The International Telecommunications Union, which seeks uniformity and agreement among nations on communication needs, has recommended discontinuance of AM in favor of single sideband (SSB) for maritime operations. Several nations have already converted to SSB exclusively. The United States has established deadlines for conversion of voice communication to SSB in the 4- to 22.5 MHz (high-seas) band. After 1969, no new AM transmitters will be licensed. After 1974, no AM transmitters may be used at all.
This impending change in maritime radio communication to SSB is going to result in a major overhaul of the equipment used. One area to be affected more than most is that of frequency accuracy. The reason for this, of course, is that in SSB the carrier is not transmitted; so the receiver must reinsert a carrier within a few cycles of the frequency of the carrier that was eliminated at the transmitter. To do this, both receiver and transmitter must have accurate frequency generating equipment.
Most maritime communication in the HF range has been by AM. Frequency stability requirements for AM are relatively loose - 15 to 50 parts per million (ppm) for coast stations and up to 200 ppm for shipboard equipment. This has permitted the use of a separate crystal for each channel, and changing channels meant simply switching crystals (and perhaps oscillators, for widely separated frequencies). With the change to SSB, however, frequency control becomes more critical. Attempts to simply update the old approach, a crystal for every channel, run into a new set of problems, and it behooves one to take a close look at the various alternatives. There are three approaches that may be considered:
1. Crystal control, with a separate crystal for each channel or set of harmonically related channels. 2. Frequency synthesis, deriving all desired frequencies from a single reference oscillator. 3. Crystal control by crystal mixing methods.
Approach 3 is impractical because of the 0.1-kHz channel spacing, the wide range of frequencies covered, and the fact that only a small percentage of the available frequencies are used.
The FCC requirements for frequency accuracy in the SSB mode are plus or minus 20 Hz for coast stations and plus or minus 50 Hz for shipboard operation. It would be logical to design all equipment to meet the coast station requirements, thus providing additional margin in shipboard operation for the more stringent environments and for fewer frequency calibrations. This plus or minus 20-Hz requirement thus converts to 10 ppm at 2 MHz and 0.73 ppm at 27.5 MHz. The problems at the low end of the range, therefore, would be not much worse than for the AM equipment. It's at the high end of the frequency range that stability becomes a problem, at least in approach 1.
To meet the accuracy requirement of 0.73 ppm will require stability versus temperature on the order of 0.2 ppm. This assumes that temperature variations will have a greater effect on frequency than other conditions, such as shock, vibration, humidity, line voltage variations, etc. This has been found to be generally true. Restricting variations due to temperature to 0.2 ppm leaves 0.5 ppm for aging. With a good crystal, aging rates of approximately one-half ppm per year can be achieved. This trade-off between temperature and aging is subject to debate, of course, but any relief for the designer by loosening the temperature specification puts that much more burden on the user by requiring frequency calibration more often.
Achieving an accuracy of 0.2 ppm over an appreciable temperature range is not too difficult in any given oscillator, using either temperature compensation or temperature control. It requires a reasonably good oven, on the order of a few tenths of a degree over the temperature range, or a reasonable amount of care in compensating the crystal; but it has been and is being done and is quite feasible. It might even be practical to use this approach of separate temperature-controlled or temperature-compensated oscillators for operations requiring only a few channels, up to six or eight. When channel requirements are greater than this, separate temperature-compensated oscillators are no longer practical. Separate temperature-controlled crystals also begin to become impractical, because one large oven would be required for the crystals and oscillator or oscillators. Maintaining the required temperature stability becomes more difficult and more expensive, because of the inevitable temperature gradients across the oven. If a separate oscillator is not used for each crystal, then care must be taken to ensure that the crystal switch does not affect the frequency.
Compounding the problem is the fact that at the higher frequencies where the accuracy requirements get more stringent, the crystal and oscillator stability become poorer, both with temperature and with time. Add to these difficulties the problem of periodically calibrating the frequency of each crystal at random frequencies spread over a wide range and it becomes evident that, as the number of channels increases, the advantages of the multiple-crystal approach decrease.
With the frequency synthesizer approach, the opposite is true. The advantages of this approach increase as the number of channels increases. Here are the advantages:
a. Only one reference oscillator is needed for any number of channels. It would be practical, therefore, to build better stability into that single oscillator, along with a better crystal, than could be done with a multitude of crystals and oscillators. b. The same percent of stability is available on all channels. The lower frequency channels thereby benefit by the high frequency stability requirements. c. All channels are available at any time. No changing of crystals or waiting for new crystals if a frequency assignment is changed or added. Also, no requirement for recalibration on a new frequency assignment. d. Only one oscillator need be calibrated and that could be checked quickly, at any time, by tuning to any of the standard frequency transmissions in the HF band. These standard frequencies are broadcast by most of the major countries and are available practically anywhere in the world. e. No switching of the crystal or reference oscillator is required. They remain undisturbed during channel switching. f. Performance capability on all channels can be upgraded by changing only the reference output.
These advantages exist for any number of channels greater than one. At some relatively small number of channels, probably in the neighborhood of 20 to 30, cost also becomes a factor favoring the frequency synthesizer, as indicated in Figure 1. Where this crossover occurs will depend on the quality of the separate crystal oscillators. If each of these oscillators is designed to compete in stability and aging with the reference oscillator in the synthesizer, the crossover will occur sooner than if the design of the separate crystal oscillators is economized. Regardless of where the crossover occurs, beyond that point there are no longer any advantages favoring the separate-crystal approach.
The simplified block diagram in Figure 2 shows the basic elements of frequency synthesizers. A tunable voltage controlled oscillator (VCO) can cover the desired range or can switch in steps to cover various bands in the range. Selection of the desired frequency is accomplished by a coarse adjustment of the VCO to approximately the desired frequency and by selecting the necessary division ratio in the variable frequency divider. This is all done automatically by the channel selection switches.
As the output signal, obtained from the VCO, is phase-locked to a reference frequency derived from the reference crystal oscillator, the frequency stability on any channel will be equal to that of the reference crystal oscillator. In present Collins equipment, two choices of reference oscillators are provided: a temperature-compensated oscillator with frequency stability of 1 to 5 parts in 10 to the 7th, depending on the temperature range; and a temperature-controlled-oven standard with stability of 1 part in 10 to the 8th.
The operation of a temperature-controlled frequency standard is probably well understood. To stabilize the frequency by temperature control, the crystal is placed in a temperature-controlled oven and the temperature is adjusted to the turning-point temperature of the crystal where there is a minimum frequency change with temperature. The degree of frequency stability achieved is then a function of the temperature control of the oven and how close to the turning point the temperature is adjusted. Thermostatic temperature control can be used where stability requirements are not too high. The on-off cycling of the thermostat, the temperature differential between on and off, aging of the spring tension and contacts, and (usually) lack of provision for temperature adjustment limit the frequency stability obtainable with a thermostatically controlled oven.
Higher precision crystal frequency standards almost invariably use a proportionally controlled oven.
In either temperature-controlled or temperature-compensated reference oscillators, the nominal frequency is usually chosen in the 2.5- to 5.0-MHz region. Lower frequencies require larger crystal holders and crystal blank sizes more susceptible to shock and vibration. Above 5 MHz the aging rate begins to increase, unless an overtone crystal is used. Higher frequencies also mean dividing further to get down to the reference frequency.
To summarize, the frequency stability requirements for single sideband have been met with increasing ease over the past few years. Usually, it has been found advantageous to generate the required frequencies by means of a reference oscillator and frequency synthesizer because of the relative ease of attaining the desired stability on all channels, immediate availability of every channel, lower cost in multi-channel applications, and ease and simplicity of providing for frequency calibration of all channels.