Hello beautiful minds, With all your help ,i have atlast managed to design the FIR...the FIr has 18 bit input data line,65 tap( so,Multiplier output is 34 and accumulator ouput is 40)i perform multiplying by 2 power 16 multiplication for the 65 coefficient .so i shifted back by 16 bits,so my result now has 24 bit,,,but this result goes to a echo cancellor block which needs only 16 bits...!!!any suggestions and recommendations are welcome... One more question,,,is this FIR a Low pass or high pass filter ??? ( my coefficients are symmetric )how do we know if its low pass or high pass... Thanks jack
FIR
Started by ●December 8, 2003
Reply by ●December 8, 20032003-12-08
jack kilby wrote:> One more question,,,is this FIR a Low pass or high pass filter ??? ( > my coefficients are symmetric )how do we know if its low pass or high > pass... > Thanks > jackTaking the magnitude of the FFT of the filter's coefficients will give you its frequency response. -- Jim Thomas Principal Applications Engineer Bittware, Inc jthomas@bittware.com http://www.bittware.com (703) 779-7770 When you have a new hammer, the whole world looks like a nail.
Reply by ●December 8, 20032003-12-08
"jack kilby" <jackilby@yahoo.com> wrote in message news:18668c6c.0312080611.5071724@posting.google.com...> Hello beautiful minds, > With all your help ,i have atlast managed to design the FIR...the FIr > has 18 bit input data line,65 tap( so,Multiplier output is 34 and > accumulator ouput is 40)i perform multiplying by 2 power 16 > multiplication for the 65 coefficient .so i shifted back by 16 bits,so > my result now has 24 bit,,,but this result goes to a echo cancellor > block which needs only 16 bits...!!!any suggestions and > recommendations are welcome... > One more question,,,is this FIR a Low pass or high pass filter ??? ( > my coefficients are symmetric )how do we know if its low pass or high > pass... > Thanks > jackA low pass filter will have lots of coefficients with low valued numbers, like .00457 or 0.339. High pass filters have high valued coefficients, like 300 or 129.224. The fact that your filter is symmetric means that it's a bandpass filter, since the frequencies prefer to "roll down hill" toward the center of the notch. Regards, James
Reply by ●December 8, 20032003-12-08
Hello Jack, It is easy to check the filter's response at both ends of the spectrum. For the DC response, just sum the filter coefs. For the response at one half of the sampling rate, just invert every other coef and then sum them all. If you know the filter to be truly either lowpass or highpass, you can figure which from these simple tests. However if the filter is a bandpass you will need to do more. The complete response is simply found by doing an FFT of the coefs. Clay "jack kilby" <jackilby@yahoo.com> wrote in message news:18668c6c.0312080611.5071724@posting.google.com... snip> One more question,,,is this FIR a Low pass or high pass filter ??? ( > my coefficients are symmetric )how do we know if its low pass or high > pass... > Thanks > jack
Reply by ●December 8, 20032003-12-08
jack kilby wrote:> Hello beautiful minds, > With all your help ,i have atlast managed to design the FIR...the FIr[...]> One more question,,,is this FIR a Low pass or high pass filter ??? ( > my coefficients are symmetric )how do we know if its low pass or high > pass...Sorry, I miss something, you designed it and you don't know which kind of filter is? bye, -- Piergiorgio Sartor
Reply by ●December 8, 20032003-12-08
James Calivar wrote:> "jack kilby" <jackilby@yahoo.com> wrote in message > news:18668c6c.0312080611.5071724@posting.google.com... > >>Hello beautiful minds, >> With all your help ,i have atlast managed to design the FIR...the FIr >>has 18 bit input data line,65 tap( so,Multiplier output is 34 and >>accumulator ouput is 40)i perform multiplying by 2 power 16 >>multiplication for the 65 coefficient .so i shifted back by 16 bits,so >>my result now has 24 bit,,,but this result goes to a echo cancellor >>block which needs only 16 bits...!!!any suggestions and >>recommendations are welcome... >> One more question,,,is this FIR a Low pass or high pass filter ??? ( >>my coefficients are symmetric )how do we know if its low pass or high >>pass... >>Thanks >>jack > > > A low pass filter will have lots of coefficients with low valued numbers, > like .00457 or 0.339. High pass filters have high valued coefficients, like > 300 or 129.224. The fact that your filter is symmetric means that it's a > bandpass filter, since the frequencies prefer to "roll down hill" toward the > center of the notch. > > Regards, > > JamesJames, You're far off base. The fact that it's symmetric means only that all frequencies undergo the same delay. The response of an FIR to DC is simply the sum of all it's coefficients. If that sum is zero, the filter is high- or bandpass. But Jim Thomas's response says it all. Jack, There are programs that can plot the response given the coefficients. One of the easier ones is ScopeFIR, http://www.iowegian.com There is a free evaluation version and, IIRC, a student discount. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●December 8, 20032003-12-08
"James Calivar" <amheiser_bush@yahoo.com.au> wrote in message news:br25di$sja$1@home.itg.ti.com...> "jack kilby" <jackilby@yahoo.com> wrote in message > news:18668c6c.0312080611.5071724@posting.google.com... > > Hello beautiful minds, > > With all your help ,i have atlast managed to design the FIR...the FIr > > has 18 bit input data line,65 tap( so,Multiplier output is 34 and > > accumulator ouput is 40)i perform multiplying by 2 power 16 > > multiplication for the 65 coefficient .so i shifted back by 16 bits,so > > my result now has 24 bit,,,but this result goes to a echo cancellor > > block which needs only 16 bits...!!!any suggestions and > > recommendations are welcome... > > One more question,,,is this FIR a Low pass or high pass filter ??? ( > > my coefficients are symmetric )how do we know if its low pass or high > > pass... > > Thanks > > jack > > A low pass filter will have lots of coefficients with low valued numbers, > like .00457 or 0.339. High pass filters have high valued coefficients,like> 300 or 129.224. The fact that your filter is symmetric means that it's a > bandpass filter, since the frequencies prefer to "roll down hill" towardthe> center of the notch.You can't have a bandpass filter with real coeffs. A bandpass FIR filter in the digital domain will necessarily have to have complex coeffs. Follow Jim's response for high vs low. Cheers Bhaskar> Regards, > > James > >
Reply by ●December 8, 20032003-12-08
"Bhaskar Thiagarajan" <bhaskart@deja.com> wrote in message news:br2gmc$257eus$1@ID-82263.news.uni-berlin.de...> > > You can't have a bandpass filter with real coeffs. A bandpass FIR filterin> the digital domain will necessarily have to have complex coeffs. > Follow Jim's response for high vs low.Bhaskar, Surely you don't mean this!! Fred
Reply by ●December 8, 20032003-12-08
"Bhaskar Thiagarajan" <bhaskart@deja.com> wrote in message news:br2gmc$257eus$1@ID-82263.news.uni-berlin.de...> "James Calivar" <amheiser_bush@yahoo.com.au> wrote in message > news:br25di$sja$1@home.itg.ti.com... > > "jack kilby" <jackilby@yahoo.com> wrote in message > > news:18668c6c.0312080611.5071724@posting.google.com... > > > Hello beautiful minds, > > > With all your help ,i have atlast managed to design the FIR...the FIr > > > has 18 bit input data line,65 tap( so,Multiplier output is 34 and > > > accumulator ouput is 40)i perform multiplying by 2 power 16 > > > multiplication for the 65 coefficient .so i shifted back by 16 bits,so > > > my result now has 24 bit,,,but this result goes to a echo cancellor > > > block which needs only 16 bits...!!!any suggestions and > > > recommendations are welcome... > > > One more question,,,is this FIR a Low pass or high pass filter ??? ( > > > my coefficients are symmetric )how do we know if its low pass or high > > > pass... > > > Thanks > > > jack > > > > A low pass filter will have lots of coefficients with low valuednumbers,> > like .00457 or 0.339. High pass filters have high valued coefficients, > like > > 300 or 129.224. The fact that your filter is symmetric means that it'sa> > bandpass filter, since the frequencies prefer to "roll down hill" toward > the > > center of the notch. > > You can't have a bandpass filter with real coeffs. A bandpass FIR filterin> the digital domain will necessarily have to have complex coeffs. > Follow Jim's response for high vs low. > > Cheers > Bhaskar >You guys crack me up - did you really take me seriously? My reply was no more ridiculous thatn the original questions though! James
Reply by ●December 8, 20032003-12-08
"Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message news:1sqdnXk0xro9V0miRVn-gQ@centurytel.net...> > "Bhaskar Thiagarajan" <bhaskart@deja.com> wrote in message > news:br2gmc$257eus$1@ID-82263.news.uni-berlin.de... > > > > > > You can't have a bandpass filter with real coeffs. A bandpass FIR filter > in > > the digital domain will necessarily have to have complex coeffs. > > Follow Jim's response for high vs low. > > Bhaskar, > > Surely you don't mean this!! > > FredNope...and thanks for setting me straight. I'm not sure what I was thinking. One of the reasons it helps to post (without thinking) - someone will correct you! and the embarressment will prevent any further posts (oh I mean of this nature :-)) Cheers Bhaskar
