JP5146188B2 - Music signal generator - Google Patents

Description

  The present invention relates to a waveform memory type musical tone signal generation apparatus, and in particular, it is possible to improve the access speed of the waveform memory by improving the method of reading data from the waveform memory and improving the data storage method of the waveform memory accordingly. It is a thing.

  2. Description of the Related Art Conventionally, a musical sound signal generating apparatus configured by a waveform memory type sound source (waveform memory sound source) is known. The waveform memory sound source generates a musical tone signal based on waveform data stored one sample at each address of the waveform memory, and performs a time-sharing operation for each sound generation channel, so that a predetermined frequency is obtained every sampling cycle. Musical sound signals for multiple sound channels could be generated simultaneously. That is, the waveform memory is accessed for each time slot corresponding to each sound generation channel obtained by equally dividing one sampling period, and waveform samples are read from the waveform memory. Interpolate between samples using several waveform samples to generate waveform samples for one sampling period for each sounding channel, and accumulate waveform samples for multiple sounding channels for each sampling period Thus, musical tone signals for a plurality of tone generation channels are generated as a whole.

  In the waveform memory sound source having the above configuration, a “single readout type” method has been conventionally used as a method for reading a waveform sample from the waveform memory. In the case of the “single readout type”, one address signal is output for each access to the waveform memory, one address is accessed, and the address signal is stored in one address specified by the address signal. Read waveform samples. A method of specifying a storage position (address) on the waveform memory by an address signal and directly accessing the specified address is generally referred to as random access.

  As another example of a method of reading waveform samples from the waveform memory, there has been a “continuous reading type” method. In the “continuous readout type”, waveform samples are stored in the waveform memory in units of frames in which a plurality of consecutive addresses are one frame, and by one access to the waveform memory, a plurality of addresses corresponding to a plurality of addresses constituting one frame are stored. A plurality of waveform samples are read (see, for example, Patent Document 1 below). Reading a plurality of waveform samples in one access in this way is called burst reading.

Further, in Patent Document 2 below, as a configuration for performing burst reading, a plurality of waveform samples are stored for each address of the waveform memory, and a plurality of samples are stored for each access to the waveform memory. It is described that reading is performed in
JP 2004-126127 A Japanese Patent Laid-Open No. 10-20860

  By the way, as described above, the waveform memory sound source realizes generation of tone signals for a plurality of sound generation channels by time division operation. Therefore, the number of tone generation channels for each sampling period of the sound source is limited by the access speed of the sound source waveform memory. That is, the access speed to the waveform memory is an element that is a bottleneck that limits the number of sound generation channels. Therefore, if the access speed to the waveform memory is improved, the number of sound generation channels can be increased as a result.

  However, the conventional “single-reading type” waveform memory sound source reads waveform samples from one address for each access by random access, so there is a limit to improving the access speed to the waveform memory. It was. As a result, the “single readout type” waveform memory sound source has a disadvantage that the number of sound generation channels is significantly limited.

  Further, when the single readout type waveform memory sound source does not have a buffer for inter-sample interpolation, a waveform memory is used to read out a plurality of waveform samples to be used for interpolation calculation at each readout timing. Must be accessed multiple times. For example, when two-point interpolation (linear interpolation) using two samples is performed, it is necessary to access the waveform memory twice for each reading timing of each sound generation channel in each sampling period. That is, the cost can be reduced by not providing a buffer for inter-sample interpolation, but on the other hand, the waveform sample reading efficiency is deteriorated and the number of sound generation channels is further limited.

  On the other hand, since the conventional “continuous readout type” waveform memory sound source reads a plurality of waveform samples for each access, the access speed to the waveform memory can be improved. A buffer having a data capacity equal to or larger than the data amount (frame size) corresponding to a plurality of waveform samples read out from the waveform memory at every access must be provided in the sound source unit. That is, the conventional method of reading out a plurality of samples in burst requires a buffer having a data capacity corresponding to the data size to be read out in burst, which has the disadvantage that the configuration is complicated and the circuit scale is increased.

  The present invention has been made in view of the above points, and an object of the present invention is to provide a musical tone signal generating apparatus capable of improving the access speed to the waveform memory and thereby increasing the number of sound generation channels.

  The present invention is a musical tone signal generating apparatus for generating musical tone signals of a plurality of sound generation channels for each sampling period based on waveform data stored in a waveform memory, wherein the address space is divided into pages by page size p, and each page The waveform memory is capable of high-speed page access to the data stored in the address, and the waveform data is stored one sample at each address, and the sample stored at the last address of each page is the next The waveform memory that is also stored in duplicate at the top address of the page and the read rate of each sound channel are accumulated for each sampling period, and a sample address consisting of an integer part and a decimal part is generated for each sound channel. Address counter and integer of sample address generated by the address counter for each sampling period The waveform using a memory address of each tone generation channel generated by the memory address generator generated by the memory address generator for each sampling period, and a memory address generator that converts a value corresponding to the final address of each page A page access to the memory, a read unit for reading two samples stored at two consecutive addresses from the memory address, and two for each tone generation channel read by the read unit for each sampling period Are interpolated between samples based on the fractional part of the sample address generated by the address counter, and an interpolator that generates the interpolated sample for each sound generation channel, and the interpolated by the interpolator Multiple samples per sampling period based on samples for each sound channel A musical tone signal generating apparatus characterized by comprising a tone signal generator for generating a musical tone signal of the sound channels.

  The waveform memory is a waveform memory in which the address space is divided into pages by the page size p, and the data stored in the address in each page can be accessed at high speed page, and the waveform data is stored in each address one sample at a time. The sample stored at the last address of each page is also stored redundantly at the top address of the next page. For each sampling period, the readout rate of each sound channel is accumulated to generate a sample address consisting of an integer part and a decimal part for each sound channel, and the integer part of the generated sample address is used as the final part of each page. The value corresponding to the address is converted into a jumping memory address, the waveform memory is page accessed using the memory address of each tone generation channel, and two samples stored at two consecutive addresses from the memory address are read out. Interpolate between the two samples for each read sound channel based on the fractional part of the sample address, and multiple sound channels for each sampling period based on the interpolated samples for each sound channel A musical tone signal is generated. Therefore, two waveform samples required for interpolation can be read by page access of each sound generation channel for each sampling period. Further, since the memory address skips the value corresponding to the final address of each page, two consecutive addresses from the memory address are always addresses in the same page. The memory address no longer points to the final address of each page, but the sample stored at the final address of each page is also stored redundantly at the top address of the next page. And the next sample can be interpolated.

  As an example of the configuration of the memory address generation unit, if the predetermined number of addresses of the page size is p, the integer part SAi of the sample address generated by the address counter and the function of the integer part SAi [SAi / (p−1 )] To generate the memory address MA based on the result of addition of the integer part, the integer part of the sample address can be converted into a memory address that jumps over the value corresponding to the final address of each page.

  The present invention is also a musical sound signal generating device for generating a musical sound signal of a plurality of tone generation channels for each sampling period based on waveform data stored in a waveform memory, wherein the address space is paged by page size p, A waveform memory capable of high-speed page access to data stored at addresses in each page, storing the waveform data one sample at each address, and n-1 addresses from the last of each page The stored waveform data samples are accumulated in the n-1 address from the beginning of the next page, and the readout rate of each sound channel is accumulated for each sampling period. An address counter that generates a sample address consisting of an integer part and a decimal part for each tone generation channel, and for each sampling period, A memory address generator for converting the integer part of the sample address generated by the address counter into a memory address that jumps over a value corresponding to n-1 addresses from the end of each page; and the memory address generator for each sampling period A read unit that performs page access to the waveform memory using the memory address of each tone generation channel generated by the unit, and reads out n samples stored in consecutive n addresses from the memory address; and each sampling period Each time, n samples for each tone generation channel read by the reading unit are interpolated between samples based on the fractional part of the sample address generated by the address counter, and the interpolated samples are obtained. Interpolated for each tone generation channel and interpolated by the interpolator Based on samples of each sound channel, a musical tone signal generating apparatus characterized by comprising a tone signal generator for generating a musical tone signals of plural tone generation channels in each sampling period.

  When reading a predetermined n number of samples in the page by page access, the waveform data samples stored in the n-1 addresses from the end of each page of the waveform memory are read from the top of the next page. The n-1 addresses are also stored redundantly. The memory address generation unit converts the integer part of the sample address generated by the address counter into a memory address that skips values corresponding to n-1 addresses from the end of each page. Then, at each sampling period, the waveform memory is page-accessed using the memory address of each tone generation channel, and n samples stored in consecutive n addresses from the memory address are read and read. The n samples for each sounding channel are interpolated between samples based on the fractional part of the sample address, and a musical sound signal of a plurality of sounding channels is generated based on the interpolated samples for each sounding channel. .

  When reading a predetermined n number of samples in a page by the above page access, as a configuration example of the memory address generation unit, if the predetermined number of addresses of the page size is set to p, the sample address generated by the address counter A memory address is generated based on a result of adding the integer part SAi and a value obtained by multiplying the integer part of [SAi / (p−n + 1)], which is a function of the integer part SAi, by (n−1). By doing so, the integer part of the sample address can be converted into a memory address that jumps over the values corresponding to n-1 addresses from the end of each page.

According to the present invention, as the waveform memory, the address space is divided into pages by the page size p, and the waveform memory capable of high-speed page access to the data stored in the address in each page is used. Samples stored at n-1 addresses from the first page are also stored redundantly at the top address of the next page, and at the end of each page (n-1 from the last) in each tone generation channel. 2 addresses (n addresses) in the same page of the waveform memory by using page access at the access timing for each sound generation channel within each sampling period. ) To two samples (n samples) can be read out at high speed.
The samples used for inter-sample interpolation of each sound channel are read from the waveform memory at each sampling period, so there is no need for a buffer memory for each sound channel, and the configuration of the tone signal generator is simplified accordingly. Can be
Therefore, according to the musical tone signal generating apparatus of the present invention, there is an excellent effect that the access speed to the waveform memory 9 can be improved and the number of sound generation channels can be increased with a simple configuration.

  Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. In this embodiment, an example in which the waveform memory sound source (musical sound signal generator) according to the present invention is applied to an electronic musical instrument will be described.

  FIG. 1 is a block diagram showing an overall configuration of an electronic musical instrument to which a waveform memory sound source (musical sound signal generating device) of the present invention is applied. In FIG. 1, an electronic musical instrument includes a CPU 1, a flash memory 2, a RAM 3, a performance operator 4, an operator 5, a display 6, a data input / output interface (I / O) 7, a sound source unit 8, a waveform memory 9, and a sound. The system 10 is configured. The CPU 1 is connected to the flash memory 2, RAM 3, performance operator 4, operator 5, display 6, input / output I / O 7, and tone generator unit 8 via the data and address bus line 1 B. Various data including address signals are communicated with the devices connected via the bus line 1B.

  The CPU 1, the flash memory 2, and the RAM 3 are control units that control the operation of the electronic musical instrument. Various programs executed by the CPU 1 are stored in the flash memory 2, and the CPU 1 reads the program stored in the flash memory 2 into the RAM 3 and executes the read program.

  The performance operation element 4 includes an operation element for a user to input a performance operation and a detection mechanism for detecting the performance operation, and is configured by a keyboard composed of a plurality of keys, for example. The user can instruct the musical sound to be generated and stopped by using the performance operator 4 and can instruct the pitch and velocity (volume) of the musical sound to be generated. The CPU 1 receives a detection signal corresponding to the operation of the performance operator 4 via the bus line 1B, and generates performance event data corresponding to the operation of the performance operator 4 based on the received detection signal.

  The operation element 5 includes an operation element for the user to select a tone color, input values of various parameters of the selected tone color, and a detection mechanism for detecting the operation. The various parameters of the timbre include, for example, a volume parameter for controlling the volume of the musical tone, a parameter for setting various effectors for applying an effect to the musical tone, and the like. The user can operate the operator 5 to set various parameters including selection of timbre, adjustment of volume, adjustment of parameter values of various effectors, and the like. The CPU 1 receives a detection signal corresponding to the operation of the operation element 5 via the bus line 1B, and generates various data corresponding to the operation of the operation element 5 based on the received detection signal.

  The display device 6 is a display device that is arranged on the operation panel of the electronic musical instrument and displays various types of information based on display control by the CPU 1. The electronic musical instrument is connected to various external devices via the input / output I / O 7 so as to be able to perform data communication. The input / output I / O 7 is configured by a data input / output interface of an appropriate standard known conventionally, such as a MIDI interface for transmitting / receiving a signal of the MIDI standard or a communication interface of the Ethernet (registered trademark) standard.

  The waveform memory 9 is a memory that stores a plurality of waveform data corresponding to a plurality of timbres. As will be described in detail later, the waveform memory 9 is a memory that can be accessed in “high-speed page mode” (page access) as a data reading method. Composed. The waveform memory 9 may be configured by a rewritable nonvolatile memory (flash memory) or dynamic RAM (DRAM), or may be configured by a non-rewritable nonvolatile memory (ROM). The waveform data is data obtained by sampling (sampling) an analog sound signal at a predetermined sampling rate and recording a sample value indicating the instantaneous amplitude at each sampling point as digital data having a fixed number of bits. In this specification, a sample value (a sample of waveform data) indicating an instantaneous amplitude at each sampling point is referred to as a “waveform sample”.

  The sound source unit 8 is a waveform memory sound source that accesses the waveform memory 9 every sampling period, reads out waveform samples from the waveform memory 9, and generates a musical sound signal using the read waveform samples. The sound source unit 8 has a plurality of predetermined sound generation channels, and can generate music signals for a plurality of sound generation channels at the same time for each sampling period by performing a time division operation. The musical sound signal generated by the sound source unit 8 is supplied to a sound system 10 including an amplifier and a speaker and is generated by the sound system 10.

  According to the present invention, as will be described in detail later, the method for accessing the waveform memory 9 is improved by improving the method for reading data from the waveform memory 9 and improving the method for storing waveform samples in the waveform memory 9 accordingly. This makes it possible to apply access, thereby improving the access speed of the sound source unit 8 to the waveform memory 9 and, as a result, increasing the number of sound generation channels for each sampling period.

  FIG. 2 is a diagram illustrating a configuration example of the waveform memory 9. The waveform memory 9 stores a plurality of N waveform data (“waveform data 1”, “waveform data 2”, “waveform data 3”... “Waveform data N”). The user of the electronic musical instrument can select one to be used for performance from a plurality of prepared timbres (tone color data) by operating the operation element 5 or the like. Each tone color data includes waveform selection information for selecting waveform data, a volume envelope parameter for controlling volume change, and the like. Based on the waveform selection information, different waveform data is selected from the N waveform data according to the pitch and intensity of the musical tone to be played.

  In FIG. 2, the configuration of one waveform data (“waveform data 2”) is shown in detail. One waveform data is composed of a plurality of WS waveform samples (“sample 0”, “sample 1”, “sample 2”... “Sample (WS-1)”), and each waveform sample is stored in the waveform memory 9. Address numbers “0”, “1”, “2”... “(WS-1)” are stored for each address. Each waveform sample is basically stored in chronological order so that a musical sound signal (waveform data) as a sampling source can be reproduced by reading it in the order of address numbers. That is, the order of the address numbers corresponds to the temporal evolution of the instantaneous phase of the waveform data. In FIG. 2, address numbers “0”, “1”, “2”... ((WS-1)) indicate the address number at which each waveform sample is stored, and the first waveform sample “ It is expressed in decimal notation by a relative address number starting from the address of sample 0 ”(“ 0 ”).

  WS is a parameter indicating the data size (waveform size) of one waveform data, and corresponds to the number of waveform samples constituting one waveform data. Since one waveform sample is stored in one address in the waveform memory 9, the waveform size WS of one waveform data corresponds to the number of addresses storing the waveform data. Therefore, if the waveform size WS is known, the final address of the waveform data can be specified. In FIG. 2, since the address number of the first waveform sample of the waveform data 2 starts from “0”, the final address of the waveform data 2 is expressed as “(WS-1)”. Here, in order to simplify the explanation, the waveform data is assumed to be of a type that is read one shot. The present invention can also be applied to other types of waveform data, such as waveform data that includes an attack portion and a loop portion, and the loop portion is repeatedly read after the attack portion is read out. .

  The head address TA is a pointer that specifies on the waveform memory 9 the address at which the top waveform sample of the waveform data is stored. Accordingly, an absolute storage position of a certain waveform sample of a certain waveform data on the waveform memory 9 can be specified by a combination of the start address TA of the waveform data and the address of the waveform sample.

  Since the waveform memory 9 is a memory capable of “page access”, its address space is divided into a plurality of pages for each page size p of a predetermined number of addresses. As shown in FIG. 2, in this embodiment, the address space of the waveform memory 9 is divided into pages every 8 addresses per page. For example, eight addresses up to address numbers 0, 1, 2,... 7 constitute a first page p1, and eight addresses up to address numbers 8, 9, 10,.

  In the “high-speed page mode” access, after accessing any address in the waveform memory 9 for the first time, an address in the same page as that address can be accessed sequentially in a shorter access time than the first access. It is. In this specification, the first access and the subsequent n-1 accesses are collectively referred to as “page access”. “N” is a positive integer. In this embodiment, as an example of page access, two waveform samples stored at two addresses in one page are read for each access. A technique for reading data at high speed by “page accessing” the memory has been conventionally known as a data high-speed reading technique in a mask ROM, a flash memory, or the like.

The waveform sample stored at the final address of each page of the waveform memory 9 is also stored redundantly at the top address of the next page. In this way, storing the same waveform sample at the last address of each page and the top address of the next page is necessary for interpolation in page access of each sounding channel in each sampling period. It is a device that makes it possible to read two waveform samples.
For example, “sample 7” stored in the address number 7 that is the final address of the first page p1 is also stored in the address number 8 that is the start address of the next second page p2, and The waveform samples after “sample 8” following “sample 7” are sequentially stored after the subsequent address number 9. Similarly, “sample 14” stored in the address number 15 which is the final address of the page p2 is also stored in the address number 16 which is the head address of the next page p3. Waveform samples after “sample 15” following “14” are sequentially stored.
In other words, except that the waveform sample stored at the last address of each page is stored redundantly at the top address of the next page, each waveform sample has the instantaneous phase of the waveform data in the order of the address number. They are stored side by side in chronological order as time progresses.

FIG. 3 is a block diagram showing a detailed configuration of the sound source unit 8 shown in FIG. The sound source unit 8 includes an address counter 11, a memory address generation unit 12, an inter-sample interpolation unit 13, a volume envelope addition (volume EG) unit 14, a mixer 15, a digital / analog conversion unit (DAC) 16, and a control register 17. It is configured. Each unit of the sound source unit 8 operates with a system clock generated by a clock generator (not shown) as an operation reference. The sampling period in which the sound source unit 8 processes the waveform data is generated by dividing the system clock.
Further, the sound source unit 8 divides the sampling period equally for each time slot corresponding to a plurality of sound generation channels based on the system clock, and at each time slot channel timing for each time slot, By executing the process, a time division operation of a plurality of sound generation channels is performed. Specifically, the address counter 11, the memory address generation unit 12, the inter-sample interpolation unit 13, and the sound volume EG application unit 14 each perform a time division operation for a predetermined plurality of sound generation channels. Then, the mixer 15 accumulates the data of the predetermined plural sound generation channels for each sampling period, so that the sound source unit 8 performs a process of generating musical sound signals for the plural sound generation channels for each sampling period as a whole.

  The control register 17 is a register that stores values of various parameters used for the operations of the address counter 11, the memory address generation unit 12, the inter-sample interpolation unit 13, the volume EG unit 14, and the mixer 15, and is a bus line 1B. It is connected to the CPU 1 via The control register 17 is provided with a storage area for each of the predetermined plural sound generation channels, and various parameter values for each sound generation channel generated by the CPUI by note-on event processing described later are stored for each predetermined plural sound generation channel. Set to area. Various parameters set in the control register 17 are a read rate (F number) corresponding to the frequency (pitch) of a musical tone to be generated, a waveform data size WS, a waveform data start address TA, and a temporal change in volume. For example, a coefficient of a volume envelope parameter AEG for controlling the sound volume.

  The address counter 11 accumulates the read rate (F number) of each tone generation channel set in the control register 17 for each sampling period, and outputs the accumulated value as the sample address SA of the tone generation channel. It is. The F number is a parameter that controls the waveform sample readout rate (advancing rate of the sample address SA) in accordance with the pitch of the musical tone signal generated in each tone generation channel. The address counter 11 is reset at the start of sound generation for each sound generation channel, and thereafter, by accumulating the F number of each sound generation channel for each sampling period, “0” is set as the initial value, and each sound generation channel is set. Using the final address (WS-1) specified by the waveform data size WS as a count upper limit value, a sample address SA that progresses at a rate corresponding to the value of the F number is generated for each tone generation channel. The sample address SA includes an integer part SAi and a decimal part SAf. The integer part SAi is supplied to the memory address generation unit 12 and the decimal part SAf is supplied to the inter-sample interpolation unit 13.

The memory address generator 12 sets the sample address integer part SAi of each tone generation channel generated by the address counter 11, the start address TA of each tone generation channel, and the page size p of the waveform memory 9 for each sampling period. Based on this, an operation for obtaining the tone generation channel memory address MA is performed, and the memory address MA is output to the waveform memory 9. The memory address MA is data indicating the address number of the waveform memory 9 to be accessed.
The arithmetic expression performed to obtain the memory address MA is
MA = TA + SAi + Int [SAi / (p−1)] (Formula 1)
It is.
In this arithmetic expression (1), “TA” is the top address of the waveform data, and is adjusted to be the top of any page. “SAi” is an integer part of the sample address supplied from the address counter 11. “P” is the page size (the number of addresses per page). Int [SAi / (p-1)] is a function of the sample address integer part SAi, and is a function for obtaining the integer part of the calculation formula [SAi / (p-1)]. According to (Expression 1), the sample address integer part SAi is converted into a memory address MA with the start address TA as an initial value and Int [SAi / (p−1)] added.
By generating the memory address MA based on the sample address integer part SAi by this (Equation 1), as will be described in detail later, it becomes a value corresponding to the final address of each page so that it becomes the head address of the next page. The memory address MA can be generated by shifting backward and converting it so as to skip (avoid) the value corresponding to the final address. As a result, the two samples accessed using the memory address do not cross the two page boundaries of the waveform memory 9. This conversion is essential for realizing reading of two samples by page access.

  The sound source unit 8 performs page access to the waveform memory 9 in a time-sharing manner using the memory address MA of each tone generation channel obtained by the above calculation in the memory address generation unit 12 for each sampling period. And waveform samples (memory data MD) corresponding to two samples stored in two consecutive addresses of the memory address (MA + 1) and the subsequent memory address (MA + 1) are read out for each sound generation channel. Thereby, two waveform samples of each sound generation channel can be read out for each time division channel timing of each sampling period. The two waveform samples of the sound generation channel that are page-accessed in the time slot of the sound generation channel for each sampling period are output from the waveform memory 9 at the first latch L1 indicated by reference numerals 18 and 19, respectively. It is latched by the second latch L2.

The inter-sample interpolation unit 13 is latched by the first latch L1 and the second latch L2 indicated by reference numerals 18 and 19 on the basis of the sampling address decimal part SAf of each tone generation channel supplied from the address counter 11 for each sampling period. Inter-sample interpolation is performed using the two waveform samples of the sounding channel, and an interpolated waveform sample corresponding to the sampling point indicated by the sample address SA of the sounding channel is output.
The sound volume EG unit 14 generates a sound volume envelope waveform AEG indicating the time change of the sound volume of the sound generation channel based on the sound volume envelope parameter for each sound generation channel set in the control register 17 for each sampling period. The volume of the interpolated waveform sample of the tone generation channel output from the interpolation unit 13 is controlled by the volume envelope waveform AEG.

  The mixer 15 accumulates (mixes) the waveform samples of each tone generation channel output from the volume EG unit 14 at each sampling period, and a musical tone signal including waveform samples for a plurality of tone generation channels obtained as a result of the mixing. Are output to the DAC 16 at every sampling period. The DAC 16 converts the musical sound signal output from the mixer 15 into an analog acoustic signal for each sampling period, and outputs the analog converted analog acoustic signal to the sound system 10 for each sampling period.

  In the tone generator unit 8 having the above-described configuration, various parameters used for generating a tone signal for each tone generation channel are set in the control register 17 by note-on event processing executed by the CPU 1 in response to an event of performance data (MIDI data). FIG. 4 is an example of such event processing, and is a flowchart showing a procedure of note-on event processing executed by the CPU 1 when a note-on event instructing pronunciation of a new musical tone is supplied. When the CPU 1 detects a sounding operation of the performance operator 4 (for example, pressing a key on the keyboard), the CPU 1 generates a sound corresponding to the pitch (pitch of the pressed key) corresponding to the performance operator 4 that has been sounded. Note-on event processing is executed on the assumption that a note-on event has occurred.

  In step S1, the CPU 1 captures the pitch instructed to generate sound (the pitch of the key pressed) into the note number register NN and the intensity instructed to generate sound (key pressing speed) into the velocity register VEL.

  In step S2, the CPU 1 selects (assigns) a sound generation channel to be used for generating a musical sound signal according to the detected sound generation instruction from the plurality of sound generation channels of the sound source unit 8, and the channel of the selected sound generation channel. The number is assigned to the assigned channel register a.

  In step S3, the CPU 1 selects waveform data corresponding to the note number NN and the velocity VEL from the N pieces of waveform data based on the waveform selection information of the currently selected tone color. This is because even if the waveform data has the same tone color, the shape of the waveform differs depending on the difference in pitch and volume.

  In step S4, the CPU 1 generates an F number based on the original pitch of the waveform data selected in step S3, the note number NN, and the pitch difference (cent unit). The original pitch is an original pitch (pitch) when waveform data is sampled, and is stored in the waveform selection information. The F number is determined according to the amount (cent unit) of shifting the pitch of the original pitch of the selected waveform data in order to generate a tone signal having a pitch corresponding to the note number NN in each tone generation channel. Read rate. The F number value controls the waveform sample readout rate (advancing rate of the sample address SA) for each tone generation channel.

  In step S5, the CPU 1 stores the F number determined in step S4, the start address TA of the waveform data selected in step S3, the storage area of the tone generation channel a of the control register 17 according to the value of the register a, The values of various parameters necessary for generating a musical tone signal including the waveform size WS of the waveform data and the volume envelope parameter of the currently selected tone are set.

  In step S6, the CPU 1 instructs the tone generator unit 8 to start generating a tone signal (sound generation start) in the sound generation channel a according to the value of the register a. As a result, the sound source unit 8 starts processing for generating a musical sound signal based on various parameters set in the control register 17 using the tone generation channel a to which the musical sound signal is assigned.

  That is, the count value SA (a) of the sound generation channel a of the address counter 11 is reset at the start of sound generation of the sound generation channel a, and thereafter, the F of the sound generation channel a set in the control register 17 for each sampling period. The number is accumulated to generate a sample address SA (a) of the sound generation channel a that proceeds at a rate corresponding to the F number. The memory address generation unit 12 converts the sample address integer part SAi (a) of the sound generation channel a into the memory address MA (a) by the calculation according to the above (Equation 1) for each sampling period. That is, the memory address generation unit 12 sets the start address TA (a) set in the control register 17 as an initial value, and adds the added value Int [SAi (a) / (p−) to the sample address integer unit SAi (a). 1)] is added to generate a memory address MA (a). Since the address counter 11 is reset at the start of sound generation of the sound generation channel a and proceeds at a rate corresponding to the F number, the memory address MA (a) of the sound generation channel a is set to F with the head address TA (a) as an initial value. Progress at a rate according to the number.

  Therefore, the tone generator 8 uses the memory address MA (a) of the sound generation channel a generated at each sampling period to page access the waveform memory 9, and the address number MA (a) from the waveform memory 9. And the process of reading waveform samples for two consecutive addresses of address number (MA (a) +1).

  At each sampling period, the two waveform samples of the sound generation channel a output from the waveform memory 9 by the page access are first and 19th indicated by reference numerals 18 and 19 for each time division channel timing of each sampling period. 2 latches L1 and L2. The inter-sample interpolation unit 13 uses the two waveform samples of the sound generation channel a latched in the first and second latches L1 and L2 for each sampling period, and uses the fractional unit SAf supplied from the address counter 11. By performing the inter-sample interpolation calculation (two-point interpolation) based on the above, a waveform sample that has been interpolated at the sampling point corresponding to the sample address SA (a) is generated. As described above, the sound source unit 8 is configured to access the waveform memory 9 for each sampling period and read out waveform samples necessary for inter-sample interpolation of each sound generation channel. It is not necessary to provide a buffer memory for each tone generation channel.

  In the above, the waveform sample readout speed (F number) corresponds to the pitch difference between the pitch (note number NN) instructed by the note-on event and the original pitch of the waveform data assigned to the sound channel. Since the waveform data stored in the waveform memory is read out from the waveform memory 9 and is interpolated at two points, the pitch shift is performed so that a tone signal having a pitch instructed to be generated is generated. Is done. The waveform memory reading method here is a pitch asynchronous method synchronized with a certain sampling period, but the present invention may be applied to a pitch synchronous method.

  For each sampling period, the waveform sample of the sound channel a output from the inter-sample interpolation unit 13 is generated in the sound volume EG unit 14 based on the sound volume envelope parameter of the sound channel a set in the control register 17. The temporal change in volume is controlled by the volume envelope waveform AEG and is output to the mixer 15. The mixer 15 accumulates waveform samples of a plurality of sound generation channels (including the sound generation channel a) output from the volume EG unit 14 for each sampling period, and accumulates waveform samples for the plurality of sound generation channels. The signal is output to the DAC 16. The DAC 16 converts the multi-tone signal output from the mixer 15 into an analog sound signal for each sampling period, and outputs the analog sound signal to the sound system 10. Thus, the sound source unit 8 can reproduce the musical sound signal instructed to be generated by the user at a pitch specified by the user.

  FIG. 5 is a diagram for explaining the waveform sample read operation by page access according to this embodiment in association with the value of the sample address SA. With reference to FIG. 5, the operation of reading the waveform data of one sounding channel will be described in detail. In FIG. 5, in the rightmost column, waveform sample 0, waveform sample 1, waveform sample 2,... Waveform sample (WS-1) constituting one waveform data of waveform size WS are drawn vertically. Corresponding to each waveform sample, address number 0, address number 1, address number 2... Address number (WS-1) where each waveform sample is stored are shown. As described with reference to FIG. 2, the waveform memory 9 is divided into page units in which the number of addresses of one page is eight, and the same waveform sample is used for the last address of each page and the head address of the subsequent page. Are stored in duplicate.

  In the leftmost column of FIG. 5, the sample address integer part SAi output from the address counter 11 is vertically arranged, and on the left side of the SAi column is an addition value Int [SAi for each sample address integer part SAi. / (P-1)]. Also, the left side of the Int [SAi / (p-1)] column is output as a result of substituting SAi and Int [SAi / (p-1)] shown in the same row into equation (1). The value of the memory address MA is shown. That is, each row in FIG. 5 includes the value of the sample address integer part SAi, the value of the memory address MA corresponding to the value, the address number of the waveform memory 9 corresponding to the value of the memory address MA, and the memory address MA. Correspondence between waveform samples for two samples read out by page access using.

For example, when the value of SAi is “0”, since the addition value for SAi is Int [0 / (8-1)] = 0, the value of the memory address MA is “0”. Note that the value of the memory address MA actually output from the memory address generation unit 12 is a value having the initial address TA of the waveform data of the current tone color as an initial value, that is, “SAi + Int [SAi / (p−1)]”. The value is obtained by adding the head address TA, but here, for convenience of explanation, the value of the head address TA is described as “0”. Since the start address TA is adjusted to be the start address of the page, even if the start address TA is other than “0”, only the memory address MA is shifted. Is basically the same.
For each sampling period, the sound source unit 8 accesses the waveform memory 9 in a time-sharing manner for a plurality of tone generation channels. In the page access of the sound generation channel, first, the value “0” of the memory address MA is output as an address signal in the first access, and the lower address “0” of the page selected by the upper address “0” (fourth bit or more). ”(3 bits), that is,“ sample 0 ”of the address number“ 0 ”is read out. The memory address “1” is output in the next second access, but since it is an address in the same page as the first access, access in the high-speed page mode is possible, and the access time is shorter than the first access. “Sample 1” with address number 1 is read out. In FIG. 5, the ellipse image clearly shows waveform samples for two addresses read from the waveform memory 9 by one page access to the waveform memory 9, and when “0” is output as the memory address MA. Reference numeral 30 is given to the ellipsoidal image showing the two waveform samples read out.

  The sample address SA of the sounding channel advances by the reading rate (F number) of the sounding channel every sampling cycle. For example, if the SAi value remains unchanged at “0” even in the next sampling period, the above-described operation is repeated, and “sample 0” and “sample 1” are read by page access. When the value of SAi proceeds to any one of “1” to “6”, the addition value for SAi is Int [SAi / (8-1)] = 0, so the value of the memory address MA is the value of SAi. Any one of values “1” to “6” corresponding to. For example, when the SAi of the sound generation channel is “2”, “sample 2” and “sample 3” are read by page access.

As described above, since page access is a method of reading waveform samples at two addresses in one page at a high speed, waveform samples at two addresses across two page boundaries cannot be read by one page access. . For example, data cannot be read from the last address (address number 7) of the first page p1 and the start address (address number 8) of the second page p2 in one access.
For this reason, in the arithmetic expression (1) for obtaining the memory address MA, by adding Int [SAi / (p−1)] to the sample address integer part SAi and proceeding in order, the final address of each page As a result, the value corresponding to the last address of each page is not generated as the memory address MA.

  Referring to the numbers shown in the column of the memory address MA in FIG. 5, it can be seen that the MA value does not include “7”, “15”... Corresponding to the final address of the page. As a result, the memory address MA does not become the final address of each page, and the memory addresses MA and MA + 1 become addresses in the same page. That is, two samples of each sound channel can be read out with one page access. The memory address MA no longer points to the last address of each page. Even in this case, the memory address MA is set to the top address of the next page so that interpolation can be performed between the waveform sample of the last address and the next waveform sample. The waveform sample at the final address is stored in duplicate. In the method of reading out two samples in the same page for each tone generation channel by page access, a waveform address continuously read and interpolated is generated only by generating a memory address MA avoiding the final address of each page from the sample address SAi. However, if the waveform sample at the final address of each page is also stored at the top address of the next page, such inconvenience does not occur.

  For example, when SAi has progressed to “7”, the value added to SAi is Int [7 / (8-1)] = 1, so the value of the memory address MA is “8”. In the waveform memory 9, the second page p2 including the address number 8 corresponding to the value “8” of the memory address MA is selected as an access target, and the address number 8 in the selected page p2 is selected. Here, since the waveform memory 9 also stores “sample 7” stored at the last address number 7 of the first page p1 also at the top address number 8 of the next second page p2, the sound source section 8 is a page access to the waveform memory 9 using the value “8” of the memory address MA, and “sample 7” stored at the top address number 8 of the second page p2, as indicated by reference numeral 31, is continuous with it. Two waveform samples “sample 8” stored at address number 9 to be read are read out.

  The address numbers 15, 23... Of the final addresses of the other pages are not generated as the memory address MA, and are the same as the waveform samples stored at the final addresses of these pages. Waveform samples are stored at the top address of the next page. As a result, even when SAi is advanced after “8”, the generated MA value does not become the final address of the page, and inter-sample interpolation is performed by one page access of the waveform memory 9. Two necessary samples can be read out.

  FIG. 6 is a timing chart showing an example of access in the high-speed page mode. Reading of two samples per sounding channel is performed using the high-speed page mode. In FIG. 6, the access time is represented by the number of system clocks of the sound source unit 8. At the first timing 40, since this is the first access to a certain page, the low order bits (3 of the MA) of the page designated by the high order bits (fourth bit or more) of the MA by the address designation using the memory address MA. Waveform sample at the address specified by (bit) is read. This first access takes 8 clocks equivalent to the access in the normal mode. At the next second timing 41, the waveform sample at the address specified by the lower bits (3 bits) of MA + 1 in the same page is read by the second access using the memory address MA + 1. Since this is an access within the same page in the mode, this only takes 3 clocks. Hereinafter, as long as the address is within the same page, the access after the third timing 42 can also be made within the same page in the high-speed page mode, and can be read in 3 clock times.

As described above, in the page access, the access time per waveform sample can be shortened as the number of waveform samples read in the same page is increased. Since the time for reading out waveform samples necessary for inter-sample interpolation is reduced for each sound generation channel, the number of sound generation channels in the sound source unit 8 can be increased as long as the sampling period is the same. Conversely, if the number of sound generation channels is the same, the sampling frequency of the generated musical sound in the sound source unit 8 can be further increased.
For example, if the system clock of the sound source unit 8 is 14 nanoseconds (14 ns) and the number of system clocks in one sampling period is 1536 clocks (sampling frequency is 46.5 kHz), in the example of FIG. Since 3 clocks are required for the second access, the number of clocks required for page access of each sound generation channel is “8 + 3” = “11 clocks”. Therefore, under this condition, the maximum number of sound generation channels provided in the sound source unit 8 is 1536 / “8 + 3” = 139 channels.
On the other hand, in the access in the conventional normal mode, since it takes 8 clocks for each of the two accesses, “8 + 8” = “16 clocks” is required per sounding channel, and the maximum number of sounding channels is 1536 / “8 + 8” = 96. Become a channel. From this, it can be understood that the number of sound generation channels of the sound source unit 8 can be increased by shortening the access time per sound generation channel by page access.

As described above, according to the present embodiment, the waveform memory 9 uses a page accessible memory, the waveform samples at the two addresses at each page boundary are stored in both the previous and next pages, By generating a memory address MA that avoids the final address of each page in the sound generation channel, two consecutive 2 in the same page of the waveform memory 9 using page access at the access timing for each sound generation channel in each sampling period. Waveform samples can be read from the address at high speed.
In addition, since the waveform samples used for inter-sample interpolation of each sound generation channel are read from the waveform memory 9 at every sampling period, a buffer memory for each sound generation channel is not required, and the sound source unit 8 correspondingly. The configuration can be simplified.
Therefore, according to the sound source unit 8 according to the present embodiment, the access speed to the waveform memory 9 can be improved and the number of sound generation channels can be increased with a simple configuration.

In the above-described embodiment, an example in which two waveform samples are read from two consecutive addresses every time the waveform memory 9 is accessed has been described. The number of samples read for each access by page access is not limited to two, and it may be configured to read n consecutive waveform samples from consecutive n addresses for each page access (n Is a positive integer). When designed to read n waveform samples for each page access, the waveform memory 9 stores (n-1) addresses from the end of each page and the beginning of the next page (n -1) The same waveform sample is stored in duplicate at the address. Then, the memory address generation unit 12 generates a memory address MA by the following arithmetic expression (2).
MA = TA + SAi + Int [SAi / {p− (n−1)}] × (n−1)
= TA + SAi + Int [SAi / (p−n + 1)] × (n−1) (Formula 2)
As a result, page access is made to the waveform memory 9 using the memory address MA, and n consecutive waveform samples are obtained from n consecutive addresses from the memory address MA to (MA-n + 1) for each access. Can be read out for each tone generation channel. In this case, the sound source unit 8 includes n latches for latching waveform samples, and uses the n waveform samples read for each sound generation channel for each sampling period, to the sample address fractional unit SAf. Inter-sample interpolation based on the result, an interpolated waveform sample corresponding to the sample address SA of the sound channel is obtained, and further, a time change in volume corresponding to the sound volume envelope parameter of the sound channel is given. Form a waveform sample. In each sampling period, the waveform samples of the plurality of tone generation channels formed are accumulated and output to the DAC 16 as tone signals for the plurality of tone generation channels. In this case, the number of clocks required for page access of each sound generation channel is “8 + 3 × (n−1)”. For example, when n is 4, “8 + 3 × 3” = “17 clocks”, and the maximum number of sound generation channels is 1536/17 = 90 channels. On the other hand, the maximum number of sound generation channels when page access is not used is 1536 / “8 × 4” = 48 channels. In the present invention, 42 channels can be increased as compared with this.

  In the above embodiment, the example in which the page size p of the waveform memory 9 is 8 addresses per page has been described. However, as the waveform memory 9, the page size p is, for example, any other address such as 16 addresses per page. A number of memories can be used.

  In the above embodiment, one waveform sample stored at one address of the waveform memory 9 is read and written as significant data, so one address corresponds to one word. Therefore, in the description of the above embodiment, for example, “reading n waveform samples from consecutive n addresses for each page access”, “page size p is 8 addresses per page”, etc. If the word “address” is replaced with the word “word”, “reading waveform samples of n consecutive words for each page access” or “page size p is 8 words per page” Expressed.

  Further, in the above-described embodiment, an example in which the musical sound signal generation device (waveform memory sound source) of the present invention is applied to an electronic musical instrument has been described. However, the musical sound signal generation device (waveform memory sound source) according to the present invention is not limited to an electronic musical instrument. In addition, the present invention can be applied to any device that mounts a musical sound signal generation device (waveform memory sound source) as a sound source.

In the waveform memory of the above-described embodiment, a memory having a read function in the high-speed page mode for performing address designation even after the second access is used. However, after the second access according to the present invention, data is sequentially read from the address continuous to the first access. Therefore, instead of the memory having the high-speed page mode, the data at the address following the first access is transferred to the second access. Thereafter, a memory having a burst read function that can be read without an address signal can also be used.
In the above embodiment, the sample address SA is reset to “0” at the start of sounding. However, the sounding may be started by setting a value (−Toff) other than “0”. . In this case, the top address TA of the waveform data is “TA + Toff”.
In the above embodiment, the top address TA of the waveform data is the top address of any page. However, it is not always necessary to use the top address of the page.

The block diagram which shows the example of whole structure of the electronic musical instrument to which the musical tone signal generation apparatus which concerns on this invention is applied. FIG. 2 is a diagram illustrating a configuration example of a waveform memory in FIG. The block diagram explaining the structural example of the sound source part of FIG. The flowchart which shows an example of the procedure of the note-on event process which CPU performs according to a pronunciation instruction. The figure for demonstrating the waveform sample read-out operation by the page access of the sound source part of FIG. 3 corresponding to the value of sample address SA. The timing chart which shows an example of the access by high-speed page mode.

Explanation of symbols

1 CPU, 2 ROM, 3 RAM, 4 performance controls, 5 controls, 6 display, 7 input / output I / O, 8 tone generator, 9 waveform memory, 10 sound system, 11 address counter, 12 memory address generator , 13 Interpolation section between samples, 14 Volume EG section, 15 Mixer, 16 Digital-analog conversion section

Claims (4)

A musical tone signal generating device that generates musical tone signals of a plurality of sound generation channels for each sampling period based on waveform data stored in a waveform memory,
A waveform memory in which the address space is divided into pages by page size p, and the data stored in the addresses in each page is capable of high-speed page access, and the waveform data is stored in each address one sample at a time. The sample memory stored at the last address is a waveform memory that is also stored redundantly at the top address of the next page,
An address counter that accumulates the readout rate of each sounding channel for each sampling period and generates a sample address consisting of an integer part and a decimal part for each sounding channel;
A memory address generation unit that converts the integer part of the sample address generated by the address counter into a memory address that jumps over a value corresponding to the final address of each page for each sampling period;
At each sampling period, the waveform memory is page accessed using the memory address of each tone generation channel generated by the memory address generation unit, and two samples stored at two consecutive addresses from the memory address are obtained. A readout unit to read out,
At each sampling period, the interpolated samples are interpolated between two samples for each sound channel read out by the readout unit based on the fractional part of the sample address generated by the address counter. An interpolation unit that generates a sample for each sound channel;
A musical tone signal generating apparatus comprising: a musical tone signal generating unit that generates musical tone signals of a plurality of tone generation channels for each sampling period based on the samples for each tone generation channel interpolated by the interpolation unit. When the predetermined number of addresses of the page size is p,
The memory address generation unit generates a memory address based on a result of adding the integer part SAi of the sample address generated by the address counter and the integer part of [SAi / (p−1)] which is a function of the integer part SAi. The musical tone signal generating apparatus according to claim 1, wherein: A musical tone signal generating device that generates musical tone signals of a plurality of sound generation channels for each sampling period based on waveform data stored in a waveform memory,
A waveform memory in which the address space is divided into pages by page size p, and the data stored in the addresses in each page is capable of high-speed page access, and the waveform data is stored in each address one sample at a time. The waveform data samples stored in the n-1 addresses from the last of the waveform memory are also stored in duplicate in the n-1 addresses from the top of the next page,
An address counter that accumulates the readout rate of each sounding channel for each sampling period and generates a sample address consisting of an integer part and a decimal part for each sounding channel;
A memory address generation unit that converts an integer part of the sample address generated by the address counter into a memory address that jumps over a value corresponding to n-1 addresses from the end of each page for each sampling period;
In each sampling period, the waveform memory is page-accessed using the memory address of each tone generation channel generated by the memory address generation unit, and n number of addresses stored in consecutive n addresses from the memory address are stored. A readout section for reading the sample;
For each sampling period, the interpolated samples are interpolated between n samples for each tone generation channel read by the reading unit based on the fractional part of the sample address generated by the address counter. An interpolation unit that generates a sample for each sound channel;
A musical tone signal generating apparatus comprising: a musical tone signal generating unit that generates musical tone signals of a plurality of tone generation channels for each sampling period based on the samples for each tone generation channel interpolated by the interpolation unit. When the predetermined number of addresses of the page size is p,
The memory address generation unit is a value obtained by multiplying the integer part SAi of the sample address generated by the address counter and the integer part of [SAi / (pn + 1)], which is a function of the integer part SAi, by (n-1). 4. The musical tone signal generating apparatus according to claim 3, wherein a memory address is generated based on a result obtained by adding together.

Images