I’d advise against using white noise as a benchmark for true peak measuring. White noise is a random value each sample and it doesn’t really make sense to interpolate between values, because the values in between should be random too. It is quite different to normal waveforms!
White noise is also a full spectrum signal. Its squirting energy into the roll-off region of your decimating filters. That means you’ll get some aliased reflections into the main band, which can lead to an upwards tilt at higher frequencies.
All in all, it’s a pathological case for oversampling!
Follow-up: Convert it back from 48kHz to 192kHz. My guess is that the peaks would stay around -18.
So when we applied the filter for down-sampling we may have actually made the peaks higher!
Glad you say that. I was hoping for such an answer. I was using white noise because it has a lot of high frequency energy in it.
But what is a good test signal for testing true peak measuring?
I generally test with a set of audio samples and oscillators that I’ve collected - but for this you should focus on tracks which have been mastered for loudness. Find a section where a limiter is really being pushed.
