The Model Railway Handbook

Power, DCC and wiring

The rails carry the power. With DCC they carry the conversation too.

What the difference between DC and DCC actually is at the rail, what a decoder does, how to wire a layout so it stays reliable, the handful of CVs worth knowing, and why a reverse loop shorts out.

DC versus DCC, at the rail

On DC, the controller varies the voltage between about 0 and 12 V and flips its polarity to change direction. The locomotive is a dumb motor: more volts, more speed. Every locomotive on that piece of wiring does the same thing at the same time.

On DCC, the rails carry a constant, full-voltage square wave that swings between + and −. That waveform is both the power supply and a stream of digital packets. A short half-cycle means a 1, a long one means a 0. Each locomotive has a decoder that reads packets addressed to it and ignores the rest.

DC — smooth, variable, one train per wired section +12 V, forward, fast 0 V, stopped −12 V, reverse DCC — full voltage, always; the timing carries the message short pulses = 1 1 1 1 long pulses = 0 0 0 +15 V −15 V A DCC “1” half-cycle is 58 microseconds; a “0” is at least 100. The motor sees the average — which is zero — so a decoder is required to do anything at all.
This is why an ordinary DC locomotive placed on a DCC layout sits and hums: it is being fed 15 V of alternating nothing.

DC still makes sense when…

…you run one train at a time, you like the simplicity, you have a fleet of older models that would each need a decoder, or you are building something small and temporary. It is not a wrong answer.

DCC pays for itself when…

…you want two trains moving independently, you want a locomotive to sit idling in a siding with its lights on, or you want sound. The wiring saving alone — no block switches, no section wiring — is substantial.

What a decoder actually does

rails15 V bipolar + data rectifiermakes usable DC processorreads packets for its address CV memoryaddress, momentum, curves motor driver (PWM)+ function outputs for lights, smoke, couplers The processor drives the motor with pulse-width modulation — which is why a good decoder can creep a locomotive at one scale mile per hour and a DC controller usually cannot.
Back-EMF, the feature worth paying for

Between motor pulses the decoder measures the voltage the motor generates as it coasts, works out how fast it is really turning, and corrects. That is why a good decoder holds a steady crawl up a gradient and through a turnout. It is sometimes called load compensation or “silent drive”.

Bus and feeders — the wiring that never gives trouble

Run a pair of heavy wires around the layout underneath — the bus — and drop a pair of thin feeders from every length of rail down to it. Do not rely on rail joiners to carry current between sections: they are mechanical connectors that happen to conduct, until they don't.

rail joint rail joint rail joint baseboard bus + bus − command station bus: 16–14 AWG (1.5–2.5 mm²), twisted a few turns per metre feeders: 22–24 AWG, as short as possible, soldered to the outside web of the rail
Keep the colours consistent all the way round — red to the same rail everywhere — and a reverse-loop short becomes obvious instead of mysterious.
The quarter test

Lay a coin across both rails anywhere on the layout. The command station should shut down instantly. If it merely grumbles, or nothing happens, the wiring at that spot is too thin or too far from a feeder — and a real short there would cook something rather than trip the protection.

Power districts and boosters

A starter command station supplies 2–3 A, which runs perhaps four or five HO locomotives — fewer with sound. Beyond that you add a booster: its own transformer and output stage feeding its own section of track, taking commands from the same bus.

Even on a small layout it is worth dividing the track into two or three power districts, each behind its own electronic circuit breaker. A derailment in the yard then stops the yard, and the train on the main line keeps going.

Rule of thumb  1 A per non-sound HO locomotive, 0.5 A per sound one at idle
Districts  main line · yard · engine terminal · each reverse section
Gap both rails  between districts, and never bridge them with a metal-wheeled train alone

Reverse loops: why they short, and the fix

Follow the left-hand rail round a loop that turns a train back on itself and you will find it arrives at the right-hand rail. Two different polarities, one piece of metal — a dead short, and it happens on DC and DCC alike.

insulated gaps in BOTH rails, at both ends of the loop auto-reverser flips polarity in ~10 ms from the bus feeds only the isolated loop section The loop section must be longer than your longest train, so that no vehicle bridges both gaps at once.
On DC the same job is done by a DPDT switch and a steady hand. On DCC the module does it before the locomotive notices.

Wyes and turntables have exactly the same problem and exactly the same cure. A turntable bridge is a reverse section that rotates.

The CVs worth knowing

configuration variables

A decoder's settings live in numbered memory slots. There are over a thousand; you will use about eight. Program them on a dedicated programming track (low power, reads back values) or on the main line with POM — programming on the main — which writes to one address while everything else keeps running.

CVWhat it isTypical use
1Short address, 1–127The number you call the locomotive by. Set it to the last two digits of the running number.
2Start voltage (V-start)Raise it until the model just creeps at speed step 1. The single most satisfying adjustment there is.
3Acceleration rateMomentum. A heavy freight should take several seconds to get moving; a light railcar should not.
4Deceleration rateThe same for braking. Set it high and you must plan your stops — which is the point.
5Maximum voltage (V-high)Cap a model that runs absurdly fast at full throttle, or match two locomotives for double-heading.
8Manufacturer IDRead-only — but on most decoders writing 8 to it performs a factory reset. Remember this one.
17 + 18Long address, 128–9999Set together; most command stations do the arithmetic for you when you type the number.
29Configuration bitsDirection of travel, 14 vs 28/128 speed steps, whether the long address is used, whether DC running is allowed.
19Consist addressPuts several locomotives under one throttle for double-heading, keeping their own lighting behaviour.
Write down what you changed

An hour of tuning is undone by one accidental reset. Keep the address, decoder type and any altered CVs on a card with the model's box.

Sound, and the capacitor that saves it

A sound decoder carries recorded samples of a specific prototype and plays them against the motor's actual load: notch up on a climb, chuff in time with the drivers, brake squeal on a stop. It needs a speaker with a sealed enclosure — the enclosure matters more than the speaker — and it draws current even when the locomotive is standing still.

A keep-alive (stay-alive) is a bank of capacitors that keeps the decoder powered for a second or two across a dirty patch of rail. It stops the humiliating reboot mid-whistle, and it improves slow running even without sound. Fit one if there is room.

Volume is a modelling decision

Prototype sound at scale distance is quiet. A locomotive that can be heard across the room is roughly ten times too loud. Turn the master volume down to about a third and the illusion improves enormously.

Powering turnouts

Solenoid

A coil that snaps the blades over with a bang. Cheap, fast, noisy, and it needs a capacitor discharge unit to fire reliably without cooking the coil. Peco PL-10, Seep, Hornby.

Slow-motion / stall motor

A geared motor that pushes the blades across over a couple of seconds and then stalls harmlessly against them. Quiet, prototypical, and usually has spare contacts for frog polarity and signalling. Tortoise, Cobalt.

Servo

A hobby servo plus a small driver board. The cheapest slow-motion option, fully adjustable in throw and speed, and easy to control from a microcontroller if you enjoy that side of things.

On DCC, a stationary (accessory) decoder lets you throw turnouts from the same handset that drives the trains, and lets a route be set with one button. Very satisfying — and entirely optional. Wire in wire-in-tube from a knob on the fascia works perfectly and never fails.

When it shorts: finding it in five minutes

  1. Take everything off the track. If the short persists, it is wiring or trackwork, not a train.
  2. Disconnect districts one at a time. The one that clears the fault contains it. This is the entire argument for having districts.
  3. Look at the last thing you changed. A feeder soldered to the wrong rail, a metal wheel bridging an insulated gap, a dropped screw across the rails.
  4. Check every insulated joint with the meter. Continuity across a gap that should be isolated is the classic reverse-loop error.
  5. Push a wagon slowly through each turnout. On a live-frog turnout, a wheel touching the wrong-polarity blade produces a short only in one direction of travel — which is why it seemed intermittent.