Unusual outcome with timestretch … Gain increase (2-3 dB)

[Sundown]

Hey all,

 

I’m trying to finish a track and as I listened to it, I felt like it needed another 2 BPM to hit the sweet spot. So to test my hypothesis I brought a mastered/finalized demo into Wavelab 11 and used the factory timestretch function to raise the song from 102 BPM to 104 BPM. I figured if I’m right, it’s not too hard to go back into Cubase and re-capture the parts at a slightly higher tempo (it’s all MIDI-based performances).

 

Interestingly when I did the operation in Wavelab, it raised the overall gain by 2 to 3 dB. I was below 0 dBFS with the original demo (about -0.3), and after the timestretch operation it jumped 2+ dB and significantly exceeded 0 dBFS.

 

I figured it was operator error and I must have a hidden plugin on the buss, but I cleared the Master section, reboot the software, and the same outcome occurred.

 

This track starts with a highly resonant synth bass part (from HALion 7) and my hypothesis is that the time compression stacked slightly more energy into less time, resulting in a gain change. I tried the operation again but I lowered the gain by 3 dB before the time stretch and it worked OK. And it did confirm that I needed that additional 2 BPM to satisfy my wants. 😁

 

It’s an interesting phenomenon, but I don’t think it’s a bug. I think it’s a genuine artifact of the time compression process when using a highly resonant sound.

 

Todd

[Anderton]

This is something I've talked about before in the context of looping. If a loop isn't playing at its native tempo, then stretching comes into play. The gain boost occurs because time stretching involves cutting up the audio and pasting it back together again. The pasting process involves crossfading, so if the audio adds up instead of subtracts, there's a gain boost. This doesn't happen only with resonant sounds, although that certainly is a prime condition for this kind of boosting to occur.

 

The amount of boost varies with the tempo. So, if you're creating a loop for commercial release, you can't just assume that accommodating peaks for one tempo will accommodate peaks for other tempos. I found that dropping peaks by -3 to -6 dB could handle pretty much any tempo change.

 

Hmmm...one more thing to include in the next update of "The Musician's Audio Handbook"