What’s the problem?

To control the lighting of coaches, several solutions are possible. First, no control at all! The coach stays lit in daylight. Why not, if the light intensity is moderate. Then, control by DIP switch placed under the chassis. This is not always very discreet, and this requires passing wires between the coach’s floor and ceiling. In addition, it is obviously necessary to remove the coach from the track to actuate the switch.

After, you have the solution using a reed switch, operated by a magnet. Placed under the roof, it is easily actuated. The problem is that for most reeds, the closing is momentary: when you move the magnet away, the switch reopens. So some have imagined more or less complex electronic circuits, based on flip-flops or bistable relays, or even microcontroller, to memorize the state of the reed. Forgetting that at the slightest power cut, this memorization is lost! Not to mention the whole bulk.

And then there is THE solution: the latching reed switch. It works as follows: a tiny magnet, unable to close the contact by itself, is stuck on the reed switch. When approaching the control magnet, with its poles oriented in the same direction as the tiny one, the magnetic fluxes add up, and the contact closes. When the control magnet is moved away, the tiny magnet is able to hold the contact closed by itself.

Et voilà! We memorized the order! Yes, but how to reopen the contact? Well, nothing easier: only approach again the control magnet, but with its poles reversed. This time, the magnetic fluxes will be subtracted, and the tiny magnet will release the contact that will open again.

So, I’ve been using such latching reed switches for a long time, see for example Turning the red front lights off in the Picasso Mistral, the lamp control in the Dd4s Roco luggage coach, that of the lamps of a Jouef cereal hopper, or that of many coaches lighting.

This is definitely good, you say, but why want to make it yourself? Well, it started with the fact that the model I was using, the PMC-1424THX I bought at Conrad, was definitely sold out. So I looked for a size equivalent, to avoid redesigning the circuit board on which it was placed. For example, I found the Meder KSK-1E66, but it is available nowhere, and it is expensive, almost €10 each!


PMC-1424THX model


KSK-1E66 model

So I wondered whether I could make it myself, without believing it was possible because I thought that the setting of the magnet had to be particularly fine and precise. But I tried, and the first result was encouraging. So I’ll explain how I make latching reed switches very easily, for a price of just one euro each, which is not its least interest.

Necessary equipment

The main problem is to find magnets small enough, still for a question of size, knowing that the diameter of the reed used is about 2 mm, and sufficiently weak to not be able to operate the reed by themselves. I have found some, I have to say, by chance. These are small parallelepiped magnets measuring 1.6 × 1.6 × 12 mm, probably ferrite.

The other indispensable element is obviously the reed switch. I chose a small model (length 14 mm for a diameter of 2.2), quite sensitive and cheap. This is the Meder KSK-1A66, the sibling of the unobtainable KSK-1E66.

Here are the components involved.

Magnets and reed switches

Heat-shrink tubing is still required to hold the magnet on the reed, with a diameter of 3.5 mm before shrinking. Larger, the magnet would not hold in place for adjustment; smaller, it would be very difficult to slip the magnet on the reed’s body.

Making the switch

A whole 12 mm magnet is too strong: it would keep the switch permanently closed; it should be cut ideally into pieces of 3.5 to 4 mm (one third). To do this, I break it in a vice, using flat pliers.

Breaking a magnet

Here are the pieces obtained; the longest can still be broken in two. If a piece is more than 4 mm, it must be shortened. For this, a diamond disc can be used.

Broken magnet

Next, find and colour the north (e.g. blue) and south (e.g. red) poles of each piece, in relation to a known magnet, here a Jouef “tournebroche” motor magnet. The pole identification is important to determine the direction of actuation of the future latching switches.

Colouring the poles

Then comes the assembly of the different elements. I put the magnet on a wire of the reed. This wire is magnetic, which is handy to hold the magnet, while I slide a piece of heat-shrink tube, then the whole, to the middle of the reed body.

Assembling the elements

Adjusting the magnet’s position

To do the adjustment, I use a very simple LED-based circuit that will turn on when the switch is closed.

Test diagram

For assembling, I use a breadboard. Here is the layout. Attention to handling reed switches: the glass bulb is fragile, especially at the wire outlets.

At the beginning, the LED is lit because the magnet closes the reed. It must be moved slowly down until the LED lights off.

Adjusting the magnet position

In blue dashes, the breadboard internal electrical connections.

Next, approach the control magnet. If we direct its poles in the same direction as the tiny magnet (same colour on the same side), the fluxes add up and the contact must close again: the LED lights up. When moving the magnet away, the LED should stay on. Otherwise, it is necessary to go up the small magnet slightly towards the middle.

Obviously, we must also do the opposite test, that is to say, approach the control magnet with its polarity inverted, to turn off the LED. The setting is correct when switching on and off occurs at about the same distance from the switch. Do not try too much refinement, because it may be necessary to alter the setting once the switch implanted in its circuit.

If the magnet piece is really very small (less than 3 mm), the LED may not light when the magnet is in the middle of the reed. But this may not be lost. Approach your control magnet: the LED will light up. Move away: it may stay on. If so, we won! If not, discard the small magnet piece.

All we have to do now is to apply — moderately — hot air to shrink the tube, which will not prevent the magnet from moving, but still avoid an involuntary displacement.

With a little training, I made ten latching reed in half an hour, then the next ten in fifteen minutes!

Shrinking the tube

Example of placement on a REE UIC coach

The control circuit (22 × 14 mm) is easily placed in the toilets of the coach.

Click the image
to get a closer view
of the control circuit.

Placement on control circuit

Does it work well?

I have already mounted these “DIY” latching reed switches into about fifteen coaches. Actuation is certainly less accurate than with commercial devices. Some are too sensitive: it happens that two coaches are controlled at the same time if their switch, placed at the ends, are face to face. But this also happens with trade switches. Overall, the operation is very satisfactory.

However, I noticed incidentally a curious phenomenon, which occurs as well with trade latching reed switches: some locomotives, Roco in particular, operate the switches of a train parked on a parallel track when they pass along. It must be believed that their motor radiates a large magnetic field. It’s not very embarrassing, but it’s pretty surprising at first…

After investigation with twenty locos of different brands (Roco, Jouef, Electrotren, Mabar, REE, LS Models, Van Biervliet), I strongly suspect open motors, that is to say, those whose rotor can be seen, to be responsible for this malfunction. And, indeed, most Roco models are so equipped, while other brands mostly use closed motors (can motor) instead.

Useful addresses

Let’s talk about prices…

With the quoted suppliers, on the basis of 50 switches manufactured, with a 12 mm magnet for 3 switches, here is the bill:

Designation Unit price (€) Quantity Price (€)
KSK1-A66-1015 reed 0.42 50 21.00
CB-HFT2X-32-BOX tube 2.80 1 2.80
CDF 8041946 magnet 6.50 / 5 20 26.00
    Total 49.80

As promised, a unit costs less than €1.00 (excluding shipping). And I counted all the tube, while it takes only a small piece.

Could we still do better? We see that the price of magnets raises the cost price. We can find, on the Internet, more and more shops that offer an infinity of magnets of all kinds, some at prices much lower than at CDF. So I looked for magnets that could be suitable.

To make a choice, we must determine some characteristics, including the bulk (about 2 × 5 mm max.) and the pull force. What is the pull force of magnets used above? To get an idea, I tried to lift a 25 g ballast steel plate with an entire 12 mm magnet: we’re just getting there. I deduce that the proper magnet should have a force of about 8 g — one third.

In the neodymium magnet category, we can find some very tiny, €0.03 each, against €0.43 at CDF. But their force is 25 g! It’s way too much! So, for the moment, I haven’t found anything better than CDF’s magnets.

A few months later, continuing reflections

If we look at the characteristics of the Meder KSK1A66 reed switch as they appear in the TME online catalog, we see one named range, with AT as unit.

Meder KSK1A66

Explanation: this is a range of possible values of the magnetomotive force (MMF), which is expressed (or rather was expressed, see Wikipedia) in ampere-turns (At).

Expressed in a simpler way, the question is: if we want to operate the reed with a coil (a solenoid), we need to drive a current of how many amps into how many turns? For example, for the KSK1A66-1015, we can wind 10 turns and drive 1.5 A (that’s a lot!) or 100 turns with 150 mA (this is better), etc. The important thing is that the product gives 15 each time: I amps × n turns. It can therefore be considered that the value of the magnetomotive force gives an idea of ​​the actuating force required for the control magnet.

What am I getting at? Well, with the model chosen above, the operation was disrupted by some locomotives passing by the reed. By choosing a model requiring a greater magnetomotive force, this disadvantage may be avoided. However, it will probably require a somewhat stronger holding magnet — perhaps a half bar instead of a third — and to get the control magnet a little closer to operate the reed.

Given the price of these reed switches, it is worth trying. In a future order at TME, I will not fail to buy the model 2030 to test it.

New, less sensitive, reed switches

Well here we are, the new KSK1A66-2030 reed switches have arrived; it’s time to do some tests.

First observation: the holding magnet can be the same as before, i.e. one third of a 12 mm magnet. Simply, you have to position it more to the middle, which is not surprising, since the switch needs a larger MMF.

So I installed on a breadboard two latching reed switches, a very sensitive “old” and a “new” one, less sensitive, each with an LED and a series resistor. I start by doing some tests with different magnets, to find that indeed, we must get the magnet closer to operate the new switch, about 25 to 30 %.

Then I start the tests with a loco that disrupted the operation of the older switches. This is the ROCO ref. 62995, NMBS 6003. I will pass it along my breadboard, at shorter and shorter distances, to see whether it actuates the switches. I recall that the goal is to prevent the loco from changing the lighting or the tail lights of a train that would be on an adjacent track, which is simulated by my breadboard.

The actuation distances are in millimetres, measured from the axis of the switch to the loco’s. The values are average. The switch-off action means when the switches were previously closed (LED on). This can only happen when the loco presents its motor’s north pole to the switch. For the switch-on action, we have to turn the loco around so that it presents the south pole. The following photos show only switch-off actions.

Note: the most annoying action is the switch-off, because more visible at night than the switch-on at daylight.

Test with horizontal switches, perpendicular to the track

This is the most common case, with the control circuit placed under the roof of the coach. The N label on the loco indicates its motor’s north pole.

At this first distance, neither of the two switches is actuated.

Perpendicular test 1

At this second distance, the most sensitive (left) is activated, its LED is off.

Perpendicular test 2

At this third distance, both are actuated, both LEDs are off.

Perpendicular test 3


Action Older switch Newer switch
Light on 60 45
Light off 90 85

Test with vertical switches

This is the case when the control circuit is placed in the toilet of the coach.

Vertical test


Action Older switch Newer switch
Light on 50 30
Light off 70 50

We see that the actuation occurs more closely: this arrangement is better than the previous one.

Actuation test with switches parallel to the track

The configuration with the switches parallel to the track is not my most usual, but I used it for example on my M2 NMBS coaches.

Test parallel to the track

We can see that, while passing along, the loco turned off the most sensitive switch (the older one), here on the right.


Action Older switch Newer switch
Light on 35 20
Light off 50 40

It’s even better here.


As I said, the most annoying action is light-off, and this is unfortunately the one that occurs from further afield.

That being so, we can conclude that:

What would ultimately be the optimal arrangement of a reed switch for lighting control? Well it would be, we just saw, parallel to the axis of the coach, and in the centre of it. Indeed, on the one hand, this point is easier to locate to position the control magnet, and, on the other hand, it is further away from contiguous coaches’ switches. There is less chance of interference between them.

Before these tests, I almost always placed the reed switches at an end of the coach, simply because the anti-flashing capacitor is placed more easily in the toilet, without the need of long connection wires. I’ll have to review this arrangement in the future.

2-pole DIP switch

Reed switch

Neodymium magnet ø 1 × 1 mm,
pull force 25 g
€0.03 each,
at e-shop.magsy.fr