Les Bases de l'Analyse en Temps Réel ( RTA )

Introduction au Concept de Mesure > Les Bases de l'Analyse en Temps Réel ( RTA )

A real-time spectrum analyzer (RTA) is the single most common and widely used analysis tool in audio and acoustics, so we will presume that most users have at least a basic familiarity with what an RTA does. But there are still a couple of points of interest about the RTA in smart that merit pointing out. For one thing, smarts RTA's multichannel. You can analyze as many signals simultaneously as you have inputs to capture and/or computer power to process. There is no built-in limit. You can also perform real-time averages of spectral data from multiple microphones or other signal sources and display the results as a single spectrum measurement. For details on how to set up spectrum measurements and averages, please refer to the topics on the Group Manager and Configuring Spectrum Measurements.

The RTA in Smaart is FFT-based, which historically has been something of a double-edged sword. The banding accuracy of hardware-based RTA's tends to be limited by the design constraints of analog or digital band pass filters. In a discrete Fourier transform (DFT or FFT) each frequency data point or "bin" is essentially a very narrow bandpass filter function. When using larger FFT sizes where the frequency bands are tightly spaced, if the FFT data is properly aggregated and allocated the effective filter functions for each fractional octave band can approach ideally rectangular. Doing this rigorously and accurately is a nontrivial undertaking however, and doing it efficiently in real time is more difficult still. Quite frankly, this is something that a lot of FFT-based RTAs simply don't do very well, but Smaart is exceptional in this respect. The fractional-octave banding algorithm in Smaart 7 Is actually one of the most frighteningly complex sections of code in the entire program and relies on performance optimizations developed and refined over a period of several years, but it works quite well.

The frequency resolution of an FFT is a function of the sampling rate and FFT size used to record and process incoming audio signals. The display resolution of the RTA is independent of the FFT frequency resolution – all fractional octave band options are available for even the smallest FFT sizes, where individual FFT bins may be spaced an octave or more apart in some cases. But it's still critical to use large enough FFTs to get good base resolution at the lowest frequencies you care about.

The trade-off in using a larger FFT size is that finer frequency resolution means coarser time resolution but Smaart gets around this limitation to some extent by performing larger FFTs using overlapping time domain data. Even at 32k, the largest FFT size smart currently supports for real-time frequency domain analysis, you will still get up to 24 updates per second even though the time constant of an FFT that size is around 0.7 seconds, assuming a 48k sampling rate.