Parallel Compression – what’s really going on?

[ReasonTalk]

By Giles Reaves.

Parallel compression, aka Detroit/Motown Compression, aka New York Compression, is simply splitting the audio into two channels and compressing one of them and not the other. Folks give many reasons for using this technique over putting the compressor in an insert; from the idea that it’s a form of upwards compression, to the concept that it somehow preserves the transients, etc. But before we continue, a quick quiz based on information and claims I’ve heard made about parallel compression over the years.

Read the article here!

[Ahornberg]

This is interesting. Thanks!

[MassiveSoundStudios]

Great article and video Giles! I never thought about parallel compression as a way to affect the ratio. I've always used it as a way achieve upward compression and/or preserve transients (and not to mention creative effects, but that's not what this article is about).

The scenario you tested had both the dry and parallel channels sitting at the same level to achieve the "null". Changing the relationship between the dry and parallel blend, for me at least, allows for upward compression techniques and preserving transients.

Using brickwall limiters and/or maximizers in parallel is also a great way of beefing up a source and can result in far greater control and variation, opposed to just using an insert. Again, this is considering that we can change the relationship between dry and parallel channels.

Of course this is all without using EQ in the parallel path.

[Redster]

I'll freely admit that I don't understand that at all. hahaha.

Are you trying to show that parallel compression is pointless unless you have a compressor that does not have a ratio control?

Is there a musical application to any of this?

//picks up a guitar//

[selig]

Great article and video Giles! I never thought about parallel compression as a way to affect the ratio. I've always used it as a way achieve upward compression and/or preserve transients (and not to mention creative effects, but that's not what this article is about).

The scenario you tested had both the dry and parallel channels sitting at the same level to achieve the "null". Changing the relationship between the dry and parallel blend, for me at least, allows for upward compression techniques and preserving transients.

Using brickwall limiters and/or maximizers in parallel is also a great way of beefing up a source and can result in far greater control and variation, opposed to just using an insert. Again, this is considering that we can change the relationship between dry and parallel channels.

Of course this is all without using EQ in the parallel path.

You can't get upwards compression from a downwards compressor IMO, the two work (and sound) totally different. With downwards compression and makeup gain, you are still yanking down the peaks, while upwards compression leaves the peaks alone and thus will sound completely different in most cases. You can't turn one into the other just by turning up the level…

As for preserving transients, same thing - you preserve the transients by not compressing them or by using upwards compression and leaving the transients alone.

Using brick wall limiters etc in parallel just reduces the ratio, as I've shown in my examples. If your compressor has a variable ratio, you get the same 'variation' and control as parallel compression.

Of course, when you add ANY other process, there are huge advantages IMO to parallel compression. I typically use saturation along with limiting on parallel drum bus channels. Sometimes EQ too, but often it sounds "phasey" to me if I do too much parallel EQ. There are better ways to do parallel EQ IMO (more on that later…).

[MassiveSoundStudios]

Great article and video Giles! I never thought about parallel compression as a way to affect the ratio. I've always used it as a way achieve upward compression and/or preserve transients (and not to mention creative effects, but that's not what this article is about).

The scenario you tested had both the dry and parallel channels sitting at the same level to achieve the "null". Changing the relationship between the dry and parallel blend, for me at least, allows for upward compression techniques and preserving transients.

Using brickwall limiters and/or maximizers in parallel is also a great way of beefing up a source and can result in far greater control and variation, opposed to just using an insert. Again, this is considering that we can change the relationship between dry and parallel channels.

Of course this is all without using EQ in the parallel path.

You can't get upwards compression from a downwards compressor IMO, the two work (and sound) totally different. With downwards compression and makeup gain, you are still yanking down the peaks, while upwards compression leaves the peaks alone and thus will sound completely different in most cases. You can't turn one into the other just by turning up the level…

As for preserving transients, same thing - you preserve the transients by not compressing them or by using upwards compression and leaving the transients alone.

Using brick wall limiters etc in parallel just reduces the ratio, as I've shown in my examples. If your compressor has a variable ratio, you get the same 'variation' and control as parallel compression.

Of course, when you add ANY other process, there are huge advantages IMO to parallel compression. I typically use saturation along with limiting on parallel drum bus channels. Sometimes EQ too, but often it sounds "phasey" to me if I do too much parallel EQ. There are better ways to do parallel EQ IMO (more on that later…).

It will be interesting to read your thoughts about parallel EQ!

Regarding preserving transients and upward compression using parallel processing. By "crushing" your parallel channel and mixing it in lower than the source channel, you're effectively bringing up all the sustain and tails in your audio. By keeping the parallel channel below the source "peaks", you're also preserving transients. So I wasn't saying you can turn a downward compressor into an upward compressor. I was saying using a downward compressor on a parallel channel and blending it in can achieve a similar sound to upward compression. Am I wrong?

This has been my impression for quite sometime. Leaving the parallel channel at the same level as the source is a different story, as outlined in your example. It's fascinating that this can demonstrate how compression ratios work.

Are you saying that (with just a compressor alone), parallel processing doesn't make sense, because the results can be achieved with just a compressor in the insert path? Or, are you saying that it cannot achieve upward compression?

Your leveler is a good example of upward compression, no? I do remember you saying something about how it actually works. Where everything under the target threshold is left alone, and everything above is brought up by the curve amount. Is that it, by definition upward compression has to leave a lower level target untouched?

[Audiotic]

interesting!
Thanks for the effort!

[EdGrip]

This is something I have wondered about quite a bit. This makes sense of my ponderings.

[Marco Raaphorst]

Excellent Giles! Never new the absolute relatioship between ratio and amount of parallel channels. Makes sense now! Thanks!

[selig]

You can't get upwards compression from a downwards compressor IMO, the two work (and sound) totally different. With downwards compression and makeup gain, you are still yanking down the peaks, while upwards compression leaves the peaks alone and thus will sound completely different in most cases. You can't turn one into the other just by turning up the level…

As for preserving transients, same thing - you preserve the transients by not compressing them or by using upwards compression and leaving the transients alone.

Using brick wall limiters etc in parallel just reduces the ratio, as I've shown in my examples. If your compressor has a variable ratio, you get the same 'variation' and control as parallel compression.

Of course, when you add ANY other process, there are huge advantages IMO to parallel compression. I typically use saturation along with limiting on parallel drum bus channels. Sometimes EQ too, but often it sounds "phasey" to me if I do too much parallel EQ. There are better ways to do parallel EQ IMO (more on that later…).

It will be interesting to read your thoughts about parallel EQ!

Regarding preserving transients and upward compression using parallel processing. By "crushing" your parallel channel and mixing it in lower than the source channel, you're effectively bringing up all the sustain and tails in your audio. By keeping the parallel channel below the source "peaks", you're also preserving transients. So I wasn't saying you can turn a downward compressor into an upward compressor. I was saying using a downward compressor on a parallel channel and blending it in can achieve a similar sound to upward compression. Am I wrong?

This has been my impression for quite sometime. Leaving the parallel channel at the same level as the source is a different story, as outlined in your example. It's fascinating that this can demonstrate how compression ratios work.

Are you saying that (with just a compressor alone), parallel processing doesn't make sense, because the results can be achieved with just a compressor in the insert path? Or, are you saying that it cannot achieve upward compression?

Your leveler is a good example of upward compression, no? I do remember you saying something about how it actually works. Where everything under the target threshold is left alone, and everything above is brought up by the curve amount. Is that it, by definition upward compression has to leave a lower level target untouched?

Good point about lowering the level of the compressed track, which does indeed create a slight variation on the compression curve - further explorations are required to see exactly what that's doing to the sound.

Remember that even if you keep the compressed track lower than the peaks of the uncompressed track you are STILL adding to the transients, just like if you lower the level of two different sounds in a mix you STILL hear the softer sound. So I wouldn't say you're preserving the transients, you're simply affecting them slightly less than if you left the levels the same. If you want to totally preserve your transients, use upwards compression with the threshold set well below the highest peak.

By definition, upwards compression is simply raising the gain of signals BELOW the threshold - the Selig Leveler is the only device I'm aware of that gives you the ability to leave the lowest signals untouched. Other upwards compressors I've seen DO limit the amount of gain increase, but it's applied to ALL signals below the threshold (including the noise floor).

[Marco Raaphorst]

It will be interesting to read your thoughts about parallel EQ!

Regarding preserving transients and upward compression using parallel processing. By "crushing" your parallel channel and mixing it in lower than the source channel, you're effectively bringing up all the sustain and tails in your audio. By keeping the parallel channel below the source "peaks", you're also preserving transients. So I wasn't saying you can turn a downward compressor into an upward compressor. I was saying using a downward compressor on a parallel channel and blending it in can achieve a similar sound to upward compression. Am I wrong?

This has been my impression for quite sometime. Leaving the parallel channel at the same level as the source is a different story, as outlined in your example. It's fascinating that this can demonstrate how compression ratios work.

Are you saying that (with just a compressor alone), parallel processing doesn't make sense, because the results can be achieved with just a compressor in the insert path? Or, are you saying that it cannot achieve upward compression?

Your leveler is a good example of upward compression, no? I do remember you saying something about how it actually works. Where everything under the target threshold is left alone, and everything above is brought up by the curve amount. Is that it, by definition upward compression has to leave a lower level target untouched?

Good point about lowering the level of the compressed track, which does indeed create a slight variation on the compression curve - further explorations are required to see exactly what that's doing to the sound.

Remember that even if you keep the compressed track lower than the peaks of the uncompressed track you are STILL adding to the transients, just like if you lower the level of two different sounds in a mix you STILL hear the softer sound. So I wouldn't say you're preserving the transients, you're simply affecting them slightly less than if you left the levels the same. If you want to totally preserve your transients, use upwards compression with the threshold set well below the highest peak.

By definition, upwards compression is simply raising the gain of signals BELOW the threshold - the Selig Leveler is the only device I'm aware of that gives you the ability to leave the lowest signals untouched. Other upwards compressors I've seen DO limit the amount of gain increase, but it's applied to ALL signals below the threshold (including the noise floor).

A compression parallel channel when set at a lower level than the original channel will increase the lower dynamics. This sounds like an upwards compressor too me. Or am I mistaken?

[Ostermilk]

So, it's possible to achieve a null when parallel processing vs a compressor on an insert? That isn't the point of parallel processing though as you are going to be altering the relationships between the processed and unprocessed result. Isn't that's a fundamental aspect of using parallel compression?

The terms 'transient preservation' or 'upward compression' are not exactly what is happening when using parallel compression either although for all practical purposes those terms do at least reflect something of what is happening even if they are not accurate enough to pass a pedantometer..

Basically the compresser will reduce the dynamic range of the unprocessed signal, often by using downward compression, so now you can compensate the gain reduction even up to the point where peak levels match on each signal path, so the lower levels of the processed audio are now present in greater proportions than they are in the unprocessed track. A kind of upward compression and something of the original transient peaks are still maintained although they are now likely somewhat flattened, by how much will be dependent on the gain relationship when blended with the compressed track.

So yes, A and B are untrue but I wouldn't be using those terms to describe what happens in a parallel compression setup anyway, and C may well be true but I wouldn't be using a parallel setup to try and achieve a null under any practical circumstance.

This technique became very popular before bedroom producers started to copy it so I'd say there's a bit more going on than ratio control, especially given the non-linear behaviour of the vintage compressors being used, and although I'm not a regular user of straight parallel compression (as opposed to more general parallel processing), certainly there are times where I have used it to end up with a result I couldn't at the time arrive at by any other means.

Besides which a simple parallel wet/dry control can be the quickest way to exactly dial-in the required amount of goodness even if all you were adjusting was ratio.

[Ostermilk]

Here's a good article on how using downward compression in a parallel setup willl create certain aspects of 'upward compression' AND keep original transients intact.

Ignore the information about using multiple parallel compressors though as you will be achieving the same effect by just using more makeup gain on a single compressor.

http://www.soundonsound.com/techniques/ ... ompression

[Marco Raaphorst]

I think we should do a listening blind test. Simply create different files with and without parallel compression.

[Ostermilk]

I think we should do a listening blind test. Simply create different files with and without parallel compression.

For what purpose?

I don't doubt the exact 'test' setup Giles used actually nulls. I don't doubt that the specific ratio is all that's changed in this very specific single example either.

I just don't think the test supports the premise that all you are doing with parallel compression is adjusting the compression ratio in anything other than the specific circumstance shown.

I also have never thought of what happens when using parallel compression can be called 'upward compression' although some of the characteristics of 'upward compression' can clearly be seen in one of the graphics in the SOS article.

It's also my experience in the use of parallel compression that results can be obtained that cannot be replicated by simply using the same compressor as an insert.

So either the article is a brain fart the OP wanted everyone to enjoy the fragrance of, or it is fruitless diversion for the sake of conversation. The OP can take the time to setup various parallel setups and level relationships homself to see if his test holds water through a wider range of parallel compression scenario's but I'm not convinced he'll support the original premise, over and above what has already been proved in a very specific instance.

Unless, of course, I'm missing something.

[Ostermilk]

This is exactly what I meant about 'Upward Compression' being a misnomer in relation to parallel compression.

From the SOS article:

Given that description, you might logically assume that parallel compression somehow achieves a form of upwards compression — and a great many people do, indeed, often refer to it in that way (myself included on occasions). It's all a bit more complicated than that, though, because it really depends on how the parallel compressor is set up. In strict technical terms, a parallel compressor setup is actually a form of downward compressor, but it exhibits a unique and very useful characteristic that allows it to behave exactly like an upwards compressor over a defined dynamic range.

[Marco Raaphorst]

Listening tests are always best imo to prove the effect it has on sound. Some things might look significant "on paper" but are not in pratice.

[selig]

Listening tests are always best imo to prove the effect it has on sound. Some things might look significant "on paper" but are not in pratice.

I'm up for a blind listening test!


Sent from some crappy device using Tapatalk

[Marco Raaphorst]

Listening tests are always best imo to prove the effect it has on sound. Some things might look significant "on paper" but are not in pratice.

I'm up for a blind listening test!


Sent from some crappy device using Tapatalk

Might be able to do one later this week.

[Ostermilk]

The thing I like about the SOS article is that it clearly illustrates 'what's REALLY going on' already.

Also someone there challenged the authors point of using multiple compressors, not that it was wrong, but that it was unnecessary as you could achieve the same result with a single compressor instead. The author then acknowledges the guy that challenged the original assertion about using multiple compressors was indeed correct and the article was then amended to reflect that.

Here though, we have a guy saying that only C is true in the contrived questionairre, when C is the flakiest assertion above A and B as it's only true in a singular situation and even then it would be totally impractical as a way of adjusting compression ratio. This guy also invites comments on where the test might be flawed but then doesn't acknowldge even after it's been pointed out that in a real world parallel compression scenario the gain relationship between the processed and unprocessed signal will most likely be varied to taste, at which point A and B become true (save for nomenclature) and C is completely untrue..

Simply put, by using parallel compression it is indeed possible (and usually desired) to arrive at a result that cannot be obtained by using the same compressor as an insert. Indeed likely you will also end up with a different result than you would using an actual upward compressor.

Given that fact a blind test would be totally pointless in relation to this particular discussion, as there's nothing to discuss if you look at the transfer plots and read the SOS article which is much more comprehensive and closer to describing "what's really going on' than this one, although it (a blind test) may be useful in testing a preference for using parallel compression in comparison to a bona fide upward compressor.

As it stands this article is misleading at best.

[Ahornberg]

I also do parallel compression on a send where I route from multiple tracks to a single compressor. Because of the difference between the send levels of the tracks and their dry levels this method helps blending tracks together in a very subtle way. And because I leave the drums aside it doesn't affect the transients.

[Marco Raaphorst]

While creating a Dry/Mix combinator to compare both settings I discovered a Spider bug.

Download this file:

Spider Bug.reason.zip

(92.91 KiB) Downloaded 1141 times

open the Reason song
press play
listen to the sound
now connect Spider Output 2 Left to Line Mixer 1 Channel 2 Left input
connect Spider Output 2 Right 1 to Line Mixer 1 Channel 2 Right input

Listen to how only the left channel is effected! flip the rack and see that Line Mixer 1 Channel 2 is even muted!

I reported the bug to the Props. I am afraid it will not be fixed because they are afraid it will affect people using such a config in old songs...

[Marco Raaphorst]

Here's a simple Combinator which will show you that the Pulveriser will sound differently using the Dry/Wet compared to a Parallel configuration:

Parallel versus Mix - Pulveriser.reason.zip

(96.68 KiB) Downloaded 354 times

This shouldn't happen. Something strange (no funky) is going on.

That's why I love listening tests Just listen and your ears will guide you.

[Marco Raaphorst]

Might be caused by audio Spider doing some weird stuff by the way...

[Ostermilk]

That's why I love listening tests Just listen and your ears will guide you.

Listening tests are great for checking what preference you or others may have between a number of choices.

Unfortunately a listening test will not tell you 'what's really going on' in the context of the subject of parallel compression, other than you might prefer one particular setting over another.

Certainly though I'd be using my ears to confirm to myself whether parallel compression or a compressor on an insert or even some other form of dynamic control will best suit the task at hand when clearly parallel compression will give me more flexibility than a ratio adjustment on an in-line compressor..

WRT the spider bug that is certainly a long-standing issue that the Props have been made aware of several times. We've got plenty of 3rd party RE spiders now though that don't suffer from the same issue.