Hi All I have implemented Adaptive Decision Feedback Equaliser using LMS Algorithm. Now I am using 4-QAM Modulation scheme. In the training mode, step size for feedforward filter is 0.01 and for feedback filter is 0.003. And in decision directed mode the step size for feedforward filter is 0.001 and for feedback filter is 0.003. I have got these values after playing thru different values. Now if I change the feedforward filter step size from 0.01 to 0.001 in training period and rest of the setup is same then I am getting big difference in terms of BER performance. So, is there any perticular method to select step size? Thanks. Chintan P. Shah
step size selection for LMS algorithm
Started by ●May 8, 2008
Reply by ●May 8, 20082008-05-08
cpshah99 wrote:> Hi All > > I have implemented Adaptive Decision Feedback Equaliser using LMS > Algorithm. Now I am using 4-QAM Modulation scheme. > > In the training mode, step size for feedforward filter is 0.01 and for > feedback filter is 0.003. And in decision directed mode the step size for > feedforward filter is 0.001 and for feedback filter is 0.003. I have got > these values after playing thru different values. > > Now if I change the feedforward filter step size from 0.01 to 0.001 in > training period and rest of the setup is same then I am getting big > difference in terms of BER performance. > > So, is there any perticular method to select step size?If you make it too big, the worst case noise will bounce the filter adaption all over the place. If you make it too small, the filter won't pull in adequately within the alloted time. Finding if a workable compromise is possible, and optimising it, is what the designer gets paid for. Steve
Reply by ●May 8, 20082008-05-08
>cpshah99 wrote: >> Hi All >> >> I have implemented Adaptive Decision Feedback Equaliser using LMS >> Algorithm. Now I am using 4-QAM Modulation scheme. >> >> In the training mode, step size for feedforward filter is 0.01 and for >> feedback filter is 0.003. And in decision directed mode the step sizefor>> feedforward filter is 0.001 and for feedback filter is 0.003. I havegot>> these values after playing thru different values. >> >> Now if I change the feedforward filter step size from 0.01 to 0.001 in >> training period and rest of the setup is same then I am getting big >> difference in terms of BER performance. >> >> So, is there any perticular method to select step size? > >If you make it too big, the worst case noise will bounce the filter >adaption all over the place. If you make it too small, the filter won't >pull in adequately within the alloted time. Finding if a workable >compromise is possible, and optimising it, is what the designer gets >paid for. > >Steve >%%%%%%%%% Hi Thanks Steve for your reply. But this is very old topic and there must be something to optimise the step size. As in real time system we just decide these parameters for once and the system is left to run. Any direction, if not direct answer, where I can find something. Thanks Chintan P Shah
Reply by ●May 8, 20082008-05-08
>Thanks Steve for your reply. But this is very old topic and there must be >something to optimise the step size. As in real time system we justdecide>these parameters for once and the system is left to run. > >Any direction, if not direct answer, where I can find something.I'm in a similar boat with some adaptive work I've been doing and the answer I got: it is possible to derive bounds, but "best" is likely situationally dependent (perhaps based on input variance, or similar?). I will be doing the research over the next couple months for a very different algorithm, but similar overall problem (and I actually need to balance two step parameters). If I find anything, I'll be sure to post. Mark
Reply by ●May 8, 20082008-05-08
On May 8, 1:59�pm, "markt" <tak...@pericle.com> wrote:> >Thanks Steve for your reply. But this is very old topic and there must be > >something to optimise the step size. As in real time system we just > decide > >these parameters for once and the system is left to run. > > >Any direction, if not direct answer, where I can find something. > > I'm in a similar boat with some adaptive work I've been doing and the > answer I got: it is possible to derive bounds, but "best" is likely > situationally dependent (perhaps based on input variance, or similar?). �I > will be doing the research over the next couple months for a very different > algorithm, but similar overall problem (and I actually need to balance two > step parameters). �If I find anything, I'll be sure to post. > > Markyou can also consider a bigger step size during aquisition and a slower one during tracking... (a different) Mark
Reply by ●May 8, 20082008-05-08
markt wrote:>> Thanks Steve for your reply. But this is very old topic and there must be >> something to optimise the step size. As in real time system we just > decide >> these parameters for once and the system is left to run. >> >> Any direction, if not direct answer, where I can find something. > > I'm in a similar boat with some adaptive work I've been doing and the > answer I got: it is possible to derive bounds, but "best" is likely > situationally dependent (perhaps based on input variance, or similar?). I > will be doing the research over the next couple months for a very different > algorithm, but similar overall problem (and I actually need to balance two > step parameters). If I find anything, I'll be sure to post.All of which again brings home forcefully that engineering is an art. Hitting on the best use of science is part of that art. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●May 8, 20082008-05-08
>On May 8, 1:59=A0pm, "markt" <tak...@pericle.com> wrote: >> >Thanks Steve for your reply. But this is very old topic and there mustbe=> >> >something to optimise the step size. As in real time system we just >> decide >> >these parameters for once and the system is left to run. >> >> >Any direction, if not direct answer, where I can find something. >> >> I'm in a similar boat with some adaptive work I've been doing and the >> answer I got: it is possible to derive bounds, but "best" is likely >> situationally dependent (perhaps based on input variance, or similar?).=>=A0I >> will be doing the research over the next couple months for a verydifferen=>t >> algorithm, but similar overall problem (and I actually need to balancetwo=> >> step parameters). =A0If I find anything, I'll be sure to post. >> >> Mark > >you can also consider a bigger step size during aquisition and a >slower one during tracking... > >(a different) >Mark >%%%%%%% HI Mark I am doing the same thing as you said. But if I change the channel then these setting fails to give good results. Chintan
Reply by ●May 8, 20082008-05-08
>All of which again brings home forcefully that engineering is an art. >Hitting on the best use of science is part of that art. > >JerryAn ugly art that creates heartburn, sometimes. ;) (the first) Mark
Reply by ●May 8, 20082008-05-08
On May 8, 3:10�pm, "cpshah99" <cpsha...@rediffmail.com> wrote:> >On May 8, 1:59=A0pm, "markt" <tak...@pericle.com> wrote: > >> >Thanks Steve for your reply. But this is very old topic and there must > be= > > >> >something to optimise the step size. As in real time system we just > >> decide > >> >these parameters for once and the system is left to run. > > >> >Any direction, if not direct answer, where I can find something. > > >> I'm in a similar boat with some adaptive work I've been doing and the > >> answer I got: it is possible to derive bounds, but "best" is likely > >> situationally dependent (perhaps based on input variance, or similar?). > = > >=A0I > >> will be doing the research over the next couple months for a very > differen= > >t > >> algorithm, but similar overall problem (and I actually need to balance > two= > > >> step parameters). =A0If I find anything, I'll be sure to post. > > >> Mark > > >you can also consider a bigger step size during aquisition and a > >slower one during tracking... > > >(a different) > >Mark > > %%%%%%% > > HI Mark > > I am doing the same thing as you said. But if I change the channel then > these setting fails to give good results. > > Chintan- Hide quoted text - > > - Show quoted text -when you change channel, you have to re-aquire...and go back to high speed larger step. right? What kinkd of demod are you working on? Mark
Reply by ●May 9, 20082008-05-09
>On May 8, 3:10=A0pm, "cpshah99" <cpsha...@rediffmail.com> wrote: >> >On May 8, 1:59=3DA0pm, "markt" <tak...@pericle.com> wrote: >> >> >Thanks Steve for your reply. But this is very old topic and theremust=> >> be=3D >> >> >> >something to optimise the step size. As in real time system wejust>> >> decide >> >> >these parameters for once and the system is left to run. >> >> >> >Any direction, if not direct answer, where I can find something. >> >> >> I'm in a similar boat with some adaptive work I've been doing andthe>> >> answer I got: it is possible to derive bounds, but "best" is likely >> >> situationally dependent (perhaps based on input variance, orsimilar?).=> >> =3D >> >=3DA0I >> >> will be doing the research over the next couple months for a very >> differen=3D >> >t >> >> algorithm, but similar overall problem (and I actually need tobalance>> two=3D >> >> >> step parameters). =3DA0If I find anything, I'll be sure to post. >> >> >> Mark >> >> >you can also consider a bigger step size during aquisition and a >> >slower one during tracking... >> >> >(a different) >> >Mark >> >> %%%%%%% >> >> HI Mark >> >> I am doing the same thing as you said. But if I change the channelthen>> these setting fails to give good results. >> >> Chintan- Hide quoted text - >> >> - Show quoted text - > >when you change channel, you have to re-aquire...and go back to high >speed larger step. > >right? > >What kinkd of demod are you working on? > > >Mark >%%%% Hi Mark I am changing channel not during transmission...if u know there are three channels in the book by proakis.....i am using channel B for now and the above values of step sizes are for this channel B..but if i use channel C the this LMS fails to converge......even for high SNR values..... And I am using soft demodulation........