Summing Noise Sources

"Pommy" means "Englishman" in Australian slang. Phil is Australian, so he isn't a pommy. I can't see any problem with the rest of the description - except perhaps that you might have included "autistic" since Phil usually inculdes it in his expressions of dissatisfaction.

-- Bill Sloman, Nijmegen

I think it may be a type of cheap white wine.

John

My mistake. Unprovoked rage is harder than it looks.

-- Joe

"John Jerkin Off"

** Fuck off - you pathetic WANKER !!

...... Phil

** I never mentioned Gaussian - only real noise sources that have finite amplitude limits.

Whatever approximation method of assigning a peak amplitude one adopts - that number will double when two similar (equal RMS value) noise sources are summed.

** So you too agree that CF increases when un-correlated noise sources are summed.

Fine.

It is no anomaly.

...... Phil

That's what you always say when you don't understand something. Curious.

John

NOBODY here seems to agree with you.

John

When I noticed a thread that was seemingly written by you alone, I immediately knew that Phil Allison must have had a hand in it. Thanks to my killfile, he's sort of the dark matter of s.e.d: He's invisible, but he's needed to provide an explanation for some phenomena, like John Larkin replying to himself over and over.

robert

"Pommeroy's plonk"?

robert

They all clip when they hit the rails, but in real life that happens once every few million years even people who know what they are talking about save time by ignoring it.

If you say so Phil. Anybody for whom it matters should consult a statistician or run an experiment - the experiment is probably easier and quicker than getting a straight answer out of any statistician you can afford (my cousin in CSIRO was out of my price range that only time I tried to give him some business).

That isn't what he said.

If you say so, Phil.

-- Bill Sloman, Nijmegen (but under cover in Sydney at the moment).

Strange, but Phil started this thread by - prepare yourself for a shock - actually asking a technical question in a civil manner. But it turns out that he already thought he knew the answer (he was wrong, of course) and launched into his usual insults and profanities when people tried to tell him the truth.

I sure wouldn't let a klutz like this fix my broken stereo. He can't understand a normal distribution curve, despite the fact that, from where he stands, he can see almost all of it.

John

How do those Mexican ladies manage to wrap a super burrito in a tiny piece of foil, so smoothly and elegantly, with a minimum of motion, but it looks like hell, and the foil breaks, and it leaks, if I eat half and try to re-wrap it?

That makes me SO MAD.

John

Practice, practice, practice.

Work-hardening. :-)

Oh, you're just impatient. ;-)

Cheers! Rich

Sounds like a confused school boy mindlessly regurgitating some rules with no rhyme or reason...Your so-called RMS is a simple standard deviation of two independent random variates, and as anyone knows, the standard deviation of the sum is the RSS of the component deviations. These statistics have absolutely nothing to do with spectral power density, it is a characteristic of the single sample amplitude distribution only.

Talking to yourself?

whatever...

There is no peak value for the idealized amplitude statistics, there may be a practical limitation in the real world, but the idealized model usually fits with less than 1e-24 chance of deviation from reality, that's why people consider it useful.

Pretty bold generalization from someone who knows so little. Your definition of crest factor is arbitrarily defined as that amplitude four standard deviations removed from the mean ( which is zero), or in the case of normal-like naturally ( statistical mechanical) occurring noise, amplitudes occurring less than 0.01% of the time.

arbitrary...

Newp...not at all. The new arbitrarily defined peak, or PK, is 4x RSS of the standard deviation of the sum. Each addend of the RSS is PK/4 of the individual distributions, so that PK of the sum is, in general, PK,sum= sqrt(PK1^2 + PK2^2 + ....+ PKn^2), or PK total= RSS of component PK's. But, from above, the standard deviation of the sum is 1/4*sqrt(PK1^2 + PK2^2 + ....+ PKn^2), making the CF of the sum to be 4. So the crest factor remains constant at 4.

You say 'this' because you are confused.

Not really, you are confused; by your reasoning the universe would explode.

You are anomalous. Calm down and study the arithmetic I showed you...keep going over it, again and again and again as required by your limited mental faculties, until you get it.

You lack the sense to realize that is the 'peak' represents a limit above which the noise rises less than 0.01% of the time, then the sum of two sources will rise above 2x peak less than 0.0001% of the time. This means 2xPK is not the peak of the sum as you assumed so simplemindedly....This was your 'basic issue.'

As soon as the math confuses him, which doesn't take long at all, he cuts over to cursing. Really, it's not worth trying to help him; all you get is perplexity followed by abuse.

John

The question probably relates to various standards audio systems testing, maybe for power handling, such as IEC or whatever that specify the test signal as pink noise with CF, crest factor, of 4. In practice the pink noise is generated by linear filering of a white noise source to shape the energy spectral distribution into one of equal energy per octave bandwidth , or what is known as pink. The amplitude distribution of the resulting pink noise amplitude distribution remains Gaussian because of the simple fact that any linear operation on a Gaussian distribution remains Gaussian. Therefore, the Gaussian characteristic of the source amplitude not only survives the pink noise transformation process, but also the linear superposition with other independent Gaussian noise sources. The point of all this background is that the

4-sigma on the composite distribution represents the exact same frequency of occurrence as in the sources of which it is composed and therefore CF remains invariant at 4. I think this is what the OP wants to know.