Nu ) ) (Z J,: R' (0 RW P' and G' (0 : 2 GFT™ '
Nu ) ) (Z J,: R' (0 RW P' and G' (0 : 2 GFT™ '
Nu ) ) (Z J,: R' (0 RW P' and G' (0 : 2 GFT™ '
. I was disappointed tofindthat the two methods of transmission were so similar in this respect and I wonder if the authors have other reasons for thinking that the quadrature signal method is, in fact, the better with regard to linear waveform distortion. Professor D. G. Tucker: Messrs. Gouriet and Newell state that Levy's paper is different. The only difference seems to be that he used reflections whereas the authors use taps. The reflection method is more restrictive with respect to the amplitude of the various components, but I cannot see any other difference in principle. Mr. B. M. Sosin: Referring to the paper by Messrs. Gouriet and Newell, there is one important point which has not been mentioned. If the vestigial-sideband signal is produced at the high power output of the transmitter, the limitation of the double-sideband transmitter as against the offset carrier transmitter must be considered. The power of the transmitter depends on its boundaries. For a double-sideband transmitter there is at least a 50% wider band, and the power is normally three-quarters to two-thirds of that obtained with the offset carrier. Mr. J. M. Layton (communicated): If, as Mr. Howson and Professor Tucker assume, r(0 is a function of carrier voltage alone, all circuits considered are linear and a cisoidal input e^?1 may be assumed (as in normal circuit theory), *t instead of the authors' cos u>gt. All the results of the paper may then be obtained in a more general form, unshackled by the assumption of resistive terminations, without in any way complicating the analysis. Indeed, when dealing with rectangular wave variation
283
of r(0, this method, suitably developed, renders the whole of Section 6.2 unnecessary. For instance, in the series modulator, if the source voltage is , then
2 gft*'
_ oo
(n odd), the condition for the absence of non-zero even-order sideband components in i is found to be nu>p)] = (ZO - I ) * oc rj, Zn =
'0'
where Z = ZR + Zs + r0 is the total mesh impedance apart from r'(0- The constant of proportionality in this relation gives Z o l/i 0 and, if Z o is known, this gives i0. The other current components are found to be given by / = g'n(\ Zo/O). In the parallel case considered by the authors, r'(/) is a rectangular wave of amplitude \(rb r/) and therefore r 'n\s'n = i(rb ~ r/)2> s o tri at Zn must be constant (independent of n). In practice this means that Zn is resistive for all n, owing to the difficulty of designing a circuit to have a constant vector impedance at all odd sideband frequencies (from oo to -|-oo) unless its phase is zero. This makes Z o l//0 positive real: if, however, Z o is complex, so also is /0, and the supply-frequency current component is phase shifted with respect to the supply voltage.
that this frequency would be chosen for a conventional cable circuit. We do not disagree with the definitions of 'single-sideband' and 'vestigial-sideband' as stated by Mr. Jacobsen and have used these terms not in order to suggest that they are synonymous but to stress that the treatments of both systems are identical by the method described in the paper. Single sideband is the limiting case of the vestigial sideband as regards the length of the delay line required. The question of signal/noise ratio has been effectively dealt with by Mr. Monteath. The method suggested by Dr. White for halving the length of delay line required necessitates the use of a loss-free line. This method prevents the possibility of compensating for loss along the line by the adjustment of the taps because the loss will not be the same at any one tap for the direct and the reflected signal. The use of a transversal filter, for the removal of the unwanted components of a double-sideband signal, must be restricted to low power because of the inherent insertion loss of such filters. The problem therefore remains of obtaining efficient, and sufficiently linear, power amplification of a vestigial-sideband signal. We thank Dr. Maurice for his interesting contribution. The criticism by Mr. Sosin is most important and if, in fact, it is true that the transmitter power is reduced to twothirds as a result of increasing the bandwidth by 50%, this is sufficient to nullify any economic advantage in the method we propose. As Professor Tucker states, the main disadvantage of the system described by Levy is the restriction imposed by the reflection method. These reflections must be kept small in order that secondary reflections are kept to reasonably small amplitude.
284
This aggravates the problem of subtracting the two direct components in order to obtain the relatively small quadrature component. In addition the appropriately delayed original signal is not so readily available as in our method. Mr. D. P. Howson and Professor D. G. Tucker (in reply): In reply to Mr. Monteath, the assumptions we have made are basically the same as those made by Peterson and Hussey, and the generality of our approach is therefore the same. However, (like several other authors), we have extended the scope very greatly to consider more complex circuits, i.e. the ring modulator, and more complex terminating arrangements. The use of the equivalent circuit has no merit other than that of giving a more easily appreciated illustration of the circuit properties than is given by the equations; in all cases the equivalent circuit is deduced from the equationsand not vice versaso that it is not very suitable as a basis for calculation. Moreover, in our
cases the circuit would have to include an infinite number of branches. There is no doubt that our main interest in avoiding even-order modulation terms is to eliminate them from the equations in order to make solution simple, and therefore we may often eliminate even-order currents or voltages without necessarily eliminating both together. It is not possible to show that, in general, avoidance of even-order products leads to maximum efficiency. There are obvious practical advantages, however, in not having frequency components such as 2o>p cog to limit the available bandwidth of the cop + coq product. We thank Mr. Layton for his interesting contribution, and agree that the use of the cisoidal-signal representation has advantages. Whether his method is as convenient as ours in solving the various problems we have discussed is not at present clear. We look forward to seeing his work in more detail.