Vocals and downsampling

[Brett - K A L I S I A]

Hey everybody,

I've noticed the following thing : when I mix (24/48) I set the vocals at a certain level. But when I bounce the song for CD burning (16/44.1) I always notice that the vocals suffer from it, being lower than what they were on the mix, so I have to go back to the mix and make them "too loud"... I know Andy never had this issue, but what about you guys ? Never noticed anything ? I guess this has something to do with the hi mid frequencies...

Thanks for your comments.

 

[black sugar]

Brett! I hear ya, brother.

I swear that the mixdown function in Cubase really does alter the mix, even if you're not changing the bit depth or sample rate. I've heard on pretty good authority that it soft limits any "clips," but it doesn't do it very well so that's what's causing the mix to sound different. It could also be that the Cubase summing engine works differently in playback and mixdown. I dunno.

Some people think I'm crazy when I say it sounds different, and they're probably right, but if you're crazy too, there are a few things you can try:

I just open a .wav editor (I like the way Samplitude "sounds" for this) and record the mix in realime from my Audiophile's monitor mix, then "save as" to 16/44.1 and dither the new file to 16bit.

If you don't have enough CPU overhead to do that (I don't sometimes) you can run Silverspike's TapeIt plugin on your Cubase master buss and it will render your mixdown in realtime to a new file. Pretty transparent from what I can tell.

Or you can just lower the volume of the mix and hope that the mixdown function doesn't munch it.

Whatever works, eh?


Wait, I hope that's what you're talking about! Haha.

 

[Brett - K A L I S I A]

Aaaah thank god I'm not crazy Another solution would be to stop using Cubase and go for another DAW (I know I'll go for Samplitude as soon as $$$...).

Anyway thanks a lot for the tips, I've just checked TapeIt and one of its use is cool : being able to use more than one instance of TapeIt simultaneously and do multiple track bounces at once, something not possible in Cubase... Ok, it would be realtime bounces (slower), but that's nice anyway. Oh by the way, do you think that not bouncing in realtime (faster) has an effect on the sound ?

 

[DURBANS]

I suffer from a different issue yet it's kinda the same.

When I mix down the song to .wav and then convert it to .mp3, most of what's in the middle of the stereo spectrum (vocals, bass) are lower in volume.

 

[axeman720]

you do lose some upper harmonics when downsampling. since the highest recorded frequency in a 48k session is 24khz when you bounce to 44.1 your highest reproduced frequency becomes 22khz. the downsampling will cause frequency loss and some aliasing from foldover, dither may help a bit if you have a good limiter with dither or just dither plugin. otherwise downsampling willl slighty affect your mixes and also you lose some dynamic range by going from 24 to 16 bit. but its more likely that you are just suffering the effects the of the way cubase processes audio. almost always you will notice slight differences between a two track and a full mix unbounced no matter what the format, its happens everywhere. can i suggest however, being patient and bouncing in real time. use cubases internal busses and create a stereo audio track and route all the tracks to that. i think you may notice the bounce to be a bit more lively, i could be wrong though. i do this with pro tools instead of using bounce to disc for my serious mixes and i certainly notice a difference.

good luck

 

[James Murphy]

Blacksugar, am i to understand that you convert to 16 bits and then dither? maybe you just wrote that unclearly... if not, let me say that you will have far better results by converting with dither. for instance if you bounce your mix to disc (whether by doing it with all your tracks or by doing it with a stereo mix you've recorded by bussing all you tracks to a new stereo track) keep your session file 24 bits... but put your dither plug as the very last insert on your master bus and set it to 16 bit with whatever noise shaping you prefer. after the bounce your file will be 16 bit.... dithered to 16 bit... and after that you should not run anymore processes at all, because it could cause the dither noise to be revealed..just take it from there to your CD compiling software. thus, be sure to sample rate convert prior to dithering using the best (and usually slowest) setting your software provides.... or better yet, record at CD rate to begin with and avoid dodgey SR conversions.... you're not hearing the "improvements" of the higher sample rates anyway... fooling yourself if you think you are. maybe with a small acoustic jazz trio or a string quartet, but not metal. you'll hear the effects of bad SR conversion though.. bet on that.

 

[Ermz]

Isn't there adverse effects when processing a signal at 44.1khz though? Surely there has to be some sort of degradation, which is why most folks record at 96k... right?

 

[axeman720]

to me recording at 96k at this point is useless. our delivery formats are still 44.1 and most dvd audio still ends up at 44.1 or 48 due to the bandwidth the video and seperate audio mixes take up. its rare that you see a truly 96k release out there. Also 96k converted to 44.1 is gonna have some weird foldover and downsample degradation. its much better to record at 88.2 (twice the sample rate of 44.1) because when you downsample all upper harmonics are exactly half of what was originally there causing less distortion and aliasing. make sense? if it doesnt do a google search on the nyquist theorem and digital audio basics, it may make some thigns clear that i have missed.

by the way moonlapse, ALL digital audio has degredation, it literally cuts out parts of what we are recording. its hard to explain but digital audio at 44.1 is literally taking a snapshot of the audio at 44,100 times a second. which means there at tiny spaces of audio and important harmonics that never make it to record, hence the harshness and immediacy of digital audio. 88.2 and 96k sound better because they are taking more snapshots per second and therefore recording MORE information, it sounds better by nature. there is no audio recording medium available today that doesnt cause degradation of some sort.

 

[Ermz]

Hehe, yes I understand the concept of sampling. I was always under the impression though, that provided an engineer works a mix at 96khz, ultimately less of the harmonic information would be lost when we downsample to 44.1khz. I figure since we're working with a higher quality of 'snapshot' per second, less of the information would get lost during the various signal processes that would occur during the mixing and mastering phases, which would then ultimately make the signal degradation from converting from 96k to 44k less quality-costly than mixing and mastering at 44k.

I mean, that's just the impression I was under. I'm sure when we cover digital audio concepts at uni more thoroughly I'll be a bit more qualified to speak my mind here.

 

[SPLASTiK]

And this my friends is why Analogue Tape still rules

Barbabatch is a pretty good SRC.

 

[Hewy80]

The arguements on sample rates always seem to occur in these sorts of forums and nobody ever agrees.
From what I've heard there can be problems with the clocks at the higher sample rates on cheaper equipment, dont know how true that is.
Either way it all ends up at 44.1, I remember Andy saying he uses 44.1 and if its good enough for him its definitely good enough for me

 

[Ermz]

Doesn't Andy use 48k?

 

[egan.]

And this my friends is why Analogue Tape still rules

. . . if you can find it.

 

[Brett - K A L I S I A]

Doesn't Andy use 48k?

No, he said that he didn't find any sample rate converter that sounded as good as directly recording at 44.1, that makes sense.

 

[ThomasT]

its much better to record at 88.2 (twice the sample rate of 44.1) because when you downsample all upper harmonics are exactly half of what was originally there causing less distortion and aliasing. make sense?

No, it doesn't.
The math is exactly the same. The samples will alwayes recalculated complety.

 

[James Murphy]

andy records at 44.1 , as he has stated many times here.. so do it.. 44.1/24 bit.. that is the way to go.. especially for metal.. you will NEVER hear the difference in a blind A/B test of the exact same song recorded and mixed at 96 and again at 44.1... you won't..

the Harshness of digital audio is largely a thing from the first 10 years.. today's converters are much, much. better.. even when operating at 44.1.

ananlog tape is leaving us.. the last plant has closed months ago... get over it.move forward.

you guys come to this forum of Andy's to learn from him... he has told you.. i have told you...

keep recording and mixing at higher rates if you want.. you will just spend more on hardrives and be able to run less tracks and plugs... your mixes won't be better.

good luck with it though.

 

[Nitronium Blood]

andy records at 44.1 , as he has stated many times here.. so do it.. 44.1/24 bit.. that is the way to go.. especially for metal.. you will NEVER hear the difference in a blind A/B test of the exact same song recorded and mixed at 96 and again at 44.1... you won't..

the Harshness of digital audio is largely a thing from the first 10 years.. today's converters are much, much. better.. even when operating at 44.1.

ananlog tape is leaving us.. the last plant has closed months ago... get over it.move forward.

you guys come to this forum of Andy's to learn from him... he has told you.. i have told you...

keep recording and mixing at higher rates if you want.. you will just spend more on hardrives and be able to run less tracks and plugs... your mixes won't be better.

good luck with it though.

I'm curious to know how you work with 44.1/24, especially [assuming you work with] ASIO 2.

Whenever I select ASIO (anything else causes tremendous lag) My only options are either 96/24 or 48/24.

This applies of course to my laughable SB Audigy 2 ZS [Yes, I know, I know]. I don't know if this is an issue with other sound cards.

 

[Kazrog]

I'm curious to know how you work with 44.1/24, especially [assuming you work with] ASIO 2.

Whenever I select ASIO (anything else causes tremendous lag) My only options are either 96/24 or 48/24.

This applies of course to my laughable SB Audigy 2 ZS [Yes, I know, I know]. I don't know if this is an issue with other sound cards.

This must be an issue specific to your sound card. I have never had this problem with any audio hardware I've ever used.

I myself work at 24 bit/48khz. I'm able to get better reverb trails at 48khz than at 44khz, and when I dither down things sound better to me than if I recorded at 44khz.

James, just out of curiosity, if you operate at 44khz, why bother with 24 bit for that matter, since our target media is obviously 16/44?

Edit:
I have removed the rest of this post, because I feel that some of the "information" I shared here is questionable. I apologize for any confusion my post has caused anyone. I will follow up with more research on "bounce to disk" features and potential bugs, but until I have done more experimenting on my own, I don't feel qualified to elaborate on the subject. My apologies to Andy and James.

 

[DecTV]

About the 16/24 question.

If we use the aproximation that signal-to-error (largest magnitude value/largest quantizing error) ratio in desibels is (S/E) = 6.02n+1.76 where n is the number of bits. (The error comes from converting analog to digital by quantizing)

With 16 bits we get 98.08 dB.
With 24 bits we get 146.24 dB.

The error is cumulative so every time you mix tracks together you get more error. And I bet there are a few tracks mixed together by the time you get to the final stereo mix.

So it would make sense to keep the tracks at 24 bits until the final stereo mix.

 

[Brett - K A L I S I A]

But still The Gathering was recorded in 16 bits / 44.1 KHz... So all in all, I'm pretty confident that if your record (takes, mix, etc) doesn't sound good in this format, it won't sound any better in 24/48... Next project I'll try to work at 44.1 to try and see... But I'll stick with the 24 bits (much more headroom, so you're less concerned about noise and digital clipping).

Concerning the bounce to disk, this is really a big problem, because if you can't trust what you hear then this sucks big time... Recording the output of the mix bus seems a good option, if you have the resources (I'm not sure I could, most of the time my CPU is dying... but I didn't work with the UAD-1 yet). One quick question : if you work on a 24/48 mix, how would you "bounce record" it ? 16/44.1 using dithering in the DAW ? Or 24/48 and convert later (longer) ?