setup/hold windows are greater than the receivers setup/hold requirements, then the timing
margin is positive. Furthermore, the ringback margin can be determined simply by observing
how close the signal rings back toward the threshold voltage. It should be obvious that eye
diagrams are not particularly suited for sweeping, or for evaluation of flight time or flight-time
skews. However, many engineers prefer to utilize them in the later stages of the design.
Figure 9.24: Example of an eye diagram.
Phase 2 ISI.
The evaluation of ISI can be tricky. To do so, it is necessary to perform targeted simulations
at the edge of the design space using a long pseudorandom pulse train. If pattern-dependent
timing or signal quality violations occur, steps should be taken to minimize reflections on the
bus; this will reduce the ISI. Usually, the best way to limit reflections on the bus is to tighten
up the impedance variations and minimize discontinuities by shortening stubs and
connectors and matching impedances between packages and PCBs. The results of the
ordered sweeps and the significance list should be quite helpful at this stage. Also, be sure
to include any startup strobe ISI in the analysis. Usually, strobes do not exhibit ISI because
they operate in a single pattern. However, often the strobe is turned off until it is time for the
component with which it is associated to drive signals onto the bus. It will take several
transitions for the strobe to reach steady state. If the bus protocol allows valid signals to be
driven onto the bus before the strobe can reach steady state, the strobe ISI can significantly
affect the timings by increasing skew.
The worst-case ISI can be evaluated using the following steps:
1. Simulate the longest net in the bus with the most impedance discontinuities using a
long pseudorandom bit pattern for each of the worst-case conditions (i.e., they should
be at the edge of the solution space so far defined).
2. Take the flight time of the first transition of the ISI simulation as a baseline.
3. Determine the rising and falling flight times for each bus transition (using the standard
load as a reference).
4. Subtract the minimum and maximum delays from the baseline delays and find the
worst-case difference.
5. Take the smallest negative and the greatest positive difference. This should be the
worst-case ISI impact on timings.
6. Simulate the maximum and minimum flight time and flight-time skew corners with a
very slow switching rate that settles completely prior to each transition and add the
numbers from step 5 to the results. This should represent the worst-case phase 2
flight time and flight-time skew numbers that should be included in the spreadsheet.
Remember to account for the startup ISI on the strobe if necessary.