Logic Pro 11 - Stem Splitter

[keith.rogers]

Well, obviously not a lot of Logic Pro users here, but besides the addition/expansion of their "players" for drums, keys and bass in the latest version, there's a "Stem Splitter" feature that's kind of interesting. NB: This feature requires Macs with Apple Silicon - it will be disabled on Intel-based systems.

At first, I was wondering what I might ever use it for, but on another forum someone mentioned that it had been used to fix some bleed problems. I don't know if "fix" was the word used, but often that's the case. So, I tried it on a mix I did (video posted in MP3 sub-forum). Just a simple 2 guitars + vocal, 3 mics, and a little bit of everything everywhere. I took the same Logic Project, did the stem-splitting on each original track, splitting Vocal and Other (bass and drums are the other specific things you can try). This results in track stacks (a submix thing in Logic) with your stems, and in the guitar ones I muted the vocal stem, and in the vocal stack I muted the "other" stem. I.e., I removed the obvious bleed. I had to apply pan and reverb send to the stack (submix) tracks, though I probably could pull the stems out, those items in the original tracks were not carried over. Volume automation on the original tracks was carried over. And, I had to lower the threshold on the Ozone Maximizer on the stems mix because some [bleed] content was now removed/muted, but just about 0.2dB to lift the stems mix up to the same LUFS as the original. Here are the two mixes - see what you think. (I didn't label them, so picking the stems mix is up to you, but might be pretty obvious.)

SoundCloud - Hear the world’s sounds

Explore the largest community of artists, bands, podcasters and creators of music & audio

on.soundcloud.com

SoundCloud - Hear the world’s sounds

Explore the largest community of artists, bands, podcasters and creators of music & audio

on.soundcloud.com

p.s. (edit) Does this subforum really need to have "Emagic" in its name? I mean, who uses that? Wouldn't GarageBand, now running on Logic's core, be a better fit, in case someone stumbles into this forum to ask a recording question...

 

[Papanate]

I wouldn’t have thought to use StemSplitter this way - it’s okay - Ii think the Stem Splitter is B because everything is louder.

 

[keith.rogers]

Well, the LUFS for the tracks only differed by 0.2dB (LUFS) - I assumed it would not be noticeable… seems like this was not an interesting topic, or perhaps it was my content!

The stem-split remix is actually the A mix. I may go back and readjust the limiting to make them match exactly, but that’s for the entire track, so certain parts likely will still be louder/softer between them. What I noticed most was the separation between the parts. Not sure I liked it better, though it is easier to manage without the bleed. Still, I think I’ll keep working on mic placement to reduce that before resorting to this.

 

[Papanate]

Well, the LUFS for the tracks only differed by 0.2dB (LUFS) - I assumed it would not be noticeable… seems like this was not an interesting topic, or perhaps it was my content!

The stem-split remix is actually the A mix. I may go back and readjust the limiting to make them match exactly, but that’s for the entire track, so certain parts likely will still be louder/softer between them. What I noticed most was the separation between the parts. Not sure I liked it better, though it is easier to manage without the bleed. Still, I think I’ll keep working on mic placement to reduce that before resorting to this.

That’s intresting - it seems like with more sonic space of separate tracks the volume would be louder - if the tracks only differed by 0.2dB - something must be going on with the Codec or how it encodes - or maybe my systems is whonky -

 

[keith.rogers]

Well, I did record direct off of the headphone out of the Soundcloud playback and their streaming is a lot more manipulative than I thought - certainly nothing like what I'd call "HD." because the loudness difference, peaks and dynamic range were all in variance with the originals (which I downloaded off the SC site to make sure I wasn't losing my mind!). This is going to take some more experimentation, but I don't have any time to do this right now, and I want to have a "clean" track or two to compare, just to see if I can articulate what's happening, and maybe compare to YouTube streaming. Once I have that kind of set up, I'll try the stem-splitting test again, but looking like later this year .

One thing I've been ignoring, or careless about, that I noticed was that the DC offset, if non-zero on the submitted track does get set to zero on the stream, so there's some filtering happening. I've noticed (and ignored) that, as it's really tiny, like +.02dB. Is an artifact of using the Zoom F8n? Or what?? So, I'm going to start by being more particular about that on any "final" mix so it's zero when uploaded, to remove that variable.

 

[Papanate]

One thing I've been ignoring, or careless about, that I noticed was that the DC offset,

I haven’t dealt with DC Offset in 10 years - something must be off in your system if you are getting any DC Offset - are you going to tape or hardisk first?

 

[keith.rogers]

I would have to wade though a lot of audio files to see - first I thought it was only with the Zoom, because of an article I found suggesting it's (mostly?) a field recording problem, but I sure haven't looked at everything I've done direct through the interface. I'll try switching power [supplies] on the Zoom to see if that's the problem - I only run off of a battery pack when recording like the one I showed.

Just a quick look at that last project and the offset in the individual tracks is .001%, so pretty small, though it seems like by the time they're mixed, it creeps up, i.e., sort of additive. Still, TINY, in relative terms.

One article I found: https://www.creativefieldrecording.com/2022/04/06/how-to-fix-dc-offset/

I used the EQ/HPF in iZotope's Ozone to remove the offset in the mix, but there's still a difference in what record, though how much of that is introduced by my interface's headphone amp is another variable I'm not sure how to eliminate. I can play the non-streamed audio locally, record that off the headphone jack and compare it to the file to see what the headphone amp is doing. From what I saw in Julian Krause's spreadsheet, that's a weak spot in most interfaces.