Switching Signals
When an output buffer switches, it generates noise to the power network. In addition, there are other effects such as cross-talk and power noise inductive and capacitive coupling to the signals. This, in turn, creates the fluctuation on current. After a IO switching simulation, the designers can perform FFT analysis on the outputs of the IO drivers (Near End) and the inputs of the receiver IOs (Far End). The resulting noise spectrum indicate the potential problematic frequency components. In the meantime, the designers should run the identical simulation except idealizing the power and ground supplies to the drivers and receivers IOs. This way, we can isolate the PG noise and the crosstalk noise. Afterward, the designers can perform targeted correction on package and board layout as well as IO re-design.
Although the primary factor is the board design, the source of the noise is the IO. The higher frequency components of the power noise can only be addressed with on-chip decap or package decap as the on-board decap is too large to address the high frequency range.
Non-switching Signals
Even thought an output buffer is idle, it may contribute to EMI radiation. Assuming an output buffer is idle and the current output state is high. If there are noise to the power supply connecting to the last stage of the output drivers, the output of this buffer will vary along with the noise. The source of this power supply noise may be external or internal. External source may be global power delivery network noise and internal source may be inter-domain coupling (core power supply to IO power supply) and/or SSN.
The first phase of the Apache EMI development is to deliver a CPM encapsulating the inter-domain coupling noise. Subsequently, the designers can incorporate the IO spice circuit to the CPM along with the package and board model to perform a system-level simulation. This implementation is available now for vectorless CPM and it will be available in Sept. for vector CPM.
The first phase doesn't consider the potential noise coupling within the IO cells. The second phase is to continue to complete the model by characterizing the IO cells. This is planned for early October.
When an output buffer switches, it generates noise to the power network. In addition, there are other effects such as cross-talk and power noise inductive and capacitive coupling to the signals. This, in turn, creates the fluctuation on current. After a IO switching simulation, the designers can perform FFT analysis on the outputs of the IO drivers (Near End) and the inputs of the receiver IOs (Far End). The resulting noise spectrum indicate the potential problematic frequency components. In the meantime, the designers should run the identical simulation except idealizing the power and ground supplies to the drivers and receivers IOs. This way, we can isolate the PG noise and the crosstalk noise. Afterward, the designers can perform targeted correction on package and board layout as well as IO re-design.
Although the primary factor is the board design, the source of the noise is the IO. The higher frequency components of the power noise can only be addressed with on-chip decap or package decap as the on-board decap is too large to address the high frequency range.
Non-switching Signals
Even thought an output buffer is idle, it may contribute to EMI radiation. Assuming an output buffer is idle and the current output state is high. If there are noise to the power supply connecting to the last stage of the output drivers, the output of this buffer will vary along with the noise. The source of this power supply noise may be external or internal. External source may be global power delivery network noise and internal source may be inter-domain coupling (core power supply to IO power supply) and/or SSN.
The first phase of the Apache EMI development is to deliver a CPM encapsulating the inter-domain coupling noise. Subsequently, the designers can incorporate the IO spice circuit to the CPM along with the package and board model to perform a system-level simulation. This implementation is available now for vectorless CPM and it will be available in Sept. for vector CPM.
The first phase doesn't consider the potential noise coupling within the IO cells. The second phase is to continue to complete the model by characterizing the IO cells. This is planned for early October.
| Matt C. Elmore is a Certified Yoga Instructor, Internet Marketer, and Electrical Engineer. His passions include health, love, the pursuit of happiness, yoga, dance, and travel. Matt Elmore Yoga |
Posts
No comments:
Post a Comment