JPS6046845B2 - digital envelope detector - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a digital envelope detector that prevents deterioration of S/N due to aliasing during sampling.

In recent years, with the development of digital LSI technology, digital modulators, filters, etc. have been developed.

Particularly in demodulators for amplitude modulated waves, the phase-locked loop for carrier extraction can also be digitized, and as a result, synchronous detectors can also be digitized relatively easily. However, there are also amplitude modulated waves for which synchronous detection is not possible. For example, in a white/black binary facsimile signal, the carrier is transmitted only in the black part, so stable carrier extraction cannot be performed on the receiving side, and synchronous detection is impossible. Therefore, when demodulating such a signal, envelope detection must be used. The principle of envelope detection is as follows.

The DS adjustment signal which is the received amplitude modulated wave is expressed as p()=A(
) c0sωct (1) However, A()>0.

When this is full-wave rectified, the full-wave rectified output/power Q ( ) is Q ( ) = 1 p ( ) 1 ■ A ( ) ICOS (J) C
tHere, if f(t)=1C0S(X)Ctl is related to the Fourier expansion, then 1 Therefore, A.

=! (-1)'+1,9yo,,9-,,...
(5) This is the frequency spectrum of (t). From this, if the signal shown in FIG. 1b is passed through a low-pass filter with a pass band of 0 to 1 KHz, a demodulated output can be obtained. Incidentally, in the envelope detection described above, a digital method has recently been considered.

Figure 1 is a diagram showing an example of the configuration of a digital envelope detector.
The received amplitude modulation waveform p(t) is sampled by a sampling pulse with a frequency of F by the sampling and hold circuit 11, and the obtained sample value is sent to the A/D converter 12.
The absolute value of this code is taken by a digital absolute value circuit 13, and a digital low-pass filter 14 and a D/A are used to extract only the baseband signal component.
By passing it through the converter 15, the baseband signal A・(
t) is extracted as a demodulated output. Note that if the demodulated output remains digital, the D/A converter 15
can be omitted. The digital absolute value circuit 13 is a circuit that converts the sign of the digital output of the A/D converter 12 into a positive value (if it is a positive value, the magnitude remains unchanged); if it is a negative value, the magnitude remains unchanged.

For example, if the output of the A/D converter 12 is a 5-bit two's complement digital signal, the output will be ゜“0001F” for an input sample value of ゜“+3゛” if the MSB is the sign bit. At this time, the output of the digital absolute value circuit 13 is also "0001 digits". Also, when the input sample value is "-4", the output of the A/D converter 12 is expressed in two's complement as "11100". The output of the digital absolute value circuit 13 is ゜゜00100 representing "+4゛"
This conversion is well known as the positive/negative conversion of two's complement representation, for example, by taking the exclusive OR of the sign bit and other bits and adding the size of the LSB to it. It is realized by ゜゜1110.
If we take the exclusive OR of the MSB ゜゜l゛, which is the sign bit of 0゛, and the other 4 bits ``1100゛, we get 6'001゛.If we add the LSB size ゜4000r' to this, we get ``゜0100゛. becomes. Since the digital absolute value circuit 13 always converts the input MSB to ゜0゛, which represents a positive value, its output becomes "゜00100゛. Note that an error of the size of the LSB is allowed in the output of the digital absolute value circuit 13. If so, the addition of the LSB magnitude can be omitted.Generally, if the frequency spectrum S(f) of the signal s(t) is IfI>F
If m and S(f)=0, then if we know the sequence s(IT) obtained by sampling s(t) at a sampling frequency Fs of F, > 2fm (where i = integer, T = 1/F,) It is known as the sampling theorem that s(t) can be reproduced.

If 1fI>Fs/2 and S(f) is half 0, then
Since this component is mixed into the component of S(f) at Ifl>F,/2 due to sampling, the original correct waveform cannot be restored from the sampling value S'(IT). As is well known, the component of S(f) in this waveform causes waveform distortion called aliasing distortion. (1) p(
Since the spectrum of t) is finite, the value P(1T) sampled at a sampling frequency Fs that is more than twice the maximum frequency of the spectrum (1 = integer, T = 1/Fs)
p(t) can be reproduced more easily.

Further, the absolute value 1P(1T)l of P(IT) is equal to the sampling value Q(1T) of Q(t) in equation (2). However, as shown in equation (6) and FIG. 1(b), Q(t
) exists up to infinite frequency, so
Q(1T) cannot accurately reproduce Q(t). That is, the sample hold circuit 11 and A/D converter 12 in FIG. By obtaining (same as obtaining the sampled value Q(IT) of Q(t)),
Q(t) cannot be reproduced. This means that the IfI>F components of Q(t) all fall into the frequency range of IfI×F, due to aliasing due to sampling, and mix with the components of Q(t) in the true IfI×F. This is because they become inseparable. When this aliasing component enters the band of the low-pass filter 14,
The baseband component in Figure 1b, that is, (6
) is added to AO/2A(t) in the equation, so a correct envelope detection output A(t) cannot be obtained. That is, when there is a component of frequency Mfs (m: integer) in the spectrum of the sampled signal with respect to the sampling frequency F, this component is folded back by the sampling and falls to the frequency 0.

Similarly, the frequency component of Mfs±Δf near Mfs falls to the frequency Δf. Since such aliasing components cannot be removed by the low-pass filter 14, these aliasing components become noise and cause a decrease in S/N.
In order to increase the S/N, the sampling frequency F can be made sufficiently high, but there is a limit to how high F can be made due to factors such as the calculation speed of the digital low-pass filter 14.

By appropriately selecting the sampling frequency, the present invention enables
The present invention aims to provide a digital envelope detector that can effectively increase N.

The present invention sets the sampling frequency Fs so that at least a modulated wave component (the second term AlA(t)COS2ωCt in equation (6)) having a frequency twice that of the carrier of the received amplitude modulated wave enters the band of the digital low-pass filter 14. This objective is achieved by selecting such that there is no such thing.

In other words, among the modulated wave components at a frequency of 2nfC that may enter the band of the low-pass filter 14 due to folding, it is the component at the frequency of n=11, that is, the lowest frequency, that has the greatest effect on the S/N. Since the amplitude becomes smaller, even if it falls within the band of the low-pass filter 14, the influence on the S/N becomes smaller. As previously mentioned,
The component that falls into the band of the low-pass filter 14 due to folding is a frequency component near Mf, so by appropriately selecting the sampling frequency Fs, the modulation component of the cast can be prevented from entering the band of the low-pass filter 14. By doing so, the S/N ratio can be significantly increased. The present invention will be specifically described below with reference to the drawings. FIGS. 3 to 5 are diagrams for explaining folding states of the spectrum of the output signal of the digital absolute value circuit 13 when various sampling frequencies are selected. As described above, the full-wave rectified output (that is, the output of the digital absolute value circuit 13) corresponds to a collection of signals obtained by DSB modulating a carrier of frequency Mfs with the original signal. Therefore, in Figures 3 to 5, the original signal 0 to F
, /2 are shown in the first stage, and the frequency components of the original signal F
, ~F, /2 frequency component modulates the carrier of Fs, original signal Fs - Fs + F, /2 frequency component F
, are also in the range of 0 to Fs/2, so they are shown in the second and third rows. Similarly, the fourth and fifth rows in the figure are falling frequency components in the range of O-Fs/2 when a 7s carrier is modulated with the original signal. The components when modulating a carrier of 3 fs or more are omitted in the figure, but the sum of these components is O-F, / of the output signal spectrum of the digital absolute value circuit 13.
The components fall within the range of 2.

Figure 3 shows the sampling frequency Fs=8k in Figure 2.
FIG. 3 is a frequency spectrum diagram showing a folding state when Hz is selected. That is, in this case, 8KHz 16KHz
, ... are folded back and fall to the vicinity of frequency 0, but 7. The modulation component of =4 KHz (the second term in equation (6)) is not folded back and does not fall within the band of the low-pass filter 14. In this case, if the band of the low-pass filter 14 is set to 0 to 1 kHz, the above table shows that for the baseband signal component (1.2732) 2 = 1.621, the KH due to aliasing that enters the band of the low-pass filter 14
The noise component which is a modulation component due to sampling of z, l6kHz, ... is (-0.0849)2x2+(-0.
0202)2×2+(-0.0089)2×2+...
=0.0152, and the S/N is as follows, and a good value is obtained.

FIG. 4 is a frequency spectrum diagram showing the folding state when Fs=6 kHz is selected.

In this case, the noise components due to folding that enter the band 0 to 1 kHz of the low-pass filter 14 are 12 kHz, 2 kHz.
The modulation component of 4kHz, (a). 0364) 2×
2+(-0.0089)2×2+...=0.0028
1, the S/N is as follows, which is further improved than when Fs=8kHz. Furthermore, if F, = 5.5 kHz is selected, only half of the 12 kHz modulation component enters the band 0 to 1 kHz of the low-pass filter 14 as shown in Fig.
/N improves. What is common in the examples shown in Figs. 4 and 5 is that the sampling frequency F, is adjusted to the frequency of the 0 part of the spectrum between each modulation component in Fig. 1b, and this part O is The first point is that the signal falls into the baseband signal band. In this way, 2fs=4k
Hz (when n = 1) 4fs = 8 kHz (when n = 2) components can also be prevented from entering the band of the low-pass filter 14, so the S
/N is improved. In the above example, we have explained the case where the spectra of each modulation component are relatively far apart as shown in Fig. 1b, but when they are relatively close together,
For example, if the baseband signal component spans the band from 0 to 1.0 kHz, the spectrum of Q(t) will be as shown in Figure 6, but even in this case, the sampling frequency F, can be selected in the same way. Therefore, the S/N can be increased. That is, the band of the low-pass filter 14 is set to 0 to
If it is 1.5 kHz and F,=8 kHz, the S/N will be 20.8 (DB) as described above. Also, Fs=6.5k
Hz, as shown in FIG. 7, the noise component that enters the band 0 to 1.5 kHz of the low-pass filter 14 for the baseband signal component (1.2732) 2 = 1.621 is, and the S/N It will continue to improve even further.

Next, when the range of the sampling frequency Fs according to the present invention is quantitatively determined, it is as follows.

If the relationship between sampling frequency FSl carrier frequency FO baseband signal band FBl and cutoff frequency FL of low-pass filter 14 is, and if and are satisfied, the modulation components of 2nfc and 2(n+1)FO are within the band of low-pass filter 14. Don't get into it.

That is, if F is selected within the range of , the S/N can be increased.

Also, if Fc-FB<FL・(11), some of the aliasing components will necessarily enter the band of the low-pass filter 14, so if Fs is selected to minimize this component, Similarly, the S/N can be increased.

[Brief explanation of the drawing]

Fig. 1 is a frequency spectrum diagram for explaining the principle of envelope detection, Fig. 2 is an explanatory diagram of the configuration of a digital envelope detector according to the present invention, and Figs. A frequency spectrum diagram showing the folding state due to sampling when selected based on the present invention in relation to the spectrum in Figure 1. Figure 6 shows another example of the frequency spectrum of the full-wave rectified output of the received amplitude modulated wave. The figure shown in FIG. 7 is a frequency spectrum diagram showing the aliasing state due to sampling when the sampling frequency in FIG. 2 is selected based on the present invention in relation to the spectrum in FIG. 6.

Claims (1)

[Claims] 1. A received DSB modulated signal waveform with a carrier frequency f_C and a baseband signal bandwidth f_B is sampled at a sampling frequency f_S, the sample value is digitally encoded, and then the absolute value of this digital code is cut off. In a digital envelope detector that performs envelope detection by passing through a digital low-pass filter of frequency f_L,
The sampling frequency f_S is 2(n+1)f_C-
f_B−f_L≧f_S≧2_nf_C+f_B+f_
A digital envelope detector characterized in that L (n: any integer) is selected to a satisfying value. 2 The sampling frequency f_S is especially f_C−f_B<
When f_L, f_S≒2_nf_C+f_B+f_L
A digital envelope detector according to claim 1, characterized in that the digital envelope detector is selected as follows.