Does anyone know when (where in the processing chain) Cubase 5 resamples when exporting at a sample rate different than the session (recorded) sample rate?
I.E. if a plugin with dither is inserted in slot 7 or 8 of the main (master - mix) out and export is set to create a file at a different sample rate (i.e. 48kHz > 44.1kHz), does Cubase resample, then dither, or does it dither, then resample?
- Mix > Master Out EQ > Master Out Insert Slots 1 - 6 > Master Out Fader > Resample > Insert Slots 7 - 8 (dither plugin)
- Mix > Master Out EQ > Master Out Insert Slots 1 - 6 > Master Out Fader > Insert Slots 7 - 8 (dither plugin) > Resample
This is an important distinction, as the resampling caculations would negate the effectiveness of the dithering.
The first scenario would make it possible to export a higher resolution recording (48 kHz or higher; 24 bit or higher) to CD standard (44.1 kHz - 16 bit) in one pass with proper dithering; the second would require exporting to a high bit depth first (24 or 32 bit f.p.) at 44.1kHz, then dithering while exporting to a 16 bit file.
Dither’s only required when bit depth is involved, not samplerate. It always goes on the Master Out channel on Insert slot 7 or 8 so it’s post fader.
Why would resampling make any difference to dither noise? would it not be a function of the word size that matters?
Resampling is a calculation intensive process; like any process, wordlength is expanded to the resolution of the platform (Cubase is at 32 bit floating point) or the plugin. Therefore, if dither is applied before resampling, you would still have truncation after the resampling. Think about it… As Bob Katz states more than once in his excellent book ‘Mastering Audio: The Art and Science’: Resampling should be the next to last process; dither to 16 bit should be the last process (paraphrased). And: Dithering to 16 bit should occur one time, as the last stage of processing.
I never make final mixes this way (resampling and dithering straight out of Cubase in one go); I always export to a 32 bit floating point file at the sample rate the session was recorded at; then bring the file into Wavelab for resampling, then dither (Incidentally,this is where you can see that resampling expands wordlength - take a 16 - 24 bit file, run it through a resampling plugin and look at the bit-meter!).
Occasionally though, I take the ‘quick and dirty’ route to make a rough mix CD for clients by exporting to the CD standard of 16/44.1 straight out of Cubase (with dither in insert 8 of the main out bus) - to save time. The results never sound as good as the Wavelab method, but I have not checked to hear if truncation is part of the lesser quality sound…
I was hoping someone had thought of this and checked to see if the low level sections of a mix made this way exhibited truncation or not.
Plenty of detailed dither discussions at the old forum.
Dither should be added before any quantization or re-quantization process, in order to de-correlate the quantization noise with the input signal and to prevent non-linear behavior (distortion); the lesser the bit depth, the greater the dither must be.
The act of resampling will not affect the noise added, a random noise over the bottom three bits of a 24bit word @ 96KHz will still ocupy the same bits at 44.1KHz