CA1293027C - Method of, and demodulator for, digitally demodulating an ssbsignal - Google Patents

Abstract

ABSTRACT
A method of, and demodulator for, digitally demodulating an SSB signal. In the demodulator a band limited signal from a roofing filter is digitised in an ADC before being applied to quadrature related signal paths including mixers and low pass, decimating filters. The output from the decimating filters are applied to a Hilbert transform pair. The upper or lower sideband signal is obtained by adding (or subtracting) the outputs of the Hilbert transform pair. Thereafter the digital signal obtained is reconverted to an analogue signal. By choosing the decimation factor to be a numbered integer greater than 1, the periodic transfer function of the Hilbert transform pair has alternate passbands and stopbands. This relaxes the requirements on any analogue filters in the receiver and greatly enhances the adjacent channel selectivity of the receiver.

Description

~293027 1 20104~ 57 l~ethod of, and demodulator for, digi~ally demodulatillg an SSB
signal.

The present invention relates ~o a method of, and demodulator for, digitally demodulating a single sideband (SSB) signal.
Analogue demodulation of an SSB siynal is known Norgaard, Proc. IRE Vol. 44 Dec. 1956, pp 1703-1705. Digital demodulation of an FM signal has been reported in Frequenczr lg83, 37, pages 16 to 22 and ntz Bd 36 (1983) Part 12, pages 806 to 808.
Typically in a digital implementation an IF signal is applied to an analogue to digital converter (ADC), processed digitally and reconverted to an analogue signal in a digital to analogue converter (DAC). An advantage of processing a signal digitally is that a more flexible or general purpose demodulator results.
However the kno~n methods require particularly stringent filtering which means that dedicated circuits have to be made for each stage of the demodulator.
An object of the present invention is to simplify the filtering in a digital demodulator whilst maintaining good selectivity.
The present invention provides a method of demodulating an SSB signal, comprising digitising an analogue signal to provide a digitised signal, applying the digitised signal to quadrature mixing means, decimating the signals from the quadrature mixing means using an odd numbered integer factor greater than 1, applying the decimated signals to a Hilbert trans-~2930Z7 PHB 33.179 2 5.4.1986 form pair, and arithmetically combining OUtpUtS or theHilbert transform pair to produce an SSB signal.
According to another aspect of the present in-vention there is provided an SSB demodulator comprising means for digitising an analogue signal, quadrature related mixing means coupled to the digitising means, means coupled to eachof the quadrature relatedmixing means for decimating the signals therefrom by an odd numbered integer factor greater than 1, a Hilbert transform pair coupled to the deci-mating means, and means for arithmetically combining theoutputs of the Hilbert transform pair to produce an SSB sig-nal.
The present invention is based on recognition of the fact that selectivity in a demodulator can be improved by initially analogue filtering an input signal and then subsequently digitally filtering the filtered analogue signal.
In particular digitally filtering by decimating signals using an odd numbered integer factor greater than 1 and applying the decimated signals to a Hilbert transform pair enables a passband/stopband characteristic to be obtained that is characterised in that a passband is flanked by stop-bands. This means that the filtering requirements of the de-modulator can be relaxed. This relaxation means that it is possible to implement the digital part of the demodulator using a general purpose processor.
In an embodiment of the present invention the digitised signal is mixed with quadrature components of a local oscillator signal which is mathematically related to the sampling frequency, f5~ used in digitising the ana-logue signal. The mathematical relationship may be fs(n-+x/4) where n is zero or an integer and x is 1 or 3. Frequencies f fs/4 and fs (11 + -) have been found to be convenient.
The analogue signal, which may comprise an IF sig-nal, may be band limited to prevent aliasing and this also provides far-out adjacent channel isolation.
The band limited signal is obtained by bandpass filtering in a roofing filter and improved selectivity can lZ93~Z7 PHB 33.179 3 5.4.1986 also be obtained by adjustlng Ihe nomlnal IF lre~uency ~o tnat it corresponds to one end of the passband of the roofing filter. Alternatively improved selectivity can be obtained by coupling a digital pre-processing filter to the output of the means for digitising the band limited signal and adjusting the nominal IF signal so that it corresponds to the centre of the passband of the roofing filter.
The present invention will now be described, by way of example, with reference to the accompanying drawings, wherein, Figure 1 is a block schematic circuit diagram of a superheterodyne digital radio receiver having an output stage for demodulating an SSB signal, Figure 2 is a block schematic circuit diagram of an FM output stage which can be used with, or insteadof, the SSB output stage, Figure 3 is a block schematic circuit diagram of an AM output stage which can be used with the SSB and/or FM output stages or independently of one or both, Figures 4 to 7 are tabular summaries of no deci-mation and decimation by 4, 3 and 5 respectively of the VSB
or LSB signals, Figures 8 to 12 are SSB selectivity charts for a receiver not including a digital pre-decimation filter, in the case of Figures 9 to 12 the various passbands have been overlaid one on the other, Figure 13 is an SSB selectivity chart for a re-ceiver including a digital pre-decimation filter, and Figure 14 is a flow chart relating to implementing the digital processing part of the receiver using a micro-computer.
Referring to Figure 1 the receiver comprises an antenna 12 connected to an r.f. front end 14 which may con-sist of the antenna 12, an amplifier or at least one IF
stage. The output of the r.f. front end 14 comprises a sig-nal on a nominal IF frequency which in the case of SSB is suppressed. This output signal is connected to a bandpass lZ93027 PHB 33.179 4 5.4.1986 filter, termed a roofing filter 18, to provide a band limited output in which aliasing is prevented and adjacent channel isolation is achieved. The filtered signal is applied to an analog-to-digital converter (ADC) 20 which is sampled at a frequency f5, which is greater than twice the signal bandwidth, typically fs = 40 kHz. The bandwidth (BW) of the roofing filter 18 in the present embodiment is fixed and is determined by the widest sW signal to be demodulated. However the filter 18 may have a variable bandwidth to ease the ana-logue to digital conversion.
The digitised output from the ADC 20 is applied toquadrature related digital signal paths 22 and 24. The block shown in broken lines comprises a digital filter 21 which is optional but if present it has an input coupled to the ADC 20 and an output coupled to the paths 22 and 24. The paths 22 and 24 comprise mixers 26, 27 to which digitised versions of sin 2~rfCt and cos 2~rfCt are applied, fc being typically 4 fs in order that a simple approximation can be made to obtain digital quadrature feed to the mixers 26, 27 in the paths 22 and 24. Alternatively by making fc equal to 114fS, then for fs - 40kHz, fc is 450kHz which is very close to a standard supernet I.F. frequency. The outputs of the mixers 26, 27 are baseband signals which are applied to res-pective decimating filters 28, 29. The pass bands of the filters 28, 29 are narrower than that of the roofing filter 18, and in operation low pass filter the signals to remove any possible aliasing components, and then decimate the sig-nals using a predetermined odd-numbered integer factor greater than 1. The outputs X and Y of the decimated filters 0 28, 29 are applied to a Hilbert transform pair 30, 31.
The stage 30 has the characteristic that it gives a -90 phase shift for a signal lying in the frequency band from 0 to fs/2 and +90 for a signal lying in the frequency band -fs/2 to 0. The stage 31 of the Hilbert transform pair is a zero phase shift stage formed by a delay network to delay the signal by an amount corresponding to the delay in the stage 30. The decimating filters 28, 29 and the Hilbert transform stage 30 are implemented as finite impulse response (FIR) ~Z~3(~27 PHB 33.179 5 5.4.1986 filters such as transversal filters. By decimating the sig-nals in the filters 28, 29. then the Hilbert transform stage 30 can be operated at a lower sampling rate and there-fore requires fewer stages. The outputs from the Hilbert transform pair 30, 31 are applied to respective inputs of a summing network 32 if lower sideband demodulation of an SSB signal is required, or of a subtracting n~twork 33 if upper sideband demodulation of a signal is required. The par-ticular demodulation being selected by operation of a switch 10 34, coupled to the networks 32, 33.
After the particular arithmetic operation the di-gital sum (or difference) signal is reconverted to an ana-logue signal by a digital to analogue converter (DAC) 36 and the output therefrom is applied to a low pass filter 38. If desired an interpolating filter 35 may be connected in the signal path to the DAC 36. An advantage of using the inter-polating filter 35 is that the specification for the low pass filter 38 can be relaxed as the periodic repeat fre-quency components are further away from baseband.
The digitally operating section of the receiver can be implemented by a suitably programmed microcomputer such as a Texas Instruments TMS 320. By operating digitally one achieves perfect tracking between the signal paths 22, 24.
In determining the operation of the demodulator made in accordance with the present invention one selects a local oscillator frequency and decimation factor such that the Hilbert transform stage 30 provides a bandpass filtering characteristic within the bandwidth of the roofing filter 18.
Figures 4 to 7 are tabular summaries showing the composite performance of the LSB and USB channels for different incident SSB frequencies. Figure 4 shows the periodic filter characteristic when there is no decimation in the decimating filters 28, 29 and Figure 5 shows the case when the decimation factor is 4. The upper tabular summary relates to the upper side band (USB) subtraction channel and the lower tabular summary relate to the lower side band (LSB) addition channel. The other legends used in the tabu-lar summaries are RUSB - rejected upper sideband, RLSB - re-12~3(~27 PHB 33.179 6 5.4.1986 jected lower sideband and INV - inverted.
Figures 6 and 7 show the periodic filter character-istics when the decimation factors are 3 and 5, respectively.
An examination of these filter characteristics shows that the frequency length of the segments becomes shorter for higher values of the decimation factors. Also there are points in the tabular summaries where it is possible to note some advantageous operating points. For example when a parti-cular wanted sideband is flanked between two rejection bands then this considerably eases the analogue prefiltering re-quirements of the dem~dulatoL Such advantageous operating points exist in regions corresponding to the nominal carrier fc = (n + -)fs but only for schemes incorporating decimat-ion by an odd integer factor (as exemplified in Figures 6 and 7).
There can be a disadvantage in making the passband/
stopband segments too small in that the Hilbert filter cha-racteristic may revert too soon back to passband within the pass bandwidth of the roofing filter 18 leading to reduced selectivity of these points with subsequent desradation of the receiver performance. Another reason for not making the passband segments too narrow bandwidth is that they may be narrower than that of the wanted signal.
Figure 8 shows the filter response 39 for a demo-dulator of the type shown in Figure 1 without a digital fil-ter 21 and in which the decimating factor is five and the nominal IF frequency of fs(n + -) is set centrally within the roofing filter passband 40. The decimating filter pass-band 42 is narrower than the roofing filter passband 40 and is disposed symmetrically with respect to the passband 40. The characteristic 44 of the Hilbert filter is also shown together with the legends relating to what is happen-ing in the respective segments which have been identified at the top of Figure 8 by the letters A to F. In the situation outlined an apparently satisfactory demodulator performance is degraded by a small portion of the next but one channel (dictated by the Hilbert filter characteristic 44), segment lZ93(1 27 PHB 33.179 7 5.4.1986 B, being insufficiently attenuated by the decimating filter and giving only 40dB adjacent channel rejection. This si-tuation can be mitigated by the provision of the digital filter 21. Alternatively the value of fs can be adjusted slightly so that the wanted USB lies in the centre of the roofing filter passband characteristic 40 butthe unwanted channel B is further down the slope of the characteristic 40 and so is attenuated.
Figure 9 shows that adjusting fs in this manner leads to a 60dB rejection over the adjacent channels. How-ever this is at the expense of not being able to have fast switching between VSB (segment D) and LSB (segment C) (Figure 8) simply by reversing the sign in the SSB phasing algorithm as this does not give equal adjacent and neigh-bouring channel isolations. Furthermore adjusting fs re-quires altering the division ratio in a frequency synthe-siser and this does not always lead to convenient round numbers thus fs (n + 4) cannot always be positioned as de-sired.
Figure 10 illustrates the overall characteristic when the local oscillator frequency is adjusted so that the nominal IF frequency fs(n + 4) lies against the cut-off point of the roofing filter characteristic 40.
Figures 11 and 12 illustrate the selectivity of the modulator to obtain VSB and LSB for an arrangement in which the passband of the decimating filter has been relaxed so that it reaches its stopband when the Hilbert filter enters its adjacent passband. To switch between upper and lower sidebands the nominal IF frequency is required to switch two channel widths so that the frequency of fs(n + 4) remains positioned against the edge of the roofing filter characteris-tic 40, at the lower frequency edge for the USB (Figure 11) and at the higher frequency edge for the LSB (Figure 12).
Reducing the decimating factor from 5 to 3 may improve the overall demodulator performance because the seg-ments are wider and in consequence the unwanted channel either is further down the slope of the characteristic 40 or lies outside the passband of the roofing filter.

~293a27 PHB 33.179 8 5.4.1986 The alternative arrangement for obtaining an SSB
signal with a satisfactory sideband rejection includes the digital preprocessing filter 21 (Figure 1). Additionally as shown in Figure 13 the passbands of the roofing filter, response curve 4~, and the decimating filter, response curve 42, are different from the arrangement not having the fil-ter 21, see Figure 8 for example. The response of the digital pre-processing filter 21 is referenced 45 and as shown the passband corresponds to the width of one of the passbands of the Hilbert filter and has a sharp cut-off which is desirable for good selectivity with SSB.
In operation the analogue output of the R.F. front end 14 is applied to the roofing filter 18. The filtered analogue signal is digitized in the ADC 20 and the digital output is filtered in the digital filter 21. In mixing the output of the filter 21 in the mixers 26, 27 the frequency fs which in this example is four times higher than fc, is selected so that fs(n + 4) is symmetrically disposed with respect to the passbands of the roofing and decimation fil-ters 18 and 28, 29 respectively. After decimating the sig-nals in the paths 22, 24, the Hilbert transform is obtained and the appropriate sideband is obtained, as before, by addition or subtraction.
The foregoing description is concerned with de-modulating an SSB signal.
The circuit shown in Figure 1, omitting the digi-tal pre-processing filter 21 can be made universal to AM, FM
and SSB by connecting different end stages to the points A
and A' in the paths 22 and 24, respectively. Figure 2 illustrates an arrangement for recovering a digitised FM
signal. This arrangement comprises delays 50, 52 providing a delay ~ = m/f5 where m is an integer, coupled to the ter-minals A and A', respectively. Mixers 54, 56 are provided, each mixer has two inputs, one input of each mixer being coupled to a respective delay 50, 52 and the other input of each mixer being coupled to the non-delayed signal in the other path. The outputs of the mixers 54, 56 are coupled to " 1293(~z7 PHB 33.179 9 5.4.1986 respective inputs of a subtractor 58 from which a digital version of the FM signal is derived. This signal is applied to the DAC 36 and filtered in the low pass filter 38.
Figure 3 illustrates that AM output can be derived by squaring the signals at A and A' in squaring circuits 60, 62, the outputs of which are added in a summing stage 64. After which the sum signal is applied to a square root stage 66 to obtain a digital version of the demodulated AM signal. This digital signal is converted to an analogue lo signal and low pass filtered in the stages 36, 38, res-pectively.
The present invention is based on the concept of analogue filtering an input signal and digitally filtering the filtered analogue signal by decimation using an odd numbered integer factor greater than 1 and ~ilbert filtering the decimated signal to obtain the required selectivity. By this approach the specification of the roofing filter 18 can be relaxed. If the demodulator is designed for demodu-lating SSB signals only then the passband of the roofing filter 18 is a gently sloping single channel one, the digi-tal filtering being used effectively to cut-out a slot having sharp or fast sloping sides. By relaxing the specifi-cation of the roofing filter then alignment and local oscillation drift problems are avoided.
Alternatively if the demodulator is designed for multi-mode operation then the passband of the roofing filter 18 is wide enough to pass full deviation FM. The digital filtering in such a case provides an effective way of ob-taining further filtering especially when demodulating SSB
signals. In practice there is a relationship between the bandwidth of the roofing filter 18 and the sampling rate in the ADC 20 (Figure 1) which is that the bandwidth of the roofing filter must be less than 2 the sampling frequency (fs/2), that is the Nyquist frequency. The bandwidth of the roofing filter 18 is dictated by the intended appli-cation, that is single mode or multi-mode, and this indi-cates the required sampling rate~ In the illustrated multi-lZ93~Z'7 PHB 33.179 10 5.4.1986 mode case the bandwidth of the roofing filter 18 is chosento pass full deviation FM and this corresponds to about four SSB sidebands.
Figure 14 illustrates a simplified flow chart for the various digital operations implemented by a micro-computer such as a Texas TMS 320 which is running in real time. It will be assumed that the pre-decimation filter 21 has been omitted and an SSB signal is being recovered.
Figure 15 demonstrates the fundamental operation of the demodulator program. The diagram shows how a loop counter is used to allow decimation by five. A sample rate of 4OkHz is to be used for the data input which would make the design of the Hilbert filter impracticable for any reasonable performance. However in choosing a sample rate of 40kHz then the quadrature mixing can easily be performed with 1OkHz thus giving a nominal input carrier frequency of 1OkHz. The bandwidth of the two channels is 3kHz. After the initial quadrature mixing decimation by five is used to reduce the sample rate down to 8kHz where it is possible to design practical Hilbert fi]ters and make better use of the available processing time.
In Figure 14 the hlock 70 is a programme timing block which ensures that sampled data from the A to D con-verter is ready every 25/usec. This data is read into the processor step 72. Step 74 denotes operating on the input sampled data to give two quadrature channels. This is achieved by multiplying the incoming signal with a 1OkHz locally gene-rated pseudo local oscillator. One channel is derived from direct multiplication while the quadrature channel is de-rived from multiplication with a 90 degrees phase shiftedversion of the pseudo local oscillator. The next step 76 is to take the signals in the mixer outputs and store them by moving data in a data memory storage and updating vacated memory locations. The step 78 denotes increasing the count in a counter by one. Step 80 denotes checking whether the count is 5, if it is not (N) then the cycle is repeated un-til the count is 5, that is (Y) at which time the program PHB 33.179 11 lZ93~Z7 5.4.1986 exists to step 82. In step 82 the counter is reset to zero and the data, which is in the correct place, is operated on by the decimation filter coefficients, step 84. Step 86 denotes the Hilbert transform operation, and the step 83 s denotes the arithmetic operation to obtain the appropriate sideband.
The operational steps to obtain digital versions of the FM and AM signals are self evident from a consideration of Figures 2 and 3 and accordingly separate flow charts will not be described.
In the case of using the digital pre-processing filter 21, this is configured as a 30 stage transversal filter which receives time samples from the analogue to digi-tal converter 20. After a newly received time sample has been stored, each of the stored time samples is multiplied by its individual filter coefficient and the results are accumulated to provide an output to the quadrature related mixers 26, 27. The multiplication and accumulation operat-ions are carried out again after the receipt of another time sample.
Since the use of the pre-processing filter 21 enables the specification for the decimation filters 28, 29 to be relaxed, then each filter comprises only eight stages.

Claims (18)

1. A method of demodulating an SSB signal, comprising digitising an analogue signal to provide a digitised signal, applying the digitised signal to quadrature mixing means, decimating the signals from the quadrature mixing means using an odd numbered integer factor greater than 1, applying tile decimated signals to a Hilbert transform pair, and arithmetically combining outputs of the Hilbert transform pair to produce an SSB signal.

2. A method as claimed in claim 1, in which the digitised signal is mixed with quadrature components of a local oscillator signal which is mathematically related to the sampling frequency used in digitising the analogue signal.

3. A method as claimed in claim 2, in which the local oscillator signal is fs (n + x/4), where fs is the sampling frequency, n is zero or an integer, and x is 1 or 3.

4. A method as claimed in claim 3, in which n is zero and x is one.

5. A method as claimed in claim 3, in which n is eleven and x is one.

6. A method as claimed in claim 1, in which the analogue signal is band limited before being digitised.

7. A method as claimed in claim 6, wherein the band limited signal is obtained by bandpass filtering in a roofing filter and a nominal IF frequency of the band limited signal is arranged centrally of the roofing filter passband and the passband of filtering means coupled to the output of each of the quadrature related mixing means.

8. A method as claimed in claim 6, wherein the band limited signal is obtained by bandpass filtering in a roofing filter and a nominal IF frequency of the band limited signal is arranged at a cut-off point of the roofing filter passband and substantially centrally of the passband of filtering means coupled to the output of each of the quadrature related mixing means.

9. A method as claimed in claim 6, 7 or 8, further comprising digitally filtering the digitised band limited signal prior to applying it to quadrature related mixing means.

10. A method as claimed in claim 1, 2 or 3, wherein the signal is filtered to avoid aliasing prior to being decimated.

11. An SSB demodulator comprising means for digitising an analogue signal, quadrature related mixing means coupled to the digitising means, means coupled to each of the quadrature related mixing means for decimating the signals therefrom by an odd numbered integer factor greater than 1, a Hilbert transform pair coupled to the decimating means, and means for arithmetically combining outputs of the Hilbert transform pair to produce an SSB
signal.

12. A demodulator as claimed in claim 11, in which a local oscillator means coupled to the quadrature related mixing means are arranged to produce a signal whose frequency is fs (n + x/4) where fs is the sampling frequency used in digitising the analogue signal, n is zero or an integer and x is 1 or 3.

13. A demodulator as claimed in claim 11, in which a local oscillating means coupled to the quadrature related mixing means is arranged to produce a frequency fs/4, where fs is the sampling frequency used in digitising the analogue signal.

14. A demodulator as claimed in claim 11, 12 or 13, further comprising means for band limiting the analogue signal.

15. A demodulator as claimed in claim 11, 12 or 13, further comprising filtering means for filtering the outputs of the mixing means, said filtering means having a bandwidth such as to prevent aliasing.

16. A demodulator as claimed in claim 11, 12 or 13, further comprising a digital pre-processing filter connected to the output of the digitising means.

17. A method as claimed in claim 1 or 2, wherein the decimation and Hilbert filtering functions are carried out using a programmed microcomputer.

18. A receiver including an SSB demodulator as claimed in claim 11, 12 or 13.