JP2576120B2 - D / A converter - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a D / A converter suitable for use in converting a digital signal into an analog signal (hereinafter referred to as D / A conversion).
The present invention relates to a conversion device.

[Summary of the Invention]

The present invention processes input digital data, shapes noise included in the processed data, alternately D / A converts the shaped data using a plurality of D / A converters, and adds the respective outputs. By outputting, digital data can be D / A converted at a speed higher than the conversion speed of the D / A converter itself, and the quantization noise after D / A conversion is reduced. It is.

[Conventional technology]

Generally, the sampling frequency of a CD etc. is 44.1kHz, 48kHz
A value slightly higher than twice the audible band, such as z, is taken, and the basic D / A converter performs D / A conversion at this speed. In this case, an analog filter (low-pass filter) is used after the D / A converter to remove aliasing noise. However, since a sharp characteristic is required, there are many problems in sound quality.

To improve this, an oversampling digital filter is used before the D / A converter to interpolate the data and calculate a new sample value between one sample value and the next sample value. Insertion has been done.

[Problems to be solved by the invention]

However, when oversampling is performed, the sampling frequency increases, and accordingly, a D / A converter having a high conversion speed is required. However, the one with a high conversion speed is technically more difficult as the speed becomes higher, and the cost becomes higher. In other words, it is no exaggeration to say that the upper limit of oversampling is determined by the limit of the conversion speed of the D / A converter.

Note that using multiple low-speed D / A converters and input data
Japanese Patent Laying-Open No. 57-140026 discloses that a D / A converter is performed at a high speed by alternately performing the timing of the / A conversion. However, this method cannot reduce quantization noise after D / A conversion.

The present invention has been made in view of such a point, and it is possible to D / A convert digital data at a speed higher than the conversion speed of the D / A converter itself, and to reduce quantization noise of the D / A converter. An object of the present invention is to provide a D / A conversion device that can reduce the number.

[Means for solving the problem]

The D / A converter according to the present invention has a sampling frequency f
processing means for oversampling input digital data of s at the sampling frequency Nfs, a noise shaping circuit for shaping noise included in data from the processing means so as to be reduced within a signal band, and a noise shaping circuit for the noise shaping circuit D / A converters that alternately D / A convert data from N / A at Nfs / 2 conversion speed, and the multiple D / A
And an adder for adding the outputs of the converter in a state shifted by 1 / (Nfs) to obtain an output analog signal.

[Action]

According to the present invention, the input digital data of the sampling frequency fs is oversampled by the sampling frequency Nfs by the processing means, and the noise included in the data of the processing means is reduced within the signal band by the noise shaping circuit. The data from the noise shaping circuit is alternately D / A-converted at a conversion speed of Nfs / 2 by a plurality of D / A converters, and each output of the plurality of D / A converters By the adder, 1 / (Nfs)
By adding in a state shifted by the time, an output analog signal in which noise in the signal band and quantization noise whose frequency is near Nfs / 2 is reduced is obtained.

〔Example〕

Hereinafter, an embodiment of the present invention will be described in detail with reference to FIGS. In the present embodiment, as an example, a case where D / A conversion corresponding to 8 × speed is performed using a D / A converter corresponding to 4 × speed will be described.

FIG. 1 shows a circuit configuration of the present embodiment. In FIG. 1, (1) denotes an input terminal, and this input terminal (1)
For example, digital data reproduced from a CD (not shown) is demodulated by a signal processing IC, error detection is performed, and error correction is performed if necessary. The digital data of the sampling frequency fs from the input terminal (1) is supplied to a digital filter (2) as processing means, where it is oversampled at a predetermined frequency, for example, 8 _fs .
Data of 8f _s is output.

Data for this 8f _s adder (3a), the frequency of the quantization noise is shaped by the quantizer (3b) and the delay circuit 1-order noise shaping circuit comprising (3c) (3). The frequency characteristic of the quantization noise when shaping is not applied by the noise shaping circuit (3) is substantially constant over the entire region as shown by the solid line a in FIG. 3, but the frequency characteristic when shaping is applied. in is as shown by the solid line b in FIG. 3, the quantization noise in the audio band increases reversed 4f _s near one half of the even last sampling frequency or the oversampling frequency is greatly reduced problems is there. Therefore, in this embodiment, the plurality as the 4f _s near the quantization noise is below D /
An A converter is used to reduce the aperture effect.

The noise shaping effect NS of the noise shaping circuit (3) is given by the following equation.

In the above formula (1), f ₁ is the bandwidth, f _s in the case of the sampling frequency for example CD of the original data, 44.1KH
z. Here, NS at each double speed when f ₁ = 20 kHz and f _s = 44.1 kHz are as follows.

When 2f _s NS = - NS = time of 2.1 dB 4f _s - The NS = -43.8dB present embodiment when NS = -13.8 dB 16f _s when NS = -19.8dB 256f _s when 7.8 dB 8f _s 8f _{Since this is} the case of _s , an improvement of NS = -13.8 dB can be obtained.

The data (8f _s ) obtained by shaping the frequency of the quantization noise in the noise shaping circuit (3) in this way is supplied to the switch circuit (4) as a distribution circuit, and the speed between the contact points a and b is 8 _fs . It is switched alternately. As a result, data of substantially 8 _fs is sampled every other data, and data of 4 _fs is obtained at each of the nodes a and b.

These data are supplied to D / A converters (5) and (6) operating at a conversion speed of, for example, 4 _fs , where they are alternately D / A converted. As a result, analog output signals are obtained at the output sides of the D / A converters (5) and (6) as shown by the dashed lines a and b in FIG. 2, respectively. These output signals are supplied to an adder (7), where they are added with their phases shifted from each other by 1/8 _fs and averaged. As a result, an averaged output signal as shown by a solid line c in FIG. 2 is obtained at the output side of the adder (7), which is led out to the output terminal (8).

Each analog output signal of the D / A converter from Figure 2 (5) and (6) is also a 1 / 4f _s cycle, the addition by averaging by adding it like with a shift 1 / 8f _s together analog signal output after becomes 1 / 8f _s period, this is equivalent to the result obtained by one after the D / a converting the average value data for 8f _s.

In the frequency domain of the analog output signal added after the D / A conversion, the frequency characteristic H (f) due to the aperture effect is given by the following equation.

For example, the return frequency of a 1 kHz signal is 4 _fs
In the case of the D / A converter, the noise occurs at 175.4 kHz, which is about 44.9 dB from the above equation (2) due to the aperture effect, and further attenuates due to the attenuation of the digital filter (2) and almost disappears.

Characteristics indicated by the solid line a in Fig. 4, D / A converter to work in the conversion rate of 8f _s data 4f _s as described above (5) and (6) in alternating aperture effect in the case of D / A conversion FIG. 3 shows frequency characteristics. In the present embodiment, the frequency characteristics are combined with the characteristics shown by the solid line b in FIG.
The characteristic as shown by can be obtained. Due to this characteristic, not only the quantization noise in the audible band after the D / A converter is significantly reduced, but also the quantization near 4 _fs which is half the frequency of the last sampling frequency (oversampling frequency). It can be seen that the noise is also significantly reduced.

Incidentally, the frequency characteristic of the aperture effect when 8 _fs data is D / A converted by one D / A converter operating at a conversion speed of 8 _fs is as shown by a solid line b in FIG. By combining with the characteristic shown by the solid line b in FIG.
A characteristic as shown by a broken line d in the figure is obtained, and this characteristic can significantly reduce quantization noise in the audible band,
4f _s near the quantization noise is seen that shows still high value.

As described above, in this embodiment, D / D which operates at the conversion speed of 4 _fs is used.
By using two A converters, digital data of 8 _fs can be D / A converted. Also, by combining the frequency characteristics of the noise shaping circuit and the frequency characteristics of the aperture effect by the two D / A converters, quantization noise in the vicinity of 4 _fs, which is 1/2 of the final sampling frequency, can be reduced.

In addition, by using multiple D / A converters, nonlinear distortion and noise are averaged, and the distortion rate after D / A conversion, S
The S / N ratio is improved and the S / N ratio is ₁₀ log ₁₀ N (d
B) Be improved.

In the above embodiment, two D / A converters are used. However, more D / A converters may be used. For example, three D / A converters may be used.
When a converter is used, the sampling frequency becomes 12 _fs , and aliasing noise is suppressed in the vicinity of 8 _fs in addition to 4 _fs.
When using D / A converters, the sampling frequency is 16f
_s, and the other 8f _s of 4f _s, near 12f _s also aliasing noise is suppressed.

〔The invention's effect〕

According to the present invention described above, the sampling frequency is fs
Processing means for oversampling input digital data at a sampling frequency Nfs, a noise shaping circuit for shaping noise included in data from the processing means so as to be reduced within a signal band, and a noise shaping circuit. D / A converters that alternately D / A convert data at Nfs / 2 conversion rate, and the D / A converters
It has an adder that adds each output of the A converter in a state shifted by 1 / (Nfs) to obtain an output analog signal, so that the digital speed of the D / A converter is higher than that of the D / A converter itself. Data can be D / A converted, and an output analog signal can be obtained in which both noise in the signal band and quantization noise near the frequency of Nfs / 2 are reduced.

[Brief description of the drawings]

FIG. 1 is a circuit diagram showing an embodiment of the present invention, and FIGS. 2 to 4 are views for explaining the operation of the present invention. (2) is a digital filter, (3) is a noise shaping circuit, (4) is a switch circuit, (5) and (6) are D /
The A converter and (7) are adders.

Claims (1)

(57) [Claims] 1. A processing means for oversampling input digital data having a sampling frequency of fs at a sampling frequency Nfs, and a noise for shaping noise included in data from the processing means so as to be reduced within a signal band. A shaping circuit, a plurality of D / A converters for alternately D / A converting data from the noise shaping circuit at a conversion speed of Nfs / 2, and each output of the plurality of D / A converters is 1 / ( An adder for obtaining an output analog signal by performing addition in a state shifted by Nfs).