Using The Limiter On Vocals

D

Doctor Varney

Cave dwelling Luddite
Me again, with the audio spoken word...!

In trying to speed up my game here, I decided to experiment with a limiter algorithm and wonder if I'm wise to use it. Let me first explain the situation...

When I record my speech into the audio logger, despite my experience with voice work, there are always some inevitable peaks which cause a limitation to the normalisation algorithm, in that the waveform won't normalise anywhere past 0dB (which is the whole idea of normalisation, right?). So I do often find myself manually attentuating some of the naughty spikes. Then, when normalisation takes place, I get a louder recording overall.

Before anyone says "Stop looking and listen..." let me say that the visual waveform is my indicator of what's going on, before I hear it. I've learned to associate what I see with what I hear in a reasonably accurate manner. I DO listen more than I look at waveforms on the screen and the question surrounds what I should be aiming for in audible terms.

Now I've just discovered that I can apply a limiter algorithm to the waveform which saves a lot of manual work. It could actually mean the difference in turnaround for my business, if you add all the minutes I spend manually editing. Of course, I'm not aiming to remove all of the variance in the waveform to undynamic blandness but some form of conformity of the peaks is what I understand as the job of a limiter and this is what I want, only to a less aggressive degree.

So far I've got as far as applying the default setting and what I end up with is a very mechanical looking waveform. On first blush, it doesn't sound bad but it's hard to tell if this would suit the discerning ear or the equipment it's likely to be played back on. I remember someone in a previous discussion suggesting that the full dynamics of the voice are needed in the spoken word and to an extent, I would tend to agree. It's just that these excessive spikes are causing the rest of the recording to be rather quiet. I figured I've got some technology here, so I might as well learn how to use it to speed up the process.

What I'm attempting to ask is whether limiting vocals is a standard practice and begging for some tips on how to approach it. I've heard of 'hard' and 'soft' limiting and I presume 'hard' is what I've achieved here, with the even 'comb' effect; while 'soft' limiting is going result in more variance (dynamics) to the sound? I'm guessing here, so have I got that right?

Also I read somewhere that a limiter is simply a compressor which limits everything to +0dB gain. If so, then I presume I've chosen the right tool, as +0 is pretty much where I want to start at and then drop levels later in the mix, to give some overhead.

I don't feel 100% confident with my use of the limiter yet, but I think I'm getting it. I just want to learn more about it's proper uses in the industry (esp. with respect to vocals). If anyone could offer any rules of thumb; dos and don'ts; things to look out for, etc, I'd be most grateful. I've heard a few jokes about people who abuse the compressor and the limiter and so obviously I would prefer to get off on the right foot with this concept.

And yes, I am looking towards purchasing a better mic in future - but for now, I must stick with what I have and make the best use of it for the time being.

Thanks.
 
Yes, a limiter is a pretty standard tool in voice recording. In most professional studios there's one across the output to prevent clipping on an otherwise good take.

However, if I read your post correctly, where you may be going wrong is attempting to track at 0dBFS. I still use a limiter on most things I record..but my gain structure is set so the tracking level is averaging around -18dBFS with peaks going above that...generally to somewhere in the -12 to -10dBFS range. This is equivalent to something like +6 or even +8 in the analogue world. Working this way, the limiter is a safety net that kicks in a few times a year, not something relied on while recording.

My suggestion is to track lower and get your levels up to somewhere near zero as a last (i.e. mastering) step.
 
Unless there's something very wrong with your recording chain, raising the level won't add any distortion...the worst it could do would be to make distortion that's already there more audible. On the other hand, recording at 0dBFS risks clipping distortion if even one sample goes a fraction of a dB higher than it should since 0dB is the point at which a digital recording system runs out of bits to go any higher.

To generalise horribly, I tend to use a limiter as a protection device while compression is a production tool.. The sort of place I use a limiter is to prevent any peak going into clipping while recording (but NOT driving it so I'm always into limiting. Or, in broadcast work, there tends to be a limiter on a studio output to avoid a peak over driving the transmitter...or, in a live sound situation, it's pretty common to have a limiter before the amp/speakers so you can't overload those. Although there are exceptions, I don't often use a limiter as a standard part of a mix.
 
The simple answer is don't use a peak normalizer. Use editing and compression to make things sound even, then use a mastering limiter to get your signal up to whatever RMS level is appropriate.
 
Peak normaliser? Is that the same as 'normalisation' or is that another word for limiting?

I'm starting to get confused. Can anyone please give me an idea, from start to finish, how to record and process my vocals? I'm reading and reading and trying to learn this but I'm still not sure how to apply any of this properly.

The biggest problem is that I can't see what these plugins are doing and all I'm hearing is a difference in volume when I adjust things. I need to see what they are actually doing to the spikes, otherwise I'm just knob twiddling with no understanding and that's bad. Now, I have the option to apply a limiter algorithm to the .wav itself before any post processing begins. That's the only way, at this stage in my education, that I'm going know whether or not I'm doing this right.

What level should I end up with in the end? Should I aim for +0dB on the main output or what?
 
Last edited:
bouldersoundguy said:
The simple answer is don't use a peak normalizer. Use editing and compression to make things sound even, then use a mastering limiter to get your signal up to whatever RMS level is appropriate.
Click to expand...

How will I know when they sound even? It's just occurred to me, all what I said about listening was me believing my own bullshit. I admit, I need to see waveforms or I have no idea what I'm doing. I just can't tell what's going on, just by listening. I've absorbed all this theory... and when I sit down to listen, it either sounds louder or quieter and that's all I can say about it.
 
Bobbsy said:
On the other hand, recording at 0dBFS risks clipping distortion if even one sample goes a fraction of a dB higher than it should since 0dB is the point at which a digital recording system runs out of bits to go any higher.
Click to expand...

Well, could I just forget all about the limiter on the output and just use this instead (as that 'safety measure)?
Fruity Soft Clipper - Effect Plugin

This is what I've been using as a noise gate, directly after my recording, to kill the hiss.
Fruity Limiter - Effect Plugin
It does the job really well. I don't think I'm using it as a compressor or limiter, but as an expander...?

So then is it alright to use the gain knob on the soft clipper to bring my overall output volume up to +0dB?
 
Quite seriously, set your file playing, push your chair back from your keyboard and close your eyes. You may be surprised how different things sound without the visual clues and clutter. If it sounds right, then it's probably right.

I wonder if you're over-thinking things a bit too much. With the spoken work, less is more in terms of processing.

Spend a lot of time on mic placement and aiming (and any noise/acoustic issues in your room) to get the sound as close as you can to what you want to hear.

Record so your PEAKS are at least a few dB below the zero mark--say between -6 and -12.

Add any EQ you want--but don't go overboard, you shouldn't need much if you've recorded properly. Maybe a subtle boost in the upper mids to aid intelligibility, maybe some low cut since you probably don't need those anyway.

Then use a compressor to even out the dynamic range. How much? It depends on both your target audience and what you're recording. For example, audio books (which are often listened to in cars, on trains, etc.) can't tolerate to much difference between the loud and the quiet bits. And that's probably it for "production" processing.

The final step is to bring the levels you have at this point up to the maximum they can be for burning to CD. If levels are pretty consistent, just normalising to -0.3dB or so might do it. If there are still a few big peaks (for whatever reason) maybe the mastering limiter as suggested earlier. To use a limiter, at this point DO use your eyes. Note on the waveform were the vast majority of your peaks are sitting. Let's say they're at -5dBFS. Tell the limiter to limit at -0.3dB and to add 5dB to your levels. Voila...the few random peaks are gone and the whole file is as loud as it can be.

...but, basically, do the least processing you can get away with.

Edited to Add: Your post about the fruity limited popped up while I was typing the above.

If you have hiss to get rid of, then something is wrong with your recording chain...maybe the gain structure or maybe a faulty piece of equipment. Gates/noise reduction should never, ever be a "standard" part of your process. They're there for restoration of old stuff, not a normal part of the chain. Time for going right back to the basics rather than messing up your product more and more with extraneous processing methinks.
 
Thank you so much, Bobbsy, for taking the time to explain that. I really appreciate it. So, I think I will have to start from scratch (again).

The reason I have the noise gate there, is because I'm using a crappy mic and that's all I can afford for now. I just blew all my budget on FL Studio so I could have a DAW. It was either that or carry on using the illegal crack, which was breaking down on me and made me feel like a freeloading scumbag, anyway. But I need to get started, to recoup the cost of the DAW so I'm planning to buy a better mic later on. If the books sell, that is. If they don't, then I've got a nice hobby DAW to play with...

Cheers, mate. I value your opinions and hope I can ask a few more questions as I go along?
 
Would it be okay to apply compression to the waveform and save it to limit those spikes? With the limiter I find I can't really tell what's going on. It would, however, be a faster process to just have the limiter/ compressor on the output so that everything going into it is soft clipped but for me, it's a confusing as hell piece of kit to get my head around.

And how loud should I record?
 
Bobbsy said:
Record so your PEAKS are at least a few dB below the zero mark--say between -6 and -12.
Click to expand...

The only way I can get that is to use some kind of amplifier on the way in and people have been telling me to record dry. That just leaves the pre-amp on the mixer and it doesn't make any difference because if I hit 'normalise' on the audio logger, it goes to the same level anyway. I figured I might as well record with the preamp somewhere in the middle or quite low to reduce the noise.

Bobbsy said:
Tell the limiter to limit at -0.3dB and to add 5dB to your levels. Voila...the few random peaks are gone and the whole file is as loud as it can be.
Click to expand...

Any chance you could take a look at this thing and tell me how to adjust the knobs to tell it to do that?
Fruity Limiter - Effect Plugin

Sorry to be a pain.
 
Doctor Varney said:
Peak normaliser? Is that the same as 'normalisation' or is that another word for limiting?
Click to expand...

I added "peak" to normalization to make it clear what it does, make the peaks of the different parts (files, blocks, whatever) the same. But as you know from your experience peaks can vary. If you have two pieces of audio, one with peaks 6dB above the average and one with peaks 12dB above the average, and you normalize them, the one with the higher peaks will sound quieter by 6dB. It's the average level of the audio that you need to match up, not the peaks.

If you're getting that much noise then there is some sort of gain structure problem or equipment problem. It can be hard to help with these things remotely since there is so much information lost by putting it into writing.

The tracking process should be fairly straightforward. Set your preamp gain so when you speak the peaks never even get close to 0dBFS. Something like -12dBFS peak level would be fine.

After that you can edit and/or compress to get things sounding fairly even. Some eq could be done, preferably after editing and before compression. Don't overdo it. Leave plenty of headroom at this stage. It might be a good idea to note your monitor volume level so you have a consistent reference point.

Once the recording and editing is done and you can comfortably listen to the whole thing without adjusting the volume it's time to master. This is when you get your final level. Use a mastering limiter because the controls are designed to make it easy compared to using a generic limiter.
 
THE BEST PLACE TO USE A LIMITER WITH VOCALS..

is during tracking.

this is where you impart the most sonic fingerprint of the sound you want.

it does no good to apply a limiter AFTER you've already hit the convertors, unless it's just a mixdown level thing.

you need a good outboard limiter.
 
GONZO-X said:
THE BEST PLACE TO USE A LIMITER WITH VOCALS..

is during tracking.

this is where you impart the most sonic fingerprint of the sound you want.

it does no good to apply a limiter AFTER you've already hit the convertors, unless it's just a mixdown level thing.

you need a good outboard limiter.
Click to expand...

It's "just a mixdown level thing" in this case.

Sure, it you're an experienced engineer and you're not also the speaker and you have a Manley ELOP or something else with a distinctive sound then go ahead. If you're an inexperienced engineer trying to record yourself through a cheap mic then there's nothing particularly advantageous about using a hardware limiter. Besides, his interface may not have inserts to accommodate hardware processing.
 
bouldersoundguy said:
If you're getting that much noise then there is some sort of gain structure problem or equipment problem. It can be hard to help with these things remotely since there is so much information lost by putting it into writing
Click to expand...

Thank you.

It's hiss and it's down to the cheap mic I'm using. There's no way around this until I find the money for a better one, other than using plugins to compensate. Then there is some room treatment and eliminating computer fan noise, which will take time and money, I just don't have right now. I've found that the limiter, set to a noise gate preset, just after the tape deck eliminates the hiss almost completely. The reverb for the voice then covers up what hiss is left during the speech. Silent areas, where I'm not speaking are absolutely hiss free.
 
Last edited:
GONZO-X said:
THE BEST PLACE TO USE A LIMITER WITH VOCALS..

is during tracking.

this is where you impart the most sonic fingerprint of the sound you want.

it does no good to apply a limiter AFTER you've already hit the convertors, unless it's just a mixdown level thing.

you need a good outboard limiter.
Click to expand...

The is spoken word, not music. If he's tracking at an appropriate level, there's no need (indeed, no likelihood) of ever overloading the converters.

Doctor Varney said:
Thank you.

It's hiss and there isn't anything I can do about it until I can get the room treated, get a better mic and minimize computer fan noise. These are things I'm working on but they'll take money, which is something I just don't have at the moment. I've found that the limiter, set to a noise gate preset, just after the tape deck eliminates the hiss almost completely. The reverb for the voice then covers up what hiss is left during the speech. Silent areas, where I'm not speaking are absolutely hiss free.
Click to expand...

The thing is, there's LOTS you can do to improved the recordings without spending money.

From your description, your "hiss" is, in fact, computer fan noise and some room atmosphere.

First, get a longer cable and move farther away from the computer. The physics of sound dictates that every time you double the distance from a sound source the amount reaching you (or your mic) is quartered. If your mic is 2 feet from the computer fan now and you can move to 4 feet, you will only have as much fan noise. Move to 8 feet and it'll be only a sixteenth what it was.

Second, use the mic's pick up pattern to your advantage. Assuming it's a cardioid, directly behind the mic will have practically no pickup so arrange things so the computer/fan is directly behind the mic.

Third, minimise room atmos by placing yourself in front of a set of heavy curtains that don't reflect the noise back into the mic. No heavy curtains? Hang a quilt or something behind you. You'll be amazed how much this can help.

Fourth, work close to the mic and speak up. Just as doubling the distance quarters the fan noise, halving your distance gives you four times as much voice level (compared to any noise).

(Fifth, next time buy Reaper instead of FL studio and save money for a good mic! But that's unfair of me to say! So I won't. Ooops, I did.)

Being honest, you're trying to convert this into a business and make money. As such, you're in competition with a lot of other audio books recorded in proper studios. I'm trying to imagine what a voice recording sounds like heavily gated and with reverb added to disguise other faults and, no matter how I imagine it, it ain't pretty.

Sorry for being a bit blunt--I've woken up in a grumpy mood and haven't had sufficient caffeine yet today!
 
Bobbsy, change that description from 'Boring Old Git' to something more positive, mate. That is golden advice which I'm taking as we speak. To be honest, I wish my common sense had told me to try getting further away from that whirring box of fans.

Don't apologise for being blunt, it sounds alright to me. Even if it were, honesty is more valuable than pussy-footing around and politeness never put food on the table. If my audience doesn't like it, it won't matter how much I convince myself I'm doing a good job, if they complain about the quality of the recording. I want it to be as good as it can be, given the limitations.

Ah, fair comment but I chose FL Studio because I just love working with it. The thought did go through my head to buy the mic first - but then this half price offer came up on the full Signature Bundle which ended up cheaper than a good mic. I'd have been a fool to miss it. The Mic I'm using cost £14.99 from a Cash Converter second hand gear shop. It retails for about £39 I think.

Finally... thanks! :D
 
Bobbsy said:
If he's tracking at an appropriate level, there's no need (indeed, no likelihood) of ever overloading the converters.
Click to expand...


well, i'm not talking about only level, i'm also including color.

this is important.

besides, a lot of folks simply don't understand that putting a software limiter on AFTER you've already committed levels to virtual "Tape" isn't going to get you anywhere.

why not nail your take, with the limiting for color AND level, while you're actually doing it?

if you're not a recording purist, then forget everything i said.
i'm just saying, it's silly to try to solve a problem after the fact.
much better, to solve it right at the get go.
sometimes it takes the 'right' equipment, and some folks want to know that.
 
You must log in or register to reply here.
Back
Top