Each transmitter consisted basically of three cubicles, close-bolted: three such transmitters, one for each service, formed a continuous row of units. There were two such rows, separated by a short corridor for rear access.

Working from left to right, the three cubicles contained

• Power supply and control systems, and a drop-coil mains regulator
• Stages 1 and 2 of RF amplification
• Stage 3 of RF amplification.

The RF circuits were quite straightforward. Stage 1 was a grounded-cathode tetrode amplifier: stages 2 and 3 grounded-grid triode amplifiers. All stages used a resonant-line system for the anode circuit. This consisted of a closed box, divided about three-quarters of the way up by a horizontal partition. One side of the box was a hinged door, held shut by thumbscrews, giving access to the valve and stage components.

The anode line was a vertical copper tube centred in the lower section of the box, and the air-cooled valve anode sat in the top of said tube. For the grounded-grid stages, the grid-ring of the valve made contact with a set of fingers located on the partition, and the upper part of the valve poked through into the upper box section, containing the input tuned circuitry. The anode circuit was tuned by a short-circuit loop which could be adjusted in position, so as to alter its coupling into the cavity - the usual mode of explanation as to how this worked was that the loop 'destroyed some inductance' in the equivalent circuit of the cavity, thus altering its tune frequency. Output power was taken from the cavity by another loop, also adjustable in position.

Unusually these stage cavities were mounted in the rear of their cubicle, and accessed from the back of the transmitter. All controls and metering for the stage were however conventionally placed on the front of the cubicle. This meant that an operator had to tune each stage by some method of remote control. I think that most of the staff felt that it would have been a good thing if S.T.&C. had kept things simple, and used remote operating shafts or possibly Bowden cables to do the tuning: what we got instead was a motor-driven system. Small reversible motors drove reduction gearboxes and thence a crank: this operated a pull-rod to another crank, which rotated the tuning loops through a limited arc. A Post-office type switch key on the unit's front controlled the motors, with a pushbutton for motor braking. There was a variable resistor coupled to the drive shaft that operated a meter on the front of the transmitter, to give the operator some idea of where the loop was positioned. Altogether it was an overcomplicated system that gave much trouble. The main problem was that over a period of time both gearboxes and loops became stiff. The main output loop could on occasion become so stiff that the coupling pins on the connecting rods would shear, or a loop would slip on its shaft and start moving through an incorrect arc. A lot of time was wasted freeing up or refurbishing these systems.

The control system for this design was distinctly odd. Most transmitters are burdened with contactors: supply contactors, and two- or three-step sequencing contactors for both filament and H.T. supplies. The purpose of these was to ensure that at switch-on, excessive surge currents didn't flow. A cold valve filament has a very low resistance compared to its running value, often only one-tenth or so, and excessive currents will flow at switch-on, which can damage both valve and supply circuits. So the current is limited for a short period, either by switching-in a reduced voltage at start and later switching to a higher voltage, or by using a series resistance, which after a period is shorted-out. Similar conditions apply to the HT. Circuits - massive surges can occur if full volts are immediately applied to the smoothing capacitors, damaging rectifiers and other components, and so a similar step-switch is used. Additionally, there are filament-delay devices: applying H.T. volts to a cold valve can much shorten the life of the valve, so one has to wait for anywhere between thirty seconds and ten minutes (depending on valve design) before HT can be applied to a powering transmitter.

The S.T.& C. design avoided all of this - there were no contactors at all and no filament delay as such. Instead, the transmitter was effectively driven from a mains supply that could be slowly turned up from zero to operating voltage over a period of about one minute. This was the purpose of the 'drop-coil regulator' mentioned earlier.

This drop-coil regulator was a feared device. It was basically an oil-filled tank transformer with separate primary and secondary magnetic circuits. The coupling between these circuits was varied by the 'drop-coil' - a short-circuit coupling loop that could be raised or lowered within the tank The raising was done by a super bicycle-chain which was driven via a sprocketed wheel by a motor-driven gearbox. In this mechanical circuit was the device that caused most problems with these transmitters - a magnetic clutch. The purpose of the clutch was to enable the drop-coil to be very quickly lowered in the event of a fault occurring. The D.C. supply to the clutch was simply removed whereupon the coil descended rapidly under gravity, with a sickening thud.

In line with this manufacturers traditions in regard to motors, the clutch system did not work terribly well. It was satisfactory when the clutch was new: but when the friction surfaces became worn with use, the clutch would not transmit the torque required and began to slip - usually after the transmitter had got nice and hot. The drop-coil would then slide back slightly under gravity, resulting in low supply voltages to the transmitter, and so output power would fall. The very first thing to check for with these units when low power was encountered was, 'are the supply volts OK?' and if not you would then mosey off round the back to witness this slipping clutch. The cure would have been to replace the clutch, of course, but this was an operation fraught with hazard. If you got it wrong, the bicycle chain could slip off and fall into the recesses of the oil tank. This was a very good way of making yourself deeply unpopular, since fixing it usually involved dismantling the regulator, a procedure we shall shortly be making an acquaintance with. Additionally, replacing a clutch only gave a short respite, since these devices lost their youthful tenacity quite quickly.

Fortunately, there was a quick fix available: trot off to the SME's office and get the can of magic juice. This can was a squeeze-side oilcan, filled with Carbon Tetrachloride ('Thawpit', a degreasing agent beloved of dry-cleaners in years gone by). This highly toxic liquid would then be squirted into the innards of the slipping clutch, so cooling it and removing any grease that may be making life difficult for the friction plates. This usually worked like a charm (for a while) and the clutch would grab up and wind the regulator back to its proper position in no time.

Any occurrence that caused an interior fault on the drop-coil regulator necessitated its complete removal from the transmitter. This was always bad news, as you will see.

Disconnecting the regulator itself was a doddle, since all connections were made by heavy-duty (Niphan) plugs and sockets. Having done that, you had finished with the easy stuff. To get it out of the transmitter meant dismantling the front of the cubicle - more plugs and sockets, all front panels to be removed and some framework members to be dismantled. The regulator could then be rolled out of the transmitter. It had small non-steerable cast-steel wheels mounted on axle shafts; the whole affair was over seven feet tall and must have weighed around two tons. If you rolled it onto the transmitter-hall floor, it would promptly sink into the floor tiles and become quite immovable, so it had to be rolled onto quarter-inch thick steel plates to spread the weight.

Having got it out into the hall, it couldn't be dismantled there since there was no lifting gear and insufficient headroom anyway - it had to be transported to the designated place, which was underneath a lifting gantry, outside, round the back of the building. The route lay as follows:

First, the unit had to be swung round by 90 degrees so as to face down the hall. Since there was no steering on the wheels, this meant skidding it round on the plates with crowbars, moving the plates about to suit. Then it could be trundled off down the hall. This involved setting up a plateway, lifting the steel sheets vacated by the trundling and laying them in front of the juggernaut.

At the end of the hall, there was another 90-degree turn to get it into the corridor beyond.

At the end of this corridor, there was a difficult doorway - too low to get the unit through. Simple enough; take the top bit of the doorframe out (it had previously been modified especially for this purpose) and then again on our way, but not for long: another 90-degree turn was needed into the next corridor. At this point the regulator would be blocking the door to the Rest Room, so anyone in there would either have to wait, or climb out of the window. The regulator would pretty well fill the corridor, causing acid comments from other members of staff trying to go about their lawful business. Along this corridor for a bit, pausing only to dismantle another low doorframe, and another turn took us through a storeroom and so into the open air. Then down a concrete ramp, steep enough to require everybody available to be hanging onto the unit to make sure that it didn't run away, because if it did then it would probably turn over when it stopped suddenly at the bottom, as its front wheels sank into the tarmac beyond.

Good news, we're almost there, only two more turns and a good deal of activity with the steel plates would bring us to the lifting gantry.

Repairs to the unit would have to be carried out in the open air and in the event of bad weather could take several days, in practice. Good news though, that large object to the side of the gantry is the spare regulator, sheeted up against the rain. All we have to do is to get this spare unit back into the transmitter.

With good luck and a following wind, if you started changing this unit before 9 a.m. you could have the spare back in, though not necessarily working, by the time the evening shift arrived. You were allowed to stop for lunch.

The room behind the transmitters was where all of the pipe-work lived. This mostly comprised the two combining units, mounted on the rear wall of the room. These were ring-type combiners and mercifully gave next to no trouble.

The same could not be said of the test-loads: each transmitter had its own private test-load into which the transmitter could be switched when it was being, well, tested, and so was therefore likely to radiate sounds that no listener should hear.

Most test loads consist of some form of resistive element to dissipate the power, and are cooled by air, oil or pumped water, depending on the power level involved. These test loads were oddballs, however: there was no resistive element as such, the feeder simply led into a cylindrical chamber through which a suitable electrolyte flowed. It was in effect a short but immensely lossy line.

The electrolyte used was a solution of sodium bicarbonate in water. If the concentration and temperature of this were kept right, the test load would look like a resistive 50 ohms. The liquid was pumped round a closed circuit comprising the load, a motor-driven pump, a heater, a flow meter and a heat exchanger. The exchanger gave up its heat to an external plain-water cooling circuit. The purpose of the (thermostat-driven) heater was to get the electrolyte up to its design temperature before the load was used.

The big problem was keeping the concentration of the electrolyte correct. The tendency was for the concentration to change: nobody ever came up with the definitive reason why this was so, but the best guess was that, some of the connecting pipe-work being of rubber and therefore slightly porous, water would slowly pass through the pipe walls and evaporate but the bicarb, having larger molecules, wouldn't; so causing the concentration to rise. Either way, getting it right involved tedious work with a jug of water to lower the concentration (and then maybe adding back some concentrated bicarb solution if you overdid things).

The test-load itself was provided with thermometers on the liquid inlet and outlet pipes, so the temperature rise of the electrolyte when passing through the load could be found. If this difference was multiplied by the metered flow-rate, and again by a suitable fudge-factor, the transmitter output power was the result. Amazingly, this figure was considered sufficiently accurate to set the transmitter output-power meters up by: in fact, there was no other way of doing it.

On the ceiling of the room were mounted a series of filters. These were of two types: intermodulation filters and harmonic filters.

The intermodulation filters were necessary because the combining units were not perfect. This is no slur: no combiner is ever perfect. What happens is, for example, that some of the output power from the Radio3 transmitter which should be going to the aerial actually appeared at the outputs of the Radio 2 and Radio 4 transmitters. These unwanted signals mixed with the wanted signals, producing unwanted signals at other frequencies. Let's illustrate it with numbers, considering the Radio 4 signal:

Radio 4 was at 92.7 MHz, and Radio 3 at 90.5 MHz. In the Radio 4 transmitter, the unwanted Radio 3 signal mixed to produce sum and difference signals, namely 92.7 + 90.5 = 183.2 MHz, and 92.7 - 90.5 = 2.2 MHz. These are first-order products.

The 2.2 MHz signal was no problem as such: the aerial was hopelessly inefficient at this frequency and radiated next to none of it. The 183.2 MHz signal was no problem either: it was removed by the harmonic filter (see later). But the problem doesn't end there.

This wretched 2.2 MHz signal that hypothetically floated around but caused no trouble, made its presence felt by mixing again with the 92.7 MHz of the wanted output: this produced sum and difference signals of 92.7 + 2.2 = 94.9 MHz, and 92.7 - 2.2 = 90.5 MHz. These are second-order products.

The 90.5 MHz one was no problem: it was squashed by the legitimate and massively greater Radio 3 signal on 90.5 MHz. The 94.9 MHz one was quite otherwise: it was effectively a new radiated signal, sounding like a mangled mix of Radios 3 and 4, and if it got to the aerial through the combining unit, as it may well have done, then the listeners would be treated to a new, and rather confusing, station to listen to. Although this signal was not in any way powerful, it still comfortably exceeded levels permitted for spurious radiations in this band. There would be other signals produced as well, by Radio 2 interacting with Radio 4 and so on. So all these had to be removed.

This was fairly easily done by connecting coaxial notch filters, tuned to each of the other service frequencies, across the output feeder of each transmitter. So, the Radio 4 output had notch filters tuned to Radios 2 and 3, short-circuiting these frequencies and preventing the products from being created in the first place. (Interestingly, the Radio 3 transmitter had no notch filters: both of its unwanted products, 88.3 and 92.7, were squashed by Radio 2 and Radio 4).

The harmonic filters were there for a simple reason: any Class C amplifier such as these will produce second, third and so forth harmonics, which due to the tightly-coupled output stage will be excessive. So they have to be removed. This was done with a coaxial low-pass filter in the output feeder.

There was quite a bit of this harmonic power, so the filters ran hot. To alleviate this, they were blown with air to remove the heat: they each had their own special little blower unit, and since it is not a good idea to blow dirty air into a highly-stressed filter of this type, their own air-filter.