JPH06338797A - Bit length extension device - Google Patents

Abstract

PURPOSE:To improve distortion and a noise when an infinitesimal level is reproduced. CONSTITUTION:Difference between samples of input data is taken out, and the fact that it is smaller than a prescribed level is detected by a level change detecting part 3. Thereby, a variable cutoff frequency filter 2 can be controlled, and the band area of the low-pass filter 2 can be narrowed, and a high-pass band noise and high-pass band distortion can be reduced. In such a case, the bit length of the output of the filter is outputted up to the one less than an input bit, and a high-pass band noise such as quantization distortion and noise shape can be eliminated when the input data are an infinitesimal level or are an infinitesimal change.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to improvement of noise and distortion at a very small level when used for data to be converted into digital analog such as digital audio.

[0002]

2. Description of the Related Art Conventionally, sample data having a limited bit length is used in the case of digitally transmitting / recording an analog signal through an AD converter, reproducing it, and performing DA conversion. Accompanied by quantization distortion.

For this reason, recently, in compact discs (CDs) and the like, the format uses 16 bits, but 18-bit or 20-bit data is processed by Japanese Patent Publication No. 55-28445, which is called noise shaping. It is recorded in 16 bits to improve the distortion of a minute level of the audio signal. However, in these cases, the noise in the high frequency range increased and the sound quality was impaired.

[0004]

As described above, the distortion is increased or the high frequency noise is increased at the minute level.

[0005]

In the present invention, these minute levels of high frequency noise are reduced, the bit length is increased, and D
The A conversion is made possible. To achieve this,
Difference between sampled data of reproduced digital signal is 1 LS
A low-pass filter means for detecting a portion below a minute level change such as B or 2LSB and removing high-frequency noise in a section within these ranges is provided. Here, in the output of the low-pass filter means, the output is less than the input bit length. Noise and distortion are improved by increasing the output to a level below the LSB of the input and producing a smooth signal without high frequency noise.

[0006]

FIG. 1 is a block diagram showing the whole. Variable low-pass filter 2 for input data 1 consisting of n bits
Put in and output. At a general level, the low-pass filter 2 is a 20 KHz cut-off digital filter or the like for through or CD.

Further, the low-pass filter 2 is selectable so as to be attenuated from several KHz by switching the filter coefficient, and the operation result of the filter is truncated by N bits of the input and is output to the lower bits.
(N + m bits) Generally, even 20 bits or 18 bits has a great effect.

When the detector 3 detects a slight level change of the input data by detecting the difference between the samples of the input data, the purpose is achieved by using the low pass filter. That is, the level is gently changed in a low level portion or a particularly low frequency, and the output at this time can be a gentle reproduction signal without high frequency noise.

FIG. 2 shows the concrete means of FIG. The input data 1 is sampled and held by the shift register 3a, and the difference between the next input sample data and the output thereof is subtracted by the subtractor 3b.
The subtraction process is performed with to extract the difference data. This output is input to the comparator 3d, the reference data 3c, that is, the value for determining the detection of the minute level, 1LSB or 2LSB, etc. is input, and the change of the value smaller than this value is extracted.

On the other hand, the input data 1 enters the low pass filter 2 and is output. Low pass filter 2
Uses a register of coefficients required for the filter operation in the shift register 2a. From the output of each register 2a, the coefficients Ko to Kn required for the filter are multiplied by the multiplier 2b, taken out, and added by the full adder 2c to obtain the filter output 4.

Here, the coefficient K for the central register
If o is set to 1 and the other is set to 0, through data can be obtained, or a low-pass filter with an arbitrary cutoff can be realized by changing the coefficients Ko to Kn. This filter coefficient is switched by the output of the comparator 3d for level detection. Here, if the output timing of the filter and the output timing of the comparator 3d are different, it is necessary to match them, and in FIG. 2, the delay circuit 3e is included in the output of the comparator 3d.

FIG. 4 shows an explanatory diagram of the operation. A corresponds to the waveform when the filter output data is DA converted. When the data between each sample greatly changes like A1, the filter is a switching signal of a predetermined wide band filter such as 20 KHz. As shown in A3, 1
In the small change portion such as LSB change, the filter is set to a low frequency, and it is possible to omit the output of the solid line A2 and the high frequency noise of the change of LSB as shown in the conventional example of A3.

In the case of general high frequency signal reproduction, 1LS
It is a large level of B or 2 LSB or more, and there is no deterioration in frequency characteristics.

FIG. 3 shows another embodiment of the present invention. Figure 2
In the above, the filter calculation needs to be performed for 16 bits of all input data, and the filter becomes large, but this is easily performed in FIG. That is, the difference between the input data is extracted, and the filter operation is performed using the data only for the level change for switching the filter. The limiter 5 makes the output of the subtractor 3b, which is a difference output, equal to or higher than a predetermined level. That is, when the filter is used for changes of 1 LSB or less, all changes of 1 LSB or more are output as ± 1. This output is put in the shift register 6. At the output of the register, the lower bit of the input n-bit data is generated and added to the upper bit to obtain the low pass characteristic.

FIG. 5 shows an explanatory diagram of this waveform. When the high-order input n-bit data changes by 1 LSB as shown in (a), the difference data becomes as shown in (b-1), and while this (b-1) shifts the shift register, it is passed through the coefficient unit 7. The coefficients Kn to K1 are passed to the output of the full adder 8 to give a coefficient to be output as shown in (c-1), and when the center is passed, the output is reversed as shown in (c-2). Then, the output obtained by adding the output of the full adder 8 and the high-order n data by the adder 9 is filtered as shown in (d-1) of (2), and becomes a smooth signal because the lower bits are increased. . Similarly, when the input data is data including high-frequency noise in the slowly changing part as shown in (a-3), a large number of difference outputs are output as the difference output, and this output is the full adder as shown in (c). 8 and the output of the adder 9 becomes a smooth signal without high frequency noise as shown in (2).

[0016]

As described above, at the reproduction minute level,
Not only the distortion can be reduced, but the recorded data by noise shaping can be reproduced even in the transmission path of a bit with a small output close to the original bit length with high frequency noise removed. Further, there is no deterioration in frequency characteristic at a normal level.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a specific block diagram showing an embodiment of the present invention.

FIG. 3 is a block diagram showing another embodiment of the present invention.

FIG. 4 is a diagram showing a waveform processing result.

FIG. 5 is a diagram showing a waveform processing result.

[Explanation of symbols]

 1 Input data 2 Low pass filter 3 Detector 3a Shift register 3b Subtractor 3c Reference data 3d Comparator 3e Delay circuit 4 Output data 5 Limiter 6 Shift register 7 Coefficient multiplier 8,9 Adder 21a-n Shift register 22 Multiplier 23 Addition vessel

Claims (2)

[Claims] 1. In digital data processing intended for digital-analog conversion, means for detecting a predetermined minute level change or less between sample data and low-pass filter means having a cutoff characteristic lower than a normal frequency band. And a data processing means for outputting the output bit of the filter calculation result to a bit output lower than the minimum bit of the input, and switching to the output of the low-pass filter means lower than a normal band by the detection output. Bit length extension device. 2. The bit length expansion device according to claim 1, wherein the low-pass filter means performs arithmetic processing using data of a slight change in the difference of the input data and synthesizes the data with the upper input bits.