Hyperlynx far end crosstalk simulation issue

Sorry for the messy code. Have no idea what happened. Resent this email...Hi experts,It's known that compared with microstrip, an added benefit of stripline routing is zero far-end crosstalk (FEXT) because of a uniform Er surrounding the signal traces. However, the Hyperlynx simulation results tell me otherwise.I use Hyperlynx to simulate 5 stripline parallel signals (4 aggrs +1 victim) using point to point topology. The aggressor drivers are toggling simultaneously (all in phase) and the victim driver is stuck at high (let's say 1.2V). Theoretically the victim receiver should also stuck at high at 1.2V. However the Hyperlynx simulation results show that there is noise coupled to the victim receiver and the noise is related to the distance to the adjacent signals. Is my simulation wrong?Thanks and regards,Amanda
lmt8885 5 years 10 months 3 days

7 answers


The best answer


You can select the best answer for current question!
Answered bylmt8885 5 years 10 months 3 days
Sorry for the messy code. Have no idea what happened. Resent this email...Hi experts,It's known that compared with microstrip, an added benefit of stripline routing is zero far-end crosstalk (FEXT) because of a uniform Er surrounding the signal traces. However, the Hyperlynx simulation results tell me otherwise.I use Hyperlynx to simulate 5 stripline parallel signals (4 aggrs +1 victim) using point to point topology. The aggressor drivers are toggling simultaneously (all in phase) and the victim driver is stuck at high (let's say 1.2V). Theoretically the victim receiver should also stuck at high at 1.2V. However the Hyperlynx simulation results show that there is noise coupled to the victim receiver and the noise is related to the distance to the adjacent signals. Is my simulation wrong?Thanks and regards,Amanda
Answered byjeff.loyer 5 years 10 months 3 days
Hello Amanda,Your results agree with mine.  Adding multiple victims makes your "homogenous dielectric" non-homogenous.  If you repeat your simulations with only a single victim and aggressor, I think the FEXT would go to essentially zero.  Add more victims, and FEXT can become significant.  It will never get as large as for a corresponding microstrip configuration, but it's not zero.  The effect gets more pronounced for differential traces (especially dual stripline).On top of this, if you add different material properties for the core (below the trace) and the prepreg (above the trace), you'll find FEXT becomes even larger.If you then add a "resin pocket" (resin-only layer adjacent to the traces) to your simulations, you'll find FEXT magnitude and sign becoming dependent on your assumptions (both electrical properties and dimensions) for the core, prepreg, and resin pocket portions of the model.For more information on this, please see the work of Brandon Gore on slides 24-33 of the presentation available at the link below.https://www.filesanywhere.com/fs/v.aspx?v=8b6f628f5e61707ca3a5I believe that stripline FEXT is a topic we'll start hearing about more in the future.  Perhaps something we measure and screen for in future designs.Cheers,Jeff Loyer-----
Answered byRicky.Lai 5 years 10 months 3 days
Totally agreed with Scott.  It is the reflection of NEXT that you are seeing.You can easily do 4 experiments on the victim line to verify:Experiment (1): [un-terminated at near-end]============50ohm-trace==============[un-terminated at far-end]Experiment (2): [terminated with 50 ohm]===============50ohm-trace==============[un-terminated at far-end]Experiment (3): [un-terminated at near-end]============50ohm-trace==============[terminated with 50 ohm]Experiment (4): [terminated with 50 ohm]===============50ohm-trace==============[terminated with 50 ohm]-----
Answered bydbrooks9 5 years 10 months 3 days
AmandaYou might find these three articles interesting, especially the second one. It has animated simulations of what happens with forward and backward crosstalk signals travel and reflect.http://iconnect007.com/index.php/article/63809/crosstalk-why-its-difficult-to-understand-part-1/63812/?skin=designhttp://iconnect007.com/index.php/article/64264/crosstalk-part-2-what-it-looks-like/64267/?skin=designhttp://iconnect007.com/index.php/article/65059/crosstalk-part-3-more-on-shapes-and-amplitudes/65062/?skin=designDoug BrooksAt 02:55 AM 3/20/2015, 163 wrote:>Sorry for the messy code. Have no idea what happened. Resent this email...>>Hi experts,>>It's known that compared with microstrip, an added benefit of >stripline routing is zero far-end crosstalk (FEXT) because of a >uniform Er surrounding the signal traces. However, the Hyperlynx >simulation results tell me otherwise.>>I use Hyperlynx to simulate 5 stripline parallel signals (4 aggrs +1 >victim) using point to point topology. The aggressor drivers are >toggling simultaneously (all in phase) and the victim driver is >stuck at high (let's say 1.2V).>>Theoretically the victim receiver should also stuck at high at 1.2V. >However the Hyperlynx simulation results show that there is noise >coupled to the victim receiver and the noise is related to the >distance to the adjacent signals. Is my simulation wrong?>>Thanks and regards,>Amanda>
Answered byscott 5 years 10 months 3 days
AmandaMost likely you are seeing reflected reverse crosstalk.  That is NEXT thatpropagates back to a mismatched driver, reflects and becomes FEXT.  In mycrosstalk calculations for stripline, I always assume a 10% reflectioncoefficient of NEXT from a well-matched driver, more for a poorly matcheddriver, if I am doing a first cut on spacing design rules.ScottScott McMorrowConsultant - R&D16 Stormy Brook RdFalmouth, ME 04105(401) 284-1827 Businesshttp://www.teraspeed.comOn Fri, Mar 20, 2015 at 5:55 AM, 163  wrote:> Sorry for the messy code. Have no idea what happened. Resent this email...>> Hi experts,>> It's known that compared with microstrip, an added benefit of stripline> routing is zero far-end crosstalk (FEXT) because of a uniform Er> surrounding the signal traces. However, the Hyperlynx simulation results> tell me otherwise.>> I use Hyperlynx to simulate 5 stripline parallel signals (4 aggrs +1> victim) using point to point topology. The aggressor drivers are toggling> simultaneously (all in phase) and the victim driver is stuck at high (let's> say 1.2V).>> Theoretically the victim receiver should also stuck at high at 1.2V.> However the Hyperlynx simulation results show that there is noise coupled> to the victim receiver and the noise is related to the distance to the> adjacent signals. Is my simulation wrong?>> Thanks and regards,> Amanda> 
Answered byistvan.novak 5 years 10 months 3 days
Amanda,Only a well matched homogeneous structure will create no far end crosstalk.Reflections, for instance, will create 'noise' at the far end from reflected near end crosstalk.Non-homogeneous behavior also creates far-end crosstalk, but I assume you did not simulate those details.Regards,Istvan NovakOracleOn 3/20/2015 5:55 AM, 163 wrote:> Sorry for the messy code. Have no idea what happened. Resent this email...>> Hi experts,>> It's known that compared with microstrip, an added benefit of stripline > routing is zero far-end crosstalk (FEXT) because of a uniform Er surrounding > the signal traces. However, the Hyperlynx simulation results tell me > otherwise.>> I use Hyperlynx to simulate 5 stripline parallel signals (4 aggrs +1 victim) > using point to point topology. The aggressor drivers are toggling > simultaneously (all in phase) and the victim driver is stuck at high (let's > say 1.2V).>> Theoretically the victim receiver should also stuck at high at 1.2V. However > the Hyperlynx simulation results show that there is noise coupled to the > victim receiver and the noise is related to the distance to the adjacent > signals. Is my simulation wrong?>> Thanks and regards,> Amanda> 
Answered byweirsi 5 years 10 months 3 days
Zero FEXT does not mean zero NEXT or zero noise coupling.Steve.On 3/20/2015 2:55 AM, 163 wrote:> Sorry for the messy code. Have no idea what happened. Resent this email...>> Hi experts,>> It's known that compared with microstrip, an added benefit of stripline > routing is zero far-end crosstalk (FEXT) because of a uniform Er surrounding > the signal traces. However, the Hyperlynx simulation results tell me > otherwise.>> I use Hyperlynx to simulate 5 stripline parallel signals (4 aggrs +1 victim) > using point to point topology. The aggressor drivers are toggling > simultaneously (all in phase) and the victim driver is stuck at high (let's > say 1.2V).>> Theoretically the victim receiver should also stuck at high at 1.2V. However > the Hyperlynx simulation results show that there is noise coupled to the > victim receiver and the noise is related to the distance to the adjacent > signals. Is my simulation wrong?>> Thanks and regards,> Amanda>