JP2591159B2 - Waveform selection and synthesis device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a waveform selecting / synthesizing apparatus suitable for use in creating a tone signal by synthesizing waveform data read from a plurality of waveform memories.

2. Description of the Related Art An electronic musical instrument has been developed in which waveform data of a natural musical instrument or the like is stored in a waveform memory for a predetermined period, and the waveform data is read to form a musical tone. As a basic configuration, there is a configuration in which one waveform memory is provided for one tone color, and waveform data is read out at a cycle corresponding to the pitch of a key. However, many natural musical instruments have slightly different waveforms in the high, middle, low, and loud ranges, and if a musical tone is formed with the same waveform data, the musical tone becomes unnatural and monotonous. Therefore, an electronic musical instrument provided with a plurality of waveform memories storing different waveforms for each tone range or each volume has been developed.

For example, as shown in FIG. 25, the keyboard 1 is divided into predetermined areas, and waveform memories M0, M1, M2, M3, and M4 storing different waveforms for each division are allocated. Then, a tone is formed based on different waveform data for each section. When a plurality of timbres are set, allocation is performed as shown in FIG. 25 for each timbre. In this case, the number of waveform memories may differ depending on the timbre.

By the way, the electronic musical instrument shown in FIG. 25 has a problem that the waveform is abruptly different at the boundary between the sections, which is unnatural. Therefore, there has been developed a device in which waveform data of adjacent waveform memories are appropriately combined so that the waveform naturally changes with the shift of the sound range.

Further, since the waveform of a piano or the like also differs depending on the string striking intensity, a device provided with a plurality of waveform memories corresponding to the touch intensity has been developed.

[Problems to be Solved by the Invention] By the way, when allocating a waveform memory to a keyboard, there is a case where it is necessary to allocate more waveform memories to a high frequency side (or a low frequency side) and to change a waveform in such a sound range finely. . After setting such characteristics, there is a demand to maintain the same characteristics even when the number of waveform memories changes due to switching of timbres or the like. This is because players often perform while recognizing the characteristics of the keyboard, and if the characteristics of the keyboard change due to changes in tone color or the like, the desired performance effect may not be obtained.

However, the allocation of the waveform memory in the conventional electronic musical instrument is appropriately determined according to the number of the waveform memories. Therefore, there is a drawback that if the tone color or the like is changed, the characteristic of the waveform change on the keyboard fluctuates. Was.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problem, and it is possible to make a characteristic of a waveform change on a keyboard constant irrespective of the number of waveform memories, and to change a waveform based on a change in pitch or touch. It is an object of the present invention to provide a waveform selecting / synthesizing apparatus capable of performing the processing very smoothly.

[Means for Solving the Problems] In order to solve the above problems, according to the first aspect of the present invention, a plurality of tone data memory groups each including a plurality of tone data for forming a tone are stored. Memory means for selecting one of the musical tone data memory groups; a conversion means for normalizing parameter values within a predetermined range according to a predetermined conversion characteristic to convert the parameter values into selected data; and The numerical value corresponding to the number of musical tone data included in the musical tone data memory group selected by the group selecting means is multiplied by the selected data, and based on the integer part of the multiplied value, the selected musical tone data memory group is Tone data selecting means for selecting two tone data; and the two tone data selected by the tone data selecting means. And synthesizing means for synthesizing at a rate corresponding to the decimal part of the multiplied value.

In the invention according to claim 2, each of the plurality of tone data is constituted by a memory block including a plurality of waveform memories each storing waveform data for forming a tone. , Two of the two tone data selected by the tone data selecting means.
The waveform data stored in each of a pair of waveform memories included in each of the memory blocks is combined at a ratio corresponding to the decimal part of the multiplied value.

In the invention according to claim 3, each of the plurality of tone data is constituted by a plurality of waveform memories each storing waveform data for forming a tone.
The synthesizing means synthesizes the waveform data stored in a pair of mutually corresponding waveform memories constituting the two tone data selected by the tone data selecting means at a ratio corresponding to the fractional part of the multiplied value. It is characterized by:

[Operation] First, one of the memory groups is selected by the memory group selecting means. On the other hand, the value of the parameter is normalized in a predetermined range according to a predetermined conversion characteristic by the conversion means, and is converted into selected data. Any two pieces of tone data in the tone data memory group are selected according to the normalized selection data. Then, these two tone data are synthesized by the synthesizing means in accordance with the value of the selected data. Therefore, regardless of the number of tone data in the tone data memory group, selection and synthesis of tone data are performed according to a predetermined conversion characteristic.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

A: First Embodiment (1) Overall Configuration FIG. 1 is a block diagram showing the overall configuration of a first embodiment of the present invention. In the figure, reference numeral 1 denotes a keyboard composed of a plurality of keys, and a key press detection unit 2 detects which key is pressed. The key press detection unit 2 outputs a key-on signal KON indicating that a key has been pressed and a key code KC indicating the pitch of the key. The initial touch detection unit 3 detects an initial touch of a pressed key and outputs it as touch data TD. 5
Is a tone selection operator operated when selecting a tone, and when its output signal is supplied to the tone number generator 6,
The timbre designation signal TC is output from the timbre number generator 6.

Reference numeral 10 denotes a tone signal generator, which has a waveform memory section M therein. Of the waveforms stored in the waveform memory section M, those corresponding to the touch data TD, the tone designation number TC, and the key code KC are appropriately processed. Read. The tone signal generator 10
Creates a tone signal based on the read waveform data,
This is supplied to the sound system 12 via the D / A converter 11. Thus, a predetermined musical sound is generated from the sound system 12.

Reference numeral 19 denotes a normalization curve control unit that creates a normalization curve for writing to the normalization table 15 in the tone signal generation unit 10. The details of the normalization curve control unit will be described later.

(2) Configuration and Individual Operation of Each Circuit Part Next, each of the above-described circuit parts will be described.

Waveform memory unit M The waveform memory unit M is provided for each tone TC = 0 to 0 as shown in FIG.
Multiple memory groups that store waveforms corresponding to N
It consists of MG1 to MGN. And each memory group
MGi (i = 0 to N) includes waveform memory blocks Mo to Mn that store different waveforms for each predetermined range.
Here, the reason why the waveform is different for each sound range is that, even if the musical instruments generally have the same timbre (for example, piano), the waveform slightly changes when the sound range changes. Note that the number of memory blocks is appropriately set for each tone color. Each of the memory blocks Mo to Mn includes a plurality of waveform memories Mij that store different waveforms for each keystroke strength. The reason for such a configuration is that the waveform differs depending on the sound intensity even in the same tone color and the same tone range, and each waveform memory is selected by the touch data TD. Note that j
Is appropriately set for each tone color.

Next, each waveform memory Mij (i = 0 to n, j = 0 to m)
Will be described with reference to FIG. In the figure, the sections demarcated by broken lines are predetermined reading sections, the first and second sections correspond to the attack part of the waveform, and the third section corresponds to the loop part. The attack portion is a waveform portion when the sound rises. The loop portion is read following the attack portion, and thereafter, the loop portion is repeatedly read. That is, after the reading of the reading sections 1 and 2 is completed, the reading section 3
Are repeatedly read. It should be noted that the number of sections of the attack portion varies depending on the timbre, the range, the keying strength, and the like.

Normalization Table 15 The normalization table 15 is a table for allocating any range to the waveform memory.
This will be described with reference to the drawings. It is assumed that n is 4 (the number of memory blocks is 5) taking the memory group MG1 as an example.

First, the waveform memory blocks M0 to M4 are assigned to the keyboard 1 in order from the low band. In this case, waveform memory blocks M0 and M1 are assigned to key area # 0, waveform memory blocks M1 and M2 are assigned to key area # 1, and two memory blocks are similarly assigned to each key area. Assigned. In this case, the range of each of the key ranges # 0 to # 3 is determined as follows. First, the value of the key code KC is converted into any value between “0” and “1” by the normalization table 15 shown in FIG. In this conversion order, a value “memory block number−1” is multiplied, and the integer part of the multiplication result is α,
The decimal part is β. That is, assuming that the converted value of the normalization table 15 is CDA, the integer part [X] of the calculation result X of CDA × (the number of memory blocks −1) = X (1) is α, and the decimal part (X −α) is assumed to be β.

Then, based on α, the memory blocks Mα and M (α +
Two of 1) are selected. Therefore, the range of each key range is determined by how the value of α changes with respect to the value of the key code KC. That is, the range of each key range is determined by the normalization curve in the normalization table 15. In other words, the waveform memory blocks M0 to M4
Is determined at what density from low to high. For example, when the conversion curve is a linear straight line as shown in FIG. 6A, allocation is performed at a uniform density over the entire area (the range of the key areas # 0 to # 3 is uniform). When the normalized curve line is as shown in FIG. 6B, the density on the low frequency side is low and the density on the high frequency side is high. In other words, the range of the low range key range # 0 is wide, and the range of the high range key ranges # 2 and # 3 is narrow.

The calculation of the above equation (1) is performed in the musical tone signal generator 10. Here, FIG. 8 is a block diagram showing the configuration of the tone signal generator 10. In the figure, reference numeral 30 denotes a timbre control unit. When a timbre designation signal TC is supplied, a signal for selecting a corresponding memory group MG0 to MGN is supplied to a memory selection unit 32, and a signal in the selected memory group is provided. The value obtained by subtracting 1 from the number of memory blocks is written into the number register 31. A multiplier 35 multiplies the output data CDA of the normalization table 15 by the output data of the number register 31. This multiplication is the operation of equation (1),
The upper side of the calculation result corresponds to α, and the lower side corresponds to β.

Next, synthesis of the waveform data read from the two waveform memories selected by the equation (1) is performed by the following equation (1−β) Wα + βWα + 1 (2). Here, Wα, Wα + 1 are
These are waveform data read from one waveform memory (corresponding to the touch data TD) in each of the waveform memory blocks Mα and Mα + 1. According to the above equation (2), for example, the effect of the waveform data Wα increases as the pitch decreases for a certain key range, and the effect of the waveform data Wα + 1 increases as the pitch increases. When the pitch is the lowest in the range, β = 0, so that a tone is formed only by the waveform data Wα. For the highest note in the entire key range, the output data CD of the normalization table 15 is used.
Since A is “1”, the result of the above equation (1) is 1 × 4 = 4, and α is 4 and β is 0. Accordingly, the waveform memory block M4 and the waveform memory block M5 are selected. However, since β is 0, the following equation (2) is used.
A tone is formed only by the waveform data W4 in the memory block M4. In this case, the waveform memory block M5 does not exist, but no problem occurs by the combining method of the equation (2). Here, a specific key for forming a musical tone by the waveform data Wα is shown by a broken line in FIG. The calculation of the equation (2) is performed by an interpolation unit 36 (see FIG. 8) described later.

Next, the normalized curve control unit 19 will be described with reference to FIG. Reference numeral 20 shown in FIG. 1 denotes a normalization curve setting operator which is operated when setting a normalization curve in the normalization table 15. A feature point data memory 21 stores four feature points characterizing the normalized curve as shown in FIG. 7 (a). The writing control unit 22 writes this feature point based on the output signal of the normalization curve setting operator 20. Also, feature point data memory 21
Is supplied with a tone color designation signal TC, so that a feature point can be stored for each designated tone color. Note that the same feature point can be stored for different timbres.

Next, reference numeral 23 denotes a normalization curve creation unit which creates a normalization curve by interpolation based on the feature points in the feature point data memory 21. For example, as shown in FIG. 7 (b), a normalized curve is created by linear interpolation. This creation is performed when a predetermined command signal is supplied, and this command signal is output at the start of the performance, for example, by the operation of the player. In this case, interpolation is not limited to linear interpolation and may be performed using a quadratic curve or a cubic curve. The normalization curve created by the normalization curve creation unit 23 is written to the normalization table 15 by the write control unit 24.

The waveform memory selection unit 32 refers to the waveform memory directory 33 based on the touch data TD, the upper-side output α of the multiplier 35, and the signal for selecting the memory group supplied from the timbre control unit 35. The waveform memory Mij is selected, and its head address and the like are supplied to the read control unit 34.

Here, the memory selection process of the waveform memory selection unit 32 with reference to the waveform memory directory 33 will be described. For example, when the timbre control signal TC is "2", the waveform memory group MG2 corresponding to the timbre "2" is selected as shown in FIG. Next, the higher-order output α of the multiplier 35 is used as a memory block designation signal WSLT, and a memory block is selected according to this signal. If the memory block designation signal WSLT is "1", the memory blocks M1 and M2 are selected (see FIG. 9 (c)). These memory blocks
As described above, there are a plurality of waveform memories corresponding to the touch data in M1 and M2 (see FIG. 2). In each memory block M, scale data S is stored.

Here, the scale will be described. Waveform memory Mij
For example, as shown in FIG. 3, waveform data is stored, and the reading of this waveform data is performed at a pitch corresponding to the key code KC. That is, as the pitch indicated by the key code KC increases, the read pitch is increased to shorten the read cycle, thereby increasing the tone generation frequency. Therefore, the waveform data on the high frequency side is coarser than that on the low frequency side, so that the memory capacity can be reduced. Therefore, the scale of the waveform memory is sequentially set smaller as the frequency becomes higher, thereby reducing the number of addresses. For this reason,
In the reading process of each waveform memory, it is necessary to use address data corresponding to the capacity. Therefore,
Scale data is provided for each memory block to create address data corresponding to the memory capacity.

Next, as shown in FIG. 9, one waveform memory is selected from each of the memory blocks M1 and M2 based on the touch data TD. In the figure, the case where the touch data TD is 5 is shown, whereby the address of the waveform memory is selected. In the WAD2-1-5 shown in FIG.
WSLT (= α) indicates the head address of the waveform memory corresponding to “1” and touch data TD to “5”. Also EA
D2-1-5 is the end address of the same memory. The end address is an address at the end of the attack part, for example, the address EA shown in FIG. Ninth
WAD2-2-5 and EAD2-2-5 shown in FIG. 4D also indicate the start address and the end address, respectively.

As described above, the waveform memory selection unit 32
One of the waveform memories is selected, and the corresponding start address, end address, and scale data are output.

Read Control Unit 34 Next, a read control unit 34 shown in FIG. 8 reads the waveform data from the waveform memory Mij, and has a configuration shown in FIG. In the figure, 40 is an address counter,
When the key-on signal KON (see FIG. 1) is supplied, the F number is sequentially accumulated in synchronization with the clock φi2, and the count output is output as the lower side of the address data.
Here, the F number is data indicating a frequency corresponding to the pitch indicated by the key code KC. Reference numerals 41, 42, and 43 denote selectors, and the first address input terminal is supplied with the start address data WAD, the end address data EAD, and the scale data S from the waveform memory selection unit 32. . Each of the selectors 41, 42, 43 outputs a signal wp1, wp2,
When wp3 is supplied, the first bit input terminal is selected. These signals wp1, wp2, wp3 are output when various data output by the waveform memory selection unit 32 are written to the read control unit 32. Reference numeral 45 denotes a start address shift register, which is a start address WA supplied from the selector 41.
D is sequentially shifted in synchronization with the clock φi1. The clock φi1 is a clock having a half cycle of the clock φi2. End address shift register
Similarly to the above, the 46 and the scale data shift register 47 perform a shift operation in synchronization with the clock φi1. In this case, the end address shift register 46
Is the end address data EAD supplied from the selector 42.
And the scale data shift register 47 shifts the scale data S supplied from the selector 43. The output data of the end address shift register 46 and the scale data shift register 47 are supplied to the selector 42,
43 is supplied to the 0th input terminal.

50 is the output data of the shift register 45 and the shift register
The comparator compares the output data with the output data of 46, and outputs a "0" signal when the coincidence of both data is detected. An AND gate 51 supplies an output signal of the comparator 50 to one input terminal and a carry signal of the address counter 40 to the other input terminal. The output signal of the AND gate 51 is supplied to the control terminal of the selector 52. The output signal of the head address shift register 45 is supplied to the 0th bit input terminal of the selector 52, and the output signal of the adder 53 is supplied to the 1st bit input terminal. The adder 53 adds "1" to the output data of the head address shift register 45. The output data of the selector 52 is supplied to the 0th bit input terminal of the selector 41.

Reference numeral 55 denotes a shifter which outputs the output data of the head address shift register 45 as upper address data and the output data of the address counter 40 as lower address data to the waveform memory Mij, and outputs the scale data supplied from the scale data shift register 47. The address data is shifted to the lower side by the number of bits corresponding to the value of S.

 The operation of the read control unit 34 is as follows.

First, the read address data WAD, end address data EAD, and scale data S of the waveform memory corresponding to WSLT (= α) are supplied to the head address shift register 45 and the end address shift register via the selectors 41, 42, and 43, respectively.
46 and the first stage of the scale data shift register 47. Next, the shift registers 45, 46,
47 performs a shift operation on the supplied data, and data on the waveform memory of WSLT (= α + 1) is stored in the first stage of each of the shift registers 45, 46, and 47 in the same manner as described above. As a result, the address data output by the shifter 55 becomes the head address of the waveform memory of WSLT (= α), and the shift amount in the shifter 55 corresponds to the number of addresses of the waveform memory. When the next clock φi1 is output, each shift register 45, 46,
Since the shift operation is performed by 47, the address data and the shift amount output from the shifter 55 correspond to the head address and the number of addresses of the waveform memory of WSLT (= α + 1). Then, the output data of the start address shift register 45 is sequentially passed through the selectors 52 and 41 to the same shift register.
The output data of the scale data shift register 47 is supplied to the first stage of the shift register 47 via the selector 43. Therefore, each time the clock φi1 is output thereafter, the address data output from the shifter 55 alternates between WSLT (= α) and WSLT (= α +
This corresponds to 1).

Also, every time the clock φi1 is output twice, the clock φi1 is output.
i2 is output once, whereby the address counter 40 accumulates the F number. Therefore, the address counter 40
Is updated once each time the clock φi1 is output twice.

Next, when the address counter 40 overflows,
The carry signal CY is output,
51 outputs a “1” signal. As a result, the first bit input terminal of the selector 52 is selected, and the output data of the adder 53 is supplied to the first stage of the head address shift register 45 via the selectors 52 and 41 sequentially. As a result, the upper side of the address data is incremented by one. This increment is performed each time the address counter 40 overflows and the carry signal CY is output. Thereby, the waveform of each section of the attack section (see FIG. 3) is sequentially read. When the incremented upper address data matches the end address EAD output from the end address shift register 46, the comparator 50
Outputs a “0” signal, which causes the AND gate 51
Is closed. When the AND gate 51 is closed, the AND gate 51 does not output a "1" signal even if the carry signal CY is output, and as a result, the selector 52 always selects the 0th bit input terminal. As a result, the addition process by the adder 53 is canceled, and the increment of the upper address data is stopped. Therefore, thereafter, only the lower address changes, and the loop portion of the waveform data (see FIG. 3) is repeatedly read. The comparison operation in the comparator 50 is performed on the end address EAD of the waveform memory corresponding to WSLT (= α) and WSLT (= α + 1) alternately supplied from the end address shift register 46. The transition to the loop section is also controlled for each waveform memory.
As described above, from the read control unit 34, the waveform memory selection unit
Address data for the two waveform memories selected at 32 are output alternately. When the address data is supplied to the waveform memory unit M, the waveform data in the corresponding waveform memory is sequentially read.

Interpolation unit 36 Next, the interpolation unit 36 shown in FIG. 8 is a circuit that performs an interpolation operation on the waveform data supplied from the waveform memory unit M, and has a configuration as shown in FIG. Complement selector 60 in the figure as the value β is a lower output of the multiplier 35, or a circuit for outputting taking its complement, clock phi _T is adapted to output a complement when "0" ing. This complement is approximately equal to the value of (1-β). Note that if "1" is added to the least significant bit of the complement value, the value completely matches (1-β).

A multiplier 61 multiplies the waveform data Wα and Wα + 1 alternately read from the waveform memory M by the output data of the complement selector 60. An adder 62 adds the output data of the multiplier 61 and the output data of the gate 63. A shift register 64 stores output data of the adder 62 in synchronization with the clock φ. The output data is supplied to the gate 63. Numeral 65 is a latch for latching the output data of the adder 62, and fetches the data when the output signal of the AND gate 66 is "1". This And Gate 66
It is adapted to ANDing the clock phi and phi _T. Here, the clock φ is a clock having the same frequency as the above-described clock φi1, and is synchronized with the clock φi1 in a predetermined manner (see FIG. 12 (a)). Also, the clock φ
_T is a clock having a period twice as long as the clock φ,
The relationship with the clock φ is shown in FIGS.

Next, the operation of the interpolation unit 36 will be described. First,
At time t ₁ shown in FIG. 12, when the waveform data Wa from the waveform memory unit M is supplied, the output data of the interpolation selection unit 60 is multiplied by the waveform data Wa. Output data of the complement selector 60 at this time is because the clock phi _T is "0" signal (1-β). As a result, the output data of the multiplier 61 becomes Wα (1−β), and this data is
The input ends of the shift register 64 and the latch 65 are reached via 62. Since the output signal at time t ₁ the AND gate 66 and the gate 63 are both "0" (Figure 12 (c) and (f)), the addition is not performed in the adder 62, The latch 65 Do not import data. Meanwhile, time
In t _1, since the clock φ rises, the operation result Wa (1-beta) are incorporated in the shift register 64 (Figure 12 (e) refer). Next, when the clock phi _T rises at time t _2, the value gate 63 is stored in the shift register 64 in the open state (1-β) Wα is output, and the output data of the complement selector 60 is beta Becomes next,
When the clock φ rises at time t _3, the waveform data Wa + 1 in synchronization with the waveform memory unit M to the clock φ is outputted. As a result, in the multiplier 61, β (Wα +
1) is performed, and further in the adder 62,
The addition of (1−β) Wα + βWα + 1 is performed. Further, at time t _3, with the rise of the clock phi, since the output signal of the AND gate 66 rises, the latch 65 captures the operation result of the adder 62. Then, the data fetched into the latch 65 is the calculation result of the above-described equation (2). Thereafter, similarly, every time the output signal of the AND gate 66 rises, the data corresponding to the calculation result of the equation (2) is taken into the latch 65. Further, the read addresses of the waveform data Wα and Wα + 1 in the obtained calculation result are sequentially updated by the processing of the read control unit 34.

As described above, the synthesis of the two selected waveform data is performed. The waveform data after this synthesis is
At 37, the data is multiplied by the output data of the envelope waveform generator 38, whereby the musical tone waveform is enveloped. The envelope waveform generator 38 creates envelope waveform data based on the timbre data TC, the key code KC, and the key-on signal KON, and supplies this to the multiplier 37.

(3) Overall Configuration First, the performer operates the timbre designation operator 5 to designate a timbre, and operates the normalized curve setting operator 20 to input feature point data. Then, at the start of the performance, the normalization curve creation unit 23 creates a normalization curve based on the above-described feature point data, and the normalization curve is written into the normalization table 15 by the writing control unit 24.
Next, when the player starts playing using the keyboard 1, a key code KC, a key-on signal KON, and touch data TD corresponding to the pressed key are generated by the key pressed detecting unit 2 and the initial touch detecting unit 3. And the key code KC
Is converted into data CDA by the normalization table 15, and the value of the data CAD in the number register 31 is multiplied by the multiplier 35 to calculate α and β. Two waveform memories are selected based on the α and the touch data TD. Next, the read control unit 34 for the two selected waveform memories
Creates address data sequentially and performs read processing alternately for each waveform memory. Then, the waveform data read from each waveform memory is synthesized in the interpolation unit 36 according to the equation (2). The combined waveform data is provided with an envelope by an envelope waveform generator 38, and supplied to the sound shifter 12 via the D / A converter 11 shown in FIG. Then, a musical tone is generated from the sound shift item 12 by a composite waveform of the two waveform memories. Thereafter, similar processing is performed for each pressed key, and musical tones are sequentially generated. In this case, the selected two waveform memories and the ratio of the waveform synthesis change depending on the pitch.
As described above, it is determined by the normalization curve.

(4) Modification In the above embodiment, the normalization curve is created by the normalization curve control unit 19. However, some normalization curves are preset in advance and selected before the performance. It may be configured as follows.

B: Second Embodiment Next, a second embodiment of the present invention will be described. The second embodiment differs from the first embodiment in that two waveform memories are selected using the keystroke intensity (touch data) as a parameter.

(1) Overall Configuration The configuration of the second embodiment is almost the same as the configuration of the first embodiment described above, but differs in the following points.

Instead of the tone signal generator 10, a tone signal generator 80 having the configuration shown in FIG. 14 is provided.

A touch synthesis circuit 70 shown in FIG. 13 is provided between the initial touch detection unit 3 and the tone signal generation unit 80.

The waveform data stored in the waveform memory consists of only one cycle of the loop section (see FIG. 3). In the case of the same tone color and the same keystroke strength (touch), the size (scale) of the waveform memory is the same.

(2) Configuration of Each Unit and Operation of Each Unit Next, the configuration and operation of each unit of this embodiment will be described.

Touch Combination Circuit 13 Reference numeral 71 shown in FIG. 13 denotes an aftertouch detection unit that detects an aftertouch of a key and outputs it as touch data ATD. Reference numerals 72 and 73 denote sensitivity tables in which a plurality of sensitivity curves corresponding to tone colors are stored in advance. These sensitivity curves are selected by the operation of the table selection unit 74 based on the tone color designation signal TC. Sensitivity table 7
2 and 73 convert the touch data TD and ATD into touch data TD 'and ATD' according to the selected sensitivity curve. The reason why the sensitivity tables 72 and 73 are provided is that the touch sensitivity characteristics differ depending on the type of musical instrument (piano, organ, etc.). These touch data T
D ′ and ATD ′ are multiplied by the output signal of the crossfade control unit 75 in multipliers 77 and 76, respectively. Then, the output signals of the multipliers 77 and 76 are added by the adder 78 and output as the combined touch data CTD.

In the above configuration, for example, as shown in FIG. 15, the crossfade control unit 75 outputs a signal that attenuates linearly with time to the multiplier 77, and outputs a signal that linearly attenuates with time to the multiplier 76. It outputs a signal that gradually increases. As a result, the composite touch data CTD is initially strongly affected by the initial touch, and thereafter, the influence of the after touch is increased over time.

Music signal generator 80 The normalization table 81 shown in FIG. 14 is a table in which a plurality of normalization curves corresponding to timbres are stored. When the timbre designation signal TC is supplied, the waveform selection control unit 82 writes a value obtained by subtracting 1 from the number of waveform memories (see FIG. 17) corresponding to the timbre into the waveform number register 83, and outputs the key-on signal KON. It is supplied to the address counter 84. Further, the waveform selection control unit 82 outputs the head address of the waveform memory, the scale information S, and the normalized curve selection data according to the address information in the internal memory.

Here, the output processing of the various data will be described. FIG. 16 is a conceptual diagram showing address information in an internal memory. Assuming that the timbre “2” (TC = 2) is selected and the gamut is determined to be 1 based on the value of the key code KC, as shown in FIG. Selected. The contents of the address information include waveform head address data TWAD, normalized curve selection data CCAD, waveform memory number data NW, and scale information S, as shown in FIG. In the figure, as for the reference numeral 2-1 attached to each data, "2" indicates a timbre number, and "1" indicates a gamut number.

FIG. 17 is a memory map of the waveform memory unit M in this embodiment. As shown in the figure, addresses are sequentially increased for each area corresponding to tone numbers 0 to N,
Further, each tone color area has a storage area corresponding to the gamut 0 to M. In this storage area, the address is sequentially increased as the gamut number increases. Further, the area of each tone range is composed of a number of waveform memories corresponding to the value of the synthesized touch data CTD, and each waveform memory stores waveform data for one cycle as described above. In the figure, reference numeral 2-1-0 indicates waveform data when the timbre number is "2", the gamut number is "1", and the synthesized touch data is "0". The same reference numerals also apply to the above. The head address data TWAD is the head number of the area of the gamut number “2”, that is, (2-1-0)
3 shows the start address of the waveform memory in which the waveform data of the waveform data is stored. This start address data TWAD is
This is supplied to an adder 89 shown in the figure.

Next, the normalization curve selection data CCAD is data supplied to the normalization table 81. In the normalization table 81, a normalization curve corresponding to this data is selected. The scale data S is supplied to the shifter 88 as a shift signal.

Next, 85 is an adder that multiplies the output data CDA of the normalization table 81 by the numerical value in the waveform number register 83,
The result of the multiplication corresponds to equation (1) described above. Therefore, the upper side of the multiplier 85 is α, and the lower side is β. Reference numeral 86 denotes an adder for adding “1” to α, and reference numeral 87 denotes a selector to which α is supplied to the 0th bit input terminal and (α + 1) which is the output signal of the adder 86 is supplied to the first bit input terminal. The selector 87 selects the 0th bit input terminal when the clock φi2 is “0”, and selects the first bit input terminal when the clock φi2 is “1”. The output data of the selector 87 is input to the upper side of the shifter 88, and the count output of the address counter 84 is input to the lower side of the shifter 88. The output data of shifter 88 is added to adder 89
Is added to the top address data TWAD in step (1), and the addition result is supplied to the waveform memory unit M as address data. The address counter 84 is the address counter described above.
This counter performs the same operation as 40 (see FIG. 10).
Here, the address data output from the adder 89 will be described.

As an example, the start address TWAD is the waveform data (2-
1-0) is assumed to be the head address of the waveform memory in which the shifter 88 is shifted for simplicity. First, since α is data of the upper bits of the shifter 88, the upper bits of the address data are determined by the value of α. In the case of this embodiment, the waveform data (2-1-0), (2-1-1), (2-1) each time the value of α increases to “0”, “1”, “2”. -2) ...
The start address of the waveform memory that stores
See figure). Since the count output of the address counter 84 is on the lower side of the address data, the waveform data is read from the waveform memory specified by the value of α in accordance with the increment of the count output of the address counter 84. By the operation of the selector 87, α and (α + 1) are alternately placed on the upper side of the address data every half cycle of the clock φi1, so that the waveform data in the adjacent waveform memory is alternately read. That is, similarly to the first embodiment, the waveform data Wα and Wα + 1 are read alternately. In the above example, reading is performed on the waveform memory in which the waveform data (2-1-1) and (2-1-2) are stored. Address, then
The address 0 of the waveform memory of (2-1-2) is read out, and thereafter, the first address of the waveform memory of (2-1-1), and then the first address of the waveform memory of (2-1-2). The reading process is performed in this order.

As described above, since the waveform memory is uniquely selected according to the value of α, the density of the waveform memory assigned to the change in the keying strength (initial touch and after touch) is determined by the normalization in the normalization table 81. Determined by the curve. Therefore, when it is desired to make various waveform changes in an area where the keying is weak (the performance of a weak sound such as piano, pianissimo, etc.), the density of the waveform memory allocated to this area is increased, for example, as shown by the solid line in FIG. Such a normalization curve is set. Conversely, when it is desired to change the timbre variously in an area where the key is strong (playing a strong sound such as forte, fortissimo, etc.), a normalized curve as shown by a dotted line in FIG. 18 is set.

Next, the waveform data Wα read from the waveform memory unit M
And Wα + 1 are supplied to an interpolation unit 36 shown in FIG. The interpolation unit 36 synthesizes the waveform data by performing the calculation of the equation (2) as described above. Then, the waveform data synthesized by the interpolation unit 36 is supplied to the D / A converter 11 after being multiplied by the envelope signal output from the envelope signal generation unit 90 in the multiplier 37. The envelope signal generator 90 is substantially the same as the envelope signal generator 38 shown in FIG. 8, but generates an envelope signal based on the synthesized touch data CTD, the chion signal KON, and the timbre designation signal TC.

(3) Overall Operation When the player presses a key, the initial touch and the after touch are detected, and these touch data are synthesized by the touch synthesizing circuit 70 to become synthetic touch data CTD. The synthesized touch data CTD is converted into data CDA by the normalization table 81, and is further multiplied by the numerical value in the number-of-waveforms register 83 in the arithmetic unit 85 to become α and β. Then, the selection of the waveform memory and the ratio of the waveform synthesis are controlled based on α and β. In this case, the combined touch data CTD changes over time according to the characteristics determined by the values of the initial touch and the after touch by the processing of the crossfade control unit 75 (see FIG. 13).
As a result, the data CDA changes, α and β change, and the selected waveform memory and the rate of waveform data synthesis change with time. That is, waveform data corresponding to the intensity of the initial touch and the after touch is smoothly selected with the passage of time. At this time, since the normalization curve itself does not change, the waveform data can be changed according to a preset characteristic.

(4) Modification In the above embodiment, a plurality of normalization curves are stored in the normalization table 81 in advance.
As shown in the figure, a normalization curve control unit 19 may be provided so that the player can appropriately set the normalization curve.
Further, as shown in the figure, a timekeeping timer 95 is provided in place of the touch synthesizing circuit 70, a signal that changes over time after the key-on signal KON is output is created, and this signal is replaced with the synthetic touch data CTD. May be used. Clock timer 9
The output characteristic of No. 5 can be an arbitrary one such as a characteristic that changes drastically in an attack part and a characteristic that changes largely in a sustain part.

As shown in FIG. 19, when the timer 95 is used, it is not necessary to detect the initial touch and the after touch, so that an advantage that the configuration is simplified can be obtained.

C: Third Embodiment Next, a third embodiment of the present invention will be described with reference to FIG. 20 and FIG. In these figures, parts corresponding to the respective parts in FIG. 14 are denoted by the same reference numerals, and description thereof will be omitted.

(1) Overall Configuration The overall configuration of this embodiment is almost the same as the configuration of the second embodiment, except that a musical tone signal generator 100 shown in FIG. . The waveform data stored in the waveform memory is a waveform data having an attack part and a loop part as shown in FIG.

(2) Configuration and Individual Operation of Each Unit 101 shown in FIG. 20 is a memory configuration information generation unit,
Various signals for selecting a waveform memory are created based on the tone specifying signal TC and the KICODE KC. Next, the creation process will be described.

FIG. 21 is a conceptual diagram showing a configuration of data stored in memory configuration information generating section 101. Now, assuming that the tone designation signal TC is “1”, as shown in FIG.
Is specified, and the range is specified based on the coded code KC. Assuming that this designation is the range "1" as shown in the figure, the address information stored in this portion is read. This address information is composed of start address data ATAD, end time data EDT, change curve selection data CCAD, waveform memory number data NW, and scale data S. End time data ED above
T is data for indicating the end of the attack unit, and indicates the number of cycles of the attack unit. The other data is the same data as in the second embodiment. The address information is stored in the same manner as described above for other tone ranges and other timbres, and the address information corresponding to the timbre control signal TC and the key code KC is appropriately read.

Next, the waveform selection control unit 102 shown in FIG. 20 creates address data to the waveform memory unit M, and has a time division table reading unit 103 and a waveform address table 104. The waveform address data table 104 is a table for converting the table address TAD into a physical address on the upper side of the waveform memory unit M. Time division table reading unit
A circuit 103 creates a table address TAD for alternately reading the waveform data Wα and Wα + 1. 108
Is an address counter for creating the lower address data of the waveform memory section M, and the count output thereof is
The signal is supplied to the waveform memory unit M via the line 9. The shifter 109 and the address counter 108 have the same functions as the shifter and the counter in each of the embodiments described above.
The address counter 108 supplies a carry signal to one input terminal of the AND gate 107 every time an overflow occurs, and the output signal of the AND gate 107 is
The clock signal is supplied to the time counter 106. The count output of the time counter 106 is supplied to the time division table reading unit 103 and the comparator 105 as time data TIME. The time-division table reading unit 103 knows the end of the attack unit from the time data TIME. Comparator 105
Outputs a “0” signal to close the AND gate 107 when the time data TIME matches the end time data ETD, and outputs a “1” signal to close the AND gate 107 otherwise. State. The time counter 106 and the address counter 108 are supplied with a reset signal from the waveform selection control unit 102. Next, a process of creating the table address TAD in the time division table reading unit 103 will be described.

FIG. 22 is a map based on upper addresses of the waveform memory unit in this embodiment. However, the map shown is
This is a map of an area in which waveform data corresponding to a combination of the tone color designation signal TC and the tone range (in this example, the tone color is “1” and the tone range is “1”). It should be noted that the other combinations have maps similar to those shown in the drawings.

As shown, in this embodiment, a series of waveform data is sequentially divided into areas A0, A1,..., And each area A0, A1,. …
…. In addition, each area A0, A1
... Are divided into areas corresponding to the time data TIME. In the area A0, the section of the time data TIME0 to TIME3 corresponds to the attack section, and the waveform of the attack section is continuously stored in each section in one cycle. Time data TIME
The waveform data of the loop section is stored in the section 4.

For area A1, time data TIME0 to TIME3
Corresponds to the attack part, and the section of the time data TIME4 corresponds to the loop part. Thus, each area A0,
The length of the attack portion of A1 ... is common. However, when the timbre or the tone range is different, the length of the attack part may be different.

Next, the time-division table reading unit 103 creates a table address TAD based on the following equation.

TAD = ATAD + α × (ETD + 1) + TIME (3) Here, since the end time data ETD is the number of periods of the attack part, and ATAD is the head address of the entire selected waveform data, the ATAD is α × (ETD + 1). ) Is the start address of the area Aα (α = 0, 1,...). Then, the value obtained by adding the time data TIME to the head address, that is, the calculation result of Expression (3) is an address corresponding to the value of the time data TIME in the area Aα. As shown in FIG. 20, the time division table reading unit stores α and (α +
1) and (3) are alternately input.
Expressions are calculated. For example, when α = 0, the address of TIME = 0 in area A0 shown in FIG. 22 is first obtained, and then the address of TIME = 0 in area A1 is obtained. Next, TIME = 1 in area A0, TIME = 1 in area A1
Are sequentially obtained, and thereafter the table addresses are alternately obtained in the same manner.
They are going to look for TAD. The cycle of this alternate calculation is set to a half cycle of the waveform data read cycle.

Next, the process of creating the above-described table address TAD will be described in more detail.

First, the time division table reading unit 103 shown in FIG. 20 supplies a reset signal to the time counter 106 and the address counter 108 to reset them. Then, the calculation of the expression (3) is performed using the head address data ATAD, the end time data ETD, and the α supplied from the multiplier 85.
As a result, the head address of the area Aα is supplied to the waveform address table 104 as table address data TAD. Thus, the waveform address table 104 creates an upper address corresponding to the table address data TAD,
It is supplied to the waveform memory unit M. The upper address in this case is the number of bits and the bit position according to the scale of the corresponding waveform memory.

On the other hand, the address counter 108 sequentially counts the F number, and supplies the count output to the lower address waveform memory unit M via the shifter 109. At this time, the shifter
109 performs a shift operation according to the scale data S. As a result, both the upper and lower address data have the number of bits corresponding to the size of the waveform memory. This allows
The upper and lower address data are determined, and the data stored at the head address of the section TIME = 0 of the area Aα is output to the interpolation unit 36 as the waveform data Wα.

Next, the time-division table reading unit 103 performs the operation of Expression (3) using (α + 1) supplied from the adder 86, and sets the head address of the area Aα + 1 to the table address data.
Output to the waveform address table 104 as TAD. As a result, the waveform address table 104 creates higher-order address data corresponding to the area Aα + 1 and supplies it to the waveform memory unit M. At this time, since the address counter 108 has not yet performed the next count, the lower address data has not changed. Therefore, the section of area Aα + 1
The data stored at the start address of TIME = 0 is output to the interpolation unit 36 as the waveform data Wα + 1.

Next, the address counter 108 counts the F number, and the lower address data is updated. Then, under the new lower address data, the same upper address data generation processing as described above is performed, and the waveform data Wα and W
α + 1 are sequentially output. In this case, there is no change in the value of the table address data TAD created by the operation of the equation (3), so that the value of the upper address data is the same value as described above.

Thereafter, the lower address data is sequentially updated each time the address counter 108 is incremented. And
When the address counter 108 overflows, the carry signal is supplied to the time counter 106 via the AND gate 107, and the count value of the time counter 106, that is, the value of the time data TIME changes from "0" to "1". As described above, when the value of the time data TIME is incremented by 1, the table address data which is the result of the operation of the equation (3)
The value of TAD is also 1 for both areas Aα and Aα + 1.
To increase. As a result, the upper address data output from the waveform address table 104 is also incremented by one. Then, the lower address data is updated based on the count output of the address counter 108 with respect to the higher address thus incremented. Thereby, the areas Aα and Aα + 1
The waveform data in the section of the time data TIME = 1 (waveform data in the second section of the attack section) is alternately read and supplied to the interpolation section 36 as waveform data Wα and Wα + 1. Thereafter, each time the address counter 108 overflows,
The time data TIME is incremented, whereby the waveform data of each cycle constituting the attack part (see FIG. 3)
Are sequentially read.

When the time data TIME becomes equal to the end time data ETD, the comparator 105 shown in FIG. 20 detects these coincidences and outputs a "0" signal. As a result, even if the AND gate 107 is closed and the address counter 108 overflows thereafter, the time counter 106 does not count up. Therefore, the value of the table address TAD does not change thereafter.

For example, in the example shown in FIG. 22, since the end time data ETD is "4", the table address data TAD is fixed when the time data TIME is equal to the end time data ETD, that is, when TIME = 4. . The fixed address indicates the upper address of the loop section. Therefore, once the waveform of the loop portion is read, the upper address data is fixed, and thereafter, the waveform data of the loop portion is repeatedly read according to the count output of address counter 108.

Next, the waveform data Wα and Wα + 1 read out as described above are synthesized by the interpolation unit 36, and further,
An envelope is added by the multiplier 37. This operation is the same as in the above-described second embodiment.

(3) Overall Operation The overall operation of this embodiment is substantially the same as the operation of the second embodiment described above, and a waveform memory corresponding to the keystroke strength is selected according to a normalization curve. However, when a match between the end time data ETD and the time data TIME is detected, it is determined that the operation has entered the loop portion, and thereafter, the waveform data of the loop portion is repeatedly read. Therefore,
Even when waveform data having different loop lengths are mixed, a tone signal can be generated without any problem.

(4) Modification The map of the waveform memory is not limited to the map shown in FIG. 22, but may be as shown in FIG. FIG. 23 is a map based on the upper address of the waveform memory unit, and is a map of an area in which waveform data corresponding to any combination of the timbre designation signal TC and the gamut is stored. It should be noted that the other combinations have maps similar to those shown in the drawings. Also,
The example shown in FIG. 23 is an example when the number of waveform memories is five.

As shown, in this embodiment, an area A in which a first section, a second section,.
It is divided into 0, A1 ... sequentially, and each area A0, A1 ...
Correspond to TIME = 0, TIME = 1,... Respectively. Each area A0, A1,... Is divided into areas corresponding to α = 0 to 4. Now, if the number of the area from 0 to M, the area A0~A _M-1 is the attack portion, the area A _M corresponds to the loop portion.

In the case of a map as shown in FIG. 23, the calculation of the table address TAD is performed by the following equation.

TAD = ATAD + TIME × NW + α (4) Here, NW is the number of waveform data, and ATAD is the start address of the entire selected waveform data.
The value obtained by adding ME × NW becomes the start address of the area A _TIME (TIME = 0, 1,...). Then, the value obtained by adding α to the head address, that is, the calculation result of equation (4) is
_The address corresponds to the value of α in _TIME . Then, the time-division table reading unit 103 performs the operation of the expression (4) on α and (α + 1) alternately. For example, when TIME = 0 and α = 0, first, in the area A0 shown in FIG.
Then, the address of α = 1 in the area A0 is obtained. Then, after the time data TIME has been incremented, the respective addresses of α = 0 and α = 1 in the area A1 are sequentially obtained, and thereafter, in the same manner as above, the time data is updated. Find the address of α = 1. The cycle of this calculation is set to half the cycle of reading the waveform data as in the embodiment. Also in this case, when TIME = ETD, only the waveform of the loop portion is repeatedly read.

Further, in the above-described embodiment, a plurality of normalization tables are written in advance in the normalization table, but the normalization curve control unit 19 may be provided so as to be rewritable.

D: Modifications of Embodiments Instead of the normalization curves used in the above-described embodiments, those shown in FIGS. 24A and 24B can be used.

24A and 24B, the horizontal axis indicates parameters such as the key code KC and the touch data TD and CTD, and the vertical axis indicates the normalized data CDM.

Then, the normalized curve shown in FIG.
The value of CDM does not take values near the upper and lower limits. Therefore, the waveform memory assigned to these values is not used. That is, an unused area can be set depending on how to set the normalization curve. Also, in the case of the normalized curve shown in the figure, the value of the data CDM may be the same even with different parameter values (the straight line l).
1). This is effective when it is desired to use the same waveform data for a predetermined parameter value even if the pitch and the hitting strength are different.

Next, the normalized curve shown in FIG. 24 (b) is discontinuous at the parameter value a. As a result, around the value a, the waveform changes significantly,
The tone of the musical tone changes greatly. Such characteristics are set when it is desired to greatly change a musical tone in a certain part of a parameter.

Further, in each of the above-described embodiments, the waveform data is stored in the waveform memory for use.
Of course, the present invention can also be applied to a case where tone parameters such as FM and parameters of a high frequency synthesis filter are stored in a memory.

[Effects of the Invention] As described above, according to the present invention, parameter values are normalized within a predetermined range according to predetermined conversion characteristics,
The conversion means for converting the selected data, the numerical value corresponding to the number of music data included in the music data memory group selected by the memory group selection means multiplied by the selected data, based on the integer part of the multiplied value, Tone data selecting means for selecting two tone data from the selected tone data memory group; and synthesizing means for synthesizing the two tone data selected by the tone data selecting means at a ratio corresponding to the fractional part of the multiplication value. Is provided, the waveform change characteristics can be made constant regardless of the number of musical sound data. In addition, it is possible to arbitrarily set the density of musical tone data allocation in the parameter change range.

Therefore, when applied to an electronic musical instrument, waveform selection and interpolation processing are performed with one parameter, so that a variety of timbre changes can be realized with a simple configuration, and even if the number of tone data differs depending on the timbre, a preset conversion is performed. A constant waveform change characteristic according to the characteristic can be obtained. Further, it is possible to extremely smoothly change the waveform based on a change in pitch or touch.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a first embodiment of the present invention, FIG. 2 is a conceptual diagram showing a configuration of a memory unit M in the embodiment, FIG. 3 is a diagram showing an example of waveform data, 4 and 5 are conceptual diagrams showing a waveform memory block allocation process in the embodiment, FIG. 6 is a characteristic diagram showing an example of normalized data, and FIG. 7 is a graph showing a process of creating normalized data. FIG. 8 shows a tone signal generator in the embodiment.
10 is a block diagram showing the configuration of FIG. 9, and FIG. 9 is a waveform memory selection unit.
FIG. 10 is a conceptual diagram showing a selection process in 32, FIG. 10 is a block diagram showing a configuration of a reading control unit 34 in the embodiment, FIG. 11 is a block diagram showing a configuration of a complementing unit 36 in the embodiment,
FIG. 12 is a timing chart for explaining the operation of the complementing unit 36, FIG. 13 is a block diagram showing the configuration of the touch synthesizing circuit 70 in the second embodiment of the present invention, and FIG. 15 is a block diagram showing a configuration of the generating unit,
FIG. 16 is a diagram showing control characteristics of the cross-fade control unit 75 shown in FIG. 13, FIG. 16 is a diagram showing stored data in the waveform selection control unit 82 in the embodiment, and FIG. FIG. 18 is a conceptual diagram showing a selection process, FIG. 18 is a characteristic diagram showing an example of normalized data in the embodiment, FIG. 19 is a block diagram showing a configuration of a modification of the embodiment, and FIG. FIG. 21 is a block diagram showing the configuration of the third embodiment. FIG. 21 is a conceptual diagram showing the processing of the memory configuration information generation unit 101 in the third embodiment.
FIG. 2 is a map of a waveform memory unit in the embodiment, FIG. 23 is a map of a waveform memory unit in a modification of the embodiment, FIG.
FIG. 24 is a characteristic diagram showing normalized data used in a modification of each embodiment, and FIG. 25 is an explanatory diagram showing allocation of a waveform memory in a conventional device. 5 ... tone selection operator (memory group selection means), 6
…… Tone number generator (memory group selection means), 15,8
1 ... normalization table (conversion table), 31 ... number register (waveform memory selection means), 32 ... waveform memory selection section (waveform memory selection means), 33 ... waveform memory directory (waveform memory selection means), 35,85 ... Multiplier (waveform memory selection means), 36 ... Interpolation unit (synthesis means), 82 ...
Waveform selection control unit (waveform memory selection means), 83 ... number of waveform number register, 101 ... memory configuration information generation unit, 103 ... time division table reading unit (waveform memory selection means), MG0 to MGN
…… Waveform memory group, KC …… Key code (parameter), TD… Touch data (parameter), CTD …… Synthetic touch data (parameter).

Claims (3)

(57) [Claims] A storage means for storing a plurality of tone data memory groups each including a plurality of tone data for forming a tone; a memory group selecting means for selecting one of the tone data memory groups; A conversion unit for normalizing the value within a predetermined range according to a predetermined conversion characteristic and converting the value into selected data; and a numerical value corresponding to the number of musical sound data included in the musical sound data memory group selected by the memory group selecting unit; Music data selecting means for multiplying the selected data by the selected data, and selecting two music data from the selected music data memory group based on the integer part of the multiplied value; Synthesizing means for synthesizing two musical tone data at a ratio corresponding to the decimal part of the multiplied value. Waveform selection synthesizer to. Each of said plurality of tone data is constituted by a memory block including a plurality of waveform memories each storing waveform data for forming a tone; said synthesizing means being selected by said tone data selecting means; And synthesizing the waveform data respectively stored in a pair of waveform memories corresponding to each other included in the two memory blocks constituting the two generated tone data at a ratio corresponding to the decimal part of the multiplied value. The waveform selection / synthesis device according to claim 1. 3. Each of the plurality of musical tone data is constituted by a plurality of waveform memories each storing waveform data for forming a musical tone, and the synthesizing means includes a plurality of waveform memories selected by the musical tone data selecting means. The waveform selecting / synthesizing apparatus according to claim 1, wherein the waveform data stored in a pair of waveform memories corresponding to each other and composed of one musical tone data are synthesized at a ratio corresponding to a decimal part of the multiplied value.