Why MP3Gain Changes Volume in 1.5 dB Steps

MP3Gain changes volume in 1.5 dB steps because that is the finest adjustment the MP3 format itself allows without re-encoding. Every MP3 frame stores a gain value, and that value can only move in fixed increments: each step multiplies the signal by 2^1/4 (about 1.19×), which works out to 1.5 dB. So MP3Gain rounds each track’s ideal adjustment to the nearest available step, landing at most 0.75 dB — half a step — from the exact target. Finer steps exist only on the other side of a decode-and-re-encode, which is precisely what MP3Gain refuses to do.

Inside an MP3 frame

An MP3 file is not one continuous stream of sound; it is a chain of thousands of frames. Each frame contains 1,152 samples of compressed audio — roughly 26 milliseconds at CD sample rates — plus a small block of “side information” telling the decoder how to reconstruct that slice: which compression tables to use, and at what level to play it. That playback level is the global gain field, and it is the only volume control the format exposes without touching the audio data itself. MP3Gain’s entire lossless method, explained in Is MP3Gain Lossless?, consists of rewriting this one field, frame by frame.

Why exactly 1.5 dB?

The global gain is a counter, not a dial. Increasing it by one does not add a fixed amount of amplitude — it multiplies the decoded signal by the fourth root of two (≈1.189). In decibel terms, that multiplication is 20 × log₁₀(1.189) ≈ 1.51 dB — which everyone rounds to 1.5 dB. The MP3 designers chose this exponential scale deliberately: it gives a wide range of adjustment (the field spans far more range than any normalizer needs) while keeping the side information small. But it hard-codes the granularity. There is no setting, in any tool, that makes a frame-level gain change of 0.5 dB or 1.0 dB, because the format has no way to write it down.

What “within 0.75 dB of target” means in practice

Say analysis decides a track needs −5.2 dB to reach the 89 dB target. MP3Gain cannot apply −5.2. It applies −4.5 dB (three steps) or −6.0 dB (four steps) — whichever is nearer, here −4.5, leaving the track 0.7 dB hotter than the ideal. The worst possible case is a requirement falling exactly between two steps: an error of 0.75 dB, half a step.

Is that audible? A level difference of about 1 dB is roughly the smallest most listeners can detect in a direct, instant A/B comparison of the same recording — and MP3Gain’s worst-case error is below that, applied uniformly to a whole track, in a library where every track carries a similar small rounding. In real listening across different songs, a ≤0.75 dB residue is not a difference anyone can hear as a volume jump. It is also worth noting what the step does not disturb: in Album Gain mode, the same number of steps is applied to every track, so the carefully preserved level differences within an album remain exact.

The alternative MP3Gain refuses

Any tool can change volume in 0.1 dB increments — by decoding the MP3 to raw audio, scaling it precisely, and encoding a new MP3. The cost is a second generation of lossy compression: artifacts compound, and the change can never be undone. MP3Gain’s designers made the opposite trade in 2001 and it remains the right one for library work: accept a rounding of at most 0.75 dB, in exchange for changes that are lossless, reversible, and honored by every player ever made. If a track absolutely must sit at an exact level for professional delivery, that is an editing job with different tools and different costs — not a normalization job.

So the 1.5 dB step is not a limitation of MP3Gain the program. It is a property of the MP3 bitstream, and MP3Gain is simply the tool honest enough to work within it rather than quietly re-encode your music.

You can watch the steps at work in MP3Gain itself. Apply Track Gain, then nudge the target volume up by half a decibel and apply again: most tracks do not move at all, because their nearest available step has not changed. Move the target by a full 1.5 dB and every applied value shifts by exactly one step. The “Track Gain” figures MP3Gain reports are always whole multiples of 1.5 dB — the frame format’s fingerprint, visible right in the interface once you know to look for it.

Why not just store a precise value in a tag?

A fair question: ReplayGain tags store adjustments at full precision — “−5.23 dB,” not a rounded step. If precision were the only goal, tagging would win. But a tag is only an instruction, and it works only where a player chooses to read and obey it. Plenty of players, car stereos, and hardware devices ignore ReplayGain tags entirely, and on those devices a precisely tagged file plays at its original, uneven volume. MP3Gain’s founders chose the opposite trade: write the change into the frames, where every decoder must obey it, and accept the format’s 1.5 dB granularity as the price. Precision you can’t hear, traded for compatibility you use every day.

Steps do not accumulate

One more worry worth retiring: processing a file twice does not double the rounding error. MP3Gain re-analyzes the track’s loudness on each pass and adjusts the gain field to the value the current target requires — it does not blindly stack another rounded change on top of the last one, and the undo data in the tag always remembers the original state. An MP3Gain-processed file can be re-targeted next year, or restored entirely, with the same ≤0.75 dB rounding it had on day one. Compare that with repeated re-encoding, where every generation permanently discards more of the audio. The step size is a fixed property of the container; unlike encoding loss, it never compounds.