HOMODYNE AND SYNCHRODYNE

8. Separating Overlapping Amplitude-Modulated Signals

Although the ordinary conception of separating signals in a spectrum by means of filters leads to the idea that when the upper sideband of one modulated signal overlaps the lower sideband of another signal, then interference between the signals is inevitable, this is, in fact, far from true. In the simplest case, for instance, when the overlap is small, interference can be avoided by sacrificing that part of the modulation frequency band corresponding to the overlap--i.e., the higher-frequency part. Even better, if r.f. filters are used which accept each signal except the overlapping portion, then on detection the whole modulation frequency band of each signal is obtained separately, the only distortion being a falling-off of up to 6 db over that portion of the band where only one sideband has been accepted. This can easily be corrected if desired, but will often be unimportant.

However, when the overlap is considerable, and especially when the sideband of one signal overlaps the carrier of the other, r.f. filtration is impracticable or useless. In such cases, the synchrodyne technique can be used to obtain theoretically perfect separation of the two signals. Two methods have been published, one by Cabrilovitch as a patent specification with a convention date of September 1936 and a date of acceptance of April 1939, and the other by the present author," in 1948. Both methods use the synchrodyne principle in the first stage as a rejector instead of the usual acceptor. Gabrilovitch's system rejects, in this stage, the signal which is ultimately wanted; the author's system rejects the unwanted signal. Unfortunately, Gabrilovitch's patent is very difficult to understand in detail, so the account given here is really the author's personal interpretation of the main principles, and the suggested schematic arrangement does, in fact, differ considerably from that given by Gabrilovitch; the latter, in the author's opinion, will not work. This may account for the fact that Gabrilovitch claims that his system will separate signals whose sidebands overlap the adjacent carrier by a large amount, whereas the author thinks that Gabrilovitch's system, even as re-arranged, will not work unless the overlap of the sidebands is restricted so that they do not overlap the adjacent carrier. On the other hand, in the author's own scheme, the spacing between the carriers of two adjacent signals need not exceed one-half of the highest modulation-frequency, although this system has its own limitation in that it can deal with only one overlapping signal at a time. The author's version of Gabrilovitch's system is shown in block schematic form in Fig. 9, which should be largely self-explanatory.

The first synchrodyne path acts as a rejector because the local oscillation is in quadrature with the incoming signal which is to be rejected. The second path accepts all incoming signals. One of the paths is then given a phase shift of 90 deg, and on combining the two paths suitably, all signals cancel out except for the wanted signal which was contained in only one path. This signal is undistorted and free of the signals which overlapped it. It is clear that the wanted signal must not overlap the carrier of any adjacent signal, because, if it did, the spectrum of the lower sideband of the output of the low-pass filter in the first path would be doubled back on itself, and the 90 deg phase shift would have the wrong· sign for the doubled-back portion.

In the author's own scheme, shown in outline in Fig. 10, the signal rejected by the first stage is actually the unwanted signal whose spectrum overlaps the wanted signal. The wanted signal then remains on a carrier equal to the difference between the carriers of the wanted and unwanted signals. If its modulation-frequency band is greater than this difference, then the lower sideband is doubled-back on itself. This signal is demodulated by first filtering out the doubled-back band with a high-pass filter, and then applying the remaining signal (which is now a vestigial-sideband signal) to a synchrodyne stage, arranged as an acceptor, with the oscillator synchronized, in phase, to the new carrier frequency. The output of this stage is the wanted modulation-frequency band, complete, but with a small distortion of frequency response due to the inclusion of the contributions from two sidebands over the very low-frequency part of the band. This can be equalised if desired.

Synchrodyne principles can also be used to enable a system to be made in which two different channels of information occupy the same frequency band and use the same carrier frequencies. The channels are separated by virtue of the fact that their carriers, although of the same frequency, are in phase quadrature. If the carriers are transmitted, then at the receiver a local oscillator can be synchronised to their resultant, and from this, two quadrature carriers can be obtained by means of phase-shifters for use in demodulating the two channels into separate output circuits. The basic principles of this "carrier-phase duplex" or "two-phase" system were set out (except for the synchrodyne application) by Nyquist in 1928, and a fuller discussion of it was given by the present author in 1948. It is used in the American N.T.S.C. colour television system on the synchrodyne basis.

NEXT PAGE : The Coherent Detector