cdaata%taux01.nsc.com@nsc.nsc.com (02/08/90)
Hello, I am looking for a digital filter design package which allows the user to specify the filter's group delay. I'd appreciate if someone could point me to such software if it exists. Thanks, David. David Almagor, National Semiconductor (IC) Ltd., P.O. Box 3007, Herzliya B 46104, Israel. Tel: 011-972-52-522255 e-mail: cdaata%taux01.nsc.com@nsc.nsc.com
bill@videovax.tv.tek.com (William K. McFadden) (02/10/90)
In article <3274@taux01.UUCP> cdaata%taux01.nsc.com@nsc.nsc.com writes: >I am looking for a digital filter design package which allows the >user to specify the filter's group delay. I'd appreciate if someone >could point me to such software if it exists. Actually, they all do, in a sense. If you implement a linear phase FIR filter by direct convolution, the group delay will be I for a filter with 2I+1 taps. Hence, if you want a group delay of 50 samples, then the filter should have 101 taps. Every filter design program I've seen lets you specify the number of taps. You can convert the group delay to seconds by dividing by the sample rate. For example, with a 50 KHz sampling rate, a group delay of 50 samples would be 1 mS. In an IIR flter, the group delay is nonlinear. In this case you would only be able to specify its value at a single frequency. I don't know of any filter design packages that do this, although many will plot group delay vs. frequency. -- Bill McFadden Tektronix, Inc. P.O. Box 500 MS 58-639 Beaverton, OR 97077 bill@videovax.tv.tek.com, {hplabs,uw-beaver,decvax}!tektronix!videovax!bill Phone: (503) 627-6920 "The biggest difference between developing a missle component and a toy is the 'cost constraint.'" -- John Anderson, Engineer, TI
cdaata@tasu81.UUCP (David Almagor) (02/11/90)
The point might have been missed in my short posting, but I obviously meant that the designer will be able to specify a variable group delay as a function of frequency (and not a fixed delay as in a linear-phase FIR). If Anyone knows of such a package, please let me know by posting or e-mail. Thanks, David. ---------------------------------------------------------------------- David Almagor, National Semiconductor (IC) Ltd., P.O. Box 3007, Herzliya B 46104, Israel. Tel: 011-972-52-522255
dmcgrath@lennon.austek.oz (David Mcgrath) (02/12/90)
In article <3288@taux01.UUCP> you write: > >The point might have been missed in my short posting, but I obviously >meant that the designer will be able to specify a variable group delay >as a function of frequency (and not a fixed delay as in a linear-phase FIR). >If Anyone knows of such a package, please let me know by posting or e-mail. > We have built a fairly general-purpose filtering box at Austek that we have used to implement such a filter. The box implements a 32768-tap FIR filter on CD audio data (in real time). The coefficients are downloaded by a PC. One demo we usually run on this system is a 1/3 octave graphic equaliser, using the PC and mouse to implement the 1/3 octave display. As a variant of this, someone here modified the equalizer coefficent calculation code to do a variable group delay equalizer, so that you can alter the group delay of the signal by over 1/2 second with 1/3 octave frequency resolution. We compute the filter response in the frequency domain, so that you can download into the box a vector of 32768 complex numbers that represent the complex response of the filter at 32768 frequency bins (from DC to 22.05 KHz). Since group delay is simply the derivative of the phase response, all you have to do is integrate the group delay as you count throught the frequency bins, and generate 32768 complex numbers, all with unity magnitude and phase that is the integral of group delay. (You can also specify filter responses as time-domain impulse responses, which is handy for building reverberation filters etc...) If anyone out there wants more info on our box, please get in touch with me. David McGrath, Phone : +61-8-260-0114 Austek Microsystems, Email : dmcgrath@austek.oz Technology Park, South Australia, 5095, Australia.