Forums

OFDM Multipath Time-Varying Channel

Started by fmtzbalt September 11, 2006
Hi there.
I have a question regarding the wireless channel model to be used on an
OFDM system simulation. I have investigated for a while and most of the
people use an ARMA Filter to simulate the time variations of the wireless
channel, but this approach is for modeling a multipath channel. The thing
is that I don't know if this model includes the distortion on the receiver
do to the echoes. OFDM is "inmune" to these echoes, so what I need is a
channel, obviously multipath.. but that doesn't take into account this
echo distortion, but only the somehow random fading do to scattering,
refraction, etc. But again.. not the echoes. 

Rayleigh fading process can do this, by only considering one path, but how
can I simulate the time variations of the channel, because I can't just
generate a new rayleigh channel, the channel impulse responses(CIR) on
each time slot have to be related and evolve accordingly in time. I hope I
stated my doubt clearly enough. I would appretiate any kind of help. 

Thank You.
-fmtzbalt-


On Mon, 11 Sep 2006 20:18:00 -0500, "fmtzbalt" <fmtzbalt@gmail.com>
wrote:

>Hi there. >I have a question regarding the wireless channel model to be used on an >OFDM system simulation. I have investigated for a while and most of the >people use an ARMA Filter to simulate the time variations of the wireless >channel, but this approach is for modeling a multipath channel. The thing >is that I don't know if this model includes the distortion on the receiver >do to the echoes. OFDM is "inmune" to these echoes, so what I need is a >channel, obviously multipath.. but that doesn't take into account this >echo distortion, but only the somehow random fading do to scattering, >refraction, etc. But again.. not the echoes.
The taps represent the delays due to long paths, i.e., scattering form reflecting objects. Is this what you mean by "echoes"? If so, the taps, usually arranged in a decaying model, represent "echoes" for longer and longer paths as the tap index increases. Since the longer delay paths are longer propagation paths, the power is assumed to decay as well (on average).
>Rayleigh fading process can do this, by only considering one path, but how >can I simulate the time variations of the channel, because I can't just >generate a new rayleigh channel, the channel impulse responses(CIR) on >each time slot have to be related and evolve accordingly in time. I hope I >stated my doubt clearly enough. I would appretiate any kind of help. > >Thank You. >-fmtzbalt-
Maybe what you are thinking is that each tap can have a Ralyeigh statistic, which is true, but how do all the taps relate? In other words, how do you generate a new channel instance given the previous one? This can be handled in different ways, but depending on the system that you are simulating, if it is packet-oriented then often what is done is generate a different, completely random and uncorrelated, channel instance for each packet. If time coherence is not important in your simulation, only packet statistics, then there's really no reason to not do it this way. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
> >Hi there. > >I have a question regarding the wireless channel model to be used on an > >OFDM system simulation. I have investigated for a while and most of the > >people use an ARMA Filter to simulate the time variations of the wireless > >channel, but this approach is for modeling a multipath channel. The thing > >is that I don't know if this model includes the distortion on the receiver > >do to the echoes. OFDM is "inmune" to these echoes, so what I need is a > >channel, obviously multipath.. but that doesn't take into account this > >echo distortion, but only the somehow random fading do to scattering, > >refraction, etc. But again.. not the echoes. > > >Rayleigh fading process can do this, by only considering one path, but how > >can I simulate the time variations of the channel, because I can't just > >generate a new rayleigh channel, the channel impulse responses(CIR) on > >each time slot have to be related and evolve accordingly in time. I hope I > >stated my doubt clearly enough. I would appretiate any kind of help. > > > >Thank You. > >-fmtzbalt- > > Maybe what you are thinking is that each tap can have a Ralyeigh > statistic, which is true, but how do all the taps relate? In other > words, how do you generate a new channel instance given the previous > one? > > This can be handled in different ways, but depending on the system > that you are simulating, if it is packet-oriented then often what is > done is generate a different, completely random and uncorrelated, > channel instance for each packet. If time coherence is not important > in your simulation, only packet statistics, then there's really no > reason to not do it this way. >
On the other hand, if you're interested in doing a bit error simulation at the sample or symbol level (vs. the packet level) the correlation between different samples will probably be important to you. Assuming you want to simulate flat fading (no delay spread, no "echos"), you could just generate a single random variable (e.g. white gausssian) and pass it through a filter. The spectral shape of the filter will affect the time coherence - for example the classical two dimensional model has a U shaped spectrum and bessel function autocorrelation. There are approximations for this power spectrum to make simulations run fast that use a sum of sinusoids, e.g do a google search for "jakes power spectrum". Tim