lukhasj

suppose you were to build a ramp, over an x amount of samples.

and the point it`s plain stupid to use more.

if you were to use "two" samples ramp for instance,

the sound qualty from the wavefile would decrease dramaticly.

^^so my guess for quality playback use loads of samples ramp.

on the other hand youd want to maximise portability and speed.

by optimising the number of samples used in the ramp, to 

as low as needed to make no audible difference.

^^so my guess for optimising use no more samples than needed.

and is 512 samples enough to do this with

suppose you were to build a ramp, over an x amount of samples.
and want to use this for waveplaying.

the amount of samples used in the ramp.
and the point it`s plain stupid to use more.

if you were to use "two" samples ramp for instance,
the sound qualty from the wavefile would decrease dramaticly.

on the other hand youd want to maximise portability and speed.
by optimising the number of samples used in the ramp, to 
as low as needed to make no audible difference.

is there a key number to aim for..
and is 512 samples enough to do this with
pitching down to fhz/60 minimum, 
without loss of sound quality.


yust-curious

That's an interesting question. Shouldn't an object such as cycle~ interpolate between the buffered samples regardless of the amount of samples and the playback speed?

I cobbled up a quick test but I get strange results (see the attached screen capture):

Is there someone that can explain this behaviour?

Quote: lukhasj wrote on Fri, 20 July 2007 15:24

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

> suppose you were to build a ramp, over an x amount of samples.

> the amount of samples used in the ramp.

> and the point it`s plain stupid to use more.

> if you were to use "two" samples ramp for instance,

> the sound qualty from the wavefile would decrease dramaticly.

> ^^so my guess for quality playback use loads of samples ramp.

> on the other hand youd want to maximise portability and speed.

> by optimising the number of samples used in the ramp, to 

> as low as needed to make no audible difference.

> ^^so my guess for optimising use no more samples than needed.

> and is 512 samples enough to do this with

Quote: lukhasj wrote on Fri, 20 July 2007 15:24
----------------------------------------------------
> suppose you were to build a ramp, over an x amount of samples.
> and want to use this for waveplaying.
>
> there is a balance between
>
> the amount of samples used in the ramp.
> and the point it`s plain stupid to use more.
>
> if you were to use "two" samples ramp for instance,
> the sound qualty from the wavefile would decrease dramaticly.
>
> ^^so my guess for quality playback use loads of samples ramp.
>
> on the other hand youd want to maximise portability and speed.
> by optimising the number of samples used in the ramp, to 
> as low as needed to make no audible difference.
>
> ^^so my guess for optimising use no more samples than needed.
>
> is there a key number to aim for..
> and is 512 samples enough to do this with
> pitching down to fhz/60 minimum, 
> without loss of sound quality.
>
----------------------------------------------------


On 20 juil. 07, at 16:04, Mattijs Kneppers wrote:

> That's an interesting question. Shouldn't an object such as cycle~ 

> interpolate between the buffered samples regardless of the amount 

Nope. Cycle~ plays a buffer of 512 samples (513 with the 513th which 

> That's an interesting question. Shouldn't an object such as cycle~  
> interpolate between the buffered samples regardless of the amount  
> of samples and the playback speed?

Nope. Cycle~ plays a buffer of 512 samples (513 with the 513th which  
have the same value as the first one).

Quote: lukhasj wrote on Fri, 20 July 2007 07:24

when i understand you right you forgot about one thing:

when you send a ramp from 0. to 1. using [phasor~] 

or [cycle~] this signal is always running at the same 

signal rate as your [wave~] or [play~] object.

and like mattijs says, of course they are interpolating.

and btw [wave~] is interpolating too, so a shift of -1

octave - half speed replaying - preserves the nyquist

frequency/runtime signal rate for further processing.

are you building a synth? yes 512 should be enough.

the waldorf microwave uses 64, the access virus A uses 900.

a number like 900 is cool, because it fits into 

44,100 exactly 49 times - so when you create 

wavetables, you can use any editor program @44,100

easier to biuld harmonics series and all.

thats a matter of taste and technology but i find 

pitching up always the more difficult part in a sampler

if you really need chrystal clear HF content when 

you request a note of 10Hz you might want to make 

longer wavetables, for example 22,050 long.

but when you fill the room you have now for HF 

content at low keys in such a wavetable, this w

avetable will alias even more at high keys ....

so changing the lenght of the wavetable is always 

relative to the speed its beeing played later.

if you want to raise the quality of the sound, you

have to raise the samplingrate of the whole runtime 

(and hardware), longer wavetables dont mean much.

longer wavetables only give you a bigger frequency

range than very short ones, but always relative to 

runtime signal rate. (relative to the base note of

the synth if you want to call it like that)

imagine a wave of only 7 samples lenght: you cant 

play a proper triangle wave with it - at play speeds

below 44.100/7 Hz it will sound worse than the

Quote: lukhasj wrote on Fri, 20 July 2007 07:24
----------------------------------------------------
> suppose you were to build a ramp, over an x amount of samples.
> and want to use this for waveplaying.
>
> there is a balance between
>
> the amount of samples used in the ramp.
> and the point it`s plain stupid to use more.
>
> if you were to use "two" samples ramp for instance,
> the sound qualty from the wavefile would decrease dramaticly.
>
> ^^so my guess for quality playback use loads of samples ramp.
>
> on the other hand youd want to maximise portability and speed.
> by optimising the number of samples used in the ramp, to 
> as low as needed to make no audible difference.
>
> ^^so my guess for optimising use no more samples than needed.

when i understand you right you forgot about one thing: 

when you send a ramp from 0. to 1. using [phasor~] 
or [cycle~] this signal is always running at the same 
signal rate as your [wave~] or [play~] object.

and btw [wave~] is interpolating too, so a shift of -1
octave - half speed replaying - preserves the nyquist
frequency/runtime signal rate for further processing.

> is there a key number to aim for..
> and is 512 samples enough to do this with
> pitching down to fhz/60 minimum, 
> without loss of sound quality.

a number like 900 is cool, because it fits into 
44,100 exactly 49 times - so when you create 
wavetables, you can use any editor program @44,100
with a basefrequency of 49.0 Hz. 
easier to biuld harmonics series and all.

thats a matter of taste and technology but i find 
pitching up always the more difficult part in a sampler
or wavetable synth.

if you really need chrystal clear HF content when 
you request a note of 10Hz you  might want to make 
longer wavetables, for example 22,050 long.
but when you fill the room you have now for HF 
content at low keys in such a wavetable, this w
avetable will alias even more at high keys .... 

so changing the lenght of the wavetable is always 
relative to the speed its beeing played later.
if you want to raise the quality of the sound, you
have to raise the samplingrate of the whole runtime 
(and hardware), longer wavetables dont mean much.

longer wavetables only give you a bigger frequency
range than very short ones, but always relative to 
runtime signal rate. (relative to the base note of
the synth if you want to call it like that)

imagine a wave of only 7 samples lenght: you cant 
play a proper triangle wave with it - at play speeds
below 44.100/7 Hz it will sound worse than the
signal theoretially allows.

it`s a granular build, so it does run subsonic.

not a real problem becouse i kinda like the 

it`s a granular build, so it does run subsonic.
not a real problem becouse i kinda like the 
uninterpolated sound with a phasor~.


yust curious