GB2231218A - FM interference reduction - Google Patents

Abstract

A method of reducing interference between the detected modulations of a strong and a weak received FM signal, of angular frequencies OMEGA 1, and OMEGA 2 respectively, forming together a composite input signal wherein the input signal is applied to an amplitude detector 9 followed by a low-pass filter 10 and amplifier 11 to produce a beat signal of frequency equal to OMEGA 2- OMEGA 1, and is also applied to an amplitude-limiter 7 and filter 71, and wherein the beat signal and the signal from filter 71 are multiplied together in a multiplier 8 whose output is low-pass filtered by filter 12, the output of which is detected via amplitude limiter 13 in a frequency discriminator 14. The arrangement is shown to give detection of the weak-signal modulation. The composite signal may be fed directly, and with a quadrature phase-shift, to two multipliers of a single- sideband generator, the multipliers also receiving phase-quadrature separated versions of the beat signal. Refinements of this technique to give improved performance, and arrangements for giving improved detection of the strong-signal modulation, are described. <IMAGE>

Description

FM interference reduction This invention relates to reducing interference between received frequency-modulated (FM) signals, both adjacentchannel and co-channel interference. In the former there is no overlap of the sidebands of two interfering signals; in the latter there is overlap.

Ideally, as is known, a wanted strong FM signal should completely "capture" a receiver's detector and thus eliminate the resulting demodulation of any weaker, ie smaller amplitude, FM signal which is also receivable, whether adjacent or co-channel. In general, although ideal performance is not obtained, demodulation from strong wanted signals having large modulation indices is satisfactory.

Unfortunately, those features which provide such near-ideal performance are disadvantageous if the wanted signal is weaker than the interfering signal. In this situation, the wanted demodulation is suppressed as efficiently as the unwanted modulation in the former situation.

The present invention provides improved techniques which can not only give demodulation of the weaker signal over a wide range of weak-to-strong signal ratios, but can also reduce interference to any wanted demodulation whether arising from weaker or stronger signals.

According to the present invention a method of reducing the interference between the detected modulations of stronger and weaker received frequency-modulated (FM) signals constituting together a composite input signal comprises: amplitude-detecting the composite input signal to derive a beat signal of frequency equal to the difference between the carrier-frequencies of the two input signals and including both FM modulations; multiplying a version or versions of said composite input signal with a version or versions of said beat signal to derive a product comprising a first signal at the frequency of one input signal and which includes the original FM modulation thereof; said first signal being thereby either separated from the other input signal, or being accompanied by a second signal at a difference frequency and which is thereby readily separable from said second signal.

Said composite signal may be amplitude-limited before multiplication with said beat signal, and the product may be low-pass filtered before detection of the modulation by frequency-discrimination. Alternatively said composite signal may be amplitude-limited and thereafter fed directly and with a quadrature phase-shift respectively to the two multipliers of a single-sideband generator, said multipliers also respectively receiving phase-quadrature separated versions of said beat signal, the subtracted contents of said two multipliers being subjected to detection of the modulation. In either case said amplitudelimiting of the composite signal may be omitted.

Said phase-quadrature separated versions of said beat signal may be produced by: squaring the composite input signal, amplitudedetecting the second harmonic of said square, and removing the DC component of said amplitude-detected second harmonic to derive one said version; and squaring said one version, amplitude-detecting said square and adding the amplitude-detected output to said square, and taking the square-root of said sum to derive the other said version.

In order to alleviate phase ambiguities, said other said version may be further processed by detecting the nulls therein, feeding a signal marking alternate nulls to two AND circuits respectively held at opposite logic levels, feeding the AND circuit outputs via an OR circuit to two analogue gates which receive respectively said other said version and an inverted version thereof, summing the outputs of said analogue gates, and utilising said summed output as said other said version. For alleviating phase ambiguities liable to rapid reversal said composite input signal may be conventionally detected by frequency-discrimination and also detected by a method as claimed in claim 4, said two detections being multiplied together, and the low-pass filtered product applied to generate a trigger pulse which is added to said signals marking alternate nulls.

The present method may also comprise: amplitude-limiting the composite input signal to derive a signal at the carrier-frequency of the stronger signal and having the FM modulation thereof; phase-modulating said amplitude-limited signal to substantially eliminate its phase component constituted by the difference between the phases of the stronger signal and the composite signal; and detecting the modulation of the resultant signal by frequency-discrimination.

Said phase-modulation may be effected by feeding the amplitude-limited signal to a phase-modulator comprising two multipliers to which said amplitude-limited signal is fed directly and with a quadrature phase-shift respectively, said multipliers also receiving respectively phasequadrature separated versions of said beat signal, to one of which is added a DC component corresponding to the squared amplitude of the larger input signal, the difference of the multiplier outputs providing said stronger signal minus said phase component. Said stronger signal minus said phase component may be fed to a single-sideband generator as hereinbefore defined and processed therein as hereinbefore defined thereby to provide as output said weaker signal minus said phase component.

The invention also provides apparatus for performing the above method.

The present invention will now be explained by way of example, with reference to the accompanying drawings wherein: Fig 1 is a block schematic diagram of a known circuit for enabling the detection of a weaker FM input signal in the present of a stronger one; Figs 2 and 3 are diagrams illustrating the operation of the circuit of Fig 1; Fig 4 is a block schematic diagram of a circuit embodying the present invention for obtaining the effect sought by the circuit of Fig 1; Fig 5 is a similar diagram of another circuit embodying the present invention; Fig 6 is a similar diagram of a circuit for obtaining wide-band quadrature signals to use in the circuit of Fig 5; Fig 7 shows waveforms of beat-frequency signals producible in the circuit of Fig 6 with co-channel interference; Fig 8 is a similar diagram of a circuit for resolving phase ambiguities in the circuit of Fig 6;; Fig 9 is a similar diagram for resolving rapid phase changes in the circuit of Fig 8; Fig 10 is a similar diagram of a circuit for improving the capture of either the stronger or weaker FM input signal.

The prior art The known circuit shown in Fig 1 was described by E J Baghdadb in IRE Trans on Communication Systems, pp 147-161, 1959. In it, a composite input signal comprising a wanted stronger FM signal and an unwanted weaker FM signal are applied to a conventional amplitudelimiter 1 followed by a low-pass or band-pass filter 70 to remove any harmonic products of limiting which could produce distortion, and an amplitude-adjuster 2, eg an amplifier or attenuator. The output of circuit 2 and the composite input signal are subtracted in circuit 3, whose output, after passing through an RF filter 4 to remove unwanted interference (as will be explained), passes via a second amplitude-limiter 5 to frequency-discriminator detector 6.

In order to explain the operation of Fig 1 it is necessary first to explain the so-called "capture effect" with reference to Fig 2. Fig 2 shows two FM signals together with the resulting composite signal c. The two signals are: K cos(Qlt + A P1) + k cos(Q2t +#P2) ... (1) where: ##lt and Q2t are the RF (or IF) phases of the strong and weak signals respectively; QP1 and A P2 are their respective frequency modulations expressed in phase-modulated form; K and k are the respective amplitudes of the stronger and weaker signals.

It can be shown that the composite signal can be written as c = (K2 + k2 + 2kK cos (p) cos (Qlt + A P1+ #) ... (2) whence, if c is amplitude-limited, c(limited) = cos (Qlt + A P1 + ~) ... (3) where (neglecting any harmonics produced by the hard limiter or clipper 1, which are removed by the filter 70): c(limited) amplitude = unity for convenience tan# = k sintp K+k cos# # = #2 A P2 + A P2 The geometrical identities of c, # and ç are shown in Fig 2.

As is known, for # < o/4 one can write

where k = k/K.

If k < 1, the higher-order terms in equation (4) can be ignored, and equation (2) can be rewritten as approximately c = (K+kcos#)cos[#1t+ AP1+k sin(#2-#1t+ #P2- #P1)] ... (5) After removal of the amplitude term (K + k cos) in equation (5) by hard limiting, producing unity amplitude, the modulation term #P1 can be detected by conventional frequency discrimination. This is the capture effect.

However, an interference term k sin (#2-#lt +#p2 ~ A P2 #P1) is also produced.

This interference term is usually small owing to the small value of k compared with usual values of # P1, and this virtual elimination represents the final outcome of the FM capture effect. This elimination is generally very desirable except when the weaker signal is required. Even for larger values of k less than unity, where equation (3) is more applicable (ie where # is not subject to the approxi mation applied above to equation (4)) capture still occurs; however, the amount of interference, as represented by in in equation (2), may be undesirably high, especially for low values of #p1.

The amplitude term in equation (2), viz 7 + k2 + 2kK cos#) #, represents a beat (Q2 - Q1) between the two signals of frequencies #2 and Q1' for if this term is expanded for K k it becomes

Upon conventional amplitude-detection, the second term provides a beat signal whose use in the present invention will be described subsequently.

Reverting to Fig 1, it has been shown above that the output of the hard limiter 1, which receives the aforesaid composite signal c, can be expressed by equation (3), viz as the captured signal cos(#1t + ssP1 + 0) ... (3) where ~ is a term containing residual interference from the weaker signal.It has also been shown that, for k < 1, after hard limiting and adjustment of the limited-output amplitude by a factor K in circuit 2, equation (3) can be re-written see equation (5)) as

Upon expansion of equation (7), its subtraction in circuit 3 from the unlimited composite signal, ie from k cos (Qlt + A P1) + k cos (Q2t + A P2) ... (8) gives a nett output of approximately

A spectral view of equation (9) is shown in Fig 3.It is seen that the first term of equation (9) is not only much weaker than the original strong unwanted signal, ie than the first term of equation (8), but is spaced by double the frequency difference from the wanted weaker signal, ie by 2(Q The second terms of equations (8) and (9) have the same frequency, but the amplitude ratio of the second term to the first term in equation (8) is the small value k, whereas in equation (9) it is unity. Hence the RF filter 4 is much more effective in reducing the unwanted first term of equation (9) on two scores, viz amplitude and frequency, than it would be in reducing the first term of equation (8).

After such reduction equation (9) becomes

where kl < k. Amplitude limitation of equation (10) in limiter 5 produces a signal similar in form to equation (5), namely

cos R2t + k 2 + kl sin 2 (#2- Qlt + # P2 ~ A P ) ] (11) where k 1 where k1 = k ie less than unity because of the reduction of k to kl.Thus detection in frequency-discriminator 6 allowed bp to be determined (in its rate of change of frequency form), ie the weaker signal has now captured the detection process. Any amplitude noise inherent in k is removed by the limiter 5 and so allows the usual noise-reduction benefits of FM detection.

The above technique is mainly useful for the reduction of adjacent-channel interference, since co-channel interference does not provide much, if any, scope for the required reduction of kl by filter 4, ie 2#1 - #2 will not differ sufficiently from #2. It is seen that the kl term of equation (11) corresponds in form to k sin (#2- #1t + #P2 - #p1) in equation (7) or, more exactly, to # in equation (3). Since in the above technique k must be small, it is possible, because of the limitations of filter 4, that kl will end up greater than k, and hence the interference upon detection of AP2 in equation (10) will be greater than upon detection of AP1 in equation (7). However, the doubled frequency of the 2(Q2-Q1) term in equation (11) may permit some reduction in interference by post-detection filtering.

In spite of its ability to detect the weaker signal modulation, the technique of Fig 1 has not found much favour because of difficulties in matching the two signals fed to circuit 3 for amplitude level (via adjustment of K in equation (7) by circuit 2) and time delay. The usual performance as reported in the literature appears to be about 10dB reduction, although one reference reports 15-30dB by taking extra care in controlling amplitude variations.

Adiacent-channel interference In the present invention the above matching problems are eliminated or reduced, and one form, suitable for adjacent-channel interference reduction, is shown in Fig 4.

The composite input signal is the same as in Fig 1, and is fed simultaneously via a hard amplitude-limiter 7 and a filter 71 (corresponding to filter 70 in Fig 1 ) to a multiplier 8 and in unlimited form to an amplitude detector 9. As shown earlier, the amplitude-detection of the unlimited input signal produces a beat signal Q2-Q1 derived from equation (6). This be t signal is k2 cos (#2 + #1t + #P2 - #p1)...(12) where

Additional terms of cos 2(#2 - #1t + #p2 -#p1) are produced, but, because of the relatively high value of 2(Q2 n,), these are removed by the low-pass filter 10 following detector 9.Filter 10 also blocks the DC components from detector 9. The beat signal of equation (12) is amplified by an amplifier 11 of gain g and then fed to the multiplier 8 whose product is

Equation (13) represents, as its third term, the wanted signal, plus interference terms similar to equation (9) but with the addition of S. S is constituted by a small proportion of the input of equation (3) arising from any imbalances in the proportions of the two inputs to multiplier 8 and typically will have a level which is 20-40 dB down on the RF input to multiplier 8. Adjustment of amplitude is not required as in the prior art, ie by circuit 2 and Fig 1.

The output of multiplier 8 is fed through a low-pass filter 12, which reduces the amplitude of the first term of equation (13) to k3, and thence to an amplitude limiter 13 whose output, provided gk2 S, is equivalent in form to equation (11), viz cos (#2t + hP2+~ + k3 sin 2 (Q2 Qlt + 2 1 P1)...(14) where k3 = k3/gk2 and is < 1 because of the aforesaid reduction of gk2/2 to k3. This output is fed to frequencydiscriminator detector 14 to yield the desired t P2. The retention of ~ in the detected output in this arrangement is frequently tolerable.

Due to the presence of ~ in equation (14), the latter may at first appear more equivalent in form to equation (3).

This might be true if k/K is large, since then ~ > k3; however, if k/K is small the k3 term dominates and the equivalence to equation (11) becomes more apparent. In any case, this is a second-order effect.

The Fig 4 arrangement thus enables detection of the weak-signal modulation P2 (+#) as in the Fig 1, but without the disadvantageous need for amplitude and delay matching.

Furthermore, detection is not restricted to small values of k/K, ie the arrangement does not depend on the approximation inherent in equation (7) and equation (9) since it utilises the full limited equation (3). Again, inclusion of limiter 13 (cf limiter 5 in Fig 1) allows the usual FM noisereduction benefits to be obtained.

Although the gain g can be made sufficiently high to prevent S being significant in equation (13), there could be dynamic range problems if k/K increased. Ideally g should be automatically adjusted in inverse proportion to k/K, so that a similar performance is obtained independently of k/K, or of any manual adjustments. This can be achieved by applying the beat signal of equation (12) from filter 10 to a detector (not shown) having a post-detection time-constant sufficiently long to give an output of k2, ie k for k K.

This output is then used in a conventional manner to control the gain of amplifier 11. More precisely, this beat signal can be fed into a normaliser (not shown), eg as described in copending GB Appln No. 8712556 (Publn No 2,191,321A) whose constant output is fed to an amplifier 11 of constant gain.

In a modified arrangement the limiter 7 is omitted and the unlimited composite input signal of equation (1) is multiplied by the output of amplifier (12), as before, to give an output which can be shown to be:

This modification is only possible if k/K is small, thus allowing the gkk2/2 group of terms to be ignored compared with the gk2K/2 groups. The detection performance is then similar to that of Fig 1, but without the amplitude and delay-matching difficulties of that arrangement. With a suitably small value of k/K this modification is preferable as simpler because limiter 7 is omitted and filter 10 may be no longer strictly necessary as small k-values may not generate an unacceptable level of harmonics from equation (12).

Fig 5 shows an alternative arrangement which, although more complex than Fig 4, has some advantages. It uses a single-sideband technique for eliminating or reducing the unwanted 2Q2-Ql component of equation (13), instead of the filter 12 in Fig 4, and can also eliminate the interference component ~ in equation (3), as explained later. In Fig 5 the composite input signal is fed to a hard amplitudelimiter 15 and to a beat detector 16. The latter comprises an amplitude detector 9', low-pass AC filter 10', and amplifier 11' as in Fig 4. As before, the output from limiter 15, via a filter 72 corresponding to filters 70 and 71, is given by equation (3) and the output from the beat detector 16 by equation (13).

The output from limiter 15 is fed to a known form of single-sideband generator 17 in which it feeds a multiplier 18 directly and a multiplier 19 via a 90C phase-shifter 20, ie the signal fed thereto is sin (Qlt + AP1 + ~) ... (16) The beat signal from detector 16 is fed to a lowfrequency wideband quadrature phase-shifter 21, whose two outputs form the second inputs to multipliers 18 and 19 respectively, viz equation (13) and

where SA is the out-of-balance component corresponding to S in equation (13).

The output of multiplier 19 is equation (16) times equation (18) and the output of multiplier 18 is equation (3) times equation (13). These products are subtracted in subtractor 22 to give S - SA + gk2 cos (Q2t + AP2 + #) ... (19) which is fed via an amplitude-limiter 33 to a frequencydiscriminator detector 23 to yield #P2.

It will be seen that equation (19) does not contain the interference represented by the second term of equation (13). In practice, however, some residual interference may remain because of imperfections in the subtraction process leading to equation (19), and hence a filter to remove the residual (2#1 - #2# component may be required between circuits 22 and 23.

There may be some difficulty in realising the wideband phase-shifter 21 in the above arrangement. In an adjacentchannel application this difficulty is alleviated by the relatively high value of (Q2 ~ Q1) in equations (12 and (17) compared with the frequencies inherent in ssP2 and #p1.

Provided the maximum and minimum frequencies in the use of a basic 90 phase-shifter or a differential 90 phase-shifter of conventional type often used in singlesideband generators (in which the differential phase-shift remains close to 900 although the phase-shift of either output relative to a reference phase may have any common value as determined by its network components).

In the latter case, ie using a differential phase-shifter, equations (12) and (17) are replaced by k2 cos (#2 - #1t + #p2 - #p1 + #)...(20) and k2 sin (#2 - #1t + #p2 - #) ... (21) where # is the common time-variant phase-shift between the two output signals introduced by the phase-shifter.

The resultant output from circuit 22 is then S - SA + gk2 cos (Q2t + P2 + # + #) ... (22) Equation (22) is similar to equation (la) but now the term ~ is increased by the increment #. If the separation of the aforesaid maximum and minimum frequencies is not too great W tends to zero and the performance tends to that of equation (19). However, if the separation is large## becomes large, thus degrading the performance.

Nevertheless, the use of a differential phase-shifter does achieve the main object of detecting AP2 free of strongsignal capture. As in Fig 4, the amplitude limiter 15 can be omitted for low values of k/K.

Co-channel Interference The arrangements so far described are more suitable for reducing adjacent-channel interference, although some co-channel reduction would be obtained provided there was some reduction, via filtering prior to detection, of the (2n, Q2) term in equations (13) and (18) as already mentioned in relation to Fig 5.

A more suitable but more complex arrangement for reducing co-channel interference is shown in Fig 6. In Fig 5, the low-frequency wideband quadrature phaseshifter 21 theoretically gives complete suppression of interference. In practice, difficulties in achieving such a phase-shifter reduce the performance, particularly in the co-channel case, where the (Q2-Q1) beat difference could be small and even zero, as well as potentially having both positive and negative values, ie where the frequencies overlapped. If the beat difference is unidirectional the arrangement of Fig 5 can be used, but even then the large separation of minimum and maximum frequency will give a large value of ~, and thus severe distortion of the detected signal.

An associated difficulty, arising when (Q2-Ql)is small, is that the beat term, as expressed in equation (12), is not easily obtainable because of the difficulty of filtering-out harmonics of this beat via filter 10. The Fig 6 arrangement resolves both these difficulties, though with some limitations as regards the quadrature phase-shift requirement.

In Fig 6 the composite input signal of equation (2) is squared in a squarer 24 to produce a second harmonic version in which the stronger signal Q1 prevails, viz a signal proportional to (K2+k2+2Kk cos P 1 2 P1)) cos 2(Qlt+APl+~) ....(23) This second harmonic is selected by a filter 25 and is amplitude-detected in detector 26 to give K2 + k2 + 2Kk cos (#2-#lt + t - 1 The third term of equation (24) provides a beat component corresponding to that of equation (12). This signal, as it passes close to, or even through, zero frequency, must not be removed or unduly distorted by any processing used to eliminate the DC terms K2 + k2.If the processing amounts to filtering at sufficiently close to zero frequency, the amount of unwanted removal of signal will be small if the beat does not dwell too long around zero frequency. With two different modulation waveforms it is unlikely that a a long dwell time would occur. A very long time-constant low-pass filter can thus be safely used, but a preferable arrangement is to use a long time-constant to detect the average DC content K + k and then subtract it from the equation (24) signal. In Fig 6 this is effected by the long time-constant integrator 27 and subtraction circuit 28.

The thus-produced beat signal 2kK cos k (#2-#1t +#P2-#P1)......... (25) is uncontaminated by harmonics, unlike those potentially present in equation (12). Hence the beat frequency can go to DC values as well as represent both +ve and -ve values.

The thus-derived beat signal is then divided into two parts. One part, taken direct from circuit 28, forms a non-phase-shifted beat output. The other part is processed to form a quadrature-shifted beat output; this processing comprises squaring the output of circuit 28 in a squarer 29, amplitude-detecting this square in a detector 30 followed by long-time integration in circuit 68, and subtracting the integrated output from the squarer output in a subtraction circuit 31.The resulting difference is fed to a square-root circuit 32, eg of the kind described in our copending GB Application No.8623236 (Publication No.2,195,796A), whose output is thus

= 2kK sin (#2 - #1t + #p2 - #p1)...(26) Equation (26) appears to give a complete quadrature phase-shift relative to equation (25) independent of frequency, these two signals after 6dB attenuation by attenuators 67 and 34 respectively being suitable for feeding to multipliers 19 and 18 of Fig 5. However, square-root circuit 32 introduces a phase/amplitude ambiguity, in that the wanted bi-polar polarities of the quadrature-shifted beat signal are removed and only one polarity becomes available. The effect is shown in Fig 7.

Fig 7(a) shows the true required form of a typical phase-shifted beat signal for certain values of (Q2 and and #p1 whereas Fig 7(b) shows the output from circuit 32. One way to revert to a bi-polar output is to phase-invert the unipolar output of Fig 7(b) at alternate instants when this output become zero. The resultant, however, is now phase-ambiguous and, via one of these phases, will give an unwanted output from Fig 5 of S - SA +2gKk2 cos (#2 - 2#1 t + #p2 - 2 #P1 - #).... (27) which would lead on detection to the wanted output #P2 being contaminated by 2hP1.

It might be thought that the above contamination is unlikely because of the precise occurrence of the zeros in equation (26), ie that once the exact phase of the quadrature beat signal is established, this phase-value is unlikely to be subject to further change so that an equation (19) rather than an equation (27) output is ensured.

However, this ignores factors such as other interference (noise etc) and waveform points such as X in Fig 7. The latter points can tend towards and even equal zero, but do not need a phase inversion to give the required beat output.

The nett effect is shown in Fig 7(c), which is to be compared with Fig 7(a), ie the phase beyond point X is improperly reversed. Thus, in practice, an unwanted phase inversion of the quadrature beat, and hence an increase in interference, is possible. However, there is still a nett gain in that a previously captured signal which would not give a usable output is still present for some of the time.

Since the contamination is only one of modulation, some post-detection filtering may suffice to greatly reduce the effects of phase inversions, particularly if the strong signal is eg a jamming signal whose modulation spectrum is different from the wanted weaker signal. In such cases conversion of a Fig 7(b)-type waveform to a bipolar one is required, together with a manual means of selecting the required phase. This means will be described with reference to Fig 8, which can also provide automatic selection.

In Fig 8 the unidirectional output from attenuator 67 of Fig 6 is led to a conventional null (zero) detection circuit 35 which feeds a bistable divider 36. The two resultant antiphase logic waveforms feed respective AND gates 37 and 38 which are also fed with DC control voltages representing logic "1" and "0" respectively, so that only one gate output (shown from gate 37 in Fig 8) is fed to a following OR gate 39. Hence the output from gate 39 is one of two possible antiphase outputs from divider 36. The OR gate output is fed to two analogue gates 40 and 41 which also receive respectively antiphase outputs directly from attenuator 67 and via a phase-inverter 42 (Fig 6). One analogue gate opens on a "1" output from OR circuit 39 and the other on a "0" output.The outputs of these two gates are summed in an adder 43, whose output is thus a bi-directional signal which is phase-shifted with respect to the original beat signal of equation (25) by +900 or -90 . The waveforms'shown in Fig 8 illustrate the the formation of these two possible phases, ie f Kk sin (writ + P2 - The choice of which phase is fed to multiplier 19 of Fig 5 can be made in several ways, including manual selection of the logic voltages fed to AND gates 37 and 38.

However, the selection of these logic states must be based upon observation of the final output of discriminator 23 (Fig 5). Manual selection may suffice where the ambiguous quadrature output does not change very rapidly. An arrangement for following rapid changes automatically is shown in Fig 9.

In Fig 9 the composite input signal of equation (2) is fed to a conventional FM detector arrangement 44, ie comprising an amplitude-limiter followed by a frequency-discriminator detector, in which the stronger signal captures the detection process and yields L P1. This output, together with the output from the subtraction circuit 22 of the Fig 5 detection arrangement 45, is fed to the multiplier 46 of a correlator 47 in which a low-pass filter 48 integrates the multiplied product. If there is a reversal of quadrature phase from the adder 43 of the Fig 8 circuit, resulting in a large AP1 interference term from the Fig 5 detection arrangement, there is a consequent increase in the output of correlator 47. This increased output level operates a trigger circuit 49 to generate a short pulse which is fed to the divider 36 of Fig 8 so as to reset the divider.This pulse is equivalent to one fed from detector 35 and acts to phase-invert the divider output.

The output of correlator 47 then declines but does not produce an output pulse from circuit 49. Thus the control is as rapid as is permitted by the integration time of filter 48 which may be as long as about 0.1 sec.

The arrangements of Figs 6, 8 and 9 are fairly complex and are not perfect in operation. Perfection is impossible in the co-channel case because a priori the two input signals, ie the wanted and interfering signals, can have very similar carrier frequencies and relative modulations which produce very frequent changes in the ambiguous output phase from Fig 8. These rapid changes make separation of the two signals very difficult because the filter 48 must have a long integration time. Any of the arrangements of Figs 6, 8 10 and 9 for co-channel interference have the same advantage as that of Fig 4, viz they can operate over a wide dynamic range of k/K, ie k, values.Some simplification is possible if k is restricted to very low values, but this is not really helpful because very low k values imply very weak signals whose signal-to-noise ratios are likely to be too small for satisfactory detection anyway.

The arrangements of Figs 6, 8 and 9, which provide an alternative to using the differential wide-band 900 phase-shifter 21 in Fig 5, can therefore also be used to reduce adjacent-channel interference. In the adjacent-channel case positive and negative beat frequencies are not present and hence the combination of Figs 5, 6 and 8 is sufficient, as the arrangement of Fig 9 is not needed to sense any frequent phase-reversals.

Improved capture Alternatives to differential phase-shifter 21 are also useful where it is desired to decrease the effect of the interference terms ~ and 0 in equation (22). The prior-art arrangement of Fig 1 and the so-far described embodiments of the present invention all have the main purpose of detecting the weaker signal. They exploit the capture effect, but do not reduce the detected interference from a weaker signal upon a stronger signal which is detected by a simple amplitude-limiter plus frequency-discriminator by operation of the capture effect, ie ~ in equation (3) is not reduced.

Similarly, the so-far described embodiments do not reduce the detected interference produced by a weak signal upon a strong signal. Indeed the modified value of ~ for the weaker signal, viz p + ~ in equation (22), can be larger than that applicable to the stronger signal, viz the original value of ~.

It might appear that one way of improving the capture effect for a stronger, wanted, signal would be to use any of the above-described arrangements to extract the weaker signal and then subtract it from the composite input signal to produce an even stronger wanted signal. However, this would require an RF subtraction process even more difficult than that described with reference to Fig 1.

Fig 10 includes an arrangement for improving the capture effect for a stronger, wanted, signal, in addition to providing for the detection of the weak signal in a manner similar to that already described. The composite input signal of equation (1) is fed, via a hard amplitude-limiter 50 and a harmonic-removing filter 73 corresponding to filters 70, 71 and 72, to a phase-modulator 51 whose phase-shift is arranged to be -~ in order to eliminate the ~ term in its output, which would otherwise be equation (3). Thus the capture is complete and a pure version of AP1 is obtained by feeding the modulator output via an amplitude-limiter 65 to a frequency-discriminator detector 66.

Fig 10 also shows an arrangement for obtaining the -~ phase-shift in modulator 51, which in this example is a known form of modulator similar to the single-sideband generator 17 in Fig 5, although other arrangements are possible. The phase-shift is produced by feeding to the respective multipliers 18" and 19", two beat-frequency modulating inputs, viz -Kk sin (#2 - #1t + #p2 - #p1)......................(28) and K2 + Kk cos (#2 - #1t + #p2 -#p1)............... (29) The connection of these inputs to the respective multipliers is selected to give the -~ phase-shift (if these connections are reversed an unwanted +~ phase-shift is obtained).

The tangent of the phase-shift is given by the ratio of equations (28) and (29), which is seen to be identical, apart from the sign, to the form for tan # following equation (3). Equation (28) is derived from circuit 42 of a Fig 6 arrangement. Equation (29) requires a factor K which has not so far been derived; it is obtained by feeding the outputs of circuits 67 and 42 in Fig 6 to amplitudedetectors 55 and 56 respectively to yield +Kk and -Kk, to which are added (K + k), obtained from circuit 27 of Fig 6, in adders 57 and 58. The outputs thereof are (K + k) and (K - k)2 which are fed to square-root circuits 59 and 60 (similar to circuit 32 in Fig 6).

The outputs of the latter are summed in adder 61 to yield 2K, which is reduced to K by attenuator 62. This value is converted to K by squarer 63 and fed to adder 64, which also receives the output of block 34 of Fig 6. The output of adder 64 is thus equation (29). (Instead of including detectors 55 and 56 fed as described, the terms Kk and -Kk can be derived by detecting the output of eg block 34 in Fig 6 to obtain one term, followed by polarity inversion to obtain the other term). The quadrature inputs to multipliers 18' and 19' are likewise obtained from Fig 6 block 34 and Fig 8 block 43 respectively.

The production of the ~-free detection of AP1 but also eliminates # in equation (13) which leads to detection of the weaker signal. This is effected by feeding the modulator output to a single-sideband generator 52 similar to the single-sideband generator 17 in Fig 5.

In this case the multipliers 18' and 19' are fed with quadrature signals obtained from Fig 6 circuit 67, and Fig 6 circuit 34 respectively. The output of subtractor 22' is fed via an amplitude-limiter 53 to a frequency-discriminator detector 54 to yield a pure version of Asp2. Although the strong signal cos(Q2t + AP1) fed to circuit 52 from circuit 51 is devoid of the weak signal cos(#2t + #p2), the information in the latter is contained in the beat signals sin and cos(#2 - #it + QP2 - AP1) fed to the multipliers 18' and 19' in circuit 52. Thus the invention can provide substantially perfect capture detection of the stronger and weaker signals, as required.As in Figs 1, 4 and 5, the amplitude-limiters 53 and 65 provide full FM suppression of the AM noise on the input.

Where detection of the weak signal only is required, the phase-modulator 51 can effectively be combined with the single-sideband generator 52, the phase-modulator 51 itself being omitted. It can be shown that this effect is achieved by feeding multipliers 18' and 19' of generator 52 with the following two inputs instead of those given above, viz and -K3ksin(1t+AP2 -AP1)-K2k2sin2(2-1t+AP2-AP1 )...(31) All four terms in these two equations can be derived by appropriate additional circuitry; in particular the two second-harmonic beat-frequency terms (the second term in each equation) can be obtained via squaring the beat-frequency outputs from Fig 6, circuits 67 and 34.

However, considerable additional circuitry, corresponding to that in the lower half of Fig 10, is required and this arrangement is not preferred.

All the above-described arrangements for improved capture ideally require perfect wide-band quadrature phase-shifters in order to obtain the terms necessary to cancel #, and such arrangements as shown in Figs 6 and 8 should therefore be used to approximate to this ideal. In the adjacent-channel case, however, a simple narrow-band 900 phase-shifter can be used provided the maximum and minimum frequencies are not too far apart. The two quadrature beat-frequency inputs to multipliers 18' and 19' can then be obtained from attenuator 34 of Fig 6, which yields one input direct, the other being obtained by feeding this signal through the aforesaid simple 900 phase shifter (not shown).

This phase-shifted signal also provides the inputs to detector 55 and 56 and to multiplier 18". The remainder of the Fig 6 circuit can be omitted.

As already explained, in the co-channel case the quadrature phase-shifting arrangements of Figs 6 and 8 will produce occasional ambiguities if the +ve and -ve beat-frequencies produce too rapid phase-reversals. If, in consequence, such ambiguities are not resolved, on occasion the ~ term will increase to 2# instead of decreasing to near zero.

Claims (21)

Claims

1. A method of reducing the interference between the detected modulations of stronger and weaker received frequency-modulated (FM) signals constituting together a composite input signal comprising: amplitude-detecting the composite input signal to derive a beat signal of frequency equal to the difference between the carrier-frequencies of the two input signals and including both FM modulations; multiplying a version or versions of said composite input signal with a version or versions of said beat signal to derive a product comprising a first signal at the frequency of one input signal and which includes the original FM modulation thereof; said first signal being thereby either separated from the other input signal, or being accompanied by a second signal at a difference frequency and which is thereby readily separable from said second signal.

2. A method as claimed in claim 1 wherein said composite signal is amplitude-limited before multiplication with said beat signal, and the product is low-pass filtered before detection of the modulation by frequency-discrimination.

3. A method as claimed in claim 1 wherein said composite signal is amplitude-limited and thereafter fed directly and with a quadrature phase-shift respectively to the two multipliers of a single-sideband generator, said multipliers also respectively receiving phase-quadrature separated versions of said beat signal, the subtracted contents of said two multipliers being subjected to detection of the modulation.

4. A method is claimed in claim 2 or claim 3 wherein said amplitude-limiting of the composite signal is omitted.

5. A method as claimed in claim 3 or claim 4 wherein said phase-quadrature separated versions of said beat signal are produced by: squaring the composite input signal, amplitudedetecting the second harmonic of said square, and removing the DC component of said amplitude-detected second harmonic to derive one said version; and squaring said one version, amplitude-detecting said square, and taking the square-root of said sum to derive the other said version.

6. A method as claimed in claim 5 wherein, in order to alleviate phase ambiguities, said other said version is further processed by: detecting the nulls therein, feeding a signal marking alternate nulls to two AND circuits respectively held at opposite logic levels, feeding the AND circuit outputs via an OR circuit to two analogue gates which receive respectively said other said version and an inverted version thereof, summing the outputs of said analogue gates, and utilising said summed output as said other said version.

7. A method as claimed in claim 6 for alleviating phase ambiguities liable to rapid reversal wherein said composite input signal is conventionally detected by frequencydiscrimination and also detected by a method as claimed in claim 4, said two detections being multiplied together and the low-pass filtered product applied to generate a trigger pulse which is added to said signals marking alternate nulls.

8. A method is claimed in claim 1 comprising amplitudelimiting the composite input signal to derive a signal at the carrier-frequency of the stronger signal and having the FM modulation thereof; phase-modulating said amplitude- limited signal to substantially eliminate its phase component constituted by the difference between the phases of the stronger signal and the composite signal; and detecting the modulation of the resultant signal by frequency-discrimination.

9. A method as claimed in claim 8 wherein said phasemodulation is effected by feeding the amplitude-limited signal to a phase-modulator comprising two multipliers to which said amplitude-limited signal is fed directly and with a quadrature phase-shift respectively, said multipliers also receiving respectively phase-quadrature separated versions of said beat signal, to one of which is added a DC component corresponding to the squared amplitude of the larger input signal, the difference of the multiplier outputs providing said stronger signal minus said phase component.

10. A method acclaimed in claim 9 wherein said stronger signal minus said phase component is fed to a singlesideband generator as defined in claim 4 and processed therein by a method as claimed in claims 4, 5, 6 or 7 thereby to provide as output said weaker signal minus said phase component.

11. Apparatus for reducing the interference between the detected modulations of stronger and weaker received frequency-modulated (FM) signals constituting together a composite input signal comprising: means for amplitude-detecting the composite input signal to derive a beat signal of frequency equal to the difference between the carrier-frequencies of the two input signals and including both FM modulations; means for multiplying a version or versions of said composite input signal with a version or versions of said beat signal to derive a product comprising a first signal at the frequency of one input signal and which includes the original FM modulation thereof; said first signal being thereby either separated from the other input signal, or being accompanied by a second signal at a difference frequency and which is thereby readily separable from said second signal.

12. Apparatus as claimed in claim 11 comprising means for amplitude-limiting said composite signal before said multiplication with said beat signal, means for low-pass filtering product and means for detecting the modulation of the filtered product by frequency-discrimination.

13. Apparatus as claimed in claim 11 comprising means for amplitude-limiting said composite signal and for thereafter feeding it directly and with a quadrature phase-shift respectively to the two multipliers of a single-sideband generator, said multipliers also respectively receiving phase-quadrature separated versions of said beat signal, means for subtracting the outputs of said two multipliers and means for detecting the difference between said two outputs by frequency-discrimination.

14. Apparatus is claimed in claim 12 or claim 13 wherein said amplitude-limiting means is omitted.

15. Apparatus as claimed in claim 13 or claim 14 wherein said phase-quadrature separated versions of said beat signal are produced by means for squaring the composite input signal and for amplitude-detecting the second harmonic of said square, and means for removing the DC component of said amplitude-detected second harmonic to derive one said version; and means for squaring said one version, for amplitudedetecting said square and for adding the amplitude-detected output to said square, and means for taking the square-root of said sum to derive the other said version.

16. Apparatus as claimed in claim 15 wherein, in order to alleviate phase ambiguities, said other said version is further processed by: means for detecting the nulls therein and for feeding a signal marking alternate nulls to two AND circuits respectively held at opposite logic levels, means for feeding the AND circuit outputs via an OR circuit to two analogue gates which receive respectively said other said version and an inverted version thereof, means for summing the outputs of said analogue gates, and connections for utilising said summed output as said other said version.

17. Apparatus as claimed in claim 16 for alleviating phase ambiguities liable to rapid reversal comprising means for detecting said composite input signal by conventional frequency-discrimination and means for detecting said signal as claimed in claim 13, means for multiplying the outputs of said two detections together and means for low-pass filtering the product, said product being applied to means for generating a trigger pulse and and a connection for adding said trigger pulse to said signals marking alternate nulls.

18. Apparatus as claimed in claim 11 comprising means for amplitude-limiting the composite input signal to derive a signal at the carrier-frequency of the stronger signal and having the FM modulation thereof; means for phase-modulating said amplitude-limited signal to substantially eliminate its phase component constituted by the difference between the phases of the stronger signal and the composite signal; and means for detecting the modulation of the resultant signal by frequency-discrimination.

19. Apparatus as claimed in claim 18 wherein said phasemodulation is effected by feeding the amplitude-limited signal to a phase-modulator comprising two multipliers to which said amplitude-limited signal is fed directly and with a quadrature phase-shift respectively, said multipliers also receiving respectively phase-quadrature separated versions of said beat signal, means for adding a DC component corresponding to the squared amplitude of the larger input signal to one of said versions, and means for deriving the difference of the multiplier outputs to provide said stronger signal minus said phase component.

20. Apparatus as claimed in claim 19 comprising means for feeding said stronger signal minus said phase component to a single-sideband generator as defined in claim 4 and processed therein by apparatus as claimed in claims 14, 15, 16 or 17 thereby to provide as output said weaker signal minus said phase component.

21. A method or apparatus substantially as hereinbefore described with reference to Figs 2 to 10 of the accompanying drawings.