Subject: [linux-audio-dev] learning DSP?
From: Stuart Allie (Stuart.Allie_AT_hydro.com.au)
Date: Fri Dec 21 2001 - 06:54:09 EET
Okay, Paul's initial explanation and Lamar's comments make it (eq via delay
lines) fairly obvious in hindsight, thanks guys.
To switch topic a bit, I just came across "The Scientist and Engineer's
Guide to Digital Signal Processing" on the web - does anybody have any
comments about the quality of this as an intro to DSP? Any recommendations
for other DSP sources?
TIA
Stuart
> On Wednesday 19 December 2001 08:39 pm, Paul Davis wrote:
> > >Just curious, but could somebody explain *how* delay lines
> can be used
> > >implement EQ? I have a strong maths background, but no DSP
> experience if
> > >that helps.
>
> > i'm not a dsp programmer, but its really quite simple. if you
> > feedback with a delay of just 1 sample, and attenutate both the
> > current and previous sample by 0.5:
>
> > the actual details are extremely hairy though - there is a lot of
> > sophisticated math that goes into really good filter design, plus a
> > lot of subjective, non-double blind tested "opinion" :)
>
> In short: Z transforms are your friend. Once I grasped what
> the Z transform
> did for you in control systems theory, I immediately realized
> that filtering
> is a natural for Z transform math.
>
> But, in a nutshell:
>
> Delaying a set time and adding back to the original produces
> a 'comb' filter.
> The amount of the delay and the depth of the readdition
> together produce
> various degrees of filtering. Notch filters are easiest with
> delays -- delay
> one half cycle at the notch and add back one hundred percent.
> You get a
> rather tight notch. Along with other neat effects. :-) Like
> the peak at
> twice the notch frequency.... :-) And the secondary notch at
> thrice the
> notch frequency. Even multiples peak, odd multiples notch --
> thus a comb
> filter.
>
> Comb filters are used rather nicely in chroma/luma separation
> in Never The
> Same Color video.
> --
> Lamar Owen
> WGCR Internet Radio
> 1 Peter 4:11
>
This archive was generated by hypermail 2b28 : Fri Dec 21 2001 - 06:54:38 EET