Reply by Mike Yarwood December 7, 20052005-12-07
"Mark" <makolber@yahoo.com> wrote in message 
news:1133978287.696157.94570@f14g2000cwb.googlegroups.com...
> yep, the delay profile IS the impulse response... > > Mark >
so what is the average power delay profile? Best of Luck - Mike
Reply by Mark December 7, 20052005-12-07
yep, the delay profile IS the impulse response...

Mark

Reply by Frank Benett December 7, 20052005-12-07
Thanks for your suggestion. If one would use this model in a realistic way, 
several factor should be
considered (delay, doppler (fading))..etc). I solved the problem, as I 
reduced the complexity of my model:
I consider only the simplest static model which doesn't vary in time. For 
this I simply calculated the powers from the
given amplitude which is equivalent to the impulse response h, which i 
needed for my matlab simulation!
Frank

"Fred Marshall" <fmarshallx@remove_the_x.acm.org> az al&#4294967295;bbiakat &#4294967295;rta a 
k&#4294967295;vetkez&#4294967295; h&#4294967295;r&#4294967295;zenetben: d5KdnR2_O4rjjQrenZ2dnUVZ_tudnZ2d@centurytel.net...
> > "Frank Benett" <frank_b@gmail.com> wrote in message > news:dn6g0u$bsc$1@namru.matavnet.hu... >> Thanks for all of your answer and effort. I'm wondering why I can't find >> any information >> about this in the internet at all however, most of the channel models are >> given with in this form. >> With this power delay profile (PDP) the impulse responses are given with >> their delay and the >> corresponding amplitude. (i.e : the average received power).. I have >> looked the IT++ which >> was mentioned, it's a really good project and I find channel models >> there, I hope form this I can figure out the >> solution, but I'm still curious, if somebody could tell me the >> mathematics behind this... >> regards, >> Frank > > Frank, > > I'll go out on a limb because I don't generally work in rf comm's and this > technique is used in that area. Nonetheless, it seems clear enough what > the motivation is: > > There is a multipath environment which means that signal arrives at a > receiver perhaps line-of-sight and also by bounce paths and possibly ray > bending in transmission through a variety of materials. > > Now, one method for modeling multipath is to use a delay / attenuation > value for each path. When the attenuation values are lined up according > to delay time you have what looks like a FIR filter in continuous time. > i.e. the temporal resulution is infinite. As long as the wavelengths are > reasonably long relative to the path characteristics then this works fine. > As the wavelengths get shorter then the temporal location / delay involved > in any path begins to be difficult to determine relative to a wavelength. > Rough location is still known of course but the ability to resolve delay > into phase is lost. > > One method for looking at signals in a situation where phase is > indeterminate is to randomize the phase for each path and look at a large > sample of possible outcomes to yield system statistics. I've done this. > > But, this doesn't account for the situation where the reflectors in a path > are distrubuted and perhaps slowly time-varying. If reflectors are > distributed they can be modeled as a cluster of reflectors. With any > wavelength variation there will now be a distribution of apparent > reflector amplitudes as well as phases. With wideband signals it seems > one can only talk about average reflectivity and random phase. It's as > though *each* reflector could be treated as the distributed / stable > reflector situation above and some statistical measure would make sense. > > So, while a FIR model might still apply and the statistical method might > still be used, is that necessarily the most convenient method for analysis > purposes? Often it isn't. Looking at a curve might be better. > > Given that each path has an average strength then it makes some sense to > measure it in dB and to plot the resulting path / delay structure in dB vs > time. > This is no longer a FIR filter representation although it certainly has > some resemblance. > > Make sense? > > Fred > > >
Reply by Fred Marshall December 7, 20052005-12-07
"Frank Benett" <frank_b@gmail.com> wrote in message 
news:dn6g0u$bsc$1@namru.matavnet.hu...
> Thanks for all of your answer and effort. I'm wondering why I can't find > any information > about this in the internet at all however, most of the channel models are > given with in this form. > With this power delay profile (PDP) the impulse responses are given with > their delay and the > corresponding amplitude. (i.e : the average received power).. I have > looked the IT++ which > was mentioned, it's a really good project and I find channel models there, > I hope form this I can figure out the > solution, but I'm still curious, if somebody could tell me the mathematics > behind this... > regards, > Frank
Frank, I'll go out on a limb because I don't generally work in rf comm's and this technique is used in that area. Nonetheless, it seems clear enough what the motivation is: There is a multipath environment which means that signal arrives at a receiver perhaps line-of-sight and also by bounce paths and possibly ray bending in transmission through a variety of materials. Now, one method for modeling multipath is to use a delay / attenuation value for each path. When the attenuation values are lined up according to delay time you have what looks like a FIR filter in continuous time. i.e. the temporal resulution is infinite. As long as the wavelengths are reasonably long relative to the path characteristics then this works fine. As the wavelengths get shorter then the temporal location / delay involved in any path begins to be difficult to determine relative to a wavelength. Rough location is still known of course but the ability to resolve delay into phase is lost. One method for looking at signals in a situation where phase is indeterminate is to randomize the phase for each path and look at a large sample of possible outcomes to yield system statistics. I've done this. But, this doesn't account for the situation where the reflectors in a path are distrubuted and perhaps slowly time-varying. If reflectors are distributed they can be modeled as a cluster of reflectors. With any wavelength variation there will now be a distribution of apparent reflector amplitudes as well as phases. With wideband signals it seems one can only talk about average reflectivity and random phase. It's as though *each* reflector could be treated as the distributed / stable reflector situation above and some statistical measure would make sense. So, while a FIR model might still apply and the statistical method might still be used, is that necessarily the most convenient method for analysis purposes? Often it isn't. Looking at a curve might be better. Given that each path has an average strength then it makes some sense to measure it in dB and to plot the resulting path / delay structure in dB vs time. This is no longer a FIR filter representation although it certainly has some resemblance. Make sense? Fred
Reply by Frank Benett December 7, 20052005-12-07
Thanks for all of your answer and effort. I'm wondering why I can't find any 
information
about this in the internet at all however, most of the channel models are 
given with in this form.
With this power delay profile (PDP) the impulse responses are given with 
their delay and the
corresponding amplitude. (i.e : the average received power).. I have looked 
the IT++ which
was mentioned, it's a really good project and I find channel models there, I 
hope form this I can figure out the
solution, but I'm still curious, if somebody could tell me the mathematics 
behind this...
regards,
Frank


"Mark" <makolber@yahoo.com> az al&#4294967295;bbiakat &#4294967295;rta a k&#4294967295;vetkezo h&#4294967295;r&#4294967295;zenetben: 
1133908052.438971.210770@g43g2000cwa.googlegroups.com...
> > Frank Benett wrote: >> Thanks for your answer In power deleay profile I mean the plot of the tap >> delay (us) on the x axis, >> and the corresponding amplitude (dB) on the y axis.. >> All of the channel models (rural,hilly..etc) are given in this form, and >> I >> need to calculate the impulse response value (h). >> thank you, >> regard, >> Frank >> >> > > If the tap values you are talking about are in the DFE equalizer, then > I think (I'm not sure) the tap values are simply equal to the channel > impulse response (in dB of course) > > In other words, if the DFE tap at 5 us is activated at -10 dB, then > the channel has an echo delayed by 5 us that is -10 dB and this (along > with the main path) is the impulse response. > > (It's a little more compoicated for an FFE) > > No??? > > Mark >
Reply by Mike Yarwood December 6, 20052005-12-06
"Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote in message 
news:FMednWw5yuHuRgjeRVn-sA@centurytel.net...
> > "Mike Yarwood" <mpyarwood@btopenworld.com> wrote in message > news:dn4dbv$64o$1@nwrdmz03.dmz.ncs.ea.ibs-infra.bt.com... >> >>> >> Hi Frank - have you looked through the models and tutorials at >> http://itpp.sourceforge.net/3.8.0/index.html ? >> > > Mike, > > How should we interpret the claim that IT++ is "wildly used"? >
Hah yes! - english ain't their first language so it's a bit of a hoot in places but interesting for the different ITU and COST models they've released. Best of luck - Mike
Reply by Mark December 6, 20052005-12-06
this is the magnitude of the tap coefficient expressed in dB?  why
would
anyone do that?  to show exponential trend as linear?



for the same reasons we typically use dB for SNR etc....

to show a wide dynamic range on a single graph and because the value
relative to the main tap is important..i.e the relative level of the
reflection is important.....

they are typically shown  as dB relative to the main tap...



M

Reply by Mark December 6, 20052005-12-06
Frank Benett wrote:
> Thanks for your answer In power deleay profile I mean the plot of the tap > delay (us) on the x axis, > and the corresponding amplitude (dB) on the y axis.. > All of the channel models (rural,hilly..etc) are given in this form, and I > need to calculate the impulse response value (h). > thank you, > regard, > Frank > >
If the tap values you are talking about are in the DFE equalizer, then I think (I'm not sure) the tap values are simply equal to the channel impulse response (in dB of course) In other words, if the DFE tap at 5 us is activated at -10 dB, then the channel has an echo delayed by 5 us that is -10 dB and this (along with the main path) is the impulse response. (It's a little more compoicated for an FFE) No??? Mark
Reply by Fred Marshall December 6, 20052005-12-06
"Mike Yarwood" <mpyarwood@btopenworld.com> wrote in message 
news:dn4dbv$64o$1@nwrdmz03.dmz.ncs.ea.ibs-infra.bt.com...
> >> > Hi Frank - have you looked through the models and tutorials at > http://itpp.sourceforge.net/3.8.0/index.html ? >
Mike, How should we interpret the claim that IT++ is "wildly used"? Fred
Reply by Mike Yarwood December 6, 20052005-12-06
"robert bristow-johnson" <rbj@audioimagination.com> wrote in message 
news:BFBA750F.CA9D%rbj@audioimagination.com...
> in article dn3092$mcm$1@namru.matavnet.hu, Frank Benett at > frank_b@gmail.com > wrote on 12/05/2005 22:24: > >> Thanks for your answer In power deleay profile I mean the plot of the tap >> delay (us) on the x axis, >> and the corresponding amplitude (dB) on the y axis.. > > this is the magnitude of the tap coefficient expressed in dB? why would > anyone do that? to show exponential trend as linear? > >> All of the channel models (rural,hilly..etc) are given in this form, and >> I >> need to calculate the impulse response value (h). > > so, after throwing away the sign information of h[n], you want to recover > h[n] including the signs of the negative taps? can we assume that the > first > h[n] is > 0? if so, we might be able to infer zero crossings and toggle > the > sign every zero crossing. >
Hi Frank - have you looked through the models and tutorials at http://itpp.sourceforge.net/3.8.0/index.html ? Best of Luck - Mike