CH666584A5 - METHOD AND DEVICE FOR DEMODULATING HIGH FREQUENCY MODULATED SIGNALS BY MEANS OF DIGITAL FILTERS AND DIGITAL DEMODULATORS, AND USE OF THE METHOD IN A REMOTE CONTROL RECEIVER. - Google Patents

Description

DESCRIPTION

The invention relates to a method for demodulating 15 high-frequency modulated signals by means of digital filters and digital demodulators, in which the modulated baseband signal having a continuous spectrum of useful and interference signals is first band-limited and then sampled and processed 20 at a specific sampling frequency.

Digital filters and demodulators are widely used today and are used, among other things, to modulate discrete-time digital signals using computers, e.g. Filter and demodulate microcomputers or signal processors. The underlying technology is known as "digital signal processing" or "digital signal processing", see for example the book "Theory and Application of Digital Signal Processing" by L.R. Rabiner and B. Gold, Prentice Hall, New Yersy.

30 In the theory of digital signal processing, a fundamental rule applies to the sampling or clock frequency, the so-called sampling theorem. This means that the minimum sampling frequency with which a continuous signal may still be sampled must be at least twice as high as the highest frequency, which is still noticeable in the spectrum of the signal.

The book "Semiconductor Circuit Technology" by U. Tietze and Ch. Schenk, Springer Verlag Berlin, Heidelberg, New York, 1980 shows that a discrete-time signal (e.g. one with the

Period Ta

1

fa sampled signal) has a periodic spectrum in fa and because of this periodicity the continuous

45 che spectrum Xa (£ 2) on | Q | fa must be limited.

If you ignore this rule, there is a so-called “aliasing” effect, which affects those

1

thus portions of the continuous spectrum which are higher than —fa

2nd

lie, after grading, slide into lower frequency ranges and interfere there.

The level of the sampling frequency together with the length of the processing program determines the minimum computing speed of the computer which must be able to work through the entire signal processing program between two samples. Since the computing speed of microcomputers and signal processors is limited, the use of 60 digital signal processing is very often restricted.

The invention is intended to provide a method which substantially increases the possible uses of signal processing by enabling the use of microcomputers with a previously inadequate computing speed for digital signal processing.

This object is achieved according to the invention by the measures specified in claim 1.

The invention is based on the new finding that the

3rd

666 584

mentioned limitation of the continuous spectrum XA (Q) can not only be introduced in the frequency range fa, but

1

also in every frequency segment from | m — fa | <| £ 2A | <| (m + 1) —fa | with m = 1, 2, 3 all of which are so limited

Spectra after scanning lead to a clear periodic spectrum. The otherwise disturbing “aliasing” effect can be used by repeating the spectra Xa (£ 2) by ± m-fa periodically. A spectrum lying in an upper frequency range in accordance with the above inequality is therefore unchanged into the range | f | by the “undersampling” with fa fa mixed down.

The method according to the invention thus makes it possible, under certain conditions and when the specified measures are taken, to sample, filter and demodulate high-frequency modulated signals with low sampling frequencies than was previously permitted on the basis of the sampling theorem.

The invention further relates to a device for performing the above-mentioned method, with a filter for band limiting the modulated signal and with means for sampling the band-filtered useful signal.

The inventive device is characterized by the features specified in claim 5.

The invention also relates to an application of the method mentioned in a remote control receiver according to claim 7.

Ripple control receivers require high-quality narrow-band filters in order to filter out control signals sampled in amplitude from the low-voltage network. So far, for example, a two-stage digital filter of the fourth order has been used (CH-PS 559 983), which can solve the task, but significantly compared to a microprocessor, for example an 8-bit microcomputer, which has the greatest practical importance today is more expensive. However, an 8-bit microcomputer requires a computing time of approx. 300 jxs for the necessary arithmetic operations (approx. 9 multiplications with filter constants, 8 additions with overflow monitoring and 8 register manipulations) and could therefore be clocked at a maximum clock frequency of 3300 Hz. However, since control frequencies of up to 2000 Hz occur and a sampling frequency of at least 4000 Hz is required according to the sampling theorem, problems with the computing speed have so far arisen and 8-bit microcomputers could not be used as digital filters for ripple control receivers.

With the method according to the invention, this task can now be solved for the first time by allowing lower sampling frequencies in the case in question, which means that the microcomputer has sufficient computing time available.

The invention is explained in more detail below using an exemplary embodiment and the drawings; show it:

1, 2 diagrams for explaining the inventive method,

Fig. 3 is a block diagram of a selective receiving part of a ripple control receiver, and

Fig. 4, 5 diagrams to explain the function.

1 and 2 show diagrams to explain the method according to the invention in a general way. FIG. 1 shows in line a a baseband signal modulated with fm with a given useful and interference signal spectrum XA (β). The useful signal spectrum is denoted by N and the interference signal spectrum is denoted by S. The signal from line a is first band limited according to line b with an analog band pass Ha (£ 2), which results in the useful signal spectrum Xa '(ß) with the stop band limit frequencies Qmi "and ßmax according to line c.

A sampling frequency a> a is now selected for sampling the useful signal spectrum Xa '(£ 2), which must satisfy the following two conditions:

C0a ^ 2 (Qmax - ^ min)

- The useful signal spectrum must be completely within

1

of a period section m — coa (m = 1, 2, 3, ...). Out-

2nd

Spectral components of interference signals may only occur within this period. However, this is only permissible if it is guaranteed that these components will be in the blocking range of subsequent digital filters.

It can be seen from line c that the sampling frequency coa may also be below the occurring signal frequencies.

If these two conditions for the sampling frequency coa are met, the periodic spectrum X (ei0) T) of the signal sampled with e> a results in line d. It is from n

Row d shows that in each period - the complete in-

T

formation is included.

The signal sampled with coa can now be processed according to known principles of digital signal processing. In particular, it can be digitally filtered - with the clock frequency K) a - and demodulated in a digital manner. Line e shows the total frequency of a digital filter H (ejmT) for carrier frequency suppression.

In FIG. 2, a representation has been selected to explain the method according to the invention, as described in chapter 2.12 “Relation Between Continuous and Discrete Systems” of the book “Theory and Application of Digital Signal Processing” by L.R. Ra-biner and B. Gold, Prentice Hall, New Yersy. The method is based on the knowledge that the limitation of the continuous spectrum Xa (£ 2) described in this chapter

n 1

not only in the frequency range] £ 21 or | Q | caa can be introduced, but also in each frequency section

1 1

from | m-jcoa | <| Qa | <| (m + 1) —a> a | with m = 1, 2, 3, ... As in

Line a and b of FIG. 2 for m = 2, all spectra limited in this way lead to a unique periodic spectrum after the scanning.

It was also recognized that the otherwise so disruptive “aliasing” effect can be used in this form by using the formula

2IÏ

(2.65) of the mentioned chapter the spectra Xa (ü) by ± m -

F

or ± mcoa are repeated periodically. This means that one in an upper frequency range | m — a> a | <| Qa | <| (m + 1)

1

—Tt> a | horizontal spectrum due to the "subsampling" with raa

2 1

unchanged in the area | co | <yCOa is mixed down.

The relationships for m = 1 are shown in lines c and d of FIG. As a comparison of the lines a and b for m = 2 on the one hand and the lines c and d for m = 1 on the other hand shows, the fact must be noted that, depending on whether m is even

1

or is odd, in the range co <—coa the positive or the ne-

2nd

negative analog frequency spectrum appears.

Of course, the sampling theorem must be observed for signal processing. However, this requirement is auto5

10th

15

20th

25th

30th

35

40

45

50

55

60

65

666 584

4th

matically fulfilled if the further signal processing takes place in the approximate cycle. In the event that the negative spectrum is processed further

1

is to be taken into account that the frequencies at —coa

2nd

are mirrored. In general, this does not pose any problems for AM systems and FM systems for digital data transmission (e.g. FSK systems). In audio applications, of course, the spectrum must not be mixed upside down, provided that frequency shifts are permitted at all.

The method described is particularly well suited to implementing digital filters in ripple control receivers. Ripple control receivers, as is known, require high-quality narrow-band filters in order to be able to filter out control signals sampled in amplitude from the low-voltage network. In European patent application 83 105 834.2 (publication number 0 105 087), a fourth-stage two-stage digital filter is presented that can fundamentally solve this task.

However, if you want to implement the digital filter as an inexpensive 8-bit microcomputer, there are problems with the computing speed and the clock frequency.

On the one hand, the necessary arithmetic operations (approx. 9 multiplications with filter constants + 8 additions with overflow monitoring + 8 register manipulations) require a computing time of approx. 300 ns, so that the digital filter can be clocked at a maximum clock frequency of 3300 Hz, and on the other hand writes that Sampling theorem because of the occurring control frequencies up to 2000 Hz a sampling frequency of at least 4000 Hz. This means that the digital filter cannot be operated with the sampling frequency prescribed by the sampling theorem.

3 shows the block diagram of a selective receiving part of a ripple control receiver, with which the task of implementing the required digital filter as an 8-bit microcomputer can be achieved using the method according to the invention.

As is known, the selective receiving part designated by reference number 1 in FIG. 3 serves to selectively receive a remote control signal with the signal frequency fs from the frequency mixture offered by the network and to emit a pulse sequence corresponding to the remote control commands. The structure and mode of operation of a ripple control receiver are assumed to be known; in this connection, reference is made to the already mentioned European patent application 83 105 834.2 and to CH-PS 559 983.

The receiving part 1 has an input terminal 2, which is connected to a connection point 3 of a power line 4, which is superimposed on the signal frequency fs. The input voltage at the input terminal 2 is fed to a pre-filter 5, which is followed by an analog / digital converter 7 and a digital filter 8. An AM demodulator 9 is arranged after the digital filter 8, the output of which is connected to the output terminal 10 of the receiving part 1. The receiving part 1 also contains a frequency generator 6 having a quartz crystal for generating the clock frequency for the individual stages of the receiving part 1. The clock frequency could also be derived from the network by means of a control circuit called PLL.

The receiving section 1 and its mode of operation will now be explained with reference to FIGS. 3 to 5, with FIGS. 4 and 5 showing the signal profiles in the individual stages of the receiving section 1: FIG. 4 shows in line a the limitation of the received signal spectrum with the Prefilter 5 (Fig. 3) and in line b the digital spectrum of the sampled signal. 5 shows in line a the filter characteristic of the digital filter 8 (FIG. 3), in line b the spectrum of the output signal of the digital filter 8, in line c the amplitude response of the filter chain pre-filter 5 + digital filter 8 and in line d the attenuation the interfering passages of the filter chain through the pre-filter 5.

The pre-filter 5 is formed by a second-order analog bandpass filter with the quality Q> 15. 4,

1

Line a with fa and —fa an attenuation of -20 dB on and

2nd

limits the received useful and interference signal spectrum of the control frequency fs. The clock frequency of the frequency generator 6, which corresponds to the sampling frequency fa of the A / D converter 7, is chosen such that the useful signal spectrum XA (jf) into one period

1

m — fa of the frequency axis comes to lie. Outside of this

2nd

Interference frequencies should be sufficiently attenuated so that the digital filter 8 is not disturbed in the pass band by original or down-mixed interference frequencies. 4, line a, the sampling frequency fa is 3000 Hz

1

half the sampling frequency — fa is 1500 Hz and the signal

2nd

1 1 lies in the frequency period between m — fa and (m +1) —fa,

2 2

where m = 1.

After sampling the band-filtered network signal, a digital spectrum X (ei2nfT) results according to FIG. 4, line b. The dashed curve A shows the sum of all overlapping periodic spectra. Since the bandpass filter according to line a has only a finite attenuation (-20 dB), somewhat disturbing «aliasing» still occurs.

The output signal of the A / D converter 7 (FIG. 3) is obtained with the digital filter 8, which can be of the type described in European patent application 83 105 834.2, for example, and a filter characteristic according to FIG. 5, line a,

1

has filtered. Because this filter dampens strongly at —fa and fa

(-20 dB), the disturbing «aliasing» effects are strongly suppressed.

Line b of FIG. 5 shows the spectrum of the output signal of the digital filter 8 (FIG. 3): Y (ej2nfr) = X (ei2nft) • H (ej2nfT). The spectrum is characteristic in the areas fa - fs and fa + fs, where new frequencies have arisen due to the periodicity of the signal spectrum X (ej2nfT) and the filter transfer function H (ei2nft) («aliasing» by scanning). In the fa - fs area, the spectrum is mirrored in fa compared to the original spectrum. However, this has no influence on further processing, since only the amplitude of the signal is to be evaluated. With FSK systems, however, the reflection would have to be taken into account.

The output signal of the digital filter 8 is evaluated in terms of amplitude by the AM demodulator 9 (FIG. 3), which is preferably implemented digitally. This evaluation can be carried out as follows: The digital signal Y (e'2nfT) is rectified per scan, so the absolute value of Y (nT) is formed, Y (nT) = | Y (nT). This absolute value is given to a digital low-pass filter, which is also clocked with fa and has a cut-off frequency that is adapted to the frequency of the baseband signal, but which of course

1

half of -fa.

2nd

5, line c, the implemented single-tone transmission function, that is to say the amplitude response of the filter chain of analog bandpass filter 5 and digital filter 8 (FIG. 3), is shown. The transmission is not true to frequency. Because when a control frequency fs is applied to the filter, the fundamental frequency fa - fs (dashed

5

10th

15

20th

25th

30th

35

40

45

50

55

60

65

5

666 584

line B). The same frequency fa - fs also appears when the frequency fi = fa - fs is fed in, but this frequency fi is attenuated by 25 dB, which is indicated by point C. This attenuation is achieved solely through the prefilter 5 (FIG. 3). Likewise, the frequency fa - fs appears at the output of the digital filter 8 (FIG. 3) when driving with any frequency m (fa ± fs), m = 1,2, .... The damping of all these periodic frequencies is likewise given exclusively by the pre-filter 5.

Since, according to the invention, the regulation for the sampling frequency merely states that the useful signal spectrum within one

1

Frequency period must lie between -fa, exist in the election

2nd

the sampling frequency still some freedom. In addition,

1

visible that is outside the frequency period —fa

2nd

Interference spectra only have to be attenuated by the prefilter 5 to the extent that they are not suppressed by the following digital filter 8 (FIG. 3).

This results in the task of optimally designing the sampling frequency fa together with the analog bandpass filter 5 and the digital filter 8. In the present case there is a particularly useful solution if the sampling frequency fa fulfills the following condition:

2nd

fa = -fs

5 ^

It then turns out that the two critical “pass bands” of the filter system: fa - fs and fa + fs, which are to be damped by the analog prefilter, are in a ratio of 1: 2 and 2: 1, respectively. As is shown in FIG. 5, line d, this ensures that the second-order analog bandpass filter attenuates the two interfering passbands equally. The ratios 1: 2 and 2: 1 lead to equidistant frequency spacings on the logarithmic frequency scale.

Since with ripple control receivers the control frequency fs is up to 15 2000 Hz, this results in a sampling frequency fa of 3000 Hz. This sampling and clock frequency is still sufficiently low even for simple 8-bit microcomputers, whereas the sampling frequency of at least 4000 Hz required by the sampling theorem would be clearly too high. There are also conceivable cases 20 where the method described results in an even higher “undersampling gain”, for example in the case of narrow-spectrum modulated with high frequency. These can be limited with a narrow bandpass filter and transformed into a lower frequency range by sub-sampling, where they can be finely filtered.

4 sheets of drawings

Claims (9)

666 584 1 fa or - fa a strong, adapted to the interference signal spectrum 1 to lie within a period m-fa of the frequency axis 1 gnalen outside of the mentioned period period m — fa das 1 period section m-fa, where m is a natural number. 1. A method for demodulating high-frequency modulated signals by means of digital filters and digital demodulators, in which the modulated baseband signal, which has a continuous spectrum of useful and interference signals, is first band-limited and then sampled and processed with a specific sampling frequency, characterized in that the modulated Baseband signal with the continuous useful and interference signal spectrum XA (Ü) with an analog bandpass filter (5) band-limited, which results in a useful signal spectrum Xa '(ß) with a lower and upper stop band frequency Qmjn or ßmax, and that one selects a sampling frequency fa , which is not less than twice the value of the difference between the upper and lower stop band frequency, i.e. fa> 2 (Qjnax - £ imin), the useful signal spectrum at least except for spectral components of interference signals within a pe 2nd 2nd Has damping and is operated with a clock frequency corresponding to the sampling frequency. 2nd is coming. 2nd The filter (8) for the digital filtering of the sampled signal is designed so that the parts mentioned come to lie in its blocking area. 2. The method according to claim 1, characterized in that the signal sampled with the sampling frequency fa is digitally filtered with a clock frequency corresponding to the sampling frequency and demodulated in a digital manner. 2nd 2nd PATENT CLAIMS 3rd ± 20-30% fulfills the condition fa = - fs. 3. The method according to claim 2, characterized in that when occurrence of spectral components of interference 4. The method according to claim 3, characterized in that the digital filter (8) attenuates spectral components of interference signals which are not suppressed by the analog bandpass filter (5). 5. Device for performing the method according to claim 1, with a filter for band limitation of the modulated signal and with means for sampling the band-filtered useful signal, characterized in that the filter is formed by an analog bandpass filter (5) of the second order, and that the sampling frequency fa is selected such that on the one hand it is not less than twice the value of the difference between the upper and lower stop band frequency of the useful signal delimited by the bandpass filter and on the other hand the useful signal spec- 6. The device according to claim 5, characterized by a digital filter (8) for filtering the sampled useful signal, which is at the whole and half the value of the sampling frequency 7. Application of the method according to claim 1 in a remote control receiver, in particular in a ripple control receiver, with a digital filter for filtering amplitude-sampled single-tone control signals from the low-voltage network, characterized in that an 8-bit microcomputer is used as the digital filter (8) and this with a Clock frequency fa samples, which is below the minimum sampling frequency prescribed by the sampling theorem. 8. Application according to claim 7, characterized in that the sampling frequency fa for the sampling of the received signal with the control frequency fs and thus also the clock frequency is chosen such that it is within a bandwidth of 9. Application according to claim 8, characterized in that the pass bands fa - fs and fa + fs of the filter system analog bandpass filter plus digital filter (8) are equally attenuated by the analog bandpass filter (5).