What survives a phone speaker

Your mix will be heard on something worse than what you mixed on

A phone speaker is a few millimetres across and sealed into a case. Its useful output typically begins somewhere around 500 to 700Hz and falls away sharply below that. A laptop speaker does a little better, often reaching 200 to 300Hz. A pair of earbuds and a decent television soundbar do better still.

Whatever you mixed on, a substantial share of your audience will hear your work through one of those. Nothing below the speaker’s limit exists for them: the bass is not quiet, it is absent.

The missing fundamental

Play a 100Hz note on a phone speaker and you will still perceive the pitch, even though the speaker cannot produce 100Hz.

The reason is that a real sound at 100Hz is accompanied by harmonics at 200, 300, 400Hz and up. Those harmonics do come through, and the auditory system infers the fundamental from the spacing between them. This is the missing fundamental effect, and it is why a small speaker can convey a bass line at all.

Two consequences for mixing:

• A pure sine sub-bass disappears entirely on a small speaker, because it has no harmonics to infer from. If low-end content matters, it needs harmonic content above 500Hz, which is what saturation and distortion on a bass are actually for.
• Low-frequency energy you cannot hear on a phone is still in the file, eating headroom in the limiter and making the mix quieter after normalisation. This is the practical argument for high-passing everything that does not need low end, not just the voice.

What survives, and what to protect

The range from about 1 to 4kHz survives every playback system and carries speech intelligibility. If the mix works there, it works everywhere. If the mix depends on anything below 300Hz to sound full, it will sound thin to most of your audience.

This tells you where to spend care: consonant clarity, the relationship between dialogue and everything else in the midrange, and whether the music bed is competing in that band.

Checking, in three steps

1. A phone speaker, at arm’s length, at low volume. Not earbuds: the built-in speaker. This is the most brutal and most representative test available, and it costs nothing. Listen for: can you follow the words, does the music vanish or dominate, is anything unpleasantly harsh.

2. Mono. Covered in the capture module for phase reasons, and it applies again here because most small speakers are single. One click.

3. Quietly. Equal-loudness contours mean the ear is far less sensitive to bass and treble at low levels than in the middle. A mix balanced at a comfortable monitoring level will sound midrange-heavy when quiet, and most listening is quiet. Mixing at moderate to low levels most of the time produces mixes that translate better, and it protects your hearing, which is the other reason to do it.

Reference tracks

The most reliable single technique: import two or three professionally finished pieces in the same genre, level-matched to your mix, and switch between them.

Level-matching is the step people skip and it invalidates the comparison if you skip it: louder sounds better, so an unmatched reference always wins or always loses for the wrong reason. Match by integrated LUFS using the meter from the loudness lesson, then compare.

Listen for the relationships rather than the tone: how loud is the voice against the music, how much low end is there, how bright are the consonants. You are checking your decisions against people who had better rooms and better monitoring.

Your room is lying to you

Every domestic room has strong low-frequency modes: peaks and nulls of 10dB or more at specific frequencies, determined by its dimensions. If your listening position sits in a null at 80Hz, you will add bass to compensate and your mix will be bass-heavy everywhere else. If it sits in a peak, the opposite.

Two cheap mitigations. Headphones have no room, and while they misrepresent stereo width and bass extension in their own ways, they are consistent, and a known error is more useful than an unknown one. Multiple systems average out individual errors: monitors, headphones, a laptop, a phone, the car. Anything that sounds right on four different systems is right.

The free path

Everything here is free. A phone you already own, headphones you already own, ffmpeg or any editor for level-matched references, and a mono button. REW measures your room’s frequency response at the listening position for the cost of a measurement microphone, and even a rough measurement tells you where the big dips are.

The limitation

Translation checking tells you what is wrong and not what to do about it. A mix that sounds thin on a phone might need more midrange, less low end, a different arrangement, or nothing at all, because some material cannot survive a phone speaker and pretending otherwise ruins it everywhere else.

There is also no such thing as a mix that is optimal on every system. Decisions that help a phone (more midrange, controlled dynamics, less sub) cost something on a good system. Professional practice is to mix for the primary listening context and check that nothing breaks elsewhere, rather than to average across all of them and satisfy none.