On Mar 26, 11:40&#4294967295;am, Vladimir Vassilevsky <antispam_bo...@hotmail.com>
wrote:
> sasuke wrote:
> > The question which I actually wanted to ask was, is it true that Gaussian
> > noise is the worse kind of noise to affect a communication channel? In
> > other words is the capacity of a communication channel least when the
> > channel is corrupted by an additive Gaussian noise??
>
> You are correct. For the given noise power, the AWGN corresponds to the
> minimum capacity of the communication channel.

Agree.

>Any non-AWGN case can be
> converted to the equivalent Gaussian by means of some linear or
> non-linear operation with the corresponding increase in the channel
> capacity.

Not really. You may be able to convert the channel to a gaussian one,
but you can not increase the channel capacity by the data processing
theorem in information theory.
>
> Vladimir Vassilevsky
> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com

Vladimir Vassilevsky  <antispam_bogus@hotmail.com> wrote:

>Steve Pope wrote:

>> Suppose the non-AWGN case consists of impulse noise arriving
>> randomly (say, with Poisson statistics).  How would you
>> convert that to the AWGN case?

>First, I will limit those pulses to the level of the useful signal, then
>I will smear the residue by some sort of dispersive filter.

Sounds good.  I think one could argue that if you apply enough
pseudorandom interleaving, the impulse noise becomes gaussian,
using some sort of central-limit-theorem argument.

Steve

Steve Pope wrote:
> Vladimir Vassilevsky  <antispam_bogus@hotmail.com> wrote:
> 
> 
>>Any non-AWGN case can be converted to the equivalent Gaussian
>>by means of some linear or non-linear operation with the
>>corresponding increase in the channel capacity.
> 
> 
> Suppose the non-AWGN case consists of impulse noise arriving
> randomly (say, with Poisson statistics).  How would you
> convert that to the AWGN case?

First, I will limit those pulses to the level of the useful signal, then
I will smear the residue by some sort of dispersive filter.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

Vladimir Vassilevsky  <antispam_bogus@hotmail.com> wrote:

> Any non-AWGN case can be converted to the equivalent Gaussian
> by means of some linear or non-linear operation with the
> corresponding increase in the channel capacity.

Suppose the non-AWGN case consists of impulse noise arriving
randomly (say, with Poisson statistics).  How would you
convert that to the AWGN case?

(I agree that the capacity is higher given the same noise power
for the impulse-noise case; I'm unclear on how one could do the 
conversion.)

Steve

sasuke wrote:

> The question which I actually wanted to ask was, is it true that Gaussian
> noise is the worse kind of noise to affect a communication channel? In
> other words is the capacity of a communication channel least when the
> channel is corrupted by an additive Gaussian noise??

White or colored?

bye,

-- 

piergiorgio

sasuke wrote:

> The question which I actually wanted to ask was, is it true that Gaussian
> noise is the worse kind of noise to affect a communication channel? In
> other words is the capacity of a communication channel least when the
> channel is corrupted by an additive Gaussian noise??

You are correct. For the given noise power, the AWGN corresponds to the 
minimum capacity of the communication channel. Any non-AWGN case can be 
converted to the equivalent Gaussian by means of some linear or 
non-linear operation with the corresponding increase in the channel 
capacity.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

On Wed, 25 Mar 2009 17:34:22 -0500, sasuke wrote:

>>On Mar 26, 7:54 am, "sasuke" <a.ssjg...@gmail.com> wrote:
>>> Does Gaussian Noise have the highest variance among all possible noise
>>> distributions? Is there a way to prove it??
>>
>>Isn't that a bit like asking - Does a 24kHz sine wave have the largest
>>amplitude of all frequencies?
>>
>>
> Sorry. I messed up the question completely. Please disregard. The
> question which I actually wanted to ask was, is it true that Gaussian
> noise is the worse kind of noise to affect a communication channel? In
> other words is the capacity of a communication channel least when the
> channel is corrupted by an additive Gaussian noise??

It's certainly the easiest kind to analyze, and I _think_ it has the 
lowest possible values for all of the higher-order moments for a given 
variance (dunno -- have to check).

It's hard to say what's "worst", because if you know the noise properties 
before hand you can design for it.

Certainly if you're designing for Gaussian noise and you're hit with 
noise that has a Cauer-like density (i.e. infinite variance, as you see 
with atmospheric noise in LF and MF) your receiver will be more than a 
little confused.

-- 
http://www.wescottdesign.com

On Mar 25, 7:08&#4294967295;pm, Sebastian Doht <seb_d...@lycos.com> wrote:
> sasuke schrieb:
>
> >> On Mar 26, 7:54 am, "sasuke" <a.ssjg...@gmail.com> wrote:
> >>> Does Gaussian Noise have the highest variance among all possible noise
> >>> distributions? Is there a way to prove it??
> >> Isn't that a bit like asking - Does a 24kHz sine wave have the largest
> >> amplitude of all frequencies?
>
> > Sorry. I messed up the question completely. Please disregard.
> > The question which I actually wanted to ask was, is it true that Gaussian
> > noise is the worse kind of noise to affect a communication channel? In
> > other words is the capacity of a communication channel least when the
> > channel is corrupted by an additive Gaussian noise??
>
> Actually gaussian additive noise is the ideal kind not the worst kind.
> Because gaussian distributions are simple to be analysed many filters
> are optimised for that kind of noise.
> The capacity of the channel will decrease in any case of noise, so your
> question still sounds confused...

I'll try to interpret your question as given a fixed amount of average
power, which amplitude distribution is most effective at jamming.

In that case Gaussian noise would not be the answer...

I think the answer would be an impulsive noise signal whose BW was
confined to your channel of interest and that is either off or on and
when on it is at max power and the duration and repetition rate of the
on time is such so as to do the most damage based on the FEC the
system is using.  So the answer depends on the type of FEC.  But
Gaussian noise while it has infinte peaks (in theory) does not make
the "best use" of the average power.

Mark

sasuke schrieb:
>> On Mar 26, 7:54 am, "sasuke" <a.ssjg...@gmail.com> wrote:
>>> Does Gaussian Noise have the highest variance among all possible noise
>>> distributions? Is there a way to prove it??
>> Isn't that a bit like asking - Does a 24kHz sine wave have the largest
>> amplitude of all frequencies?
>>
> 
> Sorry. I messed up the question completely. Please disregard.
> The question which I actually wanted to ask was, is it true that Gaussian
> noise is the worse kind of noise to affect a communication channel? In
> other words is the capacity of a communication channel least when the
> channel is corrupted by an additive Gaussian noise??
> 

Actually gaussian additive noise is the ideal kind not the worst kind. 
Because gaussian distributions are simple to be analysed many filters 
are optimised for that kind of noise.
The capacity of the channel will decrease in any case of noise, so your 
question still sounds confused...

On Mar 25, 6:34&#4294967295;pm, "sasuke" <a.ssjg...@gmail.com> wrote:
> >On Mar 26, 7:54 am, "sasuke" <a.ssjg...@gmail.com> wrote:
> >> Does Gaussian Noise have the highest variance among all possible noise
> >> distributions? Is there a way to prove it??
>
> >Isn't that a bit like asking - Does a 24kHz sine wave have the largest
> >amplitude of all frequencies?
>
> Sorry. I messed up the question completely. Please disregard.
> The question which I actually wanted to ask was, is it true that Gaussian
> noise is the worse kind of noise to affect a communication channel? In
> other words is the capacity of a communication channel least when the
> channel is corrupted by an additive Gaussian noise??

What would happen if your noise looked very much like your signal
rather than being Gaussian?