I'm actually a mechanical engineer, so I'm not very versed in electronic design, here goes:

I want to design an inductively-coupled network connector for the rail industry. The goal is to have a data link along the whole train.

Some key points to consider:

  • There is currently no electrical network across the train. Automatically coupling contacts connected to the mechanical coupler have been tried but are extremely difficult to make reliable. What does work is having axle generators and buffer batteries on each wagon.
  • I want to be able to "bypass" a "dead" wagon (wagon with nonfunctional electronics) inside a tunnel. This pretty much rules out wireless systems, as I believe they will have great difficulty to reliably send RF signals around the dead wagon. I was thinking of using the following network topology to solve this: Network Topology
  • I'm going for a physical connection here because when you connect a string of wagons, they all automatically register to the locomotive. Trains get built in yards with a lot of wagons, so you want to be sure that you're not incuding wagons on the adjacent track in your wagon list through WiFi.
  • I really need only very minimal data rates. Mainly you will have the loco telling the wagons every 100 ms to not brake (fail safe) or *brake at X% and the wagons reporting back that they are all still there (train integrity.) The other data is not time critical. Video is nice to have (to be able to back up the train with a rear-viewing camera without leaving the cab,) but a non-reliable link via normal public LTE would be fine.

I had a look at the electrical setup of Ethernet. I saw that there are always pulse transformers for galvanic isolation (and a commom mode choke between the cable and the board.)

I had the idea: Why not make a physically separable pulse transformer to transmit magnetically? I wanted to build a demonstrator network which looks something like this:

3Wagons setup

Here are my questions:

  • If I buy an Ethernet pulse transformer, saw open the ferrite beads & re-arrange and re-wind them to make them physically separable, is this likely to actually work (provided the bead halves are well-aligned without an air gap?) Any recommendations on what to buy and tinker with?

  • In the comments you told me that Ferrite is very brittle. Also the beads are very small. What sort of losses and what sort of frequencies would be possible with a ~10mm mu-metal sheet core? I understand that this would no longer be within Ethernet standards. Any recommendation to use an Ethernet-style slower network?

I'm very happy for any help I can get, thank you!

Here is visualization of the idea on the coupler head with 3 pot-cores set in resin inside a sturdy milled case:

enter image description here

*Here is some background to my task: I'm in an industry comittee looking into freight rail automation. We want to design a highly reliable communication network for freight railroad cars.

The main problem is that the rail industry is very conservative, so everyone goes with the purportedly "tried & tested" design. In the passenger transit area, electric couplers have been in use for around 100 years with hundreds of individual electric pins (the rail industry still a fan of hard-wired functions) which wear, have to be kept clean and are protected by flaps/doors and are often heated to avoid problems with moisture. (You can see the doors open and the bare pins on each side of the coupler at minute 1:10 in this video.)

There is just no way that this is going to work reliably in a freight environment. Even if we assume passenger-level care for the couplers, in the passenger world you couple 2-3 units maximum, not 30-100.

That is why I wanted to design a system which is powered by axle generators and buffer batteries, but transmits via a physical ink on the coupler via an inductive connector. I would like to avoid simply going fully wireless because of 2 reasons:*

  • I want a physical connection so when you connect a string of wagons, they all automatically register to the locomotive. Trains get built in yards with a lot of wagons, so you want to be sure that you're not incuding wagons on the adjacent track in your wagon list.
  • Everyone is terrified of hackers, I think it would be easier to prove functional safety & information security (against information tampering, not eavesdropping) on a physical connection.
  • \$\begingroup\$ I'm pretty sure the reason this isn't done is 1) cost and 2) reliability. 8p8c connectors are cheap, and magnetic materials would drive the cost up a significant amount. And it doesn't matter where the isolation barrier is, it just has to exist. As for reliability, getting good enough magnetic coupling to transmit data reliably would require very brittle exposed ferrite surfaces on the plug, and if those got chipped you'd lose coupling strength pretty quickly. Plus, this wouldn't be compatible with PoE. \$\endgroup\$
    – Hearth
    May 30, 2021 at 16:52
  • \$\begingroup\$ Wifi is not really an option for the reasons stated at the end of the Post (safety, security, train integrity) Concerning ferrite chipping, perhaps there is some special other material? At any rate it will be more robust than a plug because a) there are no little sprung tongues and b) it should be pretty impervious to water and dirt. Cost is not really an issue either: The electric head currently under consideration with all the contacts, doors, springs, seals, bells & whistles costs a lot. \$\endgroup\$
    – SWKRail
    May 30, 2021 at 16:58
  • 1
    \$\begingroup\$ How much data per wagon, how many wagons and how quickly? I would be thinking of an industry hardened one-wire protocol. See ti.com/lit/pdf/spma057 for the standard one-wire enumeration method. \$\endgroup\$
    – Transistor
    May 30, 2021 at 17:05
  • 1
    \$\begingroup\$ @SWKRail No, I think the solution is a well-designed system of electrical contacts, made of beryllium copper with a thick gold plating. Perhaps beryllium nickel if you need high-temperature performance. Use self-wiping contacts. NEMA connectors or other household power connectors might be a good inspiration, they're cheap, self-wiping, and good ones are very durable. \$\endgroup\$
    – Hearth
    May 30, 2021 at 17:06
  • 1
    \$\begingroup\$ RF bypassing of a single wagon should be 'easy enough'. | Magnetic coupling connectors with clearances of some mm also easy enough - allowing outer plastic or other shells for protection. I have done simulate with power and data transfer. \$\endgroup\$
    – Russell McMahon
    May 31, 2021 at 13:26

7 Answers 7


My answer is forcibly high-level as the problem is multi-faceted and allows different solutions.

Objective: transmit data between wagons of a diesel train that has no electric train lines , including reachability of wagons despite one in the middle may be without electric supply. (hope it is correctly synthesized)

Solution1. Wireless connection. It works, there are several examples i the literature and in practice of such connection, e.g. for passenger WiFi and onboard cameras (safety). Even in tunnels propagation at some tens of meters is viable: there are publications in the literature for propagation characteristics inside passenger wagons (yours I think are solid freight wagons) or in the air gap between train and tunnel wall. At 900 MHz and 2.45 GHz I did myself tests of the channel response. -- as for the mentioned mis-registration, there are code words, different networks (SSID), etc., so I would not say it's an issue when doing train composition. When the train is assembled, though a wifi connection the operator can confirm the wagon code and attach it to the so created network.

Solution2. Wired coupling, using e.g. an automatic coupler. Done tests in Cairo last week where the Rotem trains have mechanical and electrical couplers (separated) working in automatic mode. The signals are passed through a 74 pin connector. There is some work for protection against environmental conditions, but it can be done and there are already examples.

Solution3. Wireless transmission in the near field, using a magnetic transmitter that works e.g. within 1 m near the mechanical coupler. Or a radio transmitter, such s those working at 430 MHz, 868 MHz, or the multi GHz range that was mentioned above. What you need is to modulate your signals to pass them through the so established channel.

The overall scenario should include the time for development of solution and the "market size", that is if you want to develop a specific solution, or something that can be reused (and thus need to be a bit more flexible and with more performance than you need now, e.g. in terms of number of channels and throughput).

Last brick: there is always a safety aspect, that is reliability, availability and safety, because some of the exchanged signals may disrupt circulation or implement safety-relevant functions. If a CBTC works using wireless, then no issue, it will work also this one. But should be engineered for at least safety aspects (e.g. watchdog, packet/message loss, lack of response, etc.).

apologies for the high-level approach, but it's 30 years of working with trains and metros: it's not rocket science, but some aspects must be dealt with some care.

  • \$\begingroup\$ Thank you for your response. In Solution 3 you are talking about a near-field magnetic transmitter that works within 1m of the mechanical coupler. How does that work? Can you really transmit enough energy through a 1m air gap twice (before and after the "dead" wagon) to still have a viable signal at the end? I would have thought you'd need ferrite cores to shape the flux and 1 mm gap maximum to keep losses low enough to have something detectable after the 2nd air gap. \$\endgroup\$
    – SWKRail
    May 31, 2021 at 12:28
  • \$\begingroup\$ You can use simply two coils as for wireless power transfer (to transmit power you need shorter distance similar to electric vehicles, and you would supply the dead wagon from adjacent ones). Or a RX/TX system such as a Bluetooth, or a 430/868 MHz short-range radio: it will work over 1 m (inter-wagon gap). The radio TX/RX can also bypass the dead wagon, but some tests should be made with a mockup. All these solutions need some modulation/demodulation. (Of course you need beforehand a good description of the scenarios and environment to derive the specifications.) \$\endgroup\$
    – andrea
    May 31, 2021 at 14:17
  • \$\begingroup\$ So basically what you are suggesting would be to put an active powered component on the spare "bypass" line and power it via inductive resonant separable transformer in case the node in the wagon fails, instead of trying to push a signal through several separable pulse transformers? \$\endgroup\$
    – SWKRail
    May 31, 2021 at 14:29
  • \$\begingroup\$ That's a good solution,but it depends on the amount of power. Maybe you can separate non-interruptible loads and supply through this system, wheres the rest can be only supplied locally through the alternator. A pancake coil as for electric vehicles may work well and they can transfer a lot of power. -- For signals only you can use a RX/TX at radio freq and you can even skip the dead wagon, because it would be a stretch of max 30 m or so. \$\endgroup\$
    – andrea
    May 31, 2021 at 15:26
  • \$\begingroup\$ As stated previously I don't really want to be able to send signals long distance because that it creates a lot of extra headaches with having to discriminate between signals by the adjacent wagon and other signals. I would prefer it to have a max range of below 1m. Maybe there is a kind of two-way NFC for this? Actually having the adjacent wagon power the active node in the middle of the wagon allows to jump 2 adjacent non-functioning wagons, right? \$\endgroup\$
    – SWKRail
    May 31, 2021 at 15:42

Mainly you will have the loco telling the wagons every 100 ms to not brake (fail safe) or *brake at X% and the wagons reporting back that they are all still there (train integrity.)

Ok so this is safety-critical so you have to conform to some functional safety standard, IEC 61508 or rather the train-specific spawn-off (I don't remember the number). This rules out a whole lot of technologies - for example Ethernet isn't real-time so it can probably not be used. Nor can you use wireless unless the link has been explicitly designed for safety-critical applications.

I don't know all that much of this area of application, but there will already be standards for this, so you shouldn't need to re-invent the wheel. https://en.wikipedia.org/wiki/Train_communication_network. That site suggests something called "MVB" or alternatively CANopen. The major advantages of using some existing bus system is that you'll save tons of effort in design and safety classification. They will also have solutions for how to connect the wagons together, which will be one of the main problems in the project.

CAN bus sounds fairly ideal if you can solve the problem of reliably connecting it between wagons. It's very rugged, multi-drop and the distances won't be a problem. Redundancy can easily be solved by two parallel buses sending the same info.


Congratulations for taking the effort to investigate new solutions to existing problems. Even if your effort doesn’t yield anything useful, you will have learnt something.

My suggestion is to use something like the ST60 ( st.com) very short range wireless link. These operate at 60GHz and have centimeter range and very high speed data transfer. This is an emerging technology , so its on the bleeding edge ( like the leading edge but might have some rough edges ).

I’d envisage you would have these embedded in some form of coupling that you could quickly mate and unmate which would also give a degree of shielding to take care of any leakage of the radio signal (at 60GHz this doesn’t take much). Then have some form of indication that the link is good.

Then there’s the work to ensure reliability and you’ll need some form of redundancy. For secondary validation you could have rfid tags - this would mean that for a hack to occur they would have to fake two reasonably secure systems. This would make the likelihood a very small number.

Of course, if you make millions from this, be sure to send me my 5%!

Note, we’re not supposed to recommend products here so the usual disclaimer- i am not affiliated with ST and I have not used that product. The link is to illustrate the type of technology.

  • \$\begingroup\$ Is a 60GHz wireless link involving a whole bunch of electronics and processing overkill when a simple transformer might do? \$\endgroup\$
    – user253751
    May 31, 2021 at 8:26
  • 2
    \$\begingroup\$ Thank you for the Idea @Kartman , I suppose I forgot to state why I want to go for passively wired instead of active Wireless: You have to assume that a single car might go "dead" inside a tight tunnel. In this case, short range wireless will not work because it cannot bridge the dead car gap, long range wireless might have difficulty getting around the wagon and adds the possibility of connecting adjacent cars during train building. \$\endgroup\$
    – SWKRail
    May 31, 2021 at 8:33

Something like a coaxial Ethernet?

See here https://en.wikipedia.org/wiki/10BASE2 but I am sure you will need different connectors in railroad conditions.

Sure, it is not of much use today, but has the proper topology for the task.

  • Only 2 pins to care for.
  • You can stack segments and devices as long as the end is properly terminated (the terminator may double as a cap)
  • "dead" devices don't interfere with the rest of the network as long as the cable is not damaged.
  • 10mBit/s is pretty much enough for a rear camera and still doesn't impose hard requirements over the cable
  • 100m / 300ft range - and this is not really a hard limit and can be extended if you are not latency-obsessed. Do you have longer trains?

What can be a problem: Damage some segment (short or open) and the whole network goes down, not just the part after the damage. A "dead" wagon may be electrically dead, but it has to be at least properly wired for this to work.


I was thinking of using the following network topology to solve this: Network Topology

Fault tolerant addressable LED strips use the same topology to bypass one dead chip. It works.

In the comments you told me that Ferrite is very brittle. Also the beads are very small. What sort of losses and what sort of frequencies would be possible with a ~10mm mu-metal sheet core?

Using exotic and expensive materials like mumetal is not a good idea.

There is already a big chunk of iron connecting the two trains, so the first thing I'd try would be to wrap a coil along the green line:

enter image description here

Then wrap a coil in the same place on the other side of the coupler. Two coils around a chunk of iron make a transformer that can be used to transmit signals. Measure what sort of coupling you get between the two coils. Optimize number of turns and pick a frequency. It will probably be pretty low, which is fine.

The magnetic circuit is closed loop and returns through the rails.

If this works, the nice thing is the magnetic coupling is done through a huge chunk of metal that has already been proven, instead of tiny coils with tight tolerances that will be worn down by vibrations and will have to be tested and redesigned until they last for a million miles, which will take years.

The protocol should be something really simple, like slow AM-modulated serial, with error checking code. Really low tech on the physical level, but with smart software.

This does not solve your "skip wagon with no power" problem. But, since the frequency is low, you can use a relay to bypass the unpowered board. If relay switch reliability is a problem, a solid state MOSFET relay with depletion FETs can be ON without any driving voltage.

  • \$\begingroup\$ "The magnetic circuit is closed loop and returns through the rails." - I'm a little bit concerned about this, although my background is with rapid transport and old railway museum stuff so I might be completely off-base. If the train is crossing an insulated rail joint (which I'd expect in places with non-trivial signalling like yards, interlockings, and between blocks), or worse, is stopped across the joint, you might not be able to get any signal through just the rail joint capacitance alone. This could be a problem if wagon brakes cannot be released once the train is stopped there \$\endgroup\$
    – nanofarad
    Jun 4, 2021 at 22:03
  • \$\begingroup\$ Extending that, if the coupler could be doubled up with another mechanically robust coupler below (probably swinging together with the existing coupler as one unit), that could create a return circuit but seems mechanically costly and complex \$\endgroup\$
    – nanofarad
    Jun 4, 2021 at 22:05
  • 1
    \$\begingroup\$ This would be magnetic coupling through a coil, not capacitive. So there would probably be slightly lower coupling if one wagon and the other sit on unconnected rails. But both cases should give a usable signal. You'd have to test it. My point is to dodge the design effort and cost of a new reliable signal coupling. A few hours testing this would probably be worth it, even if it doesn't work. \$\endgroup\$
    – bobflux
    Jun 4, 2021 at 22:15
  • \$\begingroup\$ Yep, sorry, complete brain fart there when typing. I don't have the gap length and permeability numbers for the joints offhand, so I agree that further testing would be required. \$\endgroup\$
    – nanofarad
    Jun 4, 2021 at 22:32

I understand that the main problem is unreliable conductor contacts here. I would use the wires but with frequency modulated signals. This way it has the possibility that dirt's capacitance will keep its impedance low and signaling will go on. That being the physical layer, a basic master-slave scheme will fulfil the mission. As the transceiver circuits are parallel nodes on the line, the nonfunctional wagon will simply serve as the part of the line.

For the braking system, there also would be a backup mechanism indeed, which might be called "braking runaway", maybe, I just made it up :) If the wagon has acceleration sensor, it is easy to detect a negative acceleration and decide to activate the brakes, thus what runaway implies, it should be enough braking the main locomotive for the braking runaway. Maybe the wagon would prefer to use this mode as a fallback when it choose not to trust the electrical communication.


is this likely to actually work?

No. You have suggested a passive receiver at one end of the 'dead' car, and a passive transmitter at the other end of the 'dead' car. So the passive receiver doesn't just need to receive low-data-rate information, it needs to receive enough electrical power to achieve the transmission at the other end of the car.

It is possible to put a very high power transmitter at each transmitter, and a very sensitive receiver at each receiver, coupled by your dead-car passive system: satellite systems have that kind of technology. But it doesn't just work, and it can't just work by splitting open and rewiring an ethernet socket.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.