Clarification on optimum Mix level

[masterblaster]

When submitting mixes to pro mastering houses, what is the optimum level?

Is this an absolute, or is there an acceptable range? For example-- what would be considered too low and too high for modern rock?

Do optimal levels vary at all by genre?

Is the Massey Tools meter on my PT MasterFader insert a reliable way to measure?

Thanks for clarifying these issues.

[masterizer]

Originally Posted by masterblaster

When submitting mixes to pro mastering houses, what is the optimum level?

Is this an absolute, or is there an acceptable range? For example-- what would be considered too low and too high for modern rock?

Do optimal levels vary at all by genre?

Is the Massey Tools meter on my PT MasterFader insert a reliable way to measure?

Thanks for clarifying these issues.

You want to leave us plenty of headroom, -3db or more, no limiting/heavy comp and no clipping please. The meters will give you an idea of where you're at, the only true way to measure loudness is with your ears and we'll take care of that. If you want it "louder" at the mix stage, turn up your monitors.

Records that come in too loud sound like crap coming in no matter who mixed it, its not impressive! The loud releases that are listenable are the result of great tracking/performance, mixing, and mastering, leave the "loud" to us if its desired in the first place. Load some pre 1990's records in your DAW and listen and watch the meters, then load in a 2000's record and compare (this is what not to send to mastering!)

We wish records came in too low

[philip]

I don't get why some wants -3dB, as long as there is no overs it's perfectly fine.

[vx2000]

I believe most MEs are OK with mixes at -20 dbfs RMS to -14 dbfs RMS with peaks less the -6 dbfs.
This is probably an ideal target range, but they (of course) get hotter mixes at times (-12 to -10 dbfs RMS)
with peaks at -3db. This does not offer a ME much room if needed.

[Lagerfeldt]

Originally Posted by masterblaster

When submitting mixes to pro mastering houses, what is the optimum level?

Is this an absolute, or is there an acceptable range? For example-- what would be considered too low and too high for modern rock?

Do optimal levels vary at all by genre?

Mixdown for Mastering Tips (PDF)

[Bob Olhsson]

You don't want to peak limit to increase the level simply because there is no way to know what the final level needs to be. Without the final level being defined, there is no way to know how to best limit it. Doing it one time the best way generally sounds better than doing it more than once.

I also try to peak between -6 and -10 so as to make absolutely certain that my analog monitor path isn't being stressed and unduly affecting my mixing decisions. This is less of an issue when using a high-end DAC for monitoring.

[vx2000]

Originally Posted by Bob Olhsson

I also try to peak between -6 and -10 so as to make absolutely certain that my analog monitor path isn't being stressed and unduly affecting my mixing decisions. This is less of an issue when using a high-end DAC for monitoring.

Good advice Bob.

[Key]

Originally Posted by philip

I don't get why some wants -3dB, as long as there is no overs it's perfectly fine.

I don't either. Is it some sort of rule that MEs don't like to turn down the volume before doing whatever they need to do?

[Audiop]

Originally Posted by vx2000

I believe most MEs are OK with mixes at -20 dbfs RMS to -14 dbfs RMS with peaks less the -6 dbfs.
This is probably an ideal target range, but they (of course) get hotter mixes at times (-12 to -10 dbfs RMS)
with peaks at -3db. This does not offer a ME much room if needed.

Hi!

Room for..? DAW's and analog allows you to not only up the level but also to take it down, no?


/Peter

[Never1]

Originally Posted by Key

I don't either. Is it some sort of rule that MEs don't like to turn down the volume before doing whatever they need to do?

It's simply a buffer to ensure there are no intersample peaks, and also to make sure that the mix AE doesn't slam the mix buss with an L2 or something.

You'd be surprised to find out how many mixes come in that way.

I just finished a job recently, that was smashed to death. Unfortunately, a remix wasn't an option. It's a damn shame and made my job much more difficult.

I really don't know why more people aren't educated on this topic.

I always try and get advanced reference mixes from the client to check and make sure that they aren't smashed.

Also, any type of gain change to the material is technically a distortion of the original mix, so it's best to avoid turning a mix down before turning it back up.... besides, a smashed mix at -6dbfs is no different then a smashed mix at 0dbfs.

[Key]

Oh I can believe it that you get pre-squashed mixes. I think it's too much CDs being used as references.

But I don't think gain changes should cause any distortions provided you do them at 32-bit float or above.

[Bob Olhsson]

Actually all gain changes cause distortions so minimizing the number of them can theoretically improve the sound.

Digital audio is numbers. Throwing everything into a single grand calculation would be ideal but we rarely have that luxury. Still needlessly changing the gain back and forth multiple times is pretty silly.

[Key]

Eh these distortions are negligible and in the case of 32-bit float so small they are very difficult to even measure. I think telling people arbitrary rules like -3dB just adds confusion because you are not addressing the real problem. In a lot of cases you will just get a compressed file that is limited to -3dB instead of -0.03dB because you were not straight forward enough with what you do not want. Will be back to show you just how small gain staging distortions are - below the capabilities of any DAC I know of.

[Bob Olhsson]

Originally Posted by Key

Eh these distortions are negligible and in the case of 32-bit float so small they are very difficult to even measure...

Individually, yes but we are always talking about a long succession of calculations in the real world.

[Never1]

Originally Posted by Key

Eh these distortions are negligible and in the case of 32-bit float so small they are very difficult to even measure. I think telling people arbitrary rules like -3dB just adds confusion because you are not addressing the real problem. In a lot of cases you will just get a compressed file that is limited to -3dB instead of -0.03dB because you were not straight forward enough with what you do not want. Will be back to show you just how small gain staging distortions are - below the capabilities of any DAC I know of.

Well on one hand you're right, and we take great pains to make sure the client doesn't bring in a smashed mix, but on the other you're missing the point.

Setting a number as a "maximum peak" is a fail safe for human error. To say peaks should be -6dbfs leaves tons of room for an occasional over. Now if the mixer starts going up into the -3,-2, or even, god forbid, -.03dbfs area, there's potential for some horrific things to happen to the audio. I stress the word potential.

On a 24bit system, - 18dbfs is still using 21 bits of the audio engine, so it's just plain idiocy to be shooting for hot levels.
I would wager that most MEs would want to avoid doing anything, even something as trivial as turning something down, if we can avoid it.

"Do no harm" is always the mantra and sometimes that means saving the client from themselves by educating them on digital audio systems.


In short, the -3dbfs is a buffer zone. period. No need to question it.

[Never1]

Originally Posted by Bob Olhsson

Individually, yes but we are always talking about a long succession of calculations in the real world.

And as always, Mr. Olhsson is spot on.

Never said it before, but I'm a big fan of your work, Bob!thumbsup

[masterblaster]

Originally Posted by Lagerfeldt

Mixdown for Mastering Tips (PDF)

Thanks for the link. BTW, you have a beautiful website & gorgeous room!! Love your design aesthetic! STYLE-- must be a Denmark thing.

[Lagerfeldt]

Thanks

[Key]

Really I am giving you guys the benefit of the doubt even half admitting that there will be distortions from gain adjustments in a mastering program.

I made a VST that has 10 arbitrary gain staging steps in it. Then created a test tone in 96kHz 24-bit. Ran the file through the arbitrary gain changes once in 24-bit mode and once after first converting the file to 32-bit float. The distortions are too low to even measure with a program like Right Mark Audio Analyser and I don't even think I see any on the spectrogram.

Original test tone generated at 96kHz 24-bit

(note what looks like distortions is actually the limitation of 24-bit and the noise
threshold of a 24-bit file)

96kHz 24-bit > 96kHz 32-bit Float> Arbitrary Gain Change VST> 96kHz 32-bit Float


96kHz 24-bit > Arbitrary Gain Change VST > 96kHz 24-bit



Frankly I am having a difficult time getting any distortions or seeing any in the average destructive editor in either 24-bit or 32-bit float. I CAN induce a bunch with the average multitrack - but ONLY when you render the final file as 24-bits instead of 32-bit float.

If you want to avoid unneeded tiny distortions then I would recommend you tell your clients to render at 32-bit float and send you those files instead of 24-bit. You will find much more deviations in this respect than you will if they choose to turn up the master fader so that it is maxed out around 0dB instead of around -3dB - which will only sacrifice a very small amount of resolution and not really accomplish anything else.

[philip]

In short, the -3dbfs is a buffer zone. period. No need to question it.

why? honestly I don't think either you or Bob has any clue what you are talking about.

As long as there is no peaks over 0dB it's perfectly fine. If you got inferior AD/DA or outboards that can't handle full scale, it's not the clients problem.

If there has to be a gain change, why does it matter if you or the client have to do it?

And BTW almost all plugins has some gain change involved not to talk about SRC and other things that are potetialy "worse". Not turning down the gain i a DAW cause of distortion is just... anal.

[Never1]

Originally Posted by philip

why? honestly I don't think either you or Bob has any clue what you are talking about.

As long as there is no peaks over 0dB it's perfectly fine. If you got inferior AD/DA or outboards that can't handle full scale, it's not the clients problem.

If there has to be a gain change, why does it matter if you or the client have to do it?

And BTW almost all plugins has some gain change involved not to talk about SRC and other things that are potetialy "worse". Not turning down the gain i a DAW cause of distortion is just... anal.

I would much rather a mix come in that is peaking at -6db, or -3db then one that hit 0. Maybe I AM just paranoid, but if the mixer let the master buss hit zero, then it tells me that he probably isn't overly cautious about gain staging his plugins, tracks and busses, either; all of which will create some rather minute unpleasantries on a channel by channel basis but can add up to a bigger sonic degradation, globally.

To push a DAW to its upper limits in a 24bit system is just rediculous.

And I would rather not have to trim back the mix before applying any processing. Yes, I guess I am anal...

Say what you want, but any change is a recalculation, hence a distortion (not to be confused with "distortion") of the audio and we want to avoid any unecessary steps, even one as subtle/trivial as dropping the fader down.

I like it when I can just tweak and polish a mix without having to pull any stunts like that. Shooting for 0db is just silly, though it may not necessarily do any harm on a system by system basis.

I guess I don't understand why people get all up in arms about this, when the science is there to prove that keeping levels a bit more conservative will yield cleaner results.

Whenever I've mixed with total disregard for gainstaging (maybe some red lights on some plugins lighting up and my mastrr buss was getting up close to 0)I have noticed a very real difference in the quality of the finished mix. Everything sounded cloudier and less defined, when I compared it to a mix that was "anal retentive", level-wise.

Maybe I don't know what I'm talking about, but I trust my ears on what I've heard come in to the shop, as well as on my own work.

[Bob Olhsson]

Originally Posted by philip

As long as there is no peaks over 0dB it's perfectly fine. If you got inferior AD/DA or outboards that can't handle full scale, it's not the clients problem..

The idea is for people to not be making their mixing choices based on listening to a D to A or monitor chain that can't handle full scale digital signals in the analog domain.

You can't expect to get a sharp picture when your viewfinder is blurred.

[Key]

Ah but see this is an assumption that just because the DAW peaks around 0dB that they will be monitoring with the outputs set to full scale. I do not do this myself. I lower the output on my soundcard to around -6dB when I am mixing. I know, I know, some people say this sacrifices resolution or whatever but in my experience the alternative - adding an attenuator - causes more problems than it solves.

I do see where you are coming from in a way it's just that you should expect some smart ass engineer like me to look at the rule and scratch there head. I personally think that as long as you are not clipping, not too soft, and are at an adequate bit-depth - 24-bit or higher - the premaster should be fine. Watch out for internal clipping in the plugin domain it is very easy to do without you ever realizing it on certain plugs. Oh and of course leave the Maximizer off of the master fader.

[philip]

I've got nothing else to say - just wanna apologize for being rude. Of course you all have a clue and I'll respect your thoughts and ideas.

[Bob Olhsson]

Originally Posted by Key

Ah but see this is an assumption that just because the DAW peaks around 0dB that they will be monitoring with the outputs set to full scale....

It's an assumption because that's precisely how most people work. A good reason not to turn the sound card down is that all of the ones I've seen make no provision for 24 bit dither. That can make for harsher sounding monitoring which could also lead one to make different eq. choices in mixing that only sound the way intended when listening to the one monitor chain.

As Paul demonstrated, peaking to zero buys you nothing and can only create problems. Maybe the key is understanding that it's generally better to error on the low side in the digital audio world than the high side. It's the exact opposite of analog audio. An awful lot of us have learned this lesson the hard way over the past (yikes, it's really been!) twenty years.

[Key]

I don't think there is a need to dither to 24-bits on my card since the DSP is at 64-bit float.

Personally I can't hear any difference between 0dB on the soundcard output with -6dB on the DAW vs -6dB on the soundcard and 0dB on the master fader. I think any difference perceived might be imaginary in my system. Now put an analog attenuator in the path instead and there are all sorts of things that can go wrong. I am not even sure if it's not just a digitally controlled attenuator on some soundcards DSP and not even sacrificing anything in the way of digital res.

Anyway I don't try to get everything really close to 0dB don't get me wrong. I am just saying that I think it's equally wrong to tell someone to never go above -3dB or -6dB without actually running some tests on the specific cards - it really confuses the point and is cause for all sorts of superstitions.