Hi All I did a simple simulation for adaptive DFE to replicate the results of proakis. It gives me proper results. But this simulation is just a normal one, no pulse shaping, no carrier modulation, only one sample per symbol. To add the channel effects, I am convolving my tx signal with channel. But when I tried to do the same simulation at passband with 12 samples per symbol, I am not getting the same results. Now here I am not convolving the signal with channel, but I am adding proper delay for each path and adding them together, which is more or less same. I am not able to figure it out why I am getting stange results. Thanking you. Regards, Chintan
Channel Modelling
Started by ●August 21, 2008
Reply by ●August 21, 20082008-08-21
On Aug 21, 4:35 am, "cpshah99" <cpsha...@rediffmail.com> wrote:> Hi All > > I did a simple simulation for adaptive DFE to replicate the results of > proakis. It gives me proper results. But this simulation is just a normal > one, no pulse shaping, no carrier modulation, only one sample per symbol. > To add the channel effects, I am convolving my tx signal with channel. > > But when I tried to do the same simulation at passband with 12 samples per > symbol, I am not getting the same results. Now here I am not convolving the > signal with channel, but I am adding proper delay for each path and adding > them together, which is more or less same. > > I am not able to figure it out why I am getting stange results. > > Thanking you. > > Regards, > > ChintanHave you compared the "effective channel" between scenario 1 (one sample per symbol in baseband) versus the passband case? Often the issue is that the "effective channel" seen by the baseband portion in a passband simulation is complex-valued. Julius
Reply by ●August 21, 20082008-08-21
>On Aug 21, 4:35 am, "cpshah99" <cpsha...@rediffmail.com> wrote: >> Hi All >> >> I did a simple simulation for adaptive DFE to replicate the results of >> proakis. It gives me proper results. But this simulation is just anormal>> one, no pulse shaping, no carrier modulation, only one sample persymbol.>> To add the channel effects, I am convolving my tx signal with channel. >> >> But when I tried to do the same simulation at passband with 12 samplesper>> symbol, I am not getting the same results. Now here I am not convolvingthe>> signal with channel, but I am adding proper delay for each path andadding>> them together, which is more or less same. >> >> I am not able to figure it out why I am getting stange results. >> >> Thanking you. >> >> Regards, >> >> Chintan > >Have you compared the "effective channel" between scenario 1 (one >sample >per symbol in baseband) versus the passband case? > >Often the issue is that the "effective channel" seen by the baseband >portion >in a passband simulation is complex-valued. > >Julius >%%%% Hi Julius Thanks a lot for your reply. As u said, I checked the impluse response, after down converting my received signal at baseband. The IR is complex. Where as the channel in scenario 1 is real i.e. channel=[0.407 0.815 0.407]. Now the way I am doing in passband is: ds=signal to be transmitted after carrier modulation Ns=Number of samples per symbol path1=[ds zeros(1,Ns*2)]; % Direct path path2=[zeros(1,Ns*1) ds zeros(1,Ns*1)]; % Reflected path path3=[zeros(1,Ns*2) ds]; % Reflected path ds1=0.407*path1+0.815*path2+0.407*path3; Is this correct? Thanks again. Chintan
Reply by ●August 21, 20082008-08-21
On Aug 21, 10:17 am, "cpshah99" <cpsha...@rediffmail.com> wrote:> >On Aug 21, 4:35 am, "cpshah99" <cpsha...@rediffmail.com> wrote: > >> Hi All > > >> I did a simple simulation for adaptive DFE to replicate the results of > >> proakis. It gives me proper results. But this simulation is just a > normal > >> one, no pulse shaping, no carrier modulation, only one sample per > symbol. > >> To add the channel effects, I am convolving my tx signal with channel. > > >> But when I tried to do the same simulation at passband with 12 samples > per > >> symbol, I am not getting the same results. Now here I am not convolving > the > >> signal with channel, but I am adding proper delay for each path and > adding > >> them together, which is more or less same. > > >> I am not able to figure it out why I am getting stange results. > > >> Thanking you. > > >> Regards, > > >> Chintan > > >Have you compared the "effective channel" between scenario 1 (one > >sample > >per symbol in baseband) versus the passband case? > > >Often the issue is that the "effective channel" seen by the baseband > >portion > >in a passband simulation is complex-valued. > > >Julius > > %%%% > > Hi Julius > > Thanks a lot for your reply. > > As u said, I checked the impluse response, after down converting my > received signal at baseband. The IR is complex. Where as the channel in > scenario 1 is real i.e. channel=[0.407 0.815 0.407]. > > Now the way I am doing in passband is: > > ds=signal to be transmitted after carrier modulation > Ns=Number of samples per symbol > > path1=[ds zeros(1,Ns*2)]; % Direct path > path2=[zeros(1,Ns*1) ds zeros(1,Ns*1)]; % Reflected path > path3=[zeros(1,Ns*2) ds]; % Reflected path > > ds1=0.407*path1+0.815*path2+0.407*path3; > > Is this correct? > > Thanks again. > > ChintanIt depends on what you mean by "correct" :-). In general, a multipath channel in passband will give you a complex "effective" baseband impulse response, except in special cases. This is due to the carrier modulating the effective impulse response. So maybe your baseband model taken from Proakis is a little too restrictive? Julius
Reply by ●August 21, 20082008-08-21
>On Aug 21, 10:17 am, "cpshah99" <cpsha...@rediffmail.com> wrote: >> >On Aug 21, 4:35 am, "cpshah99" <cpsha...@rediffmail.com> wrote: >> >> Hi All >> >> >> I did a simple simulation for adaptive DFE to replicate the resultsof>> >> proakis. It gives me proper results. But this simulation is just a >> normal >> >> one, no pulse shaping, no carrier modulation, only one sample per >> symbol. >> >> To add the channel effects, I am convolving my tx signal withchannel.>> >> >> But when I tried to do the same simulation at passband with 12samples>> per >> >> symbol, I am not getting the same results. Now here I am notconvolving>> the >> >> signal with channel, but I am adding proper delay for each path and >> adding >> >> them together, which is more or less same. >> >> >> I am not able to figure it out why I am getting stange results. >> >> >> Thanking you. >> >> >> Regards, >> >> >> Chintan >> >> >Have you compared the "effective channel" between scenario 1 (one >> >sample >> >per symbol in baseband) versus the passband case? >> >> >Often the issue is that the "effective channel" seen by the baseband >> >portion >> >in a passband simulation is complex-valued. >> >> >Julius >> >> %%%% >> >> Hi Julius >> >> Thanks a lot for your reply. >> >> As u said, I checked the impluse response, after down converting my >> received signal at baseband. The IR is complex. Where as the channelin>> scenario 1 is real i.e. channel=[0.407 0.815 0.407]. >> >> Now the way I am doing in passband is: >> >> ds=signal to be transmitted after carrier modulation >> Ns=Number of samples per symbol >> >> path1=[ds zeros(1,Ns*2)]; % Direct path >> path2=[zeros(1,Ns*1) ds zeros(1,Ns*1)]; % Reflected path >> path3=[zeros(1,Ns*2) ds]; % Reflected path >> >> ds1=0.407*path1+0.815*path2+0.407*path3; >> >> Is this correct? >> >> Thanks again. >> >> Chintan > >It depends on what you mean by "correct" :-). > >In general, a multipath channel in passband will give you a complex >"effective" baseband impulse response, except in special cases. >This is due to the carrier modulating the effective impulse response. > >So maybe your baseband model taken from Proakis is a little too >restrictive? > >Julius >%%%%% Hi Julius Thanks again. I think it is correct :-) So ultimately a lot of things can change depending on the scheme i.e. baseband or passband. I have another doubt, please throw some light onto this. In my normal simulation, I used QPSK and adaptive equaliser. But when I used 16 QAM, it is failing. Any hint? Thanks a lot. Chintan
Reply by ●August 21, 20082008-08-21
On Aug 21, 1:32 pm, "cpshah99" <cpsha...@rediffmail.com> wrote:> > Hi Julius > > Thanks again. I think it is correct :-) > > So ultimately a lot of things can change depending on the scheme i.e. > baseband or passband. > > I have another doubt, please throw some light onto this. > > In my normal simulation, I used QPSK and adaptive equaliser. But when I > used 16 QAM, it is failing. Any hint? > > Thanks a lot. > > ChintanCheck the "effective channel" after you use your equalizer and see how much ISI is there. Plot the channel's power density profile. Maybe there's too much ISI for 16 QAM? Hard to say for sure since there are a lot of factors to consider .... Julius
Reply by ●August 21, 20082008-08-21
>On Aug 21, 1:32 pm, "cpshah99" <cpsha...@rediffmail.com> wrote: > >> >> Hi Julius >> >> Thanks again. I think it is correct :-) >> >> So ultimately a lot of things can change depending on the scheme i.e. >> baseband or passband. >> >> I have another doubt, please throw some light onto this. >> >> In my normal simulation, I used QPSK and adaptive equaliser. But whenI>> used 16 QAM, it is failing. Any hint? >> >> Thanks a lot. >> >> Chintan > >Check the "effective channel" after you use your equalizer and see >how much ISI is there. Plot the channel's power density profile. >Maybe there's too much ISI for 16 QAM? Hard to say for sure since >there are a lot of factors to consider .... > >Julius >%%%% HI Julius Thanks again. I used the same channel for QPSK and 16 QAM. Now, in the case of 16 QAM, just to see if my receiver right or not, I removed the noise and instead of channel I juts made it awgn. It worked nice, no errors. And also after adding channel and noise, I checked the raw scatter plot after the equalisation. It was very scary.... And can u please tell the factors to be considered. Thanks again. Chintan
Reply by ●August 21, 20082008-08-21
On Aug 21, 1:52 pm, "cpshah99" <cpsha...@rediffmail.com> wrote:> HI Julius > > Thanks again. > > I used the same channel for QPSK and 16 QAM. > > Now, in the case of 16 QAM, just to see if my receiver right or not, I > removed the noise and instead of channel I juts made it awgn. It worked > nice, no errors. > > And also after adding channel and noise, I checked the raw scatter plot > after the equalisation. It was very scary.... > > And can u please tell the factors to be considered. > > Thanks again. > > ChintanUm. How about just the channel and no noise?
Reply by ●August 22, 20082008-08-22
>On Aug 21, 1:52 pm, "cpshah99" <cpsha...@rediffmail.com> wrote: > >> HI Julius >> >> Thanks again. >> >> I used the same channel for QPSK and 16 QAM. >> >> Now, in the case of 16 QAM, just to see if my receiver right or not, I >> removed the noise and instead of channel I juts made it awgn. Itworked>> nice, no errors. >> >> And also after adding channel and noise, I checked the raw scatterplot>> after the equalisation. It was very scary.... >> >> And can u please tell the factors to be considered. >> >> Thanks again. >> >> Chintan > >Um. How about just the channel and no noise? >Hi No improvement....
Reply by ●August 22, 20082008-08-22
On Aug 22, 5:57�am, "cpshah99" <cpsha...@rediffmail.com> wrote:> >On Aug 21, 1:52 pm, "cpshah99" <cpsha...@rediffmail.com> wrote: > > >> HI Julius > > >> Thanks again. > > >> I used the same channel for QPSK and 16 QAM. > > >> Now, in the case of 16 QAM, just to see if my receiver right or not, I > >> removed the noise and instead of channel I juts made it awgn. It > worked > >> nice, no errors. > > >> And also after adding channel and noise, I checked the raw scatter > plot > >> after the equalisation. It was very scary.... > > >> And can u please tell the factors to be considered. > > >> Thanks again. > > >> Chintan > > >Um. �How about just the channel and no noise? > > Hi > > No improvement....Chintan, what do you mean "no improvement". Nobody can help you unless they can understand what you are saying.
