How to read a spectrum analyser and a loudness meter without fooling yourself

Meters are the one part of a home studio that does not care about your room, your headphones or how tired you are. That is why people lean on them, and also why they mislead: a meter gives a precise-looking answer to whatever question its settings ask, which may not be the question you had. This guide covers the two you will use most, the spectrum analyser and the loudness meter, what their settings change, and how to use them as a check rather than a target.

If you mix mostly on headphones, read this alongside our headphone mixing guide. For the numbers platforms actually publish, see what LUFS to master to in 2026, and for peaks, limiting and true peak explained.

Part one: the spectrum analyser

What the picture is

Ableton's manual describes its own Spectrum device in terms that apply to almost every analyser: it performs real-time frequency analysis and draws a graph with level in dB on the vertical axis and frequency (or pitch) on the horizontal one. It also says plainly that Spectrum "is not an audio effect, but rather a measurement tool" and does not change the signal (Ableton manual). So put it anywhere you like, including after your master limiter.

Most analysers, including Ableton's and the free Voxengo SPAN, use an FFT (fast Fourier transform). The analyser chops audio into blocks, and NI's primer explains that the discrete transform it runs returns its result as discrete values, or bins, and that cutting a signal into finite blocks smears energy from one frequency into its neighbours, called spectral leakage (NI). Windowing reduces that smear; NI says the Hann window is satisfactory in about 95 percent of cases.

The four settings that change the answer

Block size (FFT size). Ableton says higher Block values "result in better accuracy, but at the expense of increased CPU load." The practical effect, by the usual FFT arithmetic, is that bin spacing is the sample rate divided by the block size: at 48 kHz, a 1,024-sample block gives bins about 47 Hz apart, too coarse to tell a 41 Hz kick from a 55 Hz bass note, while 8,192 samples gets you under 6 Hz. The cost is time: a longer block covers more audio, so short transients blur. Use big blocks for the low end, smaller ones for watching drums.

Averaging. Ableton's Avg control sets how many blocks are averaged per update. At one, you see every block, which is useful for catching short peaks; higher values give a smoother picture that Ableton calls "more consistent with the way we actually hear." For mix decisions, average. For hunting a single click or resonance, don't.

Slope (tilt). This is the setting that causes the most arguments online. Voxengo's SPAN guide notes that by default SPAN uses a 4.5 dB per octave slope, which makes the display look "considerably 'elevated' towards the higher frequencies in comparison to most other spectrum analyzers" (SPAN user guide). The same mix can look bass-heavy in one analyser and balanced in another purely because of tilt. Before comparing screenshots with anyone, compare slope settings.

Peak hold and max traces. Ableton's Max shows the accumulated maximum until you reset it, and SPAN can draw a secondary real-time or all-time maximum. A max trace tells you the loudest moment in each band, not what the song usually sounds like. Reset it after every change, or it will report the problem you already fixed.

Two smaller ones: a log frequency scale matches how we hear octaves, and Ableton notes that linear scaling is useful for detailed analysis of the highs. Range (the dB span on the vertical axis) changes how dramatic everything looks without changing anything that is there.

Reading it on a real mix

• Look at shape and change, not absolute height. Music has more energy in the lows than the highs, so a falling line is normal. What matters is a bump or hole that is not on your reference track through the same analyser and settings. Our reference track guide covers level-matching.
• Find, then listen. An analyser is good at telling you where a ringing resonance or a rumble is. Whether to cut it is an ear decision. Our EQ basics cover the sweep-and-cut part.
• Check what speakers can't show you. Sub-bass below what your monitors or headphones reproduce, and DC or rumble under 20 Hz, are where an analyser earns its keep. See low end and bass mixing.
• Check stereo separately. SPAN includes a correlation meter that, per its guide, shows the average correlation between the first two channels over 500 ms. Persistent negative readings point at phase trouble; our phase and polarity guide explains why.

Part two: loudness and level meters

Three time scales, three questions

The EBU's metering spec, Tech 3341 (revised November 2023), names three time scales (EBU Tech 3341):

The spec also says a meter must be able to show the maximum Momentary and Short-term values, reset together with Integrated. That gives a useful habit: watch Short-term while mixing a chorus, and only read Integrated after playing the whole song from top to bottom, because it is a measurement of the full programme.

The units: LUFS is absolute loudness against digital full scale, and LU is relative. Tech 3341 defines two display scales, an "EBU +9" scale running from −18 to +9 LU (−41 to −14 LUFS) and an "EBU +18" scale from −36 to +18 LU (−59 to −5 LUFS). If your meter's bar looks pinned near the top, you may simply be on the +9 scale with a modern loud master.

Peaks are a different measurement

Loudness and peak level are separate readings. ITU-R BS.1770, now in its fifth revision (BS.1770-5, November 2023), covers both loudness and true-peak level (ITU), and Tech 3341 sets test tolerances for true-peak meters, including the upsampling filter. A sample-peak meter only sees the stored samples; a true-peak meter estimates the level between them, which is what a converter will actually reproduce. If a platform spec says dBTP, use a true-peak meter, not the channel fader's peak light.

SPAN's guide is a good reminder that meters also have modes: it offers dBFS, an RMS-based "dBFS+3" mode, K-system modes after Bob Katz, and EBU R128 LUFS/LU modes, and notes that its own "Max Crest Factor" in LUFS mode is calculated from RMS with K-weighting rather than the R128 loudness value. Two meters that disagree are often both correct about different things.

A routine that keeps meters honest

1. Set the analyser once. Pick a slope, a large block size and moderate averaging, save it as a preset, and use it on every project and every reference.
2. Meter the reference too. Same chain, same settings, level-matched.
3. Find with the eye, decide with the ear. Solo and sweep before you cut anything the analyser flagged.
4. Short-term while mixing, Integrated at the end. Play the full song for the Integrated number, and read true peak alongside it.
5. Close the analyser. Do your final listening pass with no screen to look at.

What we think

A spectrum analyser is the cheapest acoustic treatment you will ever buy: in a bad room or on headphones, it is often the only honest report on the bottom two octaves. We think every home producer should have one on the master bus, and the free options are good enough.

But the screen is also the most persuasive liar in the studio. The "ideal mix curve" images that circulate online are meaningless unless you know the slope, block size and averaging used to make them, and we have rarely seen one that says. Matching a line is not the same as making a record that feels right, and a song that hits a loudness number exactly can still sound worse than one that misses it. Use meters to catch what your monitoring can't show you, compare only like with like, and when the meter and your ears disagree on something you can actually hear, trust your ears and go and find out why.