[music-dsp] A question about Dirac pulses

koen vos koen.vos at globalipsound.com
Fri Jan 9 18:24:01 EST 2004


The reason for using a Dirac pulse is simplicity: the recorded response is
immediately equal to the impulse response of the accoustic space that you're
looking for - no mathematical transformation required.

For this to work, the pulse needs to have a flat spectrum, but only up to
the Nyquist frequency. One example of such a pulse is a sinc pulse with
"period" of one sample period. It's possible to construct shorter pulses
satisfying the requirement, but those will contain wasteful energy above the
Nyquist frequency.

However, by giving up on the simplicity of mathlessness, you can use any
pulse having at least some energy everywhere in the spectrum, and deconvolve
the recorded response with the pulse, in order to get the desired
accoustical impulse response. In that case the "pulse" could also be a long
noise signal..

koen.


Glen Berry wrote:
> When capturing the impulse response of either an acoustic space,
> or a piece
> of electronic gear, some people suggest using a Dirac pulse. In theory, I
> believe these would be infinitely large, and infinitely short. In reality,
> we just try to use as short an impulse as we can, and keep the
> amplitude as
> large as practical.
>
> My question is, as long as the pulse is shorter than the duration of one
> sample period, is there any difference in the end results? Let's
> say I have
> a way of generating an acoustic pulse that could be the same duration as
> one sample period, or I could chose to make it significantly shorter than
> one sample. Would there be any advantage in the shorter period pulse? My
> intuition leans toward making the pulse the length of a sample period, in
> an attempt to generate more energy in the impulse, compared to a much
> shorter impulse. Am I thinking correctly?
>
> thanks,
> Glen




More information about the music-dsp mailing list