Question about 24 bit recording levels

[algroves]

Interesting. So how does this relate when people saying things like "24bit has loads of headroom" etc.

 

Trying to get a handle on this in my head, cheers

 

Al

[algroves]

This is what I thought.

 

Whats even better is that I'm actually doing a paper on ADDA conversion (amongst other things tomorrow, so this is a great way to revise!)

 

So when I look at a manufactuers website and it says "Input/Output level at 0 dBFS @ +4 dBu: +13 dBu" does this mean that I should be targetting my levels to that particular converter at -13dbFS in Pro Tools meters?

 

How does a piece of equipments dynamic range relate to this?

 

Thanks muchly.

 

Al

[TBush]

Originally posted by MASSIVE Master You realize that "just under 0 (dBFS)" is around 18dB into the headroom of your front end, right? If you're trying to treat it like analog (which you should) the "meat" of your levels should be around 0dBVU -- Somewhere in the realm of -18dBFS.

 

Yeah- what I meant was that the peaks would be "just under 0 (dBFS)"- digital has great headroom and no need to blow it there...

[where02190]

Originally posted by TBush Yeah- what I meant was that the peaks would be "just under 0 (dBFS)"- digital has great headroom and no need to blow it there...

 

NO, that's not correct. Again, slamming into the converters as close to 0dbfs is NOT the way to record into digital. Target your nominal, that means average input level at your converters 0dbu reference. that will leave more than sufficient headroom, typically around -6dbfs. Think about it, 24+ tracks, all peaking at -1dbfs, slamming into the summing buss....+18dbu x 24+ tracks....can you say distortion?

 

There is absolutely no need to be peaking at or even near 0dbfs. all you're doing is eating up your headroom, meaning your faders actual or virtual) will be at the bottom of their throw to prevent summing smashing, leaving you little range of movement.

[Kiwiburger]

(Donning flame proof suit ...)

 

OK - here's my take on it:

 

Digital converters (A/D and D/A) are primarily Analog devices. Think about that for a minute. Analog.

 

All Analog devices are designed with good headroom above 0dBU. That is because VU meters display a slow average level, and there are plenty of transients in audio that temporarily exceed 0dB but don't move a VU meter.

 

So in the analog realm, right up to, and including, the a/d converter, you must maintain headroom. The same applies coming out of the digital realm - although this is flaunted for the purpose of loudness (at the expense of distortion) even by mastering engineers these days.

 

I disagree with some of the terminology in some comments above:

 

Most daws don't have summing amps that can be overloaded. It's just mathematical calculations - so this statement is a little like saying you can overload a pocket calculator. Sure - you can exceed the significant number of bits, and end up with some small rounding errors, but this is not a practical concern.

 

Any damage/distortion that you might hear from running digital too hot is usually ocurring in the analog realm of your converters.

 

It is completely possible to make audio files that exceed 0dB. These will sound extremely clipped because you are hearing them through fixed bit converters. You can save the file, reimport it, drop the master fader, and the distortion goes away again. This is because the file was not actually clipped at all - you just thought it was, because of what your d/a converter was doing.

 

The statement that "all bits are used at 0dBU" is true and misleading.

 

If you record total silence at 24 bits - all 24 bits are used.

If you record total clipping at 24 bits - all 24 bits are used.

If you record at perfect 0dBU at 24 bits - all 24 bits are used.

 

That's not very helpful information.

 

The important factor is called the "Apparent Bit Depth". For example - you can convert a 16 bit file to 24 bits. You haven't improved the sound quality - you've just bulked it up with extra bits that do nothing. Some software can analyse a file and tell you the apparent bit depth. The more bits that are "used", the better the sound quality.

 

This is academic hair spitting. It should be obvious to any sensible person that the deeper the apparant bit depth, the better the audio quality (I like to just say "resolution" for short, but that upsets the academics).

 

I think it's a case of old school vs new school thinking. Both are correct, in there correct realms.

 

If you apply analog reasoning to digital, you screw up.

If you apply digital reasoning to analog, you screw up.

 

If you fully understand that digital converters are analog devices, you shouldn't screw up.

 

And don't forget that some DSP algorithmns are designed to model analog, and therefore you may need to apply analog reasoning in those cases. (E.g. saturatation plugins).

 

Flame away ...

[The Chinese]

Originally posted by Kiwiburger (Donning flame proof suit ...) Flame away ...

 

Nothing really to flame in your post.... however, It's a bit advanced for the original poster, which is why I kinda just alluded to it in my earlier post.

 

I personally think that 32bit floating math is one of the bigest reasons that our engineering skills are falling by the wayside. I mean, you can keep tossing track after track after track with almost no thought. No longer do we have to meticulously record our tracks etc.

 

bUt it all comes back to that converter....0dBFS is a real ceiling. Why not just record correctly and not pull your master down so far?

 

my .02

 

-Todd A.

[MASSIVE Master]

Originally posted by TBush Yeah- what I meant was that the peaks would be "just under 0 (dBFS)"- digital has great headroom and no need to blow it there...

I have to say that the only time I've ever had a peak near -0dBFS while the meat of the signal was riding at 0dBVU was when the microphone fell onto a wood floor.

 

I'm with NPRS... Peaks even *approaching* -0dBFS should be cause for rechecking input levels.

 

But I've been around long enough to know that I'm not going to change everyone's mind on this... Although it still keeps me in almost constant wonderment that I ever had to "change" anyone's mind about proper levels...

 

And you're right to a point - Digital *does* have a phenominal amount of headroom - More than I ever even imagined would ever be possible "back in the day" -

 

So, then question is - Why are you trying to use every single bit of that headroom up? During *tracking* no less (where you have the most to lose from poor gain staging).

 

And here's something else I have an issue with (although I'm sure it's well-intended):

 

If you apply analog reasoning to digital, you screw up.

 

If you properly structure your gain, and you use your gear the way it's designed to work, keep your recording levels around 0dBVU, etc., digital will work exactly as it was designed to work.

 

But it certainly doesn't work the other way around -

[TBush]

Originally posted by where02190 NO, that's not correct. Again, slamming into the converters as close to 0dbfs is NOT the way to record into digital. Target your nominal, that means average input level at your converters 0dbu reference. that will leave more than sufficient headroom, typically around -6dbfs. Think about it, 24+ tracks, all peaking at -1dbfs, slamming into the summing buss....+18dbu x 24+ tracks....can you say distortion? There is absolutely no need to be peaking at or even near 0dbfs. all you're doing is eating up your headroom, meaning your faders actual or virtual) will be at the bottom of their throw to prevent summing smashing, leaving you little range of movement.

 

Uh, maybe when I'm looking at the little meter in Logic and it doesn't go into the red then I figure "great- I'm just under 0db and I have headroom." Besides, those busses at the master section all have meters too, and if you're summing too hot, bring the levels down at the mixer stage, right? I mix through a digi board anyway, so I can see what's going on at every stage. Of course if one is smacking the "0" a lot, that weird digi distortion is audible. So don't do it!

[TBush]

"...and if you're summing too hot, bring the levels down at the mixer stage, right?"

 

I meant to say "bring the levels down on the individual channels."

 

That's all...

[where02190]

I mix through a digi board anyway, so I can see what's going on at every stage.

 

And here is the big misunderstanding. What you don't see on most digital interfaces is what is going INTO the mix buss, the metering on the mix bus is POST FADER.

 

Leave that master fader at 0, and see what your meters say. That's what's going into the mix buss.

 

Pulling down the master fader 10db to compensate for poor fader positions which result because of recording too hot does NOT sound the same as recording proper nominal 0dbu inputs, leaving the master fader at 0, and adjusting the individuail track faders to prevent overshooting 0dbfs on the mix buss.

 

While on the mix buss subject, again that same 0dbu reference should be observed for nominal levels going out of your DAW as well. Running your mix buss at or near 0dbfs leaves little to no room for the mastering engineer. (I'm sure Massive will have a comment on that as well.)

 

I would love to see DAW and console manufacturers that have abandoned PFL on the mix buss to return that feature. SO many times I see engineers with all the faders at 0 and the master pulled way back, and the mix is just painful to listen to. Reverse that, keep the same output level on the mix buss, but lower track faders with the master fader at 0, and the mix opens up and sounds IMHO far better sonically.

[Kiwiburger]

I realise what I said could be misunderstood ...

 

FWIW - i'm not recommending tracking too hot ... if you understand what I mean about the A/D converters being analog, you will understand that I am recommending tracking at moderate levels (nominal 0dBU for the converter, typically around -18dBFS).

 

If you also understand that the D/A converter is analog, you won't want to smash your mix into the top 6dB range either ... even though most mastering houses seem to do this on purpose these days.

 

So i'm really saying the same this as Where and Massive. But where I slightly disagree is the continuation of analog thinking once inside the box. To the point that many people think they can't use the full range of their track faders - which is daft.

 

There is no reason to be afraid of using the master fader as required.

 

There is no reason to worry about digital channels exeeding 0dB, not until you reach the D/A converter.

 

Although the exception would be certain plugins that are designed to model analog. You will hear the clipping if that is the case.

 

My basic principle with digital stuff is to only attenuate, and try to avoid boosting, at all costs. That's because when bit depth is lost, it's lost for good.

 

With analog, you can attenuate then boost, attenuate and boost, numerous times, and get away with maybe some extra noise.

 

With digital, if you attenuate - you throw away apparent bit depth. If you boost, you don't regain the lost bit depth. So if you attenuate and boost too many times, you could end up with very grainy waves.

 

That's why I think that it's healthy to have strong levels on each track, which will be attenuated later. And for the sake of processing, I think the attenuation is best done after the processing. I see (and hear) no problem with lowering the master fader.

 

It's just the expression "overloading the summing amps" that bugs me when talking about internal digital audio paths. I don't believe it's possible.

[where02190]

There is no reason to be afraid of using the master fader as required.

There is if you're turning down the master fader to compensate for overshooting 0dbfs because track faders are too hot, slamming the summing. It has been proven over and over that, while there is considerable more headroom into the summing than in an analog console, it is not infinite, and there is clear sonic and dynamic difference when paying close attention to what's going into (PFL) the mix buss.

 

There is no reason to worry about digital channels exeeding 0dB, not until you reach the D/A converter.

 

0db no, 0dbfs most definitely. excedding 0dbfs makes for a noise that is absolutely not musical in any sense.

[Kiwiburger]

You still don't want to understand this, do you?

 

Within the digital domain, you can safely exceed 0dB Full Scale, with zero damage to your audio wave.

 

I'm talking about any semi-decent DAW, like Cubase SX3. Read the makers literature about 32 bit floating point files.

 

The damage only occurs when you reach an analog converter and clip that. Or, if you use a plugin or other software application that uses fixed point data, or is designed to clip to emulate analog.

 

This is so easy to prove, that i'm surprised at experienced engineers who try to state otherwise.

 

The headroom isn't infinite, but it's so huge as to never present a practical problem.

 

I'm not saying abuse it. I'm simply saying don't be afraid to use all the available resolution of your tracks, and there is no reason to be alarmed if your master bus goes above 0dbFS prefader.

 

There is no need to panic and upset your mix by trying to lower all your track faders. Simply lower the master fader so your external analog converter does not clip. Or so your rendered file doesn't clip.

 

If you are rendering to 32 bit float (which I recommend) you can leave it 'clipping' (because it's not really clipping at all).

 

The reason I strongly believe this is,

 

A - because it's true

B - because i've proved it

C - because with digital audio, you never want to throw away bit depth if you don't have to

 

 

Everthing changes once you decide to hit an analog converter. That is when you want to optimise the sound.

 

(Or not, if your a friggin Mastering Engineer, you will probably want to clip your D/A converter on purpose).

 

How do you think the top mastering engineers manage to clip their converters? You have to exceed 0dBFS to achieve that. You don't want to be apply 12dB of useless empty zero digital makeup gain to achieve that. You want the full resolution available.

 

And before i'm flamed for promoting squashed dynamics - this is not the issue at all. I'm talking peak levels (dbFS) - not average levels. So this could be highly dynamic material i'm talking about.

 

I'm just wanting to expose what I see as faulty rhetoric about digital. Read the makers manual - and run some tests to prove it to yourself.

[where02190]

Within the digital domain, you can safely exceed 0dB Full Scale, with zero damage to your audio wave.

 

If you like the sound of nails across the chalkboard I guess.

 

I suggest you post some clips of your "music" that was recorded exceeding 0dbfs. Square waves are so pretty afterall, so nice and, well square.....

[wooden]

I also took where'2 advise with really good results.

 

just a couple questions...

 

so we are talking more or less -18dBFS RMS and aroung -6 dBFS peak right?

 

when i see my daw (cubase) meters they are peak, so if i shoot for around -6 dBFS in those meters ill be ok?

 

 

as i understood headroom until now was the diference between the strongest peak in a signal and the point and wich it distorts right? but how do you define headroom of a piece of gear, i mean i though the headroom of a pre amp was increased as i pulled down the gain and viceversa. But when i read in the specs that some gear has x dBU of headroom i dont really understan...

 

plus my emu 1616 specs says that the line inputs has a max level of 20 dBu does that mean that 0 dbU is -20 dBFS?

[Kiwiburger]

This is so frustrating ...

 

What part of "you cannot clip a 32 bit floating point file" don't you understand?

 

Anyone can prove this - not that there is any need to prove this. It's a simple digital audio fact.

 

Create a sinewave that peaks at 0dBFS. Apply a 12dB boost.

 

YES - you will hear square wave clipping, as you are overdriving the FIXED point 24 bit converters.

 

Export that 'clipped' wave to a 32 bit floating point wave file.

 

View it in a wave editor - and it will probably look like a square wave.

 

Import this 'clipped' wavefile into your DAW.

 

Lower the master fader by 12dB - the clipping goes away. You are left with a perfect sinewave. This is because the wave file was never clipped - you only thought it was because of what your converter was doing.

 

I'm not defending overly hot, squashed mixes. I'm just saying you can't clip a floating point file. And therefore there is no reason to waste bit depth, or be worried about lowering the master fader as required.

 

Like I've said numerous times - once your audio has to go through a converter, you then have to start apply analog reasoning again - and leave sufficient headroom.

[Anna Log]

You think you are frustrated huh?

Try on my shoes for a spell.

[MASSIVE Master]

Let's get ahold of ourselves here -

 

It doesn't matter a rat's ass if you can or can't clip a 32bF file.

 

This isn't about digital levels - This is about *recording* levels.

 

And NO (to a previous post) you're NOT supposed to shoot for -6dB peaks. You're supposed to shoot for a 0dBVU average level and let the peaks fall where they may (which will probably be lower than -6dBFS, which is WELL into the headroom of whatever you're using on the front end).

 

And the post with the interface that has a max level of +20dBu... You should be recording *even lower* - WHICH ISN'T LOW - IT'S FINE. There is nothing wrong with recording at these levels - This is what digital was designed for. This is where 24 bit (and 32bF) comes in handy. +20 is not a lot of headroom on a piece of analog gear. I'd play it safe and shoot for "steady" levels around -20dBFS and go from there.

 

You have everything to gain by recording where your gear is designed to run, and everything to lose by overdriving it.

[myfipie]

John,

 

You rule sir. Ever since you told me about this my recordings have improved 300%. It is so easy to understand but for some reason not so many people get it (or should I say do it). Maybe this is still thinking from the 16 bit Adat days?

I do want to say I do not disagree with any of the other points, just agreeeeeeee with you 10000000000000%.

 

Glenn

[MASSIVE Master]

Only 300%?!?

 

Something is still wrong then...

 

Well yeah, back in the day with crappy converters, I'd go *a little* hotter - But bringing preamps into the overdrive stage always bugged me...

 

BTW - Love the traps. This place is flat like a freakin' pancake.

 

I shouldn't say that - I haven't had it shot yet. But it is definitely steady - You can go around the entire room and not hear a dramatic change in the sound. Even in the corners.

[where02190]

Originally posted by gsHarmony By average level, do you mean RMS, or do you just mean that on average the signal should be around -18 dbfs?

 

Your average/nominal signal should target your converters 0dbu reference, which, if you don't know it, -18dbfs is a good place. This reference varies depending on the interface, from between -20 t0 -12dbfs. Check with the manufacturer of your ADDA for specifics.

[Kiwiburger]

I absolutely agree 1000%. Because whenever we are talking about converters, we are talking about the analog domain.

 

My comments about the digital domain were simply because I take issue with the idea brought up that you can "overload a digital summing buss". You can't.

 

Sure - if you go waay over 0dBFS you get 'rounding errors', as the floating point moves. But that's no worse than throwing away bit depth resolution - which is a similar loss.

 

Neither will be audible for practical purposes.

 

I just think it's useful to know exactly where the distortion is occuring. Filling people's heads with the nonsense idea that the distortion is occuring in the digital summing bus doesn't help the understanding of the problem.

[wooden]

since i am going directrly into my 1616m i dont have good metering options. Can i insert this voxengo's free virtual VU meter in the track to measure it? is there any other better way?

[Anna Log]

where...

I would like to get out of your hair with this topic and on to another...

thanks for your patience.

My Yamaha AW4416 has a rinky dink little meter that reads 0to-60db in normal mode, and from 0to-26db in "fine" mode.

You guys are saying to stay close to -18dbfs. On this unit, when I am around -18 db on my meter, I am still one tick below orange. It just seems soooo low. But I have to say I think it sounds better down there. Is -18db the same as -18dbfs? I am so confused by the whole thing at this point that I'm not sure if I am coming or going.

 

Please understand, I haven't spent my whole life doing this, just chasing the chance to. It is hard to understand all this {censored} when the last time I recorded we were slammin to tape, and I haven't kept up on all the details of this digital monster thingy.

I have been putting my little studio together for years now and I finally have my material together and I just want to get it as right as I can.

I certainly appreciate the help you and everyone else have given.

[Kiwiburger]

Originally posted by wooden since i am going directrly into my 1616m i dont have good metering options. Can i insert this voxengo's free virtual VU meter in the track to measure it? is there any other better way?

 

No - a digital meter is obviously post-a/d converter. It's just your analog stuff that needs to be set up around nominal 0dBVU. Up to, and including, your A/D converter.

 

Some of this advice is more academic than useful. As long as you don't clip your a/d converters, or seriously under-record, you can't go too wrong.

 

If you understand that a/d converters are an analog device - the more you push them, the more distortion you get (within it's analog electronic circuit - not talking about digital clipping, running our of numbers.

 

So the purpose of lowering your gain structure and not hitting the a/d so hard is simply to avoid unwanted analog distortion.

 

Consider that many people find that DAW recordings are waay too clean, and apply saturation plugins and all sorts of tricks to dirty up the sound and make it more analog ...

 

So forget your meters and trust your ears. Maybe you will like the analog distortion of your converters (not talking about clipping at all).

 

Also, your preamps and compressor (if used) have similar analog distortion characteristics. Don't blame your converters for distortion that might be occuring before it gets to them.

 

Ultimately - there is a wide range of taste in sounds. Some people buy certain gear to get clinical, sterile, accurate sounds. Others buy certain gear to get warm, distorted, character sounds.

 

I think it's very useful to know exactly what each piece of gear you have is capable of. Abuse it, make it distort, back off, see how clean it gets. Then you will know it's "sweet spot", or at least know it's weaknesses.

 

Understanding each link in the chain let's you make decisions that help you get the sound you want.

 

Frankly, I think VU meters are fairly useless. Sometimes to get the sound I like, I'm pegging a VU meter and I feel sorry for it. Othertimes, to get the clean sounds I want, the VU meter is hardly moving. Fairly irrelevant in my opinion.

 

It's a new digital world - peak meters are more important. That probably offends the old school, but let it be.