Appendix F: Getting Excited over FM

Now an A.M. transmitter is a relatively complicated device, driven by a simple one called a 'drive'. All the drive has to do here is produce a nice, steady continuous wave at the right frequency - all the tricky modulation stuff is done inside the transmitter.

In F.M. it's all different. Here the transmitter only amplifies the R.F. presented to it by the drive, or 'exciter' as it is sometimes called - the modulation has to be done as far back as the master oscillator. Properly speaking, this 'drive/exciter' is actually a small self-contained transmitter in its own right, producing at most a few tens of watts. If this is all you need, it can be coupled straight to the transmitting aerial (and is, at some low-power sites). That large grey box that everyone calls a transmitter is actually solely an amplifier, stepping the power up to a more reasonable 10kW or so.

It follows that in an F.M. system, the 'transmitter' (i.e. the amplifier) is a fairly straightforward piece of gear, albeit large and hot. Most of the complexity lies in the 'drive'.

The S.T.& C. drives were early examples of the genre, designed in the days when there were fewer options for crafty circuitry. All the same, they managed to make some parts of it quite devious.

The actual frequency modulator was a straight L/C oscillator (Hartley as I remember). Its output was at a fixed centre frequency of 7.5MHz, irrespective of what the actual radiated frequency was to be. A sample of the oscillator was phase-shifted by 90-degrees and fed to a balanced modulator, which was also fed with the audio programme signal. The RF current from this modulator was then injected into the tuned circuit of the L/C oscillator.

This current would have the effect of a variable reactance (sometimes inductive, sometimes capacitive, depending on the programme modulation). The effect would be to vary the resonant frequency of the tuned circuit in sympathy with the programme modulation - true frequency modulation in fact. All quite simple.

What was not quite so nice was that the frequency stability of a straight L/C oscillator used in this manner is not particularly good - certainly not good enough for a broadcast transmitter. In order to get the thing into spec, some sort of frequency controller would be needed. And that is where it all started getting rather tricky.

Having corrected your carrier frequency (and I'm skirting round that for a bit) the 7.5MHz signal would then be mixed with a signal derived from a crystal oscillator, to produce an output at one-third of the radiated carrier frequency. There would be two complete drive systems, to provide redundancy, and one of these would be selected and then split two ways. Each feed would then drive each of the two transmitters for the service via a phase corrector network (another horrid motor-driven device) and a tripler, to reach the required output frequency.

I just can't put off this frequency correction business any more. What they did was to take a sample of the 7.5MHz signal, frequency-divide it down by 500 to an output frequency of 15kHz, and then phase-compare it with a local 15kHz signal derived from a bar quartz crystal. The phase comparator would then drive a bi-phase motor connected to a gear train, which would rotate the main tuning capacitor of the master oscillator in such a direction as to correct any frequency error.

Sounds a doddle if you say it quickly enough. The real fly in the ointment was this 500-divider contraption. Frequency multiplication is fairly easy to do: division was by no means as easy in the days when this equipment was new. And the way they did it was a little hair-raising.

Firstly, make the problem simpler by breaking the division ratio down. Dividers by 5, 5, 5 and 4 will do the job if tandemed up (5x5x5x4 = 500).

What they ended up with was a device called a regenerative divider. Consider the divide-by-five part (the divide-by-four worked in a similar manner). Suppose we want an output frequency of F from this divider. Then obviously the input frequency will be 5F. So feed this 5F into a mixer, with a locally derived signal at 4F. Amongst all the unwanted output products of this mixer will be a wanted one at frequency F, so pick it out with a tuned circuit. There's your output signal: all we need now is to find some 4F. Simple, just quadruple the output signal and Bob's your uncle.

Percipient readers will have spotted the chicken-and-egg situation into which we have wandered. To get an F output we need a 4F signal, but we can't make a 4F signal without an F output. Clearly the thing won't work.

Amazingly, it does. Apparently, when the mixer isn't working, there is enough noise being generated by it (mixers are notoriously noisy devices, particularly the earlier ones) to produce some noise centred on F, which in turn produces noise centred on 4F. This highly-coloured noise is enough to start the divider working.

Well, most of the time, anyway. Correct operation does presuppose proper mixer and multiplier gains, and since this equipment was totally valve operated (and plenty of them) at any given time at least one valve would be feeling rather tired, and the entire divider could give up the ghost. At this point, all of the valve feeds would change anyway, so it was difficult to spot the culprit. To stave off this unhappy occurrence, it became the custom to check the valve standing currents on a daily basis, and replace anything that looked suspect. The only alternative was to let it fail on you, whereupon you could then either test every valve till you found the bounder (very time-consuming), or replace the lot (very expensive).

Generally it was better to avoid the trouble rather than fix it. If you were unlucky enough to suffer a component failure in the dividers, quite often the entire divider would require setting up from scratch afterwards - a lengthy business, bearing in mind the number of tuned circuits, their interaction, and the possibility of getting a particular stage tuned to the wrong division ratio. These units were by no means popular.

(Before we leave these dividers for good, I may remark that when integrated circuits that could perform digital division became available in the 'sixties, no time was lost in producing a prototype replacement for the regenerative dividers. This worked like a charm and all regenerative units were replaced with the digital types throughout the BBC. These gave no trouble at all until the drives were heaved out in the 'eighties).