Hello. Can somebody confirm if playing a song with all stems “on” is:
The reason I am asking is because I feel like the sound quality is not 1:1 compared to the original files played on a computer or how the files were playing on SC6000m before stems.
Thank you.
I plays the original file.
I agree. The quality of the stem split file has lost quality compared to original file. This is evident because if your find a track with just vocals and you use the fader to switch from original to stem but also select only vocals on stem track (even though it is just a vocal perfomance), then yes, there is a lower compression file being played by the stem track. Which also means all stem tracks are slightly lower in quality overall than the original file. It isn’t always obvious on every track, but I find that vocals are the main culperate of being obvious lower quality.
I just tested this. 2 tracks were loaded on 2 different channels with matched settings. One being an original FLAC file, the other one the stem processed version of exactly the same flac.
I first played a loop of the original file. Next, the same loop of the stem processed version. This is a spectogram from Izotope RX:
There is no loss in the high end. That would be a huge indicator that the audio is exactly the same as the source file, when stem processing and playing with all layers enabled.
If you still think you do have different results, try to test it the same way. Play a loop from an unprocessed track with a fair amount of top high, then the same loop with the stem processed version with all layers enabled. Make a recording while doing this. I’m not sure whether you can record internally on the SCx000(m) players, I couldn’t find it in the manual so maybe not. In that case, use something else to do that. Preferably use the digital out for that, although an analog recording would probably reveal any differences as well when present, since nowadays even the cheapest converters have reasonably competent fidelity over a wide enough frequency range.
If you upload it, we can take a look at any alleged differences.
An even better test is to load the 2 files into a daw, make sure they’re sample accurate in sync and flip the phase on one of the 2 files. Record/bounce to disc the audio of both routed out the same output and whatever is recorded is the difference of the 2.
That’s also a very good point, yes. It is important for such a test to make sure the keylock is disabled though, because it changes the phase not 100% consistent on both tracks. Also, the analyzing has to be perfectly accurate on the sample, otherwise the residue is not exactly zero. I never tested whether that is the case with 2 tracks where one has an inverted polarity, so I can’t say anything about whether the inverted result will be analyzed sample accurate.
Although, don’t we already know that a wav file analyzed to be stems is rendered in a lossy format? I seem to recall seeing it posted in an earlier stems thread.
Yeah, I’m pretty sure I’ve read that too. I think the original source file is used when all stems are unmuted, but it switches to the lossy stems when muting one or more stems. The 4.2 faq says:
How much space will rendering stems use? When rendering stems, you are creating additional stem files for playback of four parts of audio separation.
Since it mentions additional stem files, I suppose that confirms it. It might be possible to measure this when muting a bass stem, and then check what happens on the top end of the audio.
Audio quality might change due to stem separation. If a user perceives lower quality, it could very well be that this is because of the artifacts caused by the stem separation algorithm, rather than it being a lossy stream. Decent lossy files are usually hard to distinguish from their lossless sources.
This would be a method to test what causes the differences: instant double a stem processed track to the opposite deck. Mute different stems on both sides, make sure that all 4 layers are sent to the master once. Compare the sum of this to the original lossless file. It would be easier with 4 channel consoles to do this, because it allows for an extra layer for the lossless source. If there is indeed a noticeable difference between the original and the sum of the 2 other channels, the cause might be indeed that the stem channels are lossy. I would expect that it sounds very transparent though.
Thank you for all the responses here. I feel like these more scientific approaches regarding this subject is great and I really appreciate the time used to test this. I have concluded that my perceived quality is most likely a placebo.
