Intersymbol Interference
Lim Fung
I've read somewhere on a publication that reducing (or zero forcing)
Intersymbol Interference on a baseband transmission with the addition
of white noise may increase the noise intensity but could not find any
pointers to the proof. Would appreciate if anyone could shed light on
this.
thanks.
Bill Sloman
Lim Fung wrote:
If you take the Fourier transform of your symbol, you will see that itcan
be represented as the sum of a series of sine waves of differing
frequencies.
Intersymbol Interference can be caused either by differential dispersion
in the transmission medium, so that that the delay through the
transmission medium varies with frequency, or by differential
attenuation in which the higher frequencies are attenuated more than
lower frequencies, as well as being phase shifted.
In the former case, you can reduce intersymbol interference with an
all-pass filter, in which the gain is constant, and only the phase shift
changes with frequency. Such a filter won't affect the noise.
In the latter case, you need a filter whose gain increases with
increasing frequency, to restore the attenuated frequencies to
thier original amplitude. This filter will also amplify any noise
components at these frequencies.
Hope this helps.
Bill Sloman, Nijmegen
John Woodgate
<slo...@sci.kun.nl> writes
>n the latter case, you need a filter whose gain increases with
>increasing frequency, to restore the attenuated frequencies to
>thier original amplitude. This filter will also amplify any noise
>components at these frequencies.
If you put the filter (or some of the filtering) at the sending end, the
noise level is effectively decreased. Partitioning the frequency-
response correction ('equalization') optimally between sending and
receiving ends is an interting design exercise.
--
Regards, John Woodgate, Phone +44 (0)1268 747839 Fax +44 (0)1268 777124.
OOO - Own Opinions Only. You can fool all of the people some of the time, but
you can't please some of the people any of the time.
Jack Lee
the zero forcing algorithm for digital (discrete time) T-space
equalization.
The ZF alg will force the folded spectrum of the channel
and the equalizer to a constant value thus eliminating the
ISI. So for math convience, assume an ideal AGC was use
to set the constant value to unity. This means the
equalizer would be the inverse of the channel. Given such,
if the channel has spectrul minimums then the alg will set
the equalizer to amplify sections of the spectrum to acheive
a constant spectrum. This in turn will amplify any additive
noise in the channel, if white, then part of the flat spectrum
will be amplified. So theoritically, if the channle minimums are
zeros then the alg will force infinite noise amplification. Note,
channels can have spectrul zeros or small minimums due
to reflections.
For a better proof (meaning with equations along with greek letters)
try Digital Communication by Proakis (Mcgraw-Hill).This book is
dense and a hard read but IMO the best reference in terms of
modulation theory and equalization.
jack
Lim Fung <lim...@pacific.net.sg> wrote in article
<6t2gog$9vj$1...@newton2.pacific.net.sg>...
> hi,
>
> I've read somewhere on a publication that reducing (or zero forcing)
> Intersymbol Interference on a baseband transmission with the addition
> of white noise may increase the noise intensity but could not find any
> pointers to the proof. Would appreciate if anyone could shed light on
> this.
>
> thanks.
>
>
>
Jack Lee
John Woodgate <j...@jmwa.demon.co.uk> wrote in article
<kpx6HOBi...@jmwa.demon.co.uk>...
> In article <35F51F1C...@sci.kun.nl>, Bill Sloman
> <slo...@sci.kun.nl> writes
> >n the latter case, you need a filter whose gain increases with
> >increasing frequency, to restore the attenuated frequencies to
> >thier original amplitude. This filter will also amplify any noise
> >components at these frequencies.
>
> If you put the filter (or some of the filtering) at the sending end, the
> noise level is effectively decreased. Partitioning the frequency-
> response correction ('equalization') optimally between sending and
> receiving ends is an interting design exercise.
> --
In 10BaseT, the sending end is "filtered" such that ISI and zero crossing
jitter is minimize at the receiving end. Most people call this
predistortion.
In this system, some amount of filtering is still needed at the receiver to
limit noise but all of the waveshaping is done at the transmitter to
eliminate
ISI.
Optimum partition between the sending and receiving ends has been
analyzed for raise cosine wave shaping. It turns out that each end
should has a square root of the raise cosine function, thus the combine
filter yeilds the raise cosine waveshape. Note, this is optimum in terms
of minimum prob of error under additive white gaussain noise.
jack