volker böhm

i wasn't following the thread about stringresonator tunings etc. very 

carefully and always thought that the bad tuning was due to tapin/ 

tapout not being able to have subsample delay length - only to find 

feeding the delay time as signal, tapout~ does subsample 

only thing, and i would classify this as a bug, the specified delay 

time and the resulting delay are 1 sample off (see patch below). btw 

when accounting for that, i get pretty good tunings, even for high 

the interpolation in tapin/tapout is even better than those in comb~ 

using fractional delay times, there is heavy loss in the high freq 

range, which leads me to thinking they are just doing linear 

interpolation. would be nice to have the same as in tapin/tapout.

i wasn't following the thread about stringresonator tunings etc. very  
carefully and always thought that the bad tuning was due to tapin/ 
tapout not being able to have subsample delay length - only to find  
out yesterday, that it's not true.
feeding the delay time as signal, tapout~ does subsample  
interpolation, which is nice!
only thing, and i would classify this as a bug, the specified delay  
time and the resulting delay are 1 sample off (see patch below). btw  
this is on osx 10.4.4, maxmsp 4.5.6 --
when accounting for that, i get pretty good tunings, even for high  
notes.

the interpolation in tapin/tapout is even better than those in comb~  
and teeth~.
using fractional delay times, there is heavy loss in the high freq  
range, which leads me to thinking they are just doing linear  
interpolation. would be nice to have the same as in tapin/tapout.
cheers,
volker.

fb-delay-line-tunings-tapintapout-1-sample-off

Quote: vb wrote on Sat, 16 December 2006 09:40

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

> i wasn't following the thread about stringresonator tunings etc. very 

> carefully and always thought that the bad tuning was due to tapin/ 

> tapout not being able to have subsample delay length - only to find 

that tapout can not interpolate or can no have a 

fine resolution is only an old rumor because some 

old helpfile did show the use of integer values only.

but float, or signal like you said, is also accepted

what can not be done is very _short delays, this is where 

Quote: vb wrote on Sat, 16 December 2006 09:40
----------------------------------------------------
> i wasn't following the thread about stringresonator tunings etc. very  
> carefully and always thought that the bad tuning was due to tapin/ 
> tapout not being able to have subsample delay length - only to find  
> out yesterday, that it's not true.

that tapout can not interpolate or can no have a 
fine resolution is only an old rumor because some 
old helpfile did show the use of integer values only.

but float, or signal like you said, is also accepted
and performed.

what can not be done is very _short delays, this is where 
you need replace tap with teeth

On 17 Dec 2006, at 06:22, Roman Thilenius wrote:

> that tapout can not interpolate or can no have a

> fine resolution is only an old rumor because some

> old helpfile did show the use of integer values only.

i'm not talking about integer vs float delay times.

to specify a delay time of 100 samples, of course you need a floating 

the point is, that you can't have a delay time of e.g. 100.5 samples 

and i always thought this limitation also stands for specifing the 

not true, tapin/tapout perform a nice subsample interpolation.

the latest thread about it (with incorrect conclusion), is not long ago:

t=msg&goto=83648&rid=0&srch=stringresonator#msg_83648

the confusion concerning the wrong tuning of waveguide/karplusstrong 

etc. seems related to a BUG in tapin/tapout, where the delay time 

given as signal is interpreted 1 sample too long.

> but float, or signal like you said, is also accepted

there is a substantial difference in performance of tapin/tapout 

between a delay time specified as float and as signal.

> what can not be done is very _short delays, this is where

but interpolation in teeth~ and comb~ is not as good as in tapin/ 

tapout (heavy damping of high freq contents)

> that tapout can not interpolate or can no have a
> fine resolution is only an old rumor because some
> old helpfile did show the use of integer values only.

i'm not talking about integer vs float delay times.
to specify a delay time of 100 samples, of course you need a floating  
point value to express it in ms.
this has always worked.

the point is, that you can't have a delay time of e.g. 100.5 samples  
if you specify it with a float.
and i always thought this limitation also stands for specifing the  
delay time with a signal.
not true, tapin/tapout perform a nice subsample interpolation.

the latest thread about it (with incorrect conclusion), is not long ago:
https://cycling74.com/forums/index.php? 
t=msg&goto=83648&rid=0&srch=stringresonator#msg_83648
from october 2006

the confusion concerning the wrong tuning of waveguide/karplusstrong  
etc. seems related to a BUG in tapin/tapout, where the delay time  
given as signal is interpreted 1 sample too long.

> but float, or signal like you said, is also accepted
> and performed.

there is a substantial difference in performance of tapin/tapout  
between a delay time specified as float and as signal.
this is even mentioned in the manuals :)

>
> what can not be done is very _short delays, this is where
> you need replace tap with teeth
true, it depends on sigvs.
but interpolation in teeth~ and comb~ is not as good as in tapin/ 
tapout (heavy damping of high freq contents)
have you actually looked at my patch?

fb-delay line tunings -> tapin/tapout 1 sample off