What to pay attention to when mastering

Mastering is mostly checking. The processing takes minutes; the checking is what stops a problem reaching thirty thousand listeners.

These are the checks, ordered so the expensive problems surface first.

1. Tonal balance, against references

Level-matched, switching every few minutes.

Ask specifically:

• Is it brighter or darker than the reference?
• Is there more or less weight in the low end?
• Do the low mids feel crowded?
• Is the top end open, or harsh, or dull?

Answer in words before touching a control. A vague sense that “something is off” produces vague moves.

2. Low end

The most common failure and the most consequential.

• Is there content below the useful range that is eating headroom for nothing?
• Does the kick keep its definition when the bass plays?
• Is the low end mono, or is it spreading and losing energy?
• Does the bass still exist on a small speaker?

That last one is not optional — most listeners are on small speakers. If the bass relies on a pure sub, it vanishes there. See how to saturate bass.

3. Transients

Play the loudest section and ask whether the drums still hit.

The single easiest mastering mistake is trading punch for loudness without noticing, because the trade happens gradually as you push the limiter.

The test: bypass the limiter, match the level, and compare. If the limited version feels smaller, you went too far.

4. Stereo width

• Is the image wide enough to feel finished, without the centre going hollow?
• Are the low frequencies centred?
• Does anything sound like it is outside the speakers in a way that will collapse elsewhere?

Width added at the mastering stage is easy to overdo, and it is the most common cause of a master that sounds impressive on headphones and wrong everywhere else.

5. Mono compatibility

Sum to mono and listen.

What to expect: a slightly narrower, slightly quieter version of the same track. What to watch for: elements disappearing, the low end thinning, a wide synth becoming hollow, the vocal changing level.

Anything that vanishes in mono was relying on phase relationships that many playback systems will destroy. This is a phase problem, and it usually has to go back to the mix.

6. True peak

Sample-peak metering will tell you the file is safe when it is not. The waveform reconstructed by a converter passes between the samples and can overshoot.

Enable true-peak metering, set the ceiling with a margin below full scale, and leave the margin alone. Detail in true peak vs sample peak.

7. Loudness, in context

Check the integrated loudness — and then remember that streaming platforms normalise playback, so being louder than your reference buys nothing except less dynamic range.

The useful question is not “what number am I hitting” but “does this sound right next to the records it will be played next to”. See how loud should your master be.

8. The file itself

Boring and easy to get wrong:

• No click at the start.
• The tail is complete and not cut short.
• Correct sample rate and bit depth for the delivery.
• Dither applied once, at the end, only if reducing bit depth.
• Sensible file name.

9. Multiple systems

System	What it reveals
Phone speaker	Whether the bass exists at all; midrange balance
Laptop	Harshness, vocal intelligibility
Car	Low-end weight, real-world balance
Headphones	Detail, sibilance, stereo problems
Quiet monitors	Overall balance — the most reliable single check

If it works on all of those, it will work almost anywhere.

10. A different day

The last check. Export, sleep, listen fresh before sending it.

You will hear one thing you missed. Everyone does, and finding it the day before release is much better than finding it the week after.

Consistency across a release

If the track is part of an EP or album:

• Do the tracks feel like the same record?
• Does anything jump out as noticeably louder or quieter?
• Are the gaps between tracks intentional?

Feel matters more than measurement. A sparse ballad and a dense club track at identical measured loudness will not feel equal, and the ear is the arbiter.

The tool for the measuring half

SPC SpectraScope covers spectrum, K-weighted loudness with BS.1770 true peak, stereo field and phase correlation as an analysis-only visualiser — it never touches the audio, which is the correct relationship. The judgement stays with you; see meters vs ears.