Reply to thread

Message

<blockquote data-quote="Per SÃ¸vik" data-source="post: 114363" data-attributes="member: 1285">Re: 2.0Most FFT based RTA analyzers divide the spectrum into equal chunks of linear frequency, making white noise a flat line. To compensate, a slope of 3dB/octave needs to be introduced. &quot;Real&quot; hardware based analyzers used filtering to divide the spectrum into equal (1/3 octave or whatever) chunks, and measured the signal strength in each band, and software based variants will do the same. However, doing it in software means either occupying the bandwidth of a considerable number of channels in real time or time slicing to an extent where only steady state signal can be analyzed with any degree of accuracy. Thus, most RTA is FFT-based, and therefore linear and not logarithmic as we would want it to be. To get reasonable low frequency resolution without having too many bands to analyze, the spectrum is often sub-divided into chunks of 1 to 5 octaves, each chunk having its own FFT. The higher octave bands of this approach will have quicker response or better accuracy or a combination of both.Since each analyzer might treat signals slightly differently, some may have slopes by default etc. the same signal might look different in different analyzers. If it looks good in the analyzer but sounds wrong, chances are that it actually sounds wrong.</blockquote>

[QUOTE="Per SÃ¸vik, post: 114363, member: 1285"] Re: 2.0 Most FFT based RTA analyzers divide the spectrum into equal chunks of linear frequency, making white noise a flat line. To compensate, a slope of 3dB/octave needs to be introduced. "Real" hardware based analyzers used filtering to divide the spectrum into equal (1/3 octave or whatever) chunks, and measured the signal strength in each band, and software based variants will do the same. However, doing it in software means either occupying the bandwidth of a considerable number of channels in real time or time slicing to an extent where only steady state signal can be analyzed with any degree of accuracy. Thus, most RTA is FFT-based, and therefore linear and not logarithmic as we would want it to be. To get reasonable low frequency resolution without having too many bands to analyze, the spectrum is often sub-divided into chunks of 1 to 5 octaves, each chunk having its own FFT. The higher octave bands of this approach will have quicker response or better accuracy or a combination of both. Since each analyzer might treat signals slightly differently, some may have slopes by default etc. the same signal might look different in different analyzers. If it looks good in the analyzer but sounds wrong, chances are that it actually sounds wrong. [/QUOTE]

Verification