From: Don Pearce
Subject: Re: SwCADIII
Date: Sun, 05 Jan 2003 14:56:44 +0000
NNTP-Posting-Host: host217-39-105-124.in-addr.btopenworld.com (220.127.116.11)
X-Newsreader: Forte Agent 1.91/32.564
On Sun, 5 Jan 2003 14:02:45 +0100, "mmormota"
>I am sure that infinite number of very different time domain signals have
>the same f domain distribution curves. I just like to see one of them,
>knowing that very different signals are possible too.
No, I believe that every set of amplitudes and phases in either the
frequency or time domain has a unique solution in the other domain.
>Lets suppose, that I have to choose one of two opamps, but the shape of the
>noise profiles are different. One of them has a normal 1/f shape, the other
>has a steeper profile at low freqs, but the turning point is lower, and the
>noise at 1kHz is lower too. Wich is better? I think I am not the only person
>who not just integrate the f domain noise profile but want to see it on a
Noise profiles don't really come in different slopes - but the
frequencies at which the different slopes come into play can vary
considerably - the 1/f point being the key inflection. If the "white"
noise is lower, then for a given 1/f noise level, the 1/f frequency
will be higher - but the op-amp is not so good.
Unfortunately this kind of noise is not that easy to examine. An FFT
analyser with an averaging function is probably the best bet - a scope
is probably the last instrument you would use.
>Opamp datasheets are providing a f domain curve, and a 0.1Hz-10Hz uVp-p data
>too. As noise is a random sequence, the p-p value is questionable, but I
>like it anyway... :-)
P-p values of noise have meaning only in terms of the time domain.
Take a thin enough slice of noise, and it can be as big as you like -
it will contain very little energy though. The Gaussian distribution
is a pretty good approximation to noise from an op-amp.
Depending on what you are trying to do with your circuit, either Pk-Pk
or RMS will be the appropriate noise measurement to make - generalised
preferences aren't terribly helpful.