JP3097324B2 - Digital sound data output device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital audio data output device capable of reducing an auditory noise caused by quantization noise.

[0002]

2. Description of the Related Art It is well known that when converting digital audio data into an analog audio signal, quantization noise is generated based on a quantization error.

[0003] In order to reduce an auditory noise caused by quantization noise, a technique combining oversampling processing and noise shaping processing is conventionally used. In the oversampling process, first, N-1 appropriate samples are newly inserted between each sample in the sample sequence of the original digital data, and the total number of samples is increased N times. Next, a digital low-pass filtering process in which the cutoff frequency is the same as half the sampling frequency of the original digital data is performed on the signal whose number of samples has been increased by N times. Then, the output of this processing is D / A converted at an oversampling frequency that is N times the sampling frequency of the original digital data. As a result, the frequency band is expanded to N times the frequency band of the original digital data while maintaining the frequency characteristics of the original digital data.

[0004] Next, in the noise shaping process, a kind of high-pass filtering process is executed in the process of the above-mentioned oversampling process, so that the energy distribution of only the quantization noise component is changed without changing the entire energy. It is biased to a frequency band higher than half the sampling frequency of the original digital data. Then, at the time of D / A conversion using the oversampling frequency, a frequency component higher than half the sampling frequency of the original digital data is cut by the analog filter.

[0005] By the technique combining the above oversampling processing and noise shaping processing, the perceived noise caused by quantization noise is reduced.

[0006]

However, in the conventional example in which the above-described oversampling processing and noise shaping processing are combined, low-pass filtering processing is required as oversampling processing, and a kind of high-pass filtering processing is used as noise shaping processing. However, there is a problem that the processing becomes complicated.

Further, in the above-mentioned conventional example, if a D / A converter having good response performance is not used, the sound quality of an analog signal cannot be sufficiently ensured, resulting in an increase in the cost of the apparatus. There is a problem of inviting.

An object of the present invention is to make it possible to reduce an auditory noise caused by quantization noise using a general-purpose low-cost D / A converter.

[0009]

According to the present invention, an X signal which changes at a sampling interval synchronized with a predetermined sampling frequency is used.
Digital audio data such as digital musical sound data of 1 bit is resampled at an oversampling frequency having a frequency N times as high as the sampling frequency, and is output while being requantized with BIT bits less than X bits, and the number of converted bits Is a digital audio data output device that performs BIT conversion and digital / analog conversion with a digital / analog converter whose conversion frequency is equal to the oversampling frequency.

The first aspect of the present invention has the following configuration. First, there is provided the following offset data output means. That is, for each sampling interval synchronized with the sampling frequency, this means OFST _M = 2 ^(X-BIT) × {(M−1) / N} (1
≦ M ≦ N) N pieces of offset data OFST _M (1 ≦ M ≦ N) determined by an arithmetic expression represented by the following equation are output.

[0011] Then, for each sampling interval synchronized with the sampling frequency, the over-sampling intervals synchronized with the oversampling frequency, adding or subtracting the offset data OFST _M outputted from the offset data output means into digital audio data of X bits By doing so, there is an arithmetic means for generating an added value of X bits for each of N samples.

Data conversion means for generating BIT bit output data by discarding lower (X-BIT) bits of the X-bit addition value output from the arithmetic means and outputting the same to a digital / analog converter Having.

As a second aspect of the present invention, the following configuration is provided. First, there is provided the following offset data output means. That is, this means OFST _M = 0 (M = 1) OFST _M = 2 ^(X-BIT) / N (2 ≦ ⁾ at every sampling interval synchronized with the sampling frequency, at the over sampling interval synchronized with the over sampling frequency. M ≦ N) to output a is determined by the calculation formula the N offset data _{OFST M (1 ≦ M ≦ N} ) represented by.

[0014] Then, for each sampling interval synchronized with the sampling frequency, the over-sampling intervals synchronized with the oversampling frequency, sequentially accumulates the offset data OFST _M outputted from the offset data output means into digital audio data of X bits By doing so, there is an operation means for generating an accumulated value of X bits for each of N samples.

Data conversion means for generating BIT bit output data by truncating lower-order (X-BIT) bits of the added value of X bits output from the arithmetic means and outputting the same to a digital / analog converter Having.

[0016]

When the arithmetic means performs an N-times oversampling process, it does not generate a value oversampled by N times as a value obtained by simply linearly interpolating between the samples of the input digital audio data. Of the BIT-bit digital sound data at the time of digital / analog conversion with respect to the sample of the X-bit digital sound data of
A value obtained by dividing the quantization step width by the precision of the original X-bit digital acoustic data by 1 / N {2 ^(X-BIT) / N}
Generate an X-bit value that is oversampled N times with a value that increases fixedly by increments.

The lower (X-B) of the X-bit value
The IT) bit is truncated to generate BIT bit output data, which is output to the digital / analog converter.

The value of the BIT bit output data thus generated changes so as to follow the original value of the X-bit digital sound data. For this reason, the waveform obtained by digital-to-analog conversion of the BIT bit output data and smoothed by the low-pass filter is compared with the waveform obtained by performing the same processing on the output data generated by the above-described linear interpolation processing, and It becomes closer to the waveform of the X-bit digital sound data.

As a result, the auditory noise caused by the quantization noise is reduced.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Two embodiments of the present invention will now be described with reference to the drawings. First Embodiment <Structure> FIG. 1 is a block diagram showing an overall structure of a first embodiment according to the present invention.

First, X-bit (for example, 24-bit) digital musical sound data generated by a sound source (not shown) is input to the input register 101. X-bit digital tone data output from the input register 101 to the adder 102 changes at a sampling interval 1 / Fs synchronized with the sampling frequency Fs.

On the other hand, from the offset data output unit 103, X-bit offset data (to be described later) corresponding to a mode condition determined by a combination of the multiple N of oversampling and the number of conversion bits BIT of the D / A converter. , At the oversampling interval 1 / (N × Fs) synchronized with the oversampling frequency N × Fs.

The adder 102 converts the X-bit digital tone data input from the input register 101 and the X-bit offset data input from the offset data output unit 103 into oversampling signals synchronized with the oversampling frequency N × Fs. Addition is performed at an interval of 1 / (N × Fs), and the added value is converted to a parallel / serial
Output to

The parallel / serial converter 104 converts the number of conversion bits BIT (X>) of a D / A converter (not shown).
BIT), the lower (X-BIT) bits of the X-bit addition value described above are truncated, and the resulting parallel data whose number of bits is BIT is converted into serial data, and the converted bit number is BIT. The signal is output to a D / A converter (not shown).

For example, when X = 24 and BIT = 16, the lower 8 bits are discarded, and X = 24 and BIT = 18
, The lower 6 bits are truncated. This D / A converter performs D / A conversion with the oversampling frequency N × Fs and the number of conversion bits BIT, and converts digital musical tone data into analog musical tone signals. <Operation> The adder 102 generates X-bit digital tone data from the input register 101 which changes at a sampling interval 1 / Fs synchronized with the sampling frequency Fs, and an oversampling interval 1 / ( By adding the X-bit offset data from the offset data output unit 103 changing at N × Fs at an oversampling interval 1 / (N × Fs) synchronized with the oversampling frequency N × Fs, the input register 101 is obtained. Each time one sample of X-bit digital musical sound data is input, N samples of output data are generated.

The above-described addition processing corresponds to oversampling processing. In this case, the offset data output from the offset data output unit 103 is determined as follows.

First, the parallel / serial converter 104
Is, as described above, the lower order value of the X-bit addition value.
(X-BIT) bits are truncated and the number of bits is BIT
Generate an output value for the D / A converter that is Soshi
Processing to round down one quantization step width of the D / A converter
The value expressed by the precision of the previous X-bit data is d_XTomorrow
If d _XIs represented by the following equation.

[0028]

D _x = 2 ^(X-BIT) (binary expression) Each time one sample of X-bit digital musical tone data is input from the input register 101, the output of N samples generated by the adder 102 Assuming that offset data corresponding to each of the data is OFST _M (1 ≦ M ≦ N), the offset data OFST _M is determined by the following equation.

[0029]

OFST _M = d _X × {(M−1) / N} Now, for example, assume that the bit number X of digital musical tone data is X = 24, the conversion bit number BIT of the D / A converter is BIT = 16, and oversampling is performed. The operation when the multiple N is set to N = 4 will be described below.

The parallel / serial converter 104 has 24
The lower (24-16) = 8 bits of the bit addition value are truncated to generate a 16-bit output value. Therefore, from Expression 1, the value d ₂₄ that represents the one quantization step width of the D / A converter with the precision of the 24-bit data before truncation processing is: d ₂₄ = 2 ⁸ (binary representation) = 000100H ( (Hexadecimal representation). Note that "H" indicates that the data to which the
Indicates that the value is represented in hexadecimal for each bit.

From equation (2), the offset data OFS
T _M (1 ≦ M ≦ 4) is as shown in FIG. 2 (a). FIG. 3 is a diagram showing the above-mentioned OFST _M (1 ≦ M ≦ 4).

In FIG. 3, the horizontal line at the position indicated by “16/18” indicates a D / A where the number of conversion bits BIT = 16.
The quantization step position of the converter is shown, and the horizontal line at the position indicated by “18” indicates the quantization step position of the D / A converter when the number of conversion bits BIT = 18 is assumed.

As can be understood from FIG. 3, the offset data OFST _M (1 ≦ M ≦ 4) is the conversion bit number BIT
= 18, which corresponds to the quantization accuracy of the D / A converter, and the interval between each offset data OFST _M is d ₂₄
= 00000100H divided by a multiple N of oversampling is a value of 000040H.

FIG. 4 shows the offset data OFS described above.
FIG. 9 is a diagram illustrating an operation of the adder 102 based on T _M (1 ≦ M ≦ 4). First, from the input register 101, the sampling interval 1 / Fs synchronized with the sampling frequency Fs
For example, it is assumed that 24-bit digital musical sound data indicated by ◎ in a, b, and c of FIG.

The adder 102 converts the value and the offset data OFST _M (1 ≦ M ≦ 4) for each of the digital tone data a, b, and c into an oversampling frequency synchronized with the oversampling frequency 4Fs. By adding at intervals 1 / (4Fs), (a,
A 24-bit addition value indicated by ◎ or の of (d, e, f) or (b, g, h, i) is obtained and output to the parallel / serial converter 104.

Thereafter, the parallel / serial converter 104
Is obtained by truncating the lower 8 bits of the above 24-bit addition value to obtain (a ₁ , a ₂ , a ₃ , a ₄ ) in the figure.
Or _{1 represented} by △ in (b ₁ , b ₂ , b ₃ , b ₄ )
Obtain a 6-bit output value.

The adder 102 converts the above-mentioned 16-bit output value into the above-mentioned oversampling frequency 4F.
Oversampling interval 1 / (4Fs) synchronized with s
Then, the signal is output to the parallel / serial converter 104.

Here, the adder 102 does not perform the process of adding the offset data to the digital musical sound data, and the parallel / serial converter 104
Assuming that the value obtained by truncating the lower 8 bits of the 24-bit digital tone data that changes every Fs is simply output at the oversampling interval 1 / (4Fs), 16
The output values of the bits are (a ₁ ′, a ₂ ′,
a ₃ ′, a ₄ ′) or (b ₁ ′, b ₂ ′, b ₃ ′,
b ₄ ′).

As can be seen by comparing the 16-bit output values of △ and ● in the figure, the 16-bit output value indicated by 必 ず is always equal to the original 24-bit input value indicated by ◎. Or smaller. On the other hand, the 16-bit output value indicated by △ changes so as to follow the general shape of the original 24-bit input value waveform indicated by ◎. Therefore, the waveform obtained by D / A conversion of the 16-bit output value indicated by △ and smoothing by the low-pass filter becomes closer to the waveform of the original 24-bit digital musical tone data.

As a result, the auditory noise caused by the quantization noise is reduced. In this case, in the conventional example in which the oversampling process and the noise shaping process are combined, the output value changes quickly because the high frequency component is emphasized, and as a result, a D / A converter with good response performance is required. On the other hand, the output value obtained from the adder 102 in FIG.
Since the waveform gradually changes so as to follow the outline of the digital tone data waveform, an analog signal with good sound quality can be obtained even when a general-purpose low-priced D / A converter whose response performance is not so good is used. .

FIG. 5 (a) shows the lower 8 bits after the offset data OFST _M (1 ≦ M ≦ 4) shown in FIG. 2 (a) is added to the 24-bit digital tone data value 73DC6BH. FIG. 13 is a diagram showing 16-bit output values obtained at every oversampling interval 1 / (4Fs) by truncating. As a result, the output value is
It changes as shown in FIG.

As described above, the number of bits X =
24, the operation example in the case where the number of conversion bits BIT of the D / A converter is BIT = 16 and the multiple of oversampling N = 4 has been described. In this embodiment, the number of conversion bits BIT of the D / A converter is The oversampling multiple N can be set to four times and eight times, and the oversampling process can be set to a total of four ways by a combination thereof.

FIGS. 2 (b), 5 (b) and 6 (b) show
An operation example is shown in the case where the number of bits X of digital musical sound data is 24, the number of conversion bits BIT of the D / A converter is BIT = 18, and the multiple N of oversampling is N = 4.

FIGS. 2 (c), 5 (c), and 6 (c)
Has the number of bits X = 24 of the digital tone data, and D /
An operation example in the case where the number of conversion bits BIT of the A converter is BIT = 16 and a multiple N of oversampling is 8 is shown.

Further, FIGS. 2 (d), 5 (d), and 6 (d)
Has the number of bits X = 24 of the digital tone data, and D /
An operation example in the case where the number of conversion bits BIT of the A converter is BIT = 18 and a multiple N of oversampling is N = 8 is shown. Second Embodiment <Structure> FIG. 7 is a block diagram showing the overall structure of a second embodiment according to the present invention.

First, at an oversampling interval 1 / (N × Fs) synchronized with the oversampling frequency N × Fs, the multiplexer 701 selects the input of 24-bit digital tone data when the control signal is at a high level. When the control signal is at the low level, the 24-bit addition value from the adder 703 is selected and output to the general-purpose register 702.

The output of the general-purpose register 702 is
Is input to On the other hand, from the offset data output unit 704, X-bit offset data (described later) according to a mode condition determined by a combination of the multiple N of oversampling and the number of conversion bits BIT of the D / A converter is oversampled. It is output to the adder 703 at an oversampling interval 1 / (N × Fs) synchronized with the frequency N × Fs.

The adder 703 converts the X-bit data value input from the general-purpose register 702 and the X-bit offset data input from the offset data output device 704 into an oversampling interval synchronized with the oversampling frequency N × Fs. Addition is performed by 1 / (N × Fs), and the resultant addition value is output to the multiplexer 701 or the parallel / serial converter 705.

The parallel / serial converter 705 is provided in FIG.
As in the case of the first embodiment, D / A (not shown)
According to the conversion bit number BIT (X> BIT) of the converter,
The lower (X-BIT) bits of the X-bit addition value described above are truncated, and the resulting parallel data whose number of bits is BIT is converted into serial data. / A converter. <Operation> The operation of the second embodiment is different from that of the first embodiment in the following point.

That is, in the first embodiment, for every oversampling interval 1 / (N × Fs) synchronized with the oversampling frequency N × Fs, X bits changing at the sampling interval 1 / Fs synchronized with the sampling frequency Fs are used. X-bit offset data (see FIG. 2 or FIG. 3) that increases linearly is added to the digital tone data of FIG.

On the other hand, in the second embodiment, the output of the adder 703 is fed back to the general-purpose register 702 via the multiplexer 701, so that the output is performed only once at the oversampling interval 1 / (N × Fs). After the digital musical tone data is taken into the general-purpose register 702 by the multiplexer 701, the constant offset data from the offset data output unit 704 is sequentially accumulated three times in the adder 703 at every oversampling interval 1 / (N × Fs). Then, each accumulated power is output to the parallel / serial converter 705.

Now, the offset data OFST _M (1 ≦ M
≦ N) uses the value d _X represented by expression 1 described above, it is determined by the following equation.

[0053]

Equation 3] _{OFST 1 = 000000H OFST M = d} X / N (2 ≦ M ≦ N) FIG. 8 is a content of the offset data output from the offset data output unit 704 of FIG. 7 shown in FIG. 8 (a) In
FIG. 8B shows a case where the number of bits X of digital musical sound data is 24, the number of conversion bits BIT of the D / A converter is BIT = 16, and the multiple N of oversampling is N = 4.
24, BIT = 18 and N = 4 are shown in FIG.
FIG. 2 (c) shows the case where X = 24, BIT = 16, N = 8, and FIG. 2 (d) shows the case where X = 24, BIT =
18, and N = 8 are shown.

FIG. 9 shows the multiple N of oversampling.
6 is an operation timing chart of the second embodiment when the number is 4. Hereinafter, the bit number X of digital musical sound data is 24.
An operation example in the case where the number of conversion bits BIT of the D / A converter is 16 will be described. Here, timing M
Is an oversampling timing synchronized with the oversampling frequency N × Fs. The interval until M changes from 1 to 4 corresponds to the original sampling interval 1 / Fs.

First, at the timing when M = 4, the control signal goes high as shown in FIG.
The 24-bit digital tone data input via the multiplexer 701 is taken into the general-purpose register 702 based on the clock shown in FIG.

Next, at the timing when M = 1, the offset data 0000 is output from the offset data output unit 704.
Since 00H is output (see Equation 3), the above-mentioned digital tone data is loaded into the parallel / serial converter 705 based on the clock of FIG. 9C. Immediately thereafter, since the control signal is at a low level as shown in FIG. 9 (a), the above-mentioned digital musical tone data is transmitted from the multiplexer 701 to the general-purpose register 7 based on the clock of FIG. 9 (b).
02 is fed back.

Subsequently, at the timing when M = 2, the offset data 000040H (see equation 3) is added to the input value from the general-purpose register 702 by the adder 703.
The added value is parallel / parallel based on the clock of FIG.
It is loaded into the serial converter 705. Immediately thereafter, since the control signal is at the low level as shown in FIG. 9A, the above-mentioned digital musical tone data is fed back from the multiplexer 701 to the general-purpose register 702 based on the clock of FIG. 9B.

The same applies to the timing when M = 3, and the adder 703 sets the offset data 000040
H (see equation 3) is added to the input value from the general-purpose register 702, and the added value is loaded to the parallel / serial converter 705 based on the clock of FIG. Immediately thereafter, since the control signal is at a low level as shown in FIG. 9 (a), the above-mentioned digital musical tone data is transmitted from the multiplexer 701 to the general-purpose register 7 based on the clock of FIG. 9 (b).
02 is fed back.

Then, at the timing when M = 4, the adder 703 adds the offset data 000040H (see equation 3) to the input value from the general-purpose register 702,
The added value is parallel / parallel based on the clock of FIG.
It is loaded into the serial converter 705. Immediately thereafter, FIG.
The clock for the general-purpose register 702 rises in FIG. 9B. However, since the control signal is at a high level as shown in FIG. .

By the above operation, processing substantially equivalent to the case of the first embodiment is realized. Other Embodiments In the above-described first or second embodiment, the case where the sample value of the digital musical tone data changing at the sampling interval 1 / Fs synchronized with the sampling frequency Fs increases or decreases. The positive offset data is always added to the digital tone data, but the positive offset data is added when the sample value of the digital tone data increases, and the negative offset data is added when the sample value of the digital tone data decreases. You may make it.

[0061]

According to the present invention, the output data of the BIT bit output from the data conversion means changes so as to follow the value of the original X-bit (X> BIT) digital sound data. Therefore, the output data is digital /
The waveform converted to analog and smoothed by the low-pass filter is closer to the waveform of the original X-bit digital sound data compared to a waveform obtained by performing the same processing on output data generated by ordinary linear interpolation processing. As a result, it is possible to reduce the auditory noise caused by the quantization noise.

In this case, in the conventional example in which the oversampling process and the noise shaping process are combined, the change of the output value is fast because the high frequency component is emphasized.
A digital / analog converter with good response performance is required. On the other hand, the value of the output data obtained according to the present invention changes gradually so as to follow the outline of the waveform of the digital sound data, so that a general-purpose low-cost digital / analog converter having not so good response performance. It is possible to obtain an analog signal with good sound quality even if a device is used.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram of a first embodiment according to the present invention.

FIG. 2 is a diagram showing output data of an offset data output unit 103.

FIG. 3: 4 times oversampling, 16 bit D / A
It is a figure which shows the offset data in the case of a converter.

FIG. 4 is 4 × oversampling, 16 bit D / A
FIG. 5 is a diagram showing the amplitude of musical sound data in the case of a converter.

FIG. 5 is a diagram showing output data of an adder 102;

FIG. 6 is a diagram showing output data of a parallel / serial converter 104;

FIG. 7 is an overall configuration diagram of a second embodiment according to the present invention.

FIG. 8 shows an offset data output unit 704 according to the second embodiment.
FIG. 7 is a diagram showing output data of FIG.

FIG. 9 is an operation timing chart of the second embodiment.

[Explanation of symbols]

101 input register 102, 703 adder 103, 704 offset data output device 104, 705 parallel / serial (P / S) converter 701 multiplexer 702 general-purpose register

Claims (3)

(57) [Claims] An X-bit digital sound data that changes at a sampling interval synchronized with a predetermined sampling frequency is converted to an oversampling frequency having a frequency N times the sampling frequency. , And re-quantizes with BIT bits less than the X bits, and outputs it. The digital-to-analog converter converts the number of conversion bits to BIT and the conversion frequency is equal to the oversampling frequency. In the digital audio data output device, OFST _M = 2 ^(X-BIT) × ｛(M−1) /, at each sampling interval synchronized with the sampling frequency, at an oversampling interval synchronized with the oversampling frequency. N｝ (1
≦ M ≦ N) offset data output means for outputting N pieces of offset data OFST _M (1 ≦ M ≦ N) determined by an arithmetic expression represented by the following formula: oversampling intervals synchronized with the sampling frequency, by adding or subtracting the offset data OFST _M output from the offset data output means into digital audio data of the X bits, the sum of each X-bit N samples Calculating means for generating, and by cutting off lower-order (X-BIT) bits of the added value of the X bits output from the calculating means,
Data conversion means for generating bit output data and outputting the output data to the digital / analog converter. 2. The method according to claim 1, wherein N, M, X, and BIT are natural numbers, respectively, and X-bit digital sound data that changes at a sampling interval synchronized with a predetermined sampling frequency is converted to an oversampling frequency having a frequency N times the sampling frequency. , And re-quantizes with BIT bits less than the X bits, and outputs it. The digital-to-analog converter converts the number of conversion bits to BIT and the conversion frequency is equal to the oversampling frequency. In the digital audio data output device, OFST _M = 0 (M = 1) OFST _M = 2 ^(X-BIT ), at each sampling interval synchronized with the sampling frequency, at an oversampling interval synchronized with the oversampling frequency. ⁾ / in N (2 ≦ M ≦ N) And offset data output means for outputting N pieces of offset data OFST _M determined by computing equation (1 ≦ M ≦ N) to be, for each sampling interval synchronized with the sampling frequency, the over-synchronized with the oversampling frequency in the sampling interval, by sequentially accumulating the offset data OFST _M output from the offset data output means into digital audio data of the X bits, and calculation means for generating an accumulated value of each X-bit N samples , By cutting off lower (X-BIT) bits of the added value of the X bits output from the arithmetic means,
Data conversion means for generating bit output data and outputting the output data to the digital / analog converter. 3. The digital sound data output device according to claim 1, wherein the digital sound data is digital musical sound data.