Great question and an in depth topic if we get into the guts of a system including personal secrets or flavours, (if we want to call them that). In a nut shell however, gain staging is making sure an input and output level is even to each other, below the clip point or in its sweet spot. None of it is written in stone. You may want an overdriven transformer or tube sound for a certain tone or something very transparent.
We also plan ahead by keeping individual track levels lower (yellow is the new red). Reason being, the 2-bus can only take so much level, gain staging groups of tracks is very important to keep it from piling up once it all hit the 2-bus, a reason some people choose hybrid summing, thus... more options, more headroom, more flavour.

Gain staging is all about planing levels and making sure they are at an optimal performance level or sweet spot.

Digital recording is more critical, or less tolerant because of its 0 tolerance so we pay close attention to our levels ITB to avoid that nasty digital distortion. Plug-ins and faders are also a big part of this. When using plug-ins, we want to make sure we aren't boosting levels too much either. We try and keep things even. Faders also sound better when they stay at 0 so I personally set levels close to where my mix is and try to not move them too far off axis. I use analog hardware and to juice up a digital sound. Gain staging is all about setting levels.

audiokid, post: 384849 wrote:
We also plan ahead by keeping individual track levels lower (yellow is the new red). Reason being, the 2-bus can only take so much level, gain staging groups of tracks is very important to keep it from piling up once it all hit the 2-bus, a reason some people choose hybrid summing, thus... more headroom.

While keeping optimal headroom is a good reason for conservative levels, I think there are other reasons why keeping your levels in the green is a good idea.

Firstly, intersample distortion/peaking. This is a HUGELY complex topic as it involves conversion technology, particularly the reconstruction filters, meeting the ITB digital world. There is a good TC Electronic paper by Thomas Lund (Distortion for the People, I think) that explains this all very well. Basically, in 32-bit floating point math, levels greater than full scale are accommodated without any clipping taking place albeit only in the digital realm. However, once it hits the DA converter those levels are likely to clip as most converters are only 24-bit. This is just simple clipping, however, because of extreme levels. Regarding intersample peaks/distortion, this occurs when a peak happens in between two samples within a sampled waveform and the likelihood of this producing distortion is largely dependent on what the shape of waveform. It is most likely, however, to occur on square wave material, the most extreme cases being between 6 and 12dB above what is reported by the meter! In other words, your digital meters are lying to you and when these signals hit the DA converter you will experience clipping and distortion even though the meters in your DAW say your OK. The free SSL intersample peak meter can help you track these peaks down.

The other reason is plugins and their internal math. After chatting to a plugin engineer about this I was enlightened to the fact that not all plugins are created equal. Even though many plugins are designed with a higher internal precision than 24-bit many times they just do not handle values greater than +1/-1 (full scale) very well and will exhibit distortion based on their robustness to this signal. Here is what he informed me:

MacGregor wrote: plug-ins internally represent the (audio-) sample values with floating-point values between -1 and +1. Those 2 values, when put on the output of a plug-in, correspond to 0db (+1 is the maximum positive phase, -1 is the maximum negative phase. Think of it as the maximum values of a sine). For ease of explanation I use +1 below, but for -1 it's the same.

Depending on what the plug-in does there's a lot of mathematics inside, additions, multiplications and whatnot. This processing results often in (internal) temporary values bigger than +1. That would 'normally' be not a problem as long as after all those internal processing steps everything's back to -1 ... +1, but...

...a lot, and with that I mean a LOT, DSP algorithms are designed to work ONLY in this range. Some work for all values, some more work for values slightly above +1, some work bad outside that range, some just crash or start to oscillate.

Unfortunately a LOT of plug-in programmers don't know shit about mathematics, so they just put some fancy DSP algorithms they've found in the internet/books inside their plug-in code and call it a day, without thinking twice what their code will do when used outside its designed range.

When you leave 10db headroom the chances that the plug-in behaves properly (and sounds best) are much higher. So, if in doubt, leave more headroom. It never hurts, and you can easily make up the gain to proper loudness levels at the end(!) of the processing chain.

So there you have it.

Cheers :)

Wow, this all makes so much sense, I wish I would've known this years ago. I've already tried it on a couple of past mix's and then a/b'd them and the mixes with
proper gain staging sound more crisp (Specially the snare tracks), open, I can actually hear the reverb on everything, and I can also pick out everything that's going on in the mix.
Also it makes mixing easier and funner.

What do you guys generally let your raw tracks peek at when the fader is at 0 ?
I controlled the gain so that the raw tracks peaked at -6 but If I'm reading all your info correctly than I should be letting them peak at -10?

Thank you everyone

Back in the days of analog, zero on the meters was in a sense an arbitrary working level of at least -15. Which essentially meant, you could confidently bang the meters into the red without fear of unlistenable, horrific sounding and unusable distortion. Plus it also sounded kind of cool when you hit the tape hard which caused nonlinear saturation to occur. Ain't that way with digital, not exactly. Our digital meters simply put zero at the highest point one can print without decimation of the original waveform. For those of us that came from the early 1980s introduction of digital recording and only had 16 bits, some of us pushed the envelope. This would be like never cheating on your taxes, never cheating on your partner, never spitting on the sidewalk, never exceeding the speed limit. I think that's rather unrealistic? Don't you? And in fact, in earlier 16-bit days, your best sound came at the uppermost higher levels you could print. Because in digital, the lower you record, the less clarity there would be. And then we had quantization noise, oh God... nothing like listening to your reverb trails, trail into rice crispies. Yum. It's better with milk. Of course then it comes out your nose. Stop making me laugh! But I digest. Back to levels.

In the digital world, some recommendations were made for average recording levels at -12 & -15 & -18, even -20. But that's really for people that only drive the speed limit. I'm also a motorcyclist for over 34 years and as a result, I've had my license suspended a few times for exceeding my levels. (You've got to watch out for the audio gestapo... they're coming to take me away, ha ha, hee hee, ho ho, lather, rinse, repeat) in other words, sometimes I'll occasionally exceed zero in digital. Now that doesn't make it a horrible, not yet. It just makes it horrible when it happens too much and on the wrong instruments/vocals. For instance, not unusual to have an occasional drum transient. And you won't necessarily hear that in rock 'n roll land. It might sound like crap when the cymbal crashes along with the timpani on an orchestral recording. So don't do that. Some computer audio interfaces actually include a front end limiter to ever prevent that from happening. That's not necessarily a device I would ever want to have. God no. It's simply designed as a crutch for beginners with no legs. So for entry-level folks that still don't quite yet understand how to properly dial in the right record level, it can still be useful. But then everybody wonders suddenly why their recordings have no life, sound flat, sound dull. Well...DUH. That's when happens when you hit too many limiters too often. And that's why I could never go for one of those gizmos. So while I also love dynamic range... I don't like much dynamic range. But even my highly compressed mixes sound more dynamic than without the high compression I use. It's all in how you record and mix. So trial and error experimentation is important. There are some things you can get away with and other things you can't. Like Idaho and sushi. So while you think it's fresh, it's not only days old, it was frozen first. You just think it's fresh sushi in Idaho. And most people would agree with you. And actually freezing is good because it kills some of the microorganisms you don't want in your gut. Recording is the same thing as sushi. Only you can decide how much WASABI you want in your sound. Otherwise, without that, it's just dead fish that hasn't been cooked or mixed. And there's not much appeal in that. But there might be some scales? Yuck... if there was any distortion in sushi, it would be scales. And what did your music teacher tell you to do before you ate your sushi but to play scales. I think I need some Pepto-Bismol now? Remember Dan Aykroyd putting the fish in the blender and then drinking it? Well you don't want to do that with your digital audio. Analog maybe?

Drink a fish
Mx. Remy Ann David

Awesome, I have a good grasp on the concept now.

Here's what I don't understand..
Proper gain staging starts with your gain on the preamps, correct? Verses fixing the gain after tracking?
So if digital meters suck, what meter should I use when setting my preamp levels?
Should I use something like the SSL inter sample peak meter or PSP Vintage Meter?

No. You should just use the meters in the software. They are accurate peak reading meters. You don't have to worry, they are not of the VU ballistic types which don't show peaks. A VU or, Volume Unit, meter simply indicates average levels and not peaks. That worked well for most Americans where the Europeans generally wanted peak reading meters which have different ballistics and characteristics from average VU meters.

Yes, he surmised correctly in that, levels should be set as close to optimal as possible when recording. You don't want to go too low and you don't want to go too high. That provides proper sonic integrity and plenty of headroom for processing after-the-fact. Too low a level and recording will cause the audio to be rather blah sounding. While too high a level of course causes uncorrectable distortion. It really won't matter if your software has a overload compensation plug-in. It only lessens the awful it doesn't remove the awful.

So that's why folks tell you to look for an average between -20 to -10 for your average level of recording. Some of us push the envelope a little higher because we know how to. I ago for highest peak record levels without trying to clip on purpose though sometimes momentarily transient peaks with a slight clip won't always sound awful. You just don't want to see those peak lights in your software meter constantly flashing.

Sometimes I get a little flashy
Mx. Remy Ann David

martinblem, post: 384990 wrote:
Here's what I don't understand..
Proper gain staging starts with your gain on the preamps, correct? Verses fixing the gain after tracking?

Ah, well, herein lies the nebulousness.

This is a topic that many people don't understand fully, even some professionals that I have met. I'll try keep it in layman's terms but it all starts with two ideas called maximum input level (NOT gain) and input/output calibration. Every piece of gear has a maximum input and output level that level is usually expressed in dBu. To give my setup as an example, my Lynx Aurora has a maximum input level of +20dBu. That means that it has 16dB of headroom above +4dBu, which is also 0VU. My Chilton console's direct outs have a maximum output level of +20dBu, which exactly matches the input sensitivity of the Aurora. This means that when a 1kHz sine tone is reflecting 0VU/+4dBu/5PPM on it's meters, the input meters in my DAW show exactly -16dBfs.

Now, this is a perfect case scenario, but do you see where potential gain discrepancies can start creeping in?

Let's just say that the Chilton has a maximum output level of +24dBu. This means that it has 4dB of headroom MORE than what the Aurora can accept so the calibration, we would say, is offset by 4dB. 0dBFS is 0dBFS regardless of the analog level. 0dBfs may be +18dBu to some units and +26dBu to others and that all depends on what the manufacturer has calibrated the unit to.

Have you ever been baffled when meters just don't seem to add up and display seemingly huge disparities in level? This is the reason why. Another more serious problem arises when the preamp feeding the AD converter has got a LOWER output sensitivity than the AD converter. You may find yourself cranking the preamp into distortion just to get enough level but then end up getting more distortion than you bargained for even though you're nowhere near clipping.

Gain staging is one thing, but knowing your gear and calibrating it correctly (if possible) is the key too.

martinblem, post: 384990 wrote: So if digital meters suck, what meter should I use when setting my preamp levels?
Should I use something like the SSL inter sample peak meter or PSP Vintage Meter?

As Remy said, use your digital meters, they are fine for general metering. The mantra, however, is to be mindful of your levels and just don't record or mix too hot. That's really all there is to it. Using the SSL meter will help you, yes, but don't stress about it too much, provided your levels are conservative.

Cheers :)

You guys are awesome, I understand it now.
Thank you!

I hope I don't repeat anything that's already been said..I have read some but not all the posts in this thread. Forgive me if I have doubled up. Just thought I would add something. So here it is:
If you find that you have recorded too hot, or have been given someone else's project to mix that has been recorded with the meters slammed up close to zero, use Trim plug-ins inserted on each channel to knock back the gain. This is not the same as pulling down a virtual fader. It's kind of like using the gain control on a console rather than just pulling down a fader: ie: if your gain on a console is set too high, no amount of pulling the fader down will stop the gain from clipping.
If you have bunch of tracks you are sening to a bus in your DAW, and you insert a compressor or some other plugin on the bus and you find that the input of the plugin is going into the red, rather than just pulling down faders of individual channels, you may find you have more "nice sounding " headroom if you simply lower the gains on each channel with a trim plugin.

RemyRAD, post: 385003 wrote: So that's why folks tell you to look for an average between -20 to -10 for your average level of recording. Some of us push the envelope a little higher because we know how to. I ago for highest peak record levels without trying to clip on purpose though sometimes momentarily transient peaks with a slight clip won't always sound awful. Sometimes I get a little flashy
Mx. Remy Ann David

The only reason why they won't sound awful is because the transients are probably too short to be heard. Digital clipping does not sound good. If you look at the waveform up close you will see it is square at the point of clipping. The number of "square" clipped waves you willl see will depend on how long the clipping lasted for. I'm sure you already know that, but if some newbie reads this thread and sees you saying it's ok to clip digital signals sometimes....is that the right message to get across?
Personally I can't see any logical reason to record with levels even close to approaching the level where they clip. It certainly won't improve the sound of a recording, and you run the risk of having little headroom left for mixing, and worse having an unexpected transient peak that actually can be heard as awful when you are recording because your eyes weren't glued to the meters and didn't see it coming.

I record generally at -15 a little more or less. My converters are like 13 years old, while of high quality for the time, i find this level sounds full and pretty natural thru them. At the end of the day it's whatever sounds the best, and how much you plan to do later. For instance last night we added 20db of eq to an extremely dull djembe track, which actually turn into the driving rhythm of the song, as opposed to just an occasional feature, (we were mixing, so i can't wait to see the look on the clients face, good or bad. when they hear it.). If i hadn't left headroom when tracking, this wouldn't have been as easy to get the track bright enough. So, level increases can come in many forms, 20db of eq, is 20 db. It was all highs, which takes up far less room than,lows. Some 6-8 db of compression kept our level at an average of around -6. But the annoying enemy of a digital clip light kept going of one time in the song.
so after some fiddling around, we just left it were it clipped for an inaudible amount of time, cuz we couldn't hear the clip, and it the settings were great for the whole song.
so i (+3?) that it depends on what your recording as far as clip lights. I don't like 'em in digital, but i won't compromise a good mix, just because i see one.

For some time I have been meaning to write a tutorial on the subject of gain staging. This thread kicked me into action, so instead of just thinking about it, I got to work and did it. It has been posted here [[url=http://[/URL]="http://rocksuresoun…"]Audio Recording Gain Staging | Rocksure Soundz[/]="http://rocksuresoun…"]Audio Recording Gain Staging | Rocksure Soundz[/]

Might be worth a read or a critique.