Hello!

What are the the benefits of each one of these processes and what are their differences? When should I chose one over the others? Which one sounds more "natural"?

Ho ho homework, merry xmas.

www.samplecraze.com/tutorial.php/60/what-is-parallel-compression-also-know n-as-new-york-compression
www.soundonsound.com/sos/aug11/articles/qanda-0811-4.htm
www.producingbeats.com/How-To-Use-Parallel-Compression-AKA-New-York-Compre ssion
http://en.wikipedia.org/wiki/Parallel_compression
www.sweetwater.com/forums/showthread.php?27466-upward-expansion


Martin

--------------------
YewTreeMagic

That's some real Christmas spirit you've got there, Martin

--------------------
www.thediplomatz.com

Thank you Martin for the links! But actually I know theoretically what each one of them is and how to achieve it. I am not asking this as part of some homework! (I have finished my university music studies 11 years ago )

As I said, theoretically I know what each one of them is, but practically I don't know where I could use them.

For example, I know that if I wanted to make the transients of an audio louder then I would use a normal compressor with a slow attack and moderate release.
If I wanted to tame some loud peaks in an audio, I would use a normal compressor with the threshold set high (in order to process only the peaks), a high ratio (limiting) and fast attack.
If I wanted to to gently compress an acoustic rhythm guitar to make it more even, I would use a low ratio, low threshold (in order to process most of the sound) and a moderate attack and release.

But what about New York compression (Parallel Compression)? When to use it? Whenever I try it, it seems to emphasize the room sound of the recordings. Is it useful in any other instrument except drums and voice?

And is there any difference between Parallel Compression and Upward Expansion? I know it is a different procedure, but is the end result different too? Or they are simply two different means to achieve the same thing?

Thanks!

Quote ElGreco:

---

Thank you Martin for the links! But actually I know theoretically what each one of them is and how to achieve it. I am not asking this as part of some homework! (I have finished my university music studies 11 years ago )

As I said, theoretically I know what each one of them is, but practically I don't know where I could use them.

For example, I know that if I wanted to make the transients of an audio louder then I would use a normal compressor with a slow attack and moderate release.
If I wanted to tame some loud peaks in an audio, I would use a normal compressor with the threshold set high (in order to process only the peaks), a high ratio (limiting) and fast attack.
If I wanted to to gently compress an acoustic rhythm guitar to make it more even, I would use a low ratio, low threshold (in order to process most of the sound) and a moderate attack and release.

But what about New York compression (Parallel Compression)? When to use it? Whenever I try it, it seems to emphasize the room sound of the recordings. Is it useful in any other instrument except drums and voice?

And is there any difference between Parallel Compression and Upward Expansion? I know it is a different procedure, but is the end result different too? Or they are simply two different means to achieve the same thing?

Thanks!

---

I will leave to the masters on the board the "when/why" to use Parallel compression, and await their replies myself!!

As far as the naming: http://www.gearslutz.com/board/6732197-post20.html . The graph on the bottom is upward compression, which is what is obtained with parallel compression.

--------------------
Alexis -Cubase 6.5.0/SX3.1.1.944, XP SP2, 4GB RAM (1GB not accessible, but used just to balance the computer so it doesn't tip over); Delta 66 in Omni i/O Studio; Motif8; UAD-1

Oh boy... I will go easy on you only because I can tell you are new at this. First of all - you are talking about three entirely different things. They simply cannot be compared. One is dynamic control, one is an effect, and the other is both. There are two types of mixing for any given effect, dynamic, spectrum, or time based - corrective and creative. Corrective mixing would include controlling the dynamics to level a track, such as "normal" compression (although it's anyting but normal). It would also include upward expansion on CERTAIN FREQUENCIES to level the audio. Creative mixing would include upward expansion on certain frequencies to accentuate important parts, which can also fall into the corrective section. Creative would also include parallel compression - purely an effect, but a more transparent effect than the alternatives. This is all I will tell you - I'm hoping you will fill in the blanks, because as someone of a a former education career path, I believe that auto dedactic learning is the most powerful, so I want you to fill in the blanks yourself, but I will give you a rough outline;) much of the fill in the blanks is acheived through actual practice. Just go mix! try each method! figure it out!

Quote ElGreco:

---

I know that if I wanted to make the transients of an audio louder then I would use a normal compressor with a slow attack and moderate release.

---

With my tinsel-adorned Pedant's hat in place, this won't make the transients louder! What it actually does is allow the transients through unharmed, but pulls everything else down! A subtle, but critically important distinction when you're delving into the terrority of parallel compression and upwards expansion!

Quote:

---

But what about New York compression (Parallel Compression)? When to use it? Whenever I try it, it seems to emphasize the room sound of the recordings. Is it useful in any other instrument except drums and voice?

---

Parallel compression is a technique which lifts low level sounds while having relatively little effect on high level sounds -- hence you observation that it brings up the room tone -- and you use it when that is the effect you're trying to achieve. It is popular when thickening drums as you mention. I'm not a fan of using it on voice, but perhaps others are. I've also used it on jazz and classical recordings with very wide dynamic ranges that needed reducing in a relatively subtle way.

Quote:

---

And is there any difference between Parallel Compression and Upward Expansion?

---

Yes. Parallel compression only lifts low level information, whereas upwards expansions affects everything above the threshold. Also, there is a finite limit to how much lift can be achieved with parallel compression whereas upward expansion is theoretically limitless.

Parallel compression relies for the low level lift on the coherent addition of signals from the two paths (direct and compressed). At low levels, when the compressor isn't doing anything, the two paths therefore carry identical signals and the maximum addition is 6dB -- low level sounds are increased by 6dB. At high levels the compressor is applying a lot of gain reduction and thus its contribution to the sum is relatively small and insignificant compared to the direct signal. hence there is a negligible increase in level for high level sounds.

In contrast, and upwards expander works in the reverse way to a compressor. The gain is incrased by a pre-defined ratio dependent on the amplitude of the input signal above a preset threshold level. Anything below that level is unaffected, anything above is boosted -- and potentially without any upper limit to hope much it is boosted.

So upwards expansion and parallel compression are entirely different things.

Hope that helps

Hugh

--------------------
Technical Editor, Sound On Sound

Thank you all for your replies and your help!

First of all, after Hugh's detailed explanation and alexis' link I realized that I was saying erroneously "upward expansion" while I meant "upward compression"! Sorry for that!

Quote Hugh Robjohns:

---

At low levels, when the compressor isn't doing anything, the two paths therefore carry identical signals and the maximum addition is 6dB -- low level sounds are increased by 6dB. At high levels the compressor is applying a lot of gain reduction and thus its contribution to the sum is relatively small and insignificant compared to the direct signal. hence there is a negligible increase in level for high level sounds.

---

Does this apply only when the mix of compressed and uncompressed signal is at 50-50%? Or no matter what is the mix ratio of the two, the maximum increase of low level signals through parallel compression is 6dB? And I imagine, that upward compression (De-expander) would be able to achieve more than that. Correct?

Quote Hugh Robjohns:

---

Quote ElGreco:

---

I know that if I wanted to make the transients of an audio louder then I would use a normal compressor with a slow attack and moderate release.

---

With my tinsel-adorned Pedant's hat in place, this won't make the transients louder! What it actually does is allow the transients through unharmed, but pulls everything else down! A subtle, but critically important distinction when you're delving into the terrority of parallel compression and upwards expansion!

---

Well, if I apply some makeup gain in order to make the compressed signal to be as loud as the uncompressed one, wouldn't that make the transients louder? Is there any other way to achieve louder transients using a compressor? (not SPL Transient Designer and the like, but simple compressors)

Quote ElGreco:

---

First of all, after Hugh's detailed explanation and alexis' link I realized that I was saying erroneously "upward expansion" while I meant "upward compression"! Sorry for that!

---

No problem, but still very different things.

'Upward compression' means compression, followed by a gain stage (ie. make-up gain). The compressor pulls the loud stuff down but leaves the quiet bits alone. The gain stage pushes the overall level up so that the quiet bits are made louder and the louder bits end up more or less where they were... so the low levels are raised.

However, the dynamics of the loud bits are still being mangled by the attack and release time constants of the compressor -- and that doesn't happen in the parallel compression technique because the loud bits aren't being processed at all (the very heavily compressed path signal is being completely drowned out by the clean direct path signal).

Consequently, upwards compression and parallel compression, while having a similar effect on the low level content (clean gain only), sound completely different on the high level content (compressed versus uncompressed) -- and that's why the two techniques and are used for different things.

Quote:

---

Does this apply only when the mix of compressed and uncompressed signal is at 50-50%?

---

Technically, yes, but that is the optimum ratio. More direct signal weakens the low level lift effect, and more compressed signal sounds horridly over-compressed!

Quote:

---

I apply some make-up gain in order to make the compressed signal to be as loud as the uncompressed one, wouldn't that make the transients louder?

---

Yes, of course -- but that's just the effect of adding gain to the processed signal. The action of the compressor is what is important when comparing compression techniques. An upwards expander can be arranged to make transients louder without affecting lower level signals and that would sound very different to a compressor followed by some make-up gain.

Quote:

---

Is there any other way to achieve louder transients using a compressor?

---

Not in any practical sense, no.

hugh

--------------------
Technical Editor, Sound On Sound

Quote Hugh Robjohns:

---

I'm not a fan of using it on voice, but perhaps others are.

---

That would be me then! On the other hand I'm not a fan of it on drums (I don't really like 'roomy' drums). It's a nice way of letting low-level vocal detail and hushed notes get through without squashing an otherwise good performance.

--------------------
An Eagle for an Emperor, A Kestrel for a Knave.

Quote The Elf:

---

It's a nice way of letting low-level vocal detail and hushed notes get through without squashing an otherwise good performance.

---

Fair enough! I would achieve the same either by riding the fader during tracking, or with automation in the mix. But your application is certainly a very valid argument for the use of parallel compression.

Hugh

--------------------
Technical Editor, Sound On Sound

Quote Hugh Robjohns:

---

Quote The Elf:

---

It's a nice way of letting low-level vocal detail and hushed notes get through without squashing an otherwise good performance.

---

Fair enough! I would achieve the same either by riding the fader during tracking, or with automation in the mix. But your application is certainly a very valid argument for the use of parallel compression.

Hugh

---

The Elf and Hugh - though the peaks are not processed in the "dry" arm of parallel compression, to my understanding in the end it brings the low level signals up louder relative to the upper level ones. But one can say that about "non-parallel" compression as well ...

And I know that riding the faders or using automation, as Hugh mentions is an alternative to parallel compression, is also an alternative to "non-parallel" compression ...

In the end,I'm having a hard time seeing these as anything but two methods to achieve the same goal . But surely that's not right ...? I wonder - can one examine a compressed signal and determine which method of compression was used to generate it?

Thanks -

--------------------
Alexis -Cubase 6.5.0/SX3.1.1.944, XP SP2, 4GB RAM (1GB not accessible, but used just to balance the computer so it doesn't tip over); Delta 66 in Omni i/O Studio; Motif8; UAD-1

Every form of signal processing has (unwanted) side effects associated with the wanted effect. These side effects are different for parallel compression and normal compression with make up gain. Both will lift lower level sounds, but the side effect of the dynamic gain control element (the attack and release modulation or reshaping of the audio) affects different elements and with different levels of audibility. The normal compression approach modifies the louder elements, while the parallel compression technique affects the quieter elements, but the effect is diluted by the presence of the direct path.

Hugh

--------------------
Technical Editor, Sound On Sound

Quote Hugh Robjohns:

---

Every form of signal processing has (unwanted) side effects associated with the wanted effect. These side effects are different for parallel compression and normal compression with make up gain. Both will lift lower level sounds, but the side effect of the dynamic gain control element (the attack and release modulation or reshaping of the audio) affects different elements and with different levels of audibility. The normal compression approach modifies the louder elements, while the parallel compression technique affects the quieter elements, but the effect is diluted by the presence of the direct path.

Hugh

---

Perfect, thank you Hugh! It's the side effects that always make the difference! I will look (listen) harder and better as a result of this thread ...

--------------------
Alexis -Cubase 6.5.0/SX3.1.1.944, XP SP2, 4GB RAM (1GB not accessible, but used just to balance the computer so it doesn't tip over); Delta 66 in Omni i/O Studio; Motif8; UAD-1

My 2 cents.
Uppward compression is tricky and I never really managed to make it sound natural.
Straight compression can be sometime overkill and you ll never be able to add "tons of attack" or "tons of pumping" without actually affecting the whole signal.
Parallel compression is more subtle. You can choose to "just compress" and tuck in under in order to gain volume and thickness, or make a super pumping or super attacky signal and tuck it under the original one. You get the best of both worlds: natural and what you chose to add from the parallel signal
Here is a video where fab uses parallel processing on drums:
Mixing Bass and Drums with Parallel processing

Edited by chavernac (23/12/11 12:38 AM)

Quote Hugh Robjohns:

---

Quote ElGreco:

---

First of all, after Hugh's detailed explanation and alexis' link I realized that I was saying erroneously "upward expansion" while I meant "upward compression"! Sorry for that!

---

No problem, but still very different things.

'Upward compression' means compression, followed by a gain stage (ie. make-up gain). The compressor pulls the loud stuff down but leaves the quiet bits alone. The gain stage pushes the overall level up so that the quiet bits are made louder and the louder bits end up more or less where they were... so the low levels are raised.

However, the dynamics of the loud bits are still being mangled by the attack and release time constants of the compressor -- and that doesn't happen in the parallel compression technique because the loud bits aren't being processed at all (the very heavily compressed path signal is being completely drowned out by the clean direct path signal).

Consequently, upwards compression and parallel compression, while having a similar effect on the low level content (clean gain only), sound completely different on the high level content (compressed versus uncompressed) -- and that's why the two techniques and are used for different things.

Quote:

---

Does this apply only when the mix of compressed and uncompressed signal is at 50-50%?

---

Technically, yes, but that is the optimum ratio. More direct signal weakens the low level lift effect, and more compressed signal sounds horridly over-compressed!

Quote:

---

I apply some make-up gain in order to make the compressed signal to be as loud as the uncompressed one, wouldn't that make the transients louder?

---

Yes, of course -- but that's just the effect of adding gain to the processed signal. The action of the compressor is what is important when comparing compression techniques. An upwards expander can be arranged to make transients louder without affecting lower level signals and that would sound very different to a compressor followed by some make-up gain.

Quote:

---

Is there any other way to achieve louder transients using a compressor?

---

Not in any practical sense, no.

hugh

---

Since parallel compression is a mixture of both a dry signal and a downward compressed signal my argument is that arent the loud bits processed other than to say " that doesnt happen in the parallel technique . . . "

Quote 10ndaYii:

---

Since parallel compression is a mixture of both a dry signal and a downward compressed signal my argument is that arent the loud bits processed other than to say " that doesnt happen in the parallel technique . . . "

---

I'm not sure I can unpick your grammar enough to make sense of the question...

However, the parallel compressed path is configured to compress loud sounds by ~20dB, and thus the loud sounds are going to be about 20dB quieter than the same signal via the direct path. This makes the compressed path essentially inaudible.

Quiet sounds won't be compressed by much, if at all, and thus the direct and compressed paths will have similar signal levels, and their combined signal will be larger than either on its own. Hence upward compression where quiet signals are made louder!

Hugh

--------------------
Technical Editor, Sound On Sound

Quote:

---

However, the dynamics of the loud bits
are still being mangled by the attack
and release time constants of the
compressor -- and that doesn't
happen in the parallel compression
technique because the loud bits aren't being processed at all (the very
heavily compressed path signal is
being completely drowned out by the
clean direct path signal).
Consequently, upwards compression
and parallel compression, while having a similar effect on the low level
content (clean gain only), sound
completely different on the high level
content (compressed versus
uncompressed) -- and that's why the
two techniques and are used for different things.

---

My question is how upward compression differs from parallel compression and to my understanding the difference is that parallel compression alters slightly the transients since its a mixture or a dry and "compressed " signal and upward compression doesn't , In short what am trying to say is that with parallel compression you have a downward compressed replica of the original signal and the original combined and that means that the transients have been messed with so the loud bits have been altered .

However, the dynamics of the loud bits
are still being mangled by the attack
and release time constants of the
compressor -- and that doesn't
happen in the parallel compression
technique because the loud bits aren't being processed at all (the very
heavily compressed path signal is
being completely drowned out by the
clean direct path signal).
Consequently, upwards compression
and parallel compression, while having a similar effect on the low level
content (clean gain only), sound
completely different on the high level
content (compressed versus
uncompressed) -- and that's why the
two techniques and are used for different things.



That's the first poem about compression I've read.

--------------------
'You know it's a bad role when Nic Cage passes on it.'

Quote 10ndaYii:

---

My question is how upward compression differs from parallel compression

---

We have covered that at some length now...

Quote:

---

...to my understanding the difference is that parallel compression alters slightly the transients since its a mixture or a dry and "compressed " signal and upward compression doesn't

---

Nope. It's the complete reverse. The direct path of the parallel compression arrangement which is dominant for louder signals (including transients) preserves those transients more or less intact.

Conventional compression (with make up gain to contrive 'upwards compression'), generally mangles the transients quite badly (depending on the attack setting). Although that is also an important element of the compressed sound, of course!

hugh

--------------------
Technical Editor, Sound On Sound

Quote:

---

'Upward compression' means
compression, followed by a gain
stage (ie. make-up gain). The
compressor pulls the loud stuff down but leaves the quiet bits
alone. The gain stage pushes the
overall level up so that the quiet
bits are made louder and the
louder bits end up more or less
where they were... so the low levels are raised.

---

"" . . . The
compressor pulls the loud stuff down but leaves the quiet bits
alone . . . ""

Isn't that's what a downward compressor does (to pull down loud bits ) instead of saying " upward compression " . . . like you wrote above .
check the post by polygen

Quote 10ndaYii:

---

Isn't that's what a downward compressor does (to pull down loud bits ) instead of saying " upward compression " . . . like you wrote above .

---

A standard compressor reduces the level of signals above the threshold. This is normal downward compression -- making the loudest bits quieter. Quiet signals are unaffected. Overall dynamic range is reduced.

If you follow a standard compressor with a gain make-up stage, then you can restore the original peak levels to their original value, and in the process low level stuff is also brought up in level. This is an upward compressor, where the quieter stuff is raised in level. Signals below the threshold are simply raised in level. Signals above the threshold are compressed -- so the louder elements are squashed. Overall dynamic range is reduced.

If you run a standard compressor in parallel with a direct path, you have another form of upwards compression -- the 'parallel compressor', also know as the New York or London compressor. In this configuration the loudest parts of the signal are left more or less unaffected because the direct path dominates. The quieter parts are bolstered in level thanks to the parallel contribution of the compressor path with the direct path. So the quieter parts are squashed. Overall dynamic range is reduced.

hugh


--------------------
Technical Editor, Sound On Sound

Quote:

---

the general intention of parallell
compression and upward
compression is similar (increasing
mean program level while retaining
more natural dynamics) they are by no
means *the same*. Just think about it: with parallell compression, you simply
mix the compressed signal in with the
unprocessed signal BUT the
compressed signal has altered
transients, which also get mixed in
underneath the unprocessed transients. Upward compression on
the other hand does not do this, it
simply raises the output level when
the input level goes beneath a
threshold leaving the transients
completely untouched

---

Ok just to check my understanding is there anything wrong with the above quote .

Quote 10ndaYii:

---

Ok just to check my understanding is there anything wrong with the ... quote

---

Yes, it is not quite right!

The general intention of parallel compression and upward compression is similar -- to increase the mean program level -- but they achieve it in very different ways and with very different results.

So far so good... but neither retains natural dynamics. How can they when their entire raison d'etre is to reduce the dynamic range?

However, parallel compression tends to sound less obviously processed than upward compression.

Quote:

---

With parallell compression, you simply mix the compressed signal in with the
unprocessed signal BUT the compressed signal has altered transients which also get mixed in
underneath the unprocessed transients.

---

This is true, but in normal practice the (heavily) compressed path is mixed in about 20dB below the direct path. As a result the compression processing artefacts are largely inaudible, and the transients are preserved more or less fully intact via the direct path. Moreover, becuase the compressor will be applying a large and more or less constant amount of gain reduction for loud signals, the attenuation is largely static rather than dynamic, and there will be very few artefacts anyway.

Quote:

---

Upward compression on the other hand does not do this, it simply raises the output level when the input level goes beneath a threshold leaving the transients completely untouched

---

No. The make up gain stage raises the level of all signals, not just those below the threshold. However, since signals above the threshold are being compressed and pulled down, the net level change from input to output for higer signals will be small in comparison to signals below the threshold which will be raised more substantially.

However, whether transients are affected or not is more to do with the compressor attack time. Transients tend to be loud, and thus are likely to be affected by the compression process. A slow attack time will let transients pass largely unmolested. A fast attack time will distort them.

Hugh

--------------------
Technical Editor, Sound On Sound

Quote:

---

No . The make up gain stage raises the
level of all signals , not just those below
the threshold .

---

Arrrgh it's now starting to make sense .

But one more last question . . . If in upward compression the time constants (attack & release ) only affect the louder bits like you said here

Quote:

---

However, the dynamics of the
loud bits are still being mangled
by the attack and release time
constants of the compressor . . .

---

Does it differ with downward compression and does that mean in upward compression the low bits are not even affected by time constants .

Maybe these graphs will help:



In a normal system the output level (right hand side of 'ladder') is the same as the input level (left hand side) -- as shown in the LINEAR plot above.

A normal downward compressor pulls the level down of signals above the threshold. The elvel of signals below the threshold are unchanged. (DOWNWARD COMPRESSOR)

If you follow downward compression with some make up gain you end up with UPWARD COMPRESSION. High level sounds are output at similar levels while low level sounds are raised, with an overall reduction in dynamic range. But note the dynamic congestion towards the louder end.

In a PARALLEL COMPRESSION system you can achieve a similar reduction in dynamic range, but you'll see that the dynamic congestion now tends to be towards the lower end of the audio range.

Hope that helps

Hugh

--------------------
Technical Editor, Sound On Sound

aye , aye , aye , aye , aye methinks am loving it .
Where can I find more of those graphs .

I created them just for you

Hugh

--------------------
Technical Editor, Sound On Sound

Quote Hugh Robjohns:

---

I created them just for you

Hugh

---

Hugh Robjohns for President!!

--------------------
Alexis -Cubase 6.5.0/SX3.1.1.944, XP SP2, 4GB RAM (1GB not accessible, but used just to balance the computer so it doesn't tip over); Delta 66 in Omni i/O Studio; Motif8; UAD-1

Okay thanks .

Quote Hugh Robjohns:

---

Maybe these graphs will help:



In a normal system the output level (right hand side of 'ladder') is the same as the input level (left hand side) -- as shown in the LINEAR plot above.

A normal downward compressor pulls the level down of signals above the threshold. The elvel of signals below the threshold are unchanged. (DOWNWARD COMPRESSOR)

If you follow downward compression with some make up gain you end up with UPWARD COMPRESSION. High level sounds are output at similar levels while low level sounds are raised, with an overall reduction in dynamic range. But note the dynamic congestion towards the louder end.

In a PARALLEL COMPRESSION system you can achieve a similar reduction in dynamic range, but you'll see that the dynamic congestion now tends to be towards the lower end of the audio range.

Hope that helps

Hugh

---

So when you mess around with a compressor and leave the gain make up in peace , that would be downward compression right . . . Which makes wonder whether if one would be configuring upward compression when one touches the gain make up , I mean is it always the case .

Quote 10ndaYii:

---

Which makes wonder whether if one would be configuring upward compression when one touches the gain make up , I mean is it always the case .

---

You might understand this better if you draw the graphs for yourself, playing with different compresion ratios, thresholds and amounts of make-up gain.

A compressor on its own is always providing downward compression. Adding makeup gain always raises low level signals below the threshold by the amount of make-up gain applied. However, quite how the signals above the threshold are affected depends on the amount of make up gain and the ratio...

Hugh

--------------------
Technical Editor, Sound On Sound