The basis of every 2D field solver is that the objects really continue in both directions to infinity. This way, we need only solve Maxwell's equations in 2 dimensions, greatly speeding up the process. However, it means that we are only able to calculate loop inductances per length- signal lines carry current into the board, and some other conductor, somewhere must carry the return path. If you don't specify the return path conductor, for example as selecting it as either "ideal ground" or "surface ground", the boundary will be used as the return path.
Calibrating field solvers is a very important task. Traditionally, most folks have calibrated one field solver against another. I have found this uncomfortable. A better way is to extract the parameters from a structure where there is an exact expression of the electrical properties of the structure. Unfortunately, the only cross section where there is an exact closed form analytical expression is two cylinders.
Charles Grasso is exactly right in looking at the case of twin wires as a test case. This also is a useful exercise for new users of a field solver tool. It is simple to set up and can be compared to a known result. This example is also included in the set of example problems on the new release of Spicelink 3.0 CD. It is labeled as axtwin01, in the directory labeled 'examples'.
Ansoft has written a number of papers on the issue of calibrations and interpreting 2D field solvers. Next week, we are presenting a paper at the IPC PCB Expo in Long Beach on comparing field solver results to measurements. We show the comparison to twin wires is better than 1% absolute accuracy. We also compare the extracted results of a parallel plate example, as you make the aspect ratio really extreme. It is surprising that you must make the height to width ratio at least less than 0.001 in order to get better than 1% agreement to the parallel plate approximation. When you use a field solver a lot, you are able to calibrate your intuition on the relatively large value of fringe fields.
If anyone would like copies of the papers we have written or handouts from the presentation at the PCB Expo, please send me a note off line and I will be glad to send you copies.
--eric
At 12:49 AM 4/24/98 -0700, alterra@adnc.com wrote:
>Charles,
>
>Here is some information on your quesitons about Ansoft's 2D Extractor tool:
>
>Problem 1:
>
>The standard equation for the impedance of a symmetrical 2 wire line in a
>medium with dielectric constant of 1 is
>
>Z=(377/pi) arc-cosh(s/d)
>
>where s is the wire diameter and d is the center to center wire spacing.
>Here it is assumed that the current flows out one wire and returns in the
>other. For your values, the equation gives and impedance of 188.02 ohms.
>If you set up this problem in the Ansoft 2D Extractor and identify one wire
>as a signal conductor and the other as an ideal ground, you obtain an
>impedance of 188.03 ohms as expected. I don't know how you set up your
>boundary conditions so can't comment on how you got a value of 14 ohms.
>
>Problem 2:
>
>You cannot use the 2D Parameter Extractor to analyze the partial inductance
>of a piece of wire by itself. When you analyze a problem with a 2D solver,
>you are assuming that the geometry has the same cross section to +/-
>infinity. But the partial inductance of an infinitely long wire is
>infinite. It only becomes finite if there is a return current path on
>another conductor or the wire is of finite length. If there is a return
>current path, you can include it in the 2D model. If the wire has a finite
>length, you can use the 3D Parameter Extractor to get the partial
>inductance. I suspect that you were using the problem boundary to carry the
>return current in your calculation and, id so, your result will depend on
>the size of the boundary box.
>
>I hope this helps.
>
>Best regards,
>
>Eric
>-----------------------------------------------------------------
>Eric Wheatley Ph.D. Electromagnetic Modeling Services
>Alterra Technology Co. (760) 942-9426 (phone)
>435 Dunsmore Ct. (760) 942-2366 (fax)
>Encinitas, CA 92024 US alterra@adnc.com
>-----------------------------------------------------------------
>
>At 02:56 PM 4/23/98 -0600, you wrote:
>>This question is for anyone who uses the Ansoft 2D extractor..
>>
>>I have been using this s/w for about 4 months and, in my innocence,
>>have been attempting to "calibrate" the answers. For example, I
>>plugged in a simple microstrip and behold - I got the right answers for
>>Zo.
>>
>>Problem 1..
>>I then tried to get the Zo of a two wire configuration. The
>>wires are 20mils diameter and spaced (center to center) 50mils apart.
>>
>>Accepted calculations indicate that the Zo should = 188 ohms. The
>>2DExtractor gives me an answer of 14 ohms!!
>>
>>Problem 2..
>>I then tried to get the inductance of a 2 meter length of 28AWG wire.
>>Measurements (@1kHz) gave an answer of 2.1uH.
>>Formulae gave an answer of 3.748uH.
>>2DE gave an answer of 0.5uH
>>
>>Can somebody explain what is going on??
>>
>>Thank you
>>Charles Grasso
>>EMC Engineer
>>StorageTek
>>2270 Sth 88th Street
>>Louisville CO 80027 MS 4262
>>grassc@louisville.stortek.com
>>Tel:(303)673-2908
>>Fax(303)661-7115
>>Symposium Website URL: http://www.ball.com/aerospace/ieee_emc.html
>>
>>
>>
>
>
**********************************************************************
Eric Bogatin
Product Manager, Signal Integrity Products
Ansoft Corporation
26235 W. 110th Terr.
Olathe, KS 66061
voice: 913-393-1305
fax: 913-393-1306
email: bogatin@ansoft.com
pager: 888-775-1138
web site: <color><param>0000,0000,ffff</param>www.ansoft.com
</color>
"In God we trust, all others, show your data"
**********************************************************************