17.10.2022, 18:47
(17.10.2022, 17:29)andreas42 schrieb: Ich "schaue" ja nicht mit einem WAV-Editor rein, sondern lese einfach die rohen SamplesGenau um letzteres geht es, die meisten Editoren tun nichts anderes.
Das ist aber bei hohen Frequenzen nicht richtig. Jedenfalls gibt es ein paar problematische Frequenzen, die man meiden sollte, zB Fsample/3, also 16 kHz bei 48 kSps.
Wenn du über den Effektivwert gehst, hast du bei Frequenzen oberhalb Fsample/4 "Schwein", weil da das Signal bei vorschriftsmäßigem Gebrauch nur ein Sinus sein kann/darf. Da kannst du getrost den Effektivwert aus der Summe der Quadrate schätzen (auch wenn du nur 3 pro Periode hast). Der Spitzenwert ist dann Wurzel(2)*Effektivwert.
Einfacher ist es, sich um eine kleine Offset-Frequenz neben die Problemfälle zu setzen, weil dann die "Schwebung" zu Fsample/N durch einfachen Blick auf die Hüllkurve idR den Scheitelwert offenbart.
Dein Beispiel war wohl mit LPR90 durchgeführt.
Es gibt sicher Bänder, die leichter in die Sättigung zu fahren sind.
MfG Kai