Principles of Interlocking

Understanding locking sheets, dog sheets and facing-point locks


  1. Introduction
  2. How Interlocking Works
  3. History of Interlocking
  4. Facing Points
  5. References


The switches and signals of a railway are distributed geographically over a considerable area in the vicinity of a station. When they were operated manually and at their actual locations, the switchtender or signalman was faced with a lot of walking, so one man could handle only a limited number of them. To economize labor, switch levers were collected at one point, connected by rods with their switches, and located on an elevated platform or cabin for a better view. Signals on top of the cabin were operated by stirrups, so that constant effort was required to maintain them clear. This concentration, excellent as it was, brought with it dangerous hazards.

First of all, a switch or signal could be operated in error, since it was not evident which lever or stirrup went with which switch or signal. Secondly, the signalman was not able to observe the points of switches he operated, to see that they fit properly and were set for the proper route, since they were now remote from him. Thirdly, if more than one man was required to operate the switches and signals, they might work at cross purposes. Fourthly, if a driver did correctly identify the signal applying to him, he could still be in doubt of the proper place to stop if the signal commanded it. Of all the possible ways to operate the levers, most were useless, and many positively dangerous. The purpose of interlocking is to connect the switches and signals so that a dangerous condition cannot arise, and in doing this, to make operation clearer and more logical.

The levers operating switches and signals are placed side by side, from 3-1/2" to 6" apart (the pitch) in an interlocking frame, and are numbered from left to right. Each lever assumes two positions, normal when back in the frame, and reversed when pulled forward. The lever is latched in each position by a spring-loaded dog fitting into a notch in a quadrant. The dog is removed from its notch by depressing the catch handle, after which the lever can be moved. The pivot for a 6-7 ft lever is 3-4 ft below the floor, and the rod or wire operating the switch or signal is attached to the tail of the lever. A backtail on the other side could be provided for counterweights or electric locks. The apparatus for connecting the levers to the operating rods and wires on the ground is called the leadout. The levers are colour-coded depending on their function (some British and American colours are: black--points; blue--FPL's; red--stop signals; yellow--distant signals; white--spares). Signals are generally located on the ends, switches in the middle.

A switch can be operated with 1" ID gas pipe (the usual operating rod) at distances up to about 400 yards. The limit in Britain was set at 350 yards in 1925. A signal can be wire-operated at distances up to perhaps 1500 yards, but 900 yards is a more practical limit. Wire operation is applicable only to two-position signals. Switches and signals can be operated at up to about 800 yards with a two-wire transmission (described in another article on this website). This was never used to any extent in Britain, America or France, but was very popular in Germany. Two-wire transmission is necessary when three-aspect semaphore signals are used, as in Belgium. In America, three-aspect signals were always pipe connected, if they were not electrically worked, which was already available at the time of their introduction, after 1900. Light signals, which appeared after 1920, require only electrical connections. Levers that operated only electrical contactors were specially identified (say, by making them shorter) so that excessive force would not be used on them.

Thermal expansion must be taken into account in the connections. A length of 100 yd of steel expands by 1.7" with a temperature change of 72°F. Rods are divided into equal sections of compression and tension by reversing levers, called compensating levers. The most widely used compensator, the lazy jack, uses two right-angle cranks and a connecting link arranged so that the rod is not offset. Expansion then has no effect, since compression and tension segments expand by the same amounts. Wire is a much greater problem, if only because the runs are longer. The signalman can adjust the tension in the wire manually as temperature changes. Automatic wire compensators have been used, but no design has been completely satisfactory. The general idea is to use a weight to establish a standard tension, and then to grab the wire when it is pulled. It is easier to design a good two-wire compensator than a single-wire one.

We shall be concerned here with mechanical interlocking. The solution by mechanisms of the problems of interlocking is inherently interesting and historically important. The general principles are applicable to any type of interlocking, however. It is significant that all early power interlockings, whether electropneumatic or all-electric, used mechanical locking beds. Electrical methods at first were supplementary to mechanical methods, in devices like electrical locks, track circuits and motor-operated signals. Then, using relays and later digital logic, they eventually replaced the mechanical devices. The same problems, however, have to be solved, and a study of mechanical interlocking is excellent preparation for this.

How Interlocking Works

A simple and effective way for interlocking the levers is tappet interlocking, which is illustrated in the figure. It was invented by James Deakin of Stevens & Sons in 1870. The notched tappet blades or sword irons or plungers or driving irons are moved by the levers, either directly or by some intervening mechanism. Perpendicular to them are the bridles or locking bars, which can move sideways. Riveted to the bridles are the tappets or dogs or nibs or wedges, which can fit into the notches in the locking bars, and slide in troughs or channels. When a blade is moved, a tappet engaged with it will be cammed to the right or left, moving the attached bridle. Alternatively, if a tappet is engaged with the blade and the bridle cannot be moved, the blade will be prevented from moving. Different names may be used for the parts of an interlocking bed by different people. A "tappet" is defined in the Cambridge Dictionary of Science and Technology as a sliding member working in a guide between a cam and a push rod.

The arrangement shown in the figure might be used to interlock the levers for a home signal and its distant signal. While the home signal is normal, at stop, the distant signal should not be able to be cleared. When the home signal is reversed, then the distant signal lever can be operated if desired. One says that lever 2 (the home signal) normal locks lever 1 (the distant signal) normal. Symbolically, 2N/1N. This is a French notation. It is easy to see that if lever 1 is reversed, then lever 2 is locked reversed, or 1R/2R. Such a reciprocal relation always exists in the mutual relations of interlocked levers. Levers, therefore, need only be considered in pairs. Note that while the distant signal lever is reversed, the home signal lever cannot be returned to normal. This is called back-locking. In this case, it guarantees that a clear distant signal cannot be followed by a stop signal.

A different relation between two levers is shown at the left. Here, levers 1 and 2 may be imagined to operate signals for conflicting routes. Therefore, they should never both be reversed, but in the normal state either lever should be free to be reversed and permit the corresponding movement. We see that lever 2 reversed locks lever 1 normal, and vice versa. Symbolically, 2R/1N and 1R/2N, and again the reciprocal relation holds. The rule for finding the reciprocal relation is: interchange the lever numbers and let N ↔ R.

A third relation is illustrated at the right. Here, lever 1 normal locks lever 2 reversed, and vice versa, or 1N/2R, 2N/1R. This relation is its own reciprocal, so the locking might be termed exclusion. If all the levers can be simultaneously normal, the usual case, this relation between levers does not exist.

Tappet locking makes conditional locking possible. In the diagram at the left, when lever 2 is reversed, as shown at the left, levers 1 and 3 are free to be moved in any way. When lever 2 is normal, the movable piece comes between the two tappets. Now, lever 1 locks lever 3 normal, and lever 3 locks lever 1 normal. The relation between levers 1 and 3 is conditioned by the position of lever 2. We say that lever 1 when lever 2 is normal locks lever 3 normal. Conditional locking was very much used in later interlocking practice, especially when a variety of routes existed.

Let's suppose that we have a level crossing, as shown in the figure. To protect it properly, we need a home signal for each direction, and a distant signal for each home signal. We assume that the home and distant signals are two-aspect signals. In general, therefore, eight levers are required. For simplicity, we show only four levers, which could be combined home/distant levers. Derailing switches were sometimes used, adding four more levers, as well as track circuits so the signals could not be changed after a train passed the home signal. No two home signals should be reversed at the same time, and a distant signal should not be reversed unless the corresponding home signal is reversed. If 1, 2, 3 and 4 are the home signals, then lever 1 reversed clearly locks levers 2, 3 and 4 normal. This relation is realized by the lowest bridle. Since this means that 2 reversed will lock 1 normal, we need only require that lever 2 reversed lock levers 3 and 4 normal. The middle bridle realizes this relation. Finally, lever 3 should lock lever 4 normal, as realized by the top bridle. Lever 4 does not have to be considered now, since it already has been considered in relation to all the other levers. It is easy to check from the drawing that if any signal lever is reversed, all the others are locked normal. This example illustrates a general rule. When considering signals, it is only necessary to consider locking with higher-numbered signals. Similarly, when considering switches, it is only necessary to consider locking with higher-numbered switches. Finally, signal levers should lock switch levers. These three rules make sure that each pair of levers is considered only once, and avoids the provision of superfluous interlocking.

As a second example that includes a switch, consider the junction shown at the left. The distant signals and the facing-point lock are not shown, but they would normally be present. The distant signal would have only one arm, and would be cleared only when the straight-through route was set. Lever 1 is pulled for a straight-through movement, and lever 2 for a movement to the branch. Only one signal arm may be lowered at a time. In particular, arms 1 and 2 must not be simultaneously lowered. Arms 1 and 3 may be lowered only when lever 5 is normal, arms 2 and 4 only when lever 5 is reversed.

At the right, a locking sheet for this interlocking is shown. A circle around a number means that the lever is reversed. The sheet may be read: lever 1 reversed locks levers 2, 3, 4 and 5 normal. Lever 2 reversed locks levers 3, 4 normal and lever 5 reversed. Lever 3 reversed locks lever 5 normal, and lever 4 reversed locks lever 5 reversed. We have applied the above three rules in constructing this locking sheet. The locking sheet is realized by the interlocking in the previous diagram, as can easily be checked. Since there is no conditional locking, the "when" column is empty.

As an exercise, the reader may add a facing-point lock (this device is explained below) as lever 6. We have the additional rules that FPL levers lock switch levers normal and reverse, and switch levers lock FPL levers for higher-numbered switches. As usual, these rules help to eliminate superfluous locking. Only the first rule is needed here. Levers 1 or 2 should not be cleared unless the FPL is reversed (engaged). The FPL need not be checked for the trailing movements.

In our examples, after finding the locking sheet we have drawn the tappet interlocking that realizes it. This can be done in a schematic manner, and the result is called the "dog sheet," from which the interlocking can be constructed by a technician. Computers now make it possible to verify a locking sheet or a dog sheet without actually constructing the locking.

From the diagram at the left, we see that a movement of the locking bar through a distance of one-half the width of a tappet is sufficient to move a tappet out of a notch and lock associated levers. Similarly, associated levers are fully unlocked in the last part of the motion of the locking bar. This first and last motion is essential to the safe operation of the frame. Suppose that after one lever is moved a short distance, an attempt to move a second, conflicting, lever is made. If this second lever is not already locked, then both levers may be made to move at least part-way, creating a dangerous conflict. If the locking bars are moved directly by the levers, the considerable leverage can result in damage to the frame. It is necessary, therefore, that all levers that are going to be locked are locked in the first motion of the lever, and that all levers that are going to be unlocked, are unlocked only in the last motion. Some early designers of interlocking frames were not aware of this condition, or did not take it sufficiently into account, to the detriment of their machines.

The simplest way to provide early and late motion is to make the stroke of the locking bar a good deal larger than the width of a tappet, as in the diagrams above. This was done by Stevens and Sons in a very satisfactory small frame that was often used in ground frames, and was even used in the United States, where directly-operated locking bars were otherwise not popular. The sword irons were bent into a circle and attached rather high on the levers. John Stevens, the founder of the firm, was a gas engineer of Woolwich, originally a blacksmith.

Stevens lever locking was inconvenient for any but small frames, since it was heavy and took considerable space. An alternative was to let the lever move a cam plate that then moved the locking bars, an idea originated by George Edwards, of the Railway Signal Company, Liverpool. An S-shaped cam slot would provide early and late motion, as shown in the figure. This method was used on the Great Western Railway's HT and VT tappet interlocking frames (1904). If a lever was locked, a very small trial motion would discover it. Nevertheless, the locking still had to be made rather large and heavy. It was vertical, and beneath the operating floor. Electric locks were on the lower ends of the tappet blades, below the locking.

It is not necessary to mechanically interfere with the motion of the lever itself; it is fully sufficient to prevent the catch dog from being withdrawn from the quadrant. This idea dates from 1867, and is due to Easterbrook. Also, the motion of the catch rod when the catch handle is depressed and released can be used to drive the interlocking. Since this motion occurs at the beginning and the end of the lever motion, early and late motion is perfectly assured. Indeed, the locking is actuated before the lever makes any motion at all. Saxby ingeniously arranged for this by means of the link, or rocker, that is moved by the catch rod. When the lever is normal, depressing the catch handle moves the rear end of the link up, providing the first half of the motion. Then, the centre of curvature of the link slot is the same as that of the lever. As the lever is moved from normal to reverse, the link does not move at all. When the catch handle is released and the catch rod moves downward again, the front of the link is further depressed and the final half of the motion takes place.

The Patent Office identified two ideas here: first, that of locking a lever by locking its catch; second, the driving of the interlocking by the motion of the catch rod. Easterbrook, who was employed by Saxby and Farmer and was working on this problem, applied for a patent on his own. Saxby countered with a patent, then filed a second patent, which Easterbrook also countered. The outcome was that Easterbrook owned the locking of the catch, while Saxby owned the operation of the locking by the catch rod. Therefore, neither could offer interlocking combining the two principles until the patents expired, and so the public suffered again from the illogic of the patent process and the pleasures of litigation.

The first catch-locking machines of 1874 did not use tappet interlocking, but one called the "gridiron" interlocking. By means of cranks, the links moved by the catch handle rods rotated slotted pieces called gridirons or flops, giving the nickname of the machine. When the levers were all normal, these gridirons were horzontal. A gridiron with five slots was used with five locking bars above, five below. Only an upper locking bar is shown in the figure. Each locking bar was driven by a pin and a driver from one of the gridirons. Locking dogs on the locking bars either left a gridiron free to rotate, or prevented it from rotating. In the diagram, free and locked gridirons are shown. If a gridiron was prevented from rotating, its catch could not be released. This locking met a friendly reception in France, but a less friendly one in Britain and America, where it was subsequently replaced by tappet locking around 1888 (The S&F "Duplex" frame). By any measure, however, it was an excellent interlocking and achieved dominance. Since the catch locking was not subject to the forces exerted by the levers, it could be made much lighter and smaller than lever locking. In fact, the locking was usually just behind the row of levers on the operating floor, while heavier interlocking had to go beneath the floor in a less pleasant environment.

The principle of the gridiron locking was similar to that of John Imray's locking frame of about the same date, that used rotating cylinders that were blocked by dogs on locking bars. The principle of locking before the lever is moved is called preliminary locking, especially in the United States. All locking must be done before the lever is moved, and all unlocking after it is in its final position. In lever locking, this is approximated, while in catch locking, it is exactly achieved.

History of Interlocking

In the earliest days of railways, signals were operated by stirrups pressed down by the signalman's foot. This ensured that they were displayed only when intended, and then returned to stop by gravity when his foot was removed. Charles Hutton Gregory arranged the stirrups at Bricklayer's Arms Jct. in 1843 so that pressing down one interfered with the others, so that conflicting signals could not be given. The points were still operated on the ground independently of the signals. Saxby, in 1856, concentrated levers operating points and signals in one row, as we have mentioned. This installation was the first to arrange point and signal levers in a row, and to use the spring catch. The first frame of this type was installed at Keyham Junction in Brighton. This is generally considered the first interlocking that connected signals and points, but it operated on a different principle than modern interlocking. When the point lever was pulled for a certain route, the corresponding signals were operated at the same time. Now, of course, the points are operated first, and when this is completed the signals are given.

The 3-3/4 mile London and Greenwich Rly was opened on 14 Dec 1836. The London and Croydon Rly, opened 1 June 1839 over the 8-3/4 miles from West Croydon, ran over the L&G for 1-1/4 miles from London Bridge Station to Corbett's Lane Junction. London and Brighton, and South Eastern Rly, trains were added to the traffic that year. At Bricklayer's Arms Junction, the Croydon trains were diverted onto their own tracks, and arrived at a different platform at New Cross from the Brighton and South Eastern trains. The first congested traffic on railways was on these lines, and they appear often in the history of signalling and interlocking.

The next Saxby and Farmer machine is shown in the diagram at the right, and is of the modern type. It appeared in 1860, a great improvement on Saxby's 1856 machine for Bricklayer's Arms that moved points and signals simultaneously. The sliding locking bar at the top is driven by the motion of the lever at the left. In the top jog, the necessary levers are locked; in the middle portion the locking bar is not moved, while in the lower jog the necessary levers are unlocked. In this way, early and late motion was provided, but not nearly as perfectly as in the rocker frame. This frame was used at the new Victoria Station, Pimlico, London. The first interlocking machine imported into the United States in 1870 for use at Top of the Hill, Trenton, was of this type (Serial No. 905). the 1874 machine for East Newark, however, was a rocker type with gridiron locking.

On the North London Railway, at Willesden Junction in 1859, the Inspecting Officer, Col. Yolland, refused to approve the arrangments until interlocking was provided between points and signals. Austin Chambers, in cooperation with Stevens and Sons, managed this by arranging the signal stirrups with rods that passed through holes in plates moved by the point levers. A signal could not be cleared unless the rod passed through all the plates. This was the first interlocking in the modern sense, where points and signals are operated consecutively and independently. Chambers later devised machines where both signals and points were operated by levers.

Pierre-Auguste Vignier (1811-1891), an engineer for the Chemins de Fer de l'Ouest, devised an interlocking between points and signals that was first installed in 1855 at four junctions of that company. The earliest form used rods passing through holes. The hole had to be in the correct position for the lever moving the rod, or the lever could not be reversed. This was similar to Chambers's idea of a few years later, but was executed quite differently, the rods and holes arranged just beneath the floor of the cabin. Later developments along the same lines led to an interlocking that was widely used in France for layouts that were not too complicated.

Stevens' hook locking, shown in the diagram, appeared around 1860, at the same time as the Saxby hook locking. It was, however, quite different and not as adaptable. The locking rod was pressed against a spring when a lever was normal, as shown. If the lever was reversed, the spring pushed the locking rod to the left, locking some levers and releasing others. This was an effective interlocking, and operated when the lever was moved only a short way. Everything was complete, however, at this point, and there was no final motion. When the lever was replaced, it cammed the locking rod back to the right on the inclined surface. In this form, it could not lock a lever reversed, only prevent a lever from being reversed. However, when a point lever is reversed and a signal cleared, the signal should have to be restored before the point lever can be replaced. This back locking was provided by catches lower down on the levers.

R. C. Rapier's locking of about 1874 is shown at the right. The levers are combined with a large sector that rotates about the axis. Locking bars slide transversely, and every bar is beside some part of the sector regardless of the position of the levers. The locking takes place between square notches in the quadrant and the locking bars, as shown. The lever is shown reversed; its normal position is vertical. The locking bar corresponding to a lever is moved by means of a cam (the square object) engaging a surface on the quadrant that provides the required lateral motion. The first movement of the lever locks all levers interlocked with it, and the last motion unlocks those that are to be unlocked. This is, indeed, a very simple and direct interlocking, as Rapier intended. A frame of this type with 80 levers was installed at Lincoln, and several smaller frames were used elsewhere, though the Rapier frame was never popular. Similar locking, directly interfering with the motion of a lever in this way, was called stud locking, though the large quadrants were not used, and the studs were bolted to the locking bar.

By the 1880's, Stevens tappet locking, invented in 1870, had become very popular. Although it was a lever locking, and subject to rapid wear, these machines were well made, and earned a very good reputation. The Stevens locking was very simple and easily maintained. It even gained a small foothold in the United States, where catch locking was nearly universal. It was handled by Union Switch and Signal. In 1882, the North Eastern Railway in England had 6109 levers in 386 interlockings, for an average of 15.8 levers per interlocking. Of these, 4428 were Stevens levers. This was about three times as many levers as were in the whole United States in 1885. In 1873, the London and North Western already had about 13,000 levers.

The interest in British interlocking shown in the United States in the 1870's encouraged J. M. Toucey and W. Buchanan of the New York Central and Hudson River Railroad to design a distinctly American interlocking machine, the first example of which was installed at Spuyten Duyvil at the north end of Manhattan island in 1875, a year after the Pennsylvania imported a Saxby and Farmer machine for East Newark (Serial No. 2164). This was the only interlocking machine available on short order when the Pennsylvania wanted additional interlockings for the demonstrations in connection with the Centennial Exhibition in Phildelphia. Actually, just the patterns were furnished and the Pennsylvania did the rest. Read about this in the article on the T&B machine on this website.

The Toucey and Buchanan machine worked, but was very cumbersome and could be defeated by a concerted attmept to pull two levers at once. Each lever was locked by a rotating plate in either normal or reverse. When the signalman stepped on a pedal, the plate rotated if the lever was unlocked, and he could then move the lever. The T&B machine was acquired by Union Switch and Signal, but few sales were made. Beside the NYC&HR, only the Baltimore and Potomac (Fulton Jct.) and the Clevelend, Columbus and Indianapolis (Berea, O.) had T&B frames.

Arthur Henry Johnson (1838-1910) emigrated to the United States from England in 1886 to be a manager at Union Switch and Signal. He had worked with Stevens and Sons and Saxby and Farmer, and had been Superintendent of Signals for the Lancashire and Yorkshire Railway. In 1888, with his nephew C. R. Johnson, he founded the Johnson Signal Company of Rahway, NJ, which was sold to the National Signal Co. of Easton, Pa. in 1895. Then, with J. T. Cade he founded the Standard Railroad Signal Co., from which he retired. Standard, the Pneumatic Signal Company, and Taylor Signal Company were consolidated into the General Railway Signal Company. He designed a locking frame similar to the Saxby and Farmer rocker machine, but the link was below the floor and connected in an inverted manner. The catch rod moved one end of the link, where the S&F machine took its output, the locking rods were connected where the catch rod was in the S&F machine, and ran vertical, to a vertical tappet locking bed below the floor. The link pivot was on the lever. When the catch rod was raised, it moved the end of the link upwards until it was concentric with the lever. Then, the locking did not move while the lever was pulled over, taking the link with it. When the catch was released, the link moved the final half of its travel. It was simply the Saxby and Farmer link after a kinematic transformation. It seems to have evaded the patent well enough, and the Johnson machine saw considerable use. The vertical locking bed was later used in several power interlocking systems. One sales point was that it used up less depth in a signal cabin than a large Saxby and Farmer machine did. Johnson introduced the Webb and Thompson electric train staff machine to the United States in 1892. It was first used on the Milwaukee between Savanna, Illinois and Sabula, Iowa in 1894. The staffs were 22" long and weighed 7 pounds.

Later American interlocking machines are all descendants of either the Saxby and Farmer or the Johnson machines. The machine manufactured by the National Signal Company (founded 1884 in Brooklyn) was similar to the Saxby, but the link was beneath the lever quadrant and the tappet locking was vertical. Its patentee was J. A. Bonnell. The Pennsylvania Steel Co. (1880, Steelton, Pa.) supplied a frame designed by A. G. Cummings (U.S. Patent 226,499, 1880). The vertical locking was behind the frame, and the floor plate was flat. The Standard machine was very similar, as was the Union Switch and Signal Company's Type A vertical tappet locking of later years, which was designed by A. H. Johnson.

By 1885, Union Switch and Signal had installed about 95% of the interlocking levers in the United States. They offered the Westinghouse electropneumatic, Saxby and Farmer, Stevens, and Toucey and Buchanan machines. The National Interlocking Switch and Signal Company, the Pennsylvania Steel Company, the Wharton Switch and Signal Company, and the Hall Signal Company offered interlocking or automatic signals, or both, up to that time. There was no interlocking in Canada at all. Few grade crossings were interlocked; mainly it was junctions and large stations. It was being argued in the Railroad Gazette that interlocking at crossings actually saved money, and was not just an added expense.

John Saxby (1821-1913) and John Stinson Farmer (1827-1892), both from the London & Brighton Rly, formed a partnership in 1863. Their Kilburn works were at Canterbury Road, London. In 1920, Saxby and Farmer was acquired by Westinghouse, and became Westinghouse Brake and Saxby Signal, whose principal works were at Chippenham. Stevens and Sons had works in London and Glasgow. McKenzie, Clunes and Holland were located in Worcester. Later, Clunes was dropped and the company was then long known as McKenzie and Holland. The Railway Signal Company, previously Livesey and Edwards, was formed at Fazakerley, Liverpool, in 1881. In spite of the excellent products of these firms, some major railway companies manufactured their own unique apparatus. The Midland licensed the catch-lock patents, and made their own apparatus from 1865. The London & North Western manufactured their signal apparatus at Crewe from 1874. F. W. Webb was as interested in signalling as in compound locomotives, and developed not only an ingenious catch-locking frame but also an electric train staff.

On the Great Western Railway, Michael Lane was the engineer responsible for the "old" Great Western after the mergers of 1863 that added narrow-gauge lines in the North and Wales. In those days, signal work for the company was partly by Saxby and Farmer, and partly by the company itself. A signal works was created at Reading in 1863. Like many railway engineers of the time, Lane also took an interest in interlocking and designed a frame of his own which appeared in 1865. In the Lane frame of 1865, a rack attached to a lever moved a pinion, rotating a shaft running from left to right. The end of this shaft was threaded, and worked in a nut attached to a locking plate, which it moved transversely. Levers were locked by notches in cross members of this locking plate. The locking was quite secure, but the throw was proportional to the motion of the lever, so there was no first or last motion. If two free but conflicting levers were pulled, the result would at best be a jammed frame.

The first locking on the GWR was installed at the junction with the Metropolitan Railway at Paddington, in 1860. In 1863, a frame designed by T. Blackall and made in Reading was used on the departure side at Paddington. The Signal and Telegraph works at Caversham, Reading were founded in 1859. Signal work was done there, or contracted out with Saxby and Farmer. After 1885, the GWR made all its signal apparatus at Reading, when Thomas Blackall became signal engineer there. The GWR developed the "twist" frame, in which moving a lever caused a twisted bar to rotate, and this rotation drove the locking. In the earlier "single twist" (1872-1890), invented by Tom Gooderson in 1872, there was quicker motion at the start of the pull, supplying first motion that overcame the problem with the Lane machine. Later "double twist" (1890-1906) frames provided both first and last motion; one of these, at least, survived until the 1970's. Various forms of locking were used, from Saxby-type hooks to Stevens-type studs. The "stud" frames, for 37 levers or less, were made 1892-1908. After 1908, 4" pitch replaced 5-1/2" pitch. All of these frames, of course, were lever locking, which was rugged but subject to wear.

The final GWR mechanical interlocking was a tappet interlocking moved by cam plates, as originated by Edwards at the Railway Signal Company. This frame (the HT3) was made 1906-1926. It had horizontal or vertical locking trays driven by bell cranks. The VT5, the final version, was made 1926-1966. The number 3 or 5 refers to the number of bridles per tappet blade. If a tappet blade could not be moved, then the cam plate would not move, and consequently the lever could not be moved. This heavy locking proved very satisfactory. The largest VT5 frame, at Reading Main Line West, had 222 levers. It was installed in 1943.

We have by no means been able to describe every interlocking invented, but have mentioned the most important for later developments, and those that demonstrate basic principles. Improvements after about 1900 have also not been thoroughly treated, and we ignore the interesting development of power interlocking in the United States. The lever locking of McKenzie and Holland should be described, as well as the Midland "tumbler" catch locking. After the expiry of the basic patents, any company could offer the best technology, which was usually that of Saxby and Farmer. In many cases, adequate information is not currently available to me. However, the main aim of this paper is to describe the principles of interlocking, and the examples that are present do this fairly well.

Facing Points

At facing points, a train may take either of two routes. One route is usually straight ahead, and has no speed restriction. The other is diverging, with usually a severe speed restriction, perhaps as low as 15 to 20 mph. The hazard is that a rapidly moving train may be unexpectedly diverted, with the result of derailment, damage to the track, and unexpected appearance on an occupied track. If the points did not fit properly, derailment was almost certain. Moving the points under a train also had unpleasant consequences. These contingencies rendered facing points a hazard, and one that was soon realized.

It should be appreciated that facing points are not a hazard if they are operated manually on the spot, since then the switchman can check to see that the points fit properly, and he will not operate the switch under a train obviously passing over them. Also, if a train approaches them slowly, the driver can himself check the points and stop if all is not right. In France, trains were required to approach facing points at low speed. In Britain, however, it became usual for trains to encounter facing points on a clear track at full speed. The danger was countered by eliminating as many facing points as possible, and providing those that remained by prominent direction indicators. A typical station layout is shown in the diagram. Note that trains moving with the current of traffic (to the left here) do not encounter a facing point. Trains typically backed through crossovers and into side tracks, which was not excessively inconvenient with the short trains of those times. However, it was always time-consuming. The single-slip switch might earlier have been a normal crossover followed by a second trailing crossover to the service track. A slow train got out of the way of a fast train by backing through the crossover onto the other track. It is much safer and quicker to use a passing loop (siding), but this requires a facing point. At junctions, facing points cannot be avoided.

The use of interlockings, where points are operated remotely, required additional safeguards, since the points were no longer under the direct observation of the signalman. The simple and ingenious solution is shown at the right. A bolt fits into apertures in the stretcher bar when the points are fully one way or the other, verifying that the points are properly closed. A detector bar is linked to the mechanism so that it is moved from one extreme position to the other when the bolt is removed or inserted. Since it is arranged like a parallel ruler, in doing so it rises to the rail head, perhaps to the level of the dotted line in the figure. If wheels are passing over the points, it cannot be raised, and the bolt cannot be withdrawn, so the points cannot be reversed. This facing point lock, or FPL, eliminates the hazards of remote operation. The inside connected FPL was always used in Britain, while in the United States the outside connected mechanism, with the bolt outside the stock rail, was usual.

This form of FPL does not, however, distinguish between the two positions of the points. If the connection with the points is broken, then the bolt is simply withdrawn and replaced in the same aperture, when in fact the points have not moved. Although mechanisms can be devised that do distinguish the positions of the points, it is more usual to pass the signal wire through an interlock moved directly by the points. If the points are not in the expected position, then the signal cannot be cleared. This wire lock was invented by C. F. Whitworth in 1853.

A track circuit can take the place of the detector bar, and lock the lever working the FPL electrically. Also, the points can operate contacts detecting the position of the points. The bolt can be placed inside or outside the track, and can be replaced by wedges acting directly on the points.

In the United States, efforts were made to reduce the number of levers required by combining operations. One arrangment that was widely used was the switch and lock mechanism. When the lever was pulled, the early motion withdrew the bolt, then the points were moved, and finally the bolt was reinserted. This worked well enough, but made the pull more difficult. Another arrangement was to work both sets of points of a crossover with one lever. When combined with the lock and detector bar, this was indeed a heavy pull, and could not be used at any great distance. A less useful combination was to work a home and distant signal with one lever. This was chiefly applicable in America, where distants repeating only one home signal were common. In Britain, there was generally more than one stop signal covered by a distant signal, so this combination was not used except for elecrically operated intermediate block signals. The short levers were painted red/yellow, however. If the signals were mechanically connected, the home could be cleared while the distant was maintained at danger by holding the lever at mid-stroke. Another combination was to use a lever to clear the signal for one of several related routes. Only the signal for the route that was set would clear. This was easy when route indicators were used with one stop signal. These combinations were all available by 1890. At the Jersey City terminus of the Central of New Jersey, a machine that traditionally would have required 185 levers was actually realized with only 87 levers by the US&S.


J. Dixon, P. Kay and R. Newman, A Guide to Mechanical Locking Frames (Signal Study Group, 1989).

Signalling Record Society, GWR Lever Frames: The 5-Bar Tappet Frame (SRS, Signalling Paper No. 3, 1982), and Supplement 1, 1988.

R. C. Rapier, On the Fixed Signals of Railways (Proc. Inst. Civ. Eng. 38, 142-247 and Plates 27-31, 1874).

A. H. Johnson, The Development of Fixed Signals on Railways (Railroad Gazette, 7 April 1893, pp. 255-258).

A. D. Dawnay, Treatise on Railway Signals (1874), pp. 3-86.

____________, Railway Signalling, The Railway Engineer, Vol. 2, pp. 37-387 (1881) and Vol. 3, pp. 12-319 (1882).

The Saxby and Farmer Safety Switches and Signals, Scientific American 32, 3 April 1875, p. 207.

The Grand Central Depot Signal System, Scientific American 33, 25 December 1875, p. 401.

B. B. Adams, The Block System of Signalling on American Railroads (New York: The Railroad Gazette, 1901). Chapters XIII-XVII. pp. 173-226.

R. H. Soule, Railroad Signals and Signaling (Engineering News, 21 Dec. 1889, p. 583 and 28 Dec. 1889, p. 617).

Return to Railways Index

Composed by J. B. Calvert
Created 2 August 2004
Last revised 22 September 2008