The Great Western Railway Automatic Train Control

The GWR ATC was the first widely-used automatic train control system

The GWR was a pioneer in cab signalling, devising and applying an effective system by 1906 that survived, with slight modifications, into the 1980's, when it was finally superseded on ex-GWR lines by the British Railways intermittent inductive AWS that is still in use. Inventors had toyed, usually impractically, with cab signalling since the 1860's, and underground railways have used the simple but effective mechanical trip system since the 1890's. Several inventors, mostly in the United States, had proposed systems before 1910 based on a variety of principles, including radio, but these were usually impractical, and only a few received ephemeral trials. The objects of cab signalling are two: first, to display aspects in the cab to the driver, and second, to bring the train to a stand if a restrictive signal is misunderstood or overlooked. All this is of special value in fogs and falling snow, which was the principal reason it was explored by the Great Western. With effective cab signalling, the expensive and cumbersome system of providing fogsignalmen could be eliminated.

The system was called ATC, Automatic Train Control. It was developed in the company's Reading Signal Works as a combined effort of signal engineers and craftsmen starting in 1905, and was tested on the double-track Henley branch, where 6 ramps were installed. Later the same year, it was tried on the 22 mile, single-track Fairford branch, where the distant signals were removed. The later decision was taken to retain them in future installations, possibly because drivers were uneasy at their absence, but also to permit non-equipped engines to work trains on equipped lines. The GWR was a true pioneer; nothing like its system existed elsewhere, and it was adopted as standard, not as a mere experiment.

As I'll relate below, a similar system came into use in the United States in 1911 and was in service in its original form until 1950, while other systems were installed there on various lines after a regulatory order in 1920, one of which was a cab signal system using coded track circuits that covered all main lines of the Pennsylvania Railroad well before the second world war. It is, therefore, not true that the GWR ATC was the only system in daily use on any railway from 1906 until 1947 (not even in Britain) as stated by Adrian Vaughan in Obstruction Danger (p. 106), but his point is well taken that the GWR was first and foremost, and proceeded completely under its own steam.

The experiments were so successful that main and relief lines between Paddington and Reading were equipped by December, 1908. 168 ramps were used on the 144 track miles in this stretch. Four ramps were installed on the single-track Lambourn Valley branch in 1909, but further installations were postponed, because of wartime stresses. In 1936, ATC was being extended from Plymouth to Penzance, Swansea to Fishguard, and Wolverhampton to Wellington, already covering main lines closer to Paddington. By 1938, all the GWR main lines had been equipped with ATC. Except for the Great Central and North Eastern, no other mainline company regularly used cab signals in Britain, and these two companies only in very limited areas. The Great Central used the mechanical trip Reliostop system for a few years, until it was removed by the LNER, as was the Raven system on the North Eastern between York and Darlington. There was no enthusiasm for cab signals or automatic train stop in Britain. The LMS made a trial installation of the Hudd intermittent inductive train stop system on the London, Tilbury and Southend section just after the War, work that had begun before the war. This system was taken up by British Railways and became the later standard Automatic Warning System, AWS.

Railway companies in the United States were equally enthusiastic about cab signals, but were forced much against their will to adopt such systems by regulatory action. This regulatory activity had been festering since about 1907, when the Interstate Commerce Commission began an unsuccessful effort to compel general adoption of the block system. Being frustrated in this effort, on acquiring expanded powers in 1920 the Commission ordered installation of some system of automatic train control on randomly (and badly) selected passenger lines. This compulsion was so unpopular that very few voluntary installations were ever made. Most installations that were made were of an intermittent inductive train stop system (abbreviated ATS, automatic train stop) that applied the brakes at a restrictive aspect unless forestalled by the driver. This was the cheapest system that would satisfy the regulators, but an effective one, much like the British Railways AWS that began in 1947, except that permanent magnets were not involved.

One voluntary installation had, however, already been made in 1911, before the order, on 114 miles of double-track main line of the Chicago and Eastern Illinois Railroad, from Chicago to Danville, Illinois. The system used was the patent Miller system, which employed a ramp and shoe almost identical to the GWR ATC, except that the ramp was outside the rails on the right-hand side, and the shoe was mounted on a tender bogie journal. On the locomotive, the automatic brake valve was mechanically rotated to service application position by a chain and pneumatic piston unless the driver acknowledged a restrictive signal. This system was reliable and durable, and lasted until 1950, when it needed renewal, but the Miller company had long disappeared by that time, absorbed by GRS. In 1927, the Chicago, Indianapolis & Lousiville (Monon) installed Miller ATC between Hammond and Indanapolis, 161 miles. I suspect that Miller got the idea from the GWR ATC, and modified it slightly for American conditions. At any rate, it is further confirmation of the essential practicality of the system used in ATC.

[ATC Ramp] The following description is of the original GWR apparatus. Lineside equipment consisted of a steel T-section ramp with the web upward, supported by a timber baulk, from 40 to 60 feet long (normally 40 ft), raised 4 inches (later 3-1/2") above the rail head but tapered downward at the ends, and fastened to the sleepers slightly askew in the centre of the four-foot. The skew was to prevent wearing a groove in the contact shoe on the locomotive. The photograph shows the eventual appearance of the ramp. The ramp was energised through a switch operated by the distant signal lever (and later cut through a switch on the distant signal arm). When this lever was reversed (distant signal off) the switch was closed and the ramp was electrically connected to a 30V primary battery (later 16V) in the signal box. The other terminal of the battery was earthed. On the Fairford branch, the polarity of the voltage applied to the ramp could also be changed, since operation was bidirectional. The ramp was located 440 yards in rear of a distant signal, or right at the signal where a home and distant arm were on the same doll. In the latter case, the ramp would not be energised if the home signal were at stop, even if the distant signal lever were reversed, in a kind of electrical slotting. British Railways reduced the distance to 200 yards later.

The heart of the locomotive equipment was an electromagnet with two windings that held up an armature when either winding was energized, keeping a valve to the vacuum brake pipe closed. The diagram below shows the simple but effective electrical design. The circuit for one winding led through a 4V lead-acid accumulator (80 A-h capacity) on the locomotive and a switch operated by the contact shoe below the footplate (usually under the cab, but sometimes under the front buffer beam, as on the Saint class 4-6-0s), which was closed when the shoe was down in its normal position. The shoe was raised 1-1/2" (later 1")when on the ramp, and made electrical contact. When the shoe was thus raised, the switch was opened, current in the windings stopped, and the armature fell out, opening the train (vacuum) pipe. Air entering the vacuum pipe sounded a siren, consisting of a disc rotated by the inrushing air. The driver could hold the armature up by means of a 'release trigger' on the cab mechanism box and keep the vacuum pipe closed so that he could retain control of his train. This is called resetting, cancelling, acknowledging or forestalling, and established that the driver was alert. It is very important to take some voluntary muscular action at times such as these, to alert the mind and destroy a reverie. Once the armature fell out, the strength of the electromagnet was insufficent to draw it up again when the shoe left the ramp and the magnet was re-energised. Consequently, the brake application continued. The use of the brake vacuum to give the warning siren suggests that the system involved the brakes from the first, not subsequently, as stated by some authors.

[Circuit Diagram] If the signal was off, however, the ramp was connected to the battery in the signalbox, and a second circuit was completed through the second set of windings on the electromagnet, the locomotive frame, and the rails, holding the armature up. The current in this circuit also operated a polarised relay, whose contacts energised a slow-action relay connecting a vibrating bell across the locomotive battery. Hence, while the locomotive was passing over an energized ramp, the bell rang to give positive indication that the signal was off. The bell relay was probably a slow drop-out relay, since the shoe was on the ramp for only 0.4 seconds at 70 mph. On single track, the polarity of the ramp could be reversed so that the bell would not sound when the locomotive passed over a ramp intended for the opposite direction. Since the ramp would be energised in this case, the armature would not drop. This feature was used only on the Fairford branch, but was employed in experiments for responding to a double-yellow as well as a yellow in 1946-1947, but this was not developed further by BR. Whenever a ramp was under repair, a handsignalman was assigned to the location. Occasionally, the ball on a hanging screw coupling on "other companies' wagons" struck a ramp, and employees were enjoined to make sure these couplings were properly hung up.

The locomotive battery was continuously on circuit, so its life was a matter of concern. To save the battery as much as possible, its earth terminal was connected through a switch operated by the brake vacuum. When there was no vacuum, the switch would open in a half-hour, disconnecting the battery. When vacuum was again created, the switch would close, reconnecting the battery and again holding up the armature. It would appear that the armature would have to be held up until vacuum was established. The other point of stress was the switch on the contact shoe, which was operated at every ramp. There were telegraph codes for each of these items, ENBAT and SOITCH respectively, to facilitate messages concerning them. CABUS referred to the cab apparatus.

Telegraph codes were also available for reporting faults in the operation of the ATC. NOWBON and NOWBOFF were used for neither whistle nor bell when a distant signal was on or off, respectively. WABON and WABOFF were used when both the whistle and bell were heard. WISDAN signified the fail-safe mode of whistle when the signal was off, DISBEL hearing the bell when the distant was not properly off. Significantly, there was no code for a bell when the distant was on, a dangerous fault mode that probably never occurred. Drivers reported ATC faults on Form 4707.

A later modification in BR days added the sunburst or chrysanthemum visual display orginating with the Hudd inductive Automatic Warning System when a dead ramp was encountered, to reinforce the audible warning of the siren. This display is illustrated below. The display is, in effect, connected with the electromagnet armature, and is restored to black when the warning is acknowledged. Therefore, when a restrictive signal aspect is passed, the driver receives both visual and audible warning and must take prompt definite action, or else the brakes are applied. The use of audible signals for both proceed and warning is an excellent idea of the GWR ATC. Audible signals demand attention, unlike visual signals which must be noticed.

The BR ATC Visual Warning Display

Move the mouse cursor over the disc to see the display that appeared when the ramp was not energised (distant signal 'on'). You must have Internet Explorer 4.0 or similar browser to see the animation. The display was accompanied by a loud whoop generated by a siren driven by the air entering the brake pipe. If the ramp was energised (distant signal 'off'), a vibrating bell rang instead. These distinctive signals have now been replaced by feeble electronically-generated sounds.

The brake application was sufficient to stop a train before passing a stop signal ahead even if steam continued to be worked, if the footplate men were disabled. The accident at Milton, near Didcot, in 1955 and described by Adrian Vaughan in Obstruction Danger was one in which a train ran through signals for no apparent reason, and derailed at facing points taken at too great a speed. Every effort was made to discredit the signals and the ATC. The dirty footplate of the Britannia had clogged the siren, and the driver was on the wrong (left) side for the signal layout, but on test the ATC worked correctly. Since the driver and fireman seemed to have been alert, the ATC should have prevented the accident unless wilfully disabled. In this action, the clipping of the points leading to the diversion, normally a safety measure, prevented the signalman from changing the points when he saw what was about to happen.

There was originally a boiler steam-pressure operated switch that opened the circuit of the locomotive battery when the steam pressure dropped below 40 psi. The idea was to save the accumulator charge on dead engines. The armature would drop out, and the vacuum pipe would be opened, but there would be no vacuum anyway. There must have been a cut-out cock in the vacuum pipe that would prevent a brake application when the system was out of service or defective. A failure of either battery, or a wire breakage, would cause the system to sound the siren and apply the brakes immediately, thus failing safe.

Two important principles in the use of ATC should be appreciated. First, it essentially is a cab distant signal, warning of danger ahead and alerting the driver. Second, control is taken away from the driver only if he fails to respond. It is not, like the mechanical trip on the Underground, an enforcer of stop signals that mindlessly takes control from the driver. Although the latter is appropriate in its special environment, taking control from the driver has generally been considered undesirable with the higher speeds and heavier trains of intercity railways, and the greater variety of conditions. For one thing, an emergency stop in some cases may result in derailment or other mishap more dangerous than continuing movement. In fact, the penalty brake application of ATC is sufficient to bring the train to a comfortable stand before reaching the stop signal, and should not be an emergency application, as might be required if a stop were being enforced.

When automatic train stop, to prevent passing signals at stop, was ordered in the United States, as related above, the Pennsylvania Railroad and the Union Switch and Signal Company (Westinghouse) co-operated in the early 1920's to develop a cab signal and train stop system using coded track circuits. This system displayed three (later four) aspects in the cab, continuously changing with conditions. It was so effective that the Pennsylvania was permitted to remove the connection with the brake system on condition that cab signals were installed on all its main lines. Other companies that adopted coded cab signals retained the connection with the brake system that enforced speed restrictions of 30 or 15 mph. In some cases, these companies also removed lineside signals, enjoying considerable savings, and operated trains by cab signals alone. The system is called ATC or CCS (continuous cab signals). The driver might forestall a penalty brake application, but still had to make a sufficient brake application and keep the speed within the enforced maximum, to prevent a penalty application, which would bring the train to a halt and require going on the ground to reset. In the United States, a train on a line with ATC must have the system in operation to operate at normal speed. If the ATC apparatus fails on the road, written authorisation (obtained by radio) is necessary to proceed at less than 80 mph to the nearest point at which an equipped locomotive can be substituted or the apparatus repaired. A train cannot start with inoperative ATC. The strictness of the rules is partly a result of the warfare between the railway companies and the regulators, intended to prevent companies from operating trains without properly maintained and working ATC. Incidentally, on lines without ATC, maximum passenger train speeds in the United States must be less than 80 mph by regulatory order.


  1. M. G. Tweedie and T. S. Lascelles, Modern Railway Signalling (London: Blackie & Son, 1925), p. 199ff.
  2. General Appendix to the Book of Rules, GWR 1920 and 1936.
  3. Adrian Vaughan, Obstruction Danger (Wellingborough: Patrick Stevens, 1989).

Return to Railways Index

Composed by J. B. Calvert
Created 12 July 2004
Last revised