Crest Factor
The difference between peak and average levels is commonly referred to as the “Crest Factor”. There are some well-characterized relationships between peak and average levels for specific types of audio, and variations from the norm can reveal more subtle aspects of how the audio was processed or what it was subjected to.
When the system is in Referenced Mode, it is receiving audio from which the source is one of Frontline’s Reference Audio files, and the majority of tones in these files are single sine waves. For a sine wave, the RMS level is 0.707 times the peak level, corresponding to a 3 dB difference. Therefore, any difference greater than 3 dB can indicate an abnormal increase in amplitude such as the presences of spikes or other noise, or the presence of multiple frequencies.
Note: Some test segments are intentionally dual tone segments. While Crest Factor for a single sine wave is 3 dB, for 2 sine waves of equal amplitude it is 6 dB.
When the system is in Non-Referenced Mode, there is no pre-defined relationship between peak and average levels. However, there are well-characterized relationships established for speech and audio as listed below. These numbers represent “typical” numbers, and absolute numbers depend on the original audio content.
- Speech: A typical Crest Factor is 12 dB (assuming a signal captured with no dynamic range processing algorithms such as compression or limiting).
- Music: A typical Crest Factor is 18 to 20 dB.
The significance of clipping events reported for arbitrary audio in Non-Referenced mode can be determined by assessing the Crest Factor of the original audio, and verifying the volume settings at the source to ensure sufficient headroom is provided to accommodate that ratio.