Spectrum Analyzers (RTA)

A level meter tells you how much sound there is. A spectrum analyzer tells you what the sound is made of — bass hum, voice, hiss — as a row of dancing bars, low frequencies on the left, high on the right. Once you can read one, most audio problems stop being mysteries and become things you can point at.

Seeing sound by frequency

Every sound your microphone picks up is a single wiggling pressure waveform — voice, fan, hum and keyboard clatter all summed into one line. A spectrum analyzer undoes that summing. It splits the signal into frequency bands and displays the energy in each one, updating many times a second. The display style with a fixed set of vertical bars is traditionally called an RTA — a real-time analyzer — and it's the audio equivalent of a blood panel: instead of one aggregate number, you see where the energy actually lives.

The intellectual foundation is much older than audio electronics. In his 1822 treatise on heat, Joseph Fourier developed the idea that arbitrary functions can be represented as sums of sine waves — which means any waveform can, in principle, be decomposed back into the frequencies that make it up. Everything an analyzer does is an engineering approximation of that decomposition, fast enough to watch live.

Filter banks, room tuning, and the FFT

The first practical audio analyzers didn't compute anything — they were filter banks: a rack of parallel analog bandpass filters, each feeding its own detector and display element, typically spaced one-third of an octave apart to roughly match the ear's resolution. In the 1960s, hardware third-octave analyzers became the workhorse of room tuning: sound contractors would play pink noise through a venue's PA, watch which bands the room and speakers exaggerated, and pull those bands down on a matching third-octave equalizer. Don Davis's work at Altec Lansing on equalizing installed sound systems in the late 1960s helped turn this analyzer-plus-EQ method into standard practice, and the analyzer/equalizer pair remained a fixture of live sound for decades.

Software replaced the filter rack thanks to the fast Fourier transform. The FFT algorithm published by James Cooley and John Tukey in 1965 made Fourier analysis computationally cheap, and once general-purpose processors caught up, an analyzer became a few lines of DSP: transform a block of samples, group the resulting frequency bins into display bands, smooth over time, draw. That's what runs in every modern audio tool — the same measurement that once needed a rack of filters, now essentially free. The bar-graph RTA presentation survived the transition intact, partly out of habit and partly because it remains genuinely easier to read at a glance than a raw high-resolution FFT curve: a manageable number of bands, each answering one question.

Why the bars are spaced logarithmically

Look at any RTA and you'll notice the frequency axis isn't linear: the distance from 100 Hz to 200 Hz equals the distance from 5 kHz to 10 kHz. That's logarithmic spacing, and it matches how hearing works. Perceptually, each doubling of frequency is one octave — the same musical interval whether you double 100 Hz or 5,000 Hz. A linear axis would cram nearly all of the voice's character into a sliver on the left and waste most of the display on the top octave. Log spacing gives each octave equal screen space, so equal visual distances correspond to roughly equal perceptual distances. It's the same reason equalizers lay out their controls logarithmically: the display and the tool that acts on it share the ear's coordinate system.

How to read one

An RTA rewards a few minutes of deliberate looking. Watch it in silence first, then while you speak, and the patterns become obvious:

• A voice shows a strong foundation at the fundamental — roughly 85–180 Hz for typical male voices, 165–255 Hz for typical female voices — with harmonics stacking above it. The formants that make vowels intelligible put energy through the low-mids and midrange, and sibilance ("s", "sh") flicks the bands up around 4–8 kHz. A healthy voice is a moving mountain range, not a flat line.
• Mains hum is a steady spike in the leftmost bars — 50 Hz or 60 Hz depending on your country's grid, often with harmonics above it. It doesn't move when you stop talking, which is exactly how you catch it. (More in mains hum.)
• A fan or air conditioner looks completely different: not a spike but a raised broadband floor, a carpet of low-level energy smeared across many bands that sits there whether or not you speak.

That last distinction — narrow spike versus broad carpet — is the most useful diagnostic the display offers, because the two problems have different cures. A hum spike wants a targeted filter at one frequency; a broadband floor wants noise reduction that works across the whole spectrum. Your ear tells you "something sounds off"; the analyzer tells you which of the two it is, and where.

Analyzers for voice and calls

On a call chain, an analyzer earns its place in three ways. First, diagnosis before the meeting: thirty seconds of watching the display in a "quiet" room reveals the hum, the fan and the neighbor's compressor that your brain long ago tuned out but your microphone faithfully transmits. Second, verifying processing: when a noise reducer or EQ claims to be doing something, the spectrum shows the before-and-after shape rather than asking you to trust your memory of how it sounded. Third, setting tone by eye and ear together: it's far easier to place an EQ cut when you can see which bands a boomy room is inflating. None of this replaces listening — the analyzer can't tell you whether a voice sounds good — but it turns vague impressions into frequencies you can act on.

In DeskBroadcast

DeskBroadcast includes an RTA-style analyzer with 20 log-spaced bands from 50 Hz to 15 kHz, drawn as level-colored bars so hot bands stand out at a glance. Like the level meters, it reads the processed signal — the audio your call actually receives from the virtual mic. That makes it a live window into the chain: enable spectral noise reduction and you can watch it learn your room, the broadband floor visibly dropping across the bars over the first moments. Mains hum, if present, shows up exactly where theory says — parked in the lowest bands, unmoving while you're silent. The analyzer is also part of DeskBroadcast's guided dial calibration, where seeing your voice's spectrum helps you set the processing dials against what's really coming out of your mic rather than a generic preset.