Compressor that doesn't 'snatch' or 'click' on speech?

  • Thread starter Thread starter leecovuk
  • Start date Start date
L

leecovuk

New member
Hello everyone,

Before I pose my problem and question, I suppose it helps to paint a picture of what I'm doing with a compressor;

In an amateur/hobbyist environment, I'm compressing a program output mixture of speech and music. I need to be able to compress fairly hard, somewhere towards limiting, without causing unpleasant compression artefacts.
Along with Gain Makeup, the aim is to produce a levelled signal with little dynamic range.

I have read much advice and reviews, and tried various compressors, including currently an FMR RNC, dbx 166XL and dbx 1066. I sometimes try a couple of units in circuit together, tweaking them this way and that and using their various modes.
I _suspect_ most of the advice I have pursued may be from musicians such as guitarists, rather than people like me. ie it appears many compressors may be more suitable for dialling in a very specific setting for a certain instrument, for example.

My problem is, I have ongoing troubles with finding a compressor that doesn't 'snatch' or 'click' sometimes on speech. I find the trouble usually at the start of a sentence, and maybe more so on female voices.
Is this an issue of the attack time not being fast enough?
If so, or indeed otherwise, do you have a compressor that doesn't exhibit this issue/artefact?

Let's take budget/price range/geographic location out of the equation for now ... I'd rather save up and look hard for a suitable unit rather than continue my trawl through the budget-mid range, which in itself is proving to be a false economy.

Many thanks,
Lee
UK
 
I know you said 'no dynamic range, and you may be able to get around those artifacts, but that is the sound of extreme abrupt gain reduction.
Now a question may be- if it needs that much reduction, isn't that an indication that section (or what ever) is simple too loud and can be turned down so the comp doesn't constantly have to make these big jumps?

Is this an issue of the attack time not being fast enough?
Click to expand...

Or too fast? Sometimes compromising by giving a little time letting some of the front slip through sounds more natural. Also begs the question, is the comp being allowed to release fully for no good advantage (bringing up gain between sentences for example) just to be hit again on the next?
How about fixing large variations with up front with automation? Way more natural sounding lightening the comps load.

Aphex 320. Haven’t needed one, but highly rated.
http://www.aphex.com/320A.htm
 
I essentially agree with mixsit. There is no electronic magic that will make voiceovers sound good when they are stomped on with a limiter. It is far more likely that you are hearing the abrupt effect of severe compression, than artifact. Vocals, including voiceovers, have a natural dynamic range, and you can't prevent that with compression without making a mess of it. Most likely, you need a slower attack, a lower ratio (slope), and a slower release. In short, you need less compression, not more. I have never heard a limiter applied to vocals that did not produce the effects you are describing, and in some ways, a very good compressor will make it worse, not better.-Richie
 
You can go in and edit the speech using a wave editor...then add it back on
 
Hello again and thanks for those replies.

(Re: the wave editor - I assume that's something for use with pc audio? What I have is not related to the pc; it's a setup with analogue audio from a mixer going through a hardware compressor unit.)

I may have over-stated my talk of limiting; whilst that would be an ideal aim, I still find the problem with very moderate compression at a 2:1 ratio with only around 6-8db of gain reduction. I think I've also tried pretty much all variations of attack and release settings on the units I mentioned. Changing attack and/or release gives a slightly different result but still unattractive nonetheless on speech.

Are the compressors I have essentially no use for compressing mixed program output including speech? There comes a point after trying to further moderate every setting that the compressor may as well not be in circuit.

Do you have a compressor that can moderately compress mixed output including speech and avoid the problem I find? If so, which specific unit and settings do you find to be good?
I have indeed been looking at units by Aphex and Symetrix, for example. Maybe the higher price tag, and more of a leveller/AGC/optimod than a compressor, are necessary for what I'm wanting ....
?

Thanks again,
Lee
 
Lee, honestly at this point I'd be looking hard at software compressors. UAD makes some great stuff and so does Voxengo. Over all though, I think you may be asking too much...every single compressor ever made is going to give you some artifacts if you're hitting it for 8dB with a fast attack, even at 2:1. Most of the really smooth vocals you're hearing, VO or otherwise, are made that way AFTER they're recorded and a lot of that is manual gain control (riding faders). That's why it sounds so natural...that, combined with professional VO talent.

If you're really interested in the higher end stuff, then take a look at the Distressor or for about twice as much the Cransong Trakker.

Frank
 
Hello Frank, and thanks.

I assume levellers/AGCs aimed at broadcast operations are more sophisticated than a sub £500 compressor; presumably more like multiband compression which maybe handles speech better.
Is this true?
If so, I will continue looking for a (used) unit like a Symetrix 422 or an Aphex Compeller 320A.

Since the posts here, I have again tweaked my settings and indeed I can get better results with all controls essentially set at about 12 O'clock. This is on an FMR RNC and a DBX 166XL. I will get out my DBX 1066 later which presumably will react the same. Indeed, the 1066 is meant to be a better unit than the 166XL, isn't it?

Of course, the overall compression now is weaker, which I am trying to assist by running the path into the RNC then into the DBX, ie double compression. Both units set at their nice/over-easy modes.

Lee
 
For speech you want more of an RMS (Root Mean Squared) type of compression, as opposed to peak limiting. With RMS, you're increasing the averaging time so that the peaks are averaged in with the rest of the audio reducing overall compression. With peak limiting, the averaging times are much shorter, so the peaks get chopped off more. You can get the same effect if you take an audio track and chop off the first few milliseconds of audio. "Sandwich" becomes "Thandwhich". It's very un-natural sounding. This is what you discovered when you increased attack and release times on your comps. The problem with RMS compression is that you don't get a lot of compression. If I were you, I'd use volume automation as much as possible and compress the rest with slower times.
 
leecovuk said:
I may have over-stated my talk of limiting; whilst that would be an ideal aim, I still find the problem with very moderate compression at a 2:1 ratio with only around 6-8db of gain reduction. I think I've also tried pretty much all variations of attack and release settings on the units I mentioned. Changing attack and/or release gives a slightly different result but still unattractive nonetheless on speech.
Click to expand...
The ratio doesn't have much to do with it, it's the 6-8db of reduction that is the problem. You would need to set the release time fairly slow so that the compressor doesn't recover and then have to go back down 8db inside of 10 ms.

leecovuk said:
Are the compressors I have essentially no use for compressing mixed program output including speech? There comes a point after trying to further moderate every setting that the compressor may as well not be in circuit.
Click to expand...
Does the speach really fluctuate 16db? (8db of reduction at 2:1) You might get better results if you weren't into the reduction so hard. Try raising the threshold until you are only getting 2-4db of reduction.
 
I would be looking at

(a) a pop shield opn your mic, in case plosives at start of phrases are tricking the compressor into severely ducking. If you can't afford one, a stocking stretched over a wire coathanger between the mic and the mouth will work.

(b) fast attack, maximum release time on compressor

(c) a limiter and a compressor in tandem.

(d) a correct gain structure across all devices.

(e) downloading audacity software (free) and using its compressor and auto-maximize features on your finished track.

Just some ideas - all free.
 
I'm wondering whether you're approach is wrong... I was trying to figure out why you want to crush the heck out it...

Then I noticed that you want to record spoken word and music... then, the last post mentioned ducking, and got the wheels turning... I think you're solution might be right there... sidechain the mic audio to duck the music. That should give you great clarity on the voice... If you feel the voice still needs compression, use a different compressor with less aggresive settings on a seperate audio path... then mix the two
 
One solution may be to compress the Spoken Word separately from the Music and mix them afterwords without further compression. If your music is prerecorded “commercially” available stuff, it’s already compressed. If the Music is your own recording, you may still be better off compressing it separately from the Spoken Word.

Have you already tried this approach?

Rich Smith
 
Last edited:
Hello again and thanks for those further replies; just catching up here.

Yes, it does sound a little better when I compress the mic separately to the overall mix. However even with very light gain reduction where it is only just lighting up the gain reduction meter, the sound is still what I would call unattractive and 'sharp' as before on some of those peaks/transients. (fine when compressor(s) are in bypass)
On my RNC and DBX, it seems to me that the fasted attack time available is best for _attempting_ to reduce it, which I had previously settled on. Around a 1 second release time also seems reasonable.

I can see from the replies here, that I have been in the habit/aim of over-compressing everything where speech is involved.
Maybe it also comes down to the RMS v peak compression that Guitar Zero mentioned. I don't know much about this so I'll dig out the manuals again and also read up some more on it.

I do still wonder/suspect if the compressors I have are essentially no good for speech. I'm not currently clear though, on whether this specifically comes down to the RMS v peak issue, or the capabilities / reaction times of the units.

On the subject of seeing/analysing the compression, can any of you suggest a free and suitable piece of pc software which can display the properties of a 'live' signal? I can then plug in the compressor output and have a look at it in waveform / envelope form. I've tried a couple of softwares somewhat randomly, but it appears they cannot monitor a 'live' signal or do not show what I thought they would.
I then know roughly where to find a peak in an audio waveform, but I assume the software would need to be capable of quite detailed 'zooming up' for me to properly investigate the peaks.

Thanks again,
Lee
 
It really sounds like you are trying to control the level much more than necessary, or something else is going on. 2db of reduction should really be undetectable even on the crappiest compressor.

Is there any way you can post some of this audio? There just might be something about the speaker or the audio that is causing the problem. There might be a solution to it if we could just hear it. (without compression, if possible)
 
Hello Jay, and thanks.

I might try to capture some audio, although I'm not set up to involve the pc in my audio chain. It might be quite a challenge for me to pursue my idea of monitoring the audio on the pc! I've never done anything like inputting a signal into the pc, the signal being either analogue or digital.
Of course, if I can successfully get the audio into my computer to do that, I would also be able to capture it, preferably without adding any more rubbish to it. :)

I'm quite keen for now on trying to get that 'visual monitoring' going of the signal; can anyone recommend a suitable piece of free software which will let me look at the peaks etc? I assume anything disagreeable enough to be heard on a signal would show up there as some kind of mangled or clipped peak.
I might be goggling under the wrong search terms, but so far I'm not having a lot of success finding anything for free which seems suitable for visualising a live incoming signal.
If there isn't one, I'll have to feed in and record a certain amount of audio then analyse it afterwards ... is something like NCH's Wavepad as good for that as anything else?

As for the speakers, yes, I have wondered if they are up to the job. Uncompressed audio seems ok, but maybe the speakers are of such a (low) quality that they react badly to 'excessively processed' audio, such as that created by compression. If that is indeed possible.
My headphones are pretty old with some crackly wires, and not much of a decent guide of anything, to be honest.
I am unashamedly amateur ... or amateurish ... maybe except for my misplaced interests in compression :)

Lee
 
I don't know of anything that will allow you to see anything in real time. Even if you could, it would go by too fast to be of any use. The analog equivalent to what you want is an oscilloscope.
 
You must log in or register to reply here.
Back
Top