JP3627590B2 - Sound generation method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a sound generation method that enables a musical sound to be generated by a general-purpose processing device including an arithmetic processing device.
[0002]
[Prior art]
Conventional musical tone generators usually include a performance input unit for inputting performance information from a MIDI (Musical Instrument Digital Interface), a keyboard, or a sequencer, a sound source unit for generating a musical sound waveform, and the sound source according to the input performance information. It was comprised from the microprocessor (CPU) etc. which control a part. The CPU executes sound source driver processing (performance processing) such as channel assignment and parameter conversion in accordance with the input performance information, and converts the parameter converted to the channel assigned by the sound source unit and the sound generation start instruction (note-on) to the sound source unit. To supply. The sound source unit is composed of an electronic circuit (hardware) such as an LSI (Large Scale Integrated circuit), and generates a musical sound waveform based on the supplied parameters.
For this reason, the musical sound generating device becomes a dedicated device for generating musical sounds, and it is necessary to prepare a dedicated musical sound generating device when generating musical sounds.
[0003]
In order to solve this problem, a musical sound generation method has been proposed in which an application program is executed by a CPU and a musical sound is generated based on the application program. In this musical tone generation method, in addition to the application program that generates musical tone, other application programs can be executed, and can be executed by a general-purpose arithmetic processing device that can also execute other functions. .
[0004]
[Problems to be solved by the invention]
By the way, when generating a musical tone by executing an application program with a general-purpose device including a processing unit (CPU), conventionally, a musical sound waveform sample of each channel is taken every sampling period (conversion timing of a digital / analog converter). Calculations are generated for the sound channels. Therefore, when the CPU performs processing for each sound generation channel, first, preparatory processing such as reading various register values used for the previous calculation of the sound generation channel from the memory to the CPU register is performed. In addition, after the tone generation processing for the tone generation channel, the register value needs to be written into the memory for the next processing.
In other words, since the calculation processing of the musical sound waveform samples of each tone generation channel is generated one sample at a time, much calculation time of the CPU is spent on the preparation processing other than the musical sound generation processing for generating musical sounds (the overhead is large). However, there is a problem that the calculation efficiency is deteriorated and the response and the tone generation process are slow.
[0005]
In the musical sound generation method, the performance process is a process for creating control information for controlling the generated musical sound based on the input performance information. On the other hand, the sound source processing is processing for generating musical tone waveform data based on the created control information.
For example, normally, a performance process such as key-press detection is executed, a sound source process is interrupted and executed at each sampling period for this performance process, waveform data for one sample is generated, and then the process returns to the performance process. I have to.
[0006]
The performance information (MIDI) is generated by the performance operation of the performer or the reproduction of the event by the sequencer. When the performance information is generated, the performance information is processed by the performance processing. In other words, when the performance information is generated, the CPU must execute the performance processing in addition to the normal sound source processing. Therefore, the calculation amount is temporarily increased by the performance information generated irregularly. Become. However, conventional music generation methods do not support this, and sound source processing is regularly executed with or without performance information, and performance processing may be delayed in some cases. It was.
In order to prevent such delays in performance processing, it is conceivable to increase the priority of performance processing, but this time, the number of pronunciations may temporarily decrease or the sound waveform may be interrupted. Will occur.
[0007]
Therefore, an object of the present invention is to provide a sound generation method that can solve the above-described problems.
[0008]
[Means for Solving the Problems]
In order to solve the above-mentioned object, the sound generation method of the present invention includes a first step of receiving one or a plurality of generated instructions for generating one or more specified musical sounds, and in response to the generated instructions. And assigning each designated musical tone to each one of a plurality of tone generation channels, and writing control data of the designated tone into a channel register corresponding to each tone generation channel to which each designated tone is assigned. , A third step for sequentially issuing a calculation start command, and waveform data for a plurality of samples for each tone generation channel based on control data stored in the channel register of the channel in response to each calculation start command And the waveform data generated for each sound generation channel in the fourth step is converted into each sample of the plurality of samples. And a mixed sample generated before the generation of new mixed sample data for each of the plurality of samples in the fifth step. A sixth step of reserving the new mixed sample data for reproduction after the data reproduction is completed, and the mixed sample data at the sample points sequentially for each sampling period, And a seventh step of sequentially outputting in accordance with the reservation.
[0009]
According to another sound generation method of the present invention that can achieve the above object, a first step of issuing a generation command for generating a plurality of designated musical sounds, and each designated musical sound in a plurality of tone generation channels. A second step of assigning to one of the designated tone generation channels, writing the control data of the designated tone into the channel register of each designated tone generation channel and storing it, and issuing a calculation start command at predetermined time intervals Step 3 and the tone generation calculation for each channel are sequentially executed in accordance with each calculation start command issued in the third step, and each designated tone is generated based on the control data stored in the channel register of the channel. A fourth step for arithmetically generating waveform data for a plurality of samples for each channel, and a sound generation channel specified in the fourth step; The fifth step of mixing the waveform data generated for each sample for each sample to generate mixed sample data, and converting the mixed sample data for each of the plurality of samples into an analog signal for each sampling period In the musical tone generation calculation of the fourth step, the control data is read based on one read of the control data from the channel register for each designated tone generation channel. Waveform data for a plurality of samples of the generated sound generation channel is generated, and control data is written to the channel register after the waveform data is generated.
[0010]
According to the present invention, when generation of new mixed sample data for each of a plurality of samples is completed, reproduction of new mixed sample data is completed after reproduction of the previously generated mixed sample data is completed. Since the reservation is made so that the music is performed, the musical sound can be reproduced without interruption.
Further, based on one-time reading of the control data from the channel register, waveform data for a plurality of samples of the sound generation channel from which the control data is read is generated, and the control data is written to the channel register after the waveform data is generated. Therefore, it is only necessary to prepare each tone generation channel once for the calculation of a plurality of musical sound waveform samples, and the overhead can be reduced. For this reason, the quality of the generated musical sound can be improved, and the number of simultaneously sounding channels can be increased.
[0011]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 shows the configuration of a musical sound generating apparatus that can execute the embodiments of the musical sound generating method of the present invention.
In this figure, 1 is a microprocessor (CPU) that performs various controls such as generation of musical sound waveform samples by executing an application program and the like, 2 is a read only memory (ROM) in which preset tone color data and the like are stored, 3 A random access memory (RAM) 4 having storage areas such as a work memory area, a tone color data area, an input buffer area, a channel register area, and an output buffer area of the CPU 1 indicates the time and also instructs the CPU 1 the timing of timer interrupt processing. The timer 5 is a MIDI interface for outputting a generated MIDI event while a MIDI event is input, and 6 is a so-called personal computer keyboard provided with keys such as letters, kana, numbers, symbols and the like.
[0012]
Reference numeral 7 is a display (monitor) for the user to interact with the musical sound generating device, and reference numeral 8 is an application program such as a program for generating musical sounds, and musical sound waveform data used for generating musical sound waveform samples. The stored hard disk (HDD) 9 transfers the musical sound waveform sample data stored in an area designated by a part of the CPU of the RAM 3 directly without going through the CPU 1, and has a fixed sampling period (for example, , 48 kHz) to a digital / analog converter (DAC) 10, a reproduction unit (DMA; Direct Memory Access) 10 is a digital-analog converter (DAC) that receives data of musical sound waveform samples and converts them into analog signals, 11 was output from the DAC 10 A sound system for sound the converted tone signals to analog signals.
The above configuration is the same as that of a personal computer, a workstation, etc., and the musical sound generating method of the present invention can be implemented on them.
[0013]
As described above, the RAM 3 has an area for storing various data. Of these, the area for storing timbre data is shown in FIG. 2, the area for the input buffer is shown in FIG. 3, and the area for the channel register is shown in FIG. FIG. 4 shows the area of the output buffer.
In the area shown in FIG. 2, PD1, PD2,..., PD16 are timbre data for 16 types, and each timbre data is applied with data for designating the waveform of each tone range (designation of each tone range waveform), vibrato, etc. LFO (Low Frequency Oscillator) control data (LFO control OD), filter envelope generation control data for controlling timbre filter characteristics (FEG control OD), and envelope generation control data for controlling amplitude (AEG control OD), touch control data (touch control OD) for changing the speed of rising of a musical tone by velocity, and other data (other OD).
[0014]
Note that OD indicates original data, and the original data is processed according to touch data, pitch data, etc. at the time of sound generation instruction, and sound generation data used by the sound source is generated. ing.
In addition, WD1, WD2,..., WDn are waveform data, and any one of the waveforms depending on the tone range waveform designation data in the timbre data of PD1, PD2,. Data is specified.
[0015]
Next, MIDI event data ID1, ID2, ID3,... Of note-on, note-off and various events inputted via the MIDI interface 5 are sequentially written in the input buffer area shown in FIG. By reading the data ID1, ID2, ID3,..., The event processing is executed in the musical tone generation apparatus.
The MIDI event data ID1, ID2, ID3,... Is composed of the data contents of the MIDI event (for example, data 1 contents) and the time of occurrence of the data (data 1 occurrence time). This occurrence time can be known by capturing the current time of the timer 4 when receiving MIDI event data.
[0016]
Next, the area shown in FIG. 4 is used as a channel (ch) register for storing data for controlling the generation of a plurality of independent musical tones. In this example, 32 of 1ch, 2ch,..., 32ch. There is a channel area. The area of each channel is note number, waveform designation data (waveform designation D), LFO control data (LFO control D), filter envelope control data (FEG control D), amplitude envelope control data (AEG control D), note-on data, It consists of other data (other D) and a work area that the CPU 1 uses when executing a program.
The waveform designation D, LFO control D, FEG control D, and AEG control D are sounding data obtained by processing the original data.
[0017]
Next, an area shown in FIG. 5 is an area for a plurality of output buffers, which are alternately used as an output buffer X for generating a sound waveform. As will be described later, the output buffer accumulates and stores the channel every time musical tone waveform sample data SD1, SD2, SD3,... One of the output buffers is designated as an output buffer X for generating a sound waveform, and is used for waveform generation calculation. Two or more output buffers X are prepared. In the simplest configuration, there are two output buffers X, and the data of the next musical sound waveform sample calculated on the other side while the data stored in one side is transferred to the reproduction unit (DMA) 9 and reproduced. Can be configured as a double buffer.
[0018]
The size of the output buffer can be arbitrarily set, such as 100 words, 500 words, 1K words, 5K words, etc. However, if the size is increased, the sound generation is delayed, and if the size is decreased, the time margin is reduced. The response becomes worse when the amount of computation is temporarily increased.
Therefore, in the case where, for example, sequencer performance is not required, real-time performance can be absorbed by shifting the performance timing forward, so that the size of the output buffer can be increased. On the other hand, when a real-time performance is required, such as a keyboard performance, the buffer size is preferably 100 to 200 words in order to prevent delays in pronunciation.
The above is the case where the reproduction sampling frequency is 40 kHz to 50 kHz. When the sampling frequency is set low, it is necessary to make the size smaller in order to prevent delay in sound generation.
[0019]
Next, the first to fifth embodiments of the tone generation method of the present invention will be described. Each embodiment can be executed by the tone generation apparatus shown in FIG. is there.
In the musical tone generation method of the first embodiment of the present invention, a plurality of musical tone waveform samples, for example, 100 musical tone waveform samples, are generated together in the musical tone generation processing of each tone generation channel in which the CPU 1 executes an application program for generating musical sounds. It is a thing. That is, in the processing of each sound generation channel, musical tone waveform samples are generated collectively for 100 cycles of the sampling period of the DAC 10.
[0020]
When the tone generation processing for all the sound generation channels is executed at predetermined calculation times, and the generated plurality of tone waveform samples are set to 100 samples, the accumulated values of channels for 100 sampling periods of the DAC 10 are sequentially obtained. Accumulated and stored in the output buffer. The musical sound waveform samples stored in the output buffer are read out by the reproduction unit (DMA) 9 one sample at each sampling period after completion of accumulation for all the sound generation channels, supplied to the DAC 10 and sounded from the sound system 11. Is done.
The calculation time is controlled to be generated at intervals at which the musical sound waveform samples can be read and reproduced without interruption when the musical sound generation processing is executed using a plurality of output buffers alternately.
[0021]
According to the musical tone generation method of the first embodiment, the preparation process for each tone generation channel needs to be performed only once for the calculation of a plurality of musical sound waveform samples generated together. Thus, the ratio of the calculation time spent for the preparation process is reduced, and the overhead can be reduced. For this reason, it is possible to improve the quality of the generated musical sound waveform sample and increase the number of simultaneous sounds.
Note that one section (corresponding to the size of the output buffer) between the calculation times is further divided into n equal parts, the sound waveform calculation is performed at the corresponding time interval, and the one section completed by the last nth calculation is performed. The reproduction unit (DMA) 9 may read out a musical sound waveform sample as a unit.
[0022]
Next, the tone generation method according to the second embodiment of the present invention will be described.
In the musical tone generation method according to the second embodiment of the present invention, a plurality of musical sound waveform samples are collected in the musical tone generation process of each tone generation channel in which the CPU 1 executes an application program for generating musical sounds, as in the first embodiment. In addition, every time an input data, in this example, a MIDI event, is received by the MIDI interface 5, the sound waveform calculation up to that point is performed. Then, at a predetermined calculation time, the sound waveform calculation is performed only for uncalculated musical sound waveform samples among a plurality of predetermined musical sound waveform samples (samples for one output buffer).
[0023]
This is because each sound waveform calculation is performed for a sound channel that is sounding, and there is a key-on event or key-off event (pitch bend, volume change) that changes the sound generation mode according to the input data therein. The sound generation channel requires more calculation processing than the sound generation channel that continues sound generation without changing the correspondence.
In this case, if the calculation time is set at regular intervals, when the input data increases, the calculation time is occupied by the sound generation channels whose sound generation mode changes, and as a result, the number of sound generation channels that can be calculated decreases. It will be.
In particular, a sound generation channel for starting sound generation requires many initial setting processes such as an address counter, initial settings of various envelope generators, and F number generation, and takes a long calculation processing time.
[0024]
The second embodiment will be further described with reference to the timing chart shown in FIG. 12. The output buffer has the above-described double buffer configuration, and these two output buffers are designated as A and B. Is shown in FIG. The time required for reproduction of each of the buffers A and B is T _A And T _B And T _A = T _B It is said that.
First, the time range t to be calculated for the output buffer A ₀ ~ T ₁ At time t as shown in FIG. _a When the two MIDI events are received by the MIDI receiving unit, the sound source driver unit performs processing as shown in FIG. 5B, and further, t as shown in FIG. ₀ ~ T _a Musical sound waveform sample A ₁ Is calculated.
[0025]
The MIDI receiving unit includes a MIDI interface 5 that receives an input MIDI event, and writes the data of the MIDI event in the input buffer together with the time of occurrence as described above. The tone generator driver unit receives data from the input buffer or input from the personal computer keyboard 6 and performs conversion from voicing parameters to tone generator parameters according to the tone generator channel assignment and input. The tone generator unit receives tone generator parameters, processes waveform data, and generates musical sound waveform samples that are actually pronounced. The LPF unit removes the aliasing noise component in the generated musical sound waveform sample. Then, the output of the LPF unit is written into the output buffers A and B.
The sound source drive, the sound source unit, and the LPF unit are functions realized when the CPU 1 executes an application program.
[0026]
Then time t _b When one MIDI event is received by the MIDI receiving unit, the sound source driver unit similarly performs processing, and the sound source unit further performs t. _a ~ T _b Musical sound waveform sample A ₂ Is calculated.
Then time t ₁ When t is reached, t _b ~ T ₁ Musical sound waveform sample A ₃ Is calculated. In this case, the time t _a , T _b When a key-on event is input at this time t including the initial sound generation process, ₁ Is processed. Further, filtering processing is performed in the LPF unit, and generation of the musical sound waveform sample for the output buffer A is completed.
[0027]
Next, the time range t to be calculated for the output buffer B ₁ ~ T ₂ At time t as shown in FIG. _c The three MIDI events are received by the MIDI receiver, but the sound source is the tone waveform sample A. ₃ Is being calculated, the input data is placed in the input buffer until a calculation time is allocated. Musical sound sample A ₃ When the calculation of is finished and the filter processing of the LPF unit is also completed, the data in the input buffer is processed by the sound source driver unit, and t ₁ ~ T _c Musical sound waveform sample B corresponding to the input between ₁ Is calculated. In this case, even when the calculation process is delayed, the generation time of the input data is also written in the input buffer, so that the sound generation timing is not affected.
[0028]
Similarly, musical tone waveform sample B ₁ 4 new MIDI events are received during the calculation time of this, but this input data is also the tone waveform sample B ₁ It is calculated after the calculation of.
As a result, the musical sound waveform sample B ₂ Is t _c ~ T _d Musical sound waveform sample corresponding to the input in between, musical sound waveform sample B ₃ Is t _d ~ T _e Musical sound waveform sample corresponding to the input in between, musical sound waveform sample B ₄ Is t _e ~ T ₂ Musical sound waveform sample corresponding to the input between.
Furthermore, musical sound waveform sample A ₅ Is t ₂ ~ T ₃ Musical sound waveform sample corresponding to the input between.
[0029]
As described above, in the second embodiment, when input data is generated, the sound waveform calculation up to that time is executed at that time, so that the calculation time of the musical sound waveform sample is dispersed. Therefore, since the processing at the calculation time performed every predetermined time does not increase, even if a large amount of input data that changes the sound generation mode such as a key-on event occurs, it is possible to prevent inconvenience such as a decrease in the number of simultaneous sound generations. be able to.
[0030]
Next, the musical sound waveform generation method according to the third embodiment of the present invention will be described.
By the way, when a calculation time is generated at a predetermined timing and a predetermined number of musical sound waveform samples are calculated and generated collectively, in order to generate musical sounds continuously, before the generation of waveform samples generated in the past is completed. Then, it is necessary to supply the subsequent predetermined number of waveform samples. Then, when the number of sound generation channels to be processed is large and the sound waveform calculation amount is too large, if the calculation for all channels is performed, the musical sound waveform samples are not supplied in time, and the sound is interrupted.
[0031]
The musical sound waveform generation method of the third embodiment is intended to solve this drawback, and it is determined whether or not the supply of the musical sound waveform sample is in time for the conversion timing of the DAC 10, and it is determined that it is not in time. In this case, a sound channel to be muted is selected from sound channels having low importance. For the selected tone generation channel, only the dump waveform sample corresponding to the initial period of the predetermined number of waveform samples is calculated in a short time.
As described above, since the musical sound waveform sample in the selected sound generation channel to be muted is not calculated by the short-term dump waveform, the calculation time of this sound channel is shortened, and the supply of the musical sound waveform sample as a whole is converted by the DAC 10. Be in time.
[0032]
The important sound is
(1) A loud sound at that time.
(2) A sound that has just started sounding during the attack part.
(3) The lowest sound (bass sound) when multiple part sounds are played.
(4) When multiple parts are played, the highest sound (lead sound).
(5) Solo part sound when multiple part sounds are played.
Is generally.
[0033]
Further, a modification of the third embodiment will be described. The sound generation channels to be calculated are ranked from the important sounds prior to the sound waveform calculation, and the sound generation calculations are performed in order from the important sounds according to the ranking. When the sound waveform calculation is not in time, the sound waveform calculation is interrupted in the middle, and the sound is generated only by the musical sound waveform samples generated up to that point.
In this way, even if it is necessary to abort the operation, the channel from which the sound disappears is said to be a channel generating a musical sound having a low importance and a relatively low influence.
In the third embodiment and its modifications, the sound waveform calculation may be performed every time input data is generated.
[0034]
Furthermore, the sound waveform calculation may be finally performed for one section without performing the sound waveform calculation every time input data is generated. In this case, it is preferable to trigger the calculation time to be advanced according to the number of input data.
Alternatively, one interval between calculation times is further divided into n equal parts, and the sound waveform calculation is performed at the corresponding time interval, and the musical sound waveform sample for one interval completed in the last nth calculation is played back as a unit. The unit (DMA) 9 may read the data.
[0035]
By the way, when a calculation time is generated at a predetermined timing and a plurality of musical sound waveform samples are calculated and generated together, or when a sound waveform calculation is performed every time input data is generated, musical sounds are continuously generated. For this purpose, it is necessary to supply subsequent waveform samples before the generation of waveform samples generated in the past. Then, if the number of sound generation channels to be processed is large and the amount of sound waveform calculation is too large, or the time required for processing other than the sound generation processing (sequencer processing, etc.), the musical sound waveform sample is supplied. If it was not in time, there was a possibility that the musical sound waveform sample being processed was read out and noise was generated.
Therefore, in the musical sound generating method of the fourth embodiment of the present invention, this is solved as follows.
[0036]
In the fourth embodiment of the present invention, the CPU 1 issues a command to pass the data in the output buffer to the reproduction unit (DMA) 9. In this case, the address of the output buffer itself that stores the generated predetermined number of musical sound samples can be set to 9 in the playback unit as a single readout section, or can be set as a repeated readout section that is repeatedly read. In addition, it is possible to set a read section reservation so that the address can be read following the read section that is currently being read.
In the present embodiment, the reservation of the read section is performed, and after the musical sound waveform calculation sample is generated, reservation registration is performed in the output buffer, and the waveform already being read is read out. When the calculation of the musical sound waveform sample is not completed, the reservation registration is not performed, so that it is possible to prevent the generation of noise due to the sound musical sound waveform sample being generated.
[0037]
In this case, the sound generation is temporarily interrupted, but the effect is small if the interrupted time is suppressed to several sample times at a sample frequency of 44.1 kHz, for example. In order to reduce the time to several sample times, the number of sounded channels may be controlled as described above. Further, when the process is completed, reservation registration is performed and a sound is generated.
It should be noted that the sound waveform calculation may not be performed every time input data is generated, but finally the sound waveform calculation may be performed for one section. In this case, it is preferable to trigger the calculation time to be advanced according to the number of input data.
Alternatively, one interval between calculation times is further divided into n equal parts, and the sound waveform calculation is performed at the corresponding time interval, and the sound waveform sample for one interval completed in the last n-th operation is reserved for sound generation. You may do it.
[0038]
As described above, when a calculation time is generated at a predetermined timing and a plurality of musical sound waveform samples are calculated and generated together, or when a sound waveform calculation is performed every time input data is generated, a musical tone is continuously generated. In order to generate the waveform sample, it is necessary to supply the subsequent waveform sample before the generation of the waveform sample generated in the past.
By the way, this calculation time is designated at a timing earlier than the end timing by the time required to execute the generation of the sound waveform based on the end timing of the musical sound waveform sample generated in the past. This end timing is detected by the CPU 1 confirming the state (flag) of the playback unit (DMA) 9 and detecting that the playback section of the musical sound waveform sample has shifted to the next section. There will be a time delay from when the state (flag) of (DMA) 9 changes until the CPU 1 detects it. Furthermore, since this time delay depends on the timing at which the CPU 1 executes the above-described detection, the time delay becomes uneven depending on the timing.
[0039]
Then, if the calculation time is generated based on the timing at which this uneven time delay occurs, the accurate calculation time cannot be generated.
In particular, if a timing that is greatly delayed by one time is detected, the calculation time generated based on it is short in the calculation time from the calculation start time to the pronunciation waveform supply. Will decrease.
Therefore, the musical tone generation method according to the fifth embodiment of the present invention solves this as follows.
[0040]
The CPU 1 stores a plurality of times when the state change of the reproducing unit (DMA) 9 is detected in the past. The next detection time is predicted by taking the average of the plurality of times. Since the predicted time is obtained by averaging the detection delay from the true end timing in the reproduction unit (DMA) 9, the timing before the predicted timing is set to a substantially accurate end timing. Can be detected as Then, the calculation time is generated based on the end timing.
In this way, the detected end timing is averaged and variation is reduced, so that the calculation time secured at each calculation time is uniformized, and a stable tone generation operation is performed.
[0041]
It should be noted that the sound waveform calculation may not be performed every time input data is generated, but finally the sound waveform calculation may be performed for one section. In this case, it is preferable to trigger the calculation time to be advanced according to the number of input data.
Alternatively, one interval between calculation times is further divided into n equal parts, and the sound waveform calculation is performed at the corresponding time interval, and the sound waveform sample for one interval completed in the last n-th operation is reserved for sound generation. You may do it.
[0042]
Next, the operation of the musical sound generation method and apparatus in which the elements of the first to fifth embodiments of the present invention described above are combined will be described with reference to flowcharts.
FIG. 6 is a flowchart of the main routine. When the main routine is started, initialization is performed in step S10. In the initial setting, the timer 4 and DMA are set, all sound channels are cleared, and tone color data and waveform data are prepared. Next, keyboard processing is performed in which input from the keyboard 6 is processed in step S20, and MIDI processing is performed in which processing according to the MIDI event input is performed in step S30. Further, in step S40, a sound source process is performed in which a sound waveform calculation for generating a musical sound waveform sample is performed, and other processes are performed in step S50. The process returns to step S20, and the processes in steps S20 to S50 are performed. Repeatedly in circulation (stationary loop).
These processes are executed simultaneously with other software using a multitasking method.
[0043]
Next, a flowchart of the MIDI reception interrupt process executed by the CPU 1 is shown in FIG. This process is activated by an interrupt when the MIDI interface 5 receives some MIDI event from the outside. This MIDI reception interrupt process is a process performed with priority over other processes. When this MIDI reception interrupt process is started, the received data received by the MIDI interface 5 is fetched in step S100, and the received data is paired with the time data at the time of reception in step S110. Are written in the input buffer described above in the format shown in FIG. 5B, and the process returns to the process when an interrupt occurs. As a result, the received MIDI data is sequentially written to the input buffer together with the reception time.
[0044]
Next, FIG. 8 shows a detailed flowchart of the MIDI processing executed as step S30 in the steady loop of the main routine.
When the MIDI processing is started, the input buffer is read in step S200, and it is confirmed whether there is unprocessed received data. If it is determined in step S210 that there is unprocessed received data, the process proceeds to step S220 to perform branching according to the content of the received data. If the received data is a note-on event, the process branches to step S230 and a note-on process is executed. If the received data is note-off, the process branches to step S240 and the note-off process is executed. If the received data is other data, the process branches to step S250 and the other process is executed. When any of these processes ends, the MIDI process ends. If it is determined in step S210 that there is no received data, the MIDI processing is terminated as it is.
[0045]
Next, FIG. 9A shows a flowchart of the note-on process executed in step S230 when the received data is a note-on event in the above-described MIDI process.
When the note-on process is started, in step S300, the note number of the note-on event in the input buffer is taken as NN, the velocity is taken as VEL, and the occurrence time of the note-on event is taken as TM. Captured in a register. Next, the sound generation assignment process of the note number NN taken into the register in step S310 is performed, and the assigned channel (ch) number is taken into the register as i.
[0046]
Further, in step S320, musical tone control data corresponding to the note number NN and velocity VEL is set in the ch register of ch number i fetched into the register among the ch registers shown in FIG. The tone control data to be set is the tone data (various ODs) corresponding to the MIDI channel that received the note-on event among the tone data for 16 tones shown in FIG. 2, and the values of the note number NN and velocity VEL. The sound generation data (various D) obtained by processing according to the above.
Here, the waveform designation data D in the sound generation data is obtained by referring to the tone range waveform designation data in the timbre data shown in FIG. 2 with the note number NN, and is used to generate a musical tone corresponding to the note number NN. Any one of the waveform data WD1 to the waveform data WDn is designated as the power waveform. After the musical tone control data is set, an ich note-on flag is set in step S330.
[0047]
Next, calculation generation of the sound waveform is executed in step S340. In this case, the calculation generation is performed before the time TM among all the waveforms to be written to the buffer X that is currently being prepared and an uncalculated waveform ( The partial waveform is executed, and the calculated partial waveform is written to the output buffer X. Here, the partial waveform corresponds to a sound generation waveform in a range in which calculation and generation are possible when new received data is detected in step S210 (data is determined). The waveform to be generated is a sound waveform up to the time of occurrence of the received note-on event TM, and the tone waveform that starts sounding in response to the note-on is not included in the sound waveform and is generated later It is included in the direction. Details of this processing are shown in FIG.
[0048]
The processing of step S340 and step S350 is the same as the musical sound waveform A described above with reference to FIG. ₁ Or A ₂ In step S350, initial setting of sound generation in ich is performed in the work area of the ch register based on the musical tone control data of ich set in the ch register in step S320 described above. Here, this work area displays a plurality of data current values necessary for waveform generation for each channel, such as current address values, various envelope current values and current states, and LFO waveform current values, which are necessary for the tone generation of each tone generation channel. Remember. When this sound generation initial process is completed, the note-on process is completed.
In the initial setting, initial setting such as setting of the start address to the current value of the waveform read address, generation of F number corresponding to the note number NN, and LFO, filter EG, volume EG, interpolation calculation, filter calculation, etc. Done. This initial setting is a process that requires calculation time as described above.
[0049]
Next, FIG. 9B shows a flowchart of the note-off process executed in step S240 when the received data is a note-off event in the above-described MIDI process.
When the note-off process is started, in step S400, the note number of the note-off event in the input buffer is taken into the register as NN, and the occurrence time of the note-off event is taken into the register as TM. Next, at step S410, the sound channel (ch) sounded by the note number NN is searched, and the number of the sound sound channel found is taken into the register as i.
[0050]
Next, in step S420, the ich note-on flag is defeated, and in step S430, the calculation of the sound waveform is executed. The calculation generation process in this case is the same process as step S340 described above, and an uncalculated waveform (partial waveform) before time TM is calculated and written to the output buffer X. Further, in step S440, an ich release start process is performed, and the note-off process ends. Here, the ich release start process is a process of rewriting the state of various envelopes of the ich in the work area and changing the state of tone generation in the ich to the released state.
[0051]
Next, a detailed flowchart of the sound source processing executed as step S40 in the steady loop of the main routine will be described with reference to FIG.
When the sound source processing is started, the playback state of the playback unit (DMA) 9 is checked in step S500. If the playback section has advanced, the process proceeds to step S510. If the playback section has not advanced, the process proceeds to step S520. .
The DMA 9 uses a waveform sample in a specific area on the RAM 3 designated by the CPU 1 as a reproduction area, sequentially reads out one sample from the first sample in the specific area every predetermined sampling period, and supplies it to the DAC 10 for reproduction. Further, the DMA 9 receives a reservation for designating another area to be reproduced next from the CPU 1 while reproducing the specific area. The reserved waveform samples in another specific area are sequentially read out one by one in the same manner by the DMA 9 after the reproduction of the specific area being reproduced is completed, and supplied to the DAC 10 for reproduction. Here, the next advance of the playback section means that the playback of the playback section previously designated as the specific area has been completed and the playback has moved to another playback section reserved as the next specific area. A plurality of specific areas can be reserved for reproduction at a time, and in this case, the plurality of specific areas are sequentially reproduced in the reserved order.
[0052]
In step S510, the next detection time is predicted from the time when the current progress is detected (current time) and the time detected in the past, and a time a predetermined time before the predicted detection time is designated as the next calculation time. . As a method of predicting the next detection time, a method of predicting by calculating an approximate value with a small error by the least square method based on a plurality of detection times including the current time and the past detection time, There is a method of predicting the state of change by approximating with another function such as a quadratic function. In the DMA 9, the time from when the progress occurs until it is detected at step S 510 is caused by a non-constant time delay due to the difference in the processing step position and the situation from time to time. It is included. Therefore, the calculation of the approximate function includes a process of averaging the variation in the detection times multiple times.
[0053]
The predetermined time is a time that is secured for generating a musical sound waveform, and how long it is determined is determined on the basis of the amount of computation required for the computation generation, such as the number of pronunciations to be secured, the quality of the computation, etc. Is done. The length of the predetermined time may be a fixed value, but may be set with the keyboard 6 or may be automatically determined by the CPU 1 in consideration of a plurality of processing programs running simultaneously.
[0054]
Next, in step S520, it is determined whether or not the calculation time has been reached by comparing the next calculation time with the current time indicated by the timer 4, and if it is determined that the calculation time has been reached, the step The processing from S530 to step S580 is executed.
First, in step S530, it is determined in which order the sound channel that is currently sounding is calculated. In the waveform calculation generation process in step S550, which will be described later, a sound waveform for a plurality of samples is generated for each channel for a sounding channel, but the processing order of the channels at that time is determined here. is there.
[0055]
Here, according to the method of the third embodiment described above, ordering is performed in order from the musically important musical tone and the musical tone that is difficult to disappear. Next, in step S540, it is determined whether or not all the sound generation channels that are sounding can be calculated within the scheduled calculation time (the predetermined time described in step S510). The sound generation channels to be muted are designated for one or more channels, and the amount of computation is reduced so that computation can be performed within the scheduled computation time. This process is a specific process based on the musical tone generation method of the third embodiment.
Next, in step S550, a sound waveform is calculated. Here, the sound waveform sample is calculated for the uncalculated portion of the buffer X so that the output buffer X currently being prepared is filled with the sound waveform data and is ready, and the sound waveform is calculated and written to the output buffer X. It is. This processing is performed by the sound waveform A described above with reference to FIG. ₃ It corresponds to the operation generation processing such as.
[0056]
Each sample of the buffer X that is filled with the sound waveform and ready is further subjected to a low-pass filter (LPF) process in step S560, and the high-frequency component is cut. Next, in step S570, the area of the output buffer X that stores the LPF-processed sound generation waveform is reserved and registered in the waveform reproduction unit (DMA) 9 as a specific area that stores a reproduction waveform to be subsequently reproduced. As a result, the sound waveform of the specific area currently being reproduced and the specific area that has already been reserved is reserved to be reproduced after the reproduction is completed. In step S580, a new output buffer area different from the output buffer used as the buffer X until then is secured, all sample values are cleared to zero, and a sound waveform for the next section is created. Then, it is newly set as an output buffer X for preparation, and the sound source processing is completed.
If it is determined in step S520 that the calculation time has not been reached, the process ends.
[0057]
Next, FIG. 11 shows a flowchart of the sound waveform calculation process executed in the note-on process, the note-off process, and the sound source process. When this process is performed, the time range of the sound generation waveform in which the sound generation waveform is calculated in advance is determined as described above. That is, when this flow is executed as a MIDI data reception process such as note-on processing, the time range is the partial waveform described above. When this flow is executed during sound source processing, the buffer X This is the pronunciation waveform sample of the uncalculated part of all the samples. Note that when MIDI data is received, calculation is performed based on the calculation order determined in the immediately preceding sound source process, so the calculation order of the sound generation channels is not newly determined. When there is a new note-on, the order of all the other sound generation channels is lowered by one, and the new note-on channel is sequentially added to the first operation order.
[0058]
When the sound waveform calculation process is started, calculation preparation of the first musical sound waveform sample of the first sound channel (ch) in the calculation order is performed in step S600. The calculation preparation process is to prepare data such as the previous read address, various EG values, various EG states (attack, release, etc.), LFO value, etc. so that they can be used immediately for calculation, or to the internal register of the CPU 1 This is the process of loading. In step S610, LFO, filter G, and volume EG waveform calculations are performed to generate samples of the LFO, FEG, and AEG waveforms necessary for the calculation of the specified time range. The LFO waveform is added to the F number, FEG waveform, and AEG waveform to modulate each data. For the sound generation channel designated as the channel to be muted in step S540, a dump AEG waveform that rapidly attenuates within the above range is calculated and generated as the volume EG.
[0059]
Next, in step S620, the F number is repeatedly added using the previous read address as an initial value to generate a read address for each sample within the time range, and waveform storage in timbre data is performed based on the integer part of the read address. The waveform samples are read from the area WD, and the interpolated waveform samples are interpolated based on the decimal part of the read address to calculate all the interpolated samples within the time range. For example, when the time range corresponds to a time corresponding to 100 samples, processing is performed by this step for 100 samples. Here, the processing for a plurality of samples within the time range is repeated by adding the F number to the reading address and the processing from reading based on the address generated by the addition to the interpolation processing as unit processing. Therefore, the reading address is read into the CPU register only once as a whole, and the processing speed is increased.
[0060]
Further, in step S630, timbre filter processing is performed on the interpolated sample within the time range, and timbre control is performed based on the FEG waveform. In step S640, the timbre filter processing is performed on the filtered sample. Amplitude control processing is performed, and the amplitude control of the musical sound waveform sample is performed based on the AEG and the volume data, and the musical sound waveform samples for the time range subjected to the amplitude control processing are respectively set as samples corresponding to the output buffer X. The accumulated write process to be added is executed. In this process, for each sample within the time range, amplitude control and addition to the corresponding sample in the buffer X are continuously performed, so that the number of times the sample is taken into the CPU register can be reduced. , Processing speed has been improved.
[0061]
As described above, the musical sound waveform sample calculation generation process from step S620 to step S640 is basically performed so as to generate all the samples within the predetermined time range, but the volume EG in step S610. As a result of the waveform calculation, the range in which the level of the AEG waveform decreases and the volume is sufficiently attenuated is excluded from the calculation target, and the processing is reduced accordingly. In particular, sufficient sound attenuation is often obtained in the middle of the predetermined time range for the sound generation channel that has generated the dump AEG waveform in accordance with the instruction in step S540.
[0062]
In step S650, when it is desired to continue the waveform calculation process as it is, it is determined whether the waveform can be supplied to the DMA 9 within the time limit, and it is determined whether or not the calculation is terminated. Here, supplying the waveform within the time limit means that the DMA 9 that is reproducing the specific area for storing the previously generated sound waveform continues to the buffer X before the reproduction of the area is finished. A sound waveform is prepared, and reproduction of the area of the buffer X can be reserved. If it is not in time to continue as it is, it is determined that the calculation is to be aborted, the abort process is executed in step S670, and this sound waveform calculation process is terminated.
[0063]
If it is determined that the calculation can be continued, it is determined that it is not necessary to abort the calculation, and whether or not the calculation of the musical sound waveform for all the sound channels to be calculated in step S660 is completed. Is determined, and it is not determined that the calculation of all the sound channels has been completed, the first tone waveform sample of the sound channel assigned the next calculation order is designated in step S680, and the sound channel of the next sound channel is specified. Preparation for generating the musical sound waveform is performed. When the preparation is completed, the process returns to step S610, and the sound generation process over steps S610 to S640 is executed for the tone generation channel in the same manner as described above. In this way, the processing from step S610 to step S660 is repeatedly performed until the calculation of all sound generation channels is completed. For each sound channel processing, the generated samples for the predetermined time range are sequentially added to the corresponding samples in the buffer X in step S640.
[0064]
When it is determined in step S660 that the calculation generation process has been completed, the sound waveform calculation is ended. At this time, in the buffer X, an accumulated value obtained by accumulating the generated musical sound waveform samples of all sound generation channels to be calculated is newly stored for the number of samples corresponding to the time range.
On the other hand, if it is determined in step S650 that censoring is to be performed and the sound waveform calculation is completed through the censoring process in step S670, the generation of the calculation has been completed in buffer X by that time among all the sound generation channels to be calculated. Accumulated values of musical sound waveform samples for the sound generation channel are newly stored for the number of samples corresponding to the time range.
[0065]
For the sound generation channel assigned the calculation sequence after the truncation, the musical sound is not generated and generated, and as a result, the musical sound of the channel disappears. However, it is affected when the musical sound disappears by the processing of step S530. Since channels with fewer channels are considered to be the later calculation order, adverse effects due to truncation can be minimized. In the abort process in step S670, the channel register is set so that the channel that cannot be calculated once remains muted in the next and subsequent sound waveform calculations.
[0066]
When the sounding waveform calculation process described above is executed in step S340 of the note-on process shown in the flowchart of FIG. 9A, when the above-described sounding waveform calculation process ends, the process of step S350 of the flowchart is performed. Subsequently, the note-on process is completed.
When the sound waveform calculation process is executed in step S430 of the note-off process shown in the flowchart of FIG. 9B, when the above-described sound waveform calculation process ends, the process of step S440 of the flowchart continues. And the note-off process ends.
[0067]
Furthermore, when the sound waveform calculation process is executed in step S550 of the sound source process shown in the flowchart of FIG. 10, when the above-described sound waveform calculation process is completed, the processes after step S560 of the flowchart are continuously executed. When the process of step S580 is finished, the sound source process is finished.
These note-on processing, note-off processing, and sound source processing are repeatedly performed in circulation in the steady loop shown in FIG.
[0068]
In the musical sound generation method of the present invention, it goes without saying that data independent from each other, such as a step of instructing the generation of a plurality of musical sounds and a step of instructing the start of calculation at predetermined time intervals, for example. It is not necessary to perform the processes in the order described above, and the processes may be executed by changing the order of the processes.
In the musical sound generation method of the present invention, processing including a sound generation instruction requested from other application software can be performed using the idle time of the musical sound generation calculation processing described above. Other application software includes game software, communication software, and office processing software.
[0069]
In the above, the musical sound generating method of the present invention has been described as a program executed by the musical sound generating device shown in FIG. In addition, the musical sound generation method of the present invention may be executed in parallel with other application programs as one application program on a general-purpose computer that operates Windows (Microsoft OS for PCs) and other operating systems. .
[0070]
【The invention's effect】
The present invention is configured as described above. When the generation of new mixed sample data for each of a plurality of samples is completed, the new mixed sample data is reproduced after the reproduction of the previously generated mixed sample data is completed. Since the reservation is made so that the music is reproduced, it is possible to reproduce the musical sound without interruption.
Further, based on one-time reading of the control data from the channel register, waveform data for a plurality of samples of the sound generation channel from which the control data is read is generated, and the control data is written to the channel register after the waveform data is generated. Therefore, it is only necessary to prepare each tone generation channel once for the calculation of a plurality of musical sound waveform samples, and the overhead can be reduced. For this reason, the quality of the generated musical sound can be improved, and the number of simultaneously sounding channels can be increased.
Further, if the musical sound waveform sample is calculated every time a MIDI event is input, the calculation is distributed, and a decrease in the number of pronunciations due to the initial pronunciation process can be prevented.
[0071]
Furthermore, the performance information is accepted when performance information is generated, and the tone control or waveform generation based on the accepted performance information is included in the main step executed during the idle time of the acceptance processing. An increase in processing at the time when information is generated can be distributed within the free time, and a temporary increase in processing can be prevented.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a musical sound generating apparatus capable of executing a musical sound generating method of the present invention.
FIG. 2 is a diagram showing a timbre data area on a RAM.
FIG. 3 is a diagram showing an input buffer area on a RAM.
FIG. 4 is a diagram illustrating a ch register area on a RAM.
FIG. 5 is a diagram showing an output buffer area on a RAM.
FIG. 6 is a flowchart of a main routine of the musical sound generation method of the present invention.
FIG. 7 is a diagram showing a flowchart of a MIDI reception interrupt process of the tone generation method of the present invention.
FIG. 8 is a flowchart of MIDI processing in the main routine.
FIG. 9 is a flowchart of note-on processing and note-off processing in MIDI processing.
FIG. 10 is a flowchart of sound source processing in the main routine.
FIG. 11 is a diagram showing a flowchart of the sound waveform calculation processing of the musical sound generation method of the present invention.
FIG. 12 is a diagram showing a timing chart in the second embodiment of the present invention.
[Explanation of symbols]
1 CPU, 2 ROM, 3 RAM, 4 timer, 5 MIDI interface, 6 keyboard, 7 display, 8 hard disk, 9 playback unit, 10 DAC, 11 sound system

Claims (2)

A first step of receiving one or more generation instructions for generating one or more specified musical sounds;
In response to the generation command, each specified musical sound is assigned to one of a plurality of sound generation channels, and the control data of the specified musical sound is associated with each sound generation channel to which each specified musical sound is assigned. A second step of writing to the channel register
A third step of sequentially issuing operation start instructions;
A fourth step of generating waveform data for a plurality of samples for each tone generation channel based on the control data stored in the channel register of the channel in response to each calculation start command;
A fifth step of mixing the waveform data generated for each sound generation channel in the fourth step for each sample of the plurality of samples to generate mixed sample data;
When the generation of new mixed sample data for each of the plurality of samples is completed in the fifth step, the reproduction of the new mixed sample data is completed after the reproduction of the previously generated mixed sample data is completed. A sixth step of making a reservation so that
A sound generation method comprising: a seventh step of sequentially outputting the mixed sample data of the sample points in each sampling period in accordance with the reservation of the sixth step. A first step of issuing a generation instruction for generating a plurality of designated musical sounds;
A second step of assigning each designated musical tone to any one of the designated tone generation channels from among the plurality of tone generation channels, and writing and storing control data of the designated tone in the channel register of each designated tone generation channel;
A third step of issuing a calculation start command at a predetermined time interval;
In accordance with each calculation start command issued in the third step, the tone generation calculation for each channel is sequentially executed, and a plurality of samples are generated for each designated tone generation channel based on the control data stored in the channel register of the channel. A fourth step of arithmetically generating the minute waveform data;
A fifth step of mixing the waveform data generated for the sound generation channel designated in the fourth step for each sample to generate mixed sample data;
A sixth step of converting the mixed sample data for each of the plurality of samples into an analog signal for each sampling period;
In the musical tone generation calculation of the fourth step, a plurality of samples of the sound generation channel from which the control data is read out for each designated sound generation channel, based on a single read of the control data from the channel register. A sound generation method comprising: generating waveform data for a minute, and writing control data in the channel register after the waveform data is generated.

Images