Commercial Aspects of Single-Sideband, June 1956 Radio & Television News

Prior to phasing-based single sideband generation circuits, a brute force filtering of the unwanted sideband and carrier signals was required. Depending on how well the carrier was suppressed, more than half the total signal power could be lost. According to author Jack Brown in this "Commercial Aspects of Single-Sideband" article from a 1956 issue of Radio & Television News magazine, it had only been since the mid 1940s that wide-band audio-frequency phase-shift networks were even feasible. An ideal implementation of a single-sideband suppressed-carrier modulator (SSB-SC) would result in 100% efficiency, but typical results are in the 80% range. When I built my first SSB-SC modulator around 1990, it was by using discrete 90° power dividers / couplers, sine and cosine local oscillator signals, and a pair of matched mixers. An amplitude adjustment in the one of the paths (I and Q) allowed adjustment for carrier cancellation. At the time there was not a selection of integrated circuits available to do the job with a single part. Since then a plethora of such ICs have appeared on the market that do the job much better than with discrete components.

Part 2. A discussion of single-sideband generation using either of two current systems: filter or phase-shifting.

Typical examples of commercially-available SSB bandpass filters. From left to right: 20 kc. filter by Barker & Williamson. Inc.; a 50 kc. filter made by Burnell & Co., and a 455 kc. mechanical filter made by Collins Radio Company.

In the first article of this series we dealt briefly with the theoretical system gains of a single-sideband versus an AM system, the economic considerations involved in converting an AM system to single-sideband, as well as some of the less extensive technical considerations of single-sideband such as frequency stability requirements and the choice of single-sideband generation systems. In this article the author proposes to delve deeper into the method of single-sideband generation by the two currently available systems.

The filter system of generation was the first one that was technically available for use when single-sideband was under consideration. Since it was shown theoretically that sidebands do exist about an amplitude-modulated carrier, it then became a problem for the engineer to design satisfactory filters which could, by brute force, separate one sideband of an AM signal from its adjacent sideband. If we consider that speech frequencies on the order of 300 to 3000 cycles are necessary for satisfactory communication-type speech systems, the filter problem then is to separate the sideband signals lying either side of the carrier.

To the design engineer this means that the sideband filter must be able to discriminate between signals that are a minimum of 600 cycles apart. For many years this filter requirement could be met only by filters in the high audio or low radio frequency spectrum. Until relatively recent years all sideband filter generation was at or below 100 kc. Filters of the LC variety were in common use in the region of 15 to 60 kc. and quartz crystal lattice filters were used in the region of 50 to 100 kc. These very satisfactorily divorced the two sidebands of the double-sideband suppressed-carrier signal that had to be generated first. See Fig. 1 for a block diagram of a typical filter-type single-sideband transmitter. In recent years improved techniques in filters have made possible the manufacture of filters at higher frequencies than previously considered feasible. It is now possible to build magnetostriction or mechanical filters up to the region of approximately 500 kc. The construction of crystal lattice filters up into the region of a few megacycles is more practicable. The block diagram of Fig. 1 uses what might be a typical filter in the low radio frequency spectrum of approximately 400 kc. The speech signal is fed into the microphone and amplified in the speech amplifier and thence goes into the balanced modulator.

Fig. 1 - Block diagram of a filter-type single-sideband transmitter.

The low-frequency carrier oscillator at 400 kc. also feeds an r.f. signal into the balanced modulator. The balancing action of this modulator stage successfully eliminates the carrier signal at its output yet under application of speech signals at the microphone, a double-sideband suppressed-carrier signal is present at the output of the balanced modulator. The sidebands thus generated lie symmetrically either side of the 400 kc. r.f. carrier frequency. The identical sideband signals are then fed into the sideband filter, the filter discriminates between the sideband signals in that it passes the sideband signals lying to the low-frequency side of 400 kc. and attenuates the sideband lying just to the high-frequency side of the 400 kc. carrier frequency. At the output of the sideband filter we have a carrierless single-sideband signal whose suppressed-carrier frequency is 400 kc. and whose intelligence is contained in the 3 kc. spectrum just below 400 kc. The lower audio speech frequencies will lie closest to the 400 kc. carrier frequency while the higher audio frequencies are correspondingly farther away (to the low side) from the 400 kc. carrier frequency.

Our problem now is to translate the single-sideband signal existing at approximately 400 kc. up to a usable operating frequency of, for example, 3 mc. The single-sideband signal is then fed into a mixer or balanced modulator stage into which is also fed an r.f. voltage from a crystal controlled or stable v.f.o. source which is exactly 400 kc. removed from the desired output frequency. If operation is desired on 3 mc., for example, the oscillator frequency must be either 2600 kc. or 3400 kc. If the 2600 kc. oscillator frequency is used and the 400 kc. lower sideband is heterodyned with this oscillator frequency, a lower sideband signal with a suppressed carrier frequency of 3 mc. will exist at the output of the high-frequency balanced modulator.

If the 3400 kc. oscillator frequency was chosen and the 400 kc. single-sideband signal arithmetically subtracted, the result at 3 mc. would be an upper sideband signal. A pencil, a piece of paper, and a couple of minutes will demonstrate the reason for this. Use actual numbers for the sideband frequency existing for a 2000 cycle audio tone and a suppressed carrier frequency of 400 kc. This simple mathematical manipulation is left to the reader. If the high-frequency mixer is actually a balanced modulator, the oscillator frequency at either 2600 or 3400 kc. can be balanced out. The tuned circuits following the high-frequency balanced modulator must successfully discriminate between the sum-mixture of the sideband and oscillator frequency and the difference-mixture of the same sideband and same oscillator frequency. If the 2600 kc. oscillator frequency was chosen, the tuned circuits following the balanced modulator must pass the 3000 kc. lower sideband signal as well as attenuate the upper sideband signal existing at 2600 - 400 or 2200 kc. In this part of the spectrum two or three tuned circuits will accomplish this successfully. However, if operation in the higher frequency part of the spectrum is contemplated, more tuned circuits or a different heterodyning system must be used. It is often necessary to use multiple steps in heterodyning a low-frequency generated single-side-band signal up to some usable high frequency part of the spectrum, say 20 mc., for example.

It can be seen that we now have existing at a desired operating frequency a single-sideband, either upper or lower, as the operator desires. The problem now is to amplify this single-side-band signal to a high enough level to permit radiation by the antenna system, The amplifier stage or stages following the last heterodyning stage must be some class of linear amplifier. Linear amplifiers will be discussed in detail later in this article.

Fig. 2 - Block diagram of a "fundamental phasing" single-sideband transmitter.

Generation of a single-sideband signal by the phase-shift method has been possible only in recent years. It has only been since 1946 that it has been feasible to construct wide-band audio-frequency phase-shift networks. The successful phase-shift generation of an SSB signal places very strict requirements on certain critical parts of the sideband generator. See Fig. 2. It can be seen in Fig. 2 that two different types of phase-shift network are necessary. The audio-frequency voltages picked up by the microphone are amplified in the speech amplifier and appear in the input to the audio-frequency phase-shift network. This network must meet two very stringent requirements. Throughout the audio-frequency range used (usually 300 to 3000 cps) this network must yield two output audio voltages which are identical in amplitude and frequency but which must differ by exactly 90° in phase relationship. See Fig. 3 for a schematic of a typical audio network. The departure from exact equality of the two output voltages or exact quadrature relationship in phase will cause a deterioration of this generated SSB signal. The v.f.o. or crystal-oscillator signal is fed into a radio-frequency phase-shift network which must fulfill exactly the same conditions as that of the audio phase-shift network. The r.f. network must give two equal output voltages whose phase relationship is exactly quadrature, that is, they must be exactly 90° apart. Since the radio frequency involved is a single frequency, the problem of constructing a practical r.f. phase-shift network to work on a fixed frequency is comparatively simple. A simple combination of R, L, and C can do this quite nicely. See Fig. 4. However, if operated on any frequency other than that for which the r.f. network was designed and adjusted a deterioration of the single-sideband signal will result.

The block diagram shown in Fig. 2 is for a "fundamental phasing generator." By this is meant that the single-sideband signal is generated at the frequency of the desired transmitter output. Thus it can be seen that for each frequency of operation a different r.f. phase-shift network must be used. The dual output of the audio phase-shift network feeds into two separate balanced modulators labeled "A" and "B." The two separate r.f. output voltages from the radio-frequency phase-shift network likewise feed into the previously mentioned "A" and "B" balanced modulators. The r.f. and audio-frequency signals combine in each to produce at the individual balanced modulator output a double-sideband suppressed-carrier signal. Thus at the output of balanced modulator "A" there appears a double-sideband suppressed-carrier signal with carrier frequency of 3000 kc., likewise, at balanced modulator "B" another double-sideband suppressed-carrier signal appears also of 3000 kc., suppressed-carrier frequency. These two double-sideband signals are identical in every respect except that the instantaneous phase-shift relationship of one side-band in each balanced modulator output is exactly 180° opposed.

Fig. 3 - B & W's audio phase-shift network.

At the output of the two balanced modulators "A" and "B" the two sidebands that are out-of-phase will cancel while the other remaining sideband signals will add and reinforce each other at the output of the balanced modulators. Thus by careful manipulation of the phase relationships of sideband signals involved it is possible to successfully attenuate one sideband of an AM signal. With the fundamental frequency SSB generating system if operation on other than the alignment frequency is desired careful realignment of the r.f. phase-shift network is necessary to operate satisfactorily on the new frequency. It has generally been found satisfactory to design a radio-frequency phase-shift network that yields a constant r.f. 90° phase-shift between the two output r.f. voltages for varying frequency input but the two amplitudes of the individual outputs vary in direct proportion to the operating frequency. If a certain design tolerance is permitted in the variation of operating frequency so as not to degenerate the quality of the single-sideband signal, a certain frequency variation about the design center frequency may be used. If we restrict the change in operating frequency to ±2 1/2% of the operating design center frequency it will be found that (all other phase-shift and amplitude relations in the single-side-band generator being perfect) the unwanted single-sideband suppression will be degraded to a value of 37 db. This value of 37 db is generally considered acceptable. However, it must be remembered that all other phase-shift and amplitude regulations must be perfect in order to get 37 db sideband attenuation.

In general, this will not be the case and further degradation of the sideband suppression will result. For a commercial-type service where operation only on certain fixed frequencies is contemplated, the fundamental phasing type single-sideband generator would appear to have some merit. Since it is possible to build very stable components, the r.f. phase-shift remains relatively constant with time and temperature variations. It would appear that this would be a most economical method to use. Where a number of fixed frequency channels are required in commercial service the individual channel r.f. phase-shift networks could be made part of the channel selector switch which would normally change the frequency control system of the basic transmitter. It is also possible to build precision audio phase-shift networks to hold an angular accuracy to considerably less than 1°. The fundamental phasing sideband generation system would appear to be a "natural" for converting an existing AM transmitter to single-sideband type transmission. See Fig. 5.

Fig. 4 - Radio-frequency phase-shift network.

Fig. 5 - How an AM transmitter can be converted to SSB operation. Refer to text.

The normal frequency control components and stages of the AM transmitter in question are used to furnish the r.f. signal indicated in the block labeled "xtal. oscillator" in Fig. 2. The r.f. from the exciter stages would be fed into the r.f. phase-shift network and thence into the two "A" and "B" balanced modulators. It would appear uneconomical to make general use of the audio-frequency equipment in an existing AM transmitter since only one or two stages of audio amplification are necessary to drive the precision audio phase-shift network also illustrated in Fig. 2. The output of the dual balanced modulators after a stage or two of linear amplification in the external single-sideband generator would then feed the single-sideband signal back into the grid circuit of the existing AM transmitter. It would be necessary to make appropriate changes in the AM transmitter final amplifier to insure that it would operate as some form of linear amplifier rather than in its former class C operation. This would necessitate readjusting the operating bias on the control grid of the final amplifier to a value that would permit operation of the tube as a class A, class AB₁, class AB₂ or class B linear amplifier. The tube characteristic should be consulted for proper operating bias. If the transmitter output tube is a tetrode the screen voltage applied must be stabilized by using voltage regulation either of the vacuum-tube or gaseous-tube type. The converted AM transmitter will perform satisfactorily as a single-sideband transmitter. The normal AM speech amplifier and class B modulator stages will be unused and appropriate measures can be taken to disable them to conserve power drain on the transmitter power supplies. One manufacturer markets such a unit for converting amateur transmitters to SSB if they fulfill certain minimum requirements. Admittedly this is for the restricted range of the amateur bands, however, it would appear that the same techniques would be quite usable for the commercial services where a given number of fixed-frequency channels are used. This should be a very good interim method of obtaining single-sideband without the pain of discarding existing AM equipment.

In the preceding discussion brief mention was made of linear amplifiers. This is a subject which, in itself, could occupy many thousands of words and yet not be satisfactorily covered. It is not the intention of the author to delve deeply into the general subject of linear amplifiers. A linear amplifier is one whose output is an amplified replica of the grid input signal. The most commonly encountered linear amplifiers are the ordinary class A amplifiers in an audio system or the r.f. amplifiers employed as r.f. stages in the front end of a receiver or the high level modulator stages used in currently-operated AM equipment. The linear amplifiers used in single-sideband amplification work on the same principle except that in all cases tuned circuits or r.f. coupling devices are used as interstage and output coupling units. Since a single-sideband signal once generated is made up of amplitude variations, any amplification of the signal following generation must be of a linear nature to faithfully reproduce all of the amplitude variations involved. If linear amplification is not attained spurious adjacent channel signals are generated in the amplifier circuits themselves creating unnecessary adjacent channel interference to other services and contributing nothing as far as the intelligence transmitted by the offending station. The most common causes of non-linear distortion in amplification are: 1. Improper adjustment of operating bias on the grid of a linear amplifier; 2. Improper plate loading of the linear amplifier, and 3. Too large a driving signal causing saturation of the operating characteristics.

Anyone or all of these factors will cause serious distortion products which will cause adjacent channel interference. Several articles have appeared^2,3,4 in recent technical literature which outlined the requirements and certain techniques to be used in linear amplification.

From the preceding discussion it can be seen that the fundamental phasing generating system is not feasible for general coverage of the high-frequency range. If general or continuous cover-age of the 2 to 30 mc. high-frequency range is necessary it would appear most reasonable to make use of heterodyning techniques in order to cover the spectrum properly. This involves genera ting the single-sideband signal by either the filter or the phasing system at some fixed frequency and by either single or multiple heterodyning steps to cover the high-frequency communication spectrum. Heterodyning a fixed-frequency, single-sideband signal to any other part of the radio-frequency spectrum cannot be done without careful consideration of the frequencies involved, the selectivity of the tuned circuits following the heterodyning stage, and the possible combinations of the heterodyning frequencies. Spurious mixture outputs are the result of heterodyning of the harmonics of the two signals being fed into the mixer or balanced modulator stage. It is possible to have present at or near the desired output frequency a spurious signal either c.w. in nature, or a single-sideband signal of either upper or lower sideband. Very careful consideration must be given to this matter in the initial design of any single-sideband equipment. It can be seen that the one big advantage of the fundamental frequency generating system, using the phasing technique, is that no spurious mixture products are encountered since the single-sideband signal is generated at the fundamental output frequency. The design engineer must then decide which system is most applicable to the particular problem with which he is concerned. If the heterodyning system is to be used it is suggested that the design engineer consult the spurious mixture product charts published by Badessa⁵ and Eberhardt.⁶ These two articles will give the design engineer some very handy tools for designing the heterodyning system to be used. The author can not over-emphasize the importance of the design characteristics that must be considered in a heterodyne system.

The single-sideband equipment design engineer is faced with varying and somewhat difficult problems involving, in general, the following:

(1) Frequency stability. This would appear to be one of the most difficult problems to solve particularly when the service to be rendered must be of the highest quality. Where high quality continuously stable service is to be rendered, frequency stability on the order of ±5 cps might possibly be required. When considered at a carrier output frequency of 20 mc., this stability might be difficult to attain. It would appear that some system of automatic frequency control or frequency synthesis would be necessary. There are various methods of obtaining a frequency-synthesized master frequency⁷. Good high-frequency quartz oscillating crystals are currently available with an inherent stability on the order of 1 part in 10 per day. This is assuming that the quartz crystal is pre-aged and in a temperature-controlled oven. If operation below 10 mc. is contemplated, the frequency stability problem assumes proportions of a somewhat minor nature. There are many services that do not operate above 10 mc. and therefore the frequency stability problem is not as severe as previously considered.

(2) A second problem facing the engineer would be the design of adequate sideband generating filters. Conventional filter design can be used for filters from the high audio frequencies through approximately 100 kc. using LC type networks. For frequencies above 100 kc. it will probably be necessary to consider some means such as quartz crystal filters, or magnetostriction or mechanical-type filters. There are currently available on the commercial market filters of each type which might be considered by the equipment design engineer. If the engineer wishes to design a suitable single-sideband filter he must keep in mind that the slope or skirt selectivity of the filter used must be extremely good so as to separate the low-frequency speech components on one side of the carrier from the low-frequency speech components on the other side of the carrier frequency. This means, in general, that the skirt selectivity characteristic of the filter must drop at least 60 db in one kilocycle. This is a rather strict requirement placed upon the filter but it is possible to design and build filters to do this satisfactorily.

(3) The third problem that might confront the engineer is the design of suitable audio- and radio-frequency phase-shift networks. It is suggested that the design engineer consult the original paper by R. B. Dome⁸ as well as those by Saraga⁹ and Luck¹⁰. The design engineer must select his parameters carefully and make sure that he is considering the proper audio-frequency bandpass to be transmitted. In general, the speech spectrum from 300 to 3000 cps is satisfactory for the male voice in communication service. However, in other classes or higher grades of commercial service it might be necessary to consider a wider band of audio frequencies to be transmitted. This places stricter requirements on the audio phase-shift network and makes the design procedure more complex.

(4) A fourth problem confronting the design engineer is one already mentioned, this is the heterodyning system. The spurious mixture products that appear in the heterodyning process can be of a very serious nature. If not carefully considered, the spurious mixture products can cause interference to channels which might, conceivably, be considerably removed from the desired operating frequency. Once a design is assumed and a model built, the output should be carefully checked by a continuous-coverage communications receiver loosely coupled to the generator output to determine if spurious mixture products are being radiated.

(5) The fifth problem to be considered is that of proper linear amplification. If a high quality single-side-band signal has been generated either by the filter or the phasing system the signal can be seriously deteriorated by poor design and operation of the linear amplifiers following the generator. It would be foolish to generate a superior single-sideband signal and later cause it to deteriorate by poor linear amplification. The final criterion of the single-sideband suppression is, in most cases, the degree to which the amplifiers remain linear and do not generate non-linear distortion products in the amplification system. The use of speech compression, speech limiting, or an automatic gain control system is worth considering. The use of negative feedback around the r.f. linear amplifiers is also worthwhile and has been adopted commercially by at least one manufacturer." Grounded-grid amplifiers, in themselves, provide a certain amount of negative feedback as an inherent part of their amplification system and might well be worthy of consideration by the equipment design engineer. The problems facing the SSB design engineer are difficult but not insurmountable.

2. Lund, Rose & Young: "Amplifiers for Multichannel Single Sideband Radio Transmitters," Proceedings of the I.R.E.; July, 1952

9. Saraga: "The Design of Wide-Band Phase Splitting Networks," Proceedings of the I.R.E., July, 1958

10. Luck, D. C.: "Properties of Some Wide-band Phase Splitting Networks," Proceedings of the I.R.E., February, 1949

Commercial Aspects of Single-SidebandJune 1956 Radio & Television News

Commercial Aspects of Single-Sideband
June 1956 Radio & Television News