Audio Compressor Question.

[NealeD]

Hi guys,

I write in hope that somebody with a far greater knowledge of compressors (both real world and plugin can help me answer this question that has plagued me for years!

So here's the rub.... Lets say I'm mixing a quieter instrument such as an acoustic guitar on a track and i choose a compressor at random to balance the sound.

The little bit i know on how compression works is that if i set a threshold at a point and then set a ratio, when the guitar volume crosses the set threshold it should reduce the volume by the ratio i have specified.

Hopefully that is correct.

My question is this, lets say purely for arguments sake, (and i know this would never happen in real life, but its the only way i can think of to get my whole point of the question across). Suddenly out of nowhere on my guitar track, something loud appears, like a cymbal crash for example and enters my compressor which is set for the guitar.

Even if i were to set the compressor to a very high ratio say 20:1 or higher the compressor struggles or does relatively little to turn down the volume of the loud noise even though the compression ratio is quite large.

Why does my compressor not appear to reduce the louder noise by 20db or more as the ratio would suggest and if that is the case, then why do compressors state that this is what is happening? When to my ears, the gain reduction doesnt seem anywhere close to what the compressor is claiming?

Hopefully i have explained myself well enough for somebody infinitely more experienced than me in this area to provide a detailed answer on where might be going wrong with my thoughts on compression.

Many Thanks in advance

[Hugh Robjohns]

The little bit i know on how compression works is that if i set a threshold at a point and then set a ratio, when the guitar volume crosses the set threshold it should reduce the volume by the ratio i have specified.

Yes, the part of the input signal above the threshold is reduced by the ratio.

My question is this ... Suddenly out of nowhere on my guitar track, something loud appears, like a cymbal crash for example and enters my compressor which is set for the guitar. ... Even if i were to set the compressor to a very high ratio say 20:1 or higher the compressor struggles or does relatively little to turn down the volume of the loud noise even though the compression ratio is quite large.

In that case, it's either a broken compressor or your settings are nowhere near where they need to be! It shouldn't have any trouble pulling down the level... (bearing in mind its attack time, of course, and the inevitable transient distortion) And a ratio more than 10:1 is generally considered to be limiting, BTW.

Why does my compressor not appear to reduce the louder noise by 20db or more as the ratio would suggest and if that is the case, then why do compressors state that this is what is happening? When to my ears, the gain reduction doesnt seem anywhere close to what the compressor is claiming?

I'm not sure I really understand what you're trying to do.

The portion of the cymbal signal above the threshold will be reduced by a ratio of 20:1, provided it hangs around long enough for the compressor to dial in the appropriate amount of attenuation. A brief duration signal may not be attenuated by the full amount simply because the attack time isn't quick enough.

A 20:1 ratio means that for a steady-state signal (like a tone), for every 20dB that the input level is above the threshold, the output level is only allowed to rise by 1dB above the threshold. And most compressors I've used and reviewed manage to do pretty much that.

With dynamic transient signals, rather than steady-state tones, you have to also factor in the time-constants -- the compressor/limiter takes time to respond and recover -- and so the initial transient of the loud sound will slip through to some extent as the processor thinks and then starts to wind down the gain.

Setting a much shorter attack time obviously improves this situation -- you get less 'overshoot' -- but can also increase the amount of audible distortion (often resulting in click-like artefacts).

Alternatively, use a compressor with a 'look-ahead function' so that it can plan ahead and attenuate the level slightly before the initial transient arrives. This can be done in the analogue world (many broadcast limiters use this technique), but its difficult and expensive. However, it's dead easy to do in DAW plugins and many offer this feature.

Where you need to control the dynamics of a source which will be relatively quiet in the mix (your guitar) it is obviously better to compress it in its own track, rather than trying to do it in the mix bus where other, louder, things (your cymbals) will dominate the compressor's activities.

H

[RichardT]

Hugh has given a great explanation! A cymbal crash, for example, does have a loud transient, so it’s quite likely that your attack time needs to be shorter, as Hugh suggests.

[Wonks]

There's a SOS guide to basic compression here: https://www.soundonsound.com/techniques ... -made-easy

Also if you search the techniques section of the magazine web site, there are quite a few more detailed articles about using compressors in specific ways.

There are techniques where you use two compressors in series. One acting as a limiter with fast attack, release and a high compression ratio to catch the really loud transients, and the other as a gentler more musical compression for the main signal levels.

The 'limiter' has a higher threshold than the 'normal' compressor, so only cuts in on the loud peaks. Some people put the limiter before and some put it after the normal compressor. The results will be a bit different, but you may prefer one to the other. And you still have to play about with the threshold levels, compression ratios, attack and release values and the make-up gains to get a good result.

Presets may be a good starting point for a software compressor e.g. 'male rock vocals', but as a minimum the threshold will need adjusting, as will the make-up gain. And the attack and release settings may be close but still need tweaking.

And if you only have the occasional high transient coming through, rather than say constant crash cymbal bleed, it's often better to manually search for those, and reduce their gain in your DAW's wave editor, rather than compromise your compressor settings for the main sound.

[Luke W]

I think the most common cause of a compressor not behaving as expected that I've seen, is the bypass button. Not saying it's likely to be the case here, but I think it's probably caught out more people than would admit to it

[Hugh Robjohns]

Or in more recent types the 'Blend' control being all the way over... Fell for that one myself only recently. Took a few moments of head scratching to track it down!

[NealeD]

Thanks so much for your time to reply guys. Much appreciated. Great support and lots of valuable information here.

Now we have some dialogue on the table, can i pick up on a few points please....

There isn't actually a cymbal crash that is bleeding onto the guitar track. That was just purely an example to allow me to explain the problem i am, experiencing.

Perhaps a better example is the one given by Wonks where a Male Rock vocalist may lay on the power and scream into the mic on certain parts of a track.

Interestingly, he points to searching for high transients and reducing the gain manually rather than compromising the compressor settings.

This sums up perfectly how my plugin compressors are currently working and i cannot figure out why?

So, at present, every time a loud transient appears in my mix for whatever reason, the compressor does very little to counteract it.

Using the Male Rock Vocals as a great example, if a loud transient has appeared from a vocalist scream or shout the compressor struggles to cope and will not reduce the loud part to an acceptable level in the mix and i have to go through the entire recording manually reducing the peaks to balance the level?? Very similar to what Wonks is alluding to.

I thought that was the whole point of using a compressor?

Its job is as far as i can tell, is to balance out high and low peaks of the sound, but basically (and Male Rock Vocals is a classic example of this)

I end up manually doing what i think the compressor should be doing which is manually decreasing the gain on the high transient parts of the performance so that the levels sit correctly in the mix?

[Hugh Robjohns]

So, at present, every time a loud transient appears in my mix for whatever reason, the compressor does very little to counteract it.

Assuming you are actually routing the loud signal through the compressor and it's not bypassed, then either:

1. the threshold is set too high
2. the ratio is set too low
3. the attack and/or recovery/release times are set too slow.

The clue as to what's going on, besides the values on the controls, is what the Gain Reduction meter is doing.

If it's already showing a lot of gain reduction when the singer's peak comes along, it could be that it has no more attenuation to give...

But in a situation like this, you'd be much better off compressing the vocals separately, and then adding that controlled vocal track to the rest of the mix, rather than relying on a bus compressor to deal with everything with extreme ranges.

Perhaps it would help to (a) describe your exact setup and (b) provide a link to an audio example we could hear to help identify the problem.

[Wonks]

Yes, knowing what compressor you're using, along with all the controls settings and also the level of the signal going in to the compressor would be useful. If the input to the compressor is too low, you may not be able to set a threshold low enough for the compression to even kick in.

What you don't want to do is set the compressor so that in an attempt to tame the occasional loud bits, you squash the life out of the rest of the performance.

Sometimes a single compressor isn't going to fix everything. In the days of tape, you'd probably have to use two compressors, one for the really loud bits, and one for the general levels. And in 'the old days', the guy at the desk (once they know the tune) would be ready to pull the fader down a bit on the loud bits and push it up when it got quieter. These days, as long as there weren't too many, most pros would either find the peaks and reduce them manually, or automate the fader levels and again pull the levels down in the loud bits (far more forgiving if you want to adjust the levels a bit later) .

If the vocal performance is worth it, then it's worth taking time some over, rather than trying to get a compressor or limiter to fix it.

But first we need to make sure your compressor is doing what it should do .

[NealeD]

Based on the information you guys have just given. There answer lies there plain and simple.

Basically A single compressor cannot do the job of evening out transients of a moderate temperament and dealing with the odd super high transient.

In order to capture the super high transient i would crush the life out of 99% of the remainder of the performance.

That said, I now know that it isn't my equipment causing the issue, its the task I'm asking it to perform.

So to sum up, A compressor works very well at evening out transients on a recorded instrument or material providing they are roughly the same level to begin with.

Is that a fair statement?

I should not be trying to use a compressor to capture every transient in a very uneven performance.

Thanks again chaps for your amazing support

[RichardT]

It’s quite normal, and even expected, to use volume automation alongside compression on vocals. You are likely to need both to get a good result - and it’s common to use multiple compressors on vocals too. You might need EQ too if the screamed sections are too harsh.

[Jack Ruston]

It's also worth pointing out that the compressor can rebalance the amplitude discrepancy between the guitar and the cymbal crash overall, BUT it can not rebalance the guitar vs the cymbal crash at the moment the cymbal sounds. When the compressor pushes the cymbal down, it is pushing down the entire signal, guitar included. If the cymbal obliterates the guitar when it is loud, it will still obliterate the guitar to exactly the same extent when it's pushed down by the compressor. They will both go down. SECONDLY, if the cymbal is a harsh, upper midrange rich sound, it will possibly sound louder than the guitar at the same amplitude. In other words, if your goal is to make the cymbal go away, and the guitar come back, a compressor might not be the tool to do that. The right tool is probably to find a section elsewhere where the guitar is playing the same part, and cut it in.

J

[shufflebeat]

Two small but possibly significant points:

A piercing or focus-grabbing sound can appear to the ear/mind more "powerful" than a mellow one in such a way that even when it's being compressed appropriately according to it's electrical magnitude it still demands significant attention.

If you're compressing a mix with a single compressor and one element suddenly dominates the compressor will reduce the gain of the entire mix which maintains the balance of the elements so if you're expecting the vocal to blend better into the guitar it won't, as the guitar is also being compressed.

[Edit]

Jack Ruston - proof positive that great minds and magnificent lovers think alike.

[Jack Ruston]

Could be proof that I don't properly read previous comments.

Two small but possibly significant points:

A piercing or focus-grabbing sound can appear to the ear/mind more "powerful" than a mellow one in such a way that even when it's being compressed appropriately according to it's electrical magnitude it still demands significant attention.

If you're compressing a mix with a single compressor and one element suddenly dominates the compressor will reduce the gain of the entire mix which maintains the balance of the elements so if you're expecting the vocal to blend better into the guitar it won't, as the guitar is also being compressed.

[Edit]

Jack Ruston - proof positive that great minds and magnificent lovers think alike.

[shufflebeat]

Could be proof that I don't properly read previous comments.

We made very similar suggestions at the same time but yours made it's way through the pipeline marginally quicker.

Valid points expressed beautifully - twice - bonus!

[Mixedup]

Wise words from many fine folks here...

Also worth mentioning the side-chain (the signal that the compressor reacts to, which may or may not be identical to the one that's being processed).

First, some compressors react to signal peaks, others to an average signal. For very brief high level sounds, this could affect the compressor's action. You can also choose to filter the side-chain signal — eg use a low pass filter, high shelf cut/boost, tilt EQ or a regular boost/cut to make the compressor more or less sensitive to the cymbal. You can even automate that so that it's only more/less sensitive at certain times.

But in this specific situation, I think I'd be editing the guitar part where that cymbal strikes! Maybe taking notes from elsewhere to cover it up. Maybe using a spectral editor to remove the worst excesses of the cymbal. Etc.

[Wonks]

The cymbal was just an example. The actual problem is a rock singer who occasionally shouts/sings/screams very loud, so in this instance it's all the same sound that's being processed. The peak vs RMS setting is a good one to consider here.

[shufflebeat]

Fixing a vocal level individually with automation would be much more transparent and natural sounding than relying on an algorithm, almost like having a singer with decent mic technique. This frees up the compressor to do what it does best, coaxing rather than hammering.

Specific compressors have their own character and role in the process, of course, and it's difficult for anyone to pronounce chapter and verse without hearing the context so all advice here (particularly mine) will be pretty generic and come with caveats.

[Jack Ruston]

Yeah if this is a vocal, automate it.

[Mixedup]

Yeah. Automation. Or editing and clip gains. Or both.

I mean, if you're really lazy, you can tackle this to an extent using multiple compressors. But the results won't be as good...