[SI-LIST] problems encountered extracting radiation losses
- From: "Jan Vercammen" <jan.vercammen.jv1@xxxxxxxxxxxxxxxx>
- To: si-list@xxxxxxxxxxxxx
- Date: Wed, 22 May 2002 10:37:46 +0200
Hello SI-list,
I have been trying to use the LC FDTD code (under Linux) to extract
radiation loss data from
mircroelectronic structures (check lc.cray.com). I have put the questions
below on the LC-user forum,
but got no reply, so I am trying the SI-list.
I started working on simple structures that I can compare with a frequency
domain full-wave solver (MoM).
I have been succesful in extracting the input impedance of a small dipole
(8cm long) and have been able to extract
radiation resistance as low as 1ohm.
Whereas in antenna design one is interested in a relative large radiation
resistance, I am, from the point
of view of EMC, only interested in in-efficient antennas or low radiation
resistance structures.
With MoM code it is possible to extract radiation resistances below 0.1ohm
(or even below <1milliohm). I have
been able to do almost the same with (the LC) FDTD code, but currently I am
stuck at 1ohm as a lower limit.
I was wondering if other LC users or FDTD users have been pursuing similar
problems. The main issue is to reduce the
reflection from the boundaries. Here are my questions (note that some are
very specific to the LC code):
-1- I am using a voltage source excitation and sample the current and
voltage, a simple and direct technique.
I extract the complex voltage and current spectrum and take their
ratio. The results are one-to-one for
frequencies above 1GHz compared with MoM. Extra noise is present
(below 200-500MHz) if the ACB
reflects too much.
Are there other approaches to gauge radiation loss possible besides
the one technique above??
-2- I use a PML ABC (obviously). In order to reach the radiation resistance
level of 1ohm I have to use
a PML thickness of 20. Is it useful to make it even thicker?? What is
the limit?? --> the limit is 32 in LC
-3- I am using a PML order of 2 (square) but order 3 (cube) or 4
(quadratic) could improve the situation
further (accoring to Taflove's book), but only if you run using long
doubles (128bit??). Is this enabled
under the Linux distribution or is only order 2 useful on the
PC/Linux platform??
-4- What is exactly the effect of the PML tolerance and the relation with
the PML thickness and order?
Currently it is set at 10E-9. How do you set it accordingly (with
respect to thickness and order)??
-5- What is the optimal distance between (in cells) the structure under
investigation and the boundary?? Currently
I use about 10 cells, but a smaller value of 5 cells gives very
similar results.
-6- Does it make sense to enlarge the domain, such that, the engergy/cell
becomes smaller?? Is there a
relation with question -5- ? I have tested this but it does not
improve the situation.
-7- Because the magnetic and electric field vectors are sampled 1/2 a cell
and 1/2 a time step apart I was
wondering if the LC code interpolates the magnetic (or electric
field) in space and time. With respect to
the dipole time domain data it seems that voltage and current use the
same absolute time step.
In Taflove's book there is mentioned a (faze) correction because of
the 1/2 time step difference. But
I suspect that this is a high frequency issue and not a low one (<
200-500MHz). Is this correct reasoning??
-8- Whereas it is easy to tackle open structures (like dipoles or open
transmission lines), it is more
difficult to do the same for loop-like structures. Using a voltage
source will create a slowly decaying current (because
the time constant for the current is L/R, where L=loop_inductance and
R=loop_resistance). Inserting a resistor Rs allows damping
and reduces the run time considerable. But this poses the problem
that the loop antenna impedance
now includes an extra resistive term Rs, which can be substracted (of
course) from the total antenna impedance.
This is related to question -1-, I think. Has anyone used this
apporach in his work?? In a sense this looks
like a vector network analyser (or TDR approach).
-9- I have the impression that the LC Mur 2nd order ABC is not working
correctly under Linux. Is this correct??
To conclude: LC is a very nice and powerful tool for signal integrity and
EMC. It is a very nice
gesture to make the binary code available to a broader public. Bravo!!
Kind regards,
Jan Vercammen
EMC/SI Engineering
Agfa-Gevaert
Mortsel Belgium
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
