hi.. Can anybody tell me the structure of symmetric transpose form fir filter? Please Help..
symmetric transpose form fir filter
Started by ●November 30, 2005
Reply by ●November 30, 20052005-11-30
saras wrote:> hi.. > Can anybody tell me the structure of symmetric transpose form fir > filter? Please Help..please, tell me in what context do you hear this?
Reply by ●December 1, 20052005-12-01
hi.. I know that by exploiting the symmetry of coeffs , the number of multipliers can be reduced to half. But what I saw in many books is the direct form structure(linear phase filters).So I want to know symmetric transpose form fir filter if such structure exists. Nick Molkov wrote:> saras wrote: > > hi.. > > Can anybody tell me the structure of symmetric transpose form fir > > filter? Please Help.. > > please, tell me in what context do you hear this?
Reply by ●December 1, 20052005-12-01
saras wrote:> hi.. > I know that by exploiting the symmetry of coeffs , the number of > multipliers can be reduced to half. But what I saw in many books is the > direct form structure(linear phase filters).So I want to know > symmetric transpose form fir filter if such structure exists.On most DSPs, single cycle MACs are standard, ie. you don't save anything when you cut down the number of multiplications in an FIR. If you still need it, look at this examle: Filter coefficients h = [h1 h2 h1] (symmetric linear phase FIR with three coefficients). Typically y(n) = h1 x(n) + h2 x(n-1) + h1 x(n-2). To save multiplications: y(n) = h1 (x(n) + x(n-2) ) + h2 x(n-1). You can hopefully generalize that yourself to arbitrary (odd and even) length symmetric FIRs. Regards, Andor
Reply by ●December 1, 20052005-12-01
I can't imagine, how could it be done. When you design FIR filter, you really get the symmetry coeffs. But the only one way, that i know to decrease the mathematical operations - is to use the algorithm of fast convolution, based on fft.
Reply by ●December 1, 20052005-12-01
abariska@student.ethz.ch wrote:> saras wrote: > > hi.. > > I know that by exploiting the symmetry of coeffs , the number of > > multipliers can be reduced to half. But what I saw in many books is the > > direct form structure(linear phase filters).So I want to know > > symmetric transpose form fir filter if such structure exists. > > On most DSPs, single cycle MACs are standard, ie. you don't save > anything when you cut down the number of multiplications in an FIR.Although a multiply or MAC takes the same number of cycles as an add operation on a modern DSP, the multiplier toggles a lot more gates and wires than the adder; so there still might be a power savings in using adds over multiplies in FIR intensive operation. And power is king these days, especially in the design of battery operated products. IMHO. YMMV. -- rhn A.T nicholson d.O.t C-o-M
Reply by ●December 1, 20052005-12-01
Ron N. wrote:> > Although a multiply or MAC takes the same number of cycles as an add > operation on a modern DSP, the multiplier toggles a lot more gates > and wires than the adder; so there still might be a power savings > in using adds over multiplies in FIR intensive operation. And power > is king these days, especially in the design of battery operated > products. > > > IMHO. YMMV.Ron, you are quite rignt. During the execution of MAC instructions the DSP power consumption is several times higher then when executing a general program flow. This is the fact; I observed it with ADSP-21xx. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
Reply by ●December 5, 20052005-12-05
On Thu, 01 Dec 2005 21:15:44 GMT, Vladimir Vassilevsky <antispam_bogus@hotmail.com> wrote:> > >Ron N. wrote: > > >> >> Although a multiply or MAC takes the same number of cycles as an add >> operation on a modern DSP, the multiplier toggles a lot more gates >> and wires than the adder; so there still might be a power savings >> in using adds over multiplies in FIR intensive operation. And power >> is king these days, especially in the design of battery operated >> products. >> >> >> IMHO. YMMV. > >Ron, you are quite rignt. During the execution of MAC instructions the >DSP power consumption is several times higher then when executing a >general program flow. This is the fact; I observed it with ADSP-21xx. > >Vladimir Vassilevsky >Hi Vladimir, Now that is interesting,... very interesting. Thanks, [-Rick-]
Reply by ●December 5, 20052005-12-05
Rick Lyons wrote:>On Thu, 01 Dec 2005 21:15:44 GMT, Vladimir Vassilevsky ><antispam_bogus@hotmail.com> wrote: > > > >>Ron N. wrote: >> >> >> >> >>>Although a multiply or MAC takes the same number of cycles as an add >>>operation on a modern DSP, the multiplier toggles a lot more gates >>>and wires than the adder; so there still might be a power savings >>>in using adds over multiplies in FIR intensive operation. And power >>>is king these days, especially in the design of battery operated >>>products. >>> >>> >>>IMHO. YMMV. >>> >>> >>Ron, you are quite rignt. During the execution of MAC instructions the >>DSP power consumption is several times higher then when executing a >>general program flow. This is the fact; I observed it with ADSP-21xx. >> >>Vladimir Vassilevsky >> >> >> > >Hi Vladimir, > > Now that is interesting,... very interesting. > >Thanks, >[-Rick-] > >Not only interesting. It should be fairly obvious. Its interesting to spot the people in this group that never work on portable equipment. Lines like "MACs are really fast these days, so we can use an FIR instead of an IIR" are a dead giveaway. :-) The software in portable devices very much affects their battery life, right down to choosing the least complex instructions. Designing software for the absolute minimum power consumption is an interesting art. Regards, Steve
Reply by ●December 5, 20052005-12-05
Steve Underwood <steveu@dis.org> writes:> Designing software for the > absolute minimum power consumption is an interesting art.It's also interesting that many IT schools are teaching so abstractly about software (i.e. divorcing it from the hardware), that the graduating students won't have a snowball's chance in hell of figuring out how to do this. Ciao, Peter K.
