[PD] pd internals / block delay
IOhannes m zmoelnig
zmoelnig at iem.at
Wed Feb 18 12:56:48 CET 2004
Tim Blechmann wrote:
>>i am *very* sorry that i ever introduced this stupid [nop~] object
>>that does delay the signal by one block.
>>this should be considered a bug!
>>probably i'll change the behaviour in future versions of zexy to a
>>real [nop~] (which does exactly *nothing*)
>
> i see, so the nop~ that is used in the limiter help-patch isn't really
> necessary?
>
no, it IS necessary, but this is even more a problem.
the [limiter~] does some interpolation of the signal to get the real
maximum. eg, if you have a digital signal "... 0 0 0 0 0 0 1 1 0 0 0
..." the digital maximum is one, while the maximum of the d/a-converted
signal would be higher (somewhere between 1.0 and 2.0), due to
convolution with a sinc().
now [limiter~] tries to get the "real" maximum (inbetween samples).
due to the nature of signals, the [limiter~] object cannot look into the
future and thus has to delay the signal a bit (9 samples or so)
thus the signal that has to be limited should be delayed too.
i (unfortunately) have used the [nop~] object for this (because it was
at hand and there was no [z~] then)
you might argue, that 64 samples is a bit to much (if the
limiting-signal is delayed by 9 samples only) but this is about
masquarading and psychoacoustics...
so probably i'll change the help-patch to *not* use [nop~] but [z~] to
acchieve the delay.
mfg.a.sdr
IOhannes
More information about the Pd-list
mailing list