Anatomy of a Station

A radio station is not one thing. It is four things stacked vertically.

Each layer rests on the one below it. Together they form a continuous loop of content and data.

Most people only ever see one layer. The listener sees the player. The music director sees the songs and owns the programming. The engineer sees the tech.

Fun aside. In large radio stations you will find engineers who have no idea how to build a radio station from scratch. The tech side grows so wide that specialists end up owning a slice - software, or hardware, or transmission - without touching the rest of it.

The owner sees the bill and the regulations.

Each of them is right about their part. None of them sees radio as a whole.

1. Library (The Foundation)

The library is bedrock. The library is what plays.

Something is always playing on a radio. Even on a talk show, music sits behind the voice. Why? Pre-recorded podcasts get edited until people sound like they were waiting their whole life for the question. Live on-air is different. We pause. We think. We mess up. Music covers it.

The library is also the data about what plays.

Songs. Artist names. Album, year, country, ISRC, genre, tempo, energy. Cover art. License notes. Cold ending or faded ending. Seasonal flags. Male voice, female voice, duo. Vocalists tagged across every band they sing in. Every piece of metadata you will ever need to schedule that track later.

Take Maynard James Keenan. He fronts Tool. He also runs A Perfect Circle and Puscifer. Plus the side projects and guest spots. One human, a handful of bands, and a library row for every track he ever sang on. The tags have to carry all of it, or the rotation cannot tell them apart.

A station with a bad library has a ceiling.

You cannot program what you cannot find. You cannot find what is not tagged. You cannot tag at scale without a plan.

The library is the foundation. Everything else stands on it.

2. Programming (The Brain)

If the library is the pantry, programming is the chef. This layer decides what plays, when, and why.

The library is a pile. Programming is the order. The difference between a station and a shuffle.

Library work hurts your hands. You sit and enter data for hours. Programming hurts your brain. You aim at a target that moves while you shoot, and half your shots miss.

Good programming has rules. Good programming has no rules. Two music directors handed the same library will build two completely different stations, and neither will ever be satisfied with the result. The work is a permanent loop of refine and adjust. If your music director is satisfied, fire them.

So what is programming? Clocks for each hour, where every event that has to happen in that hour is planned in advance. Categories - songs, links, spots, sweepers - grouped into bins. Each position in a clock is either a category or a specific event. You play roughly fifteen songs an hour, depending on track length. That is where the math of "how many tracks per category" enters.

Math, by example.

Take one category. Call it A. Your hourly clock has three A positions. You schedule that same clock 24 times a day. That is 72 A plays per day. You want every song in A to spin three times a day. Divide. Twenty-four songs. Done.

Wrong.

With 24 songs and 72 fixed positions, the math is too clean. Song 1 lands at midnight today, at midnight tomorrow, at midnight the day after. Every other track does the same. The whole library plays in lockstep. A regular listener catches the same song at the same hour on their commute, and they will tell you about it.

The correct answer is 25. Not because the math demands it. Because 25 does not divide cleanly into 72. The pattern drifts by one slot every cycle. Same number of plays. No fixed timestamps.

You must know the math to plan a rotation. You must drift away from the math to make one that sounds alive. That is the fun of it.

Rotations, priorities, dayparts, turnovers, separation windows - all kinds of constraints make programming possible. Too many of them make it impossible.

I earned my stripes by trial and error. Reading the RCS Selector manual at night and testing things on Radio SWH Rock. The CTO hated my permissiveness. Since then we have built four radio stations together. As they say, there is one step between hate and love.

Programming is the mental model the whole industry runs on, even when the tool wearing the badge is something else. Most stations use maybe 10% of what their scheduler can do. Sad, but real.

Most of this site's programming chapter is written around the big tools - Selector, GSelector, MusicMaster.

Smaller independent stations save the money with open-source schedulers, freeware, or cheaper offerings. The elephant in the room: radio scheduling software has been stuck somewhere around 2010. This is where programming starts blurring into infrastructure.

3. Infrastructure (The Engine)

Infrastructure is the plumbing that turns the brain's creativity into signal.

Like programming, infrastructure has many shapes at many levels.

The basic shape is on-air automation software plus an encoder. That is the bare minimum for an internet-only, music-only station.

Going on the mic changes the game. Mixers, microphones, wiring, external controls for the automation software. Want callers? Add a phone hybrid. Just because you can hear a caller and speak back does not mean you can plug that audio cleanly into an on-air show. And by the way - digital or analog?

Then there is broadcasting. AM, FM, DAB, internet-only, any combination of them. Different technologies, different complexity.

AM and FM go out from radio towers. You have to deliver the cleanest possible audio signal to the tower. In the FM case, an FM transmitter at the site converts that audio into a radio-frequency signal. The signal moves to the antenna, the antenna spreads it, and a home radio or a car radio receives whatever is delivered. That description is the simplest one I can give.

I have a joke about radio waves. They are like a kid with ADHD. Every obstacle - a leaf, a wall, a cloud - is something to stumble into.

You cannot just transmit. You need a license, you follow regulations, you pay fees and penalties. Is there anything funnier than being in debt months before you can start broadcasting? Specifics vary by country, but the pattern is universal: you pay for spectrum long before it carries your sound.

DAB I have not worked with. It is out of my scope.

Internet-only needs a service or a server to stream to. Plenty of ways to do that, and the following articles dig in. Icecast, Shoutcast, Wowza, nginx, Liquidsoap, FFmpeg are the key tools.

One thing is common across all of it. A friend of mine put it best:

No one has ever let me down like… technology.

• @elnormous

True. You can have a rock-solid stack and on the first day of your vacation you get: "Streams are down. Can you…?"

Infrastructure proves that the library and the programming exist.

A station with a great library, a great schedule, and broken infra is silence. A station with a mediocre library and rock-solid infra still streams.

You do not have to love infrastructure. But you must respect it.

Most of the work happens at night. Restarting a streaming server or switching off a transmitter at peak listening hours is bad practice. Nobody forgives that.

4. Distribution (The Face)

Infrastructure gets the signal out. Distribution decides who finds it.

Distribution is how listeners arrive.

The web player on your site. The mobile apps. The car - CarPlay, Android Auto.

Listeners are the driving force and the data. The reason to exist.

It sounds harsh, but listeners are also the burden. Read "What is radio?" for the long version. Listeners build a personal connection to a station, and any sharp change in programming can leave you without them. If a host moves to a new station, part of their regulars move with them. An audience is on loan. Brutal on FM, where every dropped listener is a dropped sale.

I have an unpopular opinion about directories. TuneIn, Apple, the rest. They are mold.

No one in the industry counts listeners by server logs. Listeners are counted on websites and apps.

Directory listeners are invisible to a station. They say exposure. I say zero exposure, zero data, and a non-zero bill. The position is based on actual analytics, not opinion.

What makes it worse: they embed your raw stream on their site and pay no royalties on it. They sell ads against the audience you handed them. You pay the infrastructure. You pay the royalties. They pay nothing. So, mold.

Next time someone tells you to put your radio on an aggregator's site, hit them in the mouth. You do not build an audience on aggregators. You build it with social presence.

Without distribution, you are broadcasting to yourself. Finding listeners is one of the hardest parts of radio. Most stations expect listeners to find them on their own. They do not.

How they feed each other

The four layers are not steps. They are a loop.

• Library feeds programming. You can only schedule what you have, tagged correctly.
• Programming feeds infrastructure. The clock demands fast loads, accurate now-playing data, gapless transitions.
• Infrastructure feeds distribution. An unreliable stream gets dropped from directories, churns listeners out of apps, and breaks embeds.
• Distribution feeds library. Listener data, requests, and gap analysis tell you what to acquire and what to retire.

Break any one of them and the others degrade.

A station is never "done." You are always pulling on one of these four threads.

The map for the rest of this site

Every other section of this site lives inside one of these four layers.

• Library → Music Library & Metadata
• Programming → Programming & Rotation
• Infrastructure → Streaming Infrastructure
• Distribution → Web, Mobile & APIs · Listeners & Distribution

If you skip ahead and feel lost, the answer is almost always "I jumped a layer." Come back here. Find the layer. Read the section that owns it.