Making precise slow speed station stops (automation)

[David]

For the eventual layout a major component will be automation - I want a lot of other traffic moving on its own, making station stops and generally behaving in a realistic manner.

To make a station stop I need to combine 3 elements:

* Trains of varying length need to be able to come to a complete stop at a set point, with a maximum various of 5mm. If Densha de Go has taught me anything it's that you lose a lot of points for overrunning a station.

* Trains should not come to a sudden stop, they should slow down gradually and smoothly. Densha de Go has also shown that you can't slow down too fast, or you'll waste a lot of time coasting along the platform. You also lose points if you have to speed back up again because you undershot the end.

* It has to be economical. Platforms will already cost a lot due to turnouts, the platform itself, and TomyTec figures. Another $150 to install 8 detection zones is not good, assuming they even perform well packed in that close.

Obviously detection zones, based on a load (the train) being placed on an isolated section of track, are the basic building block for automation. But I think something more clever is needed for the stations. Optical detection comes to mind, but neither of the two options seem good: Embedded in the track it can detect a passing train, but that brings us right back to cost because of the quantity. In theory IR and similar optical systems can do range finding, but I think this has been shown to be impractical for precise control at this scale.

I got one idea watching video of the new TomyTec bus system and how it was able to stop precisely inside the painted lines of the bus stop. That system uses magnets and reed switches. What if we placed reed switches under the track, and a magnet in the cab cars? (the motor car might also accidentially project a magnetic field). If the microcontroller component could be done cheaply enough it could be used to provide a large number of detection spots to plot the exact position of a train (we can assume the train is traveling in one direction only) on the platform track.

[inobu]

Automation = computerization.  
Economical DCC computerization = JMRI.  
Precision speed control =speed table mapping.
Station to station automation = scripting.

You can script your trains route and manage its starts and stops by time based scripting. Scripting allows you to write DCC commands that will execute in time intervals which can be synchronized with the stations train schedules.

Programming the deceleration trim will allow you to pin point the precise time to enter the stop command resulting it a consistent stopping point along the platform.  

[Martijn Meerts]

If you only want single direction travel, and want all trains to stop at the exact same spot, all you need is 2 occupancy detectors per block. Depending on which digital system you use, 2 occupancy detectors aren't all that expensive.

The problem I've had with read switches before, is that they tended to get stuck, making them unusable. Also, if you hide them under the track, and hide the magnet inside a car, you'll probably need fairly strong magnets.


There's various options for computer control. JMRI is one, but there's also RocRail (free) and iTrain (low price). Obviously, there's bigger ones, but they're overly expensive ;)

[CaptOblivious]

I've been thinking some about this problem too. A couple things I've considered:

An array of IR-LED/detector pairs placed under the station platform, facing the track. If you used several, you could keep on the actual speed and position of the train. Might be visible, though the LED and detector would only need a small hole drilled in the side of the platform. If they are positioned so that they see the trucks only, you can get a fairly accurate and fine-grained measure of speed and deceleration.

Option two, for increased hidability: Magnets in trucks, hall-effect sensors inside platform, facing trucks. More reliable (and more expensive) than reed-switches, hall-effect sensors are pretty cool, and provide an analog output suitable for a very small amount of range-finding (maybe a few mm in either direction, which would be all that you need for precise placement at the platform end).

After that, it's up to you how you process those signals. An Arduino would be an ideal candidate, I think, for processing them and passing them on to JMRI or the like. The smallest Arduinos can easily be hidden inside the platform.

[The_Ghan]

If you only want single direction travel, and want all trains to stop at the exact same spot, all you need is 2 occupancy detectors per block. Depending on which digital system you use, 2 occupancy detectors aren't all that expensive.

Why two?  I'm setting up my layout for occupancy detection (Digitrax BDL-168s) but the wiring diagram supplied shows one detector per block.  Further, I can't fathom how two detectors per block might operate differently.

I'm also interested in this topic and plan to automate trains stopping at stations.  I was hoping that I can just do this with JMRI or TrainController.  No?

Cheers

The_Ghan

[Martijn Meerts]

If you only want single direction travel, and want all trains to stop at the exact same spot, all you need is 2 occupancy detectors per block. Depending on which digital system you use, 2 occupancy detectors aren't all that expensive.

Why two?  I'm setting up my layout for occupancy detection (Digitrax BDL-168s) but the wiring diagram supplied shows one detector per block.  Further, I can't fathom how two detectors per block might operate differently.

I'm also interested in this topic and plan to automate trains stopping at stations.  I was hoping that I can just do this with JMRI or TrainController.  No?

Cheers

The_Ghan

Just 1 would work as well, but gives you no precise control of where the trains stop. Also, trains will get different running characteristics over time, meaning that if they stop where they should initially, after several months of running they might start driving faster or slower, which in turn means they'll stop at a different spot. Some software allows you to get around the issue though, by adding 1 or more speed traps, where the train speed is constantly measured and updates in the software.

If you use 2 detectors, the first one would be the brake section where the train starts to slow down to minimum speed. The 2nd one would be the stop section, where the train stops. Obviously, with a stop section, you'll be a lot more certain where the train stops.


Not sure about JMRI with regards to complete automate a layout, but I would expect it's possible. With TrainController it's definitely possible.

[David]

The software side isn't a problem, JMRI provides an open API for using a Digitrax PR3 or other DCC-Computer interface, I can provide the rest (that's part of the fun). My main concern is what Martijn Meerts talks about, the fact that ideal stopping distance based on a series of blindly issued speed steps, and actual reality (temperate, motor wear, wheel slip, track pickup, missed packets) are not the same. It reminds me of some high school robotics - making the assumption that if you timed how long it took for the robot arm to make a 1 degree turn, you could accurately move it without any sensor feedback. There is a reason specialized motors (servos) exist.

[The_Ghan]

Just 1 would work as well, but gives you no precise control of where the trains stop. Also, trains will get different running characteristics over time, meaning that if they stop where they should initially, after several months of running they might start driving faster or slower, which in turn means they'll stop at a different spot. Some software allows you to get around the issue though, by adding 1 or more speed traps, where the train speed is constantly measured and updates in the software.

If you use 2 detectors, the first one would be the brake section where the train starts to slow down to minimum speed. The 2nd one would be the stop section, where the train stops. Obviously, with a stop section, you'll be a lot more certain where the train stops.

Not sure about JMRI with regards to complete automate a layout, but I would expect it's possible. With TrainController it's definitely possible.

Hmmm ... I'm thinking of going with TrainController, but it's expensive.  I've already downloaded the demo and I like the way it works.  With regard to speed traps I would imagine (and I'm not this far into it yet) that a speed trap can be set across multiple blocks using one detector per block ... I shall explore, but I take your point.

Cheers

The_Ghan

[Dani]

I'm doing some tests with Rocrail (it's all for free!! and also has Androc to take the control from your Android phone!!!!):

* I have one detection zone at the begining of the station (one for all tracks) (ZONE A)
* I also have one detection zone in each track of the station, all at the same distance of the end of the track (ZONE B)

With RocRail you can setup all trains entering ZONE A to reduce the speed at V_mid or V_min softly following the inertia of the loco. (V_min and V_mid is configured particulary in each loco parameters)

When the traing enters in ZONE B, Rocrail can order to stop this loco, but following the special configuration of each loco. Each loco can be configured with a delay time before stopping, so you can compensate the different inertias and speeds to make all them to stop in the same distance.

I'm testing this for my layout right now, but theoreticaly it can be done without problems.

Hope this helps.....

[Martijn Meerts]

The software side isn't a problem, JMRI provides an open API for using a Digitrax PR3 or other DCC-Computer interface, I can provide the rest (that's part of the fun). My main concern is what Martijn Meerts talks about, the fact that ideal stopping distance based on a series of blindly issued speed steps, and actual reality (temperate, motor wear, wheel slip, track pickup, missed packets) are not the same. It reminds me of some high school robotics - making the assumption that if you timed how long it took for the robot arm to make a 1 degree turn, you could accurately move it without any sensor feedback. There is a reason specialized motors (servos) exist.

I believe it's Railware that first came with the whole 1 detector per block is enough, because their software was intelligent enough to stop trains exactly where you wanted them to stop. They forgot to mention (in their marketing) that this depends on the actual speed of the train, and the speed will definitely vary. Add 2 cars and your speed will already be different, albeit very slightly.

Just 1 would work as well, but gives you no precise control of where the trains stop. Also, trains will get different running characteristics over time, meaning that if they stop where they should initially, after several months of running they might start driving faster or slower, which in turn means they'll stop at a different spot. Some software allows you to get around the issue though, by adding 1 or more speed traps, where the train speed is constantly measured and updates in the software.

If you use 2 detectors, the first one would be the brake section where the train starts to slow down to minimum speed. The 2nd one would be the stop section, where the train stops. Obviously, with a stop section, you'll be a lot more certain where the train stops.

Not sure about JMRI with regards to complete automate a layout, but I would expect it's possible. With TrainController it's definitely possible.

Hmmm ... I'm thinking of going with TrainController, but it's expensive.  I've already downloaded the demo and I like the way it works.  With regard to speed traps I would imagine (and I'm not this far into it yet) that a speed trap can be set across multiple blocks using one detector per block ... I shall explore, but I take your point.

Cheers

The_Ghan

Traincontroller is a good program, but as you said expensive. With the latest release and the bronze, silver and gold versions it seems to be better at first glance, but when you check the features for each version, you'll quickly come to the conclusion that in many cases the gold version is the only thing that works well ;) You could have a look at RocRail, which is similar to Traincontroller, but it's free. It's not (yet) as advanced as Traincontroller, and it's definitely not as easy to use, but might be worth having a look.

I'm doing some tests with Rocrail (it's all for free!! and also has Androc to take the control from your Android phone!!!!):

* I have one detection zone at the begining of the station (one for all tracks) (ZONE A)
* I also have one detection zone in each track of the station, all at the same distance of the end of the track (ZONE B)

With RocRail you can setup all trains entering ZONE A to reduce the speed at V_mid or V_min softly following the inertia of the loco. (V_min and V_mid is configured particulary in each loco parameters)

When the traing enters in ZONE B, Rocrail can order to stop this loco, but following the special configuration of each loco. Each loco can be configured with a delay time before stopping, so you can compensate the different inertias and speeds to make all them to stop in the same distance.

I'm testing this for my layout right now, but theoreticaly it can be done without problems.

Hope this helps.....

My hidden yard is set up like this. 1 common detector for each of the hidden yard tracks, and than 1 detector for where the train has to stop. The disadvantage of this system is that it only works in 1 direction (unless you add the same system for trains traveling the opposite direction, but that defeats the purpose.



A while ago I did some posts about setting up blocks for computer control, it's al fairly rough and basic, but it might help some people:

- Part 1: http://www.jnsforum.com/index.php/topic,370.0.html
- Part 2: http://www.jnsforum.com/index.php/topic,372.0.html
- Part 3: http://www.jnsforum.com/index.php/topic,403.0.html