Are Rounding Errors/Dither Audible?

So I Edited some audio files while matching the project’s Bit Depth to the audio files in order to export without Dither (The only reason I chose to export this way is because I don’t like Dithering, even though it objectively improves the quality of the audio). While I know that normalizing audio does bring Rounding Errors to the audio file and therefore add a very very tiny hiss, my questions are: Are those Rounding Errors audible? Is Dither audible as well? and if so, in what scenarios will they be heard (e.g. by normal playing, or using an Amplifier/DAC and the latest Sennheiser headphones)?

What do your ears tell you? :wink:

At 8-bits or more dither, or the lack of dither isn’t normally audible. You might want to try making some 8-bit files. With 8-bits you can hear quantization noise, and I assume you can hear dither but I never tried it.

With 16-bits you might be able to hear it if you turn-up the volume enough during quiet parts. Under normal listening conditions you won’t be cranking the volume way-way-up during quiet parts and you won’t hear it.

There is no quantization noise with pure digital silence, but normally dither noise continues.

If the recording was made with a microphone, there’s a good chance that the acoustic and preamp noise is louder than dither or quantization noise.

1 Like

Interesting, thanks for the reply. It’s funny how the two options are to have hissing or… to add hissing, although Dither seems to be the way to go.

My only two reasons why I don’t like dithering (as of right now, might use it later) are because I don’t like how the audio’s spectogram looks (Which I-- I mea-- Definetly matters when listening to music) and… Somewhere on the Internet I read that FLAC Files don’t compress dithering that well (While being a much better reason than the last, I compared both Dithered and Un-Dithered audio files and both were about the same in size)

It seems the Universe is just begging me to Dither,

I think you wrote that the wrong way round Doug. I assume that you meant:

At 16-bits or more dither, or the lack of dither isn’t normally audible.

With 8-bit audio, dither can be clearly audible at normal listening levels, as can “quantization distortion” when dither is not used.

Dither is rarely used with 8-bit audio because it sounds worse than un-dithered (Audacity does not apply dither when exporting as 8-bit), so I had to use other tools to create the 8-bit dithered version:

No dither:

With dither:

1 Like

How the spectrum looks gives a good indication of why dither is used (and not used for 8-bit audio).

Without dither, reducing the bit format introduces phantom frequencies, that can be seen as horizontal lines in the “no dither” plot.

When dither is applied, it prevents the formation of those phantom frequencies, as can be seen in the lower “with dither” plot. It may appear that the noise is just “masking” the noise, but that is not actually the case - it is actually preventing the formation of those phantom (“alias”) frequencies.

If you listen very carefully to the “with dither” example, you should be able to hear “through” the dither noise, that the original 440Hz tone is still present and undistorted.

It is also obvious that for 8-bit dither, the noise level is excessive, so this is a case of the cure being worse than the ailment.

1 Like

For comparison, I have added a view of the same audio with 16-bit dither (all tracks have the same spectrogram settings for direct comparison):

1 Like

And the spectrum plots (all with the same Plot Spectrum settings):

8-bit no dither

8-bit with dither

16-bit with dither

1 Like

I see now, it seems that Dithering is the way to go. I’ll be back to Dithering, thanks for the replies. However, I set Dithering settings to “Shaped” and my edited version of my 48000 Hz 24 bits file appears without dither. Is that normal?

This topic was automatically closed after 30 days. New replies are no longer allowed.