3 Replies Latest reply on Nov 22, 2017 7:28 AM by weston_beal

    Plane noise analysis

    tyle

      Hi expert,

       

      Recently, I have been studying plan noise simulation in Hyperlynx but I don't know how to specify an AC model for power pins.

      I have searched some materials but all of them don't mention about that

      Could you please give me a help?

       

      Thanks,

      TY

        • 1. Re: Plane noise analysis
          weston_beal

          Is your question about how to assign a model to a power pin, or is it about how to determine the parameter values for an AC power pin model?

          • 2. Re: Plane noise analysis
            tyle

            Hi Weston,

            It is about how to determine the parameter values for an AC power pin.

            Thanks,

            TY

            • 3. Re: Plane noise analysis
              weston_beal

              There is some good information in this technote on Support Center.

              https://support.mentor.com/en/product/861057060/knowledge-base/MG560965

               

              This is the typical "it depends" answer. It depends on what aspect of the PDN you want to study. The ideal situation would be to get current waveforms measured at each pin of each component, and apply that waveform as the pin model in the simulation. The biggest problem with this method is measuring the current into the pins. Since this measurement is not common, the simulator doesn't support it.

               

              Remember that the major effect of power supply noise is L * di/dt. Usually, you want to model the di/dt as closely as possible. The core current is mostly constant, so the pulses that cause problems are usually from the output buffers switching. Start with the triangle pulse model. make the pulse width (rising + falling) equal to the switching time of the output buffers, and the amplitude about equal to the current through half the switching buffers. This assumes that about half the buffers switch simultaneously. If you want to be cautious, you can increase that factor. If you have a secondary set of buffers that behave differently, then you can use the double triangle pulse to model the second bus in the second pulse. These models have a major di/dt component, but also have some lower amplitude frequency content up to high frequencies because of the sharp corners of the waveform.

               

              If you suspect a PDN resonance at a specific frequency, then use the sinusoidal current pin model to stimulate the potential resonance at a certain frequency. This model waveform has a single frequency instead of the broad frequency band of the other models.

               

              In summary, get whatever information you can about the current profile of the device you are modeling, and use the model that best approximates that profile.

              1 of 1 people found this helpful