High-resolution real-time audio spectrum analyzer

An FFT spectrum analyzer has two independent frequency axes, plus a third axis in time. They are easy to mix up, and they do different things.

Frequency resolution (bin width) is how finely the analyzer can tell two close frequencies apart. It equals the sample rate divided by the FFT size. This is the axis where bin count earns its keep: more bins at the same sample rate means a narrower bin, which separates partials that would otherwise merge into one peak. Bin count alone is not the resolution; the resolution always depends on the sample rate too.

Frequency range (maximum measurable frequency) is the highest frequency the analyzer can show at all. It is the Nyquist limit, half the sample rate. Sample rate alone controls this; the bin count has nothing to do with it. In practice this range is bounded again by the microphone, which is usually the real limit (see below).

Time resolution is the third axis: how often the spectrum updates and how long an analysis window the FFT looks at. Frequency resolution and time resolution trade off against each other (narrower bins need a longer time window), so VoceVista exposes both as independent controls so you can pick the balance that fits the task.

Put together: bin count plus sample rate set the resolution, sample rate alone sets the range, and the analysis window sets how that resolution is spent against time. At the Pro edition's maximum of around a million bins, here is what resolution and range come out to at common sample rates:

Holding the sample rate constant at 44.1 kHz (the common rate for vocal analysis and CD-quality audio), here is the bin width at each available FFT size. Sizes up to 8,192 pts (8 kpts, 2¹³) are available in all editions; sizes from 16,384 pts (16 kpts, 2¹⁴) up to 1,048,576 pts (1 Mpts, 2²⁰) require the Pro edition.

A note on range. The microphone, not the analyzer, sets the real-world upper limit. Ordinary microphones roll off at around 20 kHz, which is well covered by a 44.1 or 48 kHz sample rate. Professional measurement microphones extend up to roughly 30 or 50 kHz; anything beyond that needs laboratory-grade instrumentation or ultrasonic transducers. So while the software will happily run at 192 kHz on the right audio interface, the limit you usually hit first is the microphone. For specific picks across the price range, see our microphone recommendations.

A note on resolution. For comparison, the default setting for singing analysis is 8,192 pts (8 kpts, 2¹³) at 44.1 kHz, also the maximum in the Basic and Live editions. That gives about 5.4 Hz per bin, plenty for vocal pedagogy where harmonics are spaced apart by 100 Hz or more, but not enough to separate the closely spaced inharmonic partials of bowls and gongs in their low register. At the same sample rate, the Pro edition's 1,048,576 pts (1 Mpts, 2²⁰) gives about 0.042 Hz per bin: 128 times finer.

Dynamic range, the spread between the loudest and quietest signals the analyzer can distinguish, has several aspects that stack along the audio chain. VoceVista records at up to 32-bit floating-point internally, which is wide enough that the recording stage itself imposes no practical limit; imported FLAC files are supported at 24-bit and 32-bit (nominally 144 dB and beyond). On the analysis side, the FFT itself has a usable dynamic range of about 100 dB in practice, set by the sidelobe attenuation of the window function; the 12 selectable windows listed in the specs table trade this dynamic range against main-lobe width. On live signals the limit is usually reached well before any of that: a good audio interface gives 100 to 120 dB of dynamic range, and the microphone plus room noise typically eat another 20 dB on top of that. So the overall dynamic range you see is the minimum of these stages, and in practice the hardware sets the ceiling, not the analyzer.