How do you use signal pre-processing when tracking vocals?

What signal preprocessing techniques do you use when capturing vocals?

There are really a few parts to this question: First, what kinds of inserts do you use to pre-process vocals when you're tracking them (compressors, de-essers, EQs, noise gates, etc.)? In each case, pre- or post-fader? Also, what particular settings do you use (I'm especially interested in any compression and/or limiting settings you might use)?

posted by saulgoodman (19 comments total)

Nothing.

There is a school of people who will compress while tracking - usually only a little, though, maybe 2-3dB. It's generally an attempt to get a better signal/noise ratio.

I'm not quite what you mean by pre/post fader here - that's only ever applied to sends, generally when you're talking about aux sends.

Anything you do to the track while you're tracking is there forever. It's generally considered good practice not to commit anything to tape you're not 100% sure about. So, record totally dry, and process in the mix. That's where de-essing, gating, compression and EQ come into play.
posted by god hates math at 6:22 AM on July 9, 2008

I hesitate to answer, since I rarely feel satisfied with my vocal mixes, but I have a question for you: If you are processing everything on a computer, why not record them raw and then process afterwards? Doesn't pre-processing limit you down the line? I'm sure I'm missing something obvious.
posted by umbú at 6:23 AM on July 9, 2008

Unless I'm tracking someone who is going to WILDLY vary in volume (such as screaming on a track), I don't touch it at all. No compression, no EQ, no nothing. I do everything in the mix later.
posted by chimaera at 6:58 AM on July 9, 2008

god hates math: Just using a hint of compression is what's usually recommended and that's more or less what I do. I apply a small amount of compression/limiting because while my I/O device's on-board mic-pres aren't bad, they tend to lose some of the details of the input signal at the lower-end of the dynamic range. I apply compression to the input channel on one of the post-fader inserts as part of my gain-staging process to ensure optimal input signal gain; the higher you can get your input signal without clipping the better (while at the same time leaving yourself ample room to apply more compression later, of course), because lower gain means signal resolution loss.

I used to religiously avoid using any kind of signal pre-processing during the audio capture stage (whether recording vocals or otherwise) because of the destructive nature of processing the input signal, but then I came across a couple of audio engineering articles that suggested judicious use of pre-processing (limiting compression, in particular) on vocals, and since then, I've found that while it's risky, I almost always get better results when applying a small amount of compression during tracking. A lot of higher-end studio mic-pres come with tube compression and various EQ shaping presets right on board just for those reasons; since my mic-pres don't do much more than provide phantom power, pre-processing allows me to shape the input signal in different ways as if I were using one of those fancy-pants mic-pres. At the same time, your point about not going overboard with processing is an important one.

I'm not quite what you mean by pre/post fader here

In Nuendo (and a lot of real-world mixer configurations), an incoming signal is routed through an input channel and you can increase or lower the gain of the signal or apply pre-/post-fader insert effects to it at record-time. Pre-fader insert effects are applied directly to the signal as it enters the channel before any gain boosting takes place; post-fader effects are applied to the signal as it leaves the channel after gain boosting takes place.

Here's an article that discusses the topic of pre-processing vocals in a little more depth. It notes that EQ shaping is also a fairly common pre-processing technique when recording vocals.
posted by saulgoodman at 7:37 AM on July 9, 2008

Ahh, yes. I see what you mean. Believe me, I'm very familiar with proper gain staging.

In a real-world setting, I can't think of a single instance where I'd want to put inserts postfader. A proper channel should basically go like so -

Pad->Pre->Phase->Filters->Dynamics->EQ->Aux->Fader-> (see page 2:7 of the Neve 88RS manual).
In a smaller board, the inserts are right after the pre, and before any kind of processing (see page 28 of the Mackie 1604 manual).

So, technically, you should be putting the compressor on a prefader insert, rather than a postfader insert. While it's unlikely that you'd move your fader once you've set your pre and your compressor, in the event that you did, you'd be changing the input level to the compressor, and that would throw off how much compression you'd be adding. This is true for other dynamics processors as well.

Also, I'm not sure about nuendo, but it's entirely possible that you're doing nothing by compressing to tape. In ProTools, for example, in order to compress to tape, you've got to make your input an aux channel, compress that, and buss the whole thing to a audio track. Putting a compressor on the track while you're recording just compresses the signal in the monitor path. Nuendo may or may not be the same.

As for that article, I've never known an engineer who would EQ a track before it hit tape. I'm sure people do, I just haven't found it to be that common. I even know a few people who use those all-in-one vocal input devices (like the Manley Voxbox, which actually sounds pretty good). None of them use the EQ - they really just tend to use the compression. The article also specifically mentions EQing to tape as a way to get to use analog gear within digital systems. The only way this would be beneficial would be if you've got an interface that doesn't have sends/receives. If you've got a single in/out, you can easily use the outboard EQ while mixing.

Also, I'm a little hesitant to trust the opinion of an article that says 16bit is "perfectly satisfactory." Discarding 48dB of dynamic range is dumb, no matter what you're doing.
posted by god hates math at 8:35 AM on July 9, 2008

Also, different microphones have different dynamic characteristics. I was tracking vocals not too long ago with both an MXL condenser and a SM-58 up to the singer, press-conference-style. Predictably on the condenser, it was vastly more sensitive at both lower and higher volumes (a good signal at very low whisper and peaks out easily, but in the middle volumes, the MXL was putting about -15dB into the input but the SM-58 was putting in -12dB.

Anyway, I guess the lesson is don't forget to consider the mic you're using to also determine what you want to do about dynamics.

When I do put a compressor on while tracking, it is ALWAYS pre-fader on the track insert. My mixer sends the insert signal after the gain stage but before the EQ.
posted by chimaera at 9:06 AM on July 9, 2008 [1 favorite]

Putting a compressor on the track while you're recording just compresses the signal in the monitor path. Nuendo may or may not be the same.

No, Nuendo applies any input channel insert processing to the audio signal coming from the bus and on to tape, not just to the monitor mix. It's one of Nuendo's more popular features, and is touted as providing a shorter signal path than some other mixer architectures.

you should be putting the compressor on a prefader insert,

it is ALWAYS pre-fader on the track insert

Actually, thinking again, I'm pretty sure I do and have always applied compression on one of the pre-fader inserts in the past. It's only during mixing that I usually apply it (well, limiting compression anyway) on the post-fader insert. But I have recently been having second thoughts about doing it this way (which is exactly why it's so nice to have this forum for discussing questions like this before I go making any rash choices in the studio).

Why couldn't limiting be applied as a post-fader insert, as long as the input signal isn't too hot, so you don't accidentally end up with clipping on tape? A little limiting compression just to smooth out occasional low peaks wouldn't introduce much compression noise would it?
posted by saulgoodman at 9:50 AM on July 9, 2008

It's only during mixing that I usually apply it (well, limiting compression anyway) on the post-fader insert.

This is just as bad/worse (as you've probably figured out by now) - as you change the level on the track, you're changing the amount of signal that's reaching the compressor, which will change how the compressor acts on the signal. Definitely bad. If you're truly only limiting peaks with that plugin, that says to me that you're not compressing correctly - once you've set the amount of compression on the plugin, you should be adjusting the output of the compressor so the signal peaks just below 0db. The track fader then adjusts exactly how much of the track you want in the mix.


Why couldn't limiting be applied as a post-fader insert, as long as the input signal isn't too hot, so you don't accidentally end up with clipping on tape? A little limiting compression just to smooth out occasional low peaks wouldn't introduce much compression noise would it?

Your question is a little confusing, but I'll give it a shot.

First, if you're really just limiting, you're not going to hear the compressor very much (I'll avoid the term compressor noise, since that implies that the compressor is noisy). Peak limiting is all about having a high threshold (right below 0db, usually), and a high compression ratio. Most of the time, the signal isn't going to be reaching the threshold, so the compressor won't be actively compressing. On those few times (peaks) when it does, the compressor goes in, squashes the signal, and it doesn't clip. The compressor goes back to passing the signal, unchanged, and all is well.


The phrase "low peaks" is messing with my head, so I'm not going to touch that part of your question, yet.

Basically, the only time I would even consider using any kind of post-fader insert would be on a Master Fader, with Dither (this is why, in PT, all inserts on master faders are post). This book about Protools also mentions an Aural Exciter as something that would go post-fader, but I'm still trying to figure out why.
posted by god hates math at 11:58 AM on July 9, 2008

as you change the level on the track, you're changing the amount of signal that's reaching the compressor, which will change how the compressor acts on the signal. Definitely bad.

i think you're misunderstanding me a bit. when mixing, i might apply compression as needed to the signal on a pre-fader insert (that is, the compression is applied to the signal before it's modified by any amplification or attenuation through the channel fader), but then i also run a limiter on the post-fader insert to keep the channel output under 0db (well, -0.5db).

why? because as long as the peak levels aren't hot enough to create noticeable compression artifacts (which as you point out, is hard to do with limiting compression alone) it makes managing levels a lot easier when dealing with a complex mix (like this one i recently finished, for example), especially mixes with multiple sub-mixes in group channels. also, since i'll probably use other effects like reverb, delay, etc. on the channel insert, and those might push the channel's output signal past 0db, it just makes sense to apply a post-fader limiter on every channel.

(I'll avoid the term compressor noise, since that implies that the compressor is noisy)

by 'compressor noise' i mean compression artifacts. and even a limiter can introduce noticeable compression artifacts into your mix if the signal flowing into the limiter is hot enough.

The phrase "low peaks" is messing with my head,

c'mon--you know what i mean. signals that creep into the red but not by much.

to clarify, i haven't tried using a limiter on the input signal. that's why i asked if anyone else has, and if so, how it's worked out for them.

the only time I would even consider using any kind of post-fader insert would be on a Master Fader

i like to do mastering as a discrete step (using a mastering lab tool suite) so i don't apply any signal processing to the master output fader, as it limits my wiggle room later.
posted by saulgoodman at 1:41 PM on July 9, 2008

i think you're misunderstanding me a bit.

No, I get it - you're both compressing and limiting the track. And I'm saying that, if you're using the compressor correctly, you shouldn't need to be using the limiter to keep your peaks below 0db. The output (that is, the make-up gain, which doesn't change the amount of compression) of the compressor should be set such that nothing is peaking above 0. It seems like using a limiter for that is an unnecessary step.

also, since i'll probably use other effects like reverb, delay, etc. on the channel insert

Another weird convention that I wish DAWs didn't let people do. It's generally a good idea to use time-based effects (delay, reverb, etc) in a send->return configuration. Putting them on individual channels is basically a waste of processor time. Especially if you're putting identically-set effects on multiple channels. Also, if you've got an aux input with effects on it, you can tweak the effects - compress your reverb, create a spin (buss the output of your reverb back into the input, for a slight reverb feedback), put delay on the verb, EQ the verb, etc. It's a much more flexible setup.

and those might push the channel's output signal past 0db, it just makes sense to apply a post-fader limiter on every channel.

Nope. I'm sorry, I'm not trying to be a jerk or anything, but that still doesn't make sense. Even if I thought it was a good idea to have things like delay and reverb on your track, adding gain, the obvious solution to that would be to pull your fader down. If you're clipping pre-fader, fix it in whatever gain stage is causing you to clip. And here's the thing - even if what I just wrote wasn't true, it still wouldn't make any sense for the limter to be post-fader. You would want it prefader so that - 1. you wouldn't clip the channel, and 2. so that if you adjusted the track volume with the fader, you wouldn't change how much peak reduction is happening.

c'mon--you know what i mean. signals that creep into the red but not by much.

No, really, I didn't. I'm fairly sure that ProTools doesn't have soft clipping (and I can't find a reference to confirm or deny that right this moment), so I try to be as rigorous as possible about not clipping out channels/busses/plugins/whatever. Even if you can't hear the clipping in your signal chain, if it's there, it could be audible (it depends on the D/A stage of whatever someone's listening on) on just about any piece of equipment. So unless you're talking about a tape machine, it's either clipping or it's not.

i like to do mastering as a discrete step (using a mastering lab tool suite) so i don't apply any signal processing to the master output fader, as it limits my wiggle room later.

I'm not talking about mastering - In any session in which you're recording at 24-bit and mixing to 16-bit, there should be a dither on the last insert of the master fader - it randomizes the quantization error that arises from chopping off those last 8 bits. If you're mixing down to 24-bit, so you can master in 24 bit, then the dither needs to be on the output of the master fader in the mastering session.
posted by god hates math at 3:28 PM on July 9, 2008

If you're mixing down to 24-bit, so you can master in 24 bit, then the dither needs to be on the output of the master fader in the mastering session.

Yep. I always master my final, final mixes from 24-bit stereo mixdowns. And that's exactly how/when I apply dithering.

I'm fairly sure that ProTools doesn't have soft clipping

I use Nuendo, not ProTools, so I really, really don't care how it works in ProTools! /kidding

Also, FWIW, there are no effects sends as such in Nuendo (more on that here). You can create effects channels and route stuff to them, or you can create group channels that basically let you route multiple mixer channels to a signal channel for processing as a group (where you can cheaply apply a single reverb to a lot of different channels, for example). All this talk about effects sends and returns is disorienting to me--although they're still familiar concepts from back when I used analog gear.

I'm shamelessly derailing my own damn thread here, but let me try to explain why I think this mixing technique does actually work. By applying compression on a pre-fader channel insert in Nuendo, you can make what's otherwise a weak-sounding signal on tape sound punchier and more detailed. When processing the signal within the channel (again, in Nuendo at least), any signal peaking that occurs as the signal flows out of the pre-fader compressor is more than likely going to sound just hunky dory. In fact, a hot outgoing signal in such a routing scenario tends to impart some desirable characteristics to the sound. And that's the important point: The results in the mix are actually better (to my ear, anyway) if I don't limit the outgoing signal from the compressor to below 0db. I'm repeating myself here again for emphasis: The outgoing, processed signal sounds better if its a little hot (even as much as +2.0db) when it hits the channel fader.

That's where applying a limiter in the post-fader position comes in. As you pointed out, I could just set the channel fader so that the signal never peaks above 0db, but then, I might not be able to get the track to sound as loud as I'd like it to in the overall mix without the output signal occasionally clipping (which does cause audible signal degradation). Using fader automation as a work-around would be a big, processor-resource intensive pain. And applying another pre-fader limiter wouldn't give me the flexibility to boost the gain of the channel's outgoing signal while still retaining the desirable characteristics I got from the pre-fader compression. And finally, whether you accept this explanation or not, this mixing technique just seems to works for me, and I trust my ears over even the best arguments.

Well, to keep this thread from descending any further into the saulgoodman/god hates math show, here's a random signal-processing primer for the kids back home. (It's a pretty basic introduction, but I haven't had a chance to check out the thread on other online resources yet and have had trouble finding others that aren't behind some pay-wall.)

Nope. I'm sorry, I'm not trying to be a jerk or anything, but that still doesn't make sense.

You're not being a jerk. You're just being passionate about an interesting, if incredibly geeky subject. Nothing wrong with that (although my own experiences don't align with some of your assertions).
posted by saulgoodman at 9:04 PM on July 9, 2008

Also, FWIW, there are no effects sends as such in Nuendo (more on that here). You can create effects channels and route stuff to them, or you can create group channels that basically let you route multiple mixer channels to a signal channel for processing as a group (where you can cheaply apply a single reverb to a lot of different channels, for example).

D'oh. Actually, the above statement is pretty much completely inaccurate. I guess I've just never thought of it as using an effects send when I route a channel output to a group channel or to an effects track when mixing. My brain just more or less crapped out on me for a minute there. Sorry about that.
posted by saulgoodman at 9:17 PM on July 9, 2008

This is the reason I love analog tape; you can record raw without worrying so much about limiting or whatnot, compared to digital (at least with the gear I've used.) 100% encouraging you to record the tracks raw and process 'em later (if you have to.) If you've got a singer who doesn't like how they sound raw, you can always process their monitor feed to their liking.
posted by davejay at 11:20 PM on July 9, 2008

This is the reason I love analog tape; you can record raw without worrying so much about limiting or whatnot, compared to digital (at least with the gear I've used.) 100% encouraging you to record the tracks raw and process 'em later (if you have to.) If you've got a singer who doesn't like how they sound raw, you can always process their monitor feed to their liking.

I definitely agree with you, Davejay: Less is almost always more when tracking (and there's more fault tolerance in analog--unless, of course, you find yourself needing to make a tweak or edit further down the line). Judicious use of compression is sometimes recommended, but like chimaera pointed out, you have to pay attention to the dynamic characteristics of your gear. At the same time, none of that means it isn't sometimes fruitful to experiment. And one way or another, some amount of signal pre-processing (whether in the analog or digital domains) is inevitable before a signal hits tape. Condenser mics, for example, usually have EQ shaping built right into the circuitry (and my mic, for example, has a low-cut filter switch). That kind of stuff is a form of pre-processing, too.

god hates math, just a couple more points to belabor, because I think some misinformation/confusion has made its way into this thread. This post was originally supposed to be about pre-processing vocals, but much of the discussion we had above regarded signal processing during mixing after getting the signal to tape. Maybe some of our disagreement stems from confusion about that. It's also really important to be aware that different tools have different signal flow architectures.

Also, you commented that you couldn't imagine why anyone would apply a limiter in a post-fader position. Well, I on the other hand, can't for the life of me imagine why anyone would ever use a limiter in a pre-fader position (okay, maybe sometimes, like at the end of an effects chain). The purpose of a limiter is to prevent clipping by attenuating the gain at the upper end of the dynamic range of a signal (not to flatten its dynamic range). And if a signal is already peaking before it even hits the fader without any additional signal processing, then the levels were too hot when you tracked it, and any clipping of the signal on tape is already irreversible. And even though the signal with pre-fader limiting won't be peaking internally when it hits the fader, and the fader can be used to prevent peaking in the channel's output signal, any signal degradation caused by clipping within the signal processing chain before the signal reaches the limiter will still be there, only it could be a lot harder to figure out where in the chain the damage was actually done.
posted by saulgoodman at 8:21 AM on July 10, 2008

There are no rules. Rules of thumb? Yes. But nothing etched in stone.
posted by chillmost at 9:44 AM on July 10, 2008

yes there are no rules- just whatever sounds good.

but post-fader inserts are just wrong. I've working in a lot of recording studios (mostly tape) and I've never seen that used except for a wacky effect.
posted by bhnyc at 11:20 AM on July 10, 2008

I've working in a lot of recording studios (mostly tape) and I've never seen that used except for a wacky effect.

Well, this particular architecture (where pretty much all signal processing is done using channel inserts of various kinds instead of effects sends/returns) is unique to Cubase and Nuendo, I think. Nuendo effects inserts aren't really analogous to effects inserts in other DAWs (and they work nothing at all like the effects inserts in analog gear).

Everyone I know personally who primarily uses Nuendo uses post-fader limiting in the way I've been describing when mixing--if not on every mixer channel, then at least on hard-to-level channels. Maybe we're all doing it wrong, but it works for us.

I guess the take home point for me from all this is not to make assumptions about how techniques that work in one particular production setup (digital or otherwise) might translate to other setups. This has been a helpful and interesting (if rambling) discussion for me.
posted by saulgoodman at 2:21 PM on July 10, 2008

I don't pre-process at all. But I have problems with commitment.
posted by frenetic at 2:14 PM on July 13, 2008

I'm justa bedroom recording type but anyway - I have an Joe Meek Pre-Amp / Compressor box which also has a low frequency roll off switch (thing?? to remove low rumbles below 20or 40Hz or something.)

I usually run the mic into the Joe Meek. switch on teh rumble remover, add a little bit of compression - just in case it gets too loud, then straight into the soundcard. Record at 24bit if I'm feeling precious.
posted by mary8nne at 2:37 AM on October 20, 2008