Riccardo Santalucia

Hi all, 4 years that I'm peeking the forum without joining, time has come

I'm building some spatialization utility that i will import in M4L.

Building my own matrix~ function to have non linear fades, i'm asking myself which approach would actually be the most efficient. In my utility, when i get the time "t" in which a sound will go from speaker_a to speaker_b, a fade out occurs at speaker_a and a fade in at speaker_b, with a equal power curve obtained by a squared root.

With a square root exp i'm feeding into a coll 512 steps value from 0. to 1., building a table of data to retrieve using a line, which sends 512 values to the coll object.

Feeding 512 value/samples (always square root) into a buffer named fade with the peek~ object.

In this case to make the fade i'm using "cycle~ fade" which receive phase modulation by line~, reading it's 512 values.

Actually I tried to do this in gen without success. You can see all in the patch.

Actually if the ramp i gave to line to set che "cycle~ fade" goes up to 1. the multiplier is set to 0. ... not untedrstanding why, so in the patch i've set 0.99.

Maybe i get myself stuck in things I'm not so brilliant with, but I'm trying to figure out the best suited solution (the less expensive for cpu).

P.S. maybe the audio way is the best but with gen? Can't get it right...

512 is quite the amount of steps, u thing less would be less heavy yet enough performative? For sure gonna make tries myself

button

comment

loadmess

live.gain~

number

line~

message

cycle~

gen~

buffer

cycle

ezdac~

flonum

coll

dac~

inlet

umenu

number~

gate

line

playlist~

send~

gate~

receive~

expr

buffer~

peek~

pack

loadbang

panel

I'm building some spatialization utility  that i will import in M4L.
Building my own matrix~ function to have non linear fades, i'm asking myself which approach would actually be the most efficient. In my utility, when i get the time "t" in which a sound will go from speaker_a to speaker_b, a fade out occurs at speaker_a and a fade in at speaker_b, with a equal power curve obtained by a squared root.
I see two option :

1) Data
With a square root exp i'm feeding into a coll 512 steps value from 0. to 1., building a table of data to retrieve using a line, which sends 512 values to the coll object.

2)Audio
I've grabbed pan2 solution.
Feeding 512 value/samples (always square root) into a buffer named fade with the peek~ object.
In this case to make the fade  i'm using "cycle~ fade" which receive phase modulation by line~, reading it's 512 values.
Actually I tried to do this in gen without success. You can see all in the patch.
Actually if the ramp i gave to line to set che "cycle~ fade" goes up to 1. the multiplier is set to 0. ... not untedrstanding why, so in the patch i've set 0.99.

Maybe i get myself stuck in things I'm not so brilliant with, but I'm trying to figure out the best suited solution (the less expensive for cpu).
Any help would be really appreciated : )

P.S. maybe the audio way is the best but with gen? Can't get it right...
512 is quite the amount of steps, u thing less would be less heavy yet enough performative? For sure gonna make tries myself


Hi all, 4 years that I'm peeking the forum without joining, time has come
I'm building some spatialization utility  that i will import in M4

which-is-the-best-fade-algo-for-cpu-usage-audio-mspgen-or-data-table

i don think data is an option. using a buffer is the most straightforward and objects like cycle or play offer you all kinds of reading through it you will need, including proper interpolation.

i dont know much enough about gen to make a claim about its CPU usage versus MSP, but using MSP will give you options like downsampling or turning off unused subcircuits using poly~ while gen is not.

> not untedrstanding why, so in the patch i've set 0.99.

First of all the phase at value 1 is that same as the phase at value 0. Moreover, the cycle~ object will interpolate the end of a buffer to the beginning, therefor it is not the best tool for reading ramps. The wave~ object has the same issue of interpolating the end of the buffer back to the beginning. That can not be changed unless interpolation is turned off altogether with the interp attribute.

The cycle object within gen~ behaves the same as cycle~. The reason it doesn't work seems because of the buffer name that contains a dot. When I remove it, it works.

A good option in gen~ however is the peek object, with the index attribute set to phase.

Thanks, almost never used poly, was one big missing part I knew first or later i would have to deal with. I've implemented it with the mute function and it's working .

That was the misdeed! I've got rid of the dot.

The gen option with peek is working fine, so I've introduced this little gen brick inside the poly.

@Roman
Thanks, almost never used poly,  was one big missing part I knew first or later i would have to deal with.  I've implemented it with the mute function and it's working .

@JVKR
That was the misdeed! I've got rid of the dot.
The gen option with peek is working fine, so I've introduced this little gen brick inside the poly.
Thank you



Which is the best fade algo (for cpu usage), audio msp/gen or data table?