The No.1 Website for Pro Audio
 All  This Thread  Reviews  Gear Database  Gear for sale     Latest  Trending
compressors stereo image
Old 3rd December 2008
  #1
Gear Nut
 

compressors stereo image

Hello

1-Can anyone explain technically why using a compressor unlinked causes stereo image to fluctuate?

2-Opposed to it, i have read in a highly regarded product manual that using their compressor in link mode may keep the stereo image centered but collapse the stereo spread.

This seems incoherent but im sure it has a point. (which im missing)

that's all.

cheers
Old 3rd December 2008
  #2
Gear Addict
 
Nishmaster's Avatar
 

Hopefully this will help:

1) Using a stereo compressor unlinked is the equivalent of using two separate compressors. Each sidechain and detector circuit is operating independently, so will compress each channel (L&R) differently. As an extreme example, let's hypothesize that your program material has a rather beefy guitar come in during the chorus, but only on the right side. The right side channel will compress greater than the left, shifting the stereo image to your left.

2) When using a stereo compressor in linked mode, both sidechain signals are either summed or averaged together (or combined in other novel ways, depending on the circuit topology) before hitting the detector circuits. This presents each channel with identical control voltages and thus identical compression. Using the same example above, the beefy guitar coming in will compress both sides equally. In some designs this causes the stereo width to collapse slightly, as the Side information receives more compression than the Mid information. The corollary to that, however, is that other designs that sum the sidechain signals differently can actually expand the stereo image a little, as they compress the Mid information more so than the Side.

Is that at all helpful?
Old 3rd December 2008
  #3
Gear Nut
 

Quote:
1) Using a stereo compressor unlinked is the equivalent of using two separate compressors. Each sidechain and detector circuit is operating independently, so will compress each channel (L&R) differently. As an extreme example, let's hypothesize that your program material has a rather beefy guitar come in during the chorus, but only on the right side. The right side channel will compress greater than the left, shifting the stereo image to your left.
Thats it ?
I guess i already knew it and didnt realize then. English is not my native language so i get confused sometimes.

Quote:
2) When using a stereo compressor in linked mode, both sidechain signals are either summed or averaged together (or combined in other novel ways, depending on the circuit topology) before hitting the detector circuits. This presents each channel with identical control voltages and thus identical compression. Using the same example above, the beefy guitar coming in will compress both sides equally. In some designs this causes the stereo width to collapse slightly, as the Side information receives more compression than the Mid information. The corollary to that, however, is that other designs that sum the sidechain signals differently can actually expand the stereo image a little, as they compress the Mid information more so than the Side.
So how do i know which design does the shrinking and which one does the expansion ? try them out? cause where i live there's nowhere to do that....

Yes, it helped a lot
Old 3rd December 2008
  #4
Gear Addict
 
Nishmaster's Avatar
 

Besides listening?

The only way I know is to either see if the manual mentions it or look over the schematic to see how the sidechain summing is performed.

I would also note that this widening/collapsing effect is very subtle at best and I wouldn't use it a criteria to picking a compressor.
Old 4th December 2008
  #5
Lives for gear
 
Silver Sonya's Avatar
 

Quote:
Originally Posted by Nishmaster View Post
I would also note that this widening/collapsing effect is very subtle at best and I wouldn't use it a criteria to picking a compressor.
Agreed. Don't obsess on it. Especially if you can't hear it.

- c
Old 4th December 2008
  #6
kjg
Lives for gear
 

Quote:
Originally Posted by Nishmaster View Post

2) When using a stereo compressor in linked mode, both sidechain signals are either summed or averaged together (or combined in other novel ways, depending on the circuit topology) before hitting the detector circuits. This presents each channel with identical control voltages and thus identical compression. Using the same example above, the beefy guitar coming in will compress both sides equally. In some designs this causes the stereo width to collapse slightly, as the Side information receives more compression than the Mid information. The corollary to that, however, is that other designs that sum the sidechain signals differently can actually expand the stereo image a little, as they compress the Mid information more so than the Side.
Could you elaborate a bit on why this is so?
I always thought that running a compressor 100% linked could not result in a widening or narrowing of the sound stage. If L and R are receiving equal gain reduction, then L+R is getting the exact same gain reduction as L-R, right? So how is it that either mid or side part of the signal is more compressed?

I understand that when L and R are summed in the sidechain path, the side information is lost and the gain reduction is only controlled by the M part of the signal. But how does that affect the width when L and R get the exact same treatment?

I'd like to understand your point - please elaborate a little. Thank you

Klaas-Jan Govaart
Old 4th December 2008
  #7
Gear Addict
 
Nishmaster's Avatar
 

It's a bit difficult to explain, but I will definitely try. I'm not sure of your technical background, so I'll go from the basics. If this is stuff you might already know, then my apologies for the excess.

In a compressor, the input (in a feed forward design) or output (in a feedback design) is fed to a sidechain circuit. There, the signal is rectified and converted to a DC control voltage. This control voltage is sent to the detector circuit, which compares that voltage to the voltage set by the threshold control and sends a new control voltage to the gain reduction stage. Some designs forgo the detector circuit and use a fixed threshold.

Let us create a hypothetical compressor. For a +/-1v sine wave input, let's say the sidechain creates a +1v DC control voltage, and this +1v causes the gain reduction stage to reduce the gain by 6db. 6db is half the signal, and thus our output would be a +/-.5v sine wave.

If you sum the signal before the sidechain, and thus before it is rectified to a DC voltage, Mid signals will get a 6db higher priority in the sidechain. Why is this? Let's feed the same +/-1v signal into both the L and R inputs. If we sum these signals together before the sidechain, we now have a +/-2v summed input signal, 6db higher than each of the sides independently. Our sidechain makes this +2v DC control voltage, and now we're reducing the gain by 12db. This gives Mid signals more "priority" over the gain reduction than Side information.

However, imagine that we feed this compressor the same signal again, except one channel is of the opposite polarity. If we sum those signals before the sidechain, we get no signal at all. They cancel out, thus no control voltage and no gain reduction. However, if we sum AFTER the sidechain, after the voltages are rectified, we arrive at a different answer. Each +/-1v input signal creates a +1v DC control voltage, regardless of polarity. Adding those voltages after they are rectified now gives us +2v, and again 12db of gain reduction. In this case, the Mid channel has zero information, but the side channel is full of info. Hence, the Side channel has more "priority" over the gain reduction.

The net effect of all this is that the perceived stereo width can be altered in either direction. These effects can be mitigated in various ways, however, such as averaging the control voltages. In our example above, two identical polarity +/-1v sine waves will create a +1v control voltage in each sidechain. However, instead of adding them, we do (L+R)/2. (1+2)/2=1, so we get a +1v total control voltage, and 6db of gain reduction. Similarly, if we have two signals of opposite polarity, we still get +1v in each sidechain, and thus the averaging comes out the same. +1v total, 6db gain reduction. In that topology, equal weight is given to both M and S signals.

I know it's a freaking novel, but does that explain things better?
Old 4th December 2008
  #8
Audio Alchemist
 
Lagerfeldt's Avatar


No just kidding, that was very well written and informative, thanks.
Old 5th December 2008
  #9
Lives for gear
 
Edward_Vinatea's Avatar
 

Quote:
Some designs forgo the detector circuit and use a fixed threshold.
This worries me some as I wouldn't want that on my compression process. Do you know the name of the brands that specifically do this? Are we talking both sftware and hardware here?

Thanks in advance for your response.

Word to peeps: these processes are at best 'very subtle' and for many 'impossible' to detect by normal hearing.
Old 5th December 2008
  #10
Gear Nut
 

Quote:
Originally Posted by Lagerfeldt
that was very well written and informative, thanks.
+1


Nishmaster
Old 5th December 2008
  #11
Lives for gear
 
Adam Dempsey's Avatar
 

Verified Member
As one example, the Vari-Mu with M-S mods can be still be run in either linked or un-linked mode. See owner's manual pg 12 via here.
Quote:
Originally Posted by Edward_Vinatea View Post
Quote:
Some designs forgo the detector circuit and use a fixed threshold.
This worries me some as I wouldn't want that on my compression process.
Meaning: threshold is determined by input level. I've used a modified Philips broadcast limiter, I think some AWA's are similar.. with fast attack for HF, slower for LF, 100Hz detector side chain filt and non-linear release. A lot of headroom and control, but grabby from the ratio being variable with input level only. Fortunately here we feel all bases are covered with the flexibilities and differences of the Vertigo, LTD-2's, STC-8 and Vari-Mu.
Old 5th December 2008
  #12
Lives for gear
 
Edward_Vinatea's Avatar
 

Quote:
Originally Posted by Boomshanka View Post
As one example, the Vari-Mu with M-S mods can be still be run in either linked or un-linked mode. See owner's manual pg 12 via here.

Meaning: threshold is determined by input level. I've used a modified Philips broadcast limiter, I think some AWA's are similar.. with fast attack for HF, slower for LF, 100Hz detector side chain at 100Hz and non-linear release. A lot of headroom and control, but grabby from the ratio being variable with input level only. Fortunately here we feel all bases are covered with the flexibilities and differences of the Vertigo, LTD-2's, STC-8 and Vari-Mu.
THANKS Boomshanka!

Quote:
Originally Posted by Nishmaster View Post
In a compressor, the input (in a feed forward design) or output (in a feedback design) is fed to a sidechain circuit. There, the signal is rectified and converted to a DC control voltage. This control voltage is sent to the detector circuit which compares that voltage to the voltage set by the threshold control and sends a new control voltage to the gain reduction stage. Some designs forgo the detector circuit and use a fixed threshold (Meaning: threshold is determined by input level ). The net effect of all this is that the perceived stereo width can be altered in either direction.
A nice and elegant explanation by Nishmaster completed by Boomshanka! That clarifies it. Wow, what a user name!

Best regards to Ostralya.
Old 5th December 2008
  #13
Gear Addict
 
Nishmaster's Avatar
 

Quote:
This worries me some as I wouldn't want that on my compression process. Do you know the name of the brands that specifically do this? Are we talking both sftware and hardware here?
No need to worry. Boomshanka is correct in that the gain reduction is determined by the input level. In fact, a couple of our favorite compressors have a fixed threshold, the 1176 and the Distressor. The LA2A has somewhat of a fixed threshold as well, although the "Peak Reduction" knob function is a bit more complex than that.

I would also note that your condensed version of my summary is a bit misleading, Edward, although not intentionally. The change in perceived stereo width has very little to do with the threshold and detection method. The way the stereo link sums or compares the sidechain signals is the true cause.
Old 5th December 2008
  #14
Lives for gear
 
macc's Avatar
 

Verified Member
Quote:
Originally Posted by Edward_Vinatea View Post
A nice and elegant explanation by Nishmaster completed by Boomshanka! That clarifies it. Wow, what a user name!

'May the seed of your belly be fruitful in the belly of your woman' heh
Old 5th December 2008
  #15
Audio Alchemist
 
Lagerfeldt's Avatar
On the subject, check out the section just below the color picture of the TFPro38

tfpro p38
Old 6th December 2008
  #16
Lives for gear
 

Quote:
Originally Posted by Nishmaster View Post
It's a bit difficult to explain, but I will definitely try. I'm not sure of your technical background, so I'll go from the basics. If this is stuff you might already know, then my apologies for the excess.

In a compressor, the input (in a feed forward design) or output (in a feedback design) is fed to a sidechain circuit. There, the signal is rectified and converted to a DC control voltage. This control voltage is sent to the detector circuit, which compares that voltage to the voltage set by the threshold control and sends a new control voltage to the gain reduction stage. Some designs forgo the detector circuit and use a fixed threshold.

Let us create a hypothetical compressor. For a +/-1v sine wave input, let's say the sidechain creates a +1v DC control voltage, and this +1v causes the gain reduction stage to reduce the gain by 6db. 6db is half the signal, and thus our output would be a +/-.5v sine wave.

If you sum the signal before the sidechain, and thus before it is rectified to a DC voltage, Mid signals will get a 6db higher priority in the sidechain. Why is this? Let's feed the same +/-1v signal into both the L and R inputs. If we sum these signals together before the sidechain, we now have a +/-2v summed input signal, 6db higher than each of the sides independently. Our sidechain makes this +2v DC control voltage, and now we're reducing the gain by 12db. This gives Mid signals more "priority" over the gain reduction than Side information.

However, imagine that we feed this compressor the same signal again, except one channel is of the opposite polarity. If we sum those signals before the sidechain, we get no signal at all. They cancel out, thus no control voltage and no gain reduction. However, if we sum AFTER the sidechain, after the voltages are rectified, we arrive at a different answer. Each +/-1v input signal creates a +1v DC control voltage, regardless of polarity. Adding those voltages after they are rectified now gives us +2v, and again 12db of gain reduction. In this case, the Mid channel has zero information, but the side channel is full of info. Hence, the Side channel has more "priority" over the gain reduction.

The net effect of all this is that the perceived stereo width can be altered in either direction. These effects can be mitigated in various ways, however, such as averaging the control voltages. In our example above, two identical polarity +/-1v sine waves will create a +1v control voltage in each sidechain. However, instead of adding them, we do (L+R)/2. (1+2)/2=1, so we get a +1v total control voltage, and 6db of gain reduction. Similarly, if we have two signals of opposite polarity, we still get +1v in each sidechain, and thus the averaging comes out the same. +1v total, 6db gain reduction. In that topology, equal weight is given to both M and S signals.

I know it's a freaking novel, but does that explain things better?


Whats freaking amazing is the percentage of that paragraph ( which will be and un-named figure !!) That I get after reading X amount of times !!


Here is a explanation from a soft comp manual that might be useful....





1) No link : each channel is independant

2) Average : mixes the left and right envelopes to compute the dynamics response. The same gain is applied to both channels.

3) Min: use the minimum envelope to compute the dynamics response . The same gain is applied to both channels.

4) Max: use the maximum envelope to compute the dynamics response. The same gain is applied to both channels.

5)............. The One that may answer the OP's question about a compressor generating a stereo FX ....................

Flip : apply the computed gain of the opposite channel. This enhances the stereo dynamics of the audio channel.
Old 6th December 2008
  #17
Lives for gear
 
macc's Avatar
 

Verified Member
Quote:
Originally Posted by Lagerfeldt View Post
On the subject, check out the section just below the color picture of the TFPro38

tfpro p38
I was going to mention the p38 in this thread cos of exactly that - are there many other compressors that work that way?

FWIW I never notice any image shifting/shrinking/expansion at all when using it. Well, unless I use the balance or width controls
Old 6th December 2008
  #18
Lives for gear
 
Edward_Vinatea's Avatar
 

Quote:
Originally Posted by Nishmaster View Post
I would also note that your condensed version of my summary is a bit misleading, Edward, although not intentionally. The change in perceived stereo width has very little to do with the threshold and detection method. The way the stereo link sums or compares the sidechain signals is the true cause.
I will put it in more layman's terms and provide an extreme example: when you have 2 mono compressors inserted on a 2 channel master buss (Like I used to when I had my venerable DBX-160Xs in the old days} between your mixer and your 2 trk recorder - setting one at a higher threshold, compression rate and make up than the other one will significantly compromise the stereo field simply because all the dynamics of one channel are more reduced on one side and the imaging shifts to the center {M}. You can't expand the stereo field with a compressor unless you are doing M/S compression where only the mid is processed not the sides.

I know I am old fashion and I would never ever use a compressor to either increase or reduce a mix stereo field {I fix it in the mix itself} but correct me if I am wrong.

Best regards,
Old 6th December 2008
  #19
Lives for gear
 
Adam Dempsey's Avatar
 

Verified Member
Quote:
Originally Posted by macc View Post
'May the seed of your belly be fruitful in the belly of your woman' heh
Almost.
The Young Ones, (UK).
Old 6th December 2008
  #20
Gear Nut
 

Nishmaster

Can you explain how external sidechain relates to internal in the summing of signals in a feed forward design ?

As i understand, the external sidechain interacts with the internal just before the control circuit.

Does it affect the input signal L+R summing in any way? How is it combined with the internal ?
Old 6th December 2008
  #21
Lives for gear
 
macc's Avatar
 

Verified Member
Quote:
Originally Posted by Boomshanka View Post
Almost.
The Young Ones, (UK).

DAGH - typing error!! I don't even need to click the link heh I have the scripts in my head, a defining part of my youth!

And I forgot there was this ---> Large scale fail.

Now I have 'Vivian... this is for a toaster' in my head for some reason - LOL
Old 7th December 2008
  #22
Lives for gear
 
Sunbreak Music's Avatar
 

Verified Member
Quote:
Originally Posted by Edward_Vinatea View Post
I know I am old fashion and I would never ever use a compressor to either increase or reduce a mix stereo field {I fix it in the mix itself} but correct me if I am wrong.

Best regards,

Right or wrong, people do it all the time. heh
Old 7th December 2008
  #23
Audio Alchemist
 
Lagerfeldt's Avatar
I love the way a great analog compressor can enhance the stereo field.

Tools are just tools, never say never.
Old 8th December 2008
  #24
Lives for gear
 
Edward_Vinatea's Avatar
 

Quote:
Originally Posted by Lagerfeldt View Post
I love the way a great analog compressor can enhance the stereo field.

Tools are just tools, never say never.
I am not debating its use - it was just my point of view. If I want to increase the stereo image - I make it happen in the mix or in mastering I may resort to connect a M/S matrix and change the level of S relative to M . Not compress the M channel. That's just me though.

Regards,
Old 8th December 2008
  #25
Lives for gear
 
macc's Avatar
 

Verified Member
Quote:
Originally Posted by Edward_Vinatea View Post
If I want to increase the stereo image - I make it happen in the mix or I may resort to connect a M/S matrix and change the level of S relative to M in mastering. Not compress the M channel. That's just me though.
Depends though doesn't it? At least in some of the mixes I get (drum n bass, mono sampled funk break in the middle) it could come in handy. Could be an occasion where it kills two birds with one stone.

Not that I do it really, but I am sure it could be useful...
Old 8th December 2008
  #26
kjg
Lives for gear
 

Quote:
Originally Posted by Nishmaster View Post
It's a bit difficult to explain, but I will definitely try.

<snip>

I know it's a freaking novel, but does that explain things better?
Thank you very much Nishmaster, for typing all that.

I was actually confused: I thought you were saying that at any moment, at any instant, the width could be affected (as in unlinked MS compression). But from what (I think) I understand now, the image gets affected from moment to moment, leading to a perceived widening (or narrowing) of the stereo field.
When we have the situation that mid signals control GR because of LR summing in the sidechain, for example, this will lead to relatively more compression in one moment where there is a lot of mid content, but to less compression in another moment when there is a lot of side content (in each of these moments the width stays the same though, the whole program is just momentarily louder or quieter). But, overall, from moment to moment, the parts of the program with a lot of mid content will receive more gain reduction while everything will receive the same amount of makeup gain, so the mix will be perceived as wider.

Am I getting this now?

If so, here is another question...
How do issues like these influence the usage of compressors in linked MS mode?

If the compressor in question is normally summing L and R (averaging or not) in linked mode to get the control voltage for the GR elements, that means it would sum M and S when used in a mid/side configuration.
It sums the M (L channel in normal use) and S (R channel in normal use)... M+S = L!
The compressor is now effectively controlled by the Left channel?!

Not exactly what you want when you are patching in your analog box in mid/side mode, thinking it is ok because it is running linked...

That's at least what I would make of it intuitively.. "It runs linked, therefore mid and side will receive the same treatment.. Therefore, no need to decode back to LR.."
Except that the compression is now controlled by the L instead of by the M component... Not quite the same. Bummer!

EDIT: I guess it would be ok if the signals were rectified independently before summing, right? Then that is probably standard design - two rectifiers and then summing - and this is a non-issue?

Or am I confused again/still?

Thanks for any enlightenment!

regards,
Klaas-Jan Govaart
Old 8th December 2008
  #27
kjg
Lives for gear
 

Quote:
Originally Posted by Edward_Vinatea View Post
setting one at a higher threshold, compression rate and make up than the other one will significantly compromise the stereo field simply because all the dynamics of one channel are reduced on one side more and the imaging shifts to the center {M}. You can't expand the stereo field with a compressor unless you are doing M/S compression where only the mid is processed not the sides.
Correct me if I'm wrong but I think this would cause the image to be unstable, drifting between left and right, not "shift to the center".

It seems you can (and will, like it or not) expand the perceived width of a mix by using a compressor, depending on how the sidechain is implemented.

kjg
Old 8th December 2008
  #28
Lives for gear
 
Edward_Vinatea's Avatar
 

Quote:
Originally Posted by kjg View Post
Correct me if I'm wrong but I think this would cause the image to be unstable, drifting between left and right, not "shift to the center".
If it can drift to the other side, doesn't it have to drift center first?

Quote:
It seems you can (and will, like it or not) expand the perceived width of a mix by using a compressor, depending on how the sidechain is implemented.
If you're going to use a compressor - M/S compression is how you expand the width.

So long KJG,
Old 8th December 2008
  #29
kjg
Lives for gear
 

Quote:
Originally Posted by Edward_Vinatea View Post
If it can drift to the other side, doesn't it have to pass though center first?

If you're going to use a compressor - M/S compression is how you expand the width.
mkay...

thanks for explaining Edward
Old 8th December 2008
  #30
Lives for gear
 
Edward_Vinatea's Avatar
 

Quote:
Originally Posted by kjg View Post
mkay...

thanks for explaining Edward
You are welcome. However - what I said about image "shifting" may need further clarification as it can be misconstrued. I'll try posting again tomorrow.

Good evening,
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump