1 Reply Latest reply on Oct 15, 2008 3:44 PM by Ed Bartlett

    Optimum Time Resolution in ICX?

    Lorenz.Scheurer

      I simulate with ICX for years. But I don't found information about the value of time resolution in Checking sheet. I get extreme different crosstalk results depending on different time resolution settings from 20ps to 1ps. Has somebody the same experience with ICX?

      -Lorenz

        • 1. Re: Optimum Time Resolution in ICX?
          Ed Bartlett

          For setting the time resolution I look at the transition time of the driver rise and fall, then find the fastest of the two, and calculate time resolution = Tr/20. Next, I look at the shortest segments in my routing and calculate the approximate time delay of these segments with length ( in inches) * 180ps/in to get another candidate of time resolution. Finally, I choose the smaller of these two numbers and use that for the time resolution.

           

          If you prefer a trial and error method, for a delay simulation, keep cutting the time resolution in half until the difference in results is negligible. This method is manageable if the simulation type is delay, but not if it is crosstalk, due to the variables introduced as discussed in the paragraph below.

           

           

           

           

           

          For crosstalk simulations, the time resolution set in the checking sheet affects how traces are modeled in ICX. You can expect to see some variation in the results based on changing the time resolution setting. Traces significantly smaller than the time resolution are removed from the simulation. So by making the time resolution larger, you can exclude some small segments in the net. In addition, for crosstalk simulations, the time resolution works in conjunction with the coupling settings. For instance- if you change the time resolution from a larger setting to be smaller- this causes a smaller segment to now be included. Since this new segment is included it could couple to other nets. If the coupling settings for your crosstalk simulation allow this new smaller segment to couple in another aggressor net, the final crosstalk value could be affected by adding this new aggressor net to the crosstalk simulation. Hence the time resolution and the coupling settings can include or exclude aggressor nets and thus change the final crosstalk results.

           

           

          As a side note, you can get an idea about the complexity of the netlist used in the crosstalk (or delay) simulation by observing the Hyperlynx output (i.e. for crosstalk, set the ICX Probe Net Control to the Default Crosstalk, As for Checking, then use File > Export > HyperLynx) You can now observe the schematics in Hyperlynx, with all the coupled segments for that simulation done in ICX. If you don't have HyperLynx installed, you can still export SPICE, or ICX Pro Explorer from ICX to inspect the segments in the simulation.

           

           

          Finally, in ICX version 3.9 the trace modeling is planned to be improved for short transmission lines, which will allow for even more accurate simulations when there are lots of small transmission lines in the nets being simulated.

           

           

          I hope this is useful information for you.

           

           

          Best regards,

           

           

          Ed Bartlett

          1 of 1 people found this helpful