Start here
You do not need a plan for the finished railway. You need enough decided that the money you spend this month is still useful next year. These are the six choices that lock things in, in the order they actually bite.
Make these in order. Each one narrows the next, and only the first two are expensive to change later.
Everything else is downstream of this. HO (1:87) in North America and Europe, or OO (1:76) in Britain, is the default for a reason: the widest choice of models, the most forgiving size for adult fingers, and enough mass that trains run smoothly. Choose N (1:160 / 1:148) if space is tight and your eyesight is good. Choose O or G if you want to see rivets, and have a garage or a garden.
Measure the actual space, then subtract the aisle. A layout you cannot walk round is a layout you cannot reach into, and everything on a railway eventually needs reaching: a derailed wagon, a dead frog, a tree that fell over.
A shelf 30 cm deep along one wall beats an island in the middle of the room in almost every domestic situation.
Write one sentence: “A single-track branch in the Welsh borders, 1962, the last year of steam.” or “An industrial switching district in New Jersey, 1978, Conrail blue everywhere.” That sentence decides which locomotives, which wagons, which buildings, which colour the signals are, and — most usefully — what you do not buy.
The single most common cause of a stalled collection is a shelf of unrelated models bought because they were on offer. A brief turns browsing into shopping.
You are allowed to break the brief. Knowing you are breaking it is the point.
DC: one controller varies the voltage on the rails; every locomotive on that wiring section obeys. Cheap, simple, and it means dividing your track into electrically isolated blocks if you want two trains moving.
DCC: the rails carry a constant digital signal; each locomotive has a decoder with its own address, so it responds only to commands addressed to it. More expensive up front, far less wiring, and it is what almost everything is sold ready for.
If you are building anything with more than one locomotive, start DCC and skip the rewire — how DCC works →
Track brands differ in rail height, sleeper spacing, geometry and how their turnouts handle the frog. Mixing them is possible but tedious. Pick one family now:
Whatever you choose, buy flexible track rather than short curved sections as soon as you can. It is cheaper per metre and lets you use a radius of your own choosing.
Three honest answers, all fine:
Continuous run, long trains, scenery you sit and look at. Wants: an oval or a folded loop, generous curves, a view block so the same train reappears as if from somewhere else.
Shunting puzzles, deliveries, a timetable. Wants: turnouts, spurs, a run-round, and almost no continuous run at all. Fits a shelf.
The railway is a stage for modelling: kit-building, weathering, structures. Wants: small track plan, big scenic ambition, good lighting.
The binding constraint is not length, it is the width needed to turn a train round. A continuous run needs two half-turns, so the board must be at least twice the minimum radius plus a margin. These are the practical minima — the radii at which ordinary models behave.
All of the above is about turning round. If trains arrive from off-stage and leave the same way — a shelf switching layout with a cassette at each end — the width constraint vanishes and HO fits comfortably on 30 cm of depth, O on 45 cm.
All three end in the same place. They differ in what you learn first and what you waste.
A box with a locomotive, three or four wagons, an oval of sectional track and a controller, for $120–250.
You get: a running train in an hour, and a locomotive worth keeping.
You waste: the track. Set curves are 15–18 in radius and the rail is chunky. Treat it as a test track, not a foundation.
One good locomotive, six wagons, a length of flex track, four turnouts, a DCC starter system. $400–600.
You get: everything survives into the real layout.
You waste: two evenings of research, and you will make one wrong choice. Everyone does.
Club sales, estate sales, swap meets. A box of 1990s HO for the price of one new wagon.
You get: volume, and permission to experiment — this is the stock you learn to weather on.
You waste: time. Expect horn-hook couplers, plastic wheels and at least one motor that has died of dried grease.
For a 6 ft × 3 ft HO or OO layout. Buy the tools first — they are what turn a box of parts into a railway, and they last forever.
| Buy | Which one | Why it matters |
|---|---|---|
| Track gauge | NMRA standards gauge, or a NEM equivalent | A slotted piece of steel that tells you whether track, wheels and couplers are within tolerance. It will find faults you would otherwise chase for months. |
| Rail nippers | Xuron 2175 or equivalent | Cuts rail flush without crushing it. Side cutters from a toolbox will mangle the railhead and cause derailments at every joint. |
| Soldering iron | 25–40 W, chisel tip, plus 60/40 or lead-free rosin-core solder | Feeders soldered to the rail are the single biggest reliability upgrade on any layout. |
| Multimeter | Anything with a continuity beep | Answers “is there power here?” in two seconds instead of twenty minutes. |
| Flex track | 6–8 lengths, code 83 (HO) or code 75 (OO) | Smooth curves of any radius, fewer joints, and it looks right. Code 100 is fine if you have older stock with deep flanges. |
| Turnouts | 4 × #6 (or Peco medium radius) | The #4 turnouts in most sets look toy-like and restrict what can pass. #6 is the sweet spot for everything but the tightest shelf. |
| Cork or foam roadbed | A roll, plus a tube of latex caulk | Raises the track, gives it a shoulder to ballast against, and deadens the drumming noise from the baseboard. |
| Baseboard | A hollow-core door, or 9 mm ply on a 50 × 25 mm frame, topped with 25 mm extruded foam | Flat and rigid. Foam on top lets you carve down for water and push pins in anywhere. |
| Control | A DCC starter set (NCE Power Cab, Digitrax Zephyr, Z21 start, Roco/Hornby elite) | 2 A is plenty for a small layout. Buy one that lets you add a bigger booster later. |
| Wire | 16–14 AWG for the bus, 22–24 AWG for feeders | A fat bus under the board with thin droppers to each length of rail. Rail joiners are mechanical, not electrical. |
Fourteen steps, in order. Tick them off — the list remembers between visits. Nothing here needs a skill you don't already have; several steps are mostly waiting for glue.
Including the person who wrote this page.
Ballast glues everything solid, including the fault. Prove the track over days of running first. If you must fix something later, isopropyl alcohol softens dried PVA.
“Minimum radius” on the box means the tightest curve a short wheelbase model will negotiate without falling off — not the radius at which anything looks or runs well. Add 25% and your problems mostly disappear.
They are springs made of nickel silver; they oxidise, they loosen, and they get pushed apart by expansion. Feeders every metre at worst, every length at best.
A reverse curve throws a long vehicle one way then instantly the other, and the couplers pull it off the rails. Leave at least one vehicle length of straight between opposing curves.
A loop is fun for about three weeks. Two turnouts and a siding is fun for years, because it gives you something to do.
Static grass is more fun than solder. The railway still needs solder.