Model railroad: reduce the max speed of trains?

[Thyristor42]

Hi all. I need your advice and ideas regarding the following problem:
I'm a member of a model railroad association where we have a few dozen of different trains and locomotives driving around on our tracks. But some locomotives are much faster than other ones on the same piece of track.
The trains are all analog*. This means that a DC voltage is applied to the tracks which gets picked up by the wheels of the locomotives. The wheels are then connected to a brushed DC motor inside the locomotive. Applying the voltage to the tracks and making sure two trains can't collide is done with a central system of analog circuits. The control circuits are fairly old so they can't recognise which train is currently driving on a certain piece of track.

Some specifications:
 - Track voltage: up to ~18V when a train can drive at full speed, but it's lowered when the train needs to slow down e.g. to stop at a red signal.
 - Motor current consumption: 0.5A to 1A (depends on the type of locomotive).
 - The locomotive needs to be able to drive forwards and backwards (positive and negative polarity), and also with 'pure' DC and PWM voltages.
 - The trains are at scale H0 (1/87), so space is limited

What I am looking for is a circuit or so to add to the fast running locomotives in order to limit their speed. What's often done so far is to add a few diodes in series with the motor to reduce the voltage across the motor. But that has as disadvantage that the speed curve of the locomotive is just moved up: if a locomotive normally starts to drive for 2V then with the diodes the locomotive only starts driving at e.g. 5V.
Two ideas I had:
1. Add some PTC thermistors in series with the motor to limit the current. It's a simple circuit but I'm not 100% sure it will have the desired effect.
2. Add an LDO in series with the motor. This is more complicated because I would need to add a second one at the other side of the motor for the case when the locomotive needs to drive backwards (+ schottkey diodes in parallel to bypass the LDO's). I'm also a bit doubting the transient behavior of the LDO when a PWM signal is applied. And also the BEMF of the motor might be annoying, so I think some zener clamp should be added across the motor to limit the voltage.

What do you think of these ideas? Do you maybe know something better?

Thanks!

* Yes, I know digital (DCC) exists and nowadays is the preferred way of controlling model trains. But it's not feasible to convert all the track and trains due to the cost and effort to do the conversion.

[fourtytwo42]

So is it pure DC or PWM ?
Does it have feedback ?
HO is huge, I used to fit zener diodes in N gauge steam!!

[Thyristor42]

So is it pure DC or PWM ?
Does it have feedback ?
HO is huge, I used to fit zener diodes in N gauge steam!!

It should work with both pure DC and PWM: The track is a loop which connects at both ends to a station. On the loop the trains are seeing pure DC, while in the station PWM is applied.
Regarding feedback:
 - for the pure DC section of track: No
 - for the PWM section: I'm 99.9% sure it doesn't use feedback, but I would need to double check to be 100% sure

[Whales]

But some locomotives are much faster than other ones on the same piece of track.
[...]
The trains are all analog*. This means that a DC voltage is applied to the tracks which gets picked up by the wheels of the locomotives. The wheels are then connected to a brushed DC motor inside the locomotive. Applying the voltage to the tracks and making sure two trains can't collide is done with a central system of analog circuits. The control circuits are fairly old so they can't recognise which train is currently driving on a certain piece of track.
[...]
What I am looking for is a circuit or so to add to the fast running locomotives in order to limit their speed.

Q1. At the moment is the anti-crash system effective?  How does it work?  Turns off power to sections of rail until the next section has no current draw?  This could impact suggested solutions.
Q2. Approximately how mismatched are the train speeds when fed a normal voltage?  Eg magnitude of 2:1, 5:1, 10:1?
Q3: Approximately how matched do you want to make them?  Eg within 2:1, 1.5:1?

[RoGeorge]

Have you tried adding incandescent light bulbs in series with the motor?

Try with various miniature light bulbs you may find, of different nominal voltage and different nominal current.  This would mean a lot of trial and error, yet simple to experiment.

Incandescent light bulbs have variable resistance, very small when the filament is cold, and higher resistance when the filament is hot.  I've traced the I/V curve for a few incandescent light bulbs once, they look something like this:
https://www.eevblog.com/forum/projects/resistance-of-incandescent-light-bulbs/

Alternatively, you can design a small DC motor controller that reads the RPM of the motor, and define whatever speed vs voltage curve is needed.  Much harder and more expensive, but with predictable and reproducible results.

[Whales]

Some random thoughts about a circuit to put in each train to make the max speed (voltage) adjustable:



When powered backwards: current travels through the body diode of Q1, bypassing all speed control. 

When powered forwards: Q1 starts to turn on around 3-4V (depending on mosfet model), then starts to turn off (limit speed) once a certain voltage limit is hit.  The limit is adjustable via the variable resistor.

Problems:

• 3-4V starting voltage might be annoying (PWMing the train avoids this issue).
• Q1 needs to have a good DC safe operating area (SOA) to avoid blowing up.  Go for a big one (eg TO-220, TO-263, TO-252 or similar) and make sure the datasheet specifies a DC SOA line.
• Doesn't work how you would expect when fed PWM.  If the peak PWM voltage is high then all speeds will be reduced slightly, if the peak PWM voltage is low then no speed limiting will occur.
• Might oscillate?  3 transistors worth of gain might be too much. Capacitor on the gate of Q3 will probably help here, this circuit doesn't need to work beyond a hundred hertz anyway.
• The mosfets I suggest only survive 20V, higher than that and you will want different parts.  Might also be static sensitive?  Not sure.

EDIT: Q1 will need a large heatsink too.  Not sure if this is feasible on the trains (do they have cast metal bodies?)

[AndyBeez]

@Whales in on the right line there.

Are you looking to provide constant current or voltage to the locomotive motors? And how much inrush/start current should you allow for? A voltage or current limiter may kill the motor spin up from cold. My opinion, PWM is the only reliable option; whether that's a built-in module (that can run in both directions) or PWM on the rails.

So is there a practical/political reason why the 'loop' section cannot use PWM too? PWM means all locomotives can be controlled and no locomotives require modification. Hence locomotives can be ported to other layouts without the need for removing the mod and, [non DCC] locomotives can guest on the layout.

Alternatively, is there a mechanical solution whereby the gearing is changed to spin the wheels slower for the same motor speed? That would give the sets a higher torque.

[Whales]

This is a voltage limiting circuit.  It only kicks in (providing a constant voltage) when the supply voltage goes over a certain amount.  Otherwise it tries its best to do nothing.   Don't think of it as a traditional constant voltage or constant current power supply, it's not quite like those.

This is assuming that it even works.  It's just some ideas on a piece of paper, not tested.

And how much inrush/start current should you allow for?

I don't know.  I'd be lazy and just try it, see what happens.

If you want to go into depth then look at the safe operating area (SOA) graphs in mosfet datasheets.  Some will not have a DC line (because it's not the target market) but others will.  Also look at NPN transistors like the BD139, it might be a better suit here than a mosfet.  Measure the inrush current of your motors with an oscilloscope OR try and find someone's already published rule of thumb.  Too much effort

PWM means all locomotives can be controlled and no locomotives require modification.

Huh?  How does PWM prevent the locomotives needing any modification?  They'll still run at different speeds won't they?

So is there a practical/political reason why the 'loop' section cannot use PWM too?

It's a limitation of this simple circuit.  You can work around it by adding more complexity or some very large parts, but you probably want to keep this simple if it needs to fit inside the trains.

Also yes I voted for the non-PWM party last election cycle.  The PWM party kept going on about duty and the new government is now moaning about transition periods.  Their ideas are neither here nor there and I think they spend all of their time in parliament high.

Alternatively, is there a mechanical solution whereby the gearing is changed to spin the wheels slower for the same motor speed? That would give the sets a higher torque.

Purely mechanical: Constantly variable transmission (CVT).  Very complicated, probably quite expensive.

Mixture of mechanical and electrical: replace permanent magnet motor with one that has a field winding instead ("universal motor" and a few other names).  Add a circuit that varies the amount of current going through the field winding to vary the speed.

[AndyBeez]

Huh?  How does PWM prevent the locomotives needing any modification?  They'll still run at different speeds won't they?

You just set the dial on the speed controller for the line block > the layout side has the mod.

As for gearing, this is the realm of the model engineer - the dude with a mini lathe and mill - and these days a 3D printer.

Model train motors are difficult in other way; there are so many different types. Some were made yesterday, other's 50 year's ago. All exhibit nasty transient and back EMF issues.

+ I would like to see some photos of their setup.

[mikerj]

Wouldn't a simple voltage divider and emitter follower work ok here?  Two transistors give you bidirectional control, and the divider for forward and reverse could be different e.g. to make reverse slower.  This does give a deadband in the middle due to Vbe+schottky diode drop, but it's fairly small.  If you don't need bidirectional voltage division then leave out half the circuit.

The transistors will need heatsinking as they will be dissipating a reasonable amount of power.

[Thyristor42]

Q1. At the moment is the anti-crash system effective?  How does it work?  Turns off power to sections of rail until the next section has no current draw?  This could impact suggested solutions.
...

...
So is there a practical/political reason why the 'loop' section cannot use PWM too? PWM means all locomotives can be controlled and no locomotives require modification. Hence locomotives can be ported to other layouts without the need for removing the mod and, [non DCC] locomotives can guest on the layout.
...

The loop is using PCB's from the 80's (from a company called Trans Electronic Railways, made in Belgium ). The whole loop is cut into smaller sections which are each controlled by one of these PCB's.
I sketched a simplified schematic of the circuitry that controls the voltage on the track and does the detection of the train in the block (see attachment 610_Basic1.jpg).
The voltage on the track is a pure DC voltage, set by an NPN pass device. A PTC thermistor is added for short-circuit protection.
The detection (drawn in blue) is sensing the voltage drop over a diode. Based on this detection the PCB can tell the previous block to put the signal to red and stop any train entering the next block.
When a train is stopped in a block there is still a small voltage (~1V) applied to make the detection circuit work.

Changing the PCB's to work with PWM would make things more complicated. E.g. the detection circuit would also need to be changed. Also the PCB's are working fine so it's not really needed.
Also that wouldn't solve the problem because the PCB's can't recognise which train is currently in the block to set the appropriate voltage. Adding that would mean we end up with something that's pretty much equal to digital (DCC) control.

...
Q2. Approximately how mismatched are the train speeds when fed a normal voltage?  Eg magnitude of 2:1, 5:1, 10:1?
Q3: Approximately how matched do you want to make them?  Eg within 2:1, 1.5:1?

My estimation is a mismatch of maximum 2:1 between the fastest running train and a normal running train. The target would be to make them approximately equal.

Have you tried adding incandescent light bulbs in series with the motor?
...

No, I haven't. But it's a good idea. I think it works very similar to adding a PTC thermistor in series to limit the current, right?

Some random thoughts about a circuit to put in each train to make the max speed (voltage) adjustable:
...

Wouldn't a simple voltage divider and emitter follower work ok here?
...

I also had an idea in this direction: see attachment VoltageClamp.jpg. But I used a depletion FET to limit the deadband around 0V.
It also works like a voltage clamp: For voltages lower than the zener voltage, the FET would have a small voltage drop. For larger voltage the motor voltage would be clamped to something around the zener voltage.
Power dissipation can indeed be an issue. I need to do some further research to see if it's possible without adding an enormous heatsink in the locomotive.

...
Alternatively, you can design a small DC motor controller that reads the RPM of the motor, and define whatever speed vs voltage curve is needed.  Much harder and more expensive, but with predictable and reproducible results.

That idea crossed my mind, but it looks that this will get complicated quickly. Some tricky things that I can tell from the top of my head: powering the controller (rectifier + capacitors?), sensing PWM versus DC inputs, sensing forward/backward direction, driving the motor when PWM is applied.
But it is an option to consider.

...
Alternatively, is there a mechanical solution whereby the gearing is changed to spin the wheels slower for the same motor speed? That would give the sets a higher torque.

That's a good idea! I will discuss it with the other members of the association.

[MegaVolt]

Perhaps you will be saved by the current limiting of the power supply. Those. limit the power that it gives out and therefore the mechanical energy that the engine produces ..

[NiHaoMike]

Have you considered a simple variable power resistor?

[Ian.M]

You need to limit the max. speed on a per loco basis.   One option would be to do so by dynamically limiting the track voltage. It would need some method of identifying locos e.g. a barcode sticker on the underside of each loco providing data to a MCU or SBC managing the speed control system.  It could either use multiplying DACs between the existing controllers and the track driver boards to set the gain and thus max. speed, or it could take the existing controller speed signals as an input and do the scaling in software then output the speed control signals.   It would need an under-track mounted bar code reader (which could be as simple as a focussed retroreflective optical sensor similar to a bar code pen, but non-contact) at any point where tracks merge or it would be otherwise likely to loose track of the loco movement sequence.

[Marco]

Have you considered a simple variable power resistor?

Yeah, if the range of slowdown is limited it's hard to beat the simplicity of a couple ohm couple Watt rheostat.