Does multipath effect average out in time ?

On May 15, 9:13&#4294967295;am, "cpshah99" <cpsha...@rediffmail.com> wrote:
> >On May 15, 8:10 am, "cpshah99" <cpsha...@rediffmail.com> wrote:
> >> >On May 15, 4:28=A0am, "cpshah99" <cpsha...@rediffmail.com> wrote:
> >> >> >On May 15, 9:12=3DA0am, Rune Allnor <all...@tele.ntnu.no> wrote:
> >> >> >> On 14 Mai, 10:24, Dani Camps <danicamp...@gmail.com> wrote:
>
> >> >> >> > Dear all,
>
> >> >> >> > Consider a wireless channel where we have large scale fading
> >> (path-
> >> >> >> > loss + shadowing) and small scale fading (multipath +
> doppler).
> >> >> >> > Consider that given a certain distance between TX and RX we
> >> observe
> >> >> >> > how the channel evolves in time and frequency.
>
> >> >> >> > My question is if we can observe the frequency response of the
> >> >> channel
> >> >> >> > for a long time, and then we average the observed path loss
> >> across
> >> >> >> > frequency and time, would the small scale fading effects
> cancel
> >> out
> >> >> ?
>
> >> >> >> If you have an 'evolving path' you necesarily have a
> time/varying
> >> >> >> scenario, either because the envronment changes, the source
> and/or
> >> >> >> reciever moves, or both.
>
> >> >> >> In that sort of scenario one might say that 'small scale effects
> >> >> >> average out' but it wouldn't be very helpful.
>
> >> >> >> The big picture remains the same, multi-path effects still mess
> up
> >> >> >> the signal. Once the path goes dynamic you have lost some of the
> >> >> >> keys you might have used to tune an adaptive system to
> compensate
> >> >> >> for multpath propagation.
>
> >> >> >> The key with multipath propagation is that you get destructive
> >> >> >> interference. It does not suffice to look at a frequency
> >> >> >> response go see how it works. If you have negative interference
> >> >> >> at some frequency, you get a zero in the frequency response,
> >> >> >> and it is fairly easy to understand why the signal is messed up.
>
> >> >> >> The key is to see that the same goes for 'non-cancelling'
> >> >> >> intereference, i.e. where the interefence does not cancel
> >> >> >> the 'business signal'. If you add an interfering signal
> >> >> >> which has a random amplitude and phase, the detection task
> >> >> >> does not become very easy, even if the frequency response
> >> >> >> might indicate there is plenty of signal energy present.
>
> >> >> >> If everything is stationary and the system is well designed,
> >> >> >> you might get enough time to detect and compensate for these
> >> >> >> effects. If anything starts moving, everything changes all
> >> >> >> the time.
>
> >> >> >> You might want to play around with some simple scenarios,
> >> >> >> like Lloyd's mirror.
>
> >> >> >> Rune
>
> >> >> >Dear all,
>
> >> >> >Thanks for your answers. It is now clear to me that I should not
> >> >> >expect small scale fading effects to average out in time or
> >> frequency.
> >> >> >Maybe another way to look at it is to consider the small scale
> fading
> >> >> >modeled by a Rayleigh distribution, then if the expected value of
> >> this
> >> >> >rayleigh distribution is not zero small scale fading should not
> >> >> >average out.
>
> >> >> >Regards
>
> >> >> >Daniel
>
> >> >> %%%%
>
> >> >> Dear All
>
> >> >> I have got one question related to this.
>
> >> >> If you want to simulate doppler, how do u do that? i.e. I have
> designed
> >> my=
>
> >> >> transmitter in matlab, and i want to see if my receiver corrects
> the
> >> >> doppler and ISI. I do not know how to introduce doppler into
> system?
>
> >> >> Thanks
>
> >> >> Chintan
>
> >> >Either use "rescale" or write an interpolator (such as a Farrow
> >> >interpolator).
>
> >> %%%%
>
> >> HI,
>
> >> But the meaning of doppler is that it is frequency modulation, then it
> >> should not affect the amplitude of the signal, and the change in the
> >> aplitude will depend on the path.
>
> >> And can u please explain, the use of 'rescale', I am also planning to
> add
> >> noise and 3-tap channel.
>
> >> Your guidance will be greatly appreciated.
>
> >> Thanks.
>
> >> Chintan
>
> >I meant "resample", sorry. &#4294967295;You have to do a time re-scaling,
> >basically.
> >I don't know what you really mean by "frequency modulation". &#4294967295;To me,
> >Doppler means that the time reference at the transmitter is "t", and
> >at
> >the receiver it is "t(1+\delta)", so you are compressing (and
> >expanding)
> >time (and freq) respectively.
>
> >Modeling the above in discrete-time amounts to sampling rate
> >conversion.
>
> HI..
>
> You are correct....But I have already used this function, but the problem
> is my system works fine for more expantion then less.
>
> My original vector size is 30720;
> And I am using resample(x,10010,10000) -> works fine
>
> But resample(x,10001,10000) -> does not work.
>
> Now to correct the doppler I am using linear interpolation, a tested
> algorithm, which is adaptive. What if I set the initial Interpolation
> factor to some random value instead of setting it to 1.
>
> And see if my algorithm corrects it or not? I have tried this method and
> it gives me expected result.
>
> Thanks
>
> Chintan

Using the "resample" command does not give you arbitrary sampling
rate change, for the simple reason that it's intended for simple small
integer rate changes.  If you want to be finer, then write your own
Farrow interpolator.  With oversampling, it's easy to get very good
quality.

To check the results, first decide what you mean by "quality".  If you
mean "square error", then check your function against a parametric
function such as a sine, and check for different digital frequencies
with random phase.

Julius