JPH03137871A - Variable speed reproducing device - Google Patents

Abstract

PURPOSE:To obtain the subject device, such as a digital tape recorder, etc., of high sound quality without waveform distortion by constituting a noncyclic type digital filter using a coefft. corresponding to each double speed. CONSTITUTION:An input PCM signal 1 is added to the noncyclic type digital LPF 4, and is outputted as an output PCM signal 5. When a sampling rate ratio N/M 15 of the signals 1 and 5 is added via a rate ratio input means 14 to a control means 12, a coefft. selecting signal 2 is sent to a coefft. thinning-out cyclic means 6 by the means 12 based on the rate ratio 15, and an address signal 3 for specifying a coefft. is sent to a coefft. storage means 8. A coefft. 7 is generated by the means 8 from a coefft. table stored in advance and sent to the LPF 4. Convolution calculation of a cyclic K tap coefft. and an input PCM sample is executed by the LPF 4 according to a command of the means 12. By this method, at the time of low speed calculation, high-order interpolation can be carried out, and reproducing can be performed in superior sound quality.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a recording/reproducing apparatus such as a digital tape recorder.

2. Description of the Related Art Conventionally, when editing a tape recorder, a method is generally used in which the start of the recorded audio is checked by low-speed playback and the editing point is changed. However, in digital audio playback devices such as digital tape recorders, the sampling frequency (hereinafter referred to as sampling frequency) is 48kHz or 44.1kHz.
Alternatively, 32 kHz or the like is used to input and output PCM samples at each sampling frequency. Therefore, in the case of low-speed reproduction, the sampling rate for reproduction becomes low (low frequency components are reproduced and become noise components). Conventional examples for solving this problem are shown in FIGS. 6 and 7.

FIG. 6 is a block diagram showing the configuration of a device that changes the playback speed of PCM audio according to the rotation speed of the dial;
The figure is a waveform diagram showing the relationship between the PCM signal and the timing signal. Let the input PCM signal levels be A and B,
If D is a constant, the interpolated output C is: C:A+ (B-A)・D@j...(1)
However, j is expressed as 1 to N.

In FIG. 6, the input PCM signal 1 is latched by the latch means 52 by a rotation pulse 57 corresponding to the rotation speed of the dial, and the output 53 of the latch means 52 is applied to the latch means 54 and latched by the rotation pulse S7. The output 53.55 of the latch means 52.54 provides an input PCM signal offset by one period of the rotation pulse 57. The output 53.55 of the latch means 52.54 is applied to the subtraction means 56, and the difference signal 58 obtained by the subtraction is applied to the multiplication means 5e. This difference signal 58 corresponds to (B-A) in equation (1). On the other hand, in the counter 80, the rotation pulse 5
7 period 73 (see FIG. 7) is counted by the mask clock 72, and the output 62 of the counter 6o is stored in the ROM 8.
Added to 3. ROM 83 generates information 64 that is inversely proportional to period 73. Note that the information θ4 corresponds to D in equation (1). This information 64 is added to addition means 74. Adding means 74 generates a signal that is j times the information 64 (j is a positive integer).

The output 85 of the addition means 74 is applied to the multiplication means 59 and multiplied by the difference signal 58. By this operation, the difference for linearly interpolating the input PCM signal, that is, (B
-A)・Dllj is obtained. The output 68 of the multiplication means 59 is added to the addition means 67 together with the output 55 of the latch means 54, that is, A in equation (1), to obtain an output PCM signal 68 corresponding to C in equation (1). As can be seen from the above explanation, linear interpolation is performed in the conventional example.

The input PCM samples during low-speed playback are 16a to 1 in Figure 7.
8c, and the input signal waveform is 17.

As is clear from the above explanation, the interpolated waveform is represented by the straight line 71, and the interpolated samples are each point on the interpolated waveform 71, that is, 7
2a to 72o. It can be seen that clearly large waveform distortion occurs.

Problems to be Solved by the Invention However, in the configuration of the conventional example described above, since interpolated samples are obtained by linear interpolation from the reproduced signal during low-speed reproduction, the interpolated waveform 71 is significantly different from the input signal waveform 17, resulting in large waveform distortion. There was a problem with poor sound quality.

In view of the above problems, the present invention constructs an acyclic digital filter using coefficients corresponding to each double speed.
The purpose of the present invention is to provide a variable speed playback device with high sound quality and little waveform distortion.

Means for Solving the Problems In order to achieve the above objects, the variable speed playback device of the present invention includes an acyclic digital low-pass filter with K (K is a positive integer) taps, a sampling rate of output PCM samples, and an input PCM sample. Means for inputting the ratio N/M (N, M are positive integers, N = 1.2, ..., M) of the sampling rate of the PCM sample, the number of taps of the acyclic digital low-pass filter, and the sampling rate. The denominator M of the ratio N/M
A means for storing a group of coefficients of an acyclic digital low-pass filter with M as the number of taps; In addition to forming partial coefficient groups thinned out in stages, the partial coefficient groups are divided into (N-1
) a coefficient thinning-circulating means for circulating every other coefficient; and a first control means for controlling an acyclic digital low-pass filter and a coefficient thinning-circulating means, the acyclic digital low-pass filter A convolution operation is performed between the coefficients of and the input PCM samples.

Further, in place of the first control means, a second control means is used which controls the acyclic digital low-pass filter and the coefficient thinning cyclic means after receiving an interrupt signal, and the sampling rate of the output PCM samples is kept constant. At the same time, an interrupt control means is provided that interrupts the second control means at each sampling frequency of the output PCM sample, and for each interrupt, the acyclic digital low-pass filter convolves the coefficient of the circulated tap with the input PCM sample. It performs intensive calculations and outputs PCM samples.

Operation The present invention has the above-described configuration, so that the input P during variable speed playback is
When the sampling rate of the CM sample is lower than the sampling rate of the output PCM sample, that is, during low-speed playback, tap the product K·M of the number of taps of the acyclic digital low-pass filter and the denominator M of the sampling rate ratio N/M. It is possible to perform high-order interpolation by convolving the input PCM samples with a group of partial coefficients thinned out by (M-1) from a group of coefficients of an acyclic digital low-pass filter.

Also, it is possible to vary the coefficients of the acyclic digital low-pass filter by changing the number of partial coefficient groups to the number of taps based on the sampling rate ratio N/M and cycling every (N-1) coefficients among the M coefficient groups. It is possible.

Furthermore, by performing the convolution operation in response to an interrupt request for an output PCM sample, it is possible to easily manage the time for the convolution operation.

Embodiment FIG. 1 is a block diagram showing the configuration of a variable speed playback apparatus in a first embodiment of the present invention.

In FIG. 1, an input PCM signal 1 is applied to an acyclic digital low-pass filter (convolution calculation means) 4, where a convolution operation with a coefficient 7 of the acyclic digital low-pass filter 4 is performed. , an output signal 5 is obtained. The ratio of the sampling rate of the input PCM signal applied from the outside to the sampling rate of the output PCM signal is N.
/M.

However, since it is a low-speed playback, N=1.2. ...9M
and the output sampling rate is higher than the input sampling rate. Here - as an example 3/1
The case of 5x speed will be explained. In this case, rate ratio 1
5 (N/M=3/15) is applied to the first control means 12 via the rate ratio input means 14. Coefficient 7 of the acyclic digital low-pass filter 4 is necessary for the acyclic digital low-pass filter 4 to constitute a low-pass filter and perform an interpolation function. The coefficient 7 used for this purpose is generated as follows. Coefficient 7 is stored in coefficient storage means 8. The coefficient storage means 8 is constituted by, for example, a read-only memory (hereinafter referred to as ROM). The first control means 12 sends the first coefficient selection signal 2 to the coefficient thinning circuit means θ based on the rate ratio 15, and the coefficient thinning circuit means 6 sends an address signal 3 specifying the coefficient to be selected to the coefficient storage means 8. send. The coefficient storage means 8 sends the coefficient 7 to the acyclic digital low-pass filter 4. Acyclic digital low-pass filter 4
executes a convolution operation according to a command from the first control means 12.

Input PCM samples as X (k) (k = O−km
ax), and the coefficient is H(k) (k = 0 ~ klIl
ax), the output Y(i) is expressed by the convolution operation shown in the following equation.

Y(i)=H(k)・X(k)
=(2) Equation (2) is executed by the acyclic digital low-pass filter 4.

Now, let us consider the number of taps of the acyclic digital low-pass filter 4. Generally, the signal-to-noise ratio (hereinafter referred to as S/N) after passing through a digital filter depends on the number of taps. For example, if the S/N is 80dB,
It is known from simulations that several tens of taps are required. Here, the number of taps is set to 60.

Now, let's consider the number of taps when configuring an interpolation filter. When the minimum playback speed is 1/15x,
Fourteen interpolated samples need to be generated from one input PCM sample. Therefore, the number of coefficients required is 60 taps x 15 = 900 taps. That is, a coefficient of an acyclic digital low-pass filter of 900 taps is required. This coefficient table is shown in Table 1.

In Table 1, the vertical direction represents the number of sides of the table, which in this case is 16 with i=0 to 14. The horizontal direction is the number of taps of the acyclic digital low-pass filter 4, which in this case is 60 taps from 0 to 59. Also, H(0)~
H (899) represents each coefficient. The partial coefficient group used at each double speed is M-1=
Obtained by thinning out by 14.

Table 1 Tap order → 0 - -59 That is, H (0), H (15), ... H, (8
85) etc. Furthermore, since the speed is 3/15 times faster here, the partial coefficient group is divided into every N-1=2 pieces from H(0) to H(
899), the entire partial coefficient group is obtained, A: H(0), H(15), ..., H(885).
)B: H(3), H(18), ..., H(88
8) C: H(6), H(21), ..., I((
891) D: H(9), H(24), ・・
・, H(894)E: H(12), H(2
7), ..., H (897). In equation (2), each of the coefficients (A) to (E) may be used for each i. That is, the acyclic digital low-pass filter 4 has coefficients of 1=(L 3+ 6.
9+12 table surfaces may be used. The coefficient table shown in Table 1 is stored in the coefficient storage means 8 of FIG.

FIG. 2 shows playback at 3/16x speed.

In Figure 2, the input PCM samples are 18a to 1θC.
, and the input signal waveform is 17. By performing a convolution operation on one input sample, for example, lea, using partial coefficient groups (A) to (E), interpolation samples 18a to L8e including the input sample are generated. In this case, interpolated sample 18a is identical to input sample Lea.

At this time, the interpolated waveform is represented by 19, which is not much different from the input signal waveform 17 due to high-order interpolation, and therefore has little waveform distortion, so excellent sound quality is reproduced.

FIG. 3 shows a schematic flowchart of this embodiment. Third
In the figure, in step 1, it is determined whether i is 15 or less, and if it is 15 or less, the process proceeds to step 2, the convolution calculation. In step 2, Y(i) expressed by equation (2) is calculated by controlling the coefficient table described above. Next, proceed to step 3 and output Y(i). In step 4, i is increased by 3 and the process returns to step 1. If i is 15 or more in step 1, proceed to step 5,
After reading the next PCM sample, subtract 15 from i in step 6 and proceed to step 2.

FIG. 4 is a block diagram showing a second embodiment of the invention.

In FIG. 4, 1 to 8. 10.14-15 are the same components as in FIG. Reference numeral 21 denotes a sampling frequency signal having the same frequency as the sampling frequency, and the interrupt control means 20 generates an interrupt signal 22 based on the sampling frequency signal 21 and supplies it to the second control means 23 . When the second control means receives the interrupt signal 22, it gives commands 10 and 2 to the acyclic digital low-pass filter 4 and the coefficient spacing recirculation means 6.

This control is shown in the flowchart of FIG.

In FIG. 5, steps 1 to e are the same flow as in FIG. 3. Step 7 is a loop that waits until an interrupt for each sampling frequency is received.

As described in detail, there is provided an apparatus comprising an acyclic digital low-pass filter, a rate ratio input means, a coefficient storage means, a coefficient thinning cyclic means, and a first control means. The digital low-pass filter performs a convolution operation between the coefficients of the circulated taps and the input PCM sample, and performs high-order interpolation during variable-speed playback, when the playback sampling rate is lower than the output sampling rate, that is, during low-speed playback. This makes it possible to perform playback with excellent sound quality.

Further, it has a second control means in place of the first control means,
Further, the device has an interrupt control means, and for each interrupt, the acyclic digital low-pass filter performs a convolution operation between the coefficient of the circulated tap and the input PCM sample, and the PCM sample is stored before the next interrupt. By outputting, it is possible to reproduce with excellent sound quality during variable speed reproduction, and furthermore, it is possible to easily manage the time for outputting PCM samples and convolution calculation.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a variable speed playback device in a first embodiment of the present invention, FIG. 2 is a waveform diagram showing interpolation in the first embodiment of the present invention, and FIG. FIG. 4 is a block diagram showing the configuration of a variable speed playback device in a second embodiment of the invention, and FIG. 5 is a flowchart in the second embodiment of the invention. FIG. 6 is a block diagram showing the configuration of a conventional variable speed playback device, and FIG. 7 is a waveform diagram showing the relationship between the PCM signal and timing signal in the conventional configuration. 4...Acyclic digital low-pass filter, 6...
Counting thinning circuit means, 8...Coefficient storage means, 16
...Input PCM sample, 18...Interpolation sample, 20...Interrupt control means, 58...
Subtraction means, 59... Multiplication means, 60... Counter, 63... ROM1 67... Addition means, 74... Addition means.

Claims (2)

[Claims] (1) K-tap (K is a positive integer) acyclic digital low-pass filter, sampling rate of output PCM samples and input PCM
rate ratio input means for inputting a ratio N/M (N, M are positive integers, N=1, 2, . . . , M) of a sample to a sampling rate; and a number K of taps of the acyclic digital low-pass filter; coefficient storage means for storing a coefficient group of an acyclic digital low-pass filter whose number of taps is the product K·M of the sampling rate ratio N/M and the denominator M; A partial coefficient group is formed by thinning out (M-1) coefficients from a coefficient group with the number of taps K·M, and the partial coefficient group is thinned out with the number of taps K.
- A device comprising: a coefficient thinning circuit that circulates every (N-1) coefficients among a group of M coefficients; and a first control means that controls the acyclic digital low-pass filter and the coefficient thinning circuit. A variable speed playback device, wherein the non-recursive digital low-pass filter performs a convolution operation between the circulated K-tap coefficients and the input PCM sample. (2) Instead of the first control means, use a second control means that controls the acyclic digital low-pass filter and the coefficient thinning cyclic means after receiving an interrupt signal, and keep the sampling rate of the output PCM samples constant. and interrupt control means for interrupting the second control means at each sampling frequency of the output PCM sample, and for each interrupt, the acyclic digital low-pass filter outputs the coefficients of the circulated K taps and the input. P
A variable speed playback device characterized by performing a convolution operation with a CM sample and outputting a PCM sample.