hi.. I wnat to design a fir decimation filter with 32 taps , input width 16, coefff width 16 .plz give me suggestions to implement this.
fir decimation filters
Started by ●November 23, 2005
Reply by ●November 23, 20052005-11-23
"saras" <saras_rajgiri@yahoo.co.in> wrote in message news:1132724111.894818.212270@z14g2000cwz.googlegroups.com...> hi.. > > I wnat to design a fir decimation filter with 32 taps , input width 16, > coefff width 16 .plz give me suggestions to implement this.Implement this on what? What are your filter and decimation requirements? Cheers Bhaskar
Reply by ●November 23, 20052005-11-23
Bhaskar Thiagarajan wrote:> "saras" <saras_rajgiri@yahoo.co.in> wrote in message > news:1132724111.894818.212270@z14g2000cwz.googlegroups.com... > >>hi.. >> >>I wnat to design a fir decimation filter with 32 taps , input width 16, >>coefff width 16 .plz give me suggestions to implement this. > > > Implement this on what? > What are your filter and decimation requirements? > > Cheers > BhaskarOnce, as a summer intern, I was asked to make a breadboarding mock-up. I made a few chassis in different standard sizes out of expanded material (like flattened wire lath) instead of solid steel, so one could see through them. I took the iron cores out of a few sizes of transformers and replaced them with light spacers, then fastened magnets to them. I sawed the pins of tubes of various sizes and embedded magnets in the bases. So on for other parts: resistors in different wattages, capacitors in different types, "birdies" (tie points), potentiometers without shafts, knobs, and more. I went to the stockroom with a list of the items I wanted, identified by size but not value. A asked the clerk to give me the least expensive parts that met my needs. He said that he needed specific part numbers, so I asked him for a price list that would let me choose wisely. At that point, he lost his cool. "What are you building anyway? How can you make something work without the right components?" His sneer bugged me, and I answered, "If the circuit is really good, the actual component values don't matter." We must have had an audience. My mock-up set was both praised and well used, but people all around the plant quoted "If the circuit is really good, the actual components don't matter" at every excuse. If the decimation filter is designed right, the actual requirements don't matter. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●November 23, 20052005-11-23
"Jerry Avins" <jya@ieee.org> wrote in message news:UIednXSX7KDTNBneRVn-qw@rcn.net...> Bhaskar Thiagarajan wrote: > > "saras" <saras_rajgiri@yahoo.co.in> wrote in message > > news:1132724111.894818.212270@z14g2000cwz.googlegroups.com... > > > >>hi.. > >> > >>I wnat to design a fir decimation filter with 32 taps , input width 16, > >>coefff width 16 .plz give me suggestions to implement this. > > > > > > Implement this on what? > > What are your filter and decimation requirements? > > > > Cheers > > Bhaskar > > Once, as a summer intern, I was asked to make a breadboarding mock-up. I > made a few chassis in different standard sizes out of expanded material > (like flattened wire lath) instead of solid steel, so one could see > through them. I took the iron cores out of a few sizes of transformers > and replaced them with light spacers, then fastened magnets to them. I > sawed the pins of tubes of various sizes and embedded magnets in the > bases. So on for other parts: resistors in different wattages, > capacitors in different types, "birdies" (tie points), potentiometers > without shafts, knobs, and more. > > I went to the stockroom with a list of the items I wanted, identified by > size but not value. A asked the clerk to give me the least expensive > parts that met my needs. He said that he needed specific part numbers, > so I asked him for a price list that would let me choose wisely. > > At that point, he lost his cool. "What are you building anyway? How can > you make something work without the right components?" His sneer bugged > me, and I answered, "If the circuit is really good, the actual component > values don't matter." We must have had an audience. My mock-up set was > both praised and well used, but people all around the plant quoted "If > the circuit is really good, the actual components don't matter" at every > excuse. > > If the decimation filter is designed right, the actual requirements > don't matter.I like your story and punch line !! If I don't know what is 'right' I can't design it. I'm sure you get it but I want the OP to get it too. At a minimum one needs to know what the decimation ratio needs to be before designing the filter...I'll agree with your final statement once this is known.> > Jerry > -- > Engineering is the art of making what you want from things you can get. > �����������������������������������������������������������������������
Reply by ●November 24, 20052005-11-24
hi all.. thanks for reply!! now, the specifications are decimation ratio =2, taps=32, input and coeff width =16. I want an idea about the filter stuctures to be used ,so that the no of multiplications can be reduced.
Reply by ●November 25, 20052005-11-25
Reply by ●November 26, 20052005-11-26
saras wrote:> now, the specifications are decimation ratio =2, taps=32, input and > coeff width =16.A minimax halfband filter might be a good compromise. Our very own Fred Marshall has written a free program that computes such coefficients: ftp://ftp.mission-systems-inc.com/outgoing/Halfband/ I'm not sure whether Meteor can beat this, you might want to have a try as well: http://www.cs.princeton.edu/~ken/meteor.html As far as I remember, at an equal number of taps you can achieve even better magnitude performance when you drop the linear-phase requirement, but I don't know any programs to design such FIRs. Anybody? Other trade-offs in your design are bandwidth versus stopband attenuation. Since you haven't posted any requirements or optimality criteria, it's futile to be more specific. Regards, Andor
Reply by ●November 26, 20052005-11-26
saras wrote:> now, the specifications are decimation ratio =2, taps=32, input and > coeff width =16.A minimax halfband filter might be a good compromise. Our very own Fred Marshall has written a free program that computes such coefficients: ftp://ftp.mission-systems-inc.com/outgoing/Halfband/ I'm not sure whether Meteor can beat this, you might want to have a try as well: http://www.cs.princeton.edu/~ken/meteor.html As far as I remember, at an equal number of taps you can achieve even better magnitude performance when you drop the linear-phase requirement, but I don't know any programs to design such FIRs. Anybody? Other trade-offs in your design are bandwidth versus stopband attenuation. Since you haven't posted any requirements or optimality criteria, it's futile to be more specific. Regards, Andor
