My sharp curve wye switch..

..and MSRS v.4.1 and up

Started 29Jun2021 – updated 27May2024 (Repair. Lego, MSRS v.5.0, «rudimentary between hooks bar»). This page is in groups My Crocodile locomotive pages and MODELS and is about a model railroad switch (or turnout), scale 1 (gauge 1: 1/32) that I have DIY scratch-built. The track is based on two Hübner produced Märklin/Marklin 59035 curved track with radius 1.02 m and angle 22.5°. I cut both in 3/4 segments, having an angle of 16.875°. Built by Øyvind Teig (Trondheim, Norway) 2021. I refer to several commercial sites here; however: standard disclaimer!

Aside: The photo is of me, standing alone in a switch being built at Løkken Verk (Norway), where the Thamshavnbanen (Thamshavn Line) in the summer of 2021 was working on new museum buildings and track (*). The photo reflects how I felt while building my switch. This 1-metre line, electrified since its start in 1908, is being kept as a heritage line by the Orkla industrimuesum (Orkla Industrial Museum) (here). (*) is here, translated here. (Photo: Anders)

Fig.1 – The finished, painted and aged wye switch

The switch works quite well. I have tested it with the SBB Ce 6/8 III (Märklin 555681, here), described in My Krokis notes.

Read this first

Too narrow for some loco-wagon combinations

Update 08Apr2023. This switch is too narrow for my Märklin 55681 Ce 6/8 III locomotive to pass with my short wagon from Fine Models. The buffers will slide apart when the wagon straightens up in the adjacent short S-curve when it’s well into the switch. However, with my E71 from Fine Models, it works fine, since the buffers then have a minimum of some 2 mm overlay. Press photo to see, it’s shot just before the «upper» buffers slip apart. The couplings are standard as in real life.

Observe that when this happens the wagon will derail if the train does not stop.

This is not magic. The frames move according to which wheels that govern them. The SBB Ce 6/8 III has power on three axles, but not on the blind axles at the ends, which move freely. The snout on the SBB locomotive moves much longer out of the rail than the E71. Simple as that.

Straight into short S-curve

Even if all these rolling material are defined to run on radius R = 1.02 m I hadn’t realised before now that the S-curve is too much for them. But it’s not the S-curve per se, it’s my two adjacent 22.5° degrees Märklin 59035 times 3/4 = 16.875° S tracks which in combination is too hard on the buffer angles. (Wrong, see below.) Looking into my list below, I see that I have discussed this somewhat in point 7. See the above photo, it doesn’t look bad at all.

Full 1.02 m diameter S-curve also fails

Update 09Apr2023. When I tested my two locos and the wagon from straight line to a full S-curve with two standard R = 1.02 m full 22.5° angle curved track, then the buffers still slide apart on the Ce 6/8 III. See photo above. It’s ok by a margin of about 2 mm with the E71. When it’s that «bad», then you could just as well cut the 59035 into 3/4 and make the switch as I did. Just know what kind of train you run there. So it’s not my short wye that is the problem.

The diameter of the buffers are 14.0 mm on the Ce 6/8 III crocodile, 12.9 mm on the wagon and and 11.5 mm on the E71. The loco that needs the widest buffer indeed has the widest buffer.

Full 1.55 m diameter S-curve is just about ok

With two Märklin 59076 R = 1.55 m diameter (also 22.5°) in the S-curve the Ce 6/8 III made it through the curve with a cry of distress! The margin might have been 2 mm, but it might also have been 1. I don’t know what to say.

Too narrow, solutions?

  1. Add a «buffer plank» between the buffers on the loco. Not nice
  2. Get wider buffer plates. Not prototypical
  3. Get couplings with a defined minimum length to keep the wagon at arm’s distance. Hmm. Best?
  4. A suggestion in Communities, point 2

Rudimentary between hooks bar

This works. It’s made of steel staples that I saved from a large box that came with a chair. I made eyes in each end that attach to the hook in the wagon and the long nosed 55681.

I filed it somewhat so that it sits firmly in the hooks. Almost as firmly that it’s difficult to get them in place. This is needed so that it’s possible to push the wagon through the switch. If not, the bar would be pressed up and out of the hooks.

The photo shows a pulling situation. A pushing situation is somewhat tighter.

I also showed this picture in Communities, point 2 (3May2023) and attached this question: «Does anybody know a manufacturer who would make a «prototypical» version like this? One with a screw that could be adjusted. I guess that in the 1/1 world one wouldn’t dream up such a thing. But in the 1/32 world I would be surprised if I were the first to have this need?»

Read this second

If you plan to make a switch of the same sort, do read this first. Or rather, watch the two movies first. Then, if you think it of any interest, read the whole blog note. You will experience that it is not a precise DIY step-after-step recipe. There’s a lot left for you.

Some background

  1. I built this switch to make MSRS v.4.1 layout possible. This is a 2.1 meter long railroad: My Shelf’s Railway Strip (MSRS). Update: MSRS v.5.0 – it’s now 2.7 meter long
  2. Märklin does not have a switch for the sharpest curved track for their scale 1 rail. The sharpest curve is radius = 1.02 m with their track 59035 (here). For my MSRS v.2.0 I use these to make an «open S» or S’ish curvature. I have cut the pair of them in 3/4 to make the S’ish even shorter. Before you dismantle these tracks, use a felt tip pen and draw a red line on the outer rail. They do have different radii and must not me interchanged!
  3. However, I did buy a left switch for radius = 1.394 m Märklin 59088 (here) with end piece 59092 (here) to some times use «on the floor». But also, to be be my model to learn from. However, the thin guard rails I did not copy. Even if I did copy the elevation above the other rails
  4. Observe that the “Märklin Hübner track” after 2008 is made from a solderable nickel/silver alloy. Also see 132:[Scale 1 original track and Hübner after 2008 – compatible]
  5. I also used a drawing from PECO as a model to learn from, see [3]. Especially the guard rails, which use standard rail profile
  6. Since I started off with two 50935 cut in 3/4 only, I would not have access to the rail profile that Märklin have used on 59088. I had one track profile to work with only
  7. Disclaimer: Your loco may not do R = 1.02 m, or it may not do an S with them. But then, this is 4/3 or 16.875° which may just work if 22.5° does not work. Update: see the previous chapter!
  8. Disclaimer: I have tried to make the switch look like, feel like (for the running stock, I guess) and sound like a real switch. However, in the myriad out there, you won’t find any equal. It’s a one of a kind.
  9. I have not made it «remote control ready» since I did not need it (but there is a hook)
  10. I have not made space for any lantern (the rotating indicator box with a square and an arrow). The position of the throw bar was enough indication for me. However, remaking the throw bar and adding an indicator should be possible..
  11. Disclaimer: I would need help with terminology! Please comment below or drop me a mail (here). I have leant most on [1] and railroad switch on Wikipedia
  12. It is a wye switch (Y turnout) because it will give most rail separation per unit of length, which makes the whole MSRS v.4.1 possible. At least if all I wanted to do is to splice it in and swap it with one curved track of the same dimension
  13. Finally, this is my first encounter into this kind of building..

Late work stages

Fig.2 shows the switch before I painted and aged it. You can see the wooden sleepers as well as all the fastening clamps that I filed from aluminium.

Final tuning required a more precise means to hold the switching rail down, in the same height as the outer rails. Therefore the (green) rod would not have the middle clamp on the final switch, which instead would have a thicker part of the (green) rod below both the moving rail and the sided outer rail.

I also ended up removing all of the nails that go down through the rail and replaced them with not much visible wire. The latter were soldered and also bent underneath (as are all nails), then the surplus solder was removed with a file. These are necessary, but then, nothing that should be visible.

Fig.2 – In work, more detail seen. Observe NOT final version!

MSRS v.4.1

Fig.12 – MSRS v.4.0 with DIY short and sharp switch/turnout (mast i Y rotated in MSRS v.4.1, after this photo was taken)

Observe 201:[My Shelf’s Railway Strip (MSRS) versions].

My Shelf’s Railway Strip (MSRS) may just be one of the world’s shortest scale 1 layouts, except perhaps for those nice display cases (like Märklin 59960). It is based on 201:[MSRS v.3.0]. Also see 201:[MSRS v.4.1] where the below picture is numbered as «Fig.44». But the quite long 62.6 cm Märklin 55568 Crocodile can actually just fit in on area  R  – enough for the wye switch to operate correctly. In addition to the Crocodile, an E71 from Fine Models has joined (201:[E71, Preußische EG 511 bis 537]).

I have decided to describe MSRS v.4.0 right here since note 201 is so crammed full. But jump past it if you are interested only in the switch:

  1. Ref. 201:[Catenary system]. The new four masts (orange circles 7-10) are also Sommerfeldt #640. I could have gone for double masts #642 for (10,1), (9,2) and (8,3), but MSRS v.3.0 kind of was too settled for me
  2. Mast (7) is tricky. I wanted to keep the grandchildren friendly play-field in the front (I keep some 1/32 Mercedes-Benz A-classes there from my collection, here), and placing a mast asymmetric to (4) would have spoilt that field. Position (7) is has ok passing distance of some 5 mm for the three-part twisting Crocodile, but I wouldn’t know for equally long single-body stiff locos. But an alternative position for (7) is already discussed, I would have considered that for single box locos
  3. I did find a straight line from (5) to (7) that just fit the pantograph tops, with some 2 mm margin
  4. Since mast (5) is on the top of my 8 mm elevated hill then masts (4) and (3) have their bases at some 2-3 mm different heights. Running the loco from area  Li  (Left inner) to  R  the lower wire from (7) to (5) snapped lightly on the pantograph before it was pressed on top of it. Luckily it was a matter of tenth of mm, and just bending lightly on the wires solved the problem
  5. The Märklin 55681 pantographs could have had rounded corners also for the top wear metal. There is a tiny edge there that I wouldn’t want to file. This would make running in a switch like this less of a possible pain. But of course, Märklin don’t sell such corner case switches; it must be for some reason
  6. The new catenary wires are (10) to (9) and (8) to (7) cut Sommerfeldt #645 (from 70 cm). This is the «new Märklin catenary»
  7. (9) to (8) is old Märklin 5635 (67 cm) and (7) to (5) is old Märklin 5636 (45 cm) – both cut. They are much nicer in my opinion
  8. With Sommerfeldt wires being diameter (⌀) 1.0 mm and Märklin ⌀ of 1.2 mm, when they meet the pantographs may have to see an edge. This was easy to remove with some bending of the wires
  9. I have tested with the Märklin SBB Ce 6/8 III, model 55618, which works like a charm in this sharp curve. I am awaiting an E71 model from Fine Models  which should also tackle this curve. However, the Spur1 similar model may not as easily tackle it. However, I just assume that since that model was tested on an R = 1.02 m track in the video by Pernsteiner, it might still run ok here since I only have a 3/4 S-turn. See 201:[E71, Preußische EG 511 bis 537].
  10. Having a model pass through R=1.02 does not always come easily. The Gotthard Railway has a minimum radius of 300 m (here). Divided by 32 is 9.37 m. So 1.02 m is out of proportion with a factor of nine! In other words, I guess it’s especially difficult if there are two locomotive frames. Then it’s some times not possible to make a scale model of all the details. The only example I would know of is the Märklin 55526 of the Ce 6/8 I proto-crocodile (here). It has the tool box in the middle supplied as fake piece that may be snapped onto the front when displayed as a vitrine model. When running on a curved layout, the box (not the front) must be removed. (Source: Die «Köfferli-Lok» by Peter Pernsteiner in the magazine Faszination Spur 1 #17, Jul2021. Not mentioned in [4]. Eisenbahn-Kurier, EK-Verlag).
  11. The length of the track is (in cm) 60 + 30 + curves (2*16.875° of R=1.02m → 60.0) + 60 – bumper left (9.5) – bumper right (1.4) = 2.10 – 10.9 = 199.1 cm. Times 32 this is 67.5 m and 63.7 m

MSRS v.4.2

I have now made a new background. It’s some striped fabric made of 65% cotton and 45% polyester. The whole shelf became more present like, not having the wall panel as a curtain any more. The stripes are horisontal. I made a test photo, see My Fine Models E71 scale 1 model.

I could have bought some realistic photos or posters and placed there, which might have been more realistic. But I guess, this layout I perceive as «realistic» as I want it to be. I like this solution, it gives me miles of better visual performance.

MSRS v.5.0

Fig.46 - My Shelf’s Railway Strip (MSRS v.5.0) extended length

Fig.19 – My Shelf’s Railway Strip (MSRS v.5.0) extended length. May2023

Some here is repeated from above, some is new. For My Shelf’s Railway Strip MSRS v.5.0 the main purpose is (1) to be able to have the two trains swap their left track positions, between  Li  (Left inner) to  Lo  (Left outer). Then the right side has to hold all the rolling stock: one long and one short loco, plus one short wagon. The second purpose was to (2) disconnect the overhead wires from being hooked up to the wall.

Observe Too narrow for some loco-wagon combinations (above).

Some pictures are coming. But here’s a start: My Lego 12077 crocodile notes.


The left parts have two Märklin/Hübner 59059 (60 cm) straight tracks and a cut-in-two of the same (→ two 30 cm lengths). Then my wye switch made from two curved Märklin #59035 (22.5°), cut in ¾ length (16.875°). Then again one 59059 (60 cm) for the main 2.1 m railway, plus one of the same inside the wardrobe. I also needed several extra Märklin 59095 rail joiners.

End bumpers

On the left side green and are two Märklin 5602 track bumpers. Now two masts have been placed in the middle of them both. A short line means using all centimeters! The rightmost bumper 5602 is a cut in pieces and «slid hung» on the wall inside the wardrobe (upside down key holes and screws, you know). Once removed and the below mentioned wing nut unscrewed from the mast, the track and its plank may be removed – after having disconnected the overhead wire at the (7) to (9) connection point.

Contact wire

I use four Märklin #5636 (450 mm) and one Märklin #5635 (670 mm) plus, now, for the segment in the wardrobe: a Sommerfeldt #645 (700 mm)  contact wires. The junction between the thicker obsoleted Märklin wire (1.2 mm) and the thinner Sommerfeldt (1 mm) works ok at low speeds. I do run with the pantographs raised. I used Sommerfeld #851 isolators at the end points at (1), (7-9) and (14), ten in all. I also added springs to add some pull to the wires at (1) and (14), four in all.


Now I can lift the 2.1 meter shelf off without having to unhook overhead wires from the side walls. The part that is inside the wardrobe lives by itself, it’s almost only to lift it out. But I did add a wing nut to hold and easily remove mast 8.

  • Red: (2) – (6), (8) and orange: 11-13. Ten Sommerfeldt #640 masts with boom. For MSRS v.4.1 I had to turn the mast (11) 90° around and bore new holes in it for the boom, so that the E71 and the Crocodile may run with doors open without them having to be destroyed
  • Green: (1), (14) single end masts without boom. Two Sommerfeldt #643
  • Green (7), (5) – a pair of the same masts #643 holding traversing wires across the rail


I had Märklin track only, from that Märklin rail. I did not have anything else available, as I did not want to dismantle a factory made switch, partly because my switch would have different dimensions and bending. Therefore all parts have been bade by using that profile type only.

Fastening the rail

Fig.4 – Filing mounting brackets from 3 mm aluminium. Left sliced in on a Märklin track

I saw that i could buy fasteners on the net. But since those were from plastic, I decided to make my own from 3 mm sheet aluminium. There are several types of brackets (cramps). Common for the green numbers (below) is that they don’t press hold of the track since the thickest nail does not go though the rail. (In retrospect maybe I should have made them from two part so that they could move relative to each other and certainly press and hold the track). Because of this I in addition bored 1.3 mm holes in the track, one for almost all track/sleeper crossing point. This cost me a lot of drill bits. I soldered rather stiff wire from electronic connectors into the whole and bent them on the underside. Let me just call them wire here. I then flattened the rail with a file to remove the excess solder. This made up for a very firm switch. However, the red numbers (below) hold the track better to the sleeper. Standard track starts 1 mm above the sleepers. Here are the bracket types. I have used no glue whatsoever.

  1. 62 standard brackets (Fig.4). Like on the Märklin track. One 1.3 mm hole and two 1.0 mm holes
  2. 4 brackets with one 1.3 hole for nails that actually go through the rail. Two where the closure rails hit the frog (which is insulated and therefore needs a gap). Two for the frog ends
  3. 8 brackets for the inner parts of the guard rails. 1 + 2 holes. These lift the guard rail up 1.8 mm to leave the guard rails at plus 0.8 mm elevation relative to the rail. Simply because my Märklin model switch had this elevation
  4. 8 brackets for the space between the guard rails and the rails. One 1.3 mm hole. These raise the guard rails 1.8 mm and the rails 1.0 mm, thus being asymmetrical
  5. 2 bottom pieces, from one sleeper to the next, for the end of the heel spread tracks. One 1.3 mm hole that holds a wire for the fastened closure rail. One hole for a screw from the bottom and up into the track, to form a pin that the moving rail would «rotate on». The screw I filed on the top to become a pin (pivot joint) for the small angular movement
  6. 4 sliders for the curved movable track, the point blades. The track moves on top of them. One 1.3 mm hole
  7. 2 sliders for the printed circuit board glass fibre combined connecting/switch rod. One 1.3 mm hole. They are fastened with a wire that comes from the track above, through the slider and through the rod, so they are firmly clamped together. This turned the switch from being gravity based to friction based (more below). They slide below the outer track to keep the moving track at equal height
  8. 90 SUM of hand made filed aluminium pieces

The drawing

Fig.5 – Metrics of the switch (cm)

It was hard to decide on what R=1.02 actually meant for me. The 59035 rails with radius 1.02 m and angle 22.5° (cut into 3/4 segments with angle 16.875°), didn’t really show up to the theoretically correct when I connected the track on a board to measure. I assume that there’s some slack to the accuracy of the bending at the factory. This is discussed and concluded on in this document (where page 5 is seen in Fig.5 above):

../224_fig05_scale_1_wye_switch_by_oyvind_teig_trondheim_norway.pdf – 33 MB

Since I didn’t have an A3 printer, here’s page 5 of the above, split up into two A4 pages. Print out, cut them both and glue them together. I built my switch on top of this drawing, on a veneer board. Verify that the 10.73 cm sleeper is indeed 10.73 cm. If not, print out with appropriate scaling:

../224_fig06_scale_1_wye_switch_by_oyvind_teig_trondheim_norway.pdf – 59 KB

The frog

Fig.7 – Rails and initial frog (1-5) and towards (6) final frog (7)

The frog was difficult to make. It took me several days.

For a DC model railway the rails carry the power supply across them. Therefore the frog needs to be disconnected from everything. A power short situation may arise if a wheel is so wide that it reaches the other side’s point on the heel. The heel is the «arrow» and is part of the frog. In other words, the heel must not be too short! Mine is fine with a good margin. I did have to make the heel two times, but that was for lack of accuracy. Observe that the «slack» of the wheel may cause it to reach over if the axle is pressed towards the arrow.

I repeat: Mind the gaps for DC isolation!

I made the flangeway filler (bottom floor) in the frog from two pieces of 1 mm aluminium. I think the idea is that the wheel’s flange should exactly run on the top of the floor inside the frog. This way it will not fall into the frog and then hit the edge again. It is 2.0 mm deep. (See the The movies chapter below and read this depth discussed in the text there.)

I bent the wing rails from standard rail track, as usual. Before the bending I did some small cuts at the bending points.

I thought that the red numbered sizes were correct – until I lay down on the floor and did a test. The last third of the axles of the 55681’s first frame into the switch actually climbed the end of the wing rail’s flare part. Picture numbered 4 above. I still think this is rather strange, since to me it looks like the forces would push that axle to the other side. But I guess the hinge inside the loco also applies some force on the frame. Anyway, it was fixed when I made the flare longer, so that the wing will gently push the rail into the frog. This works like a dream, even if it looks somewhat unusual.

I made the frog (heart) into a single unit that may be dismantled at any time. There are two holes for fastening the unit to the sleepers. One in the arrow for a screw, up into it and one in slimmest part for a nail down through the sleeper. Nails are always just bent and hammered underneath. I use a piece of aluminium in the vice/vise to reach the head of the nail when I do this. This makes it hard to dismantle anything without destroying the wooden sleeper. But if it’s not done too may times it’s ok. The rail parts fastened onto the frog unit itself would also have holes bored through them. There would be two in each wing, plus one for the longer end of them sitting on the next sleeper. I used visible nails there and did not solder and filed flat, since I thought it looked rather nice. The heel / arrow also has two holes in the base, fastened onto the frog’s base. However, these are hidden below the flangeway filler floor.

I observe that the heel (arrow) may be made from two pieces, one short and one long [1]. I made it from two equal pieces of rail. I made a cross hole in them and added a wire and soldered so that the tip should never open.

A final test of the frog (or the whole switch for that matter) is to do a sound noise test. I ran my 55681 without sound at its max speed and could not hear anything more than the sound of rail (see movie, below). On my 2.1 m track I have set the max speed to rather slow 201:[Setting Vmax] so I haven’t tested the switch at higher speeds. At higher speeds of course there would be some turnout noise, since it is different from single joints. But I would think there would be no extra noise since I have been as accurate as I was able to. Doing some details twice certainly helped!

NEM standard

NEM 110 [2] gives the dimensions of switches for scale 1 (gauge 1) – plus all other gauges. My problem is that no matter how much I measured the Märklin switch 59088 with end piece 59092 (mentioned above) against the standard, I couldn’t make much sense of it!

I therefore decided to use the Märklin switch 59088 with end piece 59092 as my model, not NEM 110.

Materials list

Fig.8 – Some of the material used. Wood, template, screw and nails

  1. Wood for sleepers. My wood is from Siberia! The annual rings are extremely tight at parts of it, with a density of 3 rings per mm. I thought it ideal for the sleepers. Aside: The logs have been driving in some river and not been picked up by the Russians. However, it floated with the currents from the Barents sea, the Norwegian sea and perhaps Greenland sea before it landed on a shore in the north of Iceland. A friend gifted me a log, to put in the suit case when we left. Thanks log-saver and lumber mill owner Björn at Kópasker!
  2. I made a template (die?) from aluminium. I used this at all times. By this I get the same width all over the piece. You’d have to file the opening a little more than 2.5 mm. Of course, this is a tool, not «material»
  3. Screw and nail
    1: 1.3 * 10   mm  Fleischmann 6410 wooden screws. Here
    2: 0.8 * 11.5 mm 
    Roco 10000 steel nails. Here
    3: 0.9 * 10   mm 
    steel nails. Don’t remember where I got those
    4: 0.5 *  8   mm  Märklin 8999 steel nails. Here
    5: 1.0 * 10   mm 
    copper nails, any hardware store I guess (like Jernia here in Norway)

That ends the numbering on the picture.

  1. Track from two Märklin 59035 with radius 1.02 m and angle 22.5°. I used two 3/4 segments with angle 16.875°. The leftover 1/4 segments I used for the guard rails and frog
  2. 2.1 mm thick fibreglass (glass epoxy) for the combined connecting/switch rod
  3. 3 mm sheet aluminium for the mounting brackets
  4. 1.6 mm sheet brass for details of the throw bar
  5. 0.62 mm ϕ diameter copper wire (see below) for a detail of the throw bar
  6. A brass connector from a discarded 25A socket used for our old stove, the ground terminal, for the throw bar’s rounded head. Part 2 here. I filled it with a 5 mm steel screw. See Fig.8 photo (1)
  7. Drawn out wires from some wire-board connector headers (like this from Molex). I used them because they are almost like nails, for the holes in the rail, through the rail and sleepers, soldered and filed flat on the top. Bent underneath
  8. 4 quite stiff wires to connect the rails electrically. I opened a 2.0 mm2 green-yellow ground cable and picked out some 0.62 mm ϕ diameter copper wire. Soldered into holes in the rail, filed flat on the top and put into small grooves underneath the sleepers.
    The Märklin track has 4 visible cables between the sleepers. They are made from spring like steel and get their electrical contact from that spring action, underneath the rail
  9. 0.5 mm sheet steel from some spring metal from some box. Was hard to cut in that small dimension, but I used a standard scissors. Used to make four small «fingers» or clamps that hold the movable part of the rail by the heel spread. Each clamp pair mounted «inside» the rail. Quite critical component, as it should hold the movable rail, but not add any force to it. Must be force-wise dead in the angles used. More below
  10. Paint. First I tarred the sleepers and then removed it with a soft cloth before it had dried too much. This is the stuff that we used to treat wooden skies with, in the previous millennium. It did smell creosote, as was the intention. Some of the pictures are from that stage. Then I painted with the sleepers and the metal details with one layer of Humbrol enamel paint #173 track colour matt. I then finally used the Humbrol to weather the rails at the end. Then I painted some of if away lightly with a brush with a drop of white spirit to finish the weathering.
    Aside: A very old man (ten years older than me?) told me (when I brought him on «the tour of my switch») that his father had «worked with the railroads» and that he grew up by the line. He could tell that when the cars arrived with new , wooden sleepers, full of creosote, the men had to be especially careful when working in the hot sun. The fumes from the creosote could cause serious and painful rashes in the face and arms. However, they had used protective gloves, so young Torbjørn could not remember problems with the workmen’s hands. Thanks for telling me!

The point blades

Fig.10 – Filing and attaching the point blades. Early stage

Observe that fig.10 shows an early stage. Not all the sleepers are mounted. The combined connecting/switch rod’s fastener ended up becoming different (already described), and there are now more sliders below both the point blades and the outer rail.

Of course this was the most difficult exercise, as expected. Plus, perhaps, the one with the fewest words needed. The main point to convey is that it is possible to make this with a standard rail profile. Real switches would have a different profile. Here, too, make sure not to interchange the inner and the outer track, they do have different radii. I felt that using diamond files was not as ok as using standard steel files. Keep the inner bases as intact for as much of the point blades as possible, I could have kept mine longer. But it needs to gradually go away at the side that will meet the outer rail.. Don’t press the rails in the vise, make some soft stuff around them. I made something from wood. The metal is extremely sturdy, I was surprised, but having vise marks or deformations from it would not be nice.

I don’t know how many times I had to unmount and mount the point blades after filings. It was many times.

My point blades go all the way to the start of the switch. This is unusual, but necessary since I wanted to make the switch with minimum length and maximum angle in the shortest length possible. This is also discussed in chapter Joining the rails (below).

The switch

Fig.8 – Building the throw switch, including (12) repair

For the throw bar’s rounded head I started with the brass connector (1) from the discarded 25A contact, as already mentioned. I left it too big at 10.5 mm ⌀ and 5.9 mm thickness. The Märklin piece is 8.0 mm ⌀ and 3.9 mm thick. The reason is that I wanted as much weight as possible, because I didn’t know whether I would be able to make the whole mechanism gravity based, as in real life. It turned out, I ended up with making it friction based, as is my Märklin model.

Observe that all the red numbered photos are not the final versions. I had to do fine adjustments with a file, and then was so happy with the final result that I forgot to photograph. Plus change from a gravity based to a friction based solution, as seen in the green numbered photos.

I designed the non-linear or hysteresis based throw mechanism a little different from the model’s. Probably there is no other switch in the world like it.. The small U (2) is so wide that the throw bar (1) moves freely about one mm, enough to get the push only after some movement, when it has passed 12 o’clock. A small axle (3) in it makes the connection to the combined connecting/switch rod’s connector (4) in brass flexible. I had to experiment with attaching (4) above or below the connecting rod. The gravity based solution in photo 5 has it connected below, but the final, friction based in photo 8 has it above. But from 5 to 8 I also have some other changes. Stay tuned.

Observe how far away from the rail the throw bar is mounted (also at fig.5). The axle is 50 mm from the top of the rail’s outer edge, in the curve. I had it 3 mm shorter with the result that the loco 55681 actually touched the head when it was moving to the left, with the head at 10 o’clock.

Photo 5 is gravity based. It almost worked perfectly. The first thing I discovered was that the steel fingers holding the movable rail from the closure rails, making up the hinge, should hold with some precision, but not so much that they would have any spring action on the movable rail. Basically, the fingers need to be short enough. Plus, the screw from beneath, which becomes an axle (pivot joint) into the movable rail, had to move in a somewhat oval hole. Not 1.3 mm diameter but some more on each side. Not easy to make, but then I had to deal with tenth(s) of a millimeter only.

The idea was that low friction of the part 9 (which I had to do some precision filing of its depth) should make the weight of the rounded head hold the movable rail in position. It actually did. But I had to do some more filing on those rails to have them align in position accurately. A couple of paper’s thickness gap was enough to make the second or third axle decide on entering the gap. It was pulled out of it by the advancing loco, but it wasn’t nice. But I managed, almost.

What happened was that when one wheel pushed the hinge part of the movable rail down, the rail at the start of the switch was raised a little, enough to make it some times not work. And it didn’t look nice. I then saw a basic flaw, the piece (9) in photo 5 was nice with almost zero friction, but it didn’t ensure the movable rail’s top to never be above the outer rail’s height.

I therefore had to think more correctly. Photo 8 shows the final, friction based solution. I already had the sliders that carried the weight of the loco. But now I added a slider on top of the connecting rod, sliding against the bottom of the outer rail. Now those two bottoms are at aligned to each other. Plus, the nail through the moving rail and through the connecting rod was made not like a gap with tiny up/down movement possible (to keep friction low), but as a piece that holds firmly on the movable rail. Now the movable rail is held nicely in all positions. But now the whole mechanism became friction based. The force I need feels quite correct, actually. And the movable track stays exactly where they should and don’t show any ticking movement while the loco passes above. Time after time. After time. I am satisfied and don’t have to explain away anything for any victim that says yes to a guided tour.

Joining the rails

I am not at all impressed with one particular property of the rail joiners of the Märklin/Hübner track. Namely how they tackle being connected again and again. It doesn’t take many rounds before they fall off. They are meant to be hanging onto grooves in the rail and in the plastic sleepers, and there is a lip on the rail joiners as well – plus a spring. They should have for trying. But I did feel I needed a bag of Märklin rail joiners 59095 (here).

I used them, of course, see Fig.8, photo 8. My problem was that my Y-switch is so short that I decided to end the movable rail where the turnout ends. Contrary to my models. So I had to make sure there was no collision between the rail joiners and the moving tip of the rails. Just having a «foreign» track come there with one of its own joiners I didn’t appreciate. So I decided on mounting them with a nail through the joiners and the base of the outer tracks. The end is «in the air» so, no problem, there is more than enough space for the joiner and the bent nail. The downside is that I have to remove a joiner of the connected two rails, difficult if I don’t want to inflict any permanent damage there. However, on the single track side I standardly only nailed it for a single joiner, on the left side side, going out of the switch.

The movies

46 MB movie. The Märklin 55681 runs without sound across the switch. The movie is with sound, though.

Shot with iPhone XR as 1920 * 1080 into .mov (149.5 MB) and converted with HandBrake 1.3.3 into fast720p30 modified web optimized 1280 * 720

Compare with [4] at 13:10, where the wheels actually fall into the flangeway instead of running on them.

12 MB movie. I switch the turnout by hand several times left and right.
Same procedure as above (.mov was 47.9 MB)

Almost final views

Fig.14 – Almost final views

I guess, these are about my almost final say.
Aside: I did use polygonal lasso tool in Photoshop Elements to cut out the hardware only, and then merged the two png files.

The tools

Fig.10 – I guess, most of my tools to build the switch

In addition to the obvious, I guess I have had most help from the micrometer gauge (from my grandfather or father). Plus the new, rather light drill driver (10.8V, 2.0 Ah, 1 kg, Makita DF332D. The only problem with it is the LED light which makes a shadow on the target). But starting any hole I always did by hand with a bore holder. When I soldered I used 480°C and lead-free solder, plus a fan to dilute the air. I use rubbing alcohol when I bore in aluminium, and oil when I bore in the very hard nickel/silver rail. The bores are Proxxon 0.8, 1.0 and 1.2 mm diameter bores, like 28852 (here). I did break quite a few, though – even that type!

In the figure I forgot to include a flat-headed steel pin punch that I used when I hammered and bent flat (looks like one of these). (I later found one with a 3 mm flat head with a hex shaft – it would have sat much better in the vise. I had to file a little flat area on the one I had). I also, in the picture, forgot the aluminium piece I fastened in the vise to «reach up» by the sides of the rail profiles. They both are visible in in the greyed parts if you know what to look for (bottom, right).

Finally, for the sleepers, my Proxxon 27070 FET table saw. (As always, standard disclaimer.)

No remote control

As mentioned previously I have not made any lantern. I also have made no space for a digital turnout motor, as I simply don’t need it. However, the connecting/switch rod I did extend to the other side of the switch, and also added a hook there. I guess nothing would be impossible in the future.


Fig. 19 – Repair

See Fig. 8, where point (12) shows my fork tips to pin soldering that turned out, after three years of low usage, was not solid enough. I ended up with a new and hopefully better design. I think the root problem is that the handle (throw bar) has no easily visible end stop points. It stops naturally when the rail tails are in position, just like on any manual switch mechanism. So when eager, small hands have been pushing the handle, it just might have been pressed too eagerly, causing too large a pull on my soldering.

  1. The old pin was short (12). I made a new, longer pin that would allow more solder
  2. I found some copper from my box of everything, from a battery compartment
  3. I cut it and soldered a U onto one end of the pin. The other end will be soldered like the old style, but with a longer pin there is more space for the solder. My 421°C Curie temperature based (Metcal) soldering iron (full effect and that temperature in a split second only when lifted from the stand. I just love it, the tips literally last forever) was perfect. It delivered enough power with that thin tip
  4. Here it is finished. I bored holes on each side, to enable a more solid soldering
  5. Soldering this onto the fork needed my old soldering iron with always its maximum at 480°C regulated, more power (but the tips are regularly oxidised away, even at much lower temperatures. The main problem is that they are always powered). After the soldering the throw bar’s mechanism looks just ugly
  6. But after some filing it looks nicer. I stain painted it. The paint is wet on this photo
  7. The underside of the mechanism

The switch looks just as old and worn as it used to. Stay tuned (with years perspective, I hope).


  1. DIY scale 1 wye switch – Radius 1.02 m 3/4 length turnout (22Jul2021). Liked by persons from USA, Denmark, Spain and Australia!
  2. S-curves and wagons – When the buffers pass each other the wrong way around and cause derailment – started by me 09Apr2023


Reading it all over, I see that my description is difficult, but rather accurate.

I should have made drawings of all the different aluminium connector pieces. I guess, if only one of you out there would need them, I would draw them [[TODO]].

It works quite ok. But please make sure that your locos can tackle this combination making an S-curve of radii R=1.02 m, in case you decide on making any.

I’d also be happy to know about any similar encounter, and would be glad to list them here.

Now, do have patience!



  • I file (verb) with a file. With the file I will then remove stuff.

Wiki-refs: Minimum railway curve radiusRailroad switch.

Numbered refs

  1. Turnouts og DCCWiki, see .
  2. NEM 110, Normen Europäischer Modellbahnen: Gleise und Weichen Spurführungs-Maße (Standards of European model railways: track and switch dimensions), see All NEM standards at
  3.  PECO  STREAMLINE. Turnouts and crossings plans. See and the one I printed out and also used as a model: SM-32 – CODE 200 and SL-E697 at
  4. Vorserienmodell-Unboxing „Köfferli-Lok“ Spur 1 Märklin 55526 SBB 14201 Ce 6/8 I Ur-Krokodil 1:32 by Peter Pernsteiner (on YouTube)

Leave a Reply

Dette nettstedet bruker Akismet for å redusere spam. Lær om hvordan dine kommentar-data prosesseres.