Post by tosiabunio on Aug 15, 2015 22:01:59 GMT
One of the decisive factors of buying Master Audio for my project is that it could allow me to use decibels for volumes.
Unfortunately I quickly learned that the current implementation is flawed and usable at all.
First of all, I was puzzled by the scale of sliders: -140dB do 0dB. In real life scenarios (and I've been doing game's audio for 15 years now) I rarely decrease volume of my individual effects to lower than -30dB.
-72dB is perceived as silence by most of the listeners - so wasting half of the volume scale for silence is even worse than using float 0.0-1.0 for volume.
But I fiddled with the volume sliders and numeric values to quickly find that the numbers shown have no real relation to what they represent (e.g. -6dB should halve the volume). I spent some time to setup audio metering pipeline to measure what I heard with my ears.
This is what I've found (the source was 1kHz sine wave at 0dB). Numbers are - Master Level volume : actual output volume as measured:
-1dB : -0.4dB
-3dB : -1.3dB
-6dB : -2.6dB
-10dB : -4.3dB
-12dB : -5.2dB
-20dB : -8.7dB
-24dB : -10.4dB
-30dB : -13dB
-72dB : -31.3dB
-90dB : -39dB
The volume changes are off but still linear up to -24dB (or something) and then calculations are way off for lower volumes. With master level at -90dB I can still hear the signal - it's just -39dB at the output - but it should be silent.
Anyway, what went wrong with those two functions?
linear-to-db(x) = log10(x) * 20
db-to-linear(x) = 10^(x / 20)
Unfortunately I quickly learned that the current implementation is flawed and usable at all.
First of all, I was puzzled by the scale of sliders: -140dB do 0dB. In real life scenarios (and I've been doing game's audio for 15 years now) I rarely decrease volume of my individual effects to lower than -30dB.
-72dB is perceived as silence by most of the listeners - so wasting half of the volume scale for silence is even worse than using float 0.0-1.0 for volume.
But I fiddled with the volume sliders and numeric values to quickly find that the numbers shown have no real relation to what they represent (e.g. -6dB should halve the volume). I spent some time to setup audio metering pipeline to measure what I heard with my ears.
This is what I've found (the source was 1kHz sine wave at 0dB). Numbers are - Master Level volume : actual output volume as measured:
-1dB : -0.4dB
-3dB : -1.3dB
-6dB : -2.6dB
-10dB : -4.3dB
-12dB : -5.2dB
-20dB : -8.7dB
-24dB : -10.4dB
-30dB : -13dB
-72dB : -31.3dB
-90dB : -39dB
The volume changes are off but still linear up to -24dB (or something) and then calculations are way off for lower volumes. With master level at -90dB I can still hear the signal - it's just -39dB at the output - but it should be silent.
Anyway, what went wrong with those two functions?
linear-to-db(x) = log10(x) * 20
db-to-linear(x) = 10^(x / 20)
