Signal failures: how they occur, and what their effects have been, with 17 examples
A false clear is a signal failure in which the signal displays any other aspect than its most restrictive, when it is supposed to display its most restrictive. At its worse, Clear is displayed when Stop is required at that immediate location to prevent accident. It can also be Clear displayed by a distant signal that should display Approach, so that a Stop aspect may be encountered with little warning, and passed due to inability to stop in good season. Other types are possible, but usually less dangerous.
Another sort of failure is the false stop, which is a "fail-safe" state, occurring much more often than a false clear. In fact, signal systems are designed so that any conceivable fault results in a false stop rather than a false clear. However, false stops must be minimized to ensure safety, because of the danger of a Stop being considered a fault rather than a true warning. In the early days, signal systems that often displayed false stops were usually ignored. Whenever a signal displays an aspect better than Stop, it must be maintained there actively. If anything breaks or fails, the signal must go to Stop automatically.
Light signals are not subject to mechanical troubles, as were semaphore signals. Semaphore signals that normally stood at Clear (normal-clear signals were promoted by Union Switch and Signal) might freeze there in an ice storm. This is one reason why it is preferable to have signals normally at Stop, which display the Clear aspect only when a train is in the block approaching them. The later motor semaphores were powerful enough to overcome most ice problems, and no cases of accidents due to frozen semaphore signals are reported from 1911-1966. On the good side, semaphore signals are easiest to read when light signals are at their worst, at sunrise and sunset with the sun directly in front of or behind the signal. There are several cases of light signals being misread under these conditions. Light signals can be obscured by snow and ice when the motion of a semaphore signal will shake it off. Phantom aspects of light signals are treated on their own page in this website. They did not cause any reported accidents, though there is one unusual case that was simply due to the difficulty of reading light signals with full sunlight on them.
It is remarkable that signal failures have been responsible for very few accidents and casualties. In the years 1911-1966 in the United States, only 11 passengers have died in such accidents, 8 in a single accident in 1920, and one each in 1915, 1926 and 1953. A computer search of ICC accident reports with the search string "false near clear" turned up only 40 documents, in most of which the phrase found was "no evidence of a false clear." In only 10 was a false clear considered possible, and of these still fewer as probable or proved. Other documents discuss other kinds of signal failures, and will also be considered here. In all, 17 incidents are discussed here, and these seem to be all that occurred in the interval 1911-1966. Note carefully that we do not consider accidents in which signals displaying proper aspects were ignored, a much more common cause.
Most false clears occur when a signal system is being installed, modified or maintained. Before the system is allowed to go into service, rigorous testing removes all of these defects and hazards. However, some accidents involving false clears have occurred because of erroneous modifications made by a signal maintainer while signal or track maintenance is being carried out. Extreme vigilance is required under these conditions to ensure safety. Other false clears occur in normal service, but do not lead to accident because they are discovered and reported in time. These are mainly mechanical failures and damage, or track circuits that do not shunt properly, and are very quickly set right. Finally, we have the unlucky cases in which a false clear results in accident, and these cases are very few. Many of the accident reports contain lists of false clears reported by signal engineers, to show how rare they are, and their characteristics.
Electricity plays a large role in all signal systems, even in those that are considered "mechanical." The fundamental circuit in signal systems is the track circuit, used to detect the presence of a train or cars in a section of track called a block, separated by insulated joints from other blocks. In an unoccupied block, it is a closed circuit through the rails supplied by a power source, a battery or transformer, that holds a track relay energized, closed, or picked-up. When this circuit is shunted by the wheels and axles of any equipment, the relay drops out, closing and opening contacts that control circuits responding to this state. The track circuit must be carefully adjusted that the track relay picks up in all weathers when the block is unoccupied, and reliably and quickly drops out when it is not.
A false clear will obviously be displayed when the track circuit does not detect a train occupying it. When the first Budd cars went into service on 1 May 1950 on the Boston and Albany (NYC) it was immediately noted that they did not reliably shunt track circuits when they made a stop. On 2 May 1950 an absolute block behind them was instituted to ensure safety while something was figured out. The track circuits were made more sensitive, and copper shoes were fitted to the trucks that rubbed against the wheel treads. The copper shoes were bonded to the truck frames. These cars had disc brakes, and it was thought that the wheel treads were not maintained in conducting condition because of the lack of brake shoes. The copper shoes were replaced by cast iron, and the bond wires were removed, on 29 Aug 1952, which made no difference. There were 2 instances of failure to shunt in November 1950, 3 instances in January 1951, 1 in April 1951, and 2 in October 1953.
Finally, on 23 November 1953 an accident due to failure to shunt occurred at Palmer, MA. Car M463 had departed Springfield at 10.04 am, and almost at once experienced engine trouble. Both 275 hp Detroit Diesels had failed at Athol Junction, just east of Springfield. No. 78, a passenger train with two units and 10 cars, leaving Springfield at 10.10 am [apparently so that passengers could not transfer to No. 602 for local stops, and so a fast train could follow a stopping one] was flagged at Athol Junction. When they got the engines running again, No. 602 went ahead and No. 78 followed. The engines ran for a while, then sputtered out. The flagman of No. 602 heaved a lighted fusee overboard 5 miles west of Palmer. The engineman coasted into Palmer with dead engines at 10.45 am, sanding lightly so the brakes would be reliable in the light rain, and it was decided to transfer the passengers into No. 78. The Central Vermont crossed here at grade, and No. 602 stopped with its rear 297' east of the crossing.
No. 78 was running easily, at about 50 mph, as it approached Palmer. Signal 8552, 1.86 miles away, was green over green (Clear). So was signal 8542, 4705' from Palmer, and finally, controlled signal 2, 516' from M463, displayed green over green over red, Clear. No. 602's flagman was back a little, as normal, and saw to his horror that No. 78 was approaching at speed. He did not have time to light a fusee. An emergency application reduced 78's speed to about 30 mph. Nothing was derailed, but the rear vestibule of M463 was crushed and the car knocked forward.
What the signals should have displayed was yellow over yellow (Advance Approach) at signal 8552, yellow over red (Approach) at signal 8542, and red over red over yellow (Restricting) at signal 2. They did not, because M463 was not shunting the track circuit. It had stopped with its wheels on a thin layer of powdered sand that coated the rails. Tests showed that this was the cause of the unreliable shunting, not anything wrong with the wheel treads. If the car were moved forward 5' so that the wheels were not on the sand-coated areas resulting from braking sand, then contact was made and the track circuit shunted. The truck wheelbase was 8' 6", incidentally, and the wheels 33". The weight on each wheel was over 7 tons.
A few years earlier, in 1945, a similar event occurred at East Portal, MA on the Boston and Maine involving a pair of diesel units, light, and three tunnel electrics coupled. The units had gone by a dwarf signal at the entrance of Hoosac Tunnel, and were stopped by an automatic train stop brake application just inside the tunnel. The three tunnel motors came up behind them, and the signal was Clear. It was a shock to discover the diesels ahead, but the collision was not severe, and there were no fatalities. It was found that there was a 1/32" coating of greasy wet sand on the rail head, and the diesels had stopped on it, so the track circuit was not shunted.
The most costly false clear accident occurred at Erie, PA on the New York Central's 4-track main line. As late as 1920, the NYC still employed a switchtender to handle the crossover switches here, though the lines were fully protected by automatic block signals. There is a kind of barbaric efficiency in using a switchtender. The installation is very cheap, no interlocking is necessary because it is obvious where the switches are and what they do, and no signals are necessary because the switchtender has a flag or lamp. On the fatal day, track and signal repairs were in progress in the area. Normally, an open switch in the main track would set the signals protecting it at Stop. The signal maintainer, to minimize unnecessary delays, had bridged the circuit contactors so that this protection was absent, not an intelligent thing to do. A light engine was making a crossover movement from track 4 to track 1, with switchtender Bell opening and closing the switches and riding on the footboard. As he passed signal maintainer Larson, he pointed to a crossover switch, and gestured to ask if Larson would close it, saving him from getting down. Larson said sure, and went over to close the switch. Unfortunately, he just closed the one switch, not the one at the other end of the crossover as well, still not exhibiting great intelligence.
As luck would have it, two passenger trains approached on parallel tracks under clear signals. One was diverted into the side of the other at about its eighth car. The locomotive sheared open the side of Pullman sleeper Peosta, where the 8 fatalities probably occurred, and left a considerable mess otherwise.
A false clear accident occured on the Chicago, South Shore and South Bend near Gary, IN in 1926. West of Gary, the CSS&SB crossed over the EJ&E on a bridge with a gauntlet track 1474' long. There was a 15 mph restriction on the gauntlet track, and signals protecting it had been installed in 1911. These used two-aspect light signals at each end, and alternating current track circuits. The track relay was an unusual motor type of relay, with contacts operated by centrifugal force. When the track circuit was clear, the motor revolved and the contacts were closed.
Train No. 63, made up of a motor and trailer, accepted a Clear aspect at the eastern end of the bridge and started up the incline. Train No. 64, made up of a motor car, approached at almost exactly the same time and accepted a Clear aspect at its end. Both cars were exceeding 15 mph, so when they saw one another they were not able to stop short and bumped, apparently hard enough to kill or frighten a passenger to death. An official in the neighborhood saw the signal at Clear. Then it went dark for a while, and finally some time later became red. It was discovered that the relay had become extremely sluggish, and worked very slowly, if reliably. This is the only case in the ICC records where a track relay was to blame, and this was a very unusual type.
The final accident associated with signals involving the death of a passenger occurred on the Baltimore and Ohio at Felton, PA in 1915. There was no evidence here that anything had not functioned properly, so this is really not a false-clear accident at all. What did happen was a following passenger train collided with a freight train leaving a siding on double track. Of course, the freight train should not have been there, and the signals were a backup. The brakeman who opened the switch said the switch indicator, which told him if a train was approaching, showed clear, while the engineman of the passenger train said the signal he accepted was Clear as well. It was thought that maybe the switch indicator went to red just as the brakeman turned around to open the switch, so is was only unfortunate timing. Actually, the design of the system may not have protected against such an occurrence as well as it should have done.
Signal maintainers have been responsible for several false-clear accidents in addition to the serious one at Erie, PA. In 1936, at Champaign, IL on the Illinois Central there was a relay interlocking at the south end, on the 100 mph main line. In this case, the maintainer got the connections reversed both at the switch and at the control machine, so when the lever was normal the switch was reversed and the signals clear. The Larson-like condition probably did not exist for very long, but long enough for a passenger train to encounter a No. 15 turnout at about 70 mph. At West Belt Junction, PA on the Pittsburgh and West Virginia in 1936 another signal maintainer had been working on the time delay at an electric facing point lock. When the handle was moved, the protecting signal went to stop in 4 seconds, and 1 minute 30 seconds later the electric lock was released, a light went on, and the switch could be opened by hand. Our maintainer wired it so that the protecting signal did not go to Stop until the lock was released. This contributed to a minor accident at that point.
At Alida, IN the B&O crossed the Monon, and there was a mechanical interlocking at that point. One day in 1947, the leverman had some trouble with a switch that he had reversed and wanted to restore to normal. So he applied a little force, and thought he had done it. What he had done is break the operating rod, so the switch had not moved, but the facing-point lock had gone in, making the machine think the switch was normal and locked. The circuit controller for the signals was operated by the operating rod, not directly by the points (a dangerous arrangement) so the signals knew no better. No. 26, with two units and 9 cars, approached the open switch at 76 mph.
In 1946, at Winford Junction, KY on the Gulf, Mobile & Ohio, IC freight Extra 1203 South collided with a northbound GM&O freight No. 28 in interlocking limits, the IC train having received signals allowing it to proceed. There had been some irregularities in the mechanical and electrical arrangments, with broken seals on time releases and other curiosities, but the most curious was a .22-caliber bullet found embedded in a cable that might have had something to do with the situation.
At Stonington Junction, CT there was a failure of a distant signal in 1912. This was controlled-manual-block territory on the New Haven, and the signals were wire-connected as is the practice in Britain. The trouble was a broken chain in the machine, which allowed the signal to continue to display Clear, so that a train was not prepared to stop at the block signal and collided, not very energetically, with another. At Columbus, OH on the Pennsylvania Railroad in 1923 a mechanically-operated distant signal also failed. Enginemen had been informed that the signal was not working, and they should approach the interlocking under control, but this precaution did not suffice.
We are now down to the only two accidents where a false clear was due to electrical circuit faults. These are of three types: failures due to foreign currents (currents originating outside of the signal system); to crosses (undesired contact between two wires in the signal system); and to grounds. Great efforts are made to prevent all three kinds of failure. Any electrical system that operates out-of-doors is subject to great stresses, and must be carefully designed. Foreign currents are best guarded against by using alternating current circuits, since most foreign currents are direct currents, from earth currents (thunderclouds), lightning, magnetic storms and street railways.
A celebrated incident occurred at Rockledge, TN in 1915 on the Nashville, Chattanooga and St. Louis on a very difficult stretch of track where block signals and operation by signal indications had been installed to facilitate movements. A freight train left a siding and entered an occupied block on the basis of a clear aspect of a dwarf signal. The false clear was definitely established, but its cause was less clear. There was work going on in the vicinity on Western Union telegraph lines, and a cross with these is most probably the explanation of the false clear. Since the cross only happened during temporary works, it was not possible to reproduce it, though it was shown that such a cross would indeed cause a false clear aspect.
The other case was in 1938, at Valley Junction, WI on the Chicago St Paul Minneapolis & Omaha. Here a cross between line wires, where the insulation had been worn off and the lines allowed to have too much slack, was responsible. The offending signals showed Clear even with trains obviously in the block. Circuits should be designed, if possible, so that a single cross or ground cannot cause a failure, but two are required simultaneously.
Finally we come to three accidents in which a false clear was claimed to have been the cause by those directly involved, but in none of these cases was any reason for a false clear discovered, though one could not be absolutely ruled out, usually because the necessary evidence had been destroyed in the accident. In many other similar cases, the claim of a false clear could be disproved beyond a reasonable doubt. Whenever an engineman is accused of passing a signal at stop, the response is nearly always that the signal was actually at Clear, supported and sworn to by others in the cab, and the fact is nearly always that it was at Stop.
The accident at Ryan, VA on the Seaboard Air Line in 1942 is a good example. This busy main line was operated by CTC, with the control machine in Raleigh. Northbound Train No. 10 was to meet a southbound freight here, and held the main track. The freight was approaching the north switch, ready to enter the siding, as No. 10 passed a signal at Stop, trailed through the siding switch, and collided with the freight train. Everyone in the cab swore that the signal in rear of the controlled signal showed green, Clear. That signal had been struck by lightning, and the yellow bulb had been shattered, so it had been dark when No. 10 passed it. Everything else was working perfectly. By rule, any such imperfectly displayed signal must be regarded as its most restrictive indication, in this case Stop. It is clear they never noticed the signal at all.
At Moapa, NV, 95 miles east of Las Vegas on the Union Pacific, there was a siding with a dual-control switch at the west end. On 20 August 1942 Second 154 took siding to be passed by No. 8. The switch was controlled by the operator at the control machine, which had a switch lever and a signal lever, and a time release for permitting changing the route. No. 8 arrived at no great speed, since the speed limit was 30 mph at that point, but since the switch was still set for the siding entered the siding and collided with the rear of Second 154. The operator was delayed in lining the switch because of some difficulty with the time lock.
The men on the engine swore both signals approaching the siding displayed Clear, and that they called the aspects to each other in a display of correctness. However, the circuit controller at the switch was open, which would put the controlled signal at Stop or Restricting, and the signal in its rear at Approach. Everything worked perfectly on test, and the signal displayed Stop after the collision. No crosses or grounds were found. What probably really happened was that everyone was snoozing on the engine in the hot August afternoon, and nobody actually noticed what was going on until the bang. There is nothing like an accident to awaken the creative and cooperative faculties.
Many years earlier, on 20 February 1912, there was a collision in the Hoosac Tunnel on the Fitchburg Division of the Boston and Maine that killed four employees, but no passengers. The 4.8-mile-long tunnel was electrified by that time, 2200 V, 25 Hz electric motors pulling the steam engines with their trains through the tunnel from North Adams to Hoosac Tunnel Station, 7.6 miles. Inside the tunnel the track circuits were 60 Hz alternating current, center fed at 10 V, with 3.5 V at the four vane-type track relays, which in turn controlled galvanometer-type AC relays. Light signals were used, with 110 V bulbs, displaying green over green, green over yellow, and red over yellow for Clear, Approach and Stop, respectively. On the eastbound track, the first signal was 7013' from the west portal, the next signal 8031' farther on. Near the east portal was a distant signal for the first signal east of the tunnel. As was typical B&M practice, the insulated joints were in the rear of the block signals by 150', so that an engineman would observe that his own train shunted the block, and be assured that his train was properly protected. White lights were placed at these points.
Extra 2633 East, pulled by Motor 5001, entered the tunnel at 3.45 pm and broke in two at the center shaft. When this was fixed, it pulled forward until the rear 1000' remained in the tunnel, and stopped. No. 2, a Troy-Boston express, entered the tunnel at 4.15 pm, towed by Motor 5004. It stopped at the first signal, which displayed Stop, then proceeded at restricted speed and picked up the flagman left by Extra 2633 near the center. A fusee was encountered, one of those that burned red for a time, and then yellow, but did not wait for it to burn yellow. After passing the second signal, the train accelerated to 35 mph and collided with the rear of Extra 2633. Everyone in the cab of motor 5004 was killed, so we do not know what they would have said. Others claimed that the second signal showed green over green, and this would have explained the acceleration. Fire broke out after the collision, but the steam engine of No. 2 was able to back the train out of the tunnel, and only two passengers were injured. Because of the fire, there was no opportunity to check that the signals operated correctly, but there is no sign that they did not.
Finally, we have a case of a failure that did not cause an accident, but removed protection that might have prevented or ameliorated it. This accident occurred at Ripley, NY on the New York Central main line not far from Erie, on 8 February 1957. By this time, the four-track main line had been reduced to two tracks signalled for movement in both directions, controlled from Erie. There was a crossover at Ripley by which a westbound train on track 2, the southern track, could return to track 1, the northern track. Extra 1704 West had been put on track 2 at Brocton to let passenger train No. 43 by on track 1. This had happened, and now Extra 1704 was crossed over to track 1 at Ripley, since No. 74 with two units, 3 baggage cars and 3 coaches was coming east on track 2.
It was late in the afternoon, and the low sun was shining brightly on signal 672E, the penultimate signal before the crossover at Ripley, which was showing Approach. The engineman, however, read it as Clear. He felt that this was confirmed when he received no warning whistle from the intermittent inductive train stop device, which he would have done if the signal were displaying Approach. The next signal was displaying Stop, and although the brakes were immediately applied, No. 74 hit the 27th car of Extra 1704 at the fouling point of the crossover at 28 mph. The time was 4.29 pm. Fortunately, there were no fatalities.
What had happened was that the wayside inductor at signal 672E was faulty. Inductors are placed about 70' to 90' in rear of the signal, outside the right-hand rail, where a matching receiver on the locomotive passes over them. An inductor without a winding will always give a train stop event, sounding an air whistle in the cab, and causing a penalty brake application if the forestalling lever is not pulled back. Such dead inductors are used for testing the ATS apparatus, and at severe permanent speed restrictions. A wound inductor with an open circuit will act in the same way. The circuit of the winding of the inductor at signal 672E should have been open. When the signal displays Clear, a current passes through the winding that produces a magnetic field that exactly cancels the one produced by the locomotive equipment, so that the automatic stop does not trip. In this case, water had entered the winding and destroyed several patches of insulation, so the winding was effectively short-circuited. Now, when the moving inductor passed by, a current was induced in the short-circuited winding that prevented the magnetic field from building up, just as it would if the inductor were energized, and so no trip resulted. This was a quite unexpected fault, showing that ATS inductors were more complicated than believed. It also shows what bad effects water can have on electrical apparatus, which is not unexpected news.
Composed by J. B. Calvert
Created 21 March 2002