Automatic Signals For Welded Track

[ChuckHackett-844]

As some may know, I have developed and sell a system for automatic signals designed for Riding Scale Railroads (http://MiniRailSolutions.com).

This system uses "Track Circuit Detection" to detect trains on sections of track. This requires that the rails in each section of track be insulated from each other (I will leave the details of track circuit detection and dealing with wet/dry track, wheel noise, etc. for another time).

The system works well but, there are many tracks that use steel bar stock with welded steel ties/gauge bars or similar construction where the rails are not insulated from each other and thus track circuit detection of trains is not possible/practical.

I know that there are many tracks that use engineer operated buttons, toggles, etc. to control signals but here I am interested only in Fully Automatic Signals.

There are methods of detecting trains entering & leaving a block that can be used in these situations, some are:

• Short sections of insulated rail to detect axle shunts
• Counting wheels
  • Mechanical (micro-switches, etc.)
  • Optical sensors
  • Proximity sensors (assumes ferrous wheels)

Some of the problems with these sensors is:

• They are subject to activation by pedestrians, animals, etc. and, in the case of optical sensors, falling leaves (which can also block optical sensors).
• All of these methods need to be able to distinguish direction
• They must handle the case where half of a train passes over the sensor(s), stops, and then backs up - thus possibly resulting in a count error (i.e.: wheel stopped on top of a sensor).
• They must be able to handle extremely short trains: e.g.: A "peddle car", "Motor Car" or other 2-axle vehicle.

Some of these issues can be discarded if a railroad says "we will just deal with that exception". The problem with this is that, if you tell visiting engineers that you have "Automatic Signals" they expect them to work in all situations. (Trust me, you can tell engineers "do not backup after starting to enter a block" but they WILL).

The problem with "wheel Counting" in general is that the system must handle the situation where there is a 'count error' where, after the train leaves the wheel count does not go to zero because of a counting error going in or out.

There is also the case where, half way through the block a car failed and was set off the track. The train then exits the block and it appears that a car is still in the block - thus the block is still marked as occupied.

Many who try this handle the 'count error' situation by having a timer on the track occupancy - i.e.: if the track has been occupied for longer than, lets say, 5 minutes the system will "assume" that the train left and that there was a count error and thus it will clear the block. That's fine in many cases but do you want to be working on your derailed locomotive and have someone come around a blind curve at you? not me ...

There are full size railroads that use wheel counting but their sensors are much more accurate, durable, ... and more expensive ... than what the average hobby railroad would use, and, if they do get a miss-count, the block remains out of service until another train traverses the block at dead-slow speed while in constant contact with a human dispatcher. After that, the block is cleared by the dispatcher. This does not seem like a procedure most hobby railroads would tolerate.

Some tracks will use a detector on the ENTRY and EXIT tracks of a block and assume (there's that word again) that, when a train trips the ENTRY detector he will not back up ... same as the miss-count situation, the block remains occupied but there is no train in the block. Also, this only works in simple passing siding situations and does not work where a block contains ENTRY points that are also EXIT points.

So, my question is, has anyone seen an approach to Fully Automatic Signals that does not use track circuit detection and deals with welded ties?

[Mark Landgraf]

Chuck,

Well you could use loop presence detection like used in highway traffic signals. The big rr's use these at quad gate grade crossings to make sure the crossing is free of autos before dropping the 3rd & 4th gates. Sensitivity can be adjusted so as to tune out the rail structure. This would require burying a wire down one side of the block and have it return back up the other side. Trains can start and stop and still remain Detected. Detached cars will remain detected where they sit.

Many transit systems, including Amtrak, use an on-board and track mounted transducers to actively transfer information between the loco and specific track side locations, usually the entry and exit of blocks. This method assumes you have your whole train, and that the train is shorter than the block length. This system requires the system to remember that a train is in the block. Trains that are stopped for a long time may cause problems. This requires all trains to be equipped. Not an ideal situation for visiting trains.

On DC electric railroads they used a AC imposed current and Weezee induction units to pull off the AC signal current and the propulsion DC going to ground. This was dependent upon the propulsion current going to ground at the ends of the track circuit. I can't see it working in your situation, even thou their rails were connected together. When AC propulsion was in use, the propulsion frequency was either 25 cycle or 60 cycle. The typical signals were 100 cycles when propulsion was 60 cycles. Obviously, 25 propulsion and 100 signal were not a good combination.

Tokens and token holders are used in India for branchline operations.

Chicago North Shore had motorman operated hand electric switches mounted on poles at that each insulated joint.

The last two don't meet your fully automatic criteria.

Personally, I like the first choice best. One of the loop detection manufacturers is Safetrans

Mark Landgraf
Albany NY

[rkcarguy]

I think you've touched on all the options, it's tough with welded track. I think the loop presence system is probably the best option, or its time to start cutting and isolating one side of your welded track. FYI, I ran into a situation on my turnout points where my point rails crossbar will conduct electricity and mess up my signal system. I ended up drawing up a bar with 4 holes and a thin section in the middle that I had laser cut. My plan is to transfer drill the 4 holes through some phenolic sheet and bolt it all together, and then cut off wheel the thin section out of the steel bar and voila I'm isolated and still supported/gaged.

[ChuckHackett-844]

Well you could use loop presence detection like used in highway traffic signals. The big rr's use these at quad gate grade crossings to make sure the crossing is free of autos before dropping the 3rd & 4th gates. Sensitivity can be adjusted so as to tune out the rail structure. This would require burying a wire down one side of the block and have it return back up the other side. Trains can start and stop and still remain Detected. Detached cars will remain detected where they sit.

To berry (bury*) a presence loop on each section of sampled track would require berrying a wire down both sides of the track and the AC excitation and sensing module would be a significant expense. For a crossing gate this might be practical but not for an entire railroad.

The sensing thresholds would be a nightmare having to handle a train made from aluminum (my 10' passenger cars are all aluminum) or steel and size down to a peddle car. This would not be practical for a railroad of any size. I deal with customers with large railroads, some 50-60 signal heads and approaching 100 sensed track sections. That would mean 100 loops and sensing modules. At that point it would be cheaper and easier to insulate the rails ... which is also impractical unless they REALLY wanted automatic signals.

Many transit systems, including Amtrak, use an on-board and track mounted transducers to actively transfer information between the loco and specific track side locations, usually the entry and exit of blocks. This method assumes you have your whole train, and that the train is shorter than the block length. This system requires the system to remember that a train is in the block. Trains that are stopped for a long time may cause problems. This requires all trains to be equipped. Not an ideal situation for visiting trains.

Years ago I worked on Union Pacific's implementation of the Advanced Train Control System (ATCS) which used RF transponder capsules in the roadbed (fancy RFID tags). RFID tags are cheap but the readers aren't. If the reader is on the loco this leaves the need to get the info back to the signal system (e.g.: radio), etc. and, as you mention, it doesn't address the 'split train' problem. Again, I doubt if any of my customers would want to go that route.

I think you've touched on all the options, it's tough with welded track. I think the loop presence system is probably the best option, or its time to start cutting and isolating one side of your welded track.

Maybe in very limited situations but, for reasons stated above, impractical for an entire railroad of any size.

FYI, I ran into a situation on my turnout points where my point rails crossbar will conduct electricity and mess up my signal system. I ended up drawing up a bar with 4 holes and a thin section in the middle that I had laser cut. My plan is to transfer drill the 4 holes through some phenolic sheet and bolt it all together, and then cut off wheel the thin section out of the steel bar and voila I'm isolated and still supported/gaged.

Yes, We have used insulating plastic shoulder washers and full plastic bridals. The full plastic bridals worked best. The shoulder washers tended to ware or break down in the hot sun.

So far it seems like the only solution for automatic signals on welded track is sensors with timeouts. If I have to, I do have some ideas on how to minimize the block hitting that timeout and possibly clearing it falsely but I am looking for a solution that gives near the reliability provided by track circuit sensing ... I have not given up, still looking for ideas ...

(*) Intended to aid in doing a search

[rkcarguy]

The only other thing I can think of is a set of garage door opener type sensors aimed down the section of track *lengthwise*. Adjust the height so it hits at coupler height and it should catch everything. Putting an adjustable delay module on it of several seconds should omit tripping from falling leaves and passing animals. If the track turns though, you'll need another bank of sensors or at least carefully aimed mirrors/optics, and I imagine at more than a certain distance apart they will stop working. Just a thought, being you are looking for ideas.

FYI, the phenolic I mentioned above is an excellent isolator and not subject to UV degradation. The only thing is once it's cut, the open edges must be sealed with resin or epoxy or the fibers in it can wick up moisture.

[ChuckHackett-844]

The only other thing I can think of is a set of garage door opener type sensors aimed down the section of track *lengthwise*. Adjust the height so it hits at coupler height and it should catch everything. Putting an adjustable delay module on it of several seconds should omit tripping from falling leaves and passing animals. If the track turns though, you'll need another bank of sensors or at least carefully aimed mirrors/optics, and I imagine at more than a certain distance apart they will stop working. Just a thought, being you are looking for ideas.

Yup, we have considered that at a concrete level crossing on a current railroad but, even with timers it's not perfect - Harry standing on the track talking to Joe, etc. ... unless we use CO2 lasers which would punch a hole through Harry's ankle

Seriously though, a lot would be required on a curve and mirrors get dirty, receivers and senders get tripped on and misaligned, etc.

But good thought ...

FYI, the phenolic I mentioned above is an excellent isolator and not subject to UV degradation. The only thing is once it's cut, the open edges must be sealed with resin or epoxy or the fibers in it can wick up moisture.

I am familiar with phenolic but didn't know about sealing it - thanks.

[John Hasler]

This has been done using rf propagating along the track rather than DC, though I can't find a link right now. I'm not sure I could make it work if the track is shorted at every tie but because of the way rf transmission lines work a short at a switch or similar simply ends the block. Lots of engineering required, though.

[ChuckHackett-844]

This has been done using rf propagating along the track rather than DC, though I can't find a link right now. I'm not sure I could make it work if the track is shorted at every tie but because of the way rf transmission lines work a short at a switch or similar simply ends the block. Lots of engineering required, though.

Yes, this is called Time Domain Reflectometry or "TDR" (https://en.wikipedia.org/wiki/Time-domain_reflectometer). It doesn't use RF but rather a very short pulse.

If the tracks were not shorted by the welded ties this method can determine not only if there is a train on the track but exactly how far down the track it is. Since these tracks are shorted TDR would always show a train present. A great system for non-shorted track but lots of electronics required for each track connection.

Before other engineers get on me for the above: Yes, it is 'possible' to see past shorted ties but it would require LOTs more electronics and I doubt it would be able to 'see' very far. This is similar to Radar trying to see through heavy rain or ultrasound mapping layers in metal, etc. by detecting multiple echos.

[John Hasler]

A "very short pulse" *is* rf. Look at it in the frequency domain.

However, the system I can't find a link for uses cw.

I can think of several ways to get rf of the proper frequency to propagate well along track shorted at every tie. A transmission line need not have two insulated conductors. Think about waveguide. Huge amount of engineering to get it working, though, and it would be tuned to that particular track construction. Not really practical here.

[Edit] A reference:

https://digital-library.theiet.org/cont ... ew032e_ch4

[kcameron]

Chuck,

As you are seeing the ways to detect with welded ties is not easy. I'd stick with trying wheel counting. Key there is they use multiple sensors to see which way each wheel is moving. That's solves the forward/backward issue. These sensors are less than a wheel width apart. Optic vs metal detection, that's a different story. Same for if the sensors should be one sided (just detecting or reflecting something) or two sided (emitter/receiver), both have good and bad tied into them.

But to balance how fancy or not to get the hardware, I'd say for our railroads, a reset button here or there for a block would be the right way to go. The other would be a normally quite long timeout reset. I know while doing operations you can be in a block for 30 minutes on a branch line pretty easy. At least an overnight timeout like Michigan Central that reboots every middle of the night, just in case somebody left a block busy (manual block system). Maybe an hour or two of no movement would be a good reset if the count was less than five or something. But you could leave that as a user config option.

[ChuckHackett-844]

I can think of several ways to get rf of the proper frequency to propagate well along track shorted at every tie. A transmission line need not have two insulated conductors. Think about waveguide. Huge amount of engineering to get it working, though, and it would be tuned to that particular track construction. Not really practical here.

Yes, but I don't think folks are going to place crossties at an exact multiple of the wavelength just for me

Yes, an interesting engineering challenge but hardly practical ...

[ChuckHackett-844]

As you are seeing the ways to detect with welded ties is not easy. I'd stick with trying wheel counting. Key there is they use multiple sensors to see which way each wheel is moving. That's solves the forward/backward issue. These sensors are less than a wheel width apart. Optic vs metal detection, that's a different story. Same for if the sensors should be one sided (just detecting or reflecting something) or two sided (emitter/receiver), both have good and bad tied into them.

Yes, that is along the same lines that I have been thinking (using quadrature encoding of multiple sensors) and sense both sides to eliminate/reduce false trips. Apply the "KISS" principal to the sensors to keep them reliable and inexpensive and make up for the shortcomings in firmware.

But to balance how fancy or not to get the hardware, I'd say for our railroads, a reset button here or there for a block would be the right way to go. The other would be a normally quite long timeout reset. I know while doing operations you can be in a block for 30 minutes on a branch line pretty easy. At least an overnight timeout like Michigan Central that reboots every middle of the night, just in case somebody left a block busy (manual block system). Maybe an hour or two of no movement would be a good reset if the count was less than five or something. But you could leave that as a user config option.

If we drop the requirement for detecting a dropped car or two some judicious logic applied to how close the count is, as well as noting that 'most' of the train just left, etc. can address most of the issues. It just irks me that I can't narrow the error to 0.

I see you are on the JMRI team - When I started this 12 years ago my thought was to use the JMRI display capabilities tied to my field logic (me providing block logic and just feeding JMRI occupancy for display and accept switch commands, etc. from JMRI). Alas, in the end it was easier for me to write my own display front-end which I needed anyway for track voltage strip charts, diagnostics, etc.

I liked the display configuration capabilities of JMRI but wanted the distributed logic capabilities I had (my system runs without central PC).

If you know anyone who would like to develop a JMRI display interface to my CAN data bus I'd be glad to provide message formats, etc. Who knows, it might replace my display. A plus for folks who want a truer CTC panel presentation.