Please correct my confusion on digital recording levels.

[umkcprof]

Hi all . . .

 

All of a sudden I'm a bit confused on recording levels. I have always thought I should record the individual tracks as hot as possible without clipping. However, on a recent (okay, last year!) posting, the advice was to record with peaks in the -6dB range.

 

Now, I can understand why one would want to record the mix at a lower level (no room for headroom, etc.), but would that apply to the individual tracks, as well?

 

Thanks so much.

 

Stephen

[toddlans]

It was fairly common I guess to slam levels to things like DAT or cheap 16bit technology to not lose any resolution. I never used DATs so I don't know. If you're using something 24bit there's no need to slam the levels in fact you'll be adding subtle (or not so subtle) distortion to your signal. It's a good idea to shoot for your convertors nominal 0 dbVU reference (should be found in your manual or on the company's website). Then again maybe you want that slight distortion. It is art so its really whatever sounds good to you.

 

Here's a recent link talking about it, there are several probably but here one.

 

http://acapella.harmony-central.com/forums/showthread.php?s=&threadid=1242887&highlight=nominal

[Kiwiburger]

With 16 bits, recording as hot as you could without clipping probably made more sense. Now with 24 bits, you can afford to relax and give yourself more headroom.

 

These discussions usually go pear shaped, because the old timers like to talk about dbVU, and the young timers like to talk about dbFS (because DAWs have peak meters, not VU meters). The Katz system is gaining in popularity too, so you'll get people talking about K-20 etc.

 

Working backwards - you want to create a mix that still has some headroom - maybe peaking at -6dBFS. There is no point tracking all your tracks this hot, because they will have to be attenuated for the mixdown. Although you could use compression, with no makeup gain, to get the track levels low enough.

 

But the main reason to leave yourself headroom - imo - is because analog converters are analog devices with a sweet spot designed around nominal 0dBVU - which might be around -18dbFS.

 

IMO - cheaper converters are more prone to this problem, but try this experiment to hear for yourself.

 

Import a sinewave into your DAW. Either clone it or apply clean gain so that you can exceed 0dBFS on the master bus. Listen to the sound as your slowly raise the fader. The sinewave should sound like a pure sinewave, but on some converters, as you approach 0dBFS it gets "crunchier" and you can hear the harmonics being added. (You can see them on a spectrum meter, because they are being added in the analog side of the D/A converter). Once you exceed 0dBFS, you hear the much harsher digital clipping. That's not the coloration i'm talking about - i'm talking about the finer coloration before that point, in the top 6dB range.

 

OK - if you hear this, you can see that it's not wise to push your audio into this range (even though it's done on most commercial CD's). With really high quality converters, like my Benchmark DAC-1, there is very little coloration right up to 0dBFS. But i'm talking about real world converters, and the converters in most affordable recording devices.

 

This experiment is about D/A. But i'm suggesting that equivalent A/D converters will have similar coloring - especially the cheap ones.

 

So imo it doesn't make sense to "track as hot as you can without clipping" - but to aim for the nominal 0dBVU of your converter.

 

Which of course makes no sense if you don't have a calibrated VU meter, and are just looking at your DAW peak meters.

 

Many people have preamps without VU meters - and I don't think they add much value.

 

IMO - aiming for peaks around -9dBFS should be fine.

[toddlans]

For reference i'm one of the young timers, I don't own a VU meter, I'm usually thinking dbFS.

[Phil O'Keefe]

But the main reason to leave yourself headroom - imo - is because analog converters are analog devices with a sweet spot designed around nominal 0dBVU - which might be around -18dbFS.

 

Sometimes it's as high as -15 dBFS, but it's usually going to be somewhere in that range. IIRC, on the AW4416, it is -18 dB. That's where you want to put your average (RMS) levels - some peaks going above that point are to be expected, and should be fine as long as you are not hitting 0 dBFS and clipping.

 

Printing levels as hot as possible without peaking might have been a better idea back when all we had was 16 bit converters, but if you're recording at 24 bit / 44.1 kHz or above, setting your levels so they average out at about -18 dBFS to -15 dBFS should give you ample level, great resolution and not overload the output (stereo) bus when combined at mixdown.

[Geoff Grace]

Oddly, someone else has just started an almost identical subject in another forum I frequent.

 

Here's the other thread in case it's of interest:

 

Should I Stay at 0dB Max or Wade in Yellow Levels when Recording Digitally?

 

Best,

 

Geoff

[where02190]

It's your nominal levels you want to watch, targeting them to the 0dbu reference of your converters. Typically this will leave at least 6dbfs of headroom over 0dbfs.

 

Recording all your tracks to 0dbfs leaves no headroom, and you either a) end up driving the hell out of the mix buss or b) with all your track faders pretty much at the bottom.

 

When mixing, leave the master fader at 0, since most all DAWs master meters are post fader and they have no pfl, so you can accurately see what is going into the mix buss. Don't put plugins on the master, as these will give you an inaccurate reading of what's going into the mix buss. Keep that mix buss levels nominal to the 0dbu reference of your converters. Don't worry about the overall volume of your mix, that's what mastering is for. Concentrate on getting a cohesive mix, and be aware of proper gain staging and levels throughout.

 

This will allow you plenty of headroom if you want, to smash the hell out of it in mastering, removing all the dynamic feel of your track, but it'll be louder than everything else. Conversely, it will give you plenty of room to work with in mastering without having to smash the hell out of it, and your music will breath naturally as it should.

[JeffLearman]

A few points:

 

1) There is absolutely nothing wrong with having faders at low levels in a digital mixer.

 

2) I would imagine that soundcards are designed to be linear up to 0dBFS. However, even if they're not (and Kiwiburger's test is a good way to tell), and instead are linear only up to -6 or -12 dB, that does NOT mean that you don't want signals to go above -6 or -12 dB. Just as with an analog mixer, you want NOMINAL levels (not peak levels) in the sweet spot. If peaks go above the linear range, that's fine. (If you're recording a very nondynamic signal, like some keyboard pad and string sounds, THEN you might want to take this "sweet spot" into consideration and peak a bit lower than normal.)

 

I suppose it depends on how much distortion you get as you approach 0dBFS. If you get a lot, you might want to consider getting a better soundcard, because commercial mixes are really going to sound bad! (As Kiwiburger points out.)

 

While I'm disagreeing with Kiwiburger's bottom line advice, it's still a great post.

 

3) With DAT units, it was more important to leave headroom, because (according to Katz) most of these units had less than perfect clip detection. This is also true of some DAWs, so test your gear to see if it detects all peaks, by recording a very short spike that the DAW reports going to nearly 0dBFS but not clipping, and then inspecting the wave file to see if there are leveled off flat peaks.

 

4) I'm convinced that leaving "headroom" in a mix is a myth. Mixing to peak at -6dB for a 24-bit mixdown is just wasting one bit per sample and delivering less information to the mastering engineers. Leaving headroom after mastering to 16 bits certainly isn't a common practice -- in fact, the common practice is to go quite overboard and compress the snot out of a lot of music. Yes, that's just a bad trend in mastering in some genres. But if it was so important to leave headroom after mastering, it wouldn't work at all.

 

I might be wrong with regards to high-end converters, though. Katz points out that peaking at 0dBFS causes peak levels to actually go well above the corresponding level in the analog gear. I don't know whether this is true for delta-sigma hardware; Katz doesn't qualify his statements with type of hardware and some of the info in his book "Mastering Audio" apply to older hardware designs. Anyway, if it's true that peaks that go to 0dBFS can go over the equivalent analog level, and if the analog gear downstream can't handle those levels, then it might make sense. But then, what do you do when listening to commercial masters? My suspicion is that Katz was correct at the time but no longer is, and if your setup can't play back commercially mastered material without problems in the analog chain, then you need to recalibrate. Or get a better soundcard.

 

In any case, you should mix to 32-bit floating point format and submit that to the mastering engineers. In that case, your peak level doesn't make ANY difference in the quality of the 32-bit mix or the ability of the mastering house to use it to its fullest advantage. The only affect your mixing peak levels have is on your monitoring chain, or on any effects you have on the master channel. For these reasons, though, it's best to mix to the optimum levels -- because you want to hear what you're mixing, of course. If your ears tell you that this level is -6 dB or nearly 0dB or even -20 dB, and if you trust your ears, then go with that.

 

If anyone can point to an authoritative and ideally quantitative article showing why we should leave headroom in mixes, I'd like to see it. The only thing I've ever found was Katz's caveat about pushing a soundcard and its consequences on the analog circuitry in the soundcard and the downstream analog chain. That applies to monitoring, but not to a digital mixdown to submit to mastering. And I suspect (and I HOPE) it's dated. Note that I'm not disputing that overshoot happens, because it's supposed to. But the analog circuitry in the soundcard should be designed to handle this.

 

Cheers

Jeff

[JeffLearman]

 

Originally posted by where02190 It's your nominal levels you want to watch, targeting them to the 0dbu reference of your converters.

 

Good point.

 

Typically this will leave at least 6dbfs of headroom over 0dbfs. Recording all your tracks to 0dbfs leaves no headroom, and you either a) end up driving the hell out of the mix buss or b) with all your track faders pretty much at the bottom.

 

Of course, there's nothing wrong with (b). If you can't adjust your faders well regardless of their level, get a DAW that works better.

 

When mixing, leave the master fader at 0, since most all DAWs master meters are post fader and they have no pfl, so you can accurately see what is going into the mix buss.

 

I say, know where in the signal chain the master fader and master channel meters are, and work accordingly. But if you don't know and don't want to bother finding out, this is good advice. Folks who use n-Track needn't worry about this problem.

 

Don't put plugins on the master, as these will give you an inaccurate reading of what's going into the mix buss. Keep that mix buss levels nominal to the 0dbu reference of your converters. Don't worry about the overall volume of your mix, that's what mastering is for. Concentrate on getting a cohesive mix, and be aware of proper gain staging and levels throughout.

 

Golden advice.

 

This will allow you plenty of headroom if you want, to smash the hell out of it in mastering, removing all the dynamic feel of your track, but it'll be louder than everything else. Conversely, it will give you plenty of room to work with in mastering without having to smash the hell out of it, and your music will breath naturally as it should.

 

No argument, except that there's no reason to leave unused bits in a digital mixdown (which is what "leaving headroom in a mix" means). But if you use 32-bit float format, the level doesn't matter anyway, so you might as well -- you'll waste no bits and lose the least information in the mixdown process. If your mastering engineer doesn't understand 32-bit float format, clue them in or find one who does.

[Kiwiburger]

I don't understand what appears to be self-contradicing advice Learjeff.

 

You say there is nothing wrong with leaving faders at low levels, and then go on to talk about not wasting bits (which is an outdated idea that should be shot to pieces). I used to think that way, and used to track too hot. I got better.

 

The advice about aiming for nominal 0dBU is totally correct. The problem is many people don't have VU meters, and are trying to guess where this might be in relation to peak meters calibrated in dbFS.

 

But as far as shooting down this wrong idea about bit resolution - these are some thoughts that helped to change my mind:

 

Comparing digital audio with digital pictures is misleading, and fuels the idea that preserving bits is more important than it really is.

 

The difference? With pictures, every bit counts. The more bits, the more pixels, the better the resolution, the less grainy the picture.

 

With audio - you never hear audio as discrete bits. Audio is a continuous waveform, which is created by the D/A converter. If you imagine a sinewave - you can represent this sine wave with a few bits, or a few thousand, and it makes no difference to the end result: a sinewave.

 

Audio can be considered to be comprised of nothing but sinewaves - the most complex harmonics can be analysed and broken down into sine waves. Obviously the frequency range that the converter can reproduce is important - and that is determined by sample rate (not bit depth).

 

Bit depth only determins the noisefloor of the audio - and with 24 bits available, this is the least of your worries compared to the analog circuitry.

 

So what this means is that trying to squeeze your audio into the maximum bit depth is simply not necessary, and you are more likely to "burn" your audio in the process. Mainly in the analog realm, but many plugins are modelled on analog and therefore color your sound if you get too hot.

[where02190]

Faders at the bottom of their travel leave little room for adjustment/automation, and severly limit your ability to adjust them within the parameters of the mix.

 

It's not a myth, you MUST leave the mastering engineer some headroom, or he will simply reduce the gain of your 2 mix in order to recapture headroom to master, which will deteriorate the quality of the mix. Since he has no choice, he is not responsible for this deterioration, you are, because you left him no headroom.

[modul8]

learjeff,

 

Read this thread, particularly starting around page 5 or 6 on. Pay close attention to what Paul Frindle and some of what Bob Katz say. Very important information regarding digital levels and potential problems with undetected peaks from processing signals that are as low as -7 dbfs.

It is a lot to wade through but, extremely informative. He outlines an experiment to show how an EQ plug can cause undetected problems.

Bottom line, if you want digital to sound good, leave headroom.

 

http://recforums.prosoundweb.com/index.php/mv/msg/4918/0/0/0

[JeffLearman]

 

Originally posted by Kiwiburger You say there is nothing wrong with leaving faders at low levels, and then go on to talk about not wasting bits (which is an outdated idea that should be shot to pieces). I used to think that way, and used to track too hot. I got better.

When mixing on a program that uses 32-bit floating point (or better yet, 64-bit floating point) accumulation (that is, "mix bus"), you do not lose nearly as much information during accumulation as you would if you recorded at a lower level to begin with. To understand this you need to understand floating point math. I could give a detailed example if you want to be bored to tears.

 

The advice about aiming for nominal 0dBU is totally correct. The problem is many people don't have VU meters, and are trying to guess where this might be in relation to peak meters calibrated in dbFS.

True; good point.

 

 

Comparing digital audio with digital pictures is misleading, and fuels the idea that preserving bits is more important than it really is. The difference? With pictures, every bit counts. The more bits, the more pixels, the better the resolution, the less grainy the picture. With audio - you never hear audio as discrete bits. Audio is a continuous waveform, which is created by the D/A converter. If you imagine a sinewave - you can represent this sine wave with a few bits, or a few thousand, and it makes no difference to the end result: a sinewave. Audio can be considered to be comprised of nothing but sinewaves - the most complex harmonics can be analysed and broken down into sine waves. Obviously the frequency range that the converter can reproduce is important - and that is determined by sample rate (not bit depth). Bit depth only determins the noisefloor of the audio - and with 24 bits available, this is the least of your worries compared to the analog circuitry.

This last bit is incorrect. Bit depth also governs dynamic range: the difference in level between the quietest information and loudest. For example, with a greater bit depth, the loudest passage can be much louder than the quietest. However, much more importantly, with greater bit depth, the loudest sine wave in a complex signal can be much louder than the quietest sine wave. This is crucially important, and is why 16 bit tracks sound incredibly better than 8-bit ones.

 

You're correct that greater bit depth also increases the S/N ratio. Just a minor point here, though: What we often think of as the "noise floor" is essentially zero for digital -- that is, silence is truly silent. However, the more important "noise floor" for digital is the level of the quantization noise (which you do get whenever there IS a signal), and in that sense you are completely correct. So, I'm not correcting you here, I'm agreeing -- just clarifying what might seem incorrect to onlookers.

[JeffLearman]

 

Originally posted by where02190 Faders at the bottom of their travel leave little room for adjustment/automation, and severly limit your ability to adjust them within the parameters of the mix.

This is a valid point. However, for me it's only a problem with lots of tracks, and in that case it's best to use groups anyway (which dilutes the problem significantly).

 

If you have two tracks (of uncorrelated sound -- that is, not miking the same instrument), when you add them together, you get a 3dB boost on average. 4 tracks, 6dB. 8 tracks, 9dB. 16 tracks, 12 dB. I generally wouldn't use more than 8 tracks without using groups, and let's add a 6 dB pad for rogue peaks, so that puts me at -15dB. My mixer works fine at that level, but YMMV.

 

Furthermore, if you know how your DAW works, and if it's designed to work the way it should, there should be no harm in using the master fader. I like to avoid it though, but just to simplify rendering "freeze" tracks -- I'm using an earlier version of software that doesn't support this built-in.

 

 

It's not a myth, you MUST leave the mastering engineer some headroom, or he will simply reduce the gain of your 2 mix in order to recapture headroom to master, which will deteriorate the quality of the mix. Since he has no choice, he is not responsible for this deterioration, you are, because you left him no headroom.

However, reducing the gain on a DAW that uses floating point (as all I'm aware of do) is far different than reducing the gain prior to encoding in a fixed point format (e.g., 24-bit wave file).

 

In the first case, you lose a lot more information than in the second case.

 

module8, thanks for the link, which I'll read with great interest.

[JeffLearman]

Oh, evidently PT uses 48-bit fixed point. My argument about leaving headroom in a mix for mastering apply to that case as well -- they apply to any case where the accumulation is better than 24-bit fixed point (or, for 16-bit tracks, 16-bit fixed point.)

 

Concerning modul8's thread, I haven't finished reading it, but Paul Frindle makes a very good case that you should keep master levels below -6 dB for monitoring when mixing, due to soundcard operation on output. (Katz has made this same point, which I think I alluded to above.) This has nothing to do with creating a mix for mastering, and also has nothing to do with record levels. However, it is an exception to overgeneral statements I made above, and is worth paying attention to.

 

Frindle says that you need to keep the peaks below -6dB at every stage. This might be true for fixed-point buses like PT, but it isn't for floating point buses like most other software DAWs.

 

When working with 24 bit tracks, there's really no harm in keeping peaks below -6 as a rule of thumb. Just don't think it's necessary to get ideal results.

 

Someone else posted another exception to my statement that is very good to make note of. They said that they can't reach nearly 0dB without overcranking their mike preamps, when recording a signal with low dynamics. Obviously, overcranking a preamp isn't a good idea. Note that a simple "peak at -6 dB" rule isn't any better to handle exceptional situations, because the same thing could happen at that level (though admittedly not as bad). The only general rule that applies here is to do what sounds best when you check the results.

 

 

[modul8]

At -6dbfs you are still at 24 bits of resolution. If you look at the waveforms of popular music from the late 80's and early 90's (Depeche Mode's Violator and NIN Pretty Hate Machine are examples I know of) the mastered tracks peak around -3dbfs. This is still 16 bits leaving a small bit of headroom for processing during D/A.

Insert plug-ins are usually pre fader (which is a good reason to meter pre fader in your DAW). Hot digital signals can cause problems with plug-ins resulting in crappy, "digital" sounding mixes. Rather than run the risk of accumulating errors over multiple tracks run through questionable plug-ins (EQs and Comps in particular), mix conservatively, leave headroom and both you and your mastering engineer will be happy.

[modul8]

Slightly off topic but, pointing to the issue of headroom and a well mixed and mastered song:

 

DM's "Personal Jesus" through Inspector XL

 

No clipping at all

 

1.12 db of headroom

 

RMS around -16dbfs

 

 

Evanescence's "Going Under" through Inspector XL

 

40 Hid Clip incidents

 

No headroom

 

RMS around -8 dbfs

 

 

Both good songs only "Going Under" left my ears fatigued with it's barely 1-bit of dynamic range. How much better could it be if it had been respected? Kean is the same way, great songs that are distorted all over the place. It's a shame. Do your music a favor and mix intelligently and if it sounds like {censored} after mastering, you know it's not your fault.

 

Sorry for the ramble but, I think it is a good illustration of how digital abuse = big-fat-poo.

[where02190]

 

Insert plug-ins are usually pre fader (which is a good reason to meter pre fader in your DAW)

 

 

On channels yes, on master faders, no. Typically master fader inserts are post fader.

[Brittanylips]

Originally posted by umkcprof Hi all . . . All of a sudden I'm a bit confused on recording levels. I have always thought I should record the individual tracks as hot as possible without clipping. However, on a recent (okay, last year!) posting, the advice was to record with peaks in the -6dB range. Now, I can understand why one would want to record the mix at a lower level (no room for headroom, etc.), but would that apply to the individual tracks, as well? Thanks so much. Stephen

As Model8 suggests, there is some fairly new thought that levels are not what everyone used to think they were. The new insight suggests that even sticking within apparently safe levels (but at the top of those safe levels) introduces distortion you may not see on your meters. Therefore, you will get better fidelity by giving yourself a cushion.

 

By the way, I have noticed that your recent posts suggest that you are apparently just starting to record but are starting with top of the line equipment. Are you a musician who is putting together a studio? Did you win the lottery? What sort of things are you looking to record? I don't mean to pry but if I connect the dots of your recent posts, I am curious about what you are putting together.

 

-peaceloveandbrittanylips

[Zooey]

 

Originally posted by Brittanylips As Model8 suggests, there is some fairly new thought that levels are not what everyone used to think they were. The new insight suggests that even sticking within apparently safe levels (but at the top of those safe levels) introduces distortion you may not see on your meters. Therefore, you will get better fidelity by giving yourself a cushion.

 

 

Yep. There's been a lot written about this in recent years. Basically, you don't need to record hot to capture the entire dynamic range of most sources with complete accuracy.

 

With 24 bit recording, I have never once had to re-record a track because it was recorded too quietly. I can almost hear people screaming, "BUT YOU'RE THROWING AWAY THE BITSSSSSSSSSS." Bull{censored}. Play my tracks in a double blind test and you will never once be able to tell which ones were recorded at lower levels.

[MASSIVE Master]

 

As Model8 suggests, there is some fairly new thought that levels are not what everyone used to think they were.

 

There is no "new" thought about this - 0dBVU has been 0dBVU for over 60 years. Line level hasn't changed. The gear hasn't changed. The operating level of the gear hasn't changed.

 

The only thing that's changed is the inexperience of the people using the gear... Getting "hot" levels in digital has *never* been "standard" in the professional audio community. Using the gear as it was designed to run is the "standard" (for lack of a better term).

 

In the "good ol' days" when you'd want to smack the tape pretty hard, you'd drive the preamps to +4 or +6. Not a big deal for a well-designed preamp circuit.

 

Getting a signal "hot" digitally would require that same preamp being overdriven to +20 or so (depending on how the converters are calibrated).

 

24-bit digital allows you to run your gear at *NORMAL* levels with less loss in fidelity and more headroom than ever before. And yet, a large percentage of people use that headroom as an invitation to CRUSH the signals being recorded.

 

And then they ask why their recordings "aren't as loud" or "don't sound like 'pro' recordings" - It's like shooting yourself in the foot and then wondering why it hurts when you walk.

[Brittanylips]

 

Originally posted by MASSIVE Master There is no "new" thought about this - 0dBVU has been 0dBVU for over 60 years. Line level hasn't changed. The gear hasn't changed. The operating level of the gear hasn't changed.

 

There is some new thought about digital distortion under 0dBVU. I think Dan Lavry wrote something up recently about this - it's floating somewhere on the web. If I have a moment I might be able to find it.

 

-plb

[littledog]

I don't know if "new thought" is the right term, but there is definitely a change in dealing with levels between 16 and 24.

 

When I first started out on 16 bit ADATs, my mixes were critiqued by some pretty reputable mastering engineers, who suggested that I record hotter, since I was wasting a lot of bits because my levels were too low.

 

Now at 24 bit, that's not nearly as much of an issue.

[MASSIVE Master]

Reconstructive distortion isn't a new concept either - I suppose it's great that people like Dan and Nika (Aldrich) are bringing it out into the open, but it doesn't really apply too much to the people who understand their levels -- And it won't convince the people that don't anything else - They're going to record "hot without clipping" until who knows when...

 

Don't get me wrong - I've steered dozens of people into recording at "normal" levels and I usually wind up with long-winded letters with "OH MY GOD!!!" all over them. But a lot of people will never be convinced, no matter how much sense it makes. Many won't even try it as an experiment.

[Zooey]

Originally posted by MASSIVE Master Many won't even try it as an experiment.

 

I like to think of that as a competitive advantage for me!