+ Post New Thread
Results 1 to 4 of 4
  1. #1
    Advanced Member level 2
    Points: 4,410, Level: 15

    Join Date
    Nov 2013
    Posts
    660
    Helped
    169 / 169
    Points
    4,410
    Level
    15

    ANSYS Totem electromigration check

    Hello everyone,

    I don't clearly understand the difference between average / RMS / peak electromigration (EM) results of ANSYS Totem.
    When I check power or signal EM I get the highest EM(%) in average EM mode, and the lowest EM comes in peak mode. It is strange for me.

    For example at a differential amplifier tail the average signal EM is 80%, RMS signal EM is 30%, peak is 0%.
    Or power EM of VDD33A rail is 40% in average, 30% in RMS, 20% in peak mode.

    How Totem solves average, rms, peak? Does it use else integration time or bandwidth limit or something else? Is it normal that the peak EM is the lowest?

    Thanks in advance.
    "Try SCE to AUX." /John Aaron/

    •   AltAdvertisement

        
       

  2. #2
    Advanced Member level 3
    Points: 5,634, Level: 17
    Dominik Przyborowski's Avatar
    Join Date
    Jun 2013
    Location
    Norway
    Posts
    743
    Helped
    311 / 311
    Points
    5,634
    Level
    17

    Re: ANSYS Totem electromigration check

    I have never used Totem tool.
    However, the EM/IR tools using dedicated technology files containing information provided by DRM, where some formulas are given together with limits. The result of simulation is shown in respect to these limits and peak limit is much higher than average/dc. It can be like 300µA/µm for average, 3mA/µm for RMS and 10mA/µm. Maybe limits provided by PDK are responsible for what you see?

    The issue can be also related to methodology. I have faced an issue with competitor tool, which uses simulation of x,y coordinated DSPF extracted netlist. If netlist was created with some post-layout optimization (like parasitic reduction) and/or the simulator uses parasitic reduction and/or doesn't preserves instances, the results were very inconsistent.



    •   AltAdvertisement

        
       

  3. #3
    Advanced Member level 2
    Points: 4,410, Level: 15

    Join Date
    Nov 2013
    Posts
    660
    Helped
    169 / 169
    Points
    4,410
    Level
    15

    Re: ANSYS Totem electromigration check

    Yes, these limits probably can answer the phenomenon, thank you very much for the clarification. I really didn't know that the PDK contain them, I will check it then.
    Methodology is another favour. Totem can product strange warnings and it doesn't like only R extraction for example and the solver fails sometimes, it more likes R+C or R+C+CC rather. The best issue was when it mixed the pins under mapping and VDD current flew into the bandgap reference input. Great software really, but I feel sometimes life was easier without it.
    "Try SCE to AUX." /John Aaron/



    •   AltAdvertisement

        
       

  4. #4
    Advanced Member level 2
    Points: 5,318, Level: 17
    Achievements:
    7 years registered

    Join Date
    Feb 2008
    Posts
    504
    Helped
    170 / 170
    Points
    5,318
    Level
    17

    Re: ANSYS Totem electromigration check

    Totem, Voltus (or VoltusFi), or any other EM / IR drop analysis tool simulates current flow in R or RC network (for power nets or signal nets).
    When it does dynamic EM / IR simulation, a current through each parasitic resistor is a function of time.
    For each resistor, the tool calculates average current (over time), RMS current, or peak current.
    These values are then compared with the critical current density values, different for each of these modes, and which may be complex functions of various parameters (such as temperature, metal line length, etc.).

    A percentage of violation (or similar characteristic) is reported by the tool.

    This is a top level description of how EM analysis is performed.



--[[ ]]--