Improving sound quality

[Lawrence]

I started doing those kinds of thing when I realized that I would obsess over (for instance) drums for an hour only to find out two hours later that the most important track (IMHO, the lead vocal) didn't really sit well in there. Sure I'd bring it in when mixing the drums and it sounded fine. But the furthur along I got with the mix, the more instruments came in, I'd start to lose "quality" of the lead vox.

Then I would compress it and eq it and twist it and generally farg it up.

Now when I have what I feel to be a really good vocal (especially if the natural tone without any eq is really good) I work on that first and eq the drums and the other tracks around that. I try to leave that good natural tone and the artificial space I created around it intact. "Here's Joe singing in my artificial room acapella". Now lets make the "band" support that without disturbing that image. Just a different approach. It works well for me when I do it.

Many engineers follow the standard method of mixing from the drums up (hey, it works fine 95% of the time) but sometimes all of that stuff has to be changed or the lead vocal has to be changed. I've seen many people eq'ing a lead vocal to fit better with a track mix that by iself, is really slamming. I disagree with that method. Opinions vary.

I often like to have a concrete image of the vocal and it's space burned into my brain (and ears) so when the other instruments come into the mix I can hear the subtle changes from the unwanted interferences. If I start with everything in I never really get to hear that vocal image. It's always affected by the early mix.

IMHO you should be always carving a space FOR the lead vocal even if you're mixing from the drums up. Not the other way around, eq'ing the lead vox to fit the music bed you just spent 4 hours on.

Make the music mix "slam" _around_ your natural and great vocal.

Just my .02. With that and another .48 cents I can buy a Snickers bar!

Lawrence

[theblue1]

Originally Posted by Hope209

... And once I bounced the format down to 44.1k 16-bit and put it on a cd ...

[EDIT: I was working from an incorrect understanding of the issues when I wrote this. It's wrong. I much regret spreading this incorrect info. Not to mention being kind of a dick about it. ]


You bounced a recording made at a 48 kHz SR down to 44.1 kHz and you have to ask why it sounds crummy?


Next time, start at the sample rate you want to end up at -- or an even multiple thereof. So, to output to CD -- and you're keeping everything in the digital realm (IOW, you're mixing in the box or via a digital mixer with no analog stage) -- work at 44.1, 88.2, or 176.4 kHz.

There is no filter or magic dither algorithm that will produce a transparent sample rate conversion from a source rate that is an uneven multiple of the target rate.



[BTW, to dither means to add noise, either random or shaped. It does NOT mean either digital word length reduction (bitrate conversion) OR sample rate conversion. Thankyuhvurrymuch.]

[lukejs]

I was told by Apogee to try to stay away from clocking via ADAT because of increased jitter caused by the signal passing through plastic, or something like that if my memory serves me right... Any possiblitiy that the clarity is because you are clocking this way ? Also, have you tried tracking at 96khz ? Maybe Í´m wrong here, but I thought clocking through bnc cables were best... just a few ideas to throw at ya.... I¨m no expert, but tryin to be helpfull....

[lukejs]

All I can say is........ Mann... I wanna meet a girl like that !!

,

[Hope209]

Originally Posted by lukejs

I was told by Apogee to try to stay away from clocking via ADAT because of increased jitter caused by the signal passing through plastic, or something like that if my memory serves me right... Any possiblitiy that the clarity is because you are clocking this way ? Also, have you tried tracking at 96khz ? Maybe Í´m wrong here, but I thought clocking through bnc cables were best... just a few ideas to throw at ya.... I¨m no expert, but tryin to be helpfull....

Hmm, I'm not sure it's possible to use BNC connections with the 002. Am I wrong, though? Right now I have two Hosa OPM330 Fiber Optic cables with toslink connectors runing in the ADAT Ins/Outs

[RainbowStorm]

Hope209,

I have a Yamaha Tyros myself and I read that you are using a Yamaha Motif, these are quite similar in terms of sound quality and articulation. One big problem with these samples is in the velocity response. If I use initial touch I get a completely different sound than if I would use the keyboards natural attack, because it's impossible to express that kind of attack on the keyboard. So I lose something like 5 - 10% of the velocity pitch only by expressing my playing on the original keyboard. Besides this, much of these samples are not sampled in stereo and the amount of layers is not very high. The result is that there is simply not the degree of "expression" in these kind of sounds that could match the real thing. Expression IS one of the most important things in music, it's what makes music musical sounding. I can record whatever with my Tyros, but I can't match the real thing. If I choose a piano or a B3 I have already lost the competition. What I need to do with these sounds to make them "fatter" is to amplify them in a certain way (for instance with a tube driven amplifier) and apply certain effect processing and then mic these with m/s miking. The mics can then interpret the expression in a certain way and for instance a multiband compressor can act as a customizer. In that way I can track acoustic instruments in a much more interesting way to minimize the "cheese". But even though I do this I have to be very selective with what element I use. For instance, I have a hard time using piano or B3 in any kind of solo context and wouldn't use it in a professional context not even as the pad element. But during the last 5 - 10 years software instruments have improved in quality and some professional session players really use samples to save time and money. Often these instruments belong to the so called "pad element" that don't take up a lot of dB in the mix and often these intruments are heavily cut in the EQ during mixing to consume the sound field most efficiently. They still sound "good enough" for professional use...! I think your Ivory piano is such an instrument that a lot of session keyboardists would find useful...

Sometimes people don't believe me when I say I have used a keyboard when I've recorded the song. That is not because of the sound quality itself, but because of the way I try to immitate the kind of expression that happens naturally when you are playing the real thing. The problem is that a musician communicates with the instrument in such a way that it sounds the way you want it to. For instance when I play real electric guitar I try to pick exactly has hard as I need in order to make some tones sound in a certain way. Imagine how difficult this whole process is when you are playing the same thing on a keyboard even with good sounding samples when it's even hard on the real thing. But this is currently the "sound" you try to match here with drums and acoustic guitar presets in your Yamaha, instruments that are extremely critical for efficient sound field consumption. It will really create the effect of "something is missing", especially without any additional processing. So when it comes to samples played on a keyboard, focus mostly on pianos and organs that are naturally played in the same way.

But still, as always in music, it's a combination of things and that's why I think you will solve this when you've got some more time. When you start focusing on efficient sound field consumption you will automatically get the right focus on how you need to deal with your decisions in recording.

[theblue1]

[EDIT: I'm wrong, wrong, wrong, here. And insufferable, too. Sorry. Sorry. Sorry. ]

I feel another rant about willful ignorance coming on...


There SHOULD be no mystery here:

People... he bounced from 48kHz down to 44.1kHz... to a target rate from a source rate that is NOT an even multiple of the target... how can one NOT get a degradation of sound? And I'm NOT talking about just losing that roughly 10% of time resolution.

It doesn't matter WHAT SRC algorithm he used - the sound will CERTAINLY be much farther from the original source audio than if he'd started at 44.1 in the first place.

It's Digital Audio 100, for cryin' out loud.

[RainbowStorm]

Originally Posted by theblue1

I feel another rant about willful ignorance coming on...


There SHOULD be no mystery here:

People... he bounced from 48kHz down to 44.1kHz... to a target rate from a source rate that is NOT an even multiple of the target... how can one NOT get a degradation of sound? And I'm NOT talking about just losing that roughly 10% of time resolution.

It doesn't matter WHAT SRC algorithm he used - the sound will CERTAINLY be much farther from the original source audio than if he'd started at 44.1 in the first place.

It's Digital Audio 100, for cryin' out loud.

Yep, that's a problem. There are a few things to think about in that context though. First of all he has good converters. Secondly some converters perform the best at certain sample rates, I'm sure there's a reason why he records at 48KHz. That compensates for the sound degrade quite well. I've also heard very good sounding mixes that were originally recorded at 48KHz and then downsampled to 44,1KHz. And is it the biggest problem? IMHO it's not his main problem here. Just listen to the Ivory piano. It's clear and natural even though it's sampled and even though it has gone through that downsample process. Prove me wrong.

[theblue1]

I'm talking about basic digital audio reality here. [EDIT: Wrong. Wrong. Wrong. Ignore me.]

He's got a fundamental problem if he's recording at 48 kHz and then bouncing down to 44.1 kHz. If his primary concern was audio for video, I'd say, cool, take the hit on the CD side and keep things good on the video side.

But that's apparently not the case.

So I'm assuming he was working at 48 kHz SR because of some misguided idea that he's going to get better fidelity that way -- but instead he is assured of getting worse fidelity. (Now, some people claim to like the 'edge' they get from the alias error implicit in such an 'uneven' SRC. But that's apparently not the case here.)

Is it possible he has OTHER problems (maybe even worse problems?)... my thinking is no doubt... since he apparently -- as apparently many/most others in this thread do -- fails to grasp an important fundamental of digital audio reality.


Did I listen to his tracks?

No, I did not.

Why? Because -- as far as I'm concerned -- someone who is recording at 48 kHz for output to 44.1 kHz has NOT performed the "due diligence" required for me to spend time trying to figure out what his OTHER problems may be.

Has he got BIGGER problems?

If you say so, I'll take your word for it.

But until he does his basic homework -- why should I or anyone else lavish time trying to figure out what ELSE went wrong?


I guess that makes me sound like a hard-nosed old curmudgeon. Hell, I probably AM. But I'm just BURNED OUT on people who ask questions they can answer by googling, people who pour money into gear they don't understand, people who call themselves "engineers" when they don't really have the faintest grip.


There are a LOT of really knowledgeable, experienced people here. And there are far more who don't know shit. Some of the latter should be paying attention to some of the former.

(And -- NO -- I am not including myself in the former group. I may have spent much of the 80s recording in studios, and the 90s in my own 16 track project studio -- but I KNOW I have a lot to learn. It's all those people who don't appear to know much of anything but talk as though they think they do that bug the crap out of me. I was talking to a buddy who used to love this place and, like me, he's really burned out on folks who can't be bothered to understand the fundamentals of their craft.)


[UPDATE: Geez, I just reread this and can't believe what a dicky mood I must have been in when I wrote this. SORRY. ]

[Kiwiburger]

Preach it brother. Sample Rate Conversion is a very dodgy area - and most DAWs have very flawed SRC algorithmns. I choose 44.1 and avoid SRC.

However - to balance this out - it is possible to get a reasonable conversion from 48 to 44.1. Just not with most DAW software.

It's possible to use a D/A and A/D converter to convert the rates. Obviously that introduces some noise and distortion - but plenty of engineers are using external analog outboard gear, via two converters, and the improvement in sound quality is worth the degredation in the converters.

In fact - it would appear that music that lacks some analog degredation is too clean and sterile anyway.

So there is no reason why you can't convert from 48 to 44.1 using converters in the analog realm and have acceptable results.

www.voxengo.com have a SRC application that offers clean conversion that is similar to back to back converters.

The argument about uneven maths isn't really valid. You can always find a common denominator between any two numbers by first multiplying them by each other.

It's possible. Just don't trust your DAW software to get it right.

[theblue1]

Thanks, Kiwi...


It's not really matter of SRC algos being flawed.

There is simply NO such thing as transparent SRC from a rate that is not an even multiple of the target (or, for that matter, the inverse upsample). There is NO magic algorithm, no magic dither component...


[EDIT: I was wrong. There is no theoretical reason you can't do a transparent SR from an 'uneven' multiple. A real facepalm moment for me, here. If they'd only had facepalm icons, then. And, if I'd only realized my error. Later, none other than Dan Lavry handed me my head (very politely) on this very issue. I stood corrected. ]

And speaking of dither... I'd go away and feel like I'd accomplished something if I could just convince the great unwashed masses here at GS that dither is NOT synonymous with either bitrate conversion or sample rate conversion...


I mean... THAT should be easy.

So, really, it's clear I'm banging my head against a wall that is bigger and much more unyeilding than my head...



PS... Let me just issue a general apology to anyone I've offended with my blunt talk. I'm afraid I'm not feeling any too diplomatic today. Mea culpa.

[Kiwiburger]

No apologies necessary - you have a very good, often overlooked point.

But I still think you are not correct in saying that all SRC from 48 to 44.1 is necessarily badly flawed. (Assuming you can accept that a digital recording at 44.1 is acceptable in the first place).

Do you accept that you can convert cleanly using a good D/A at 44.1 into a good A/D at 48? If you can accept the quality level at 44.1, then simply recording the analog output of audio previously recorded at 48 should also be perfectly acceptable.

That being that case - it is theoretically possible to have software SRC that achieves the same thing, and arguably better since there would be no noise floor.

But ordinary DAW software SRC is not at this quality level.

Obviously - there must be some loss between 48 and 44.1, but I don't think it has to be the huge, life changing, soul destroying problem that you make it out to be.

I agree that Dither is an abused word.

Sample rate conversion is not dithering, and neither is bit depth conversion.

[theblue1]

Originally Posted by Kiwiburger

No apologies necessary - you have a very good, often overlooked point.

But I still think you are not correct in saying that all SRC from 48 to 44.1 is necessarily badly flawed. (Assuming you can accept that a digital recording at 44.1 is acceptable in the first place).

Do you accept that you can convert cleanly using a good D/A at 44.1 into a good A/D at 48? If you can accept the quality level at 44.1, then simply recording the analog output of audio previously recorded at 48 should also be perfectly acceptable.

That being that case - it is theoretically possible to have software SRC that achieves the same thing, and arguably better since there would be no noise floor.

But ordinary DAW software SRC is not at this quality level.

Obviously - there must be some loss between 48 and 44.1, but I don't think it has to be the huge, life changing, soul destroying problem that you make it out to be.

I agree that Dither is an abused word.

Sample rate conversion is not dithering, and neither is bit depth conversion.

[bold added for emphasis]

Well... just talking about it is destroying my soul...


Obviously, if you have already have a 48 kHz file and you need it in 44.1 kHz you're GOING to have to do something.

I would probably try it both ways.

That's what I've always done in the past when I was faced with that dilemna -- and, whaddya know? I ALWAYS ended up going with the 'analog copy' method over the software SRC based on the results of both.

But it is an imperfect COPY of an imperfect COPY. And we all know what that means...

And, at least in my experience, the audible results from 48 kHz D/A to 44.1 kHz A/D (I had a DAT machine back at the end of the 80s and I didn't get a CD burner until '96... all of a sudden I was confronted with just this very issue) sounded 'better' have always sounded better than the results of software SRC.


Anyhow... the only place where I feel the need to turn Jeremiah is in the decision of what sample rate to start at:

It seems to me, if one is truly aware of what the issues are, one is a sap to start working at anything other than the target sample rate or an even multiple -- when one will be keeping the entire project in the digital domain.

[superjimmer]

I honestly think everyone here has a piece of the explanatin nailed. Me personally? It's mostly in the arrangement and the players and the tone of the instruments. No bass and sampled drums to start with . . .

A great performance with real players who can process and deliver the tone of their instruments on the fly and taylor it to a reasonably treated room (particularly drums on this mix) and this mix would be entirely in a different league even if it were recorded with all SM57's through digi pre's.

Basically real drums by a good drummer and a nice bass line and suddenly your a better mixing engineer. Turn down the piano and maybe treat the vocals a hair differently too would be my taste.

[KevWind]

Hope. I am no expert here but, you asked, if when you are bouncing, is the conversion taking place in the 002 or the rosetta 800 ? if you are not going to an outboard recorder and just staying in pro tools, then the answer is the conversion is happening in the 002. also it is unclear from your post if you are trying to SRC and reduce bit depth at the same time, if so, don't !!!. First do the SRC, then go from 24 to 16 bit. As some one has suggested and as I have been told by some very knowlegeable engineers, that they prefer to record at 44.1 - 24, keep that for mastering And make another copy to reduce to 16 bit, if you need to burn a CD. Also you might want to use your SPTIF in and out from the 002 to the Rosetta. Good luck

Kev

[A27Hull]

Originally Posted by Hope209

And once I bounced the format down to 44.1k 16-bit and put it on a cd and listened in my car the mix just seemed to lack SOMETHING...The quality just wasn't as good as commercial stuff.

I haven't read through the entire thread so forgive me if others have mentioned this...

In one phrase...Professional Mastering.

You've got the gear, and your learning the skill. What's next? Learning the value of a good set of objective ears to tweak your mix. Professional mastering is a major reason commercial music sounds as clean as it does. I'm not saying you get it without all the awesome engineers and producers who mixed them, but that the final step, pre-mastering, has continued to exist for a good reason.

When talking about getting a mix "radio-ready," a great producer once told me that my mixes will never sound as good as they did in the studio. This is true for me. So now my goal is to make mixes that sound great outside my studio, in most if not all audio systems, even if they are not commercial releases.

The people who mix commercial material work in some of the finest environments available. It is quite possible, (especially for those like Chuck Ainlay,) for commercial mix engineers to come away from a mix session and have something that pleases in any system. A mix that can stand on its own, in other words.

But even those engineers whose mixes sound phenomenal on there own still appreciate the pre-mastering phase.

I'm sure that someone has asked where else to you reference the mixes (aside from the car...). I also know that in my world, professional mastering only comes into play when the project is finished. So, how about the many mixes I print during the project? How can I get them to sound like post mastering commercial mixes before the mixes are in fact pre-mastered? (say for the client's desire sake...)

One thing I have done is find some comparable commercial music, sit it side by side with my mixes and compare them. Mine always fall short of course, but from the difference I can deduct where my problems are, then I do post mixdown tweaks based on what differences I find in the commercial mixes.

In a way, I feign professional mastering. I EQ, Level and Compress my own mixes so that they take on the form and balance of commercial stuff. Ideally, I'm not the one who should be doing this. The main reasons are that I do this on non commercial releases is that it helps save money and that the client always wants to hear the best thing you have to offer them. They want to hear their music sound as best it possibly can.

Not everyone can effectively do this. I understand that I could be hurting my mixes rather than helping them. I choose to do this based on my preference and limited knowledge of mastering processes. YMWV!

So, aside from the other books and resources others may have mentioned, try finding a Mastering book, by a great mastering engineer. Learn what sort of processes they do to mixes, then experiment yourself.

Hope this helps-

Andrew Wayland

[Lawrence]

People... he bounced from 48kHz down to 44.1kHz... to a target rate from a source rate that is NOT an even multiple of the target... how can one NOT get a degradation of sound? And I'm NOT talking about just losing that roughly 10% of time resolution.

It doesn't matter WHAT SRC algorithm he used - the sound will CERTAINLY be much farther from the original source audio than if he'd started at 44.1 in the first place. .

Coincentally, on the point of this topic, today a client brought in two stems at 48k for touch up. He mixed the song at home but wanted the benefit of my better monitoring space for a quick touch up. One stereo track with music and another with all the vocals premixed, recorded at 48k. As they were they sounded pretty good. I switched my master clock to 48k and went to work.

We applied some eq to both individual tracks for more clarity and added multiband compression to the overall mix (and a little L2 ) and rendered a mix at 16/44.1k. There was no audible difference in the playback of the 16-bit mix file vs. the mix of the 48k stems in the daw that we'd been listening to for the hour it took to "touch up" the audio. There was no audible loss in "quality" or "fidelity". There obviously IS a difference because SRC is not perfect with rates that are not multiples of 44.1. It did the job so well it was a non-issue.

I absolutely agree that a daw has "lots more work to do" to reliably convert 48k to 44.1 faithfully. The science behind SRC is well documented and no SRC at all should be better. But I must also state that some daws (at least SX/Nuendo) seem to be doing it very, very well. So well that it's (at least in my case) practically irrelevant. I don't mean to suggest that's the case with evey daw in every studio.

I do agree that if overall sound quality is poor then sure, attack every avenue for possible improvement.

Lawrence

[theblue1]

I'm glad that worked out well.

I'm not saying it's the end of the world if you need to do a conversion from 96 or 48 to 44.1.

I'm just saying that -- as a rule -- it's best to work at a rate that is an even multiple of the target rate so that you won't have to go through such a process. One may not hear a problem -- but such a process nonetheless introduces the near certainty of further alias error to every new sample value extrapolated -- and that's all of them. The drift in relative accuracy may not always be audible -- but I can't help but feel it's best to avoid the necessity in the first place if possible.

[EDIT: Wrong. Ignore my sorry backside.]

[u b k]

Originally Posted by Lawrence

As they were they sounded pretty good. I switched my master clock to 48k and went to work.

...

There was no audible difference in the playback of the 16-bit mix file vs. the mix of the 48k stems in the daw that we'd been listening to for the hour it took to "touch up" the audio.

so src is transparent on tracks that sound pretty good. what happens on tracks that sound breathtaking, with a silky top and depth and dimensionality for days?

src affects those things, audibly.


gregoire
del ubik

[u b k]

others have said it, but it bears repeating:

MASTERING.

by a 3rd party pro with experience, a great room, and top flight analog gear.


gregoire
del ubik

[theblue1]

And... if you do it that way (analog mastering) you can give your 48/96/192 kHz SR file to the mastering engineer knowing you won't have any sample rate conversion issues...

And with that... I think I'm moving on to the next controversy...

[RainbowStorm]

Originally Posted by theblue1

And... if you do it that way (analog mastering) you can give your 48/96/192 kHz SR file to the mastering engineer knowing you won't have any sample rate conversion issues...

And with that... I think I'm moving on to the next controversy...

Hilarious!

[Greg B]

Originally Posted by theblue1

You bounced a recording made at a 48 kHz SR down to 44.1 kHz and you have to ask why it sounds crummy?


Next time, start at the sample rate you want to end up at -- or an even multiple thereof. So, to output to CD -- and you're keeping everything in the digital realm (IOW, you're mixing in the box or via a digital mixer with no analog stage) -- work at 44.1, 88.2, or 176.4 kHz.

There is no filter or magic dither algorithm that will produce a transparent sample rate conversion from a source rate that is an uneven multiple of the target rate.



[BTW, to dither means to add noise, either random or shaped. It does NOT mean either digital word length reduction (bitrate conversion) OR sample rate conversion. Thankyuhvurrymuch.]

Using sample rates that are multiples of 44.1 is not important. It is a myth that DAWs simply use every other sample when converting from 88.2 to 44.1. That's not how it works. There are more complex algorithms for converting down to 44.1 from 88.2 which is basically the same process as converting from 96kHz to 44.1, except 96kHz starts with more information so you can end up with a slightly higher quality result with 96. Based on what I've read (and heard with my ears) there is no benefit from 88.2 over 96 except that it uses slightly less hard drive space and CPU. That said, it's not always the highest sample rate that sounds the best. Different converters perform best at different sample rates.

[theblue1]

Originally Posted by Greg B

Using sample rates that are multiples of 44.1 is not important. It is a myth that DAWs simply use every other sample when converting from 88.2 to 44.1. That's not how it works. There are more complex algorithms for converting down to 44.1 from 88.2 which is basically the same process as converting from 96kHz to 44.1, except 96kHz starts with more information so you can end up with a slightly higher quality result with 96. Based on what I've read (and heard with my ears) there is no benefit from 88.2 over 96 except that it uses slightly less hard drive space and CPU. That said, it's not always the highest sample rate that sounds the best. Different converters perform best at different sample rates.

You are 100% correct and I was dead wrong and I thank you for pointing it out.


When I realized I had my head where the digital sunshine don't shine on this issue, I went looking for places where I'd spread this misinformation but, apparently, managed to miss this one.

A huge, fat mea culpa. And a half.