Michael Gounelas

I was wondering if there are any general algorithm techniques in order to achieve an analogue sound result. I know this question is a bit general and has been discussed before, but the only sources i found in this forum are about building 'analogue' synths rather than 'analogue' compressors. 

So I would be grateful if someone could give some advices for doing my compressor patch sound a bit more 'analogue'. I could post my patch but what I'm looking for is some general advices that can be implied in many cases and not only on this specific patch.

I was wondering if there are any general algorithm techniques in order to achieve an analogue sound result. I know this question is a bit general and has been discussed before, but the only sources i found in this forum are about building 'analogue' synths rather than 'analogue' compressors. 
So I would be grateful if someone could give some advices for doing my compressor patch sound a bit more 'analogue'. I could post my patch but what I'm looking for is some general advices that can be implied in many cases and not only on this specific patch.

non-convolution-analogue-warmth

In general, various degrees of low-pass filtering, small amounts of 

randomness ('noise' in audio signals and control data), and (for 

emulating vintage opto-compressor circuits) a bit of delay in the 

side-chain (control) circuit are all going to get you a bit closer to 

an analog compressor than the purely digital starting point. Season 

On Mar 30, 2007, at 9:10 AM, Michael Gounelas wrote:

> I was wondering if there are any general algorithm techniques in 

> order to achieve an analogue sound result. I know this question is 

> a bit general and has been discussed before, but the only sources i 

> found in this forum are about building 'analogue' synths rather 

> So I would be grateful if someone could give some advices for doing 

> my compressor patch sound a bit more 'analogue'. I could post my 

> patch but what I'm looking for is some general advices that can be 

> implied in many cases and not only on this specific patch.

In general, various degrees of low-pass filtering, small amounts of  
randomness ('noise' in audio signals and control data), and (for  
emulating vintage opto-compressor circuits) a bit of delay in the  
side-chain (control) circuit are all going to get you a bit closer to  
an analog compressor than the purely digital starting point. Season  
to taste - your mileage may vary...

> I was wondering if there are any general algorithm techniques in  
> order to achieve an analogue sound result. I know this question is  
> a bit general and has been discussed before, but the only sources i  
> found in this forum are about building 'analogue' synths rather  
> than 'analogue' compressors.
> So I would be grateful if someone could give some advices for doing  
> my compressor patch sound a bit more 'analogue'. I could post my  
> patch but what I'm looking for is some general advices that can be  
> implied in many cases and not only on this specific patch.
>

Could you please be a bit more specific by saying ''various degrees of low-pass filtering''? Where should i apply the filters and what are they supposed to do inside a compressor?

hi Steven and thanks for the reply.
Could you please be a bit more specific by saying ''various degrees of low-pass filtering''? Where should i apply the filters and what are they supposed to do inside a compressor?

The low-pass filtering should be applied to the audio signal on 

input. The specific setting of the Fc would really depend on which 

unit you're trying to emulate, and/or what sort of general "analog" 

sound you're interested in. The slope would generally be a simple 

first-order 6 dB per octave. Doing a comparison on a scope of input- 

vs.-output frequency response (and phase, if you want to go there) 

might be a good 'objective' guide, but your ears are probably fine 

for most work (assuming you're going for the general sound, and not a 

For that matter, some filtering (or 'smoothing') of the side-chain 

(control) signal would also be a good idea.

The idea here is merely to reflect the reduced frequency response of 

a lot analog equipment (particularly vintage gear). It's not 

necessarily part of the original analog designs, but an attempt to 

emulate their characteristics in the digital domain.

On Mar 30, 2007, at 10:05 AM, Michael Gounelas wrote:

> Could you please be a bit more specific by saying ''various degrees 

> of low-pass filtering''? Where should i apply the filters and what 

> are they supposed to do inside a compressor?

The low-pass filtering should be applied to the audio signal on  
input. The specific setting of the Fc would really depend on which  
unit you're trying to emulate, and/or what sort of general "analog"  
sound you're interested in. The slope would generally be a simple  
first-order 6 dB per octave. Doing a comparison on a scope of input- 
vs.-output frequency response (and phase, if you want to go there)  
might be a good 'objective' guide, but your ears are probably fine  
for most work (assuming you're going for the general sound, and not a  
specific emulation).

For that matter, some filtering (or 'smoothing') of the side-chain  
(control) signal would also be a good idea.

The idea here is merely to reflect the reduced frequency response of  
a lot analog equipment (particularly vintage gear). It's not  
necessarily part of the original analog designs, but an attempt to  
emulate their characteristics in the digital domain.

> Could you please be a bit more specific by saying ''various degrees  
> of low-pass filtering''? Where should i apply the filters and what  
> are they supposed to do inside a compressor?
>

yeas, great info. i didn't really know that you can use a lp in compreossors to be honest. I'll give a try when i go back home.

PS. i'm still interesting to hear advices and techniquesabout this :)

PS. i'm still interesting to hear advices and techniquesabout this  :)

> The low-pass filtering should be applied to the audio 

> input. The specific setting of the Fc would really depend 

should i assume that the Fc is below the audible frequency bandwidth or not? If not what would be the higher value for Fc? It sounds a bit confusing to me cause if the Fc is at 80Hz (say) how can you compress a bass guitar without loosing the bottom end? or the answer is to eq it after the compression?

> The low-pass filtering should be applied to the audio 
> signal on 
> input. The specific setting of the Fc would really depend  
> on which 
> unit you're trying to emulate,

sorry about the last post. i confused LP with Hp :) but my question remains the same. 

replace 'bass guitar' with hihats and '80Hz' with 10Khz... does this make sense?

sorry about the last post. i confused LP with Hp :) but my question remains the same. 
replace 'bass guitar' with hihats and '80Hz' with 10Khz... does this make sense?

The Fc of the low-pass filter would be on the upper end, to reduce 

the high-frequency content of the overall signal - say somewhere 

between 12kHz - 18kHz or so, depending on the sound you're going for. 

I'd recommend doing the filtering before compressing, to make for 

smoother compression, all else being equal.

On Mar 30, 2007, at 11:15 AM, Michael Gounelas wrote:

> should i assume that the Fc is below the audible frequency 

> bandwidth or not? If not what would be the higher value for Fc? It 

> sounds a bit confusing to me cause if the Fc is at 80Hz (say) how 

> can you compress a bass guitar without loosing the bottom end? or 

> the answer is to eq it after the compression?

The Fc of the low-pass filter would be on the upper end, to reduce  
the high-frequency content of the overall signal - say somewhere  
between 12kHz - 18kHz or so, depending on the sound you're going for.  
I'd recommend doing the filtering before compressing, to make for  
smoother compression, all else being equal.

>
> should i assume that the Fc is below the audible frequency  
> bandwidth or not? If not what would be the higher value for Fc? It  
> sounds a bit confusing to me cause if the Fc is at 80Hz (say) how  
> can you compress a bass guitar without loosing the bottom end? or  
> the answer is to eq it after the compression?

On Mar 30, 2007, at 11:50 AM, Michael Gounelas wrote:

what would be typical for "analog" sound is that 

mixing 2 signals compresses and saturates them 

low-alias sound sources (if there are very high frequencies 

in a sound) would also help, as well as a good potion of 

what would be typical for "analog" sound is that 
mixing 2 signals compresses and saturates them 
(and not only [*~ ]s them)

low-alias sound sources (if there are very high frequencies 
in a sound) would also help, as well as a good potion of 
randomisation/incorrectness.


> what would be typical for "analog" sound is that 

> mixing 2 signals compresses and saturates them 

are you reffered to convolution method or i'm confused?

> what would be typical for "analog" sound is that 
> mixing 2 signals compresses and saturates them 
> (and not only [*~ ]s them)

get PSP vintage warmer, its totally f!@#n incredible analogue modeling compressor, sounds so amazing, I use it on everything now. VST and AU and all that jazz I think. Heres the link:

I think you can buy it from other places cheaper though.

You should read my thread on analogue/digital synthesis. There is good info in there too. Use cheby lookup tables on the low frequencies!!! it helps soooo much on making the signal "warmer" and also make an abstraction that you can plug all your values into that will add small ammounts of randomness into it, really subtle stuff goes a long way.

get PSP vintage warmer, its totally f!@#n incredible analogue modeling compressor, sounds so amazing, I use it on everything now.  VST and AU and all that jazz I think.  Heres the link: 

You should read my thread on analogue/digital synthesis.  There is good info in there too.  Use cheby lookup tables on the low frequencies!!! it helps soooo much on making the signal "warmer" and also make an abstraction that you can plug all your values into that will add small ammounts of randomness into it, really subtle stuff goes a long way.


I have psp vintage warmer and i can tell is pretty cool. I don't use it a lot tho.

However, my aim is to understand and see how far i can get with such a project rather than building the best ever made plug-in in the industry :)

I've read your threat and it's very useful although i'm not really sure how (and where) can i apply these techniques (such as look up tables, and randomisation). Also, i'm not really sure what the tollerance of randomisation should be, and in which values applied.

I have psp vintage warmer and i can tell is pretty cool. I don't use it a lot tho.
However, my aim is to understand and see how far i can get with such a project rather than  building the best ever made plug-in in the industry :)

Quote: warp wrote on Fri, 30 March 2007 13:18

----------------------------------------------------

> > what would be typical for "analog" sound is that 

> > mixing 2 signals compresses and saturates them 

> are you reffered to convolution method or i'm confused?

no to mixing. many analog "signal paths" in an analog mixer

or synthesizer do _change the signal. for example because a 

signal is boosted or mixed together in another.

often this is "imperfectness" but sometimes thats wanted.

think "tube compressor" (or PSP as a digital example...)

where saturation/distortion is adding by putting tubes

you could do the same to bring some life into your compressor,

because if your compressor only consists of average~, 

rampsmooth~, *~, and line~, it will also sound like that.

why dont you put an extra option into it where the attack time

is not perfectly constant but changes a bit? that changes the feel

or make the attack time a littl ebit depending on how much 

Quote: warp wrote on Fri, 30 March 2007 13:18
----------------------------------------------------
> > what would be typical for "analog" sound is that 
> > mixing 2 signals compresses and saturates them 
> > (and not only [*~ ]s them)
>
> hi Roman,
>
> are you reffered to convolution method or i'm confused?

no to mixing. many analog "signal paths" in an analog mixer
or synthesizer do _change the signal. for example because a 
signal is boosted or mixed together in another.

often this is "imperfectness" but sometimes thats wanted.
think "tube compressor" (or PSP as a digital example...)
where saturation/distortion is adding by putting tubes
in the (analog) signal path.

you could do the same to bring some life into your compressor,
because if your compressor only consists of average~, 
rampsmooth~, *~, and line~, it will also sound like that.

why dont you put an extra option into it where the attack time
is not perfectly constant but changes a bit? that changes the feel
of a compressor "sound" very nicely. 

or make the attack time a littl ebit depending on how much 
bass material there is in the input ... 

thank you all , i'll try to work on all these great ideas yoy gave me. Hopefully, with some good results..

Do you know any good sources around this subject? it doesn't have to be max/msp "based". Some general info would be great.

Go to the following link and download the pdf of Jim Tenney's article 

'Computer Music Experiences 1961-1964' - it has a number of concreate 

examples of adding in periodic and aperiodic modulation for 'livening 

up' the sound of digital synthesis. The basic techniques can be 

applied beyond these specific examples, however.

http://pubweb.csf.edu/~smill/courses_06_Fall/MUS332.html

On Mar 31, 2007, at 6:50 AM, Michael Gounelas wrote:

> I have psp vintage warmer and i can tell is pretty cool. I don't 

> However, my aim is to understand and see how far i can get with 

> such a project rather than building the best ever made plug-in in 

> I've read your threat and it's very useful although i'm not really 

> sure how (and where) can i apply these techniques (such as look up 

> tables, and randomisation). Also, i'm not really sure what the 

> tollerance of randomisation should be, and in which values applied.

Go to the following link and download the pdf of Jim Tenney's article  
'Computer Music Experiences 1961-1964' - it has a number of concreate  
examples of adding in periodic and aperiodic modulation for 'livening  
up' the sound of digital synthesis. The basic techniques can be  
applied beyond these specific examples, however.

>
> hi Axiom-Crux,
>
> I have psp vintage warmer and i can tell is pretty cool. I don't  
> use it a lot tho.
> However, my aim is to understand and see how far i can get with  
> such a project rather than  building the best ever made plug-in in  
> the industry :)
>
> I've read your threat and it's very useful although i'm not really  
> sure how (and where) can i apply these techniques (such as look up  
> tables, and randomisation). Also, i'm not really sure what the  
> tollerance of randomisation should be, and in which values applied.
>
> thanks,
> mike

why dont you find a friend with a tape machine, run a saw wave through it, and use the response to do some wave shaping. or just guess.

or just buy this thing from my friend greg.

why dont you find a friend with a tape machine, run a saw wave through it, and use the response to do some wave shaping.  or just guess.  

or just buy this thing from my friend greg.
http://www.anamodaudio.com/


Analog flavor in compressor is not that much about "adding" harmonics.Rather, the harmonics should comes from your transfert function and your time constants. Not by addings harmonics on top.

Otherwise you'll end up with a comp that still have digital nasty harmonics, but with added low order harmonics.

You'd better try to replace the digital artefacts by some pleasing analog artefacts.

To do so, I suggest you code some soft knee, and especially adress the smoothing/filtering on your sidechain signal (This is the most important step).

What have you done in this specific aera so far ?

Maybe you can post your patch so I can advice you further ?

Analog flavor in compressor is not that much about "adding" harmonics.Rather, the harmonics should comes from your transfert function and your time constants. Not by addings harmonics on top. 

What have you done in this specific aera so far ?
Maybe you can post your patch so I can advice you further ?

non convolution analogue warmth