Keeping your levels low in digital recording/mixing...

[Funkybot's Evil Twin]

Of course, so does people not making full use of their bit resolution...

I was just about to get to this...when recording, shouldn't I want to get as close to 0 without going over, if only to get the best performance out of my converter in relation to bit depth?

I can see why mixing at lower levels helps solve other problems, but I'm curious as to why anyone would think they'd want to record at low volumes. Sure, you're not dealing with tape hiss in digital, but you are dealing with bit depth.

[ahjteam]

Turn up the volume of your system, and mix at lower levels...

So, is that basically it? I mean, there are other things to take into account, but i'm just trying to get it straight in my head:

Mix everything low with your system volume up, and then normalise the end results?

The way i generally do it is try to tame the busses, but make them as loud as i can, this seems like quite the opposite........but i do always end up with 'bricks', so...

To make it simple: If the audio is normalized to 0dbfs, you won't make it any louder, just less dynamic. By doing it this way you will keep the sound dynamic and it doesn't compress the shit out of it. But don't get me wrong, compressors are needed.

[MacZero]

here's my way of putting it: don't mix at insane levels cuz you won't have to muck about to get a mix to give to a mastering guy.

This talk of integer audio data, what proggies use that system? We talking techno-fossils here? excuse my ignorance.

[aciddose]

jens, well yes. when i said "if you have to throw a limiter on everything" i was talking about the imagined need like you said. really the whole chain should have levels set to maintain -6 ... 0 in all places. limiters should never be used for the purpose of pushing the nominal signal levels higher than they can go.. this is simply removing all dynamics.

when i talk about -6db or -12db, i'm talking about nominal signal levels. you're not going to be losing any bits, they're part of your headroom. peaks end up in those bits. if you're mixing based upon peak signal levels what i'm saying doesnt make much sense. (although the thought of using peak levels to mix anything doesnt make sense to me either so we're even)

i never mentioned about integer anything. i was only explaining about the difference between int and float, where with float you do not have to pay as much attention to headroom since you have so much available. i wasnt talking about int vs. float anything. i was making a reference to the different states of mind when using the two formats. if you mix in floating point, you need to force yourself to treat headroom the same way you would as if you were using integer. it'll produce better results if you're more mindful of anything, especially headroom.

it would be nice to have an integer spec.. but this will have to wait for adoption of 64 bit systems since 32 bits is really just not enough for linear ints in audio. having an integer spec would FORCE you to be aware, and in my opinion integer dsp code is generally better than float dsp code (for various reasons) and this would lead to a higher average quality of software, too. that is just something for the future though, it isnt worth arguing the merits at this point since there currently are not many.

[MacZero]

oh ok, sorry for that misread acid (guilty of skimming), cheers for the info.

[The Chase]

Turn up the volume of your system, and mix at lower levels...

So, is that basically it? I mean, there are other things to take into account, but i'm just trying to get it straight in my head:

Mix everything low with your system volume up, and then normalise the end results?

The way i generally do it is try to tame the busses, but make them as loud as i can, this seems like quite the opposite........but i do always end up with 'bricks', so...

To make it simple: If the audio is normalized to 0dbfs, you won't make it any louder, just less dynamic. By doing it this way you will keep the sound dynamic and it doesn't compress the shit out of it. But don't get me wrong, compressors are needed.

How is that possibly less dynamic? If you take a signal that peaks near 0db and listen to it on a system, and take the same signal peaking at -6db but make up for the gain-loss by turning up your monitors, how is the near 0db signal less dynamic? Then only difference would be that it has the equivelent of lowered bit-depth, wouldn't it?

[aciddose]

i think what he may have been thinking when he wrote that though, chase, is of a signal peaking at +6db with a limiter set at 0db. the nominal level would be 0db with or without the peaks, but of course the limited signal would have less dynamic content.

i think a lot of people use limiters that way. i know it seems crazy but i've seen it myself. throwing a limiter on the master bus is great, but it should be activated only very rarely on the highest peak levels. a lot of people i think actually use limiters in a way that keeps them active a majority of the time - which is very bad for the dynamics of *most music.

* - very good for some

[beej]

So the problem with 'hot' mixing is that people do not realize what limiters are really for (changing the dynamics of the program-material) and instead think it's to stop a channel from exceeding 0DB (which in itlsef practially has no consequence at all due to the floating-point calculations of the summing-engine)

Yep (unless we are talking about output channels here).

I partially agree here though - for some plugins it's important that they're feed with a certain input-level (though this depends on the respective plugin - e.g. GCO-1 should be fed with mainly ~ -3 to -2 db)...

Yep.

This might help those having difficulties with understanding loud signals and clipping in a modern, well-designed DAW:
http://www.kvraudio.com/forum/viewtopic ... 29#2482229

[jens]

So the problem with 'hot' mixing is that people do not realize what limiters are really for (changing the dynamics of the program-material) and instead think it's to stop a channel from exceeding 0DB (which in itlsef practially has no consequence at all due to the floating-point calculations of the summing-engine)

Yep (unless we are talking about output channels here).

yup, to be exact what I wrote applies as long as we are talking about

a mixer-strip which doesn't feed a DA-converter

[Bonteburg]

wow ... tried the low level approach and it's doing my mixing a world of good - I can cram tons more stuff in there now and still don't lose dynamics

My default project has everything except the master bus at -6 dB now (if indeed that's how it was meant).

Marco

[ahjteam]

Turn up the volume of your system, and mix at lower levels...

So, is that basically it? I mean, there are other things to take into account, but i'm just trying to get it straight in my head:

Mix everything low with your system volume up, and then normalise the end results?

The way i generally do it is try to tame the busses, but make them as loud as i can, this seems like quite the opposite........but i do always end up with 'bricks', so...

To make it simple: If the audio is normalized to 0dbfs, you won't make it any louder, just less dynamic. By doing it this way you will keep the sound dynamic and it doesn't compress the shit out of it. But don't get me wrong, compressors are needed.

How is that possibly less dynamic? If you take a signal that peaks near 0db and listen to it on a system, and take the same signal peaking at -6db but make up for the gain-loss by turning up your monitors, how is the near 0db signal less dynamic? Then only difference would be that it has the equivelent of lowered bit-depth, wouldn't it?

Sorry, English isn't my primary language I meant normalized signal is more dynamic compared to compressed signal.

My default project has everything except the master bus at -6 dB now (if indeed that's how it was meant).

Mine has at +0db when recording, but when I start mixing, I drop everything to -inf db and start adding the channels one by one (I rarely have even 20 channels in my mix so its really not that time taking)

[bduffy]

wow ... tried the low level approach and it's doing my mixing a world of good - I can cram tons more stuff in there now and still don't lose dynamics

My default project has everything except the master bus at -6 dB now (if indeed that's how it was meant).

Nice! That's usually what I do too, and I often wind up ganging and lowering the faders one more time during a complex mix. Glad it's helping out!

[music-tech]

Of course, so does people not making full use of their bit resolution...

I was just about to get to this...when recording, shouldn't I want to get as close to 0 without going over, if only to get the best performance out of my converter in relation to bit depth?

I can see why mixing at lower levels helps solve other problems, but I'm curious as to why anyone would think they'd want to record at low volumes. Sure, you're not dealing with tape hiss in digital, but you are dealing with bit depth.

This is a common mis-conception. The RESOLUTION has nothing to do with how hot you record, or how many "bits" you are using by recording at different levels. The only thing you lose by recording with fewer bits is dynamic range, which is NOT the same as resolution (in the commonly used meaning of the word). If you record with only a few bits, you are closer to the digital noise floor (yes, digital has a noise floor), which is the noise you get from quantizing/rounding errors, etc. Resolution, in the way that most people think of the term, is more closely linked to the sample rate than the number of bits you use.

But, if you are recording at 24 bits, you have more than enough dynamic range to leave yourself plenty of headroom while recording and never ever "hear" that digital noise floor.

I didn't read all the articles in the forum from the link, but I think that possibly some people are missing the point here. Most converters were optimized to be used with a certain established reference level, common in the analog recording world. We are talking about the analog side of the converters here. If you are RECORDING a source, and always slamming your converters right up to digital zero, then your levels are pushing the limits of what those converters were designed for, and you are probably adding some distortion and other artifacts to your signal on the way in. These converters were designed so that O dBVU on a typical analog console will hit somewhere around -18dBFS on the converters (different manufacturers use different values, but somewhere around -18 seems to be average). That means your average RMS level really should be hitting only -18dbFS on your converters, and the peaks somewhere between -12 to -6, depending on the source signal. That's the optimum signal range for your converters, and is where they will sound the best. Recording with peaks hitting 0dBFS all the time is basically like slamming all your needles into the red all the time on an analog console. Sometimes that may desirable to get a certain type of distortion, but it probably isn't that desirable on all but the best A/D converters.

The same thing is true on the way back out. I think some of this is also related to the way many pros still mix, which is to split out multiple channels through D/A converters and mix through an analog console or analog summing box. If you are pushing all your signals on every channel right up to 0dBFS on the way out through D/A converters, then you are driving those converters (and the following analog gear) at a range that they really weren't optimized for.

If you are doing everything in the box, then once you get past the A/D stage without overdriving your A/D converters (which can make things sound thin and crappy with bad converters), then you still may have to watch your levels on each channel if you are using a lot of plug-ins. Just because your DAW is floating point doesn't mean that all the plug-ins are floating point. So, if you have a fixed point plug-in within your channel chain, it's still possible to clip that plug-in if you are cranking the levels. Other than that, I think you are getting into the math of each DAW and plug-in, and how well the programmers have done their job at programming and handling the signal as it goes from one extreme to another. I'm not going to get into that argument as I'm not a programmer and I haven't done any experiments on my own to test. I started in the analog world, so I'm just used to maintaining proper gain staging and signal levels throughout, and so I never slam any of my tracks too hard in digital land... plus, I do split out most of my channels to an external mixer... and, even though my external mixer is digital, the digital connections are fixed point from the DAW to the mixer, so I have to make sure I'm not clipping any channels on the way out.

Of course, you don't have to believe anything you read on the net. Try it both ways and see which way sounds best to your ears.

[kylen]

That was a really good and clear post music-tech. And the fact that there don't appear to be any veins popping out of your neck while correcting mis-conceptions is just an added bonus!

[kylen]

Ooops - hit send twice after a failure - glad this wasn't a banking transaction!