quality loss when rendering

[peeddrroo]

this has probably been asked a thousand times, but i just had a talk with a "pro" sound enginerd, and he told me that rendering files several times caused losses in audio quality.
i seriously doubt it, but i'd like to have some other feedback.

also: i have to render files in 16 an 24 bits to 24 in order to mix them. there's no need to dither, right?
does converting 16 to 24 add some artifacts of any kind?

thanks for the help

[locojohn]

this has probably been asked a thousand times, but i just had a talk with a "pro" sound enginerd, and he told me that rendering files several times caused losses in audio quality.
i seriously doubt it, but i'd like to have some other feedback.

also: i have to render files in 16 an 24 bits to 24 in order to mix them. there's no need to dither, right?
does converting 16 to 24 add some artifacts of any kind?

thanks for the help

Read this thread.

Rendering a track to disk, then loading it back and rendering it again any number of times should not cause any loss in sound quality as long as:

1) your samples are at project's sample rate and bit depth
2) your samples are not warped
3) you export at project's sample rate and bit depth with dithering OFF and Normalize OFF.

Converting to a higher bit depth should not create any artifacts. No dithering is required when converting to a higher bit depth, and there is no change in the audio quality. Converting to a lower bit depth does require dithering and noise shaping in most cases and does affect sound quality. For best results Ableton team recommends to perform such operations as resampling to a lower sample rate, reducing bit depth or dithering/noise shaping -- outside Live. SoX is a great open source multi-platform command line utility that can perform all these operations in good quality and under great control.

Before you want to render to WAV/AIFF you want to ask yourself what is the target format you need? Are you mastering for CD (44.1Khz/16-bit)? Do you give away your pre-master mix to a another company? Or do you just intend to create an mp3 demo? Depending on your final objective you should set properties in the Export Audio dialog.

I advise you to read Ableton Live's manual, a part where destructive and non-destructive operations are discussed.

Andrejs

[Captain Comeback]

This issue is driving me insane on my own projects.

With respect to the other thread that discusses this, it didn't necessarily help resolve the problem for me, other than the idea to totally bypass live's dithering process and instead just rip what's playing in my arrangment view to a cd via my firewire unit after I reimport it from 32bit/96khz.

The issue that I'm running into is that I'm losing A TON of low-end presence in a song after I dither it down from the 32bit/96khz original. It's essentially the same as running the tune through your dj mixer and turning the bass eq knob halfway down.

I have even tried throwing various saturators and dynamic tube effects into the track before dithering in order to try and keep hold of the low end. I'm cranking up the Base knob on saturator and the Tone knob in dynamic tube, especially. However, no matter what I do, it always ends up with the same result.....a dithered wav file at 16bit/96khz that has perfect overall volume but WEAKKKKKKKKKKKK low end.

And when I say low end, I mean the bassline notes and their resonance and not the kick. The kick is fine. It's basically everything from 75-200khz.

I've been using ableton for about 4 years now, but I'm not too advanced on the post-production aspect. I know about sample rate and bit depth and all, but I'm just at a complete loss with the loss of quality problem.

I know that most people will say "go master your tracks in another program" or "use some plugs", but I want to see if live can actually go from start to finish on a track with the components it already has. I'm very determined to make sure I can function in every aspect of making music with only using live. Call me crazy

So in so many words, is there anyone that can shed more light on this? Is the best option actually ripping your music to another source and NOT using live's dithering?

[locojohn]

Call me crazy

So in so many words, is there anyone that can shed more light on this? Is the best option actually ripping your music to another source and NOT using live's dithering?

You already know my opinion, right? I expressed it in the post above yours The above mentioned SoX tool will convert your exported audio tracks from Live to a lower sample rate and bit depth with minimum artifacts, and your final CD-ready 44/16 tracks will be impossible to differentiate from the exported original files.

Just try it!

Andrejs

[leedsquietman]

Live uses pow-r dithering which is pretty much industry standard and used by many other programs, Sonar is one which comes immediately to mind. The dithering is a barely audible process, even for a dog with super hearing. If you can hear dithering noise at around -80 to -90dB then you have immense golden ears and this is still not responsible for your bass issues.

More likely that poor mixing (especially too much range in the 20-40 Hz area which is often not heard on lower quality monitors without a sub, or headphones (which don't translate sub bass very well anyway typically), is overpowering the bass end (read up on high pass/lo cut filtering)) and lack of headroom when going into the 16 bit domain, often a sign of overcompression/volume maximizing and/or digital clipping, are causing your issues. While it's OK to MONITOR your mixes while redlining in 32 bit due to the huge amount of headroom, you have to take into account 16 bit formats for CD and mp3 have much less headroom and anything too hot will cause artifacts/clipping and affect the sound.

Also, try a different sample rate converter. DAWS sample rate converters are not very good (any of them). If you're on PC Voxengo's R8Brain (free) offers much better SRC. Ozone's 64 bit SRC is also highly recommended. Although this will typically have less impact than the previous paragraph even using Live/DAW SRC.

Anrejs is talking good sense here.

[Captain Comeback]

thanks for the tips guys

I'm definitely filtering my sub and hi-end bass sounds for sure. On my sub bass, I put the hi cutoff at anywhere from 150khz to 300khz. On my hi-end bass sound, I put the low cutoff at anywhere from 300khz to 1000khz. I have a fairly good knowledge of balancing the frequencies I would hope, but I can always use more info. Maybe I just answered my own question with the above statement, but do you think it's because I put my sub bass cutoff so low that I can't hear it on the final exported product?

@ andrejs - I will check our the sox tonight and thanks for the help

[cutoutcollective]

http://www.youtube.com/watch?v=BYTlN6wjcvQ

excellent talk on a lot of the myths that surround dithering, etc... worth watching.

It covers a lot of perceived differences in sound quality - to cut an hour long lecture short - if you want to hear a difference, you'll hear it.

[auditory canvas]

thanks for the tips guys

I'm definitely filtering my sub and hi-end bass sounds for sure. On my sub bass, I put the hi cutoff at anywhere from 150khz to 300khz. On my hi-end bass sound, I put the low cutoff at anywhere from 300khz to 1000khz. I have a fairly good knowledge of balancing the frequencies I would hope, but I can always use more info. Maybe I just answered my own question with the above statement, but do you think it's because I put my sub bass cutoff so low that I can't hear it on the final exported product?

@ andrejs - I will check our the sox tonight and thanks for the help

While that will help keep your mix clean and seperated, that isn't going to help the issue of the low end below 40hz taking up too much energy and causing you issues on mixdown.

Try rolling everthing off from 30hz down, that'll help stop your bass from draining all the headroom in the mix, and should result in a cleaner mixdown.

[cutoutcollective]

This thread got me thinking though - maybe i'm in the camp of people who want to not hear a difference and so don't... maybe I've been wrong about live all along and my ears are shit... not entirely impossible given the amount of time i spend in loud venues.

so... simple test to solve the question... (SPOILER FOR THOSE WHO DON'T FEEL LIKE READING MY LONG RANT: Dither induces noise at around -80 to -90dB - your computer fan is probably making more noise right now. Sort out your mix and stop worrying about dither)

I went into Live and loaded up a Justice song. I then duplicated that track 24 more times and routed the audio into another audio track with a utility on it to gain down the output before the master. So basically all I had was 25 tracks going into a 26th track with -20db of gain on that track so that the master didn't clip. (side note - to set that utility level just make just that your master level is always less than or equal to the level on each track... that way you'll know that the master isn't going to clip)

I then rendered the track with the various dither settings (including "none"). I then imported the rendered tracks back into the same project (with warping off obviously) and put them on their own muted tracks. On each track I put a utility to phase invert the audio (http://lmgtfy.com/?q=phase+cancellation) I then played back the project and un-muted the various tracks. Of course each of the dithering algorithms completely phase canceled with the tracks when listening on my laptop speakers and it was only when I got out my head phone that I could hear the -90dB noise that was not canceled. Watch that video I posted above and see if you can hear the difference that -90dB of noise makes in a track. I guarantee that it is not going to cause you to lose "A TON of low-end presence". Especially if you're using pow-r - which according to live's spectrum effect resulted in extremely low noise (ie noise below the noise floor of your sound card - no matter how expensive it is) and only started to get into the -70dB to -80dB range in the 16kHz plus region (turn operator into fixed tone mode and see if you can hear 0dB at 16kHz - its possible but bearly and not particularly pleasant. Now drop the level to -20dB... still hear it? you've got some seriously loud monitors... drop it another 40dB and you're getting to the noise produced by pow-R).

Now the more observant of you may have noticed that I was summing 16-bit audio (I don't have Justice in 32-bit) using live's 32-bit engine (which, incidentally is a complete waste - there is no need to go above 24-bit... but that's a topic for another day) and then dithering down to 16 bit again... So not exactly answering the question.

Hence, another test. This time I made 40 instances of operator (my computer didn't like this much) each playing 5 note of a saw wave so as to cover the whole audio spectrum and did the same thing. Again I rendered the different dithering algorithms and for comparison, a 32-bit, undithered file. This when phase inverted in the project gave -inf reading, as expected and the other dithering algorithms performed similarly well. None and triangle at around -90dB but uniformly across the spectrum while pow-r was around -70dB but centered in the barely audible to inaudible range and very little noise elsewhere...

Ok, I'm bored of this rant and have satisfied myself... if you want to test this for yourself feel free to do so - it only take five minutes... one thing to note though - if you're using plug-ins that involve a degree of randomness (ie from beat repeat to some amp modelers) freeze them before exporting or else you won't get phase cancellation.

[Captain Comeback]

Also, like leeds said, it could be clipping - check out SSL's X-ISM & see if your tracks have inter-sample distortion going on. I recommend reading up on gain-staging even if this isn't the problem

I appreciate the help, levi, but how the hell does this thing work? lol

It goes on and on abot what it does but it says nothing about where to put it in your project. Does this go on the master? Does it go on each individual track? Are there adjustments for it? etc

[Captain Comeback]

thanks for the tips guys

I'm definitely filtering my sub and hi-end bass sounds for sure. On my sub bass, I put the hi cutoff at anywhere from 150khz to 300khz. On my hi-end bass sound, I put the low cutoff at anywhere from 300khz to 1000khz. I have a fairly good knowledge of balancing the frequencies I would hope, but I can always use more info. Maybe I just answered my own question with the above statement, but do you think it's because I put my sub bass cutoff so low that I can't hear it on the final exported product?

@ andrejs - I will check our the sox tonight and thanks for the help

While that will help keep your mix clean and seperated, that isn't going to help the issue of the low end below 40hz taking up too much energy and causing you issues on mixdown.

Try rolling everthing off from 30hz down, that'll help stop your bass from draining all the headroom in the mix, and should result in a cleaner mixdown.

I already do that. I cutoff all of my sub bass frequencies at 30khz in the eq eight. I even cut it off on the master channel too

[Captain Comeback]

I have more to say on this subject - first of all:

On my sub bass, I put the hi cutoff at anywhere from 150khz to 300khz. On my hi-end bass sound, I put the low cutoff at anywhere from 300khz to 1000khz.

What? Did you mix your words up here or do you really mean to say you hi-pass bass @ 1k? If so you haven't lost bass my friend, you've taken it out!

ok...I'll try to clear it up...

On my sub bass, i cutoff the high on the EQ8 at around 150-300khz to eliminate any high frequency

On my hi-bass, I cutoff the low end on the EQ8 at around 300khz to 1k based on how strong the signal is from the instrument. This is based on the fact if the synth contains at least one sine wave on the oscillator bank, since sine waves produce the lowest frequency impact. If it's a saw or square wave synth, I'll put the low end cutoff near 300khz

I'll also cutoff the low end of the sub bass on the EQ8 at 30khz to make room for the kick drum.

But my problem isn't the hi-bass, it's the sub bass. My sub bass is highly prevelant in the mixing of the track, but it just about completely disappears after export.

Hope that clears it up

[jhartford]

My sub bass is highly prevelant in the mixing of the track, but it just about completely disappears after export.

Did you read my post as cutoutcollective? (my copy of Live is registered under a band e-mail account and I forgot to change back to my forum account...)

Its very easy to show that there's absolutely no difference between what you hear before and after export...

So why are you hearing a difference? First - Have yo got the normalize button active? If you have then there will be a difference. It will compensate the master level to prevent you clipping the output in live so you won't get the distortion that you get when you're playing back in Live:


That is a picture of the same +6dB sine wave exported with and without the normalize button on. Note how with normalize off it almost becomes a square wave, so its going to sound very different. In this example I would be hearing a pseudo-square wave when I'm listening in live and a sine wave when I export. Now, hopefully you're not clipping your master channel by 6 dB, but even a little clipping will change the tonality of the song once you export (if you have normalize on).

If you don't have normalize on, then it could be one of two things - either you're listening to the song at a different volume (a difference of 0.5dB will be subtle enough for you not to really notice but enough to make think that there's a difference); or you're hearing a difference that's not there. Both cases are easy to check by doing the test that I described in the cutoutcollective post.

[Captain Comeback]

That's a very useful diagram. I don't normalize anything I export because I just don't like the idea of added noise to anything I've worked on. I never put any of the levels anywhere near the red lines, either. I keep everything about 60-70% volume, until I'm going into post-production. Then, I'm slamming the limiters and saturators on it to boost the performance volume. But even then, I'm still not allowing any redline distortion to occur. I've read loads of topics here about the normalization debate, and so far there's no convincing story for me to use it. However, if it's what I'll need to help cure this problem, I'll surely use it.

[leedsquietman]

No-one should be clipping the master buss by +6dB, even listening in 32 bit float, it's a bad habit. Using normalisation has sonic consequences too, it's better to just turn your faders down so you're not clipping, IMHO

People in general need to learn to turn things down, leave proper headroom for cleaner mixes and address the overall gain/loudness issues at the mastering stage.

There are limiters with multisampling, Voxengo's Elephant 3 being a great example as it has up to 8x oversampling to reduce intersample peaks (btw - don't use 8x oversampling mode while you are tracking, or just checking a mix, it will burn your CPU and have latency issues. I always run mixes at 2x oversampling and then just turn it to 8x when rendering). The best way to avoid clipping is still to turn down your levels but sometimes a super fast transient could shoot by and cause intersample clipping.

Captain Comeback - like Tarekith, I do demo mastering. If you want to send me an example of your work PM me, I will use my ears (primarily) and visual analysis to determine how I would approach your issues, and send you back a remastered version with a note documenting the changes I made for free, as a gesture of goodwill to another forum user (I usually charge for this type of service). If your .wav file is recorded at 32 bit / 96 this will take an age to upload, so hopefully the track won't be too long !! Don't dither or normalize the file. No obligations. (Maybe also send me the previous 16 bit .wav file you processed so I can compare).