Recording Audio Levels

[Centrifuge]

I use Ableton to record drum machines and synthezisers etc. Was wondering what level to record at. I was previoulsy using an EMU 1820 and I would push the signal until it was hot but not clipping on the EMUs mixer before it went into Ableton - This gets the signal in Ableton peaking just around the 0db mark.

Now I've just read that you should ideally keep your track meters about -10 to -12db to leave headroom for the mix. Just wondering if I should record the audio to peak at -10db or to just record the strongest signal possible and then lower the fader once it is recorded?

[terminal3]

In the old days (and still today, of course, in rarer cases), in the world of analogue, you'd record hot because the sound of the natural tape compression is extremely pleasing.

In the world of 16 bit digital, when digital recording first started, you needed to record hot because the noise floor was otherwise unacceptable, and it would sound awful.

However, with 24 bit digital, the noise floor is so low and the dynamic range so great (144db theoretical maximum, in reality with prosumer converters max around 110db+), there's absolutely no good reason to not leave some headroom.

The headroom will help you out in the mix! Just lowering the fader doesn't change the bit usage, it only changes the output volume. Having those extra bits will allow you to use plug-ins and even external hardware more successfully, you'll have the 'space' to push things if you need to. If you push your level right up to -1dbFS, for example, and then you add a +1.1db EQ boost on that track, guess what's just happened?

If you need your finals hot (please don't, it hurts everyone*) then kill your dynamics in the mastering. Use a limiter if you must to get things up to the ceiling. But your mix will be able to breathe much easier if you've got that recorded headroom to start with. -10 to -12 is great. Go no higher than -6 in a worst-case scenario! You have space then still to make things loud, and lots of room to make things quiet.

* visit Turn Me Up! | Bringing Dynamics Back To Music for a good cause.

[Centrifuge]

Perfect - Just recorded my first loop @ -12db Thanks

[terminal3]

And just so you know I use commercial loops in production fairly often, and I absolutely HATE when they peak really high. Makes it very difficult to do anything with them!! So good on ya for leaving some space.

[EdRyan]

Yeah I've found that with sample libraries. Their authors seem bent on making their products as loud as possible to grab attention. You can obviously turn it down, but the annoying thing is you automatically itch to pull it back up again because you feel like something is missing..

The other good thing about modern daws is that the faders and controls are more accurate, meaning you have more control on levels. They use a data filtering system to smooth out any amplitude modulation that can occur between intervals. I've gotta re-build a vst plug-in I started last year for a project, and I remember my lecturer stressing how important this is for audio quality.. we A/B'd the same plugin without the data filters and it sounded horrible !!

[Centrifuge]

I take it you wouldn't use the normalization function when rendering? This made my -12db loop into a 0db peaking monster

[terminal3]

Originally Posted by Centrifuge

I take it you wouldn't use the normalization function when rendering? This made my -12db loop into a 0db peaking monster

No. I never use normalisation for anything, ever.

[Centrifuge]

Cheers, will avoid that in the future - peak normalisation bad

[waxxy]

i read an interview with fatboy slim and he normalises everything..he says the slightest thing he records he automatically normalizes it ..in fact a lot of the top dance producers do the same thing they like the slight distortion you get ..me personally when im working with external gear such as akai samplers i normalize everything in there too but vst samplers like kontakt,emulator x i find they make the beats sound too harsh for me so i never normalize there beats id rather use a limiter and compress them until they sound bigger.and sometimes a little bit a chorus can help fatten them up too .

[FrankGIK]

Originally Posted by terminal3

And just so you know I use commercial loops in production fairly often, and I absolutely HATE when they peak really high. Makes it very difficult to do anything with them!! So good on ya for leaving some space.

Originally Posted by terminal3

No. I never use normalisation for anything, ever.

I like the way you think. We're like brothers or somthin'.

Frank

[TrevCircleStudios]

10 to -12 is great. Go no higher than -6 in a worst-case scenario! You have space then still to make things loud, and lots of room to make things quiet.

Absolutely right. I usually average about -18 with -12 peaks while tracking. Here's an excerpt from another thread:

Quote:
"thought my audio should be "hot" ie. going into the red at times without clipping?"

This pretty widespread confusion is caused (in my view) by the difference between the analogue and digital scales. 0dbu (i.e. the analogue scale) equates to -18dbfs (the digital scale). Thus if you are recording at -18dbfs in your daw that is actually 0dbu. Engineers used to push hot signals to tape for that euphonic saturation it gave. This doesnt happen in the digital domain. odbfs is absolutely the highest you can go without clipping. Don't go there. It doesn't sound nice.

Quote:
"I understood that normalising was a natural part of the process to improve the level of the recorded audio.
What is normalisation there for?"

Nope. Not a natural part of the process at all. Normalisation is not there for any good reason at all. All it does it automatically add gain. This leaves no space to EQ, compress or otherwise process the audio before you hit 0dbfs (and remember that is digital clipping - which for the avoidance of doubt does not sound musical in any way, shape or form).


Quote:
"If I record at 24bit, do you recommend I also record at sampling rate higher than 44100 or is this ok?"

90% of professional engineers out there record at target (44.1 for audio). If it's good enough for them...

[modz1]

This should be a sticky.. (What a sticky populated sub-forum eh?!)

[Khazul]

Originally Posted by waxxy

i read an interview with fatboy slim and he normalises everything..he says the slightest thing he records he automatically normalizes it ..in fact a lot of the top dance producers do the same thing they like the slight distortion you get ..me personally when im working with external gear such as akai samplers i normalize everything in there too but vst samplers like kontakt,emulator x i find they make the beats sound too harsh for me so i never normalize there beats id rather use a limiter and compress them until they sound bigger.and sometimes a little bit a chorus can help fatten them up too .

There are several phrases that come to mind around those who think this, none of them are what I would call favourable

Yes you can get what apears to be saturation as a consequence of raising the gain of a signal that was previously clipped such that it ends up effectively normalized at the point it gets converted to analog, or where analog-like waveform reconstruction may occur, and also with mp3 encoding due to the way it approximates the signal for compression.

What actually happens with a clipped and subsequently normalized (or near enough normalized) signal is that voltage overshoots (there is a term for this that I cant remember) can occure which will be perceived as harmonics, and thus can sometimes (if you very very lucky) sound a little like analog saturation. If your unluckly then it just still sounds like digital clipping and therefore nasty.

All sounds worth a try? Well the problem also is its very hit a miss as the end result is entirely dependent upon the DACs and associated analog circuits when the analog waveform is eventually reconstructed (perhaps some bit of gear in the studio, or a random CD player etc). With mp3 encoding it generally just sounds nasty. In the case of a hardware sampler, then the situation is at least predicatble, so if it sounds good - do it, but with entirely in the box production - dont do it.

[mutilatedlip]

So when tracking in say, for instance, Cubase 5 - what's the best way to do it from scratch, for the beginners out there?

Whereabouts would we be looking to get the levels to when recording?

[FrankGIK]

Originally Posted by mutilatedlip

So when tracking in say, for instance, Cubase 5 - what's the best way to do it from scratch, for the beginners out there?

Whereabouts would we be looking to get the levels to when recording?

You're looking for RMS (average) levels around -18dBFS...that's -18 in the digital meters in Cubase. Of course, RMS levels will be higher for non-dynamic material like a hard-charging electric guitar rhythm line...that's going to eat up a lot more head room on the mix buss than a snare track that peaks at -6dB, which will have much lower RMS levels.

Frank