JP2005537506A - Sound synthesizer - Google Patents

Abstract

A sound synthesiser is provided that reduces the computational requirements of a synthesiser with a high degree of polyphony, while ensuring that audible artefacts are kept to a minimum. The synthesiser comprises a plurality of samples stored in a memory; a plurality of voices each comprising means for calculating an output using a plurality of samples selected from the plurality of samples stored in the memory; wherein a voice is active when calculating an output; wherein the number of samples selected by the means for calculating depends upon the number of active voices. <IMAGE>

Description

  The present invention relates to a sound synthesizer, and more particularly to a sound synthesizer for use in a device with limited computational resources such as a portable device.

  Modern sound synthesizers need to have a large number of voices. The number of voices that a synthesizer has is defined as the number of sounds that can be generated simultaneously.

  There are several different protocols and standards that define how an electronic sound synthesizer plays the required set of sounds.

  One common method of generating sound within an electronic device is to use a MIDI (digital interface for musical instruments) protocol. Unlike digital audio files, MIDI files (as found on compact discs) do not contain specific sound items. Instead, it contains a list of events that a device must perform in order to reproduce the correct sound. Sampled sounds are stored in the synthesizer and accessed by instructions contained within the MIDI file. Therefore, MIDI files are much smaller than digital audio files and are suitable for environments where storage memory is limited.

  In General Midi System Level 1 (GM-1), one synthesizer needs to have at least 24 voices.

  A synthesizer that generates sound from pre-recorded sounds such as a MIDI synthesizer is known as a synthesizer based on a wave table. In such a synthesizer, one or more pre-recorded sequences of a musical instrument are stored in a wave table. Each sequence contains a series of samples that are played to reproduce the sound.

  Depending on the instrument, a large number of notes can be generated, and since a large amount of memory is required to sample and record all possible notes, only a small number of notes are stored.

  Thus, if a synthesizer is required to generate a note or sound that has a different frequency than the frequency stored in memory, the synthesizer uses one of the stored sequences and a technique called “sample rate conversion”. And resample it to change the frequency to get the requested tone.

  Changing the frequency of the stored sequence is accomplished by accessing the stored samples at different rates. That is, for example, if a stored sample represents a musical note at a frequency of 300 Hz, the musical note can be reproduced at 300 Hz by sequentially accessing all samples. . If each stored sample is output twice before the next stored sample is read, this musical note played by the synthesizer will have a frequency of 150 Hz. Similarly, every other stored sample is read if a 600 Hz musical note is required.

  It is important that the rate at which samples are output by the synthesizer is constant and equal to one sample period (the time between each stored sample).

  In the example above, artifacts (distortion) are introduced into the output sound by accessing all samples twice. To overcome this distortion, the synthesizer calculates additional samples based on the stored samples. Thus, in the 150 Hz example above, instead of repeating each stored sample twice, the synthesizer outputs one stored sample and calculates the next sample based on the stored surrounding samples.

  To do this, the synthesizer requires an interpolation technique.

  The simplest interpolation technique uses a weighted average of two surrounding samples. However, this technique is often inaccurate and still results in audible distortion.

  The optimal interpolation algorithm uses a sin (x) / x function and requires an infinite number of calculations. Of course this is impractical and therefore a suboptimal algorithm has been developed.

  One suboptimal interpolation technique is Dana C. et al. As described in Chapter 8 of “Applications of DSP to Audio and Acoustics” by Massie, where several stored samples are used in the calculation. (The number of samples used for interpolation is known as the interpolation degree). The more samples that are used for interpolation, the better the performance of this synthesizer.

  In a synthesizer, each voice is implemented using one or several digital signal processors (DSPs), and the computing power of this DSP system places a limit on the number of voices that a synthesizer can generate, and for each voice Limit the interpolation order used.

  When using a sub-optimal interpolation algorithm, such as the truncated sin (x) / x algorithm, the computational complexity increases linearly with the interpolation order.

  Interpolation order 10 is often used in many commercial synthesizers because it provides a good trade-off between computational complexity and sound quality.

  It is desirable to be able to implement a MIDI sound synthesizer in a portable device such as a mobile phone so that these devices can generate polyphony ringing and higher sound quality.

  However, limitations placed on computing power in portable devices (such as cost and space available in the device) are compliant with General Midi System Level 1 (GM-1) (ie have 24 voices) 10 It is not enough to allow implementation of a sound synthesizer with front and rear interpolation orders.

  Accordingly, it is an object of the present invention to provide a sound synthesizer that reduces the computational requirements of a synthesizer with higher order polyphony and minimizes audible artefacts.

  Thus, according to one aspect of the present invention, a memory including a plurality of stored samples and means for calculating an output signal for each of a plurality of active voices, wherein the storage for each of the active voices Using the plurality of samples selected from the selected samples, and the number of samples used for each active voice by the means for calculating depends on the number of active voices Is provided.

  Preferably, each voice can calculate one output at a time.

  Preferably, the number of samples used for each active voice by the calculating means decreases as the number of active voices increases.

  Preferably, the number of samples used for each active voice by the means for calculating decreases as the active voice increases so that the maximum computational complexity does not exceed the limit.

  Alternatively, the number of samples used for each active voice by the calculating means decreases non-linearly as the number of active voices increases.

  Preferably, the plurality of samples stored in the memory include musical note samples.

  Preferably, the plurality of samples stored in the memory include musical note samples generated by various musical instruments.

  According to the second aspect of the present invention, there is provided a portable device including the music synthesizer as described above.

  Preferably, the mobile device is a mobile phone.

  In another aspect, the portable device is a pager.

  It should be noted that the term “include / include”, as used herein, should be interpreted as defining the presence of the stated function, integer, step or component, but one or It does not exclude the presence or addition of any other functions, integers, steps, components or groups thereof.

For a better understanding and implementation of the present invention, reference is made to the accompanying drawings.
FIG. 1 shows a music synthesizer according to the invention. As is conventional, the synthesizer includes a controller 2, a plurality of voices 4, a wave table memory 6, a filter table 8, a mixer 10 and an analog / digital conversion module 12.

  In the following, this synthesizer will be described as a synthesizer based on a wave table using the MIDI protocol, but for all wave tables that need to calculate one sample between two stored samples. It should be understood that the present invention can be applied to a synthesizer based thereon.

  Note that the term “sample” as used in the present invention refers to a single audio sample point.

  The total number N of voices 4 in this synthesizer defines the maximum polyphony for this system. As N increases, the polyphony increases, allowing a greater number of sounds to be generated simultaneously. For one MIDI synthesizer corresponding to General Midi System Level 1 (GM-1), the value of N is at least 24. For ease of understanding, an example using only three voices is shown in FIG.

  Controller 2 receives data through input 14. This data includes a stream of MIDI information about a particular set of music or sounds. Each MIDI file contains a list of events that describe the specific steps that the synthesizer must perform to generate the requested sound.

  If a MIDI file is stored in the portable communication device, the file may relate to short music that can be used for ringing, for example.

  The controller 2 processes the MIDI data stream and directs the appropriate part of this data to the relevant voice 4 so that the requested sound can be synthesized. For example, the requested sound may include several different instruments playing at the same time, so each voice 4 processes a portion of one monophony or polyphony instrument at the same time. Often, a MIDI file contains instructions for a particular voice 4 that is used to synthesize the next output.

  Depending on the specific content of the MIDI file, different numbers of voices may be used simultaneously, depending on the specific part of the music being played.

  Each voice 4 is connected to a controller 2, a mixer 10, a wave table memory 6 and a filter table 8.

  The wave table memory 6 contains several sequences of digital samples. Each sequence represents a musical note of a specific instrument, for example. Due to memory limitations, a small number of notes may be stored for each instrument.

  The filter table 8 contains several values for one filter. In the preferred embodiment, this value represents a sinc function (where the sinc function is (sin (x)) / x).

  Although not shown in FIG. 1, both the wave table memory 6 and the filter table 8 have a multiplexer, which allows each table to be accessed more than once per sample period. Defined as the reciprocal of the rate, ie the speed at which the original sound was sampled). Thus, each of voice 1 through voice n can share the same set of resources.

  As before, the voice 4 generates the requested output samples 16 based on the commands received from the controller 2 and the interpolation order of the system.

  Often, the sound produced by a particular voice 4 does not correspond to one frequency of the stored sequence of samples. Therefore, in order to generate sound at the required frequency, the sequence in which the voice 4 is stored must be shifted in frequency.

  For example, if the sequence of stored samples represents a middle C note on a piano, this sequence can be shifted in frequency to obtain a C # note or a D note.

  The requested sound frequency can be expressed as a multiple of the frequency of the stored sequence. This multiple is written as rational number M / L and is known as phase increment.

  Thus, if the requested frequency is twice the stored frequency, the phase increment is equal to 2. If the requested frequency is half the frequency of the stored sequence, the phase increment is equal to the order 1/2. In an example where C # notes are required, the phase increment is the 12th root of 2 (irrational number), which can be approximated by a rational number.

  Often, the requested sample is not stored in memory when the stored sequence of samples is shifted in frequency. That is, the requested sample is between two stored samples.

  Therefore, voice 4 retrieves several samples around the sample requested from wave table memory 6 and the same number of filter coefficients from filter table 8. Each sample retrieved from the wave table memory 6 is then multiplied by the appropriate filter coefficients from the filter table 8 and the products are combined to produce the output of the voice 16.

  If the coefficients of the filter table 8 are selected so that the wave table memory 6 contains the requested sample, the other samples retrieved from the wave table memory 6 are multiplied by the zero filter coefficient. The stored sample is output.

  In an embodiment in which the filter table 8 includes a value representing a sinc function, the period of the sinc function is twice the sample period.

  Each output 16 of the voice 4 is sent to the mixer 10 where all the active voice 4 outputs 16 are combined into a combined output 18 and passed to the DAC module 12.

  The DAC module 12 includes one or more digital to analog converters that convert the combined output 18 of the mixer 10 into a single analog signal 20.

  FIG. 2 illustrates a method performed by the controller 2 of FIG. 1 according to the present invention.

  In step 101, the controller 2 analyzes the MIDI data stream and determines the number of voices 4 to be active during the next sample period. That is, the controller 2 determines how many different voices 4 contribute to the output 16 to the mixer 10.

In step 103, the controller 2 determines the number of samples used by each voice 4 to calculate the next output 16 (referred to as interpolation order _ID ) and commands the voice 4 appropriately.

In step 105, each active voice 4 calculates an output 16 based on instructions received from the controller 2 using a number of stored samples in a calculation equal to the interpolation order _ID . Each active voice 4 also uses several filter coefficients from the filter coefficient table 8 equal to the interpolation order _ID .

  The process is repeated for each output cycle, i.e., the process is repeated once for each sample period.

  In the embodiment of the invention described with reference to FIGS. 3 and 4, the synthesizer has 24 voices 4 and has an interpolation order of up to 11th order.

  FIG. 3 is a table showing a method for determining an interpolation order based on the number of active voices according to the present invention. In particular, for any given number of active voices, this table gives the interpolation order used.

  For example, if the controller 2 determines that only one voice 4 is active during the next sample period, the controller 2 instructs the voice 4 to use the 11th order interpolation order.

  As the number of active voices 4 increases, the interpolation order used to calculate the output 16 decreases in a linear fashion.

  If all 24 voices 4 of the synthesizer are active, the controller 2 decides to use the fourth order interpolation order.

  As another example, if the maximum computational complexity is defined for a synthesizer, such as a synthesizer used in a mobile device, the interpolation order is selected so that the maximum computational complexity does not exceed the limit. May be.

  FIG. 4 is another table showing such a scheme. Again, as the number of active voices 4 increases, the interpolation order decreases. However, this change is not linear. Instead, the interpolation order is calculated so that the maximum computational complexity does not exceed the limit.

  For example, if a synthesizer has 24 voices, an 11th order maximum interpolation order, and processes 0.5 MIPS / order / voice (1 second / order / million instructions per voice), a normal synthesizer Requires up to 132 MIPS. This computational capability far exceeds that available with typical current portable devices such as mobile terminals.

  If the scheme shown in FIG. 4 is used, the computing power does not exceed 50 MIPS. This value is more appropriate for portable devices.

  The actual scheme used will be determined by the amount of computing power available to the synthesizer and the amount of computing power required to perform each degree of interpolation.

  FIG. 5 shows one voice of FIG. 1 in more detail. The voice 4 is indicated by a controller 2, a wave table memory 6 and a filter table 8.

  The processor 22 receives instructions related to the voice 4 from the controller 2. These instructions include MIDI information related to voice 4 and instructions regarding the interpolation order used to calculate the next output 16.

  Controller 2 indicates to each voice 4 the actual interpolation order used to calculate the next output, or alternatively, controller 2 indicates to each voice 4 the number of active voices. Thus, the processor 22 determines an appropriate interpolation order.

  The processor 22 is connected to a phase increment register 24, a counter 26 and a filter coefficient selector 28.

  A filter coefficient selector 28 is connected to the filter table 8 to retrieve the appropriate filter coefficient.

  The filter coefficient selector 28 is also connected to the counter 26.

  In accordance with the present invention, processor 22 instructs counter 26 and filter coefficient selector 28 about the interpolation order used to calculate the next output 16.

  The processor 22 sets the value of the phase increment register 24 to generate the required output. The value of the phase increment register 24 is M / L, where L and M are integers and are determined by the processor 22 based on instructions received from the controller 2.

  The phase increment value is passed to adder 30. The adder 30 is connected to a phase register 32 that records the current phase. The output of the adder 30 includes an integer part and a decimal part.

  Both the integer part and the fractional part of the phase register are fed back to the adder 30.

  The integer part of the output of the phase register is also passed to the second adder 34 where it is added to the output of the counter 26. The integer output of adder 34 is connected to wave table memory 6 to determine the sample to be read.

  Samples retrieved from the wave table memory are passed to the product-sum operation circuit 36.

  In addition to being sent to the adder 30, the fractional part of the output of the phase register 32 is sent to the filter coefficient selector 28.

  The output of the filter coefficient selector 28 is passed to the multiplication and accumulation circuit 36 where it is combined with the sample retrieved from the wave table memory 6.

  The operation of voice 4 will be briefly described.

  If the input of the phase register 32 is a non-integer value, that is, the fractional part is not 0, the required sample is between the two samples shown in the table. Therefore, the necessary samples must be calculated.

  Adder 30 operates once per sample period to add the phase increment from phase increment register 24 to the current phase (provided by phase register 32).

The integer part of the output of the phase register 32 points to the wave table memory address containing the sample stored immediately before the required sample. In order to calculate the required samples, several samples equal to _ID are read from the wave table memory 6.

The counter 26 is incremented by one each time an _ID sample is selected from around the necessary samples. Thus, when _ID is 8, the 4 samples before the required sample are read, along with the 4 samples after the required sample. If _ID is 5, three samples are read before the required sample, along with the two samples after the required sample. As another example, two samples are read before the required sample, and three samples are read after the required sample. These samples are passed to the product-sum operation circuit 36.

  Note that the counter runs once every sample period from its initial value to its final value.

The filter coefficient selector 28 obtains an appropriate filter coefficient from the filter table 8 based on the fractional part of the phase register output and the interpolation order. The filter coefficient selector 28 is controlled by the counter 26 to obtain the _ID coefficient from the filter table 8.

  When the filter coefficient 44 is obtained from the filter cable 8, the input received from the counter 26 is used to pass the filter coefficient to the product-sum operation circuit 36. Here, the sample obtained from the wave table memory 6 is multiplied by an appropriate filter coefficient 44, and the products are added to obtain the output of the voice 16.

  When the decimal part of the phase register 32 changes, the filter coefficient obtained from the filter table 8 changes.

  As the number of active voices 4 changes, the processor informs the counter 26 and filter coefficient selector 28 that the required interpolation order is appropriate.

  FIG. 6 shows a mobile phone having a music synthesizer according to the invention. Although the present invention is described as being incorporated into a mobile phone, it may be applied to any portable device such as a personal digital assistant (PDA), pager, electronic organizer, or any other device where it is desirable to be able to play high quality polyphonic sound. It should be understood that the present invention is applicable.

  As is conventional, the mobile phone 46 includes an antenna 48, a transceiver circuit 50, a CPU 52, a memory 54 and a speaker 56.

  The mobile telephone 46 also includes a MIDI synthesizer 58 according to the present invention. The CPU 52 supplies the MIDI file to the MIDI synthesizer 58. The MIDI file is stored in the memory 54 or downloaded from the network via the antenna 48 and the transceiver circuit 50.

  Thus, a sound synthesizer has been described that reduces the computational requirements of a synthesizer with a high dimensional polyphony and keeps the audible artifacts to a minimum.

1 illustrates a sound synthesizer according to the present invention. Fig. 3 shows a method performed by the controller of Fig. 1 according to the invention. The method for determining the interpolation order based on the number of active voices according to the present invention will be described. Fig. 4 illustrates another scheme for determining the interpolation order based on the number of active voices according to the present invention. One voice of the synthesizer of FIG. 1 is shown in more detail. 1 shows a mobile phone having a music synthesizer according to the invention.

Claims (16)

A memory containing a plurality of stored samples;
Means for calculating a single output signal for each of the plurality of active voices using a plurality of samples selected from the stored samples for each of the active voices;
The number of samples used for each active voice by the means for calculating depends on the number of active voices.   The synthesizer of claim 1, wherein the number of samples used for each active voice by the means for calculating decreases as the number of active voices increases.   3. The number of samples used for each active voice by the means for calculating decreases as the number of active voices increases so that the maximum computational complexity is not exceeded. The synthesizer described.   The synthesizer of claim 1, wherein the number of samples used for each active voice by the means for calculating decreases non-linearly as the number of active voices increases.   The synthesizer according to claim 1, wherein the plurality of samples stored in the memory include musical notes.   6. A synthesizer according to claim 5, wherein the plurality of samples stored in the memory include samples of musical notes generated by various musical instruments.   A synthesizer according to any of claims 1 to 6, wherein the means for calculating the one output signal includes a filter table.   The synthesizer of claim 7, wherein the filter table includes coefficients of a sinc function.   The synthesizer according to any one of claims 1 to 8, wherein the synthesizer is a MIDI music synthesizer.   A portable device including the synthesizer according to claim 1.   The mobile device according to claim 10, wherein the mobile device is a mobile phone.   The portable device according to claim 10, wherein the portable device is a pager. A method of operating a synthesizer having a plurality of samples stored in a memory, comprising:
Determining the number of voices that are active to produce a single sound;
Determining an interpolation order based on the number of active voices, the interpolation order determining the interpolation order defined as the number of samples selected from the plurality of samples stored in memory. Steps,
Calculating an output for each active voice using the number of stored samples determined by the interpolation order;
Including said method.   The method of claim 13, wherein the interpolation order decreases as the number of active voices increases.   The method of claim 13, wherein as the number of active voices increases, the interpolation order decreases so that maximum computational complexity is not exceeded.   The method of claim 13, wherein the interpolation order decreases nonlinearly as the number of active voices increases.

Images