Jim:
[snip]
Look up "noise bandwidth", it's different from the actual 3dB
bandwidth... I can't remember the ratio right off the top of my head.
...Jim Thompson
[snip]
It sure is different from the 3dB bandwidth... and it is not a constant
ratio
it varies depending upon the transfer function under consideration.
If a transfer function [of a filter] is given by:
T(p) = N(p)/D(p)
Where p = sig + jw is the Laplace transform variable[Heavisides operator
"p"] then
the "total square integral" is defined by:
I = 1/2*pi * Integral [-00 - +00] {N(p)N(-p)/D(p)D(-p)}dp
Where the integral is taken along the jw axis.
i.e. I is the total area along the jw axis under the function...
1/2*pi * |T(jw)|^2
Such an integral is evaluated using Cauchy's Residue Theorem from the theory
of complex variable
calculus. This integral, i.e. the number I, is the total power delivered
through the transfer function
T(p) by a unit value white noise source applied to the input of the
filter/transfer function.
The noise bandwidth of the system with transfer function T(p) is then the
bandwidth of a brick wall
filter of unity gain that has the same integral value, i.e. the value of I.
If the brick wall filter has bandwidth
B then B = I, etc...
So to find the equivalent noise bandwidth you gotta first evaluate and get a
number for I for your transfer
function or filter and then the equivalent noise bandwidth is revealed as
the value of that total squared integral.
The value of total squared integrals [The "I" that I mentioned.] is actually
tabulated in at least a couple of
textbooks for many transfer functions in terms of the coefficients (e.g.
N(p) = n0 + n1*p + n2*p^2 + ... nn*p^n)
of the rational polynomials of the transfer functions N(p) and D(p) up to
the order 10 or so in several textbooks.
cfr: Kaiser et all "Stochastic Control Systems".
For simple filters/transfer functions, like say first or second order then
the integral is easily evaluated by
anyone with a smattering of first year integral calculus. In fact for
simple single pole low pass filters which
is what Jim Thomson was referring to, the ratio of the total square integral
I to the "actual" 3dB bandwidth
is about 1.35, i.e. the equivalent noise bandwidth is about 35% wider than
the 3dB bandwidth. But you
can work it out for yourself. Of course the result will be different for a
second order low pass filter, and
in the general case of arbitrary N(p)/D(p) you would have to work it out
yourself or get the value of I from
the available tables.