Subject: Re: Transition of a transmission line into a lumped-elements circuit
Date: 5 Nov 2002 10:24:41 GMT
References: <firstname.lastname@example.org> <email@example.com> <firstname.lastname@example.org>
User-Agent: tin/1.4.5-20010409 ("One More Nightmare") (UNIX) (CYGWIN_NT-4.0/1.3.12(0.54/3/2) (i686))
Allan Herriman wrote:
> On 4 Nov 2002 17:21:19 -0800, email@example.com (Tom Bruhns) wrote:
>>If you know the line's characteristic impedance and delay time
>>(presumably 250psec), you can get an L and C from Z0=sqrt(L/C) and
>>Tau=sqrt(L*C). If you divide the resulting L and C values by n, you
>>can make n sections, cascading them. I would suggest actually
>>dividing the end L into two, and putting half of it on each end. In
>>the limit as n gets large, this structure should approach the same
>>behavior as a lossless line. You can add series R and shunt G to
>>simulate loss, but those, in an actual line, are frequency dependent.
>>And if n is small, the circuit won't behave very close to the same as
>>a real line...you can do better with the same number of elements by
>>tweaking on their values, but it's not something I do often enough to
>>be able to tell you off the top of my head how to best do it.
> Here's an old discussion on the issue of determining L and C
> parameters in a lumped line model:
That discussion was about modelling of transmission lines in Spice, in
which case using large numbers of L's and C's to improve the accuracy
of the model is fine. But attempt to emulate that on a real design, and
you'll find that the L and C values end up too small, and board effects
will become significant too. In any case, by using enough sections to
produce an accurate model, you've probably taken up more board area that
the OP's lamda/4 transmission line.
Your employer's application notes contain at least one reference to
lumped quarter-wave sections implemented as Pi or Tee networks. These
just require three components to attain a 90 deg shift.