AW(2): Re: Couple EE questions...

Magnus Danielson magda at it.kth.se
Thu Oct 24 15:12:39 CEST 1996


> > So, it is in fact three various types of noises, and one of them you can't
> > do anything about and the other two must be balanced to get the best 
> result.
> > Picking an arbitrary low value will not generate the best noise.
> [...]
> > Do you now grasp the noise problem a little better Juergen?
> 
> Yes - and thanks for the good explanation!

(blush) I try, I try.... I like teaching you see...

> Good to have you here !

Nice to be here :)

> > Note however that shot noise, flicker noise is not included within this =
> > discussion. Metal film and wire wound resistors are quieter in general th=
> > an
> > carbon composition and carbon film resistors when it comes to flicker noi=
> >se =(1/f noise).
> 
> Do you know - as a rule of thumb - *how much* worse a real garden-variety
> carbon or metal film resistor is, compared to an ideal one? And is the
> difference only in the 1/f noise (i.e. at low frequencies), or is there a
> a difference in the upper f range as well that is worth mentioning?

As for the rule of thumb I have these values picked out of a book:

Carbon-composition	0.10uV to 3.0 uV
Carbon-film		0.05uV to 0.3 uV
Metal-film		0.02uV to 0.2 uV
Wire-wound		0.01uV to 0.2 uV

These values are measured over an octave of frequency (don't know which).

It's a rule of thumb, but if one really cares about noise that much one must
really get the datasheets for resistors and ensure that you buy resistors of
that quality. It should by know be apparent that resistors is not only resistors...

As for the upper ranges of frequency I don't know of a specific type of noise
that is causeing a problem, but there is other issues where noises can occur
in large quanteties....

Noise is as all signals being affected by filters, but only noise being at the
input will follow the same damping curve as the signal on the input terminal.
Noise (and other signals I migth add) being introduced at other parts of the
filter circuit can however be amplified. The noise curve of a filter can be
quite diffrent from the frequency response curve and the signal to noise figure
can be quite low for some ranges. The same problem is caused by round of noise
in digital filters BTW.

It is not all that uncommon that we judge the sound of a box on how the noise
of it sounds...

I have been around optimising a PA system so that even if the top box was capable of 139 dBspl at one meter we tried to kill the noise in the system such
that you could lean your head to it and not hear much noise (with no input
signal, _full_ volume of the amps... just the crossover filter and the amps).
When you have such a system you have well beyond 120 dB dynamic in the system
looking for 130 dB... that's when 16 bits digital stuff just won't do it!!!

The world of sounds is partly a land in shadows, hard to see clear... much of
this comes from lack of real knowledge and ignorance. One must be prepared to
throw all old stuff overboard and go on a tripp which migth lead to new stuff...
Few people don't dare to consider throw the old stuff, they are just happy
haveing people invent new "problems" for them along with the "cure".

I'm not a mystics when it comes to sound... however, I migth keep some stuff to
myself... but that's a diffrent thing...

Oh gosh, now I have gotten into the political/religous arean... and I who only
wanted to say a few things about noise... the signal to noise ratio was maybe
a bit low today...

Cheers,
Magnus





More information about the Synth-diy mailing list