14 Replies Latest reply on Jun 23, 2017 5:49 AM by cathy_terwedow

    S-parameters extraction for differential traces on multi-board project

    jchou

       

      I have few questions regarding the use of S-paramters extraction for differential traces on multi-boards environments.

       

       

      1. After PCB post-layout translation into Hyperlynx, the two traces in diferential modes are not associated with each other. How do I change that ? So far, I only can simulate it by enabling crosstalk simulation. But I am not sure if this is correct.

       

       

      2. The differential trace impedance is 100 ohms. During S-parameter extraction, should I set the reference impedance to 100 ohms or 50 ohms ?

       

       

      3. I change the spice simulator in HyperLynx from Eldo/ADMS to Hspice, and the resultant S-parameters do output  with some difference (2db on some frequencies).

       

       

         How can I tell which one is correct ? Any suggestions ?

       

       

      4. For a connector on multi-board mating, the female connector is on one board while male connector is on another board. How do I use the connector model library on each board ? The connector model library is a result of both female and male connectors combined, not seperable, such as in EZ from Erni. 

       

       

        • 1. Re: S-parameters extraction for differential traces on multi-board project
          yu.yanfeng

          After assigning the model to the driver, Hyperlynx should automatically recongnizes the differantial pair. If You don't have the driver model, simply create a model which describing +/- pins.

          For reference impedece,it only related to the normalization ,so you decide which value. For Eldo/ADMS/Hspice's extacting S parameter, please take cares to the time step. For plug/socket model,

          You may use a single-pin model(RLC) to create EBD model for your board.

          • 2. Re: S-parameters extraction for differential traces on multi-board project
            Steve_McKinney

             

            Hi jchou,

             

             

            Sorry this is a late reply.  With the Christmas and New Year holiday, most Mentor people were out of the office.  Hopefully the responses below help:

             

             

            #1) the crosstalk method you are using is perfectly acceptable to create the coupling between the 2 traces.  You can also do as yu.yanfeng suggested  and apply a differential IBIS model.

             

             

            #2)For the reference impedance, I would recommend 50 ohms because this it the characteristic impedance of the traces.  The differential impedance is a mathmatical relation  between the 2 traces, however when you extract the traces, the reference for each trace will be ground and not the other trace.

             

             

            #3)There could be a couple of things happening here that could create that difference. Probably the main thing to consider is that  HSPICE and ADMS(HyperLynx's SPICE engine) use different models for dielectric loss.  I would compare the results against measurements if you have any (this could be done with a TDR and Tektronix's IConnect software if you don't have a VNA) in order to determine the validity of which one is more accurate.   Another factor causing differences may be in the handling of simulation of S-parameter data (I'm not sure if you had any s-parameters in your topology) since the 2 simulators use different algorithms for processing this data.

             

             

            #4) For a complex connector model, I would recommend exporting the net to LineSim, adding the connector model with the blackbox, and then extracting your s-parameter.  This is an easy way for you to include the full connector model for the particular net of interest.  There is some information in AppNote 5612 that describes how to use complex SPICE connector models in BoardSim which may be helpful for you.  AppNote 5612

             

             

             

             

             

            • 3. Re: S-parameters extraction for differential traces on multi-board project
              yu.yanfeng

              Hi Steve,

              When extracting S paramter in Hyperlynx, user should take to the dafault value of  signal edge in setup. I think Jchou's problem is related to this default value,normally it's 2 ns.

              If there is no ibis model there, Hyperlynx will use this default value.

               

              Yanfeng

              • 4. Re: S-parameters extraction for differential traces on multi-board project
                jchou

                Hi, Steve-

                 

                Thanks for your reply and the answers are very valuable.

                 

                I did try something as in AppNote5612.

                This method adds one more .hyp file for the connector but raise another question, what happened to original connector ?

                The original connector is still there and it still needs the same model library.

                The S-paramter extraction in this case becomes more complex too because it now has more signals to be extracted including new .hyp file.

                 

                I have one more question.

                 

                I have a HMZD connector on a PCB, and when I extract the S-parameters from the PCB using Boardsim, does the S-parameters output include the characteristic of the HMZD mounting holes on PCB ?

                 

                According to HMZD model library from Erni, the mounting hole characteristics is not included in the model.

                 

                If it is not, what are the suggestions ?

                 

                Thanks

                • 5. Re: S-parameters extraction for differential traces on multi-board project
                  jchou

                   

                  Hi, Yanfang-

                   

                   

                     Thanks for your suggestions. I did not use IBIS model and that may have some problems.

                   

                   

                    I will give it a try.

                   

                   

                  Thanks

                   

                   

                   

                   

                   

                   

                   

                   

                  • 6. Re: S-parameters extraction for differential traces on multi-board project
                    Steve_McKinney

                     

                    Hi Yanfeng,

                     

                     

                    You're right, that default edge is a factor for calculating the crosstalk coupling when no IBIS model is present.  If differential pair nets are not coupling with crosstalk turned out, then lowering this edge rate should generate the desired coupling.  You need to be a little careful though to make sure that you don't put the edge rate too low.  You can get too many nets coupled together which may not be your desired outcome.  With no model, you may have to work a little bit to find that sweet spot.

                     

                     

                     

                     

                     

                    -Steve

                     

                     

                    • 7. Re: S-parameters extraction for differential traces on multi-board project
                      Steve_McKinney

                       

                      Hi jchou,

                       

                       

                      When you setup your multiboard design, if you're using HyperLynx 7.7, you have the option to ignore connector parasitics.  If you do a "fake" board with the connector on it as in AppNote 5612, you want to select a "short" as your connector parasitics between the boards.  This means you are only seeing the connector parasitics due to the model you are using on that new .hyp file.

                       

                       

                       

                       

                       

                      As far as the more complex extraction, I recommend 2 options:

                       

                       

                      1) The easiest option is to not use the AppNote 5612 method and instead just export the net from BoardSim into LineSim and add your connector model there.

                       

                       

                      2)More complex, but still doable, is to only use a connector model for your pins of interest on that differential pair.  For example, if you have an 8 port s-parameter model, that means you have 4 signals total that could be going through that s-parameter.  Instead of using all 8 ports, just use the 4 ports that you need for the one differential pair.

                       

                       

                       

                       

                       

                      As for the mounting hole characteristics, the only suggestion I have would be to take your board to LineSim and add a via at that connector location to approximate the parasitics due to the mounting hole.

                       

                       

                      • 8. Re: S-parameters extraction for differential traces on multi-board project
                        jchou

                         

                        Hi, Steve-

                         

                         

                        Thanks for your answer.

                         

                         

                        It seems that HyperlYnx still needs some improvements over multi-board simulation environment.

                         

                         

                        If we have many signals to simulate, it will require effort to convert them from boardsim to linesim, add blackbox(connector), and add via(mounting hole) for each signal. That will be a time-consuming effort.

                         

                         

                        Is it possible to make it more friendly to use ?

                         

                         

                        Thanks

                         

                         

                        Jeffrey

                         

                         

                         

                         

                         

                         

                         

                         

                        • 9. Re: S-parameters extraction for differential traces on multi-board project
                          Steve_McKinney

                           

                          Hi Jeffrey,

                           

                           

                          Yes, there are areas for improvement here to make it more friendly and we have been taking some steps to help with this, but I'd say we still have some work to do.  The next release of HyperLynx (8.0) will address the need to create the .hyp file with the connector from LineSim and adding that extra board.  You'll be able to connect boards together with SPICE and S-parameter models right from within the Multiboard wizard which will help quite a bit.  This doesn't necessarily address all of the issues, but it is a step in the right direction to improve the ease of use.  One of the big problems we face with these large s-parameter models or SPICE models is mapping the connectivity through the model.  When HyperLynx connects 2 boards, the netlist builder has to be able to "read" through the models to determine the electrical net connectivity in order to connect the 2 nets on the different boards.  This is easy to do with an IBIS model or a series component like a resistor, but it's pretty much impossible for us to map through a SPICE model (there is no standard here and there can be a complex matrix of elements)  That means that in order to guarantee that we have electrical connectivity to connect the 2 boards together, we must include all the nodes or ports of the respective SPICE or S-parameter model.  As a result, you get a net topology with a bunch of nets in it...not ideal, but it's a limitation we have to navigate right now.  The hope would be that we can eventually either allow a user to provide the node connectivity map or we can figure it out on our own if you were using an s-parameter (maybe we can make some assumptions on connectivity between ports based on insertion loss at low frequency since it should be ~0dB for electrically connected ports).  Those are things we're looking towards in the future to make this process easier.  

                           

                           

                          If you are interested in trying out the new multiboard functionality for HyperLynx 8.0, it should be available in beta 3 in a few weeks.  Send me an email at steven_mckinney@mentor.com and I can get you setup with access to the beta.

                           

                           

                           

                           

                           

                          -Steve

                           

                           

                          • 10. Re: S-parameters extraction for differential traces on multi-board project
                            yu.yanfeng

                            Hi Steve,

                             

                            User should takes care to the default value of signal edge in steup if there is no real model there when trying to extract s-parameter. Hyperlynx 7.X's exportint S modle  is very sentitive to this value. For 2ns and 0.01ns, there is a huge difference betewen results. I think it was a bug of Hyperlynx 7.X.

                            BTW, I have seen Hyperlynx 8.0 Beta have corrected it.

                             

                            Yanfeng

                            • 11. Re: S-parameters extraction for differential traces on multi-board project
                              aleksandr.kupchik@oracle.com

                              Hello,

                               

                              Please provide an updated pointer to the Appnote 5612. Current pointer does not work.

                               

                              Alex

                              • 12. Re: S-parameters extraction for differential traces on multi-board project
                                cathy_terwedow

                                Alex - the link to appnote 5612 works. Do you have an active support contract?

                                • 13. Re: S-parameters extraction for differential traces on multi-board project
                                  milostnik

                                  Sorry Cathy

                                  It shows a non available page.

                                  I have active support. :-)

                                  Nor does it semm to find it

                                   

                                  Bye

                                    Matija

                                  • 14. Re: S-parameters extraction for differential traces on multi-board project
                                    cathy_terwedow

                                    Very strange. The link works perfectly for me. I forwarded your question internally to see if others can troubleshoot the problem. Thanks for reporting it, Matija.