1. 11 Measurements and Simulations with1.83-m RG58 cable S. Caniggia, P. Belforte April 5th 2013

2. 22 Outline Introduction Cable Studio (CS) and MicroCap10 (MC10) Measurements and simulations Conclusions References

3. 33 Introduction In [1] the signal integrity problems met by using cable lumped models in cable studio were investigated. It is shown that the non realistic fast oscillations on lossy cable waveforms can be solved by using the option “allow modal models”. The under estimations of the losses by CST 2012 are also investigated by comparison between measurements on actual 1.83-m RG58 coaxial cable and simulations. Simulations were performed by using Cable Studio 2012, MC10 (SPICE) and DWS (for more details see also [1])

4. 44 Cable Studio (CS) and MC10 (SPICE) (See also slides 12-22 of [1]

5. 55 Cable studio (cs) structure S11=VP5 S21=VP3-VP4

6. 66 MC10 (SPICE) structure Cascade of 610 3-mm unit RL-TL cell S11=VTin S21=VTout Step source with tr=100ps

7. 77 S11&S21: comparison between MC10&CS mc=Microcap; cs=cable studio Cable studio (CS) 2012 under estimates losses, see also slide 19 of [1] where lumped model was used. This did not occur with CS 2010, see also slide 18 of [1]. mc=Microcap; cs=cable studio For comparison, a delay of 0.904 ns were added to cs waveforms

8. 8 CS 2012: Dielectric losses (tanδ=0.8m) Ohmic losses S11 S21 sec Volt There are no significant slight differences

9. 99 Measurements and simulations (see also slides 23-36 of [1])

10. 10 Measurement: CSA803 setup detail 1.83-m RG58 coaxial cable

11. 11 S11 meas=measurements mc=Microcap; cs=cable studio Remark: simulations do not take into account measurement setup and discontinuity cased by cable and connector connections CS under estimates cable losses

12. 12 S21 meas=measurements mc=Microcap; cs=cable studio CS under estimates cable losses

13. 13 CSA803 setup DWS model

14. 14 S11&S22 366XRL-TL cells DWS vs actual measurement CSA803 Measurement setup effects included With actual cable S11 and S22 are not equal (cable is not symmetrical) Setup discontinuities and distributed impedance irregularities along the cable produce non-regular reflections

15. 15 S11 DWS 366xRL-TL cells and DWS Behavioral Time Model (BTM) vs actual measurement CSA803 Measurement setup effects included With a behavioral model we can take into account the cable irregularities BTM model RL-TL model

16. 16 DWS 366XRL-TL vs measure including setup S21 edge MC10 (no setup) vs measurement

17. 17 S21 edge (zoom) DWS 366XRL-TL vs measure including setup (30ps risetime of TDR waveform) MC10 (no setup) vs measurement Some differences (due probably to dielectric losses) can be pointed when setup discontinuities are taken into account in the whole model (upper plot). These differences are less evident with 100ps risetime with no setup in the model (lower plot).

18. 18 Comments on measurements and simulations CS under estimates losses in time and frequency domain simulations also taking into account dielectric effect. This is confirmed by slide 8 of [2] where the attenuation at 1000 MHz computed by 2DTL, T! and F! solvers of CST 2013 is less or equal to 0.02 dB for 5-cm of RG58 cable, while the cable used for measurement has a nominal attenuation of 51.8/100x0.05=0.026 dB at 1 GHz (see slide 28 of [1]). RL-TL model considers both skin and proximity effects and gives a conservative value of attenuation of 0.032 dB (see slide 10 of [1]) For this reason, RL-TL model provides very good accuracy with measurements for edges not faster than100 ps (1.83m cable) nevertheless dielectric losses are neglected. For edges faster than 100 ps, one should consider that above 2.5 GHz the dielectric loss dominates on skin effect and discontinuities due to the setup should modeled for good accuracy

19. 19 Conclusions The unrealistic oscillations on sending edge signal computed by CST Cable Studio (CS) when using lumped model for lossy cables [1] can be avoided by using the option of “allow modal models” in 2D (TL) modeling settings. Unfortunately, the under estimation of losses with CST CS 2012 still remains as evidenced by comparison between measurement and simulations. Taking into account dielectric losses with tan(δ) up to 0.8m (PE material), losses do not change significantly. Good agreement with measurements occurs when simulations are performed by: CST CS 2010, MC10 and DWS (50X faster than MC10) with RL-TL or BTM models (fastest DWS solution that can take into account actual measures) [1]. CST should continue to investigate these last three items for CST 2013 version because the frequency domain results shown in [2] seem to confirm that the problem has not yet been resolved (see the comments of previous slide)

20. 20 References [1] Piero Belforte, Spartaco Caniggia, “CST coaxial cable models for SI simulations: a comparative study”, March 24th 2013 [2] Leonardo Sassi, “CST CABLE STUDIO 2013, Coax Cable analysis, Report from S. Caniggia”, file: CST CS 2013_SI_Isa002(1).pdf, april 2013