Formats and bitrate: MP3, AAC, OGG, Opus, CBR, VBR, and DC offset.

An audio file is a compromise: retain everything and use more space, or discard information and use less. Lossy formats do the latter systematically, using perceptual models to discard what is least audible. Problems begin when decisions are chained unknowingly—a WAV converted to MP3 and converted again by someone else—and when bitrate is treated as a quality label instead of what it is: a budget.

This chapter covers MP3, AAC, OGG, and Opus, CBR vs VBR, how bitrate affects audio, detecting transcodes, DC offset, and format conversion without unnecessary degradation. The preceding material—spectrograms and interpreting frequency cutoffs—is in the formats, WAV, FLAC, MP3, and spectrograms guide.

MP3, AAC, OGG, and Opus

MP3 is history’s most widespread lossy format: it divides audio into blocks, uses psychoacoustic analysis to remove what your hearing will not miss, and saves the rest with an assigned bit rate. It works well, fits everywhere, and supports everyday quality levels through 320 kbps.

AAC emerged as its successor and belongs to MPEG-4’s audio standard, explaining its presence in Apple profiles, YouTube, and most video platforms. At equal bitrate it usually outperforms MP3, especially at medium rates.

OGG names the container; Vorbis is its most associated codec, openly developed under a free licence. Its official documentation gives a warning relevant throughout this chapter: converting one lossy format to another is generally a bad idea.

Opus is the newest of the four, defined in RFC 6716 and royalty-free. It combines speech- and music-oriented engines and spans 6 to 510 kbps, adapting from video calls to music at 48 kHz. It supports constant and variable bitrate plus memory between windows, so the next section’s distinction applies directly.

CBR vs VBR

CBR (constant bitrate) maintains the same bit rate second by second. Its virtue is predictability: you know the file size before export, and equipment assuming fixed bitrate encounters no surprises.

VBR (variable bitrate) does the opposite: spends more bits on complicated sections—transients, dense passages, complex stereo—and fewer on easy ones—silence, flat material. The same file size sounds better, or the same quality takes less space.

When to choose each: for distribution, high target-quality VBR is almost always the right answer. For practical needs—a backing track with guaranteed duration, size-limited delivery, an old playback device—CBR avoids questions. Many mixing applications and streaming platforms use VBR anyway, because a commercial track’s effective bitrate is rarely constant.

How bitrate affects audio

Bitrate is the number of bits allocated per second to describe audio. Fewer bits mean more quantisation of what remains after perceptual analysis: treble smearing, transient pre-echo, and a sense of slight veiling appear over time. More bits reduce these effects until, beyond a point, differences become inaudible.

Three important qualifications. First, bitrate is not an isolated indicator: a modern codec at 256 kbps can sound better than an old one at 320 kbps. Second, each format has a ceiling: MP3 stops at 320 kbps, while Opus goes higher. Third, material decides: percussion-heavy tracks with wide stereo need more bits than a vocal with flat accompaniment.

Set a working criterion: for editing or mixing, WAV or FLAC is enough; for distribution, choose codec and bitrate for the audience and platform, and stop treating the number as a guarantee.

How to detect transcodes

A file claiming 320 kbps may come from up-converting 128 to 320: size changes, but the sound is already marked. Detection matters because the second compression restores nothing and adds its own artefacts.

The classic method is spectrogram inspection. Each codec and bitrate leaves a signature: band cutoffs below the declared format’s expected reach, extra noise floor where the perceptual model worked, or no energy where the original material had it. This is exactly the principle behind automated bitrate detection: frequency-spectrum analysis classifying a file’s encoded bitrate class, with research reporting accuracy around 97 % across five common classes.

Helpful indirect clues: inconsistent metadata, names promising more than the spectrogram delivers, and provenance. Two successive conversions degrade a file at every step, and no metadata can fix it.

What DC offset is

DC offset displaces the signal’s mean amplitude from zero: the waveform oscillates around an offset value instead of the centre. It behaves like a phantom constant level added to all audio.

Common causes lie in capture and hardware calibration, but origin does not matter: the effect is the same. It moves clipping closer without the waveform deserving it, complicates otherwise clean cuts and edits, and interferes with processes relying on signal symmetry, such as some distortion types.

Manuals recommend checking and removing it as an early editing step. One detail: a severe high-pass filter may mask the problem, but the correct solution is subtracting the constant, not carving away bass.

How to convert formats without unnecessary degradation

One golden rule: convert from the best available source, once. If you have the original WAV, convert that; if you only have a 192 kbps MP3, do not raise it to 320 and call it an improvement—it remains the same audio with another label.

Pairs that make sense:

• WAV or AIFF → FLAC: retains everything and uses less space. It is the only high-quality format conversion that loses nothing.
• FLAC or WAV → MP3/AAC/Opus: the first and only loss, chosen with bitrate and codec judgement.
• MP3 → another lossy format: avoid it. Vorbis documentation states directly that converting lossy formats to other lossy formats is generally a bad idea.
• MP3 → WAV: only makes sense for tool compatibility, not quality: the resulting WAV contains exactly the already degraded audio.

A ten-second final check: after export, compare the new file’s spectrogram with the original’s. If the band cutoff dropped or a new edge appeared, something in the chain was not what you expected.

Common conversion and tagging mistakes

Renaming instead of converting. Renaming MP3 to FLAC does not make it lossless; the contents remain lossy-compressed.

Chaining compression. Every lossy pass accumulates. Always produce from the master source.

Choosing the largest bitrate number. Correct bitrate depends on codec, material, and destination, not the maximum available figure.

Ignoring DC offset because “it is inaudible”. It is inaudible, but affects clipping, cuts, and rectification processes.

Trusting metadata to judge a file. Metadata is copied and forged; a spectrogram does not lie. With this material, you have the complete series’ chain: from what the room does to sound to its packaging in a file.

Next step

You have reached the end of the fundamentals pathway: sound, level, metrics, digital audio, and formats. You can now branch towards the waveform-reading pillar or meters and gain staging, according to this week’s work.