yx6@in.gr

i include the following patch that explains what i want to do exactly.

im trying to avoid making it with a signal vectorsize dependant logic (like with send~ - receive~ and a 1 sample svs for the feedback).

i thought of using poke~ to write 1 sample per time in a buffer~ and then use index~ for playing it back 1 sample later but i cant get it to work.

hi ,anyone has solved that problem?
i include the following patch that explains what i want to do exactly.

im trying to avoid making it with a signal vectorsize dependant logic (like with send~ - receive~ and a 1 sample svs for the feedback).
i thought of using poke~ to write 1 sample per time in a buffer~ and then use index~ for playing it back 1 sample later but i cant get it to work.

changing-the-frequency-of-a-phasor-only-when-its-output-0

thanks ,but i want a signal domain solution only (sample accurate)

, because look for example what will happen if i do this :

thanks ,but i want a signal domain solution only (sample accurate)
, because look for example what will happen if i do this :

I've posted a patch within the last couple of weeks or so that 

demonstrates this sort of thing. Subject heading:

I've posted a patch within the last couple of weeks or so that 
demonstrates this sort of thing.  Subject heading:
  Re: clicks using phasor~ and wave~

stefan ,your patch outputs the attached waveforms in my pc.

those are not straight phasor lines ,that means that the frequency doesnt change when the phase = 0 ,but on any phase.

stefan ,your patch outputs the attached waveforms in my pc.
those are not straight phasor lines ,that means that the frequency doesnt change when the phase = 0 ,but on any phase.

> stefan ,your patch outputs the attached waveforms in my pc. those are

> not straight phasor lines ,that means that the frequency doesnt

> change when the phase = 0 ,but on any phase.

yes, I realised that, it seems you have to give the tapin~ an argument 

of at least 1, and then its almost there. it will still produce a signal 

pierre schrieb:
> stefan ,your patch outputs the attached waveforms in my pc. those are
> not straight phasor lines ,that means that the frequency doesnt
> change when the phase = 0 ,but on any phase.

yes, I realised that, it seems you have to give the tapin~ an argument 
of at least 1, and then its almost there. it will still produce a signal 
vector delay...

-- 
Stefan Tiedje------------x-------
--_____-----------|--------------
--(_|_ ----|-----|-----()-------
-- _|_)----|-----()--------------
----------()--------www.ccmix.com


if all you're worried about is getting rid of clicks when you change the frequency, then just use a line~ object to drive the frequency inlet of your oscillator.

if you tell me you can hear the difference between a signal that changes frequencies instantly when its amplitude is 0, and a signal that ramps from one frequency to another over half a millisecond, then i'd say you're lying.

i can hear the difference for sure ,because i use a low frequency phasor~ as an index to read from a buffer not as an audio oscillator itself.

who told you all those?
i can hear the difference for sure ,because i use a low frequency phasor~ as an index to read from a buffer not as an audio oscillator itself.


> i can hear the difference for sure ,because i use a low frequency 

> phasor~ as an index to read from a buffer not as an audio 

afaik, what you are after is not possible in plain maxmsp.

i've put together a little external for my own needs, that does 

exaclty that: a phasor~ that changes its frequency only at phase 0.

this is osx only and beta - but it works for what i'm doing.

if you are still interested you can contact me off list.

shouldn't be too hard to mock it up in java and mxj~ for 

>
> who told you all those?
> i can hear the difference for sure ,because i use a low frequency  
> phasor~ as an index to read from a buffer not as an audio  
> oscillator itself.

afaik, what you are after is not possible in plain maxmsp.
i've put together a little external for my own needs, that does  
exaclty that: a phasor~ that changes its frequency only at phase 0.
this is osx only and beta - but it works for what i'm doing.

if you are still interested you can contact me off list.
shouldn't be too hard to mock it up in java and mxj~ for  
crossplattform use.

> afaik, what you are after is not possible in plain maxmsp.

No, you definitely can do this in MSP, provided that the length in 

samples of the phasor going from 0-1 is not less than the signal 

vector. (and even if it is, it means that you grab a value every 2+ 

vectors instead of every vector. At those speeds, you probably won't 

I'd be interested in your code. Here's a way of doing it in Max. It 

handles positive and negative frequencies and has a lockout feature 

around 0, so you can specify a minimum +/- value (e.g. 0.1,-0.1) to 

>
> afaik, what you are after is not possible in plain maxmsp.

No, you definitely can do this in MSP, provided that the length in 
samples of the phasor going from 0-1 is not less than the signal 
vector.  (and even if it is, it means that you grab a value every 2+ 
vectors instead of every vector.  At those speeds, you probably won't 
notice.)

I'd be interested in your code.  Here's a way of doing it in Max.  It 
handles positive and negative frequencies and has a lockout feature 
around 0, so you can specify a minimum +/- value (e.g. 0.1,-0.1) to 
prevent uber-slow phasor frequencies

> i'm sorry to disagree. the signal feedback method (which has been 

> of at least 1 sig-vector, and therefor can't be precise.

> i do agree however, that in most cases this is not audible and works 

> well enough - as in the example you provided.

> but there are examples where it does matter (try changing loop-length 

I understand the latency factor and that there's no getting around it 

from a theoretical perspective, but this almost never matters for the 

- the played back material is almost always played with an envelope.

- If you delay the output of the phasor~ using delay~ equal to the 

signal vector size and then run the remainder of your sah~ off of the 

trigger for the sah~ of the phasor~, they will now match up to within 

- because whatever phasor~ is reading through is most likely 

interpolated, or if not interpolated, not read through at a 

sample-by-sample rate, there will already be sampling artifacts.

- the length of a signal vector can be manipulated using poly~. A 

latency of 1 signal vector when the signal vector is 4 is not much. 

The length of the signal vector is related to the maximum speed that 

you'll be using the phasor~ at. If you run at slower speeds, you can 

Don't get me wrong, I think it would be nice to have an object that 

would work this way without all the extra fussing, but I do think it's 

worth it to see if you can come up with a solution without having to 

write a C external. The other thing to take into account is that if 

you write a C external and it interpolates cycle-lengths that are not 

integer length, then you're still going to have a very slight pitch 

drift issue in that some cycles will be one sample longer than other 

even though they have the same frequency. There's always noise in the 

system from sampling-rate to bit-depth. The big question is not how to 

get rid of it, but what is a tolerance you can live with. If you want 

to write an external for it, more power to you, and I look forward to 

it. In the meantime, here's some new code that works pretty well.

At a signal vector of 4, I can't hear any artifacts, and the pitch is 

perfectly stable for frequencies -20 through 20 (haven't tried other 

vector sizes). If you keep going higher, eventually the pitch is going 

to drift because the wave is not stable enough for long enough to 

produce a decent pitch, but this is just the nature of sampling, not of 

this particular system. To further optimize it, you could wrap just 

Just out of curiosity, what do you need the precision for? Reason 

being, phasor~ by its nature is imprecise. It interpolates on both x 

and y dimensions. If you want precision on both dimensions, then you 

use count~. What would be nice on count~ is to have a way to set the 

maximum at signal rate so that it's updated on the next cycle. Throw 

in a /~ and you have a sample accurate phasor.

> i'm sorry to disagree. the signal feedback method (which has been 
> proposed before) always has a latency
> of at least 1 sig-vector, and therefor can't be precise.
> i do agree however, that in most cases this is not audible and works 
> well enough - as in the example you provided.
> but there are examples where it does matter (try changing loop-length 
> without altering playback-pitch e.g.).

I understand the latency factor and that there's no getting around it 
from a theoretical perspective, but this almost never matters for the 
following reasons:

- the played back material is almost always played with an envelope.
- If you delay the output of the phasor~ using delay~ equal to the 
signal vector size and then run the remainder of your sah~ off of the 
trigger for the sah~ of the phasor~, they will now match up to within 
one sample.
- because whatever phasor~ is reading through is most likely 
interpolated, or if not interpolated, not read through at a 
sample-by-sample rate, there will already be sampling artifacts.
- the length of a signal vector can be manipulated using poly~.  A 
latency of 1 signal vector when the signal vector is 4 is not much.  
The length of the signal vector is related to the maximum speed that 
you'll be using the phasor~ at.  If you run at slower speeds, you can 
get by with a longer vector.

Don't get me wrong, I think it would be nice to have an object that 
would work this way without all the extra fussing, but I do think it's 
worth it to see if you can come up with a solution without having to 
write a C external.  The other thing to take into account is that if 
you write a C external and it interpolates cycle-lengths that are not 
integer length, then you're still going to have a very slight pitch 
drift issue in that some cycles will be one sample longer than other 
even though they have the same frequency.  There's always noise in the 
system from sampling-rate to bit-depth.  The big question is not how to 
get rid of it, but what is a tolerance you can live with.  If you want 
to write an external for it, more power to you, and I look forward to 
it.  In the meantime, here's some new code that works pretty well.

At a signal vector of 4, I can't hear any artifacts, and the pitch is 
perfectly stable for frequencies -20 through 20 (haven't tried other 
vector sizes).  If you keep going higher, eventually the pitch is going 
to drift because the wave is not stable enough for long enough to 
produce a decent pitch, but this is just the nature of sampling, not of 
this particular system.   To further optimize it, you could wrap just 
the phasor~ and related code in a poly~.

Just out of curiosity, what do you need the precision for?  Reason 
being, phasor~ by its nature is imprecise.  It interpolates on both x 
and y dimensions.  If you want precision on both dimensions, then you 
use count~.  What would be nice on count~ is to have a way to set the 
maximum at signal rate so that it's updated on the next cycle.  Throw 
in a /~ and you have a sample accurate phasor.

Hi Pierre, in the Pure Data examples, included there is an example how to do

this. That is changing the frequency on the zero crossing.

> i include the following patch that explains what i want to do exactly.

> im trying to avoid making it with a signal vectorsize dependant logic

> (like with send~ - receive~ and a 1 sample svs for the feedback).

> i thought of using poke~ to write 1 sample per time in a buffer~ and then

> use index~ for playing it back 1 sample later but i cant get it to work.

> #P user number~ 255 61 294 76 9 139 3 1 0. 0. 0 45. 250 0. 0 0 0 221 221

> #P newex 179 208 54 9109513 pong~ 1 0 1;

> #P comment 197 343 100 9109513 my output;

> #P comment 311 109 119 9109513 change the frequency only when the signal =

> #P fasten 5 0 6 1 184 266 483 266 483 86 299 86;

Hi Pierre, in the Pure Data examples, included there is an example how to do
this. That is changing the frequency on the zero crossing.

On 22/05/07, pierre  wrote:
>
>
> hi ,anyone has solved that problem?
> i include the following patch that explains what i want to do exactly.
>
> im trying to avoid making it with a signal vectorsize dependant logic
> (like with send~ - receive~ and a 1 sample svs for the feedback).
> i thought of using poke~ to write 1 sample per time in a buffer~ and then
> use index~ for playing it back 1 sample later but i cant get it to work.
>
> any opinions?
>
> thanks
> -peter
> (yx6@in.gr)
>
>
> max v2;
> #N vpatcher 15 55 740 541;
> #P outlet 178 343 15 0;
> #P window setfont "Sans Serif" 9.;
> #P user number~ 255 61 294 76 9 139 3 1 0. 0. 0 45. 250 0. 0 0 0 221 221
> 221 222 222 222 0 0 0;
> #P newex 256 116 48 9109513 sah~ 0.01;
> #P newex 179 208 54 9109513 pong~ 1 0 1;
> #P newex 179 156 87 9109513 *~;
> #P newex 178 92 48 9109513 phasor~ 1;
> #P comment 197 343 100 9109513 my output;
> #P comment 311 109 119 9109513 change the frequency only when the signal =
> 0.01;
> #P comment 298 61 119 9109513
> #P connect 5 0 8 0;
> #P connect 3 0 4 0;
> #P connect 4 0 5 0;
> #P connect 7 0 6 0;
> #P connect 6 0 4 1;
> #P fasten 5 0 6 1 184 266 483 266 483 86 299 86;
> #P pop;
>
>
>

-- 
www.centuryofnoise.com
www.perevillez.com

changing the frequency of a phasor only when its output = 0