Re: [SI-LIST] : Effect of low Zo for unterminated lines

About this list Date view Thread view Subject view Author view

From: Vinu Arumugham (vinu@cisco.com)
Date: Thu Oct 21 1999 - 17:46:29 PDT


Raymond,

Since the average current is a function of load capacitance and operating frequency, what frequency do the chip designer's assume? Datasheets for LVC245 type devices do not usually specify a max. operating frequency.

If the driver is operated at half the assumed max. operating frequency, it looks like the load capacitance can be doubled without exceeding the electromigration limit.

Thanks,
Vinu

"D. C. Sessions" wrote:

> Raymond.Leung@qsa.idt.com wrote:
> >
> > Vinu,
> >
> > Electromigration is sensitive to average output current. Therefore if you drive
> > a line of 60 ohm in 133MHz, say, the average stressing current would be
> > 43/2=21.5mA on the pull-up device (base on your calculation, abeit the actual
> > output
> > resistance of the driver would be smaller); and of course the same
> > stress for the pull-down device. Some metal wires in the chip running with
> > alternate current is not an issue.
> >
> > What I have suggested is to use a 50PF lumpped load as a reference to calculate
> > the average current limit, and that's the usual way a chip designer handle the
> > electromigration issue when the chip datasheet specifies a 50PF loading. (The
> > original posting is taking LVC245 as an example)
>
> Yes and no. Electromigration is actually sensitive to the signed square
> of current density, sort of. It's not a simple relationship. The energy of
> a given electron-aluminum collision is proportional to the local potential
> gradient, which is proportional to the current density, and the number of
> collisions is of course proportional to the current density. The probability
> of an atom being displaced is a function of both energy and temperature and
> pretty nonlinear.
>
> All in all, you can make a linear formula work in small regions and a
> quadratic formula work over rather larger ones. Since a 50pf cap gives
> the same net charge transferred over a smaller time, the only problem
> is with the energy relationship. It's a pessimistic estimate by maybe
> 20%, which isn't all that bad.
>
> > ---------------------- Forwarded by Raymond Leung/QSA/AU on 22/10/99 09:25
> > ---------------------------
> >
> > Vinu Arumugham <vinu@cisco.com> on 22/10/99 02:43:00
> >
> > To: Raymond Leung/QSA/AU@QSA
> > cc:
> >
> > Subject: Re: [SI-LIST] : Effect of low Zo for unterminated lines
> >
> > Raymond,
> >
> > I did a quick calculation and have a question. If we assume a driver is spec'd
> > at Iol=24mA at Vol=0.4V, the driver impedance is 0.4/24mA = ~17 ohm. The dynamic
> > current when driving a 33 ohm line is 3.3/(33+17) = 66mA. The dynamic
> > current for a 60 ohm line is 3.3/(60+17) = 43 mA. If the 33 ohm line is 9.2 inch
> > long, at a frequency of ~250MHz, the driver will have to constantly supply 66
> > mA. For the 16.7 in. 60 ohm line, running at ~133MHz, the driver will
> > have to constantly supply 43 mA. Which is the electromigration limit, 43 mA or
> > 66 mA?
> >
> > Thanks,
> > Vinu
> >
> > Raymond.Leung@qsa.idt.com wrote:
> >
> > > I think the main issue as far as the driver is concerned is its reliability at
> > > high current/temperature level. I would stick to the 50PF loading spec.
> > > then there should not be any problem of overheat at high speed or
> > > electromigration at high current level. Thus the estimated max. line length
> > > should be 9.2 inches for 33 ohm and 16.7 inches for 60 ohm case.
> > > Please be reminded the loading of the inputs of the other chips has not
> > > been considered.
> > >
> > > Raymond
> > >
> > > Chris Bobek <cbobek@cadence.com> on 20/10/99 03:08:52
> > >
> > > Please respond to si-list@silab.eng.sun.com
> > >
> > > To: si-list@silab.eng.sun.com
> > > cc: (bcc: Raymond Leung/QSA/AU)
> > >
> > > Subject: Re: [SI-LIST] : Effect of low Zo for unterminated lines
> > >
> > > This is a little off subject, but I made a slight mistake, the Zo is 33ohms.
> > > Here is a stripline portion of my stackup:
> > >
> > > 1 oz copper plane
> > > 4mil FR-4 (er = 4.5)
> > > 1 oz signal
> > > 4.4mil Prepreg
> > > 1 oz copper plane
> > >
> > > When I plug those values into Specctraquest, I get 33 ohms. When I use
> > Howard's
> > > equations from his book, I get roughly 30 ohms or so (I kinda forget). So, I
> > > believe this is correct.
> > >
> > > Chris
> > >
> > > Dave Hoover wrote:
> > >
> > > > All the PCB's I've built using 29 Ohms (GTL and Ram something)
> > > > was .018" lines .003" above a plane (microstrip).
> > > >
> > > > ??????? .006" was using a high Er' dielectric??????
> > > >
> > > > Dave
> > > > -----Original Message-----
> > > > From: Chris Cheng [mailto:hycheng@3pardata.com]
> > > > Sent: Tuesday, October 19, 1999 9:12 AM
> > > > To: si-list@silab.eng.sun.com
> > > > Subject: RE: [SI-LIST] : Effect of low Zo for unterminated lines
> > > >
> > > > are you sure you have a 29ohm line. its very hard to believe
> > > > you can achieve 29ohm with 6 mil traces on PCB.
> > > > chris
> > > >
> > > > > -----Original Message-----
> > > > > From: owner-si-list@silab.eng.sun.com
> > > > > [mailto:owner-si-list@silab.eng.sun.com]On Behalf Of Chris Bobek
> > > > > Sent: Monday, October 18, 1999 5:24 PM
> > > > > To: Si-list
> > > > > Subject: [SI-LIST] : Effect of low Zo for unterminated lines
> > > > >
> > > > >
> > > > > Hi,
> > > > >
> > > > > Because of several constraints on our stackup and line width, a 6mil
> > > > > stripline trace yields a Zo = 29ohms. When I simulate a 29ohm,
> > > > > unterminated line on our board (@ 10MHz), I see a pretty nice waveform,
> > > > > albeit with slower rise/fall times (which is ok for the nets in
> > > > > question). When I increase the Zo to 60 ohms, for example, I see a lot
> > > > > of ringing on the same net.
> > > > >
> > > > > >From a theoretical point of view, my guess is that this happens because
> > > > > as you widen the trace, there is more capacitance, which slows down the
> > > > > rise/fall times. Is this a correct assumption? Are there other factors
> > > > > as well that are going on?
> > > > >
> > > > > What I am most concerned about is the effect of a low Zo net on its
> > > > > driver. For example, if I have an LVCH245 driver (3.3V) driving a 29ohm
> > > > > line, will the low Zo "hurt" the driver? Or, is the only negative
> > > > > effect the slower rise time? This question is a general question for
> > > > > most ICs.
> > > > >
> > > > > Thanks for your help,
> > > > >
> > > > > Chris
> > > > >
> > > > >
> > > > > **** To unsubscribe from si-list: send e-mail to
> > > > > majordomo@silab.eng.sun.com. In the BODY of message put:
> > > > > UNSUBSCRIBE si-list, for more help, put HELP. si-list archives
> > > > > are accessible at http://www.qsl.net/wb6tpu/si-list ****
> > > > >
> > > > >
> > > >
> > > > **** To unsubscribe from si-list: send e-mail to
> > > > majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
> > > > si-list, for more help, put HELP. si-list archives are accessible at
> > > > http://www.qsl.net/wb6tpu/si-list ****
> > > >
> > > > **** To unsubscribe from si-list: send e-mail to
> > majordomo@silab.eng.sun.com.
> > > In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP.
> > > si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****
> > >
> > > **** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com.
> > In
> > > the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP.
> > si-list
> > > archives are accessible at http://www.qsl.net/wb6tpu/si-list ****
> > >
> > > **** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com.
> > In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP.
> > si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****
> >
> > **** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****
>
> --
> D. C. Sessions
> dc.sessions@vlsi.com
>
> **** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****

**** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 11:39:17 PST