Logic Pro 8 Dither Notes                         Back to fader8's audio blog Index

These are some resulting bounces in done in Logic Pro 8.0.2 to investigate a dithering issue that I was curious about.

Thanks to Kees De Visser for providing the useful test file.

The signal encompasses three sine sweeps. The signal slowly decreases in level from -24 to -144 dBfs. 

Plot 1

Here is the spectragram for the 24 bit undithered file. All 3 sweeps show clearly down to -144.

(Time is from left to right. Note that the colour/dB scaling reflects a very large FFT size.))

Plot 2

This is the same file, but truncated in SampleManager to 16 bit without dither. As expected, as the signal drops below the 16th bit, the signal disappears. (The third sweep is completely gone.)

Plot 3

This is attempting the same thing in Logic Pro 8, bounce at unity to truncate to 16 bit, no dither. There are clearly artifacts created in this process. Some kind of noise is being blended into this process which provides a kind of "false" dither allowing some of the low level (sub 16 bit) audio to get through. Source is unknown and I don't know yet how this impacts a dithered bounce.

Plot 4

Here's a Logic Pro 8 bounce at unity, truncate to 16 bit with Pow-r#1 dithering. This looks normal in that at very low levels the dither noise would mask the signal.

Plot 5

Here I'm switching gears and trying out the Waves L2 plug-in on the output channel. It's set to 16 bit quantize, Type 1 dither, no noise shaping.

This plot is made realtime with the audio streaming into the analyzer directly from Logic. This is not a file bounce but represents what you would hear while mixing/mastering into and through the plug-in. (I've lengthened the time axis to show more detail on these.) Note the artifacts and amplitude nonlinearities before and during the third sweep. Also, the signal goes below the 16 bit threshold sooner, which leads me to think there is a general miscalculation with low level signals during live monitoring.

Plot 6

Here is the exact same settings, i.e. Waves L2 plug-in on the output channel. It's set to 16 bit quantize, Type 1 dither, no noise shaping.

But this is an analysis of the bounced file, not the playback stream from Logic. Logic's bounce settings were set to Real Time bounce, 16 bit, no dither. This plot looks normal and seems to imply that the condition above, while present during monitoring, does not manifest itself in the actual bounced file.

This prompted the Realtime vs Bounced Quantize and Dither Study

 Back to fader8's audio blog Index