Analysis: spectrum, LUFS, true peak, correlation, crest factor

Why measure

Your ears fatigue and adapt; the room lies (resonances, missing or inflated bass); monitors color the sound. Analysis tools don't replace listening, but they confirm or contradict what you hear and let you compare against references. Rule: listen, then look, then listen again.

Spectrum analyzer

• What it shows: energy per frequency (FFT). Displayed as a curve (average) or as a spectrogram (frequency over time, with color as level).
• Settings: resolution (FFT 4096–16384 to see the bass clearly; higher = slower), averaging/smoothing (long averaging for the overall shape, short for transients), display slope (slope/tilt): at 3 dB/oct pink noise looks flat; at 4.5 dB/oct (the default of the Fruity Parametric EQ 2 analyzer) a typical music mix looks roughly flat; it is only there to make reading easier.
• What to look for: narrow resonances (1/3-octave peaks sticking out by more than 6 dB), buildup at 200–400 Hz, missing or excessive sub compared to the references, highs that are too harsh (2–5 kHz) or missing (above 10 kHz), rumble below 30 Hz.
• In FL Studio: the Fruity Parametric EQ 2 analyzer (histogram or heatmap 20 Hz–20 kHz, range −60/−90/−120 dB, slope 1.5/3/4.5/6 dB/oct, monitoring of Mid, Side, Left or Right); Wave Candy in Spectrum mode (scrolling spectrogram with a piano keyboard for reading notes); Fruity Spectroman; Edison (spectral view). Free: Voxengo SPAN, MiniMeters.

Loudness (LUFS)

• LUFS (Loudness Units relative to Full Scale, = LKFS): a measure of perceived loudness according to ITU-R BS.1770 (K-weighting filter that emphasizes the upper mids and attenuates the bass, channel summing, gating).
• Momentary (0.4 s window): fast response.
• Short-term (3 s): the "section" loudness; useful for comparing verse and chorus.
• Integrated: the average over the whole track with gating (very quiet parts don't count). This is the value platforms use for normalization.
• LRA (Loudness Range): how much the loudness varies across the track, in LU (based on the distribution of short-term values). Pop/EDM 3–6 LU, rock 4–8, jazz/acoustic 6–12, classical 10–20.
• LU: a relative unit (1 LU = 1 dB of loudness difference).
• In FL Studio: Emphasis shows momentary, short-term and integrated LUFS and Range; Wave Candy in Peak meter mode with the ITU-R BS.1770 setting gives a weighted average according to the standard (without the full readout of a loudness meter). Free: Youlean Loudness Meter, TBProAudio dpMeter.

Peak and true peak

• Sample peak (dBFS): the highest sample.
• True peak (dBTP): an estimate of the peak of the reconstructed waveform (oversampling), which can exceed the sample peak by tenths of a dB or more after heavy limiting. A streaming master should stay below −1 dBTP (−2 dBTP if above −14 LUFS on Spotify).
• Meters: Emphasis (True Peak), Youlean, Pro-L 2 (TP metering), Insight 2.

Crest factor, PLR, PSR

• Crest factor = peak − RMS (in dB). It tells you how "spiky" the transients are relative to the average. A sine wave has 3 dB; an unprocessed drum kit 15–20 dB; a heavily limited master 5–7 dB.
• PLR (Peak to Loudness Ratio) = true peak − integrated loudness. Example: −1 dBTP and −9 LUFS → PLR 8 dB.
• PSR (Peak to Short-term loudness Ratio) = true peak − short-term: density section by section.

Correlation and vectorscope

• Phase correlation (−1/+1): +1 = mono (L and R identical); 0 = independent; negative = out of phase (cancellation in mono). A healthy mix: +0.2/+0.8; below 150 Hz above +0.9.
• Vectorscope (goniometer): vertical shape = center; tilted = L/R imbalance; horizontal = out of phase; width of the cloud = stereo width.
• In FL: Spreader (pre/post phase meters with values near +1 = good, near −1 = cancellation; vectorscope; balance), Wave Candy (Vectorscope, without a numeric correlation readout). Free: Ozone Imager, SPAN (correlation).

Comparing with references

1. Choose 2–3 reference tracks in the genre (quality files, not compressed YouTube rips).
2. Import them on a mixer track that does not pass through the master chain (routed directly to the output or with the master chains bypassed) or use a comparison plugin (ADPTR Metric AB, Mastering The Mix REFERENCE, iZotope Tonal Balance Control).
3. Match the loudness (e.g. everything at −14 LUFS integrated): the louder one always seems better.
4. Compare: tonal balance (average spectrum), vocal level relative to the music, weight of the low end, width, density (crest factor, LRA), shape of the sections (short-term).

Common mistakes

• Mixing with your eyes: a "flat" spectrum is not a goal.
• Comparing references at different loudness levels.
• Using a single value (integrated LUFS) as a quality judgment.
• Using low-resolution analyzers to judge the bass.
• Ignoring true peak because the sample peak is below 0.

How to use them in your workflow

• During the mix: spectrum and correlation on tracks and buses; gain staging (peak/RMS).
• At the premix: short-term per section, crest factor, correlation, spectrum against the references.
• At the master: integrated LUFS, true peak, PLR, LRA, mono, spectrum.