From: Sunil Kumar ([email protected])
Date: Sat Sep 02 2000 - 00:16:50 PDT
I would like to make one more point. In digital environment, an edge is
made up of various frequency components, the highest is given by 0.35/rise
time. Now higher the frequency, lower will be the skin depth, that means
skin resistance will be higher. Hence skin effect acts like a low pass
filter, rise time will increase. Any comments?
Sunil
On Thu, 31 Aug 2000, Andreas Cangellaris wrote:
> Gentlemen,
>
> Greetings!
>
> I second Michael Tsuk's opinion. His explanation of the reasons
> why FDTD is the wrong approach to use for the quantification
> of skin effect on resistance is excellent!
>
> Michael has contributed some of the most effective methods for
> extracting frequency-dependent resistance as a search in the relevant
> literature will convince you.
>
> Andreas Cangellaris
> University of Illinois
>
>
>
> ----- Original Message -----
> From: Tsuk, Michael <[email protected]>
> To: <[email protected]>
> Sent: Thursday, August 31, 2000 12:56 PM
> Subject: RE: [SI-LIST] : skin effect
>
>
> > Mike Khusid wrote:
> >
> > > My favorite way to solve a skin effect problem is an FDTD (finite
> > > difference time domain) solver which can actually use Maxwell
> > > equations directly and avoid approximations done to derive the
> > > aforementioned skin effect formula. Practically any FDTD solver
> > > can solve for current distribution inside a real conductor and
> > > calculate the resulting skin effect loss.
> >
> > I'm sorry, I can't think of a *worse* computational technique to solve
> skin
> > effect problems than FDTD! In order to model, say, the skin effect in
> > copper at 1GHz, you'd need cells no more than say a fourth the size of the
> > skin depth, which is about 0.08 mils. So to do a 6 mil by 1.4 mil line,
> say
> > 4 mils above a ground plane, you'd need at least 900 cells wide by 400
> cells
> > high. And you'd want to do a significant length of etch, at 50000 cells
> per
> > inch. At 25 bytes per cell, minimum, you're talking 450 GB of storage per
> > inch of etch! (You could use 2D FDTD, reducing the memory requirements
> > significantly. But then you'd be making the kind of approximations you're
> > trying to avoid by using a full-wave technique).
> >
> > But the real problem is the timestep. The Courant condition would force
> > you, with 0.02 mil cells, to a timestep on the order of a femtosecond. So
> > to do just one period of a 1GHz sine wave would take about a million time
> > steps!
> >
> > Use finite elements. Use the Partial Element Equivalent Circuit method.
> > Use a method of moments code with an impedance boundary condition (my
> > favorite). But *please* don't use FDTD for this!
> >
> > --
> > Michael Tsuk
> > Compaq AlphaServer Product Development
> > (508) 467-4621
> >
> >
> >
> > -----Original Message-----
> > From: Michael Khusid [mailto:[email protected]]
> > Sent: Thursday, August 31, 2000 3:03 PM
> > To: 'Clewell, Craig W'; 'Muhammad S. Sagarwala';
> > [email protected]
> > Subject: RE: [SI-LIST] : skin effect
> >
> >
> > Craig,
> >
> > The formula Ray and you are referring to
> > skin depth = sqrt(2/(omega * mu * sigma))
> > indeed seems to be geometry independant, but that's where the catch is.
> >
> > This formula is derived from a current/electric field penetration if an
> > electromagnetic wave is normally incident on the half-infinite slab of
> > homogenious material (metal in this case). This would normally be a
> > viewpoint of microwave textbooks.
> >
> > There are two gotchas going to high speed digital world. First, in the
> > stripline the propagation of field is parallel to the metal, not
> > perpendicular. Second, the thickness of metal comparable to field
> > penetration depth changes the current distribution, and thus, the skin
> > effect. What a digital designer ultimately cares for is an increased
> > resistance of the stripline at high frequencies, and that can be
> determined
> > from the current distribution.
> >
> > My favorite way to solve a skin effect problem is an FDTD (finite
> difference
> > time domain) solver which can actually use Maxwell equations directly and
> > avoid approximations done to derive the aforementioned skin effect
> formula.
> > Practically any FDTD solver can solve for current distribution inside a
> real
> > conductor and calculate the resulting skin effect loss.
> >
> > Mike Khusid
> >
> >
> > > -----Original Message-----
> > > From: Clewell, Craig W [mailto:[email protected]]
> > > Sent: Thursday, August 31, 2000 2:24 PM
> > > To: 'Muhammad S. Sagarwala'; [email protected]
> > > Subject: RE: [SI-LIST] : skin effect
> > >
> > >
> > > Muhammad,
> > >
> > > Funny thing....I was under the impression that the skin depth was
> > > "frequency" and "material" dependant not geometry dependant.
> > > I would be
> > > curious to know what data your boss has that says different.
> > >
> > > Craig
> > >
> > > -----Original Message-----
> > > From: Muhammad S. Sagarwala [mailto:[email protected]]
> > > Sent: Thursday, August 31, 2000 12:45 PM
> > > To: [email protected]
> > > Subject: [SI-LIST] : skin effect
> > >
> > >
> > > Hello Si Gurus,
> > >
> > > I just had an interesting discussion with my boss on "skin
> > > effect on pcb
> > > traces".
> > > I was of the opinion that increasing the trace thickness from
> > > 1/2 oz. to
> > > 1oz. would help reduce the
> > > skin effect but according to him skin effect does not reduce
> > > significantly
> > > with the increase in trace thickness.
> > > He was of the opinion ( and also had some data to back him
> > > up) that skin
> > > effect is more dependent on the
> > > width of the trace.
> > >
> > > I always thought that if one increases the overall perimeter
> > > of the pcb
> > > trace - regardless of whether it is done by
> > > increasing the width or increasing the thickness - the skin
> > > effect would
> > > reduce. I would appreciate if somebody
> > > could come up with a better explanation...
> > >
> > > Muhammad
> > >
> > > p.s.: when we talked about pcb traces we were talking about
> > > striplines in
> > > particular...
> > >
> > >
> > >
> > >
> > >
> > > Muhammad S. Sagarwala
> > > Design Engineer
> > > Schlumberger SABER
> > > Ph. (408) 586 7065
> > > Fax (408) 586 4668
> > >
> > >
> > >
> > > **** To unsubscribe from si-list or si-list-digest: send e-mail to
> > > [email protected]. In the BODY of message put: UNSUBSCRIBE
> > > si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> > > si-list archives are accessible at http://www.qsl.net/wb6tpu
> > > ****
> > >
> >
> > **** To unsubscribe from si-list or si-list-digest: send e-mail to
> > [email protected]. In the BODY of message put: UNSUBSCRIBE
> > si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> > si-list archives are accessible at http://www.qsl.net/wb6tpu
> > ****
> >
> > **** To unsubscribe from si-list or si-list-digest: send e-mail to
> > [email protected]. In the BODY of message put: UNSUBSCRIBE
> > si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> > si-list archives are accessible at http://www.qsl.net/wb6tpu
> > ****
> >
>
>
> **** To unsubscribe from si-list or si-list-digest: send e-mail to
> [email protected]. In the BODY of message put: UNSUBSCRIBE
> si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> si-list archives are accessible at http://www.qsl.net/wb6tpu
> ****
>
>
**** To unsubscribe from si-list or si-list-digest: send e-mail to
[email protected]. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****
This archive was generated by hypermail 2b29 : Tue May 08 2001 - 14:29:26 PDT