined for each waveform data of the plurality of attack portion waveform data Wa (x) that is the fourth waveform data. Further, the third point (third p) is set in the first waveform data W (f), and in the first waveform data W (f), from the head of the attack portion waveform data Wa (x) to the first point. The point located before the second point by the length of is the third point.

FIG. 14 shows a timing chart in the case where a musical sound is generated in the musical sound generating apparatus 1 including the sound source 20 according to the present invention in response to a note-on instructing the start of musical sound generation.
In FIG. 14, it is assumed that note-on occurs at timing t0. In response to this, the CPU 10 determines the waveform data Wa (x) and W (f) to be used for sound generation according to the note number and velocity, assigns the sound channel and uses the first sound channel to be used for sound generation. The second tone generation channel is determined, the parameters and link information of the first and second tone generation channels are set in the control register 38, and the note-on signal is supplied to the first tone generation channel. The parameters of the first tone generation channel set here include a start address indicating the beginning and an end address indicating the end of the attack portion waveform data Wa (x). The set parameters of the second channel include a start address indicating the second point of the waveform data W (f) and an end address indicating the end. Further, the set link information indicates that a sound generation trigger is applied to the second channel when the read address reaches the first point in the first channel.
The start address and the end address are set by causing the reading unit 32 to sequentially read the samples of the waveform data Wa (x) at the time division channel timing of the first tone generation channel and at the time division channel timing of the second tone generation channel. This is for sequentially reading the samples of the waveform data W (f). Hereinafter, the reading operation of the first sounding channel is referred to as a first reading unit, and the reading operation of the second sounding channel is referred to as a second reading unit. The first sound channel and the second sound channel are controlled so that the F number generated by the F number generator 31 is always the same. That is, at the time of note-on, the same note number NN, the same pitch data PD, and the same original pitch are set, and when the pitch data PD is changed during music generation, the same value is used by using the multiple simultaneous writing unit 39. Pitch data PD is set for both sound channels simultaneously (same sampling period).

  In response to the note-on signal from the CPU 10, the first reading unit starts reading the attack portion waveform data Wa (x) from the first sample. That is, the reading unit 32 accumulates the supplied F number at the time division channel timing of the first sound generation channel, and reads out at a speed corresponding to the F number starting from the beginning of the waveform data Wa (x). An address is generated, and samples of waveform data Wa (x) are sequentially read from the waveform memory 23 based on the generated read address. When the first reading unit (first sounding channel) reaches the timing t1 for reading the first point of the attack part waveform data, the second reading unit (second sounding channel) is set based on the set link information. On the other hand, a trigger is supplied, and the second reading unit starts reading the waveform data W (f) from the second point. When the trigger is generated, the reading unit 32 of the sound source 20 causes the second sounding channel (second reading unit) to take over the fractional part of the F number accumulated in the first sounding channel (first reading unit). A takeover function is provided, and the phase of the sample of the waveform data to be read can be completely synchronized between the first reading unit and the second reading unit. Note that this takeover function is not essential and can be omitted. In this case, the synchronization between the first reading unit and the second reading unit is only the integer part of the phase, and the interpolation sample is output with a shift in the decimal part. May have slightly different values. In this way, at the timing t1, the second reading unit starts reading the waveform data W (f) in synchronization with the reading of the attack portion waveform data Wa (x) by the first reading unit.

  Further, the Xfade control unit 34 is set to 100% (1.00) at the time of note-on (timing t0) at the time division channel timing of the first sound generation channel based on the crossfade control data set in the control register 38. When the read address of the first tone generation channel reaches the first point (timing t1), it starts to decrease (fade out), and after a predetermined crossfade time (timing t2), 0% (0.00) The first Xfade coefficient is generated and is set to 0% (0.00) at the time of note-on (timing t0) at the time division channel timing of the second sound channel, and the read address of the first sound channel is set to the first point. When it reaches (timing t1), it starts to rise (fade in), and after a predetermined crossfade time (timing t2) 100% (1.00 To generate the second Xfade coefficient. In the multiplier M2, at the time division channel timing of the first tone generation channel, the sample of the attack portion waveform data Wa (x) output from the first readout portion and given the volume change by the multiplier M1 is multiplied by the first Xfade coefficient. Thus, waveform data of the first channel is formed, which starts sounding from the time of note-on (timing t0), starts fading out at timing t1, and is muted at timing t2. Also, at the time t1 of the second sound generation channel, the second Xfade coefficient is multiplied by the sample of the waveform data W (f) output from the second reading unit and volume-controlled by the multiplier M1, and the sound generation starts at the timing t1. As a result, the waveform data of the second channel which is faded in and then continues to generate sound is formed. In the mixer 35, the formed waveform data of the first channel and the waveform data of the second channel are synthesized, and waveform data CW (x) which is synthesized waveform data related to the note-on is generated. Of the final synthesized waveform data CW (x), from note-on (timing t0) to timing t1, it is formed based on the attack waveform data Wa (x), and from timing t1 to timing t2, the attack waveform It is formed based on the crossfade waveform from the data Wa (x) to the waveform data W (f), and after the timing t2, it is formed based on the waveform data W (f). Since the attack portion waveform data Wa (x) is not read after timing t2, the first tone generation channel is released when timing t2 is reached.

In the period from the timing t1 to the timing t2 when the crossfade is performed, the attack part waveform data Wa (x) and the first waveform data W (f) have the same characteristics (or almost the same characteristics) as described above. It has become. As described above, since the crossfading is performed in the waveform data region (same characteristic portion) having the same characteristic, no problem in hearing is caused regardless of the timing, speed, change curve, and the like of the crossfade. As in the embodiment, it is not necessary to perform the crossfade over the entire period from the timing t1 to the timing t2.
In this embodiment, since the connection is performed by crossfading, even when the waveform data in the first sound channel and the waveform data in the second sound channel are slightly different at the timing of reading out the same characteristic portion, noise is generated at the time of connection. Does not occur. As described above, it has been described that the takeover function from the first sound generation channel to the second sound generation channel is not necessarily required because there is a crossfade at the time of connection. Further, when each tone generation channel is provided with a timbre filter, it is possible to make the filter coefficient different between the first tone generation channel and the second tone generation channel. On the contrary, without the crossfade, the output of the first sound channel is changed from 100% to 0% and the output of the second sound channel is changed from 0% to 100% in one sampling period immediately after timing t1, and the channel Can be considered. In that case, the handover function is essential and the filter coefficients must be the same.

By the way, when the waveform data W (f) having the maximum intensity is selected by note-on and the waveform set having the configuration shown in FIG. 3 is stored in the waveform memory 23, only one tone generation channel is assigned. In one assigned channel, the waveform data W (f) is read from the rising edge to the falling edge and becomes the combined waveform data CW (x). Further, when the waveform set having the configuration shown in FIG. 4 is stored in the waveform memory 23, as described above, two tone generation channels are allocated and the above-described processing is performed.
In the above description, at the timing when the first reading unit reads the first point of the attack portion waveform data Wa (x), the second reading unit starts reading the first waveform data W (f) from the second point. Thus, the first waveform data W (f) is synchronized with the attack portion waveform data Wa (x). As described above, since the same F number is supplied to the first reading unit and the second reading unit, the phase synchronized at a certain point in time does not shift even if the pitch is changed thereafter. Regarding the reading start timing of the second reading unit, the first reading unit starts reading the attack portion waveform data Wa (x) at the timing t0, and the second reading unit sets the first waveform data W (f) to the first time. You may make it start reading from 3 points | pieces. In this case, a note-on signal is supplied to the first and second reading units simultaneously (same sampling period), and reading is started simultaneously. A trigger for sound generation from the first reading unit to the second reading unit is not supplied, and the phase takeover function as described above is unnecessary.

Next, FIG. 15 shows a flowchart of note-on event processing executed when a note-on event instructing the start of sound generation of a new musical tone occurs in the sound source 20 shown in FIG. However, it is assumed that a waveform set having the configuration shown in FIG. 3 is stored in the waveform memory 23 for each tone range of each tone color.
When note-on is instructed to the sound source 20, note-on event processing is started. In step S10, a note number indicating the pitch at note-on is stored in the register NN, and a velocity indicating the playing intensity is stored in the register VEL. In this specification, the parameter stored in the register X is simplified and described as the parameter X. Next, in step S11, the velocity stored in the register VEL is converted into a volume value indicating the volume by referring to the touch curve (T curve) and stored in the register tvol. In step S12, among the waveform sets stored in the waveform memory 23, a waveform set corresponding to the tone range to which the note number NN belongs is selected from a plurality of waveform sets corresponding to the timbre currently selected in the part. Further, a waveform (W (f) to Wa (x)) corresponding to the velocity VEL is selected from the selected waveform set. Next, it is determined in step S13 whether or not the waveform selected in step S12 is the waveform data W (f) having the highest intensity in the waveform set.

  Here, if it is determined in step S13 that the waveform data W (f) having the highest intensity has been selected, the process proceeds to step S14, and a plurality of tone generation channels possessed by the sound source 20 are generated in order to generate a musical sound related to the note-on. One of the sound generation channels is assigned, and the start address (SA) indicating the beginning of the waveform data W (f) having the highest intensity, the end address (EA) indicating the end, and the original sound are assigned to the sound generation channels assigned in step S15. High and set. In step S16, the volume of the difference between the volume value tvol and the volume value wvol of the selected waveform is set in the sound generation channel. As a result, the sound channel is sounded at a volume corresponding to the volume value tvol. Next, in step S17, the note channel NN, pitch data PD indicating the pitch deviation, amplitude envelope parameter corresponding to the velocity VEL, crossfade control data for setting the crossfade coefficient to 100%, etc. Parameters for musical tone control are set. Then, a sounding start instruction (supply of a note-on signal) for the sounding channel is performed in step S18. As a result, in the sound generation channel of the sound source 20, the waveform data W (f) is read from the waveform memory 23 from the head indicated by the start address at a speed corresponding to the note number NN, the pitch data PD, and the original pitch. Then, the maximum intensity composite waveform data CW (f) whose characteristics are controlled based on the read waveform data and in accordance with the musical sound control parameters is generated. Next, in step S19, “1”, which is the number of sounding channels assigned this time, is set in the channel number register (NN) corresponding to the note number NN, and the note-on event processing is completed. Reading of the waveform data W (f) in the sound generation channel is continued until the amplitude envelope waveform is completely attenuated or until the read address reaches the end address.

  If it is determined that the waveform selected in step S13 is not the highest-intensity waveform data W (f) in the waveform set, the process branches to step S20 and the first channel (# 1) and second channel ( Sound channel # 2) is assigned. Next, the start address (SA) indicating the beginning of the attack part waveform data Wa (x) selected in accordance with the velocity VEL and the end address indicating the end are assigned to the first channel (# 1) assigned in step S21. EA) and original pitch are set, and the second point of the waveform data W (f) with the highest intensity of the waveform set to which the attack waveform data Wa (x) belongs is assigned to the assigned second channel (# 2). The start address (SA) to be indicated, the end address (EA) to indicate the end, and the original pitch are set. Here, the position of the second point differs according to the selected attack portion waveform data Wa (x). The set original pitches are the same for the attack portion waveform data Wa (x) and the waveform data W (f). Next, in step S22, the volume of the difference between the volume value tvol and the volume value wvol of the attack waveform data Wa (x) is set in the first channel (# 1), and the second channel (# 2). ), The volume of the difference between the volume value tvol and the volume value wvol at the third point of the waveform data W (f) is set. Thus, the volume value tvol indicates the volume of the musical tone waveform of the body part generated in the second channel (# 2) and the volume of the musical sound waveform of the attack part generated in the first channel (# 1). Can match.

In step S23, the assigned first channel (# 1) and second channel (# 2) are assigned note numbers NN, pitch data PD indicating pitch deviation, amplitude envelope parameters corresponding to velocity VEL, and the like. Parameters of the same value for various musical tone controls are set. Further, as crossfade control data, data for setting the crossfade coefficient to 100% is set for the first channel (# 1), and data for setting the crossfade coefficient to 0% for the second channel (# 2). Is set. Next, in step S24, the control register 38 has a trigger channel number indicating the first channel (# 1) on which the trigger is applied, a link address indicating the first point to be triggered, and the second channel to be triggered. Is set as link information. Then, an instruction to start sounding (supplying a note-on signal) to the first channel (# 1) is performed in step S25. Thereby, in the first channel (# 1) of the sound source 20, the attack address waveform data Wa (x) from the waveform memory 23 is indicated by the start address at a speed corresponding to the note number NN, the pitch data PD, and the original pitch. The composite waveform data CW (x) of the performance intensity (x) read from the head (first reading unit) and subjected to the characteristic control based on the read waveform data and in accordance with the parameters for tone control An attack part is generated. Next, in step S26, “2”, which is the number of sounding channels assigned this time, is set to the number of channels (NN) which is the channel number register, and the note-on event processing is ended.
When the read address in the first channel (# 1) indicated by the trigger channel number reaches the first point of the attack waveform data Wa (x) indicated by the link address, the second channel (indicated by the triggered channel number) A trigger (including a fractional part of the phase) is given to # 2). In response to the trigger, the second channel (# 2) of the sound source 20 uses the highest-intensity waveform data W (f) stored in the waveform memory 23 in accordance with the note number NN, the pitch data PD, and the original pitch. At the speed, reading starts from the second point indicated by the start address (second reading unit). As a result, in the second reading unit, reading of the waveform data W (f) to be connected to the attack unit waveform data is started in synchronization with the reading of the attack unit waveform data Wa (x) by the first reading unit. The At this time, since the crossfade coefficient of the first channel (# 1) is 100% and the crossfade coefficient of the second channel is 0%, the combined waveform data CW (x) is only the waveform data of the first channel. However, the phases of the two waveform data to be read out are controlled to be the same, and the volume of the waveform data (the waveform data before being multiplied by the crossfade coefficient) input to the multiplier M2 is also controlled. Are controlled to be the same as each other.

  When the read address of the first channel (# 1) reaches the first point indicated by the link address, the sound source 20 interrupts the CPU 10 with a link point. FIG. 16 is a flowchart of a link point interrupt process executed by the CPU 10 in response to the interrupt. When the link point interrupt is applied, the CPU 10 first sets the cross-fade control data instructing the fade-in start of the second channel (# 2) in the control register 38 in step S30, and then proceeds to step S31. Then, the cross-fade control data for instructing the start of fade-out of the first channel (# 1) is set, and the link point interrupt process is completed. Thereby, as shown in FIG. 14, the crossfade coefficient of the second channel (# 2) gradually increases from 0% to 100%, and the body portion of the synthesized waveform data CW (x) generated in the second channel Is faded in, the crossfade coefficient of the first channel (# 1) gradually decreases from 100% to 0%, and the attack part of the synthesized waveform data CW (x) generated in the first channel is faded out. The In the mixer 35, the musical tone waveform of the attack portion faded out at the link point and the musical tone waveform of the body portion faded in from the link point are mixed, and from the rise to the fall corresponding to one note-on. Composite waveform data CW (x), which is waveform data, is generated.

  When the crossfade coefficient of the first channel (# 1) generated by the crossfade control unit 34 reaches 0%, the sound source 20 applies a crossfade end interrupt to the CPU 10. FIG. 17 is a flowchart of a fade-out end interrupt process executed by the CPU 10 in response to the interrupt. When the cross fade end interrupt is applied, the CPU 10 releases the first channel (# 1) that has been faded out in step S40 so that it can be used for new sound generation. Next, in step S41, "1" indicating that one assigned channel remains is set in the register of the number of channels (NN) corresponding to the note number NN of the musical tone generated in the channel. End the fade-out end interrupt process. As described above, the first channel (# 1) is released as soon as the generation of the attack portion of the combined waveform data CW (x) is completed, so two sound generation channels are assigned to one note-on. However, the actual number of simultaneous pronunciations does not decrease so much.

Note that the waveform data of the same characteristic portion in the attack portion waveform data Wa (x) has the same shape as the second and subsequent points of the waveform data W (f), as described above, so that it does not crossfade. No noise is generated even when switching. For example, in the sound source 20, at the timing when the first channel (# 1) reaches the link point, the second channel (# 2) is triggered to start sounding, and the crossfade control unit 34 itself performs the trigger. Accordingly, the crossfade coefficient of the first channel (# 1) may be set to 0%, and the crossfade coefficient of the second channel (# 2) may be set to 100%. In this case, since the same characteristic part of the attack part waveform data Wa (x) is not necessary, it can be made shorter waveform data not including the same characteristic part. Moreover, the link point interruption process by CPU10 is also unnecessary.
Alternatively, in the link point interrupt process corresponding to the link point interrupt, the CPU 10 sets the crossfade coefficient of the first channel (# 1) of the crossfade control unit 34 to 0% and the second channel (# The crossfade coefficient of 2) may be set to 100%. In that case, the same characteristic portion corresponding to the delay of the interrupt processing is required for the attack portion waveform data Wa (x).

Next, a flowchart of the pitch change event process is shown in FIG.
When the pitch bend operator provided in the musical sound generating device 1 is operated, or when the CPU 10 performs processing for generating a vibrato waveform or a pitch envelope waveform, a pitch change event for instructing a dynamic change in the pitch of the musical sound. Will occur. FIG. 18 is a flowchart of pitch change event processing executed by the CPU 10 in response to the pitch change event. A pitch change event such as pitch bend in MIDI is an event for controlling the pitch change for each part. When a pitch change event occurs, first, a new pitch changed in step S50 is stored in the register PD. Next, in step S51, the note number (NN) currently sounded in the part is detected, and it is determined in step S52 whether or not there is a note number being sounded. If it is determined that there is no note number being sounded, the pitch change event process is terminated, and if it is determined that there is a note number being sounded, the process proceeds to step S53 and the detected note number is detected. The first note number is selected. Next, whether the number of sound generation channels set in the note number selected in step S54 is set to “1” or “2” with reference to the register of the channel number (NN) of the note number. To be judged. Here, if it is determined that the number of sound generation channels is “1” in the note number, the process proceeds to step S55, and the control channel 38 stores the sound channel assigned to the sound generation of the note number. As the pitch data, the pitch data stored in the register PD is written. On the other hand, if it is determined in step S54 that the number of sound generation channels is “2” in the note number, the process branches to step S57, and a plurality of simultaneous writing sections 39 are used to enter the control register 38. The pitch data stored in the register PD is simultaneously (same as the two pitch data of the two channels of the first channel (# 1) and the second channel (# 2) assigned to the tone number generation of the note number). (Sampling period).

  When the process of step S55 or step S57 ends, the process proceeds to step S56, and it is determined whether or not a note number that has not been processed remains. If it is determined that the note number remains, the process branches to step S58, the next note number is selected, the process returns to step S54, and the processes from step S54 to step S57 are repeated. This iterative process is performed until there is no unprocessed note number being processed, and it is determined in step S56 that all the note numbers being sounded in the part are processed and there is no unprocessed note number. Then, the pitch change event process ends. With this pitch change event processing, when the pitch is changed, all note numbers that are sounding in the part are assigned two sounding channels or one sounding channel. The pitch of the composite waveform data CW (x) can be changed with time. Thereby, it is possible to dynamically change the pitch of the musical sound waveform of all the sound generation channels that are sounding in the part. Furthermore, since the pitch data is simultaneously written in the first channel (# 1) and the second channel (# 2) that generate the composite waveform data CW (x) of the performance intensity (x), the pitch Even when is changed dynamically, the phase of the waveform data read by the first channel (# 1) and the second channel (# 2) does not shift.

As described above, the timing at which the second reading unit is synchronized may be the timing at which the first reading unit starts reading the head of the attack portion waveform data Wa (x). In this case, among the processes from step S20 to step S26 in the note-on event process of FIG. 15, steps S21, S24, and S25 may be changed as follows.
First, in step S21, a start address (SA) indicating the beginning of the attack waveform data Wa (x) selected according to velocity VEL and an end address indicating the end are assigned to the assigned first channel (# 1). EA) and original pitch are set, and the third point of the waveform data W (f) having the highest intensity of the waveform set to which the attack waveform data Wa (x) belongs to the assigned second channel (# 2) is shown. The start address (SA), end address (EA) indicating the end, and original pitch are set. In step S24, the trigger channel number indicating the first channel (# 1) on the triggering side and the link address indicating the first point to be triggered are set as link information, and the triggered channel number is not set. Therefore, the trigger from the first channel (# 1) to the second channel (# 2) is not supplied. In step S25, an instruction to start simultaneous sounding is supplied to the first channel (# 1) and the second channel (# 2) (note-on signal is supplied at the same sampling period).
Thus, in response to the note-on signal supplied in step S25, the attack section waveform data Wa (x) is read from the head indicated by the start address (first) in the first channel (# 1) of the sound source section 20. 1 reading section) and the second channel (# 2), the waveform data W (f) having the highest intensity is read from the third point indicated by the start address (second reading section). At this time, since the crossfade coefficient of the first channel (# 1) is 100% and the crossfade coefficient of the second channel is 0%, the combined waveform data CW (x) is only the waveform data of the first channel. It becomes. When the read address of the first channel (# 1) reaches the first point indicated by the link address, the link point interrupt process is executed, and the tone waveform of the attack part generated in the first channel is faded out, Instead, the tone waveform of the body part generated in the second channel is faded in. The generated musical sound waveforms are mixed by the mixer 35 to become synthesized waveform data CW (x).

In the waveform data forming method described above, when the waveform set having the configuration shown in FIG. 4 is stored in the waveform memory 23, even if the waveform selected according to the performance intensity is the highest intensity waveform. The same note-on event processing is performed even if the waveform has a performance intensity weaker than the maximum intensity. In this case, the processing of step S13 to step S19 in the note-on event processing shown in FIG. 15 is not necessary, and the processing of step S20 to step S26 is subsequently performed when the processing of step S12 is completed. As described above, if the waveform set having the configuration shown in FIG. 4 is stored in the waveform memory 23, two tone generation channels are assigned to note-on regardless of the performance intensity. Different note-on event processing depending on intensity is not performed.
In addition, when the musical sound is generated using the first and second channels of the sound source 20, the read start address of the first waveform data in the second channel is the second point to the third point in FIG. The read start address can be set to any point in the range from the third point to the fourth point, with the point after the length of the crossfade waveform data from the second point being the fourth point. In this case, the link address may be an address corresponding to the set read start address of the fourth waveform data.

In the waveform data forming method described above, the attack waveform data Wa (x) having four types of performance intensities are stored in the waveform memory 23. However, the number of performance intensities is not limited to four and corresponds to any number of performance intensities. The attack portion waveform data Wa (x) may be formed and stored in the waveform memory 23.
In the above waveform data forming method, the strongest waveform data W (f) switched from the attack waveform data Wa (x) is read from any point before the second point (end point SE). You may start. However, it is necessary to synchronize with the attack part waveform data Wa (x) at the second point.

1 musical tone generator, 10 CPU, 11 flash ROM, 12 RAM, 13 timer, 14 HDD, 15 MIDI I / O, 16 panel SW, 17 panel display, 18 writing circuit, 19 buffer, 20 sound source, 21 sound system , 22 Access management unit, 23 Waveform memory, 24 External waveform input terminal, 25 Bus line, 31 F number generation unit, 32 Reading unit, 33 Amplitude change generation unit, 34 X fade control unit, 35 Mixer, 36 DSP, 37 DAC I / O, 38 control register, 39 multiple simultaneous write unit, M1 multiplier, M2 multiplier

Claims (5)

A plurality of first waveform data having a plurality of connection points corresponding to a plurality of performance intensities and a plurality of readout start addresses corresponding to the plurality of performance intensities, all waveforms having the strongest performance intensity; A plurality of fourth waveform data corresponding to the performance intensity of the first, each of the attack feature in the waveform data of the corresponding performance intensity, and the subsequent connection point corresponding to the performance intensity of the first waveform data. A waveform memory in which a plurality of fourth waveform data including a connection portion for connection is stored;
A selection unit for selecting any one of the plurality of fourth waveform data according to the performance intensity indicated by the performance intensity data included in the supplied pronunciation instruction;
In response to the sound generation instruction, the fourth waveform data selected by the selection unit is read from the head address, and in synchronization with the reading, the first waveform data is read from the plurality of read start addresses. A reading unit for reading from a reading start address corresponding to the performance intensity;
At a predetermined point after the end of the connecting portion of the fourth waveform data read by the reading portion, the first waveform data that has been read out is connected to the fourth waveform data, and the connected waveform data is connected to the connected waveform data. A musical sound forming unit for forming musical sound waveform data based on the
A waveform memory sound source comprising: A plurality of first waveform data having a plurality of connection points corresponding to a plurality of performance intensities and a plurality of readout start addresses corresponding to the plurality of performance intensities, all waveforms having the strongest performance intensity; A plurality of fourth waveform data corresponding to the performance intensity of the first, each of the attack feature in the waveform data of the corresponding performance intensity, and the subsequent connection point corresponding to the performance intensity of the first waveform data. A waveform memory in which a plurality of fourth waveform data including a connection portion for connection is stored;
A selection unit that selects any one of the first waveform data and the plurality of fourth waveform data according to intensity data included in the supplied sound generation instruction;
When the first waveform data is selected by the selection unit, the selected first waveform data is read from the start address from the waveform memory, and any of the fourth waveform data is selected by the selection unit In this case, the selected fourth waveform data is read from the waveform memory, and in synchronization with the reading, the first waveform data is read corresponding to the performance intensity of the plurality of reading start addresses. A reading unit that reads from the start address;
When the first waveform data is selected by the selection unit, musical tone waveform data is formed based on the first waveform data read by the reading unit, and any of the fourth waveform data is selected. Connects the first waveform data read to the fourth waveform data at a predetermined point after the end of the connecting portion of the fourth waveform data read by the reading unit, and the connected waveform data A musical tone generator that forms musical tone waveform data based on
A waveform memory sound source comprising:   Each of the plurality of fourth waveform data includes the same characteristic portion having the same characteristic as that of the first waveform data at the connection portion, and the fourth waveform data and the first waveform data in the musical tone forming portion are The connection is established by cross-fading from the first waveform data to the fourth waveform data in the whole or a part of the range of the first waveform data within the same characteristic portion. Waveform memory sound source according to 1 or 2.   2. The reading of the first waveform data in synchronization with the reading of the fourth waveform data is started at a timing when the reading unit finishes reading the connection portion of the fourth waveform data. Or the waveform memory sound source of 2.   3. The waveform according to claim 1, wherein reading of the first waveform data in synchronization with reading of the fourth waveform data is started at a timing when the reading unit starts reading the fourth waveform data. Memory sound source.

Images