JP2003124785A - Device for converting sampling frequency - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sampling frequency converting device with high distortion accuracy in a configuration of a small circuit scale. SOLUTION: A sampling frequency comparator 3 compares a sampling frequency fsi with a sampling frequency fso to generate output time position information indicating a time position of output data with respect to a time position of input data to a curve interpolating means 4. The curve interpolating means 4 performs Bezier curve interpolation between input data with interpolation operation using de Casteliau algorithm on the basis of the output time position information to generate output data.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sampling frequency conversion device for converting a sampling frequency of audio data or the like, and more particularly to a sampling frequency conversion device which realizes conversion of a sampling frequency even when the sampling frequency is asynchronous between input and output. The present invention relates to a digitized frequency conversion device.

[0002]

2. Description of the Related Art At present, as a sampling frequency conversion device for converting the sampling frequency of audio data, voice data of communication, etc., there is, for example, the one disclosed in Japanese Patent Application Laid-Open No. 6-120776. As shown in FIG. 9, this stores the input data of the first sampling frequency in the storage means 101,
The FIR digital filter 102 interpolates the input data read from the storage means by multiplying the first sampling frequency by n times, and further, the linear interpolation means 103 linearly interpolates the output data of the FIR digital filter 102 to obtain the second data.
Sampling frequency output data is used to realize sampling frequency conversion even when the input and output sampling frequencies are asynchronous.

[0003]

However, in the conventional sampling frequency conversion apparatus shown in FIG. 9, since linear interpolation is performed, the distortion becomes large in a portion where the change of the input data is large and the waveform has a high curvature. In order to obtain the required distortion accuracy, the value of the magnification n when multiplying the first sampling frequency by n needs to be large, for example, 256 times. For this reason, the filter coefficient is increased, and the number of calculations is increased, thus increasing the circuit scale.

An object of the present invention is to provide a sampling frequency conversion device having a small circuit scale and good distortion accuracy.

[0005]

A sampling frequency conversion device of the present invention inputs data at a first sampling frequency, and a second sampling frequency conversion device.
Is a sampling frequency conversion device for outputting data at a sampling frequency, comparing the first sampling frequency with the second sampling frequency, and comparing the time point of the input data with the time point of the output data. Time position information generating means for generating output time position information indicating a position, and curve interpolation between the preceding and following input data, based on the output time position information,
And a curve interpolating means for calculating a value corresponding to the output time position information on the curve and outputting the value as the output data at the second sampling frequency.

It is preferable that the curve is a Bezier curve, and the curve interpolating means performs the calculation based on the algorithm of Do Castello.

The above curve is a Bezier curve, and the above curve interpolating means performs the above calculation based on the algorithm of Do Castello, and it is also preferable to weight each point used for the above calculation.

The curve is a quadratic Bezier curve, and the output time position information is 1 / s of the sampling cycle of the input data.
It is also preferable that the unit is 2 ⁿ (n is an integer of 1 or more), and the curve interpolating means performs the calculation based on the midpoint division method.

The above curve is a quadratic Bezier curve, and the position of the control point in the time direction is set to the position of 1/2 of the sampling period of the above input data from the starting point of the above curve, and the value of the above control point data is , 1 of the input data corresponding to the start point
The slope of the starting point of the curve that is approximately obtained from the previous input data and the input data corresponding to the end point of the curve, and the input data after one of the input data corresponding to the end point. And the slope of the end point of the curve that is approximately obtained from the input data corresponding to the start point, the input data corresponding to the start point, and the data corresponding to the end point. preferable.

The above-mentioned curve is a cubic Bezier curve, and the position of the first control point in the time direction is set to the position of 1/3 of the sampling period of the above-mentioned input data from the starting point of the above-mentioned curve, and the second control point. The position in the time direction of is the position of 2/3 of the sampling period of the input data from the start point of the Bezier curve,
The value of the data of the first control point is the value of the curve that is approximately obtained from the input data immediately before the input data corresponding to the start point and the input data corresponding to the end point of the curve. The value of the data of the second control point is determined from the slope of the starting point and the input data corresponding to the starting point, and the value of the data of the second control point corresponds to the input data one after the input data corresponding to the ending point and the starting point. It is also preferable to be determined from the slope of the end point of the curve that is approximately obtained from the input data and the input data corresponding to the end point.

Further, the sampling frequency conversion apparatus of the present invention is
A sampling frequency conversion device which inputs data at a first sampling frequency and outputs data at a second sampling frequency, wherein data is input at the first sampling frequency and the input data is set to a predetermined value. Storage means, a digital filter for taking out the data held in the storage means, multiplying the first sampling frequency by n times and interpolating, the first sampling frequency and the second sampling. Between the time position information generating means for comparing the frequency and the output time position information indicating the time position of the final output data with respect to the time position of the output data from the digital filter, and the output data from the digital filter. Curve-interpolating, the value corresponding to the output time position information on the curve is calculated based on the output time position information, and the value is used as the final output data. It is also preferably characterized by comprising a curve interpolation means for outputting at sampling frequency.

It is also preferable that the curve is a Bezier curve, and the curve interpolating means performs the calculation based on the algorithm of Do Castello. The above curve is a Bezier curve, and the above curve interpolating means performs the above calculation based on the algorithm of Do Castello, and it is also preferable to weight each point used for the above calculation.

The curve is a quadratic Bezier curve, and the output time position information is 1/2 ⁿ (n is an integer of 1 or more) of the sampling period of the output data from the digital filter.
It is also preferable that the curve interpolation means performs the above calculation based on the midpoint division method.

The curve is a quadratic Bezier curve, and the position of the control point in the time direction is a position from the start point of the curve to half the sampling period of the output data from the digital filter. The value of the data corresponds to a value obtained by subtracting 1/4 of the output data immediately before the output data corresponding to the start point from the output data corresponding to the end point of the curve and the start point. A value obtained by subtracting the value obtained by adding the output data and the output data corresponding to the start point from the output data that is one after the output data corresponding to the end point and having the value of 1/4 is set as the end point. It is also preferable to determine by averaging the values subtracted from the corresponding output data.

The curve is a cubic Bezier curve, and the position of the first control point in the time direction is set to the position of 1/3 of the sampling period of the output data of the digital filter from the starting point of the curve. The position of the second control point in the time direction is set to the position of 2/3 of the sampling period of the output data from the start point of the curve, and the value of the data of the first control point corresponds to the end point of the curve. The second control is determined by adding the output data corresponding to the starting point to a value obtained by subtracting 1/6 of the result obtained by subtracting the output data immediately preceding the output data corresponding to the starting point from the output data, and determining the second control. Regarding the value of the point data, the value obtained by subtracting 1/6 of the result obtained by subtracting the output data corresponding to the start point from the output data immediately after the output data corresponding to the end point corresponds to the end point. Output data It is also preferable that is determined by subtracting.

[0016]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will now be described in detail based on examples with reference to the accompanying drawings.

The asynchronous sampling frequency converter according to the first embodiment of the present invention will be described. FIG. 1 is a block diagram showing the configuration of this example. First, the configuration of this example will be described with reference to the figure. The data buffer 1 as a storage means has a sampling frequency fsi (for example, fsi = 44.1).
It holds the number of input data of (kHz) required for the filter calculation described later. The n-fold interpolation FIR filter 2 as a digital filter doubles the sampling frequency fsi by a known oversampling method and interpolates between input data by a physical operation, that is, a product-sum operation with a coefficient. .

The sampling frequency comparator 3 as the time position information generating means is composed of a phase comparator or the like and has a sampling frequency f.
The input sampling clock CLKi of si and the output sampling clock CLKo of the sampling frequency fso which is not synchronized with the input sampling clock CLKi are compared, and the curve between the input sampling timings of the data to the curve interpolation means described later is compared. The output time position information of the output sampling timing of the data from the interpolating means is calculated. That is, between the input sampling timings, in other words, the input sampling period Tf
When s_in is 1, the ratio of the interval from the input sampling timing, which is the starting point in the input sampling period Tfs_in, to the output sampling timing is obtained as time t described later, and this is output as output time position information. The input sampling period Tfs_in is Tfs_in = 1 / (n × fs
i).

The curve interpolating means 4 interpolates between n-times multiplied data input at each input sampling timing based on the output time position information by a curve interpolation algorithm using a Bezier curve, and responds at each output sampling timing. The interpolated value is output as output data.

Next, the operation of this example will be described. The input data input at the sampling frequency fsi is loaded into the data buffer 1. The data held in the data buffer 1 is sequentially sent to the n-fold interpolation FIR filter 2, and the interpolation processing corresponding to the sampling frequency of n times the sampling frequency fsi is performed by multiplying and adding the coefficients according to the filter calculation. Then, the n-fold interpolation FIR filter 2 outputs data corresponding to the sampling frequency of n × fsi.

The sampling frequency comparator 3 has a sampling frequency f
Input sampling clock CLKi of si and sampling frequency fs
The output sampling clock CLKo is compared with the output sampling clock CLKo to calculate output time position information. The curve interpolating means 4 inputs n at each input sampling timing based on the output time position information.
The data from the doubling interpolation FIR filter 2 is interpolated by a Bezier curve, and an interpolation value corresponding to each output sampling timing is output as output data.

Next, the operation of the curve interpolation means 4 using the Bezier curve will be described. 2 in each of FIGS. 2 (a) and 2 (b).
An example of a next-order Bezier curve and a third-order Bezier curve will be shown, and the concept of interpolation by the Bezier curve of this example will be described with reference to FIG. Bezier curve is one of parametric curves,
A curve from the start point to the end point is defined by changing the parameter (τ) of the quadratic and cubic polynomials from 0 to 1 using the degree-1 control point.

The curve on the time axis used in this example is the time t.
(Τ) and the value y (τ) at that time, and the start point S0 (t
0, y0), control point P1 (t1, y1), end point S1 (t
2, y2), the equation defining the quadratic Bezier curve R2 (t, y) is as follows.

[Equation 1]

Here, the equation (1) can be expressed as follows.

[Equation 2]

With reference to the time t0 of the start point, t0 = 0, the time t2 of the end point is t2 = 1, and the time t1 of the control point is t1.
= 3, the equation (3) becomes as follows.

[Equation 3]

As a result, the parameter τ and the time t can be treated as equivalent. When this condition is satisfied, the quadratic Bezier curve R2 can be treated as a function of only t, and the equation (2) can be expressed as follows.

[Equation 4]

Similarly, the cubic Bezier curve R3 has a starting point S0.
(T0, y0), control point P1 (t1, y1), control point P
2 (t2, y2) and end point S1 (t3, y3),
It looks like this:

[Equation 5]

Equation (6) can be expressed as follows.

[Equation 6]

With reference to the time t0 at the start point, t0 = 0, the time t3 at the end point is t3 = 1, and the time t1 at the control point is t1.
= 1/3 and the time t2 of the control point is t2 = 1/3,
Equation (8) is as follows.

[Equation 7]

When this condition is satisfied, the cubic Bezier curve R
3 can be treated as a function of only t, and the equation (7) is as follows.

[Equation 8]

From equations (5) and (10), respectively,
A cubic Bezier curve can be calculated, but in this example, the amount of calculation is reduced by using the algorithm of Do Castello. Interpolation by quadratic and cubic Bezier curves using the Do Castello algorithm will be described with reference to FIG. FIG. 3A shows the quadratic Bezier curve R2, and the quadratic Bezier curve R2 at time t.
When obtaining the upper point R2 (t), it is obtained as follows. First, t on the line segment from the starting point S0 to the control point P1:
A point Q1 (t) having a ratio of (1-t) and a point Q2 having a ratio of t: (1-t) on a line segment from the control point P1 to the end point S1.
Find (t). Next, when a point having a ratio of t: (1-t) on the line segment from the point Q1 (t) to the point Q2 (t) is obtained, this point becomes the point R2 (t). This calculation is realized by the calculation of the following three formulas, and the same calculation amount as when performing linear interpolation three times is required.

[Equation 9]

Similarly, a cubic Bezier curve R shown in FIG.
When the point R3 (t) at the time t on 3 is obtained, it is realized by the calculation of the following six equations, and the same calculation amount as in the case of performing the linear interpolation six times is required.

[Equation 10]

Next, how to obtain the control points will be described. As a characteristic of the quadratic and cubic Bézier curves, the vector connecting the start point and the next control point is the slope at the start point, and the vector connecting the end point and the control point immediately before it is the end point. Has become the inclination of. In this example, these slopes are approximated from the data before and after the input data to obtain the control points. This will be described by taking as an example the case where the data S1 and the data S2 are interpolated by a quadratic Bezier curve in the continuous data S0, S1, S2 and S3 of four points as shown in FIG.

As a condition that the parameter τ is equivalent to the time t, the position of the control point in the time direction is input sampling period Tf.
It is 1/2 of s_in. At this time, the slope S1 ′ of the data S1 can be approximated as follows from the difference between the previous and next data.

[Equation 11]

Similarly, the slope S2 'in the data S2 can be approximated as follows.

[Equation 12]

In the case of a quadratic Bezier curve, the control point is 1
The control point P1 is the average of the sum of the slope S1 'and the value of the data S1 and the sum of the inverted vector of the slope S2' and the value of the data S2. That is, it becomes as follows.

[Equation 13]

Similarly, when performing interpolation using a cubic Bezier curve, the positions of the control points P1 and P2 in the time direction are set to 1/3 and 2/3 of the input sampling period Tfs_in, and S1 is set according to the ratio. Adjust ', S2' to control point P1,
P2 is as follows.

[Equation 14]

When the quadratic Bezier curve is used to determine the control point in this way, only the addition / subtraction and shift are required for the calculation, and the arithmetic circuit can be simplified.
Further, even when a cubic Bezier curve is used, only an addition / subtraction is required as in the case of using a quadratic Bezier curve except that a multiplier for 1/6 operation is required. it can.

As described above, the linear interpolating means 3 of the present example uses n
Receiving the data multiplied by n by the doubling interpolation FIR filter 2, four continuous data S0, S1, S2, S3
When the control point is calculated from, for example, interpolation by the quadratic Bezier curve is performed, the control point P1 is calculated based on the above equation (22), and the data S0 and S1 and the obtained control point and control point P1 are calculated. And time t, that is, time position information from the sampling frequency comparator 3, is used to calculate the above equations (11) to (1).
The point R2 corresponding to the time t on the quadratic Bezier curve that interpolates between the data S0 and the data S1 by the calculation of 3).
The value of (t) is calculated. The calculated value is output at the next output sampling timing determined by the output sampling clock CLKo having the output sampling frequency fso. Such an operation is performed for each data input at the input sampling timing, and the data is output at the output sampling frequency fso.

The linear interpolation means 3 may use either a quadratic or a cubic Bezier curve, and in the above description, the case of using a quadratic or cubic Bezier curve is described in parallel for convenience. Needless to say, the circuit configuration actually uses either one.

Further, in the above description, the calculation processing of de castello executed by the curve interpolating means 4 is shown by software. Next, an example of a hardware configuration for realizing this calculation processing will be shown. FIG. 5 shows the above equations (11) to (13) in the curve interpolating means 4 for performing the curve interpolation by the quadratic Bezier curve.
3 is a block diagram showing a configuration of an arithmetic circuit that performs the arithmetic operation of FIG.
The arithmetic circuit includes six selectors SEL3 to SEL6, four registers REG1 to REG4, two adders ADD.
1, ADD2, multiplier MUL and inverter INV.

The selector SEL1 is a data input terminal DDA.
TA-IN, output terminal of register REG1, register R
The output terminal of EG4 is selectively connected to the input terminal of register REG1. The selector SEL2 is a data input terminal DA
TA-IN, output terminal of register REG2, register R
The output terminal of EG4 is selectively connected to the input terminal of register REG2. The selector SEL3 is a data input terminal DA
The TA-IN and the output terminal of the register REG3 are selectively connected to the input terminal of the register REG3. Selector SEL
4 selectively connects the main output terminals of the registers REG1 and REG2 to one input terminal of the multiplier MUL. The selector SEL5 has an output terminal of the register REG3 and an adder AD.
The main output terminal of D1 is selectively connected to the other input terminal of the multiplier MUL. The multiplier MUL multiplies the input data to the two input terminals and outputs the result.

The inverter INV inverts the data output from the register REG3. The adder ADD1 has "1.
The data indicating "000 ..." And the output data of the inverter INV are input. The digit "1.000 ..." matches the digit of the time position information indicating the time t (0 <t <1). The adder ADD1 adds the data indicating "1.00 ..." And the output data of the inverter INV, and adds the carry to the least significant bit to realize the operation (1-t). One input terminal of ADD2 is a multiplier MU
L is connected to the output terminal and the other input terminal is the selector S
It is connected to the output terminal of EL6, adds the input data to these input terminals, and outputs the result to register REG4. The selector SEL6 selectively connects the output terminal of the register REG4 and the terminal for supplying data of all digits of "0" to the other input terminal of the adder ADD2. The output terminal of the register REG4 is an output terminal of data indicating the value of the point R2 (t) corresponding to the time t on the quadratic Bezier curve.

Next, the operation of the arithmetic circuit of FIG. 5 will be described with reference to FIG.
The operation sequence is shown in FIG. 5 and will be described with reference to FIGS. In the “No. operation” on the left side of FIG. 1 to 8 show the operation of storing data in the register, which is performed every clock of 1 to 8 clocks. For example, No. 1 “REG3 <= D_IN
“(T)” indicates that “REG3” indicates the code of the target register, “<=” indicates that the data on the right is stored in the register, and “D_IN (t)” indicates the input terminal DATA_IN. , And the code in “()” indicates the content of the data.
Shows that data indicating time t is stored from DATA_IN. No. 2 “REG1 <= D_IN
(S0) "indicates that the data S0 from DATA_IN is stored in the register REG1." REG2 <= D_IN (P1), REG4 <= (1-
t) × REG1 (S0) ”is stored in the register REG2 by DA
Stores data P1 from TA_IN and registers REG
4, the data S0 stored in the register REG1 and (1
-T) is shown and the result of multiplication is stored.
Hereinafter, No. Similarly, data storage operations up to 8 are shown.

In the center of FIG. 6, "(Selector Co
"control)" indicates the state of each clock of the selectors SEL1 to SEL6 corresponding to No. 1 to No. 8 on the left. "Hold" indicates that the output of the register is fed back and fixed. "D-IN" is the input terminal DATA
_IN is selected, “REG1” and the like indicate the code of the corresponding register, the register is selected, and “()” indicates the data stored in the register. Regarding the selector SEL5, No.
In 1 clock of 1, the register REG3 is selected as “(REG3)”, and No. In 2 clocks of 2, it indicates that the data t stored in the register REG3 as "t" is selected and the register REG3 is selected.
No. In 3 clocks of 3, the adder A is set as "1-t".
Adder ADD1 showing data (1-t) from DD1
Is selected, and so on. Also,
Regarding the selector SEL6, "(0)" indicates a state in which a terminal supplying data having all digits of "0" is selected.

"Status_after" on the right side of FIG.
"_Clocked" indicates the data content stored for each clock of the registers REG1 to REG4 corresponding to Nos. 1 to 8 on the left. Note that "undefined" indicates that the data is undefined. .

Next, the operation of 1 to 8 clocks will be described in order.
In one clock, the selectors SEL1 and SEL2 are set to “Ho
ld ”, and the selector SEL3 is connected to the input terminal DATA_
IN is selected, selector SEL4 is set to register REG1, and selector SEL5 is set to
With the register REG3 selected, the selector SEL
6 is a state in which a terminal for supplying data having all digits of "0" is selected. As a result, the data t indicating the time t from the input terminal DATA_IN is stored in the register REG3.

At 2 clocks, the selector SEL1 is set to the state in which the input terminal DATA_IN is selected, and the selector S
Set EL2 and SEL3 to "Hold", and select SEL
4 is in a state in which the register REG1 is selected, the selector SEL5 is in a state in which the register REG3 is selected, and the selector SEL6 is in a state in which terminals for supplying data of all digits of "0" are selected. This allows register R
Data S0 indicating the value of the starting point S0 from the input terminal DATA_IN is stored in EG1.

At 3 clocks, the selector SEL1 is set to "H".
and the selector SEL2 is connected to the input terminal DATA.
With _IN selected, the selector SEL3 is set to “Ho
ld ", the selector SEL4 is in a state in which the register REG1 is selected, and the selector SEL5 is an adder ADD.
1 is selected, and the selector SEL6 is set to a state in which a terminal supplying data with all digits of "0" is selected. As a result, the register REG2 has an input terminal DAT.
Data P1 indicating the value of control P1 from A_IN is stored. In addition, the register REG via the selector SEL4
1 to the data S0 and the data (1-t) from the adder ADD1 via the selector SEL5 are multiplied by the multiplier MUL, and the result is stored in the register REG4 via the adder ADD2.

At 4 clocks, the selector SEL1 is set to the state in which the input terminal DATA_IN is selected, and the selector S
Set EL2 and SEL3 to "Hold", and select SEL
4, the register REG2 is selected, the selector SEL5 is set to the register REG3, and the selector SEL6 is set to the register REG4. As a result, the input terminal DA is connected to the register REG1.
Data S1 indicating the value of the end point S1 from TA_IN is stored. In addition, the register RE via the selector SEL4
The data P1 from G2 and the data t from the register REG3 via the selector SEL5 are multiplied by the multiplier MUL. The resulting tP1 is the adder ADD
2, the register REG4 via the selector SEL6
From the data (1-t) S0. That is,
The data Q1 of the point Q1 (t) obtained by the above equation (11) is obtained. The data Q1 is stored in the register REG4.

At 5 clocks, the selector SEL1 is set to "H".
old ”and the selector SEL2 is set to the register REG4.
Is selected, and the selector SEL3 is set to “Hold”.
Then, the selector SEL4 is set to the state in which the register REG2 is selected, the selector SEL5 is set to the state in which the adder ADD1 is selected, and the selector SEL6 is set to the state in which all the terminals for supplying data of "0" are selected. To do.
As a result, the data Q1 stored in the register REG4 is stored in the register REG2. Also, the selector S
The data P1 stored in the register REG2 via EL4 and the data (1-t) from the adder ADD1 via the selector SEL5 are multiplied by the multiplier MUL, and the result is passed via the adder ADD2. Register RE
It is stored in G4.

At 6 clocks, the selectors SEL1 and SE
L2 and SEL3 are set to "Hold", and the selector SEL4
To the state in which the register REG1 is selected, and the selector S
EL5 is in a state in which the register REG3 is selected, and selector SEL6 is in a state in which the register REG4 is selected. This causes the register R to pass through the selector SEL4.
The data S1 stored in EG1 and the selector SEL
The data t from the register REG3 via 5 is multiplied by the multiplier MUL, and the result is the data tS.
1 and the data (1-t) P1 from the register REG4 via the selector SEL4 are added by the adder ADD2. That is, the data Q2 at the point Q2 (t) in the above equation (12) is obtained. This data Q2 is stored in the register RE
It is stored in G4.

At 7 clocks, the selector SEL1 is set to the state in which the register REG4 is selected, and the selector SEL4 is selected.
2, SEL3 is set to "Hold", the selector 4 is set to the state in which the register REG2 is selected, and the selector SEL5
To the state where the adder ADD1 is selected, and the selector SE
L6 is in a state in which a terminal for supplying data having all digits of "0" is selected. As a result, the data Q2 from the register REG4 is stored in the register REG1. In addition, the data Q1 from the register REG2 via the selector 4
And the data (1-t) from the adder ADD1 via the selector SEL5 are multiplied by the multiplier MUL,
The resulting data (1-t) Q1 is stored in the register REG4 via the adder ADD2.

At 8 clocks, the selector SEL1 is set to the state in which the input terminal DATA_IN is selected, and the selector SE
L2 and SEL3 are set to "Hold", and the selector SEL4
To the state in which the register REG1 is selected, and the selector S
EL5 is in a state in which the register REG3 is selected, and selector SEL6 is in a state in which the register REG4 is selected. As a result, the register R via the selector SEL4
The data Q2 from EG2 and the data t from the register REG3 via the selector SEL5 are multiplied by the multiplier MUL. The resulting data tQ2 is added to the data (1-t) Q1 from the register REG4 via the selector SEL6 by the adder ADD2.
That is, the data R0 of the point R2 (t) obtained by the above equation (13) is obtained. The data R0 is stored in the register REG4. This data is output at the next output sampling timing.

Interpolation by the quadratic Bezier curve is repeatedly performed by repeating the above operation of 1 to 8 clocks. Further, if the next start point S0 and time t are stored in the registers REG1 and REG3 at the above-mentioned 8 clocks, the data R0 at the point R2 (t) can be obtained by the arithmetic processing of 6 clocks thereafter. As described above, by using the Do Castello algorithm, it is possible to suppress the decrease in the processing speed and the increase in the circuit scale due to the curve interpolation as much as possible.

As described above, in the asynchronous sampling frequency conversion apparatus of this example, the sampling frequency comparator 3 causes the input sampling clock CLKi of the sampling frequency fsi and the output sampling clock CLKo of the sampling frequency fso to be obtained. Since the output time position information is compared and the curve interpolation means 4 performs the Bezier curve interpolation based on the output time position information, the Bezier curve interpolation is performed even when the input sampling frequency and the output sampling frequency are asynchronous. And sampling frequency conversion with good distortion accuracy can be realized. Since the Bezier curve interpolation is performed by using the Do Castello algorithm, it is possible to reduce the calculation amount of the curve interpolation calculation and suppress the decrease in processing speed and the increase in circuit scale as much as possible. For example, in the case of a quadratic Bezier curve, interpolation calculation can be performed with a calculation amount that is only three times that of linear interpolation. In addition to this, since the sampling frequency comparator 3 serves as a supply source of parameters used in the curve interpolation calculation, a separate parameter supply source is not required, and an increase in circuit scale can be suppressed.

In this example, since the distortion error can be reduced by performing the curve interpolation as compared with the conventional linear interpolation, the oversampling magnification needs to be about 256 times in the linear interpolation, but the same accuracy is obtained. In order to realize this, in the case of this example, about 16 to 32 times is sufficient, and the magnification is 1/8.
It becomes possible to suppress from 1/16. N by it
It is possible to reduce the filter coefficient of the double interpolation FIR filter and the circuit scale.

Next, an asynchronous sampling frequency converter according to the second embodiment of the present invention will be described. In the first embodiment, the Bezier curve interpolation is performed using the Do Castello algorithm, but the present invention is not limited to this. The Bezier curve interpolation may be performed using the midpoint division method described below, which is used in this example. In the midpoint division method, as shown in FIG. 7, the midpoint R1 between the midpoint Q1 of the line segment between the start point S0 and the control point P1 and the midpoint Q2 of the line segment between the control point P1 and the end point S1 is a quadratic Bezier. Located on the curve, midpoint R
It focuses on the point that 1 can be divided into two different quadratic Bezier curves. Of the two quadratic Bezier curves obtained by division, the position in the time direction of the midpoint R1 sequentially generated by repeating the midpoint division on the side including the time t is brought closer to the time t, The point on the quadratic Bezier curve at time t is approximately obtained. FIG. 8 is a block diagram showing the configuration of an arithmetic circuit that obtains a point R2 (t) on a quadratic Bezier curve using the midpoint division method. The other structure is the same as that of the first embodiment. This is a control circuit ctl, selectors sel1, sel2, s
el3, registers reg1, reg2, reg3, shift registers sr1, sr2, sr3, adder add1,
It consists of add2 and add3.

The control circuit ctl selects the selectors sel1, sel2, sel according to the time t indicated by the output time position information.
It controls which of the two input terminals L and H of 3 the input data to which is valid.

The selectors sel1, sel2, and sel3 register the valid input data in the registers reg1 and reg1, respectively.
Output to reg2 and reg3. In the selector sel1, the input terminal L is connected to the output terminal of the register reg1 and the output terminal H is connected to the output terminal of the shift register sr3. In the selector sel2, the input terminal L is connected to the output terminal of the shift register sr1 and the input terminal H
Is connected to the output terminal of the shift register sr2.
In the selector sel3, the input terminal L is the shift register s
It is connected to the output terminal of r3 and the input terminal H is connected to the register re.
It is connected to the output terminal of g3.

The adder add1 has registers reg1 and re
The output data from g2 is added, and the adder add2 adds the output data from the registers reg2 and reg3.
The shift registers sr1 and sr2 are register re
Upon receiving the output data from g2 and reg3, the data is shifted to the lower bit side and the value is halved. Adder a
dd3 adds the output data from the shift registers sr1 and sr2. The shift register sr3 is an adder add3
Receiving the output data from, the data is shifted to the lower bit side and the value is halved. Also, the shift register s
The data of the point on the quadratic Bezier curve to be obtained is output from the output terminal of r3.

Next, the operation of the arithmetic circuit of this example will be described with reference to FIGS. Here, for convenience, the time t as the output time position information is the input sampling period Tfs described above.
_In is 1 and divided into 8 equal periods 1/8
~ 7/8. First, the time t is the input sampling period Tf
A case will be described in which the data of the point R1 on the quadratic Bezier curve at the midpoint of s_in, that is, when t = 4/8 is obtained.

First, the registers reg1, reg2, re
The data S0 at the start point S0, the data P1 at the control point P1, and the data S1 at the end point S1 are given to g3 as initial values. The data S0 and the data P1 are added by the adder add1, the addition result is output to the shift register sr1, and the shift register sr1 outputs 1/2 of the addition result.
The value of is obtained as output data. This output data is
It is the value of the midpoint Q1 of the line segment between the starting point S0 and the control point P1.
The data P1 and the data S1 are added by the adder add2, the addition result is output to the shift register sr2, and the shift register sr2 obtains 1/2 of the addition result as output data. This output data becomes the value of the midpoint Q2 of the line segment between the control point P1 and the end point S1. The output data of the shift registers sr1 and sr2 are added by the adder add3, and the addition result is the shift register s
It is output to r3, and the shift register sr3 obtains 1/2 of the addition result as output data. here,
Half the addition result of the data Q1 and Q2 of the middle point Q1 and Q2
The obtained output data is the value of the midpoint R1 of the points Q1 and Q2, and the point R1 on the quadratic Bezier curve to be obtained is
Becomes the value of.

Next, a case will be described in which the data of the point R2 on the quadratic Bezier curve when the time t is t = 6/8 is obtained. In FIG. 7, the value of the point R2 is calculated by applying the above-mentioned method to the quadratic Bezier curve on the right side including the point R2 among the two quadratic Bezier curves divided into two by the point R1. To ask.

First, the point R1 is obtained by the above-mentioned first midpoint division calculation. Next, when performing the second midpoint division operation, when the time t as the time position information is t = 6/8, the control circuit ctl selects the selectors sel1 and sel.
2. Make input data to the input terminal H valid in sel3. As a result, the data R1 that is the value of the point R1 previously obtained is input to the selector sel1, and the data R1 is output to the register reg1 via the selector sel1. The previously obtained data Q2 is input to the selector sel2, and the data Q2 is output to the register reg2 via the selector sel2. The data S1 of the register reg3 is input to the selector sel3, and the data S1 is output to the register reg3 via the selector sel3. The adder add1 adds the data R1 and the data Q2, and the shift register sr1 adds 1 to the addition result.
Data Q3 having a value of / 2 is obtained. This data Q3 is
The value of the midpoint Q3 of the line segment between the points R1 and Q2 is shown. The data Q2 and the data S1 are added by the adder add2, and 1 / of the addition result is obtained by the shift register sr2.
Data Q4 having a value of 2 is obtained. This data Q4 indicates the value of the midpoint Q4 of the line segment between the point Q2 and the point S1. Adder a
The data Q3 and the data Q4 are added by dd3,
By the shift register sr3, the data R2 having a value of 1/2 of this addition result is obtained. Data R2 are points Q3 and Q4
It becomes the value of the midpoint R2, and becomes the value of the point R2 on the quadratic Bezier curve to be obtained.

When the data of the point R3 on the quadratic Bezier curve when the time t is t = 7/8 is obtained, the point R is calculated by the first and second midpoint division operations by the above operation.
After obtaining the values of 1 and R2, in the third midpoint division calculation, the control circuit ctl has the selectors sel1, sel2,
The value of the point R3 is obtained by validating the input data to the input terminal H of sel3 and performing the above-mentioned midpoint division calculation.

Further, when the data of the point R4 on the quadratic Bezier curve at the time t = 5/8 is obtained, the points R1 and R1 are calculated by the above-described operation by the first and second midpoint division operations. After obtaining the value of R2, the control circuit ctl determines the selectors sel1 and sel for the third midpoint division calculation.
2, valid the input data to the input terminal L of sel3,
The value of the point R4 is obtained by performing the above-mentioned midpoint division calculation.

That is, the above-mentioned input sampling period Tfs_
The point of the quadratic Bezier curve at the time when in is divided into 2 ⁿ (n is an integer from 1 to 3 in FIG. 7, but is not limited to this and may be any integer of 1 or more). The value is obtained by performing the above midpoint division operation n times. If the value of the time t of the point obtained at the n + 1-th time is larger than the value of the time t of the point obtained at the n-th time when moving from the n-th to the n + 1-th midpoint division calculation, the control circuit ctl selects the selector sel1. , Sel2, sel3, the input data to the input terminals H of the sel, sel2, and sel3 are valid, and if the value of the time t of the point obtained in the (n + 1) th time is smaller than the value of the time t of the point obtained in the nth time, the control circuit ctl determines that the selector sel1, sel2, s
The input data to the input terminal H of el3 is validated. Such a control process can be realized as follows, for example, by indicating the time t with a binary number from 1 to 111.

The setting of the selector after the first midpoint division calculation compares the time t with 100, and if they match, the control circuit c
tl is the data of the shift register sr3 obtained by the first midpoint division calculation, and ends the arithmetic operation. If they do not match and the most significant bit is 1, the input terminals of the selectors sel1, sel2, and sel3. If the input data to H is valid and the most significant bit is 0, the input data to the input terminals L of the selectors sel1, sel2, and sel3 are valid and the second midpoint calculation is performed. The setting of the selector after the second point division calculation is performed as follows. First, the entire time t is shifted by 1 bit to the upper bit side, the most significant bit immediately before this shift is truncated, and 0 is given to the least significant bit. This is compared with 100, and if they match, the control circuit ctl ends the arithmetic operation with the data of the shift register sr3 obtained by the second midpoint division calculation as the data to be obtained. Do not match here,
When the most significant bit is 1, selectors sel1 and sel
2, valid the input data to the input terminal H of sel3,
In the case of 0, the input data to the input terminals L of the selectors sel1, sel2, and sel3 are valid, and the third midpoint calculation is performed. If the digit of the time t increases, the same procedure may be taken. As described above, the control circuit ctl can be realized by a simple circuit configuration including the comparison circuit and the shift register.

As described above, in this example, the time t as the output time position information from the sampling frequency comparator 3 is determined in the form of 2 ⁿ , and the arithmetic circuit is used by using the midpoint division method. Can be composed of three selectors, a control circuit for switching the selectors, three adders, and three shift registers. In other words, this example does not require a multiplier that occupies a relatively large circuit area, has the same operation and effect as those of the first embodiment, and can further reduce the circuit scale.

Although the Bezier curve is not particularly weighted in each of the above embodiments, the present invention is not limited to this. For example, weighting may be given to each point used in the curve interpolation calculation, and interpolation may be performed using a rational Bezier curve. In this case, the division calculation between each point is as follows.

[Equation 15]

Here, W (i) is the weighting in P (i), and P (i) is i (i = 0,
(1, 2 ...) The second point, for example, if applied to the equation (11), P (0) = S0, P (1) = P1, P
(2) = Q1 (t) can be set. By performing appropriate weighting, it is possible to give a desired effect to the tone color of the reproduced audio signal.

Further, in each of the above embodiments, the control point of the Bezier curve is obtained as a result of multiplying the gradient of the start point and the end point, which are obtained approximately, by a coefficient, and adding this to the start point and the end point.
It is not limited to this. For example, in the case of a quadratic Bezier curve, the control points may be intersections of straight lines extending from the start point and the end point along the inclination, instead of being multiplied by the coefficient. In this case, the equivalence between the time t and the parameter τ is broken, but this deviation does not pose a problem if the parameter τ is calculated from the time t.

[0074]

According to the present invention, the output time position information generating means compares the sampling frequency of the input to the curve interpolating means with the sampling frequency of the output from the curve interpolating means to determine the output time position. Since the information is generated and the curve interpolating means performs the curve interpolation based on the output time position information, the curve interpolation can be realized even when the sampling frequency is asynchronous between the input and the output, and the sampling frequency with high distortion accuracy can be realized. The conversion can be realized.

Particularly in a sampling frequency converter using a digital filter for oversampling, if it is applied to curve interpolation between output data of the digital filter, the distortion error is improved by the curve interpolation. If an accuracy similar to that of linear interpolation is obtained, the oversampling ratio can be suppressed as compared with the conventional one, and the filter coefficient of the digital filter and the circuit scale can be reduced.

Further, since the output time position information generating means is the interpolation parameter supply device, there is no need to provide a separate interpolation parameter supply device, and from this point also the circuit scale can be effectively reduced. Becomes

Also, using a Bezier curve for curve interpolation,
By performing the curve interpolation calculation based on the Castello algorithm, it is possible to reduce the calculation amount of the curve interpolation calculation, and to suppress the decrease in processing speed and the increase in circuit scale as much as possible.

Also, curve interpolation is performed using a quadratic Bezier curve with the output time position information as a unit of 1/2 ⁿ (n is an integer of 1 or more) of the sampling cycle of the input data, Depending on the point division method, a multiplier is not required, and curve interpolation can be realized with a simple circuit configuration such as a selector, an adder, and a level shifter, and the circuit scale can be effectively reduced. .

Further, by performing the curve interpolation, data processing is also performed such that each point used in the curve interpolation calculation is given an appropriate weighting to give a desired effect to the tone color of the reproduced audio signal. It will be possible.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an asynchronous sampling frequency conversion device according to a first embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a Bezier curve used in a curve interpolation calculation of the asynchronous sampling frequency conversion device in FIG.

FIG. 3 is an explanatory diagram showing an algorithm of de Castello used in the curve interpolation calculation of the asynchronous sampling frequency conversion device of FIG. 1.

FIG. 4 is an explanatory diagram showing control points of a Bezier curve used in a curve interpolation calculation of the asynchronous sampling frequency conversion device of FIG.

5 is a block diagram showing a configuration of an arithmetic circuit that executes a curve interpolation operation of the asynchronous sampling frequency conversion device of FIG.

FIG. 6 is an explanatory diagram showing an operation sequence of the arithmetic circuit of FIG.

FIG. 7 is an explanatory diagram showing an algorithm of the midpoint division method used for the curve interpolation calculation of the asynchronous sampling frequency conversion apparatus according to the second embodiment of the present invention.

FIG. 8 is a block diagram showing a configuration of an arithmetic circuit that executes a curve interpolation operation of the asynchronous sampling frequency conversion apparatus according to the second embodiment of the present invention.

FIG. 9 is a block diagram showing the configuration of a conventional sampling frequency conversion device.

[Explanation of symbols]

1 Storage means (data buffer) 2 Digital filter (n-fold interpolation FIR filter) 3 Time position information generating means (sampling frequency comparator) 4 Curve interpolation means

Claims (12)

[Claims] 1. Inputting data at a first sampling frequency,
A sampling frequency conversion device that outputs data at a second sampling frequency, wherein the first sampling frequency and the second sampling frequency are compared, and the output data with respect to the time position of the input data. The time position information generating means for generating the output time position information indicating the time position and the preceding and following input data are curvedly interpolated, and based on the output time position information, correspond to the output time position information on the curve. A sampling frequency conversion device, comprising: a curve interpolating unit that calculates a value and outputs the value as the output data at the second sampling frequency. 2. The sampling frequency conversion device according to claim 1, wherein the curve is a Bezier curve, and the curve interpolating means performs the calculation based on the algorithm of De Castello. 3. The curve is a Bezier curve, the curve interpolating means performs the calculation based on the algorithm of de Castello, and weights each point used in the calculation. Item 1. The sampling frequency converter according to item 1. 4. The curve is a quadratic Bezier curve, and the output time position information is 1 of a sampling cycle of the input data.
2. The sampling frequency conversion device according to claim 1, wherein the unit is set to / 2 ⁿ (n is an integer of 1 or more), and the curve interpolation means performs the calculation based on a midpoint division method. 5. The curve is a quadratic Bezier curve, and the position of the control point in the time direction is set to the position of 1/2 of the sampling cycle of the input data from the starting point of the curve, and the control point data The value is 1 of the above input data corresponding to the above start point.
The slope of the starting point of the curve that is approximately obtained from the previous input data and the input data corresponding to the end point of the curve, and the input data after one of the input data corresponding to the end point. And the slope of the end point of the curve that is approximately obtained from the input data corresponding to the start point, the input data corresponding to the start point, and the data corresponding to the end point. The sampling frequency conversion device according to claim 1, wherein the sampling frequency conversion device is a sampling frequency conversion device. 6. The curve is a cubic Bezier curve, the first curve
The position of the control point in the time direction from the starting point of the curve is 1/3 of the sampling period of the input data,
Of the control point in the time direction from the start point of the Bezier curve to the position of 2/3 of the sampling period of the input data, and the value of the data of the first control point is the input corresponding to the start point. It is determined from the slope of the starting point of the curve approximately obtained from the input data immediately before the data and the input data corresponding to the end point of the curve, and the input data corresponding to the starting point. The value of the control point data is the slope of the end point of the curve approximately obtained from the input data one after the input data corresponding to the end point and the input data corresponding to the start point. The sampling frequency conversion device according to any one of claims 1 to 3, wherein the sampling frequency conversion device is determined from the input data corresponding to the end point. 7. Inputting data at a first sampling frequency,
A sampling frequency conversion device for outputting data at a second sampling frequency, comprising storage means for inputting data at the first sampling frequency and storing a predetermined number of the input data, and storage means for storing the input data in the storage means. The extracted data, and
The digital filter that interpolates by multiplying the sampling frequency of n by n and the first sampling frequency and the second sampling frequency are compared, and the final output with respect to the time position of the output data from the digital filter is compared. The time position information generating means for generating the output time position information indicating the time position of the data and the output data from the digital filter are curvedly interpolated, and the output time position on the curve is based on the output time position information. A sampling frequency conversion device, comprising: a curve interpolating unit that calculates a value corresponding to information and outputs the value as the final output data at the second sampling frequency. 8. The sampling frequency conversion device according to claim 7, wherein the curve is a Bezier curve, and the curve interpolating means performs the calculation based on the algorithm of De Castello. 9. The curve is a Bezier curve, and the curve interpolating means performs the calculation based on the algorithm of De Castello, and weights each point used in the calculation. Item 7. The sampling frequency converter according to item 7. 10. The curve is a quadratic Bezier curve, and the output time position information is in units of 1/2 ⁿ (n is an integer of 1 or more) of a sampling period of the output data from the digital filter. The sampling frequency conversion device according to claim 7, wherein the curve interpolation means performs the calculation based on a midpoint division method. 11. The curve is a quadratic Bezier curve, and the position of the control point in the time direction is set to 1/2 of the sampling period of the output data from the digital filter from the starting point of the curve.
And the value of the data of the control point is 1/4 of the result obtained by subtracting the output data immediately before the output data corresponding to the start point from the output data corresponding to the end point of the curve. Value obtained by adding the output data corresponding to the start point and the output data corresponding to the end point to the output data immediately after the output data corresponding to the start point, and subtracting 1 / The sampling frequency conversion device according to any one of claims 7 to 10, wherein a value obtained by subtracting 4 from the output data corresponding to the end point is averaged. 12. The curve is a cubic Bezier curve, and the position of the first control point in the time direction is set to 1 / s of the sampling period of the output data of the digital filter from the start point of the curve.
3 position, and the position of the second control point in the time direction is 2 / of the sampling cycle of the output data from the start point of the curve.
The value of the data of the first control point is obtained by subtracting the output data immediately before the output data corresponding to the start point from the output data corresponding to the end point of the curve. It is determined by adding the output data corresponding to the start point to a value obtained by ⅙, and the value of the data at the second control point is 1 of the output data corresponding to the end point.
10. The value obtained by subtracting 1/6 of the result obtained by subtracting the output data corresponding to the start point from the subsequent output data is determined by subtracting from the output data corresponding to the end point. The sampling frequency conversion device according to any one of 1.