Dithering raises level slightly, should this happen?

[Masselito]

I'm creating 16-bit files from my 24-bit master files, but when applying dither, the level raises slightly. My mastered files reaches a maximum of -0.2, but after conversion and dithering, the newly created 16-bit files reaches a maximum of -0.1. Without applying dithering, the 16-bit files stays at the same maximum level as the 24-bit masters (-0.2).

 

It doesn't matter if I use Logic or WaveBurner when bouncing the 16-bit files, I get the same result in both programs.

 

Is this supposed to happen, or have I done something wrong?

[fader8]

That sounds about right. Is it a problem for you?

[David Nahmani]

Since you are adding noise to your signal, it does make sense that the overall level would be higher.

[Masselito]

fader8: No, it's not a problem for me. I just thought it was odd, and started to wonder if I had done something wrong.

 

David: I thought about that too, but I thought that the added noise is so low that a 0.1 dB raise of the level actually seemed too much (the descriptions I've read about how to hear the side-effects of dithering always involves cranking up the volume a lot)

[David Nahmani]

David: I thought about that too, but I thought that the added noise is so low that a 0.1 dB raise of the level actually seemed too much (the descriptions I've read about how to hear the side-effects of dithering always involves cranking up the volume a lot)

.1 dB is the smallest volume raise a meter can show you. That means that a meter showing you a .1 dB increase does not mean that your signal has increased by .1 dB. It means your signal has increased by an amount X where 0 dB < X < 0.2 dB. That's about the only information the meter is giving you.

[fader8]

Also, the noise is very broadband, so it has a lot of spectral density. It adds up.

[lagerfeldt]

Adding dither when you bounce from 32 bit float/24 bit to 16 bit can add 0.0x dB of extra signal.

 

If you set your maximum output to -0.22 dBFS it'll take care of things. I can't remember if or where Logic's meter rounds down, i.e. allows for 0.0x over without rounding up.

 

Most 3rd party limiters allow this kind of precision, but Logic's won't.

 

If you really want to stay on the safe side, then am output ceiling of -0.32 dBFS is the best, also in terms of avoiding intersample peaks during playback/MP3 conversion. Of course you can have intersample peaks happening at much higher levels, but a ceiling of approximately -0.3 dBFS will take care of the majority.

[Masselito]

Now I get it Thanks guys.

 

EDIT: Sorry for the editing and re-editing of my reply here.