Mike S

I'm working on simple synthesiser with effects. I've come unstuck on the routing.

I have 3 effects, which I want to be able to route the signal to in any order. I can imagine ways to do it, but when I actually come to it, I can't quite get it right

I know this is probably quite a simple question, but seeing as though I've been banging my head against the wall for quite some time I wouldn't mind being pointed in the right direction

signal-routing-using-matrix-control

a real "free routing" solution would include the 

possibilty to route effect A to effect A i.e. will

require the implementation of [send~ tralala] and

[recieve~ tralala] in the signal path for "feedback".

if you stick with 3 blackboxes, i would recommend 

you to build all possibilties A->B->C, A->C->B, and

so on as subpatches and and lod them using [poly~].

or do it the other way round: make a simple serial 

setup of 3 subpatches. each subpatch will contain

all the effects in parallel and all you have to do 

is switching between them. this would add the possibilty

of routings such as A,A,A = three time the same effect.

a real "free routing" solution would include the 
possibilty to route effect A to effect A i.e. will
require the implementation of [send~ tralala] and
[recieve~ tralala] in the signal path for "feedback".

if you stick with 3 blackboxes, i would recommend 
you to build all possibilties A->B->C, A->C->B, and
so on as subpatches and and lod them using [poly~].

or do it the other way round: make a simple serial 
setup of 3 subpatches. each subpatch will contain
all the effects in parallel and all you have to do 
is switching between them. this would add the possibilty
of routings such as A,A,A = three time the same effect.

Probably use some sort of poly~ setup. Or you can use matrix~ where you can set the gain~ ramp times to avoid clicks.

Here's a folloq-up question which will probably show my ignorance, but:

If you have two effects, what's the difference in the order of applying them? If you apply a delay~ then a filtergraph~ is the result different than if you reverse this order? From my (limited) experience they sound pretty similar, but I don't know. If the math involved is all addition and multiplication it shouldn't matter due to the commutative property right?

BTW if you don't have a filtergraph~ in there I would recommend it, it's a great effect to have. Try some automated swoops through frequency ranges. Very cool. Just watch out for levels as they can get super loud and distorted quickly... but when they're not, they're oh so sweet.

If you have two effects, what's the difference in the order of applying them?  If you apply a delay~ then a filtergraph~ is the result different than if you reverse this order? From  my (limited) experience they sound pretty similar, but I don't know. If the math involved is all addition and multiplication it shouldn't matter due to the commutative property right?

Quote: Roman Thilenius wrote on Mon, 09 July 2007 16:10

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

> a real "free routing" solution would include the 

> possibilty to route effect A to effect A i.e. will

> require the implementation of [send~ tralala] and

> [recieve~ tralala] in the signal path for "feedback".

I was wondering what this would sound like?

> If you have two effects, what's the difference in the order of applying them? If you apply a delay~ then a filtergraph~ is the result different than if you reverse this order? From my (limited) experience they sound pretty similar, but I don't know. If the math involved is all addition and multiplication it shouldn't matter due to the commutative property right?

Yes I have a filtergraph, I've actually put this in with each oscillator (5 in total). I have an effects or processes section which contains some bit-crushing, distortion, and compression. So in answer to your question, something that has been compressed, then distorted will sound different to something that has been distorted then compressed, at least I think they will :/

Quote: Roman Thilenius wrote on Mon, 09 July 2007 16:10
----------------------------------------------------
>
> a real "free routing" solution would include the 
> possibilty to route effect A to effect A i.e. will
> require the implementation of [send~ tralala] and
> [recieve~ tralala] in the signal path for "feedback".

> If you have two effects, what's the difference in the order of applying them?  If you apply a delay~ then a filtergraph~ is the result different than if you reverse this order? From  my (limited) experience they sound pretty similar, but I don't know. If the math involved is all addition and multiplication it shouldn't matter due to the commutative property right?

Yes I have a filtergraph, I've actually put this in with each oscillator (5 in total). I have an effects or processes section which contains some bit-crushing, distortion, and compression. So in answer to your question, something that has been compressed, then distorted will sound different to something that has been distorted then compressed, at least I think they will :/


I suppose one could readily do some experimentation with that, trying different combinations and re-recording, then comparing the buffers~ that are produced. Max is definitely good for that... bang thru all the signal values of each resulting buffer~ index~ using an uzi, subtract one from the other, and there's your answer... if values deviate (significantly? or at all?) from 0 they're different.

Of course, "different" is relative and perceptual, so the numbers may show a difference that's tough to hear.

hmm.... interesting thought... guess I won't get any other work done tonight!

I suppose one could readily do some experimentation with that, trying different combinations and re-recording, then comparing the buffers~ that are produced. Max is definitely good for that... bang thru all the signal values of each resulting buffer~ index~ using an uzi, subtract one from the other, and there's your answer... if values deviate (significantly? or at all?) from 0 they're different.  

Quote: Mike S wrote on Mon, 09 July 2007 16:24

> Quote: Roman Thilenius wrote on Mon, 09 July 2007 16:10

> ----------------------------------------------------

> > a real "free routing" solution would include the 

> > possibilty to route effect A to effect A i.e. will

> > require the implementation of [send~ tralala] and

> > [recieve~ tralala] in the signal path for "feedback".

> I was wondering what this would sound like?

hehe, thats the problem that one might hear it. :)

its simply neccessary to break the signal chain at one

point when you want to connect blackbox A to itself.

probably i completely under- oder over-rate your knowlegde.

i knew that you do not get that abstraction of the

english language which i posted to the forum with my

drunken head - so i made you an working example as patches

now. feel free to ask if anything is unclear.

this is the "pro" version of how to do it. ;)

put all in search path, restart maxmsp, and run "SynthABC.mxb"

it lacks initialisation atm but that doesn matter.

Quote: Mike S wrote on Mon, 09 July 2007 16:24
----------------------------------------------------
> Quote: Roman Thilenius wrote on Mon, 09 July 2007 16:10
> ----------------------------------------------------
> >
> > a real "free routing" solution would include the 
> > possibilty to route effect A to effect A i.e. will
> > require the implementation of [send~ tralala] and
> > [recieve~ tralala] in the signal path for "feedback".
>
>
> I was wondering what this would sound like?
>

its simply neccessary to break the signal chain at one
point when you want to connect blackbox A to itself.

i knew that you do not get that abstraction of the
english language which i posted to the forum with my
drunken head - so i made you an working example as patches
now. feel free to ask if anything is unclear.

Quote: seejayjames wrote on Mon, 09 July 2007 15:14

Order of effects matters a lot! You are probably thinking of convolution-based effects, which do have a commutative property so ordering does not matter in that case. But most interesting effects are not purely convolution-based. For example, a single delay is a very simple type of convolution but you cannot model delay+feedback with convolution. So two delay+feedback effects units in series would give completely different results by swapping the delay times. Order matters for many other time delay effects (chorus, flange, phaser) since they have a feedback aspect, distortion effects, all but the simplest types of reverb, etc.

Feedback seems to be the common reason order matters in my examples, but I'd imagine most effects implemented via arbitrary DSP algorithms (not convolution) would not be commutative.

Quote: seejayjames wrote on Mon, 09 July 2007 15:14
----------------------------------------------------
> If you have two effects, what's the difference in the order of applying them?  If you apply a delay~ then a filtergraph~ is the result different than if you reverse this order? From  my (limited) experience they sound pretty similar, but I don't know. If the math involved is all addition and multiplication it shouldn't matter due to the commutative property right?

you'll need dialmode though, and a graphic for the dialmode which i've attached

matrix~ should do it fine
you'll need dialmode though, and a graphic for the dialmode which i've attached

Hmmm, I sent something very much like Binray's post with a smaller (53k) attachment, and it bounced for being 'too big'? Just how does the forum bot make these decisions...?

Hmmm, I sent something very much like Binray's post with a smaller (53k) attachment, and it bounced for being 'too big'? Just how does the forum bot make these decisions...?
cheers
Roger


Very nice! I love that matrixctrl and can see lots uses beyond matrix~.

On 7/10/07 5:33 AM, "James Little" wrote:

Very nice!  I love that matrixctrl and can see lots uses beyond matrix~.

On 7/10/07 5:33 AM, "James Little"  wrote:

Cheers
Gary Lee Nelson
Oberlin College
www.timara.oberlin.edu/GaryLeeNelson

Thanks Adamj, that helped understand a lot about the processing. I had a feeling that order did matter in that lots of processes weren't commutative, as well as the feedback element (applying various levels of effect upon something will have increasing or decreasing effect upon the signal over time).

The idea that any signal is a stream of really fast numbers that one can apply virtually any kind of operation to is kind of mindblowing, it's so great that computers have caught up to what modern musicians want to be able to do.

Can't help but think how spoiled we are compared to musicians of the past... ;)

Can't help but think how spoiled we are compared to musicians of the past...  ;)

>Quote: seejayjames wrote on Mon, 09 July 2007 15:14

>----------------------------------------------------

> > If you have two effects, what's the difference in the order of 

a band pass filter after a grain synth has nothing in common with a 

>Quote: seejayjames wrote on Mon, 09 July 2007 15:14
>----------------------------------------------------
>  > If you have two effects, what's the difference in the order of 
>applying them?

a band pass filter after a grain synth has nothing in common with a 
grain synt after a band pass

Signal Routing using Matrix Control