JPS6310836B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to an envelope generator for an electronic musical instrument, and more particularly to an envelope generator for an electronic musical instrument that can form an envelope waveform for various amplitude modulation effects and control its volume level with a simple configuration. Traditionally, amplitude modulation effects have been widely used in electronic musical instruments, such as the repeat effect that repeatedly generates a decaying sound in marimba, banjo, and mandolin, and the cresciendo (gradually becoming stronger) in string instrument ensembles. There are various effects, such as attack effects that particularly emphasize the rising part of a sound. Usually, such effects are created using analog multipliers, but they generally have small dynamic ranges, poor signal-to-noise ratios, slow response speeds, and variations in characteristics between elements, making adjustments difficult. There were drawbacks such as the need for It is also possible to create such an effect using a digital multiplier, but the problem is that the circuit would be complicated and, in particular, each amplitude modulation signal would have to be generated by an independent circuit, making it expensive. Ta. SUMMARY OF THE INVENTION An object of the present invention is to provide an envelope generator for an electronic musical instrument that can form envelope waveforms for various amplitude modulation effects and control volume levels with a simple configuration. In order to achieve the above object, the envelope generator of the electronic musical instrument of the present invention uses a speed parameter to represent an envelope waveform to be generated in response to a key press, and repeatedly calculates a calculation value corresponding to the speed parameter at a predetermined timing. In an envelope generator of an electronic musical instrument that includes a calculation means and generates envelope data, the envelope data for amplitude modulation effect to be added to a predetermined envelope waveform is calculated at a preset timing different from the calculation timing of the predetermined envelope waveform. storage means for temporarily storing the calculation results, predetermined envelope data calculated by the calculation means in response to key presses, amplitude modulation effect envelope data selected from the storage means, and key presses; a multiplication means for time-sharingly multiplying pitch information corresponding to each tone and volume data set by tone color information corresponding to each tone; and a multiplication means for converting the calculation result of the multiplication means into musical waveform data. The present invention is characterized in that it includes a converting means for converting. The present invention will be described in detail below with reference to examples. FIG. 1 is a diagram explaining the principle of an envelope generator used in the present invention. As shown in the figure, in the present invention, the envelope is divided into a plurality of phases, for example, 12 phases (hereinafter referred to as phases). The attack portion of the envelope is divided into phases 1 to 4. The day-care section is from phase 5 to phase 8,
The pressed state of the key is stopped at phase 8,
It is represented by 12 phases from phase 9 to phase 12 in the release section after key release. The envelope data is then represented as a floating point number and divided into an exponent and a mantissa as shown in Table 1.

【table】

[Table] For example, if the maximum value is 1 or less, the maximum level of the envelope is represented by the exponent "000" and the mantissa "000". If 1.000 x 2 ⁰ is 0db, the minimum level is the exponent "111" and the mantissa "000". 000” is 1.000×2 ^-7 (−42db)
It is expressed as Then, the exponential characteristic of the envelope is formed using this floating point characteristic. That is, the decay and release parts can be easily formed by down-counting the mantissa part and up-counting the exponent part. The attack portion is obtained by shifting data in accordance with the phase using a binary shift circuit, which will be described later. A method of expressing and forming ADSR using binary floating point numbers as described above is described, for example, in Japanese Patent Application Laid-Open No. 54-1609 "Amplitude Generator for Electronic Organs" proposed by the present applicant.
Since it is disclosed in , the details are omitted. In the present invention, the attack, decay, and release lengths are represented as digital parameters,
This parameter is effectively applied to the envelope calculation means. Now, considering the release part, in order to be able to change the length of the release from 1 millisecond to 128 milliseconds, this is divided discretely into, for example, 32 steps, and each position is represented by a 5-bit binary number. This is sent to the speed parameter decoder as an envelope speed parameter, SL×SC/RT SL: Relative sustain level where SL = (sustain level - minimum level) SC: System clock calculation time) RT: Release time where RT/ SC indicates the number of calculations during the release time. generates a value corresponding to . This value is the release time
It means the amount of displacement per each calculation cycle included in RT. Therefore, if a different speed parameter is given for each phase during release, the calculation time included in that phase allows the release to proceed according to the displacement. The same is true for DK. For example, if the sustain level is 0dB, the level at which the release ends, that is, the minimum level, is −
Assuming 42dB, the relative sustain level SL is SL=
0-(-42)=42dB. Assuming that the envelope calculation time is performed every 1 ms, SC = 1 × 10 ^-3 (S), and the release time is 14 ms, RT = 14 × 10 ^-3 (S) SL × SC / RT = 42 × 1 × 10 ^-3 / (14×10 ^-3 )=
It becomes 3dB. For release (attenuation), system clock 1
By applying 3dB of attenuation to each point, a 14ms release can be created. Therefore, it is sufficient to select a value necessary for changing by -3 dB and give it as a displacement value. In this case, the displacement value is
It is expressed as an exponent of 000 and a mantissa of 100. This value does not correspond to Table 1. Here, the calculation for forming the envelope of the present invention will be explained. As mentioned above, for decay and release, the mantissa part is down-counted and the exponent part is up-counted. For example, if the current value is 0 dB, the exponent part is 000 and the mantissa part is 000. Assuming that 3 dB of attenuation is given in the release section as described above, the displacement value is calculated as shown in Table 2, with the exponent part being 000 and the mantissa part being 100.

[Table] In Table 2, the mantissa 100, which is converted into a two's complement in the quantile mantissa part, is added to the mantissa part 000 of the current value.
At this time, since no carry occurs from the mantissa part, it is determined that this is a borrow during subtraction, and 1 is added to the exponent part. 001 with the result of addition of the exponent part of the current value and displacement value and the borrow
becomes. Therefore, the calculation result has an exponent part of 001 and a mantissa part of 100, which is 1.5 x 2 ^-1 from Table 1, which is about -3 dB. Next, in the case of attack, if the current value is -39 dB, the exponent part is 111 and the mantissa part is 100. When increasing this by 9 dB, the displacement value is set to exponent part 001 and mantissa part 100, and calculations are performed as shown in Table 3.

[Table] In Table 3, if you add 111, which is the two's complement of the quantile index part 001, from the current value index part 111, it becomes 110.
becomes. (The carry of this result is rounded down.) Next, the mantissa part is obtained by adding the current value 100 and the displacement value 100.
000, resulting in a carry. The mantissa part of this attack is treated as a carry by addition, and the exponent part is subtracted by 1. Therefore, 110 minus 1 becomes 101. The exponent part 101 and the mantissa part 000 are 1.0×2 ^-5 from Table 1.
It becomes 30dB. As described above, attack and
Calculations are performed to form a release envelope. FIG. 2 is an explanatory diagram showing the configuration of an embodiment of the present invention, and is a block diagram of an envelope generation circuit in which the above-mentioned speed parameter is applied to the above-mentioned floating point arithmetic circuit. In the figure, in response to input of key data, the phase initial value generation circuit 11 generates the initial value of the ADSR phase of the envelope explained in Figure 1 (phase 1 in the case of key-on, phase 9 in the case of key-off), and envelope calculation control is performed. used for. Here, when the envelope phase is 1 to 8, when key-off data is input, the phase initial value generating circuit 11 is forcibly set to phase 9 and starts releasing whatever the current phase is. The value of this phase is determined by the phase decoder ADSR.
It is decoded by the control circuit 12 and used for phase transition control of required circuits within the system. One of them is a phase end value prediction circuit 19, into which sustain level data is input, and the predicted value of each phase end value is sent to a comparator 18, where it is compared with a value from an arithmetic unit (described later), and when they match, a match signal is output. It is sent to the phase decoder ADSR control circuit 12, 1 is added to the phase adder 13, and stored in the phase memory 10. It is sent to the phase decoder ADSR control circuit 12 via the phase initial value generation circuit 11, and is used for the next phase. Transition. Here, the sustain level data input to the phase closing price prediction circuit 19 is as follows:
This is variable data input from the outside, and the closing price of each phase in decay and release changes depending on this value. On the other hand, the part where the fall time changes in the release part or decay part is discretely divided into, for example, 32 parts, each position is represented by a 5-bit speed parameter, and the speed parameter at the desired position is input into the speed parameter decoder 23 and decoded. generates a displacement value of . This decoder 23 and the next binary shift circuit 20 are controlled by the phase decoder control circuit 12, and in the case of release and decay, the decoder output passes through the binary shift circuit 20 without being shifted and is sent to the addition/subtraction control circuit 21. The phase decoder ADSR control circuit 12 controls addition and subtraction to the arithmetic unit, and in the case of release and decay, it operates to add exponents and subtract mantissas as described above. An adder 16 and an adder 17 in the arithmetic section add and subtract data read from the exponent memory 14 and the mantissa memory 15, which store exponents and mantissas based on Table 1, using data from the addition/subtraction control circuit 21, and add and subtract data read from the mantissa memory 15 and the comparator 18. The data is sent to the data selection circuit 24. The comparator 18 uses the added exponent from the adder 16 and the subtracted mantissa from the adder 17 to the phase closing price prediction circuit 1.
9 and the computation is repeated until they match or exceed the final value. That is, the exponent from adder 16 and adder 17
The mantissa from . Here phase memory 1
0, the exponent memory 14, and the mantissa memory 15 have the same number of words as the sum of the number of tones and the number of envelopes for amplitude modulation effect. In this way, calculations are repeated for each phase to form a release or decay of length corresponding to the parameter expressed by the 5 bits described above. Note that as a function of the data selection circuit 24, for example, when the adder 16 overflows, the overflow is detected by the overflow detection circuit 28, and the data is transferred to the exponent memory 14.
Instead of sending this to the mantissa memory 15, control is performed to select the memory output or to select and send the initial amplitude value in the next phase from the phase initial amplitude value generating circuit 22. There are two cases of this overflow. One is when it becomes larger than the maximum value in attack calculation, and the other is when it becomes smaller than the minimum value in decay and release calculation. In the case of attack, the current value is -3dB exponent part 001 and mantissa part 100, and the displacement value exponent part 010 corresponds to 12dB.
If the mantissa part is 000, the current value + displacement value is -3 + 12
=9dB exceeds the maximum value of 0dB. Two's complement 110 of quantile exponent part 010 and current value exponent part
Adding 001 gives 111, where the exponent part is 111 and the mantissa part is 100.
According to Table 1, the value is approximately -39dB. Therefore, if no carry occurs from the exponent part at the time of attack, it is treated as an overflow. Similarly, if the current value is -3 dB and the quantile exponent part is 001 and the mantissa part is 000, which corresponds to 6 dB, then adding the two's complement of the quantile exponent part 111 and the current value 001 becomes 000, which causes a carry. However, the calculation result is exponent part 000 and mantissa part 100, and the maximum value is 0dB exponent part 000.
Since the definition of the mantissa part 000 is exceeded, this is also detected as an overflow and processed. Next, in the case of Decay and Release, the current value is approximately -
Considering the case where the exponent part 111 and the mantissa part 100 are 39 dB, and the displacement value exponent part 001 and the mantissa part 000 correspond to 6 dB, the current value minus the displacement value is -39-6=-45 dB, which is smaller than the minimum value. When the current value exponent part 111 and the displacement value exponent part 001 are added, it becomes 000 and a carry occurs. The result of this calculation is approximately +3 dB with an exponent part of 000 and a mantissa part of 100. Therefore, the exponent carry in Decay and Release is treated as an overflow. Next, in the case of an attack, the decoded data from the speed parameter decoder 23 is sent to the binary shift circuit 20. If the binary shift circuit 20 shifts the data to increase in accordance with the phase, a good attack shape can be obtained. for example,
During phase 1, the data from the speed parameter decoder 23 is shifted to the left by 2 bits, that is, shifted to quadruple. Next, in phase 2, the data is shifted to the left by 1 bit, that is, shifted to double the data. When phase 3 is reached, the data is sent without being shifted, and in phase 4, the data is set to be shifted to the right by 1 bit, that is, set to 1/2.
For example, if the quantile exponent part is 001 and the mantissa part is 000, which corresponds to 6 dB, when the phase is 1, the exponent part is 100 and the mantissa part is 000.
24dB at 000, exponent part 010 mantissa part in case of phase 2
000 has a displacement value of 12 dB, phase 3 has a displacement value of 6 dB, and phase 4 has a displacement value of 3 dB with an exponent part of 000 and a mantissa part of 100. The same applies to other displacement values. Here, the comparison values from the phase closing price prediction circuit 19 are as follows: phase 1-12 dB exponent part 010 mantissa part 000, phase 2-6 dB exponent part 001 mantissa part 000, phase 3-
3dB exponent part 001 mantissa part 100, phase 4 exponent part 000
When the mantissa part is 000, an attack having the shape shown in FIG. 1 is formed. The envelope thus formed
The amplitude of each phase of ADSR is temporarily stored in exponent memory 14 and mantissa memory 15. A predetermined normal envelope is generated in this way, and an envelope for the amplitude modulation effect is also generated by the same envelope generator. The formed amplitude modulation effect envelope is temporarily stored in the buffer memory 25. Here, for example, for a repeat effect such as a marimba, an envelope for the repeat effect can be easily obtained by inputting a predetermined repeat signal as a start signal to the key data line and setting the ADSR parameter of the shape of the attenuated sound waveform. The crescendo effect can also be obtained in the same way by inputting the logical sum of all key press signals as the start signal and setting the slow rise ADSR parameter. Next, all the calculated normal envelopes are read out from the exponent memory 14 and the mantissa memory 15 and sent to the data selection circuit 39, and at the same time, the amplitude modulation effect envelope is read out from the buffer memory 25 under the control of the execution control circuit 40. The data is sent to the data selection circuit 39. Here, the normal envelope and the amplitude modulation effect envelope are multiplied, and the multiplication method that is characteristic of the ADSR generator using floating point arithmetic used in the present invention will be described below. A=(1+a ₁ ×2 ^-1 +a ₂ ×2 ^-2 +a ₃ ×2 ^-3 )×2 ^-c (1) B=(1+b ₁ ×2 ^-1 +b ₂ ×2 ^-2 +b ₃ ×2 ^-3 ) × 2 ^-d (2) When the two numbers A and B are expressed as floating point numbers using binary numbers, they can be expressed as in equations (1) and (2). This is the number corresponding to Table 1. Now, when A and B are multiplied, it is expressed by equation (3). A×B={1+(a ₁ +b ₁ )×2 ^-1 +(a ₂ +b ₂ )×2 ^-2 +(
a ₃ + b ₃ ) × 2 ^-3 + a ₁ b ₁ × 2 ^-2 + (a ₁ b ₂ + a ₂ b ₁ ) × 2 ^-3 + (a ₁ b ₃ + a ₂ b ₂ +
a ₃ b ₁ ) × 2 ^-4 + (a ₂ b ₃ + a ₃ b ₂ ) × 2 ^-5 + a ₃ b ₃ × 2 ^-6 } × 2 ^-(c+d) (3) Expression (3) is replaced by a ₁ b ₁ ×2 ^-2 If we cut down the following terms, we get A × B ≒ {1 + (a ₁ + b ₁ ) × 2 ^-1 + (a ₂ + b ₂ ) × 2 ^-2 + (
a ₃ + b ₃ )×2 ^-3 }×2 ^-(c+b) (4) An approximate expression as shown in equation (4) is obtained. As can be seen from this approximate expression, the multiplication of A and B can be expressed by addition of the decimal parts of the respective mantissa parts and addition of the exponent parts. According to actual measurements, the maximum error in this truncation error is about ±2 dB.
It can be assumed that this method can be applied to envelope waveforms expressed by monotonically increasing or decreasing functions, and sufficiently satisfactory results have been obtained in auditory tests. In the first step of multiplication, normal ADSR data is multiplied by loudness data, which will be described later, for outputting different volumes depending on the tones or range of sounds to be produced. First, normal ADSR data is selected by the data selection circuit 39 under the control of the execution control circuit 40, and the exponent part is sent to the exponent adder 27 and the mantissa part to the mantissa adder 29. On the other hand, the data selection circuit 30 is connected to the execution control circuit 40.
selects loudness data under the control of and similarly sends it to the other inputs of exponent adder 27 and mantissa adder 29. The exponent adder 27 is a normal
The exponent parts of the ADSR data and loudness data are added and sent to the exponent subtracter 33. Similarly, mantissa adder 29 adds the mantissa and sends it to latch 34. For example, if ADSR data is -3dB exponent part 001 and mantissa part 100, loudness data is -6bB exponent part 001 mantissa part.
If it is 000, the calculation result will be an exponent part of 010 and a mantissa part of 100, which is about -9 dB from Table 1. If a carry occurs as a result of adding the exponent part, a signal is sent to the exponent subtractor 33 to subtract a numerical value 1 from the calculation result of the exponent part. The exponent subtracter 33 subtracts 1 from the exponent when there is a carry signal, and sends the data as is to the latch 34 when there is no carry signal. Latch 34 latches the results of these operations and completes the first step. Next, as a second step, the above result is multiplied by an envelope for amplitude modulation effect. Similar to the first step, the data selection circuit 39 now selects the data in the buffer memory 25 and inputs the data to the exponent adder 2.
7 and mantissa adder 29. The data selection circuit 30 selects the data in the latch 34 and inputs the data to the exponent adder 2.
7 and mantissa adder 29. Exponent adder 2
7. The exponent adder 29 and the exponent subtracter 33 operate in the same manner as in step 1, and the resulting data is sent to the latch 3.
4 and the exponent part is sent to a decoder 31 and the mantissa part to a binary shift circuit 32. Here the aforementioned
ADSR data and loudness data calculation result -9dB
For example, if the amplitude modulation envelope is -3dB
If the exponent part is 001 and the mantissa part is 100, the calculation will be performed in the same way as above, and the mantissa will be carried and the exponent part addition result will be 011.
Then minus 1 becomes 010 and the mantissa part is 000 -
It becomes 12dB. The decoder 31 decodes the data of the exponent part and controls the binary shift circuit 32. If an exponent carry occurs during the operations in step 1 and step 2, a carry signal is sent from the exponent adder 27 to the overflow detection circuit 26. Overflow detection circuit 26 holds this signal and sends it to decoder 31. This signal means a numerical value other than the dynamic range that can be expressed when, for example, the exponent part is 111+111=110 and a carry occurs. The decoder 31 therefore operates to round the numerical value to the minimum level in the presence of this signal. The binary shift circuit shifts the mantissa data based on the signal from the decoder 31, converts the data from a floating point number to a fixed point number, and sends the data to the buffer memory 35. Buffer memory 35 temporarily stores this data under the control of execution control circuit 40. The buffer memory 35 is repeatedly read out by the execution control circuit 40 at a rate different from the calculation rate of the ADSR generator. The DA converter 36 converts this data into an analog voltage to form an envelope waveform. FIG. 3 is a specific circuit example showing a loudness data input section. Loudness data is provided by what is stored in the loudness memory 37 and by a loudness data bus controlled externally. The loudness memory 37 is read out based on the frequency information of the operated keyboard and the information of the timbre to be formed, and sent to the data selection circuit 38. The data selection circuit 38 selects this data and data from the loudness data bus using a data selection signal. The selected loudness data is used as described above. The data in the loudness level memory 37 controls the level depending on the frequency of the sound being generated. The loudness data bus is data for adjusting the balance between tones. If this loudness level memory 37 stores a number of patterns smaller than the number of tones that can be produced, and performs the aforementioned floating point multiplication with the data for the balance between tones from the loudness data bus, a small memory capacity will be required, making it economical. A system is created. As explained above, according to the present invention, an electronic musical instrument is provided with a calculation means for representing an envelope waveform using a speed parameter and calculating the envelope waveform at a predetermined timing based on the speed parameter. Envelope data for the amplitude modulation effect is temporarily stored in a buffer memory, and predetermined envelope data corresponding to key presses calculated by the calculation means, the envelope data for the amplitude modulation effect, and volume data such as loudness data are stored. Multiplication is performed in a time-division manner. In particular, the envelope data calculated by the calculation means is represented by a floating-point exponent part and a mantissa part, and the multiplication means is composed of calculation parts corresponding to these parts, and the multiplication of the three components described above is performed by simple addition. , and since it is possible to perform time-divisional calculations by selecting data using a single multiplication means, it is possible to freely add a desired amplitude modulation effect, and it is also possible to arbitrarily adjust the volume level of each tone. Become. As a result, various envelope waveforms can be formed easily and inexpensively without requiring multiple multiplier circuits.

[Brief explanation of the drawing]

Fig. 1 is an explanatory diagram of the principle of the envelope generator used in the present invention, Fig. 2 is an explanatory diagram showing the configuration of an embodiment of the present invention, and Fig. 3 is a specific circuit example of the main part of the embodiment of Fig. 2. Yes, in the figure, 10 is phase memory, 1
2 is a phase decoder ADSR control circuit, 14 is an exponent memory, 15 is a mantissa memory, 16 and 17 are adders, 18 is a comparator, 19 is a phase closing price prediction circuit, 20 and 32 are binary shift circuits, and 21 is an addition/subtraction a control circuit, 23 a speed parameter decoder;
24, 30, 38, 39 are data selection circuits, 2
5, 35 are buffer circuits, 26 are overflow circuits, 27 are exponent adders, 29 are mantissa adders, 31
33 is an exponent subtracter, 34 is a latch circuit, 36 is a DA converter, 37 is a loudness level memory, and 40 is an execution control circuit.

Claims (1)

[Scope of Claims] 1. An envelope waveform to be generated in response to a key press is represented using a velocity parameter, and includes a computing means for repeatedly computing a computed value corresponding to the velocity parameter at a predetermined timing to form envelope data. In envelope generators for electronic musical instruments,
Storage means for calculating amplitude modulation effect envelope data to be added to a predetermined envelope waveform by the calculation means at a preset timing different from the calculation timing of the predetermined envelope waveform and temporarily storing the calculation results; Predetermined envelope data calculated by the calculation means in response to the key press, envelope data for amplitude modulation effect selected from the storage means, pitch information corresponding to the key press, and tone color information corresponding to each tone color. An envelope for an electronic musical instrument, characterized in that it comprises a multiplication means for selecting and time-divisionally multiplying each volume data to be set, and a conversion means for converting the calculation result of the multiplication means into musical sound waveform data. generator. 2. The envelope data calculated by the calculation means is represented by an exponent part and a mantissa part of a floating point number,
Claim 1, wherein the multiplication means comprises an arithmetic unit corresponding to the exponent part and the mantissa part.
Envelope generator for electronic musical instruments as described in Section 1. 3. An envelope generator for an electronic musical instrument according to claim 1, wherein said converting means comprises a circuit for converting a floating point number representing envelope data into a fixed point number.