Recording Levels

[waitup]

As per the discussion in the other thread, let's hear how hot you guys record and why.

I aim for my peaks to be hitting -18dB FS, occasionally one might slip through and hit -12dB FS.

Why? A few reasons.

1. A 24 bit digital recording system has a theoretical dynamic range of about 144dB. This far exceeds any of the analogue gear connected to it, so as far as I'm concerned the noise floor is a non issue. Hell, imagine 144dB mapped out on a linear scale and place a marker at -18dB... you're still a hell of a lot closer to clipping than you are to the noise floor.

2. Any higher and I'm closer to clipping than I feel comfortable. It only takes a drummer to get a bit excited and hit a bit harder and you could get a peak of 10dB or more... why risk clipping?

3. With levels around where I'm talking about, you can push up pretty much as many faders as you like and the stereo bus wont clip.

4. I don't buy into this "using all the bits" theory. As far as I understand it, weather you record complete silence or full scale clipping, every binary word is 24 (or 16) bit long.

5. I interface some hardware with my system. My converters are calibrated so that a -20dB FS tone will read 0dB on my compressors. If I had audio on a track peaking at -6dB, I'm hitting the compressor harder than I'd like.

There's my 2c, let's see yours.

[obutcher]

Just want to add that it's simple math when it comes to levels in the digital realm.

If a signal is at -50dBFS at a given sample and it's boosted by 30dB, then all the information is doing it moving that sample's level up by at amount of dB's on the digital scale. It's not 'making up extra data', it's simply moving the level's reference point up or down on that sample.

Also, in WAV format, try recording a track with NO audio input for 30sec. Then record a track with up to -0.1dBFS of signal for 30 sec. Check the file sizes, they'll be the same. Same amount of info used to store full range signal and no signal at all!

[Chris H]

I record low level into the A to D mainly treating the yellow as red especially if i'm mixing ITB so I'm not maxing out the stereo mixbus. I must admit I don't know a lot about the maths of digital signal levels etc but in reading long discussions over at the old now virtually defunct PSW forum where Terry Manning was a committed advocate at recording at low levels and people such as Paul Frindell, a principal designer in the Sony Oxford team, supporting the reasons.

A quote from Terry at the begining of that discussion:

"The stuff about "not using all the bits" approaches untrue urban myth status.

The old enemy in the analogue world was noise. We all learned recording techniques based upon that premise, whether we worked in the analogue world, or just learned from those who did so.

Noise IS NOT A REAL FACTOR in digiworld.

Keeping levels low helps almost every aspect of what people thought was bad about digital (harshness, tinniness, crunchiness, distortion).

You could even record -40 to -50 dB down, and be better! But -12 to -18 is great."


here is a link to the discussion from the forum archives:

http://repforums.prosoundweb.com/index. ... 878.0.html

[audioio]

And this:

http://people.xiph.org/~xiphmont/demo/neil-young.html

[Kurt]

When I do live recordings I tend to set the levels very low (no preamps on the drum tracks, others set to unity if possible) I run a line out of my compressors. The guitars might wind up averaging -30.. It means I can get on with doing a live mix without worrying that the recording levels are too hot. I have never had a problem with the signal being too low or noisy.

[waitup]

I like the responses in this thread. It's encouraging that no one has piped in with the "record as hot as possible without clipping" line.

[Sammas]

4. I don't buy into this "using all the bits" theory. As far as I understand it, weather you record complete silence or full scale clipping, every binary word is 24 (or 16) bit long.

"using all the bits" isn't actually far from the truth in basic terms... it is just really misunderstood. I blame it on the statement "1 bit equals 6dB". If you increase your word by 1 bit, it mathematically results in an extension of dynamic range by 6dB... 1 bit doesn't directly represent 6dB though.

Binary is just a base two counting system. 0 & 1. Our counting system is base ten. 0, 1, 2, 3, 4, 5, 6, 7, 8, 9.

Consider an 8 bit word in our base ten counting system.

0000 0005.

That represents 5.

In three bit.

005. Means the same thing.

In one bit...

5. Yup, 5 again.


For small values, the leading bits do indeed remain unused... but the secret is in the resolution.

You have 5 chickens in 8 bit - 0000 0005
I have 100 chickens in 8 bit - 0000 0100

I don't have a higher resolution of chickens because I have more of them. Your 5 chickens still cluck and lay eggs just the same as mine...

Binary is based on the same principles.

"5" in 4 bit - 0101
"5" in 8 bit - 0000 0101

Words are always transmitted in their complete word length, but as you increase the value you do use more bits to represent it. The leading bits are just padded out...

The really improvement moving from 16bit to 24bit is the increase in resolution.

[Alistair]

http://www.youtube.com/watch?v=d82ukYXj52U
That can act as the 'as loud as possible' argument for the rest of the thread.

Over the past 3 or 4 years I've been consistently lowering my levels- I record averaging peaks around -10dBfs, with singers and drummers always set a little bit lower as they are more likely to decide to play twice as loud suddenly.
As soon as I get to mix down I grab every track and pull them down to around -20dBfs and mix from there to make sure I'm not overloading any outputs. I might aim for a lower volume on tracking next time round and see the differences.
A couple of questions though-

What do you do in those rare (at least for me) situations where you want to run a pre or compressor fairly hot? Do you have a bunch of attenuators?

At what point do you bump up the volume of the tracks? Do you aim for mixdown at -20dBfs as well and make up the volume in mastering only?

[waitup]

The really improvement moving from 16bit to 24bit is the increase in resolution.

You got it.

[waitup]

Do you aim for mixdown at -20dBfs as well and make up the volume in mastering only?

Pretty much. It might come out a bit louder with dense mixes...

[Wiz]

This is such a touchy subject in the online audio community..a bit like bringing up politics and religion in one sentence at the Xmas dinner table...

Back in 2005, I was having a internet discussion with a colleague on the net. I didn't agree with something he was saying. We were discussing resolution of digital audio, and the difference between 24 bit and 16 bit with regards to resolution.

I posted the question to pro sound web.

It was quite a humbling experience, I got a reply from George Massenburg which was tactful and also really put me in my place...there was also a fantastic reply from Nika Aldrich, who wrote a book about Digital Audio.

Nika handled the explanation of my question, beautifully. He gave me a test , which I failed miserably, to gauge my true understanding, and then took the time to explain the answer gracefully.

You can read it here...

http://repforums.prosoundweb.com/index.php?topic=5739.0

that answer changed a lot for me in my understanding, and only recently I had gone back to read it again.

That in conjunction with reading Dan Lavry's white paper on sample rate differences, cleared up a lot for me.

When I got a handle on quantisation error, oversampling, nyquist etc, and why there is roughly a 6db change in dynamic range with each bit added to digital bit depth (and its still a learning area for me) it settled my mind quite a bit.

If you are interested in this kind of stuff, its worth reading about it, and mulling it over....


Now back to Xmas dinner....

[Drumstruck]

I like the responses in this thread. It's encouraging that no one has piped in with the "record as hot as possible without clipping" line.

There's my queue..... and I do record as hot as possible for electronic drums only - and pose the question: if it's not clipping then isn't recording hot better than recording "less hot"?

With 8 analog direct outs from the drum module -> 8 line ins on the mixer I let the desk peak (only just) on the hardest hits. The drum module provides 127 increments of level, but the softest hits are showing about -60 so the real dynamic range is less than 70dB.

<edit>
Generally the desk faders are at 0 though sometimes up to 10 <edit miskey =1> to push more level into PT - thence over firewire into PT so the actual recording isn't as hot as the desk.

My take is that recording hotter means I don't need to apply as much gain on mixdown - in the past I recorded at the suggested levels (nowhere near peaking) and sometimes had to double the drum tracks to get enough gain. So, after experimenting by recording hotter and hotter I got to the point where I am now recording hot - no digital distortion and no need to double track.

[waitup]

But why do you need so much level out of the drums at mixdown? Are the other elements in the mix recorded super hot too?

[Drumstruck]

hi Nick,

Mostly I'm talking about my own material so.... generally I record everything as hot as I can - probably analog habit.

When all the other tracks are combined and being mixed the poor drums are often left behind. I try to compress the drums as little as possible (in my own non-commercial material) looking for max dynamic range......

Note I'm not disputing any of the other great info in this thread - if I did more commercial / client material I would approach that more like everyone else.

[waitup]

Cool, I'm not criticizing I'm just interested

[Kurt]

Doubling tracks to get more gain seems a little odd. Maybe it's a 'tools thing, the DAWs I've used let you right click a track and normalize it, or apply whatever amount of gain you like. Reaper even let's you select multiple tracks/takes and normalise/amplify them all by the same amount.

[waitup]

Doubling tracks to get more gain seems a little odd. Maybe it's a 'tools thing, the DAWs I've used let you right click a track and normalize it, or apply whatever amount of gain you like. Reaper even let's you select multiple tracks/takes and normalise/amplify them all by the same amount.

You can apply gain/normalize to multiple regions at the same time in Pro Tools also.

[Drumstruck]

Doubling tracks to get more gain seems a little odd. Maybe it's a 'tools thing, the DAWs I've used let you right click a track and normalize it, or apply whatever amount of gain you like. Reaper even let's you select multiple tracks/takes and normalise/amplify them all by the same amount.

You can apply gain/normalize to multiple regions at the same time in Pro Tools also.

Every time I've "normalised" I've found the result a little "abnormal" in that it applies too much gain and the peaks are (visibly) flattened and I lost some dynamics (but I'll have another play with that) By doubling it allowed a "2-stage" gain - one track pretty much set and the 2nd track to provide a bit of subtler boost / cut - a bit like a manual side chain without the compression. But being largely self-taught in the digital domain, I may be doing something the hard way

[waitup]

All interesting ideas. In my experience, if all your tracks are recorded at about the same, conservative level, and you begin the mix with your faders somewhere classy, you never have to "normalise" or "apply gain" to any audio. You should have tons of room to push things up with not a clip light in sight, but YMMV.

[tunetown]

I use the analogue VU and try to peak them at zero. This usually translates to about -18dbs. It also seems to hit the sweet spot of my pres there. If I drive the pres hotter to create a louder Digital meter I usually find it less pleasing.
Also, I've noticed that when something is recorded digitally hot you can't add many digital plugins. It distorts horribly. With low input analogue recordings I have no dramas with digital distortion.

Analogue_-_Digital_scale1.jpg

[CriticalD]

Having said all of that.. what does Nick know about this topic anyway?
havent you read Stav's book?

[Thirteen]

I still can't sit for more than a few minutes listening to an MP3 without losing attention, CD's for about 1.5 songs, but a vinyl record in good condition, well I can sit there and listen attentively for a couple of albums. No-one has explained that one to me, but I find digital audio more fatiguing than analog, so I am still not sure that the numbers and Nyquists tell the whole story.

[jkhuri44]

hi end is lame....vinyl as a medium is LPF'd to some degree, so, it is going to sound better...ie. like someone turned down the treble...

Also, when u play vinyl, its not that convenient to scroll or flick through a track, making listening to music more of an activity. Same with purchasing a record, u'd be buying shit u really like...

[Thirteen]

hi end is lame....vinyl as a medium is LPF'd to some degree, so, it is going to sound better...ie. like someone turned down the treble...

Also, when u play vinyl, its not that convenient to scroll or flick through a track, making listening to music more of an activity. Same with purchasing a record, u'd be buying shit u really like...

It would depend on who mastered the record. My experience with buying 1980's sealed, unplayed 12" singles and comparing them to the CD versions of the same tracks is usually that the CD sounds like it has a blanket over it and is dull and the record is shiny and open sounding, the opposite to what I expected. A vinyl record can record and reproduce out to 40Khz, double the bandwidth of a CD.

[rick]

I make everything louder then everything else .. all the time all at once ..

well kinda ..

actually kinda not


I record at o vu = -14db

which in the washup means unless my vu meters are totally pegged I will never ever hit the red overload light

I gain stage every piece of gear to work with in its own unique "sweetspot "
so something I do get into ovu to +3 territory and some stuff also runs -10 vu
it just depends on the noise and the bit of gear

make sense ..?

Ok goodo
I have no idea what db peak average/ rms levels or any of that guff means in regard to real life levels or sound

well actually i do

but I have no idea what it means the way guys use it on the web
I tried to understand the conversations
but it made my brain hurt so I stopped trying

[Text_Edifice]

Rick do you meam 0vu=-14dBFS or dBm?

[rick]

no dbfs in my world
fs = for sure

[Text_Edifice]

Ok - so about 28-30dBm of headroom before digital clips yeah (depending on how the converters were calibrated)?

[Text_Edifice]

Unless you're tracking to tape in which case just ignore me, except why is 600Ohms relevant?

EDIT again: don't really mean relevant more wondering if that means you're impedance matching through your record chain or just using dBm as a stand-in for dBu?

( ricks edit and added info for clarity )
I have slipped an m in where it didnt belong
I guess to simplify the rabbit hole and to ignore the idea that i have lots of old stuff with tubes and transformers and 600 ohm load are still relevent although less so

and i am running the digital standard that sony proposed for prolevels back in 1984-5 and became an AES thing and is now lost to confusion .. except it still works perfectly

my digital meters read - 14 db below full scale ( the sony dmu meters)
when my vu meters read Ovu
my vu meters are referenced to + 4dbm

the peaks sit about anywhere from -4 to -6 on the digital meters on loud stuff
the program way below that

i dont use all the terms the clever guys do , i am not that clever - Rick

[Drumstruck]

.....
I gain stage every piece of gear to work with in its own unique "sweetspot "
so something I do get into ovu to +3 territory and some stuff also runs -10 vu
it just depends on the noise and the bit of gear
......

To me this is a peak in the conversation.

I suspect the majority of you pro guys also do this as part of building the mix - so the levels would be based on the particular piece of kit the signal is passing through until the signal reaches the ADAC/DAW (and vice versa on mixdown / mastering).

It would be interesting to hear the subtle differences our top engineers would get if mixing / mastering the same song and using exactly the same pieces of equipment.