Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!ucsd!pacbell.com!tandem!netcom!mcmahan
From: mcmahan@netcom.UUCP (Dave Mc Mahan)
Newsgroups: comp.dsp
Subject: Re: Having messed with FFT programs, I have a few questions...
Message-ID: <18034@netcom.UUCP>
Date: 4 Dec 90 06:06:20 GMT
References: <1811@beguine.UUCP> <17926@netcom.UUCP> <18033@netcom.UUCP>
Organization: Dave McMahan @ NetCom Services
Lines: 32


 In a previous article, By-Tor writes:
>Wouldn't it be possible to rid yourself of the "glitch" problem by crossfading
>the section you are working with?  They do this to get "clickless" loops in
>samplers, and it seems the technique might work here.  Just make a copy of the
>section, reverse it, ramp the volume up like this:  "/"

This is essentially what windowing does, but the window used is better at
filtering out the bogus harmonics without introducing problems of it's own.
Windowing requires the creation of the proper window and then multiplying
it by the data samples.  Instead of being a straight ramp, it is more of a
smooth curve.  Various windows are available, each with benefits and drawbacks
of their own.  Theoretically, it should give you better performance than just
a straight ramp-up ramp-down technique.  After all, you have to decide where
to start the ramp for the downward tailoff and where to end the ramp for the
initial ramp-up.   The ramp technique really is a trapezoidal window and
therefore introduces various types of distortion in the frequency domain.  It
does have the benefit of working well for mixing two signals together for
sending to a speaker and being quite easy to implement, but it isn't the
optimal choice for the task of looking at undistorted data in the frequency
domain.  If somebody wants me to, I'll look up a couple of different types
of windows at work tomorrow and post the formulas used to this newsgroup.
The functions can be easily plotted using a spreadsheet or other program
and you can see for yourself just exactly how each window looks.



   -dave

-- 
Dave McMahan                            mcmahan@netcom.uucp
					{apple,amdahl,claris}!netcom!mcmahan


Brought to you by Super Global Mega Corp .com