Increased gain on WAV export

Just noticed this tonight.

I rendered a WAV file out of Ableton Live to -0.3db, loaded it to Audacity, chopped dead space of the front and end and then exported as WAV.

When I load the WAV file exported out of Audacity back in to Audacity to check it and open the Amplify dialog, the amplification to take it to 0db is 0.298 which means the export to WAV changed the gain by a tiny amount.

Is this correct? I thought export to WAV would have no effect on the gain? Should be exact match?

Is that “-0.3 dB according to Ableton”?

0.298 rounded to 1 decimal place is 0.3
0.298 rounded to 2 decimal places is 0.30

Yes, assuming:

  1. Audacity’s track “Gain” and “Pan” sliders remain at zero (default settings).
  2. There has been no change to the format anywhere through the round trip.
  3. There has been no resampling (the same sample rate throughout).

Note that Audacity uses 32-bit float internally, so to avoid any change in format you would have to use 32-bit float WAV (not the more common 16-bit or 24-bit).

Audacity recognises the WAV file out of Ableton Live would need amplified by exactly 0.3db to take it to 0db in Amplify

Its 16 BIT PCM WAV I am dealing with, so is that where the tiny change comes from?

Also, what’s the effect on a 16 BIT PCM if I export as WAV from Audacity when its already had its gain set at 0db? Will this export introduce clipping then?

Because Audacity is working internally in 32-bit float format, exporting as 16-bit (less bits per sample) will add “dither”, which is a tiny amount of randomization (“noise”) to avoid quantization noise when sample values are rounded down from 32 to 16 bits.

The addition of dither can be avoided by turning off dither (set to “None”) in Preferences.
I wish there was an easier way, but here it is: Set “High-quality Sample Rate Converter > Dither” to “None”.
Turning off dither should only be done in cases where you expect the output sample values to be identical to the input sample values, (or zero / absolute silence).

Normally you would expect output sample values to be non-zero different from the input sample values due to effects / other processing in Audacity, so normally dither should be set to “Shaped” for best sound quality and lowest noise.

More information here: