JP2002518693A - Synthesizer system using mass storage device for real-time, low-latency access of musical instrument digital samples - Google Patents

Abstract

(57) [Summary] The present synthesizer system includes a CPU and a host memory that operates software routines. The software routine stores a first portion of each waveform signal in a sample pool (40) of host memory and stores the remaining portion of the selected musical sound on a hard disk drive without any audible recognizable delay. From the drive (22) to the stream cell array (42). Synthesizer systems utilize a caching system that enables low cost and high storage capacity devices to be used in audio synthesizer systems. A MIDI control signal is provided to the audio processor to select an appropriate digital waveform signal.

Description

DETAILED DESCRIPTION OF THE INVENTION

FIELD OF THE INVENTION [0001] The present invention relates generally to digital sound synthesizer systems.
In particular, the present invention relates to a sound synthesizer system utilizing a mass storage medium or device.

BACKGROUND OF THE INVENTION Digital synthesizers or other electronic systems generally generate sound or music and can be used as electronic musical instruments or electronic sound machines. Digital synthesizers are typically configured to accept input signals from a musician or operator interface and produce a digital output signal representing an analog signal in the audio frequency range. The interface may provide a keystroke signal, a mouse signal, a touchpad signal, or a keyboard signal that represents the musician's operation.

[0003] Digital output signals from digital synthesizers are converted to analog signals, directed to loudspeakers, tape recorders, mixers or other devices, and played back in the form of sound. Digital synthesizers are configured to provide an output signal that simulates the sound of a conventional known instrument. Alternatively, the synthesizer may be configured to simulate sounds emitted by a logical instrument having predetermined characteristics that are different from those of any conventionally known instruments.

[0004] The synthesis of music and sounds is a huge technical task. Real instruments produce a complex mixture of many different frequencies, commonly referred to as the "tone" of the sound. For example, percussion sounds, such as those created by drums, cymbals, etc., are non-periodic functions that cannot be completely explained by any simple mathematical formula. Thus, generating or synthesizing a digital signal representing sound as a rich and complex sound, such as the sound of a real instrument, is a very voluminous digital signal processing and sampling task. In addition, the synthesizer must respond to the nuances of manipulating the interface by the musician. For example, snare drums have many different sound characteristics when played at various locations, such as near the center of the drum, near the edge, or on the edge. Also, the characteristics of the sound of the snare drum differ depending on the striking power of the musician, the playing technique, and the methodical tone change.

[0005] To properly capture the behavior of a musical instrument requires a sufficiently large sample database for each of these features and their combinations, facilitated by large sampling capabilities. By having a sufficiently large waveform database, more realistic and expressive synthesis can be achieved. Numerous synthesizers using waveform sampling techniques have been widely used in the music and multimedia fields due to their ability to produce musical sounds that approximate and emulate the sounds of musical instruments.

[0006] Earlier synthesizer systems used a musical instrument digital interface (MIDI) to control a digital synthesizer.
Often use. The MIDI interface is a control signal or MIDI
Generate control data. The MIDI control data is stored in a specific musical tone (e.g., a central musical instrument (Mi
ddle C)).

[0007] Conventional synthesizer systems utilize large solid state memories that store digital waveform signals representing the true sound of each musical tone played on a particular instrument. This memory is a static random access memory (SRAM)
, A dynamic ram (DRAM) or a read only memory (ROM). When the musician activates a key or other interface, an appropriate waveform signal is selected depending on the strength of the key or keystroke activated. The waveform signal is converted to an analog output signal. The digital waveform signal can be combined with other tones played simultaneously before being converted to an analog output signal.

[0008] In this configuration, the synthesizer actually only plays the individual sounds of the musical sounds or digital recordings of the musical sounds. Each waveform signal is stored as a series of individual data words, each representing a single sample of the waveform at a particular time. To achieve satisfactory fidelity, any of such stored waveform signals
The stored sound must contain thousands of samples per second. The memory required to store such waveform signals is large, and therefore the solid state memory required to store all the required waveform signals is very large. Solid state memory is required because its speed allows for essentially real-time playback (eg, no audible delay). The high price per unit of sound sample storage devices of all types of solid state memory has hampered the use of large amounts of digital waveform signals for accurate representation of musical instruments and all their nuances.
A major limitation of current synthesizers is that there is not enough memory to store complete samples of the wide range of sounds associated with the instrument (eg, due to cost).

To reduce memory requirements, synthesizer systems have used techniques to store instrument samples more efficiently. These techniques generally result in lower quality sound production. The "looping" technique reuses sound samples. By duplicating and reusing groups of samples, the overall memory requirements are reduced. Other techniques store the samples in a compressed state in solid state memory. However, these systems use a decompression algorithm to implement
Requires large CPU power. Also, the decompression algorithm is lossy and degrades voice quality.

[0010] Other prior art synthesizer systems utilize a hard disk or other mass storage device to load instrument samples from them into solid state memory prior to musical tone generation or instrument play. However, these systems still require large amounts of solid state memory, since all digital waveform signals for musical instruments must be loaded into solid state memory before playing.

Some synthesizer systems utilize ROMs or other types of cheaper and slower solid state memory to store digital waveform signals. These synthesizers employ a data caching technique in order to move the waveform signal from a lower speed ROM to a high speed RAM such as a tone generating SRAM. Nevertheless, slower ROMs and faster RAMs have greatly increased the cost of synthesizers.

[0012] Other prior art systems store only one or several waveforms representing each instrument. These waveforms may be digital signal processing techniques or other electronic techniques (e.g., non-linear distortion) to reflect frequency and amplitude changes associated with different musical features, as indicated by MIDI control data. Adjusted by. For example, the frequency and amplitude of the sample waveform representing the center c of a piano can be adjusted to synthesize different piano tones and volumes. However, these types of synthesizers cannot produce complex mixtures or timbres to a sufficiently high fidelity for musically trained ears. In another example, some systems utilize digital filters to adjust the harmonic content of a particular musical tone. However, these systems require significant CPU power and can suffer from voice quality degradation.

Thus, without using a large amount of expensive solid state memory,
There is a need for a music synthesizer that can store a large number of music samples. Furthermore,
Although mass storage devices are available, there is a need for a music synthesizer that can provide real-time generation of musical sounds.

SUMMARY OF THE INVENTION The present invention is directed to an audio processor that provides a digital output signal representing sound at an output. The audio processor has a mass storage device input for receiving a plurality of digital waveform signals, a control input for receiving a digital control signal, and a control circuit connected to the control input and the mass storage device input. The digital control signal indicates at least one selected digital waveform signal of the plurality of digital waveform signals. The control circuit stores the first portion of the digital waveform signal in a host memory. The control circuit provides (provides) a first portion of the selected digital waveform signal to the output from the host memory in response to the digital control signal, followed by a remaining portion of the selected digital waveform signal. From the mass storage device input to the output. The control circuit combines the first portion of the selected digital waveform signal with the remaining portion of the selected digital waveform signal without loss of continuity between the remaining portion of the selected digital waveform signal. provide.

The present invention comprises a mass storage device means for storing a plurality of digital waveform signals, a control input means for receiving a digital control signal, and a host memory means for storing a first portion for each of the digital waveform signals. , A processor means for producing a digital sound signal at the output. Each of the digital waveform signals corresponds to a specific sound among a plurality of sounds. The digital control signal indicates a selected sound among the sounds. The processor means provides a first portion of the digital waveform signal corresponding to the selected sound to an output from the host memory and a second portion of the digital waveform signal corresponding to the selected sound to the mass storage device means. From output to provide. The processor means utilizes caching techniques to provide a second portion to the output.

The present invention further relates to a memory architecture for a digital synthesizer system. A digital synthesizer system includes a mass storage device and a processor. The mass storage device stores a plurality of digital waveform signals. Each of the digital waveform signals corresponds to a specific sound among a plurality of sounds. The processor receives the digital control signal. The digital control signal indicates a sound selected from the plurality of sounds. The audio processor generates a digital sound signal at an output in response to the digital control signal. The processor provides a first portion of the digital waveform signal corresponding to the selected sound, followed by providing a second portion of the digital waveform signal corresponding to the selected sound from the mass storage device. The memory architecture has a sample buffer and a stream buffer. The sample buffer stores a first portion of each of the digital waveform signals. The stream buffer temporarily stores a second portion of the digital waveform signal corresponding to the selected sound.

The present invention further relates to a method for digitally synthesizing sounds in a synthesizer system. The synthesizer system includes a mass storage device, a host memory, and a processor. The mass storage device stores a plurality of digital waveform signals. The processor has a digital control signal input and a digital output. The processor is connected to the host memory and the mass storage device.
The method includes downloading a first portion of each of the digital waveform signals into a host memory, receiving a digital control signal indicative of a selected one of the digital waveform signals at a digital control signal input, and selecting a selected one of the digital waveform signals. Providing a first portion of the digital waveform signal to the digital output from the host memory and providing a second portion of the selected digital waveform signal to the digital output from the mass storage device. The first part and the second part are provided such that there is no speech recognizable delay between the first part and the second part.

The invention further relates to a method for digitally synthesizing sounds in a synthesizer system. The synthesizer system includes a mass storage device, a host memory, and a processor. The mass storage device stores a plurality of digital waveform signals. Each of the digital waveform signals corresponds to a specific sound among a plurality of sounds. A first portion for each of the digital waveform signals is stored in host memory. The processor has a digital control signal input and a digital output. The processor is connected to the host memory and the mass storage device. The method receives a digital control signal indicative of a selected one of a plurality of sounds at a digital control signal input and provides a first portion of a digital waveform signal corresponding to the selected sound from a host memory to a digital output. , Providing a second portion of the digital waveform signal to a digital output corresponding to the selected sound from the mass storage device. First
And the second part are provided such that there is no lack of continuity between the first part and the second part.

The present invention also relates to an audio processor for providing a digital output signal indicative of sound and output. The audio processor includes a mass storage device input for receiving the plurality of digital waveform signals, a sample pool for storing a first portion of the digital waveform signal, a stream buffer, and a mass storage device input and stream buffer. It has a stream engine connected to it and a synthesizer connected to the stream buffer and output. The stream engine is
A second portion of the digital waveform signal is provided from a mass storage device input to a stream buffer. The synthesizer engine provides a second portion of the digital waveform signal from the stream buffer to an output.

According to one exemplary aspect of the invention, a digital synthesizer stores a first portion of the digital waveform signal in host memory and caches a remaining portion of the digital waveform signal from a mass storage device. . The digital music synthesizer utilizes a hard disk drive as a storage device and is preferably implemented in software executed by a computer. The computer has a central processing unit and host memory operating in a Windows-based environment.

In another exemplary aspect of the invention, the synthesizer system utilizes an audio processor configured in software. The audio processor includes a solid state sample pool, a stream engine, a synthesizer engine, and a solid state stream buffer.
The stream engine provides digital waveform samples or portions of the signal from the mass storage device to the stream buffer. The synthesizer engine
A portion of the digital waveform signal is provided from a sample pool and a stream buffer to a digital output. Preferably, the processor is responsive to a MIDI control signal to select a particular digital waveform signal.

In yet another exemplary embodiment of the invention, a memory architecture for a digital synthesizer utilizes a host memory with a sample pool and an array of stream cells. The sample pool stores a first portion of each digital waveform signal stored on the hard drive. The sample pool is
Each has an array of sample cells including a deck buffer and a control data buffer. Each cell of the array of stream cells preferably includes a deck buffer pointer, a stream buffer, and a control data buffer. The deck buffer stores a first portion of the waveform. The stream buffer is used to store subsequent portions of the waveform stored on the hard drive.

Preferred embodiments of the present invention will be described below with reference to the accompanying drawings. In the drawing,
Like numerals indicate like elements.

DETAILED DESCRIPTION OF THE PREFERRED EXEMPLARY EMBODIMENTS OF THE INVENTION Referring to FIG. 1, a digital synthesizer system 10 includes an audio
A processor 12, a sound selection source 18, such as a MIDI source, a sound source 22, such as a mass storage device for storing musical instrument digital waveform signals, a digital-to-analog (D / A) converter 25, and a speaker system 26. And Audio processor 12 has a digital output 28 connected to a converter 25.
The converter 25 has an output connected to a speaker system 26.

In operation, the audio processor 12 receives a digital control signal from the sound selection source 18 at a digital control input 20 and a digital waveform signal source 22
Receives a digital sample or digital waveform signal at a mass storage device input 24 from a USB device. Audio processor 12 processes a digital control signal at input 20 and a digital waveform signal at input 24 to provide a digital output signal at output 28. The digital output signal is provided to converter 25, which provides the analog signal represented by the digital output signal to speaker system 26.
When processor 12 receives a number of control signals at input 20, it combines, mixes, or mixes several different digital output signals with different digital output signals representing the digital waveform signal indicated by the control signal at input 20. At the output 28. Speaker system 26 produces sound or music that is representative of the digital output signal provided at output 28.

The converter 25 and the speaker system 26 can be replaced by other devices connected to the digital output 28. For example, a digital storage device such as a mixer, recorder, hard disk drive, or other audio device can be connected to the output 28 of the processor 12. The speaker system 26 can be a stereo system with two or more speakers that provides stereo generation of the sound indicated by the digital output signal, which preferably represents stereo sound.

The sound selection source 18 can be any device that generates control signals, most preferably MIDI data or control signals. MIDI control signals are well known in the art and may be specific musical sounds, such as music events, such as the occurrence of a central musical tone, such as the occurrence of a particular musical tone, such as a piano, horn, drum, or other device. Represents Alternatively, the audio processor 12
May be utilized with different types of sound selection sources 18 that provide control signals at input 20.

The sound selection source 18 is a mouse interface, piano keyboard, electronic drum pad, keypad, touchpad, button interface, or other musician interface that provides digital control signals to the input 20. be able to. The sound source 22 is preferably a mass storage device such as a hard disk drive, optical disk drive, tape source, floppy disk drive, or other memory device. The sound source 22 stores a wide range of digital samples of one or more instruments. For example, the sound source 22 may have approximately one unit associated with a grand piano.
100 independent digital waveform signals can be stored. The independent waveform signal is
Represents 88 keys of a piano played with six variations of stereo timbre (for example, 88 × 6 × 2 = 1,056). The number of waveform signals can be adjusted for various applications and musical instruments. The description of the use of system 10 to reproduce the sound of a grand piano is not for purposes of limitation, but merely for purposes of illustration. Music types, source types, instrument types, and sample sizes are discussed herein only in exemplary ways.

The system 10 can be implemented on a PC platform as an embedded system, as a music synthesizer, as a dedicated rack-mounted music sampler, or as another sound generator. System 10 is preferably a computer-based system with a CPU, hard disk drive, and host memory. The CPU is configured with software to perform the operations of the audio processor 12 described throughout this application. Alternatively, system 10 can be a dedicated digital signal processor system, a general-purpose processor system, or a hardware system. The host memory is preferably at least 32 megabytes of DRAM, but other types of solid state memory could be used. Host memory is
It can be ROM, SRAM, flash memory or other device. Preferably, the CPU is a 486 or higher processor.

The system 10 is configured to provide the digital waveform samples or signals stored at the source 22 to the output 28 in real time, ie, without any audible recognizable delay. When the musician operates the sound selection source 18, the source 1
With no speech recognizable latency between the operation of FIG. 8 and listening to music from the speaker system 26, the audio processor 12 provides a digital output signal at output 28 representative of the operation. Audio processor 12 includes a source 18
It is preferable that a digital output signal can be provided in less than 15 milliseconds from the above operation.

As used herein, the term audible recognizable delay refers to a lack of continuity, improper latency, stall, or hold of music or sounds audible to the human ear. The term phonetic recognizable delay is used in two different contexts. The first context relates to such a time lag between the operation of the musician and the production of sound. The second context relates to the continuity of the signal between the first and the rest of the digital waveform signal. The second context also concerns the continuity of the signal between the remaining parts of the digital waveform signal. The second context requires that the delay be less than a few microseconds (eg, so that there is no lack of continuity) to prevent artifacts detectable by the human ear. System parameters and application criteria can affect the definition of speech recognizable delay and latency adequacy. Most preferably, there is no delay between portions of the digital waveform signal when provided at output 28.

Because of the configuration of the audio processor 12, the system 10 is a slower and less expensive mass storage device (eg, a hard disk drive)
Can be advantageously used as the source 22 and still provide a digital output signal at the output 28 in sufficient time. The use of source 22 to store large amounts of digital waveform signals greatly reduces the size of solid state memory (not shown) required by processor 12 for real-time generation of sound (eg, , Thereby reducing the cost of processor 12). Unlike conventional synthesizer systems that download all of the digital waveform signals from the source 22 to solid state memory, the processor 12 operates such that most of the digital waveform signals are not stored in the processor 12 solid state memory. Use a caching system. Most of the digital waveform signal is only stored in solid state memory when actually needed.

More specifically, processor 12 utilizes its solid state memory (not shown) to store only a first portion of each of the digital waveform signals stored at source 22. . When a musician selects a particular instrument, processor 12 downloads a first portion of each digital waveform signal stored for the selected instrument. Alternatively, the musician may select one instrument or a portion of several instruments, and processor 12 may download a first portion of each digital waveform according to the musician's selection.
Thus, the first portion of the digital waveform signal stored in the solid state memory of audio processor 12 is one-to-one with the digital waveform signal stored in source 22 for the selected instrument or group of instruments. It is preferable to cope with the relationship.

Once the first portion has been loaded into the solid state memory of audio processor 12, audio processor 12 is responsive to digital control signals provided at input 20. Audio processor 12 provides a first portion of the selected digital waveform signal at its output 28 from its solid state memory in response to a digital control signal provided at input 20. Preferably, at least the first 0.5 seconds of the digital waveform signal is stored in solid state memory. The remaining portion of the selected waveform signal is then passed from source 22 to processor 1
2 and output 28. The remaining part can be searched while the first part is playing. The remainder can include portions of the digital waveform signal representing up to three minutes of sound. Accordingly, processor 12 may provide a speech recognizable delay or continuity between providing the first portion of the digital waveform signal stored in solid state memory and the remaining portion stored in source 22. Utilize a caching mechanism to reduce or eliminate the lack of

Referring to FIG. 2, the audio processor 12 has an input 24 connected to a source 22, which is shown in FIGS. 2 and 3 as a hard disk drive. The audio processor 12 includes a sample pool 40, an array 42 of stream cells 43, an instruction parser 50, an instruction queue 52, a synthesizer engine 66, a download module 64, a system thread
It has a manager 68 and a stream engine 80. Instruction parser 50
Are connected to the input 20 and the instruction queue 52. The instruction queue is connected to a control input of the synthesizer engine 66.

Input 24 is connected to download module 64 and stream engine 80. The download module 64 is connected to the sample pool 40. The sample pool 40 includes a plurality of sample pool cells 41.
A sample pool cell 41 is provided for each digital waveform signal of a particular instrument, a portion of an instrument, or a group of instruments selected by a musician. The sample pool 40 can be configured into various numbers of cells. In the example of a grand piano, pool 40 has about 1100 cells 41.

The stream engine 80 is also connected to the system thread manager 68. System thread manager 68 operates as a master to stream engine 80. The manager 68 cooperates with the computer operating system to provide the C
Ensure that the required operating time of the PU (not shown) is received by the stream engine 80. The manager 68 preferably keeps the stream engine 80 running consistently when the CPU operates in a multitasking environment, such as a Windows environment.

The operation of manager 68, engine 66, engine 80, module 64, and parser 50 is preferably performed by a CPU of a PC platform (not shown). The storage operation of the array 42, pool 40 and queue 52 is P
This is done by the C platform host memory.

The stream engine 80 is also connected to the array 42. Stream engine 80 may provide a portion of the digital waveform signal from source 22 to stream cell array 42. That portion of the digital waveform signal is stored in cell 43 of cell array 42. Referring to FIG. 4, each of the cells 43 includes a stream buffer 47 and an execution time control data buffer 49. The stream cell array 42 is preferably resizable in about 24 to 256 cells, but any size array 42 is possible. Cell 4
Each of 3 represents a single audio or stream of polyphony, where 2
One voice polyphony represents simultaneous musical tones (eg, a combination of C and E musical tones). Therefore, the relationship between the one-to-one audio and the cells 43 of the stream is utilized by the processor 12.

The stream engine 80 manages the flow of data or digital waveform signal portions to the stream cell array 42. Engine 80 maintains a partial stream of data or signals from source 22 to cell 43. The stream of data can be up to the number of cells 43 and can be maintained as a virtual stream from source 22.

The stream engine 80 is a state machine configured to move portions of the digital waveform signal from the source 22 to the array 42 so that the processor 12 can generate sound in real time or near real time. Preferably, there is. The stream engine 80 converts a portion of the digital waveform signal to cell 4 of array 42
When provided to 3, synthesizer engine 60 consumes a portion of the digital waveform signal in cell 43 and provides a digital waveform output at output 28 representing the selected sound.

When the stream engine 80 fills the cells 43 of the array 42 with a portion of the digital waveform signal, the stream engine 80 performs synthesis and error correction and decompression operations such as lossless second order decompression. In this application, stream engine 80 is shown and described as throughput from input 24 to array 42 for illustrative purposes, but in effect stream engine 80 moves data from source 22 to array 42 (e.g., , From source 22 to array 42
Perform more transmission management operations of the signal to).

When an instrument, group of instruments, or portion thereof is selected by the musician, download module 64 receives from source 22 a respective first portion of the digital waveform signal for the particular instrument. Download module 6
4 stores the first portion of the digital waveform signal in the sample cell 41 of the sample pool 40. Alternatively, the first portion of the digital waveform signal may be stored permanently in pool 40 or received from another source (not shown).

Referring to FIG. 5, sample pool cell 41 of sample pool 40 (FIG. 2) includes a deck buffer 51 and a global control data buffer 53. The deck buffer 51 of the sample cell 41 may store the first 0.5 seconds of each of the digital waveform signals (approximately 64 Kbtie, depending on system 10 characteristics such as RAM size, CPU speed and hard disk speed). preferable.
Of course, if source 22 and engine 80 can provide signals to array 42 more quickly, smaller samples can be stored.

The global control data buffer 53 contains control information or global control data. Global control data refers to control data for a particular sample. Global control data is available to system 10 at load time.
The global control data provides characteristics of the sample or digital waveform signal to the system 10, including control words for the sampling rate, sample length, playback parameters, and compression algorithm. The control words are used to decompress specific packets or portions of the digital waveform signal.

The memory space does not store the first portion of the digital waveform signal in cell 43, but merely addresses the corresponding deck buffer 51 in pool 40 with a pointer in buffer 49 of cell 43 (shown in FIG. Can be saved by using Similarly, memory space can be saved by simply providing a pointer (not shown) pointing to global control data buffer 53.

In operation, when a digital control signal, such as a MIDI control signal, is provided at input 20, parser 50 manages the assignment of all new stream cells 43. Parser 50 responds to the asynchronous digital control signal at input 20 to refer stream cell 43 to the appropriate sample cell 41. The parser 50
In response to the on / off message received at input 20, it manages the allocation and deallocation of active stream cells 43. For example, when a musician strikes a key, an event is sent to parser 50 to turn on a particular sample cell 41. The parser 50 finds an appropriate cell 41 from the sample pool 40 based on the speed at which the key was pressed, along with the current state of the synthesizer system 10. Parser 50 is responsible for stream cell 4 from available cell 43.
3 is responsible for assigning 3 to array 42.

The parser 50 first searches for an unused stream cell 43. Parser 50 uses the runtime control data stored in buffer 49 of stream cell 43 to determine its availability. If all stream cells 43 have been assigned, parser 50 uses the following rules to determine which stream cell 43 to interrupt. 1. If applicable, parser 50 looks for a voice with the same characteristics as a new voice. For example, the middle C on the keyboard is currently playing and the system 1
If the musician taps another center C while 0 is at maximum polyphony, parser 50 will cause the old center C to be interrupted by the new center C. 2. Parser 50 finds the audio with the lowest execution time volume value in buffer 49 (eg, cell 43) and selects it as the interrupted audio.

In allocating the stream cell 43, the parser 50
To set the execution time control data. The execution time control data refers to a single sample cell 41. The runtime control data also signals both the stream engine 80 and the synthesizer engine 66 to begin processing this particular cell 43. At the time of new voice allocation, parser 50 sends stream cell 4
Responsible for all initialization of 3.

Similarly, parser 50 passes all control signals to synthesizer engine 66 via instruction queue 52. These instructions are sent by the engine 66
It can be controlled to control playback as defined by industry accepted standards such as 1.0 use. For example, a natural end of speech can signal that a musician has released a key on an instrument. This off signal allows the synthesizer engine 66 to naturally lower the active sound in a manner consistent with the sound of the acoustic instrument.

The parser 50 initializes the execution time control data by setting a flag “new voice”. The flags are used with the stream engine 80 for priority scheduling and error correction. Parser 50 also initializes a progress counter for the new voice / stream cell 43. This progress counter is maintained by the stream engine 80 and the synthesizer engine 66 for priority scheduling and stream data integrity.

The parser 50 initializes the execution time control data in the buffer 49 to point to the deck buffer 51 of the associated sample cell 41. Further, the parser 50 sets the stream buffer 47 corresponding to the stream cell 43. Next, the synthesizer engine 66 starts processing from the deck buffer 51. When data is consumed from the deck buffer 51, the stream engine 8
0 starts filling the stream buffer 47 of the cell 43. Eventually, the synthesizer engine 66 consumes all data available in the deck buffer 51. Once all of the data in buffer 51 has been consumed, engine 66 dynamically switches its internal pointer to begin consuming data from stream buffer 47 in cell 43. Engine 66 signals stream engine 80 by setting the appropriate flag in the run time control data and incrementing a progress counter.

Referring to FIG. 3, a stream engine 80 having an input 24 connected to the source 22 is shown. Engine 80 is also connected to system thread manager 68. The output of the stream engine 80 is connected to the array 42 of cells 43.

The stream engine 80 includes a priority stream scheduler 82, a buffer management selector 84, a stream count manager 86,
A disk synchronizer 88 having a control output connected to the control word
It includes a parser 90, a data mover 94, and a decompressor 96. data·
The mover 94 is also connected from the instruction parser 50 to the instruction synthesis circuit 102.

The stream engine 80 transmits data to the source 22 while maintaining the real-time integrity of the sampled audio (eg, digital output signal).
Is responsible for supplying the stream cell 43 operating from Engine 80 acts as a producer to individual cells 43 while synthesizer engine 6
6 works as a consumer. Synthesizer engine 66 and stream
The engine 80 is connected by execution time control data,
Initialized by parser 50 when a new voice is assigned. Runtime control data ensures audible integrity of the entire system 10 by ensuring that consumers (eg, engine 66) and producers (eg, engine 80) of the data remain constantly synchronized.

The stream engine 80 transfers data to an n-channel (eg, n-cell 43)
Where n is the current polyphony level. Data is streamed from the source 22 to the stream buffer 47 of the appropriate stream cell 43. The order in which the plurality of stream cells 43 are processed is determined at runtime by the priority stream scheduler 82. Priority stream scheduler 82
Is based on a priority level scheduling mechanism. The cell 43 is dynamically assigned a priority level. Cell 43 is treated at the highest level processed first. Priority levels are assigned to cells 43 based on the current position of the data stream. When a new voice is assigned, the data stream is initialized to the starting position of source 22. Specific cells 43 are assigned to new stream positions. The stream cell 43 flagged at the new stream position is given the highest priority in the system 10. If there are multiple cells 43 with a new set of stream flags, a round robin method is implemented.

Stream cells 43 that do not have a new set of flags are stored in buffer 4
9 is given a priority based on the progress counter which is a component of the execution time control data. This counter increments when a single quadrant of data is consumed from the stream buffer 47 by the synthesizer engine 66. The priority level is determined by the value of the progress counter. This counter is initialized to zero by parser 50 when a new voice is assigned. If multiple cells 43 have the same priority level, a round robin scheduler is implemented.

The stream engine 80 is a state machine that is given control of the system 10 via the system thread manager 68. . The system thread manager 68 is responsible for allocating all of the remaining bandwidth of the host processor (not shown) to the stream engine 80.

The priority stream scheduler 82 determines if the stream cell 43, if any,
Determine which needs to be processed next. If the appropriate stream cell 43 has been tagged for processing, control is passed to the buffer management selector 84. The buffer management selector 84 initiates a transfer from the source 22 to the stream buffer 47 based on the currently used buffer scheme. To guarantee real-time accuracy for various hardware systems,
System 10 implements two different buffering schemes. The choice of buffering scheme is a function of system parameters, including CPU bandwidth, source 22 speed (including data access time and seek time). System 10 employs two buffering schemes, a single quadrant transfer and a two-quarter transfer.

All stream buffers 47 in cell 43 can be viewed as four quadrants. The synthesizer engine 66 consumes data every quarter.
Each time a quadrant is consumed, the progress counter increments. The stream engine 80 fills the stream buffer 47 with either a single quadrant or two quadrants at a time, depending on the buffering scheme currently employed. Stream engine 80 potentially maintains n streams of real-time data, which may be split across source 22, where n is the polyphony of system 10. In accessing multiple streams, two parameters are a problem for efficient use of source 22, setup time (eg, seek time), and data rate. Each time a stream cell 43 needs to be processed, the system 10 incurs setup time costs.
The system 10 uses a disk synchronizer 88 to set up the source 22 from the exact location of the current stream. 2/4 section buffering
In the scheme, the cumulative time required to set up the stream over the lifetime of the stream is half that required when using a single quadrant buffer.

The downside to using a two-quarter buffering scheme is that
Each time a cell 43 needs to be processed, twice as much data has to be moved from the source 22 to the stream buffer 47. Because stream engine 80 maintains single thread execution when processing stream cell 43, if single cell 45 maintains control of source 22 for too long, other streams have the potential to be depleted. Have. The buffering scheme for a particular stream cell 43 is stored in the runtime control data buffer 49. The stream count manager 80 manages a progress counter for any given stream cell 43 with the purpose of indicating the current state of the stream engine 80. The progress counter checks the stream engine 8 for more data.
Used by synthesizer engine 66 as a way to signal a zero. Similarly, a progress counter can be used to indicate that data is available.

After the buffer management selector 84 selects the appropriate buffering scheme, the disk synchronizer 80 sets up the source 22 by issuing instructions to move the reader to the correct location on the source 22. . The exact location on source 22 for any stream cell 43 is maintained in runtime control data buffer 49. The execution time control data is updated each time the data passes through the stream manager 80.

The system 10 is responsible for the compressed data. Any compression algorithm can be implemented. However, for tone generation applications, it is preferable to implement a lossless scheme. A lossless algorithm ensures that quality is not degraded due to the compression / decompression process. The music samples are compressed before storage at source 22. This embodiment implements a second order differential compression / decompression algorithm. Decompression of the stream is handled by decompressor 96. The compressed stream is packetized. Each packet contains a single control word used as control data for decompressor 96. The control words are initially stored in source 22 with the instrument. When an instrument is loaded into system 10, download module 64 transfers control words from source 22 to global control data buffer 53 of cell 41 of pool 40. The decompression algorithm can accept various packet sizes, provided it is 2 ⁿ in size. The packet size offsets the compression ratio of the sampled instrument. This embodiment uses a packet size of 2 kilowords.

The data mover is responsible for directing the sampled instrument from the source 22 to either a second order decompressor or stream buffer 47. Before the data is moved, the instruction synchronization module 102 is executed. Module 102 is responsible for handling any error correction and any synchronization between stream engine 80 and parser 50. Instruction synchronization 102 handles the aforementioned interrupts of stream cell 43. If the stream cell 43 is deallocated early because the system 10 has reached the maximum polyphony and a new voice has been started, the stream engine 80 will return to the active stream cell before proceeding. The integrity of cell 43 must be checked. This situation is the result of the stream engine 80 and the parser 50 operating asynchronously.

Referring to FIG. 2, system 10 transfers a first portion of the digital waveform signal stored in deck buffer 51 of cell 41 and pool 40 from pool 40 to output 28 via synthesizer engine 66. Rather than providing directly, it can be configured to provide to cells 43 of array 42. In this manner, engine 80 or engine 66 moves a first portion of the digital waveform signal from pool 40 to cell 43 of array 42. Next, synthesizer engine 6
6 moves a first portion of the digital waveform signal from cell 43 of array 42 to digital output 28.

It is desirable that the deck buffer 45 be composed of 64 Kbytes.
The stream buffer 47 is preferably composed of 2 to 4 quadrature phases of 32 kilobytes (Kb). The control buffer 49 is composed of 200 bytes. The source 22 has a seek time of about 7 milliseconds (ms) and a
It has a data rate of S, but other times are possible. Source 22 is preferably a one gigabyte or larger hard disk drive.

Although the detailed drawings, specific examples, and specific component values given describe preferred exemplary embodiments of the present invention, it is understood that they are for purposes of illustration only. Will be. The apparatus and method of the present invention are not limited to the details and conditions as disclosed. For example, the system is being used as a software-based computer-based synthesizer, but other hardware and software schemes can be used. Also, while specific memory sizes and periods have been discussed, they are discussed for illustrative purposes only. Further, a single line in various drawings may represent multiple leads. Various changes may be made to the details disclosed without departing from the spirit of the invention as defined by the appended claims.

[Brief description of the drawings]

FIG. 1 illustrates a digital audio device with an audio processor, according to an exemplary embodiment of the invention.
FIG. 2 is an exemplary block diagram of a sound synthesizer system.

FIG. 2 shows the audio engine shown in FIG. 1 with a stream engine, an array of stream buffer cells, and a sample pool including sample pool cells.
FIG. 3 is a more detailed block diagram illustrating a processor.

FIG. 3 is a more detailed block diagram of the stream engine shown in FIG.

FIG. 4 is a block diagram of one of the stream buffer cells shown in FIG.

FIG. 5 is a block diagram of one of the sample pool cells shown in FIG.

──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Vivo, Joseph A. USA, 78759, Texas, Austin Bluegrass Drive, 8707 F-term (reference) 5D378 AD01 AD21 BB02 BB07 BB22 XX11

Claims (10)

[Claims] 1. A digital synthesizer system having an input and an output and using a plurality of digital waveform signals, the mass storage device storing at least a second portion of each of the plurality of digital waveform signals. A memory for storing a first portion of each of the plurality of digital waveform signals; a memory connected to the memory and the mass storage device, the first one of the signals selected from the plurality of digital waveform signals; Providing a portion of the one of the plurality of digital waveform signals to the output from the memory;
And a control circuit for providing the portion from the mass storage device to the output. 2. The method according to claim 1, wherein the second part of the selected signal among the plurality of digital waveform signals is performed after the first part of the selected signal among the plurality of digital waveform signals is selected. The digital synthesizer system according to claim 1, wherein the digital synthesizer system is stored in the memory. 3. The digital synthesizer of claim 2, wherein said second portion of said selected one of said plurality of digital waveform signals is retrieved from said memory when stored in said memory. ·system. 4. The digital synthesizer system according to claim 1, wherein said control circuit includes a CPU. 5. The digital synthesizer system according to claim 1, wherein said mass storage device comprises an optical disk drive, a tape drive, a hard disk drive, or a floppy disk drive. . 6. A digital synthesizer system having an input and an output and using a plurality of digital waveform signals, storing a first portion and a second portion of each of the plurality of digital waveform signals. A first memory having a relatively slow access timing, a second memory having a relatively fast access timing, and a plurality of the plurality of memories connected to the first memory and the second memory. A control circuit for providing the first portion of a selected one of the digital waveform signals to the output and subsequently providing the second portion, the control circuit comprising: A digital synthesizer system for selectively moving the first portion of the selected signal into the second memory. 7. The method of claim 1, wherein the first portion of the selected signal of the plurality of digital waveform signals comprises: after the control circuit provides the first portion of the plurality of digital waveform signals to the output. 7. The digital synthesizer system according to claim 6, wherein the digital synthesizer system is stored in the second memory. 8. The method of claim 7, wherein the first portion of the selected one of the plurality of digital waveform signals is received from the second memory when stored in the second memory. The described digital synthesizer system. 9. The digital synthesizer system according to claim 6, wherein said control circuit includes a CPU. 10. The digital synthesizer system according to claim 6, wherein the first memory includes an optical disk drive, a tape drive, a hard disk drive, or a floppy disk drive. .