Recording levels - the real truth

Often people ask questions like "What level should I record tracks at?" or "What level should I record my mixes at?". The standard answer that comes back is: "Always record so that your maximum level is just below digital "zero" at the top of the meter." This well meaning advice is - unfortunately - completely wrong. Why? Well, that's the subject of this short article

The digital audio industry has a lot to answer for. Firstly, they lied to us about the sound quality. Early digital recorders and CD players sounded truly HORRIBLE, and many engineers pointed this out.

Amazingly - impossible as it may be to believe today, back then people didn't consider Product Marketing people to be a bunch of cheats and liars with a hidden agenda, and people - incredibly - actually believed what they said.

When people claimed that digital audio sounded horrible, they tried to say that it was simply because our ears weren't used to the "purity" of digital sound. It was "too good for our ears". In fact it wasn't. It was just crap. It has taken many years for digital audio to reach the acceptable level of sound quality that it has today. Largely this is through the use of valid, but somewhat kludgy, tricks like oversampling, one-bit digital converters, digital filtering, and noise-shaped dithering. It's a shame that this led to some engineers saying that they didn't like "the digital sound". I hope that people start to realise that when done properly, there is no such thing as "the digital sound" - only the sound of poorly implemented equipment. There is no practical theoretical limit to how good digital audio can ultimately get.

However, one area in which Digital Audio manufacturers still haven't matured, is in the advice that they give in instruction manuals. They can even be quite dogmatic about their bad advice too: "You MUST make sure that your recordings peak as close to zero as possible without the overload light coming on".

Complete nonsense.

Here's why:

What's the point in recording sound?

SO WE CAN LISTEN TO IT.

If you understand that simple point, then you are at least 50% towards understanding the rest of this article. The point is, we do NOT record sound in order to satisfy the needs of the equipment. The equipment is there in order to satisy OUR musical needs.

And one of the most important needs of music, is that it can get LOUDER.

We can call this:
Issue 1

"You shout so loud I can't hear you."

This is extremely important. Somethings need to be "louder" than others. "Light and shade" as they used to call it in Pinewood Studios Sound Department. In music for example:

1. A THRASH METAL track should sound "louder" than a pop BALLAD
2. A guitar SOLO should sound "louder" than its normal playing

If you record these things seperately, using only digital meters as your guide, you can find that ASTONISHINGLY the REVERSE can be true! You can end up with a BALLAD being LOUDER than a THRASH METAL track, and a guitar's verse part sounding LOUDER than its featured SOLO!! This is because digital recorders have to use peak meters, but our ears have more of an "averaging" characteristic when judging levels. Something that sounds "quiet" is usually less peaky and therefore has a relatively higher *average* level. On the other hand, something that is loud is often peaky, and it's these very peaks that can prevent you recording it a very high levels on a digital system.

What this means in practice, is that the listener of a compilation album containing the songs in example (1) would be forced to get out of their chair and adjust the volume levels on their Hi-fi DURING the album (which is unnacceptable - and unfortunately a common experience). In example (2), it means that whenever you play back the multitrack, you have to remember to manually turn up the solo - or even leave automation turned on to cope with it.

Wouldn't you rather just have everything recorded at the correct levels in the first place?

For example, to save tracks, it makes sense under certain circumstances on some systems (normally "real" multitracks with a fixed number of tracks), to "share" one track between two (or more) instruments. You normally only do this if the two instruments are not playing at the same point in the song. Clearly, when you are doing this, you want to record the instruments at the correct levels for the parts of the song that they play in.

But even in general, on playback, volumes should be automatic. Quiet songs should be quiet, loud songs should be loud. I'm not suggesting that you tolerate under-recorded multitracks - just that you carefully estimate how loud a multitrack recording MIGHT go, and allow yourself some headroom, and that when you lay down your mixes you should do so such that they are the right "sonic" level rather than "technical" level. The end listener will not be staring at meters to judge you - they will be listening with their ears. There is nothing worse than having one song on an album that is way too loud so you have to get out of your chair to turn down the volume on your hi-fi, only to find you need to get up and turn it back up again when the next song comes on. Records are meant to take care of themselves!

Of course, if you are taking your mixes to a mastering engineer for album compilation, then it can be argued that you should record them all at maximum (zero peak), as the mastering engineer will sort out the relevant levels between the songs (it's a standard part of their job, although some do it better than others).

But even in that case I would still question it. What are you going to do in the meantime? If you are putting together "rough cuts" of the album on CD or Minidisc for others to hear, you can COMPLETELY SPOIL the artistic impression of the album if the relative volume levels of the songs are unbalanced. If on the other hand you mix your songs with a fixed and known monitor level, and set your master fader so the song "sounds" at the right level, then you know that on playback, all of the songs will be playing back at the right volume level for the listener. Here's how you set yourself up to do this.

Set your monitor level to a FIXED and KNOWN amount by listening to a LOUD, zero-peaking song you know well at the level of loudness that it is INTENDED to be COMFORTABLY listened to (don't go too mad). Now, WITHOUT changing the monitor levels, adjust the master level of the track you're currently mixing, so that it "sounds" the right level for listening to. Check the meters to make sure that the level is not overloading. If it is, then you should consider limiting rather than turning the level down more - you have after all - just calibrated your ears with a reference track that certainly sounded loud enough without going past zero (an impossibility on digital playback).

By doing this, you can make excellent compilations collections of different material that can make up an album which can be listened to, end-to-end, without the listener ever having to change the volume knob.

Make a note of that monitor level setting - and use it as your reference.

Issue 2

"This guitar amp goes to eleven"

Related to issue one, there is also the issue of having somewhere to go when you otherwise run out of level.

Remember the movie "Spinal Tap"?

The guitarist says to the interviewer: "Look at my modified Marshal amp. It's very special, because all the knobs go up to eleven!!" (obviously the knob labels have simply been changed!). "Yeah - on stage, with other amps, you hit TEN and where can you go? NOWHERE. But on this little baby, I've always got that hidden extra! I can go that one bit more louder than anyone else!".

The interviewer looks bemused at this obvious nonsense - it is after all just the labelling on the knobs that is different, not the amp itself. "Erm.... But... On other amps", the interviewer says, "surely you just back everything down a bit, turn the master level up way high, and then you can go really, REALLY loud before the other knobs even hit ten!".

The guitarist thinks about this for a few seconds.
"But this one goes to eleven..."

It's funny, because although it's completely bogus thinking, we can all understand the guitarists logic. The psychological advantage in having a bit more headroom (even though in this case it doesn't actually exist and is just a marking on a knob) is significant.

So where's your "eleven" then?

By recording the instruments TO SOME DEGREE at a level that reflects their "playing" level, you offer yourself a real advantage. If you can't see this, then you've probably not recorded a live band before.

Imagine this: You've recorded the entire track with a live band, it's got a good feel, and the levels JUST JUST touch the zero mark. You feel very proud. Not because it sounds good - you were too busy watching the levels to care about sound quality. You're proud because the meters just touch zero without overloading. "Wow!" you say to yourself. "I bet my old tutor at "The Sound University" would be very proud of me now. What a fine student I must be. It doesn't sound any good, but who cares? I've peaked perfectly at zero. How cool am I?" - BUT, then something unexpected and disturbing happens:

"I've got a great idea", says the leader of the band. "That was a great take - can we drop in for the ending?" he says. "Erm... Yeah sure, it's easy to drop in for the end." you say. "GREAT", says the band leader, "listen everyone, I've got a brilliant idea. I think that on the ending we should play REALLY REALLY loudly! - I want those drums to really smack loudly. I want the guitar to HAMMER out those chords. The bass to really KICK you in the stomach, and that keyboard to PULSATE over long sustained chords."

"BRILLIANT" they all say, and rush out into the studio.

At this point you feel sick in the stomach. Because you know darn well there is no technical reason why the artists shouldn't be allowed to do this kind of thing if they want. But YOU can't do it properly because you've messed it up retrospectively. The rest of the song has peaked at zero so you now have nowhere to go. It's too embarrassing to try to explain to the band that this isn't going to work out properly, so you just knock back all the recording faders by three dB, and hope that will stop everything distorting. It usually doesn't. Maybe 5dB was needed.

Let's say you get away with it. It didn't distort. Phew! you're in the clear - or so you think. The band listens back. "Hey great ending guys" says the band leader. He turns to you: "Is it my imagination, or did the levels drop suddenly at the punch in point?". You try to look innocent. "Erm... Maybe". The band leader says "Oh well - it must be our fault. Come on guys, lets do it again. The level dropped last time, so make sure you're playing REALLY REALLY loud even at the drop in point - let's go". You go green and feel funny.

The hole you're digging for yourself just got a lot deeper...

This time around you knock the faders back 6dB. You have a cunning plan. Cool! You'll just match the levels at the drop in point because the band are mistakenly playing louder, and THEN you'll try to fade them back a bit further when the band really kick in so that they don't distort. You'll either have to try and move over a dozen faders precisely by hand simultaneously (not very easy) or if you're on a top-flight mixing desk you'll have some "VCA-Style-Subgrouping" that will let you do this with just one fader. You reckon you'll get away with it. Nice idea but:

You've really blown it big time now...

It doesn't matter that the "technical" levels are the same if the band is actually *playing* differently.

When the band comes back in the band leader says: "This is really weird... What's going on? At the drop in point the levels are the same but the band now "sounds" different... Very odd. Hey! What's going on at the end? We were playing really loudly but it sounds like something is "holding back" the recording." (you snatch your hand off the fader and look innocent), "It's terrible", he says "Do you have ANY idea why this is happening??"

"Nope" you say, shaking your head, whilst going red-as-a-beetroot, "no idea at all...". Then you do the most UNDERHAND get-out attempt of all: "I suggest we re-record it from scratch.. Yes - a good idea I think. You guys know the ending so well now that you are much better prepared to build up to it during the song". You little RAT!

At least this time you can record the whole thing PROPERLY with some headroom.

All this trouble because you trying to live up to the advice in poorly written student textbooks about the "importance" of peaking at zero. Maybe someone should tell those authors that the importance IS ZERO - i.e. it ISN'T IMPORTANT AT ALL. Certainly not compared to ruining a session recording!

Nor is it as important as you being able to relax on the session, knowing your levels have a margin - not just for safety, but for artistic licence too. Even if you are just recording by yourself, you might well decided that in a certain part of the song you want that part to get louder. Sure you can do it with "automated mixing" - but it's much simpler just to record the parts at the correct level for each part of the song. Then you can forget about it and mixing becomes a doddle - the track mixes itself.

So: Give yourself some slack, and make sure the guitarist can go to "eleven" if he wants to.

Issue 3

"Let's be reasonable"

Finally, remember that on a multitrack, most things simply DON'T NEED to be recorded at maximum volume. If you are mixing down to digital, then - on a pop track - there will be only TWO TRACKS that will be played back at zero level (accounting for master fader offsets that creep in during mixing). These two tracks are BASS DRUM and SNARE DRUM. Everything else will usually be substantially below this. Even the lead vocal will be much quieter than these two, and other things - such as guitars - will be quieter still, and certain other instruments like "Cabasa" will be very quiet in the mix indeed.

One argument says "What's the point in having sixteen bit dynamic range if you're under-recording your instruments!".

Urm... yes... How about (as a counter argument) "What's the point in religously trying to record your cabasa at exactly 16 bits when it is never going to represent more than 12 bits of the final stero mix?"

Anyway, there are very practical reasons why it doesn't make sense to record everything at full blast. Aside from the issue - already discussed - of having somewhere to go if things need to get louder, there are other important considerations too.

Firstly, if you are using a PC system like Cubase or n-Track, recording some instruments (hi-hat and cabasa as good examples) can lead to you having to have the fader very low on mixdown in order to get the level right, and the fader is not very easilly controllable at such low levels. Similarly if you are using a "real" mixing desk with nice long faders, you can end up using no more than the bottom inch of fader travel! - which doesn't give you much control either. Sure you can use "Trimpots" and their PC equivilent, but this isn't always available to you when monitoring (and on some desks not even during actual mixing!), and in any case it's just more hassle for you to do. Why not record things at practical levels so you don't have any such problems?

Secondly, if you are using an analogue mixing desk, having certain instruments - such as hi-hat - recorded as loud as possible on digital multitrack can cause serious mixing problems. High-frequency energy bleeds across an analogue desk easilly. If things like hi-hats and cabasas, and cymbals are all peaking away at zero, you will find that even when you switch them off, you can still hear them clearly in the background. This can affect the aux sends on an analogue desk too. You can find all that electrical leakage can result in there being reverb on the hi-hat (for example) even when you haven't put any on! - because it is somehow leaking into the aux bus. Often this is worse with headphone mixes. Artists can complain that they have asked for something to be turned off, but it is still leaking through. This problem applies not just to budget consoles, but to great big world-class consoles as well - probably due more to the cable looms that connect everything, rather than the desk itself. You simply don't NEED to record *those* things that loud! Bass Drum and Snare Drum, YES! - by all means - but not the "top kit" stuff or indeed any supposedly "quiet" sizzly instrument.

Recording every single track religuosly and painstakingly at exactly sixteen bits will not (within reason) give you significantly better sound quality than recording at REASONABLE levels that give you room to manoever, are much easier and practical to work with, and which reflect - albeit to a very *small* degree - the intended loudness of the instrument in the mix.

So, let's be reasonable!
Summary

Hopefully you now understand why trying to "always reach zero" is nonsense. Who made up this stupid rule in the first place? Certainly not a Sound Engineer that's for sure.

As I am sure many of you have already discovered, the folks at Microsoft (indeed people EVERYWHERE) stupidly record all 16bit PC sound samples at maximum peak. What's more they are so obsessive about "MUST REACH ZERO OR ELSE" they compress even the most insignificant samples to death so that they almost never leave the zero mark. The practical upshot of this, is that you can be merrilly recording away on Cubase, Logic Audio, n-Track, whatever, and be having a thoroughly good time recording wonderfully dynamic high-quality music. Then you make the simple mistake of clicking outside of a dialog box, and Windows goes:

Ding !!

Blowing your head off in the process. It's so LOUD!!!!!! If you are working on headphones this can actually cause you significant physical pain. I turn off things like the Windows "startup" tune, because after some PC audio work, I know I'll accidently be deafened next time I switch on and my PC starts up.

In the days of moving-coil VU meters, everyone - including ordinary members of the public - seemed to instinctively know that the meter vaguely reflected how "loud" things sounded - and that this was meant to be some way below maximum reading. The red section just meant "too loud" but the tape could handle it if it didn't happen too often. You knew that normal music bounced around just the middle mark, and very loud stuff was a bit higher (not much higher though, as these were not peak meters). It was a bit hit-and-miss but it worked ironically much better than the peak meters of today. I spent three years working with two Sony "3324" 24 Track Digital Tape Machines as a 48 Track setup based around a Solid State Logic desk that had ONLY MOVING COIL "VU" METERS for the outputs. However, the setup made excellent recordings. I didn't need to worry about peaks getting past the VU meters because SSL desks had a peak light in the channel, which lit up a dB or so below zero FSD on the Sony tape machine peak meters.

Ordinary people don't understand (and why should they?) that on a peak, digital meter, the readings are not directly related to how loud something actually "sounds". They are simply indicating the headroom requirements of the signal under question, and that is not the same thing.

In summary, when there are times when the relative "loudness" of something is important - such as the level of a mix of a quiet song within an album of other songs, or the level of a quiet solo instrument in the middle of a song, then stop looking at that meter and start using your EARS for a change!

I worked in the Sound Department at Pinewood Film Studios for a while. It may surprise you to learn that the entire soundtrack for movies - including academy award-winning films - is mixed COMPLETELY BY EAR. The ONLY time that the meters are briefly looked at, is when something very "loud" occurs, and they want to check that it is in acceptable range.

Sound engineering - when it comes down to it - has always been about LISTENING.

Jezar.