Whenever I export, I noticed that my mixes would sound noticeably worse on 16bit 44.1 wav (I work in Ableton at 48khz, 32bit float ofc).

Its like committing everything to wav (specifically at lower bitrate) made all the issues that I typically encounter during mixing become WAY more obvious. These issues would typically appear fixed inside the daw but once exported to 16bit (at 32 bit and 24 bit wav I did not notice these issues) they would seem noticeably worse. Issues I noticed after exporting include:

- phase cancellation on bass frequencies(kick and bass) becomes very noticeable. Any kick and bass overlap hugely decreases the power of the track
- mid frequency mud becomes way more apparent.
- stereo image placement becoming clearer (and any overlapping placement or bad placement becomes more obvious - the may be subliminal but pretty sure its related to point 1 - phase cancellation)
- issues with frequency balance becoming slightly more obvious? (if I mixed it dark, it becomes very dark upon export)

sorry to say but it feels like working in a 32 bit floating daw is all LIES!

jk haha since I noticed this issue I've gotten better at mixing and its not so much of a problem, but has anyone else had this issue?

I thought I was going crazy but then I actively started looking for it and low and behold.. I exported a track to 32 and 24 bit, everything was fine. then dropped it to 16bit and poof! a few issues I had never heard before where now blatantly obvious.

Am I crazy?

 

32-Bit Float Files Explained --> www.sounddevices.com/32-bit-float-files-explained/

 

Doesn't the bitrate just deals with your dynamic range and also the loudness of your mix??? , so I don't see why it would mess up the quality of your mix. example 1 bit is equal to 6db so 16bit is equal to 16x6db and 24bit is equal to 24x6db. Sample rate would make more of a difference in your mix since we're dealing with the quality of audio per second. So if you're having certain issues at 16bit then that means they already existed at 32bit anyway, just not that obvious because of a higher dynamic range. I could be wrong.

 

I would more suspect the sample rate conversion from 48 to 44k.
Also make sure you have no digital overs in your 32 float files. !6 bit will clip them hard.

 

Bouncing between sample rates like this you may encounter Dither up, Dither down, Truncation, and/or Retardation.

 

32bit floating is basically 24bit fixed point plus 8bit floating. So 48/32 -> 44.1/16 is samplerate conversion and dithering at the same time. If your DAW doesn't have a special setting to fine-tune these I can't recommend doing this at all.
Use a separate tool like Voxengo's r8brain PRO.

And, of course, you need to check beforehand if a plugin or a channel is clipping.

 

Wow, this just raised a very interesting point that may be the cause of my frustration!

Initially I had thought that the extra headroom afforded by 32bit float should have been completely unrelated to these sort of mix issues. right? I didn't check but my guess is that the conversion from 48 too 44.1 should be minor right? Has anyone been able to notice much difference? except maybe in high end/aliasing.

Thanks! I will confirm this with a blind test when I get some time, doing each conversion separately (32-24-16 and 48-44.1, all separate and in different combinations - as per). This should identify the issue BUT the point is that there is a noticeable difference, regardless of where the issue takes place (wether that's doing 32-16 and 48-41 at the wrong settings OR doing multiple conversions simultaneously). I normally do this one shot in Ableton with triangular dither and that's pretty much it.

PLUS I also noticed the issue gets compounded when converting to mp3.

now I can understand quality loss, but my mix translating worse and worse? why?

I suspect the issue may be what Secretworld and Havana alluded to above, if lowering bitrate reduces headroom, then wouldn't it actually incur some sort of compression? if you raise the noise floor, wouldn't that compress the relative volumes of all elements? bringing the instruments closer to digital zero and in some cases, actually clipping? would clipping cause my mix to sound worse? I thought it would just sound like digital distortion?

now that I think about it, the 16bit wav actually sounds similar to what happens when I drive the mix into a clipper or limiter too hard, some element pops out in the mix, another is pushed back, like the kick and snare become more prominent and the vocal is pushed back. but there definitely is no audible clipping or distortion. whatever happens is happening subtly and quite transparently

sorry for the monologue, I'm organising my thoughts here but clearly no one has encountered this issue? I haven't heard it anywhere. thought I'd mention it. Will confirm after testing no avenger and the rests advice to convert with proper form haha and see if that fixes the issue. thanks for input!

 

If your 16 bit wav file sounds bad, it is a warning signal, you should check your work and correct mistakes.
The quality of the source material should improve. A Wav file versus an MP3 file always sounds better.
When a Wav becomes an MP3, there is already degradation.
A Wav sounds clearer and purer compared to an MP3.

I would set everything in 24 bit (because of the generous headroom) and 48 kHz and leave these values unchanged.

Mixing and mastering properly, and in a less than ideal recording space, is not easy. I don't think 16 bit is causing the
difficulties, but your mixing technique is not perfect yet and the use of effects should be practiced and improved.

 

Thank you! I agree, this is much more apparent on any mix that I rushed.

if anything, it's kind of helpful to know that if I convert down that these issues will become more apparent cuz now I have another method of referencing. My initial thoughts were varying from "clearly I'm missing something here" to "maybe I'm just crazy"

also helpful to know that theres more to bit reduction and downsampling. I knew there was a quality reduction but had no idea the extent and what it would affect. I was expecting some loss of high end and some worsened true peak clipping on mp3, and maybe a little increase in noise floor.

just glad I was able to clear that up! Will do @BEAT16, great motivation to improve my mixing!

 

Overlooking is not enough, if mixing and mastering were so easy, we would bring our music into the charts ourselves without the sound engineers and studios. A trained sound engineer has to know a lot of math and remember a lot of parameters - plus years of listening. we call that professional practice. And with luck and only with a lot of talent, the sound engineer will become famous and be an asset for film and radio.

I think you have already come very far and you enjoy it. You want it because your heart (soul) burns for the music. Best conditions.
No you are not crazy. Just take this crazy out of the dictionary permanently. Take "I am not a robot, but a human being".

 

ALL modern audio converters are 24 bits LINEAR
So basically, when you listen back to your 32 bits float or 16bits linear export ... it is converted into 24 bits LINEAR OUTPUT

That's a fact.

Now the question is : HOW 16bits linear and 32bits float are CONVERTED BEFORE output ?
Player itself ?
Soundcard drivers ?
Soundcard hardware ?

I don't know ... and i don't care
Already lost too many time

 

Not necessarily. Very subtle digital clipping can sound like the most transparent limiter ever.

I'm more intrigued about this...

Bit reduction shouldn't be that noticeaeble if noticeable at all. BUT, differences in how things render can be a real thing, think about round robin on sampled instruments or the little differences between takes on stuff like virtual synthesizers. Most of the time these things are accounted for but sometimes it's just not. It could be that you kick is a sample and it's the same all the time but the bass is being modulated by something that changes everytime (like a free running lfo or something) and it's not resetting between exports, so you're getting noticeable cancellations.

I'd first try rendering the kick and bass individual tracks (within the project) and then export with different export settings, see if the problem is still there.

Here's an example. 16 and 24bit files of the same project, nothing changed within the project between exports. Kick is a static sample, synth is an analog emulation with some non-resetting stuff. Null them out and see what happens.

edit: just tried to explain a little bit better.

 

Last edited: Aug 4, 2022

This! I appreciate it and honestly enjoy the process of discovering these things more sometimes than the ever evolving struggle of making a living from it. I hope to one day! but until then I'll enjoy the technicalities and adventures in discovering new things like this every day thanks to the community!

Interesting! If I may ask, do you mean "Linear" asin "fixed point"? as per @BEAT16 article on 32bit float? will read on this further, thanks!

now that I know "more or less" where these issues stem from and how each conversion affects the final product I can dig a bit further and then try to mitigate these problems. haha I agree: tell me what I need to know to make better music, beyond that, couldn't be bothered!

Woah! this is crazy, I nulled them and the kick disappears entirely but somehow the bass is completely unaffected? not only is it not cancelling but there aren't even any phasey artefacts. its like its a totally different synth! jeez I need to mix in audio.

gosh thinking about it now I actually MOSTLY mix with live midi synths instead of audio. I'm just introducing more potential problems along the way. this is major motivation to mix everything in audio in another session entirely.. I seriously need to rethink my workflow. I never took this seriously enough. Thanks guys!

 

No, the bits are reduced at the bottom, not at the top (means below -96dB FS for 16bit).

A very clear hint to 16bit clipping. Means, some of your channels have more than 0dB FS. Not a problem as long as you stay within 32FP (although even some plugins can produce digital clipping when they don't work with FP) but when you export these with 16bit...

If you don't share your thoughts, how could we help?

 

Aha! first I've heard of that.

thank you! I will watch out for this. Especially during exporting stems to my new "pure audio" mixing template. Will pay special attention to metering and gainstaging now. good point.

edit: i used to just go willy nilly in the red until the master and then pull it back before limiting. will be sure to be a bit more thoughtful from now on

 

That's why I've set my DAW to pre-fader metering and peak hold. This way I can see any time if any channels are clipping, not matter the fader position.

 

nice! apparently ableton only has post fader metering (cries in "live 11") but I will be extra watchful of those little gain stage bars in between plugins in future!

 

Yeah a raw session will always sound better then a mixdown, but it gets a lot less noticeable when you get good at mixing.
Raw session always has a bigger headroom, tiny bit more low end and nicer top end this is why using oversampling can save your mix it becomes more apparent after a mixdown.

 

Why do you want to export 44/16bit? as an editing format it is outdated, it is not even a distribution format. I mean, what use do you make of it?

 

Which version of Ableton Live? Test/try to use a better SRC (sample rate converter).

http://src.infinitewave.ca/

 

Last edited: Aug 4, 2022