Wednesday, March 28, 2012

Dear SNR: Vocals In The Mix

I've been starting to publicize the blog a little now that there's a bit of content up and lo and behold it's working.  I suspected a few redditors were reading and now here's the proof.  Here's our first true "Dear Soundguy" letter so let's get right to it.
Hi there,

I recently stumbled upon your blog from a reddit link (I think it was in, which makes sense). I wanted to get your advice on something which has been bothering me for a while, if you have a few minutes to spare out of your day to help a newbie!

My basic problem is thus: when i'm mixing vocals at higher volumes (funk bands, rock bands, rockier theatre shows, etc), a lot of the time, I end up with a melange of that ear-hurty icepick-in-the-ear-sounding interference which kinda sounds like it's somewhere between 2k and 5k. I try my hardest to be aware of this and notch it out either on the FOH graphic or on a vocal subgroup, but I was wondering if you have any tried-and-true methods to keep clarity and intelligibility in the source material (i.e, a couple of vocalists singing over a full band), whilst also actually dealing with the problem? I've heard it done by other engineers, but they're fairly secretive of their mix techniques when I ask, and I don't want to be as unsubtle as to try and get a look at their EQ pages or racks.

Do you have any thoughts on this? I fully understand if you also want to try and keep your mix techniques to yourself, just looking for a starting point, really.

I also weaseled a few additional facts out of him via email and the interference he's talking about isn't feedback, but just a sharp sound.  The mics he's using are generally Shure SM58s and occasionally Beta 58s which he doesn't seem to have as much trouble with although they usually show up when he has fewer vocalists on stage.

I'd like to start by addressing the bit about trying to get info off of other pros.  It's not unreasonable that they don't want to give away any details to a random person that walked up to them.  If you've got the time it's always a good idea to try and strike up some sort of relationship with another mixer.  If you can find a minute when they're not too busy you can try to start swapping stories, offer to buy them a drink or run an errand for them. And it never fails to warm the heart of a pro if someone offers to help strike at the end of the night, but most will turn you away if it's a verbal offer.  If you're still there when the club is empty and just ask, "What can I do for you?" it's not likely you'll be sent away.  But at any rate, I'm glad you came to me because now my readers get the benefit of the exchange and you get your answer in print.

Let's start where you started, with EQ.  You've already identified a range where you think the problem is and it sounds like you've tried hunting around for it a bit.  Let me say that I'm encouraged to hear that you're trying to notch it on the FOH graphs or on a group.  The group would be my choice for a vocal EQ because it leaves those frequencies untouched for other instruments that live in that zone.  2kHz to 5kHz is a pretty big range though, more than an octave, and right in that range are some frequencies that are the most important for vocal intelligibility.  If you loose too much 2.5k then the consonants start to disappear. So I would say that you're on exactly the right track, you just need to fine tune a little.

A real time analyzer is a good place to start.  Until you have your ear really trained, looking at an RTA, even one on your phone or iPod, can help you catch on to problem areas in a flash.  It's a crutch though, and I would advise to never use things like auto-EQ functions on boxes that have them.  I don't ever want a machine to do my thinking for me.  But I'll use one as a crutch at first, then a tool, then I'll fly on my own.  There are a few good frequency training tools out there, none of which I can remember the name of at the moment. 
If you don't have an analyzer for whatever reason, you can always use a swept EQ to hunt down the offending frequencies.  I usually take a full cut on a mid EQ and sweep it back and forth until it sounds the best, then ease up on how deep the cut is, and play with the width as well if that's an available control.  In a pinch, if I'm really stumped I'll do the opposite and crank the gain instead and when I find the spot where it sounds worst then I'll cut, but that's a pretty awful thing to do to any other people in the room.  It does work but it's like hitting yourself repeatedly in the thumb with a hammer... it feels so good when you stop.  You can do the same trick with a graph as well, cutting and boosting individual frequencies until you find the right spot.

Whatever your method, once you've got the culprit you need to decide how to deal with it.  If you've been sweeping parametrics EQs around, then you've solved your problem already.  But if you're squashing a wide band of frequencies and don't have a Q (width) control then you may want to look at a graph so you can be a bit more surgical about it.  On the other hand, if you're using a 31 band graph and you're mashing four or five sliders to fix your issue then that's not the right tool for the job and you do want a nice broad parametric filter in there.  The reason for that is if you make a curve on the face of a graph, what your actually getting is let's say five notches that are very narrow at the bottom but widen out toward the tops.   So instead of that nice smooth curve you have a weird comb shape.

Now let's move on to another idea.  Compression was never mentioned in your question.  So I'm going to suggest that if you're not using any, you get some right away!  What you might be hearing when you pile on a lot of vocals is each mic sounding nice and clear, but the additive effect of a few of them together gets to be ear piercing.  There's a few different ways you can tackle this.

You could always just insert a compressor on each vocal channel, but you might not have that many and it's also a good idea to keep compression out of the monitor mixes so the musicians aren't straining at the mic without getting any louder in their wedges or IEMs.  So save some comps and just insert one on your vocal group.  That may be enough right there.  A fairly low ratio, like 2:1 or just a touch more, but that kicks in sort of early (low threshold) will engage smoothly and really sort of automate your vocal mixing for you.  If there's one singer belting away they'll barely be touching it, and have control over their own dynamics but when the others join in, the mix will only get slightly louder but much fuller and it happens without any input from the engineer, once properly set up.

To take dynamic processing a little further let's look at a little used trick.  Let's say your vocalists are pretty well trained and have good dynamic control all on their own.  (Ok all you old pros, it doesn't happen all that often but it could.)  Go ahead and insert that comp on your vocal group, but then insert a graphic EQ on the side chain with another insert cable.  Once you've identified your culprit frequencies, push those sliders all the way up on the graph, and reduce all the others a little.  What this does is split the signal inside the comp and uses the signal running through the EQ to control the signal that you actually hear at the output.  It's typically used as a "de-esser". With the highs cranked the comp only goes to work when a big S comes down the line and the rest of the time the signal is free to do what it wants.  Just pick your frequencies and it can easily be a mid-tamer or a popped P killer.

There's just one more idea I'd like to cover.  You said that the problem occurs less when you have Beta 58s on stage.  I think that may be because the pattern is a little tighter and you get less bleed from them.  I'm not big on Betas but I like them for drummers for just that reason.  So if you're getting bleed from drums and guitars on four or five vocal mics, you're also getting bleed from any vocals in the monitors.  There's another hefty dose of mids there and your problem might be fixed just by attacking those same frequencies on the monitor side, just be careful you don't go so far as to hurt intelligibility and leave your singers stranded with a mushy mix.  Let me also say that microphone selection can be huge, but you often don't have any control over this.  The venue only has what it has, and you can only afford what you can afford.  Turn to other guys for their favorite vocal mic suggestions, mine's the tried and true SM58.  It hasn't been on stages for forty years for nothing.

That's how to fix it with more EQ, and here's how to fix it with dynamic processing.  I actually wrote a whole article on gates and expanders but it's not set to drop for a few more days so here's the skinny.  A gate does just what it says, it's either open or closed and sound can either pass or not depending on where you set the threshold.  An expander does the same thing but instead of closing it just reduces volume by a set amount.  Think of it like a compressor but it's working on the quiet end of the spectrum instead of the loudest.  On a digital desk you can have one on every channel, in the analog world it can be harder to implement but way worth it.  
If you've got a quad expander/gate you could take care of quite a few vocal channels on stage.  In this case you can use the inserts because you're only going to be affecting the audio when it's not there.  When the band is playing but a singer isn't singing, get it set to reduce their gain by about 6 dB.  That should be enough to help keep monitor bleed out of the house and still let it open up smoothly for a subtle backing vocal.  Your results may vary, and it can be a little tricky to get right, but it's a technique that's well worth mastering.  I've had it save the day more than once in a nasty situation like under a big top tent on a sticky summer night.

I hope that helps Craig.  It wasn't my intention to choke you with a novel, but at least if you've got it in writing you can work though it a bit at a time until you've got it.  As with any topic I'm not trying to write the definitive article here, just pass along some of what I've learned and hopefully encourage further research and experimentation on the part of the reader.  I'm also relying on my pro readers to jump in on the comments section and straighten me out on anything I'm missing or not getting right.  Feel free to throw in your own tips and tricks at any time.  
Till next time my Brethren of the Knob and fader...

1 comment:

  1. All the above is great advice. A lesson from a guy that started in recording and then ventured to live sound is always consider your source. Not all vocalists sound great on an SM58 or any mic for that matter. If I can I always like to get vocalists on a hypercardioid mic so as I can get rid of all that bleed up front. Let's be honest though, there are so many vocalists with bad mic technique that a hypercardioid also doesn't work all the time.

    With your particular situation of harsh mid range on vocals I would take a look at your mic selections, crossover points, type of horn, and room acoustics

    I've never been a big fan of the entry level JBL product as it always seems to have a harshness in the range you describe. I have a feeling that it has a lot to do with their choice of crossover and power going to the horn. When you have more vocalists on stage you may just be exciting this range more than normal. If your system isn't already a 3-way you may want to consider that so you can have a bit more control of the drivers.

    Another friend of mine has suggested flipping the phase (Okay in reality polarity) on a couple vocal mics if they are singing in close proximity to one another.

    I would seriously take a look at crossover setup and horn/driver being used before having to make your eq look like a jack o lantern


You're the Scotty to our Kirk