Design Article
Tell us What You Think
We want to know what you thought about this Design. Let us know by adding a comment.
Removing pessimism and optimism in timing analysis
Naman Gupta - Freescale Semiconductor
10/1/2012 10:29 AM EDT
Calculating pessimism
Pessimism due to different transition sense in the common clock path
Consider the case when a positive-edge triggered FF is talking to a negative-edge triggered FF. In this case, the transition sense for the cells constituting the common clock path would be different for launch and capture. Rise and fall times of the cells can be different. Hence we need to take into account the difference between the derated rise and fall times of the cells while calculating CPPR.
Now, consider two possibilities:
Case I: Take the late delay for launch (at rise edge, in the above case) and early delay for capture (at fall edge, in the above case).
Setup Slack would be: -1.1 + 1.8 = 0.7
However, the premise for this calculation is the assumption that BUF1 exhibits late and early derates simultaneously. This is not a plausible scenario and hence adds extra pessimism in the timing analysis.
Case II: To avoid extra pessimism in timing analysis, for the same cell in the common clock path, one should take either late or early derates for both the launch and capture paths.
Two possible cases:
Assuming Late derates for both launch and capture:
Worst case for setup would correspond to the one with minimum clock skew. In the above example, early case presents are more critical scenario with lesser skew.
Setup Slack would now be: -0.9 + 1.8 = 0.9.
Hence Case II has lesser pessimism than Case I.
Pessimism due to different transition sense in the common clock path
Consider the case when a positive-edge triggered FF is talking to a negative-edge triggered FF. In this case, the transition sense for the cells constituting the common clock path would be different for launch and capture. Rise and fall times of the cells can be different. Hence we need to take into account the difference between the derated rise and fall times of the cells while calculating CPPR.
Figure 4: Different transition sense for launch and capture clock path at BUF1
Now, consider two possibilities:
Case I: Take the late delay for launch (at rise edge, in the above case) and early delay for capture (at fall edge, in the above case).
Setup Slack would be: -1.1 + 1.8 = 0.7
However, the premise for this calculation is the assumption that BUF1 exhibits late and early derates simultaneously. This is not a plausible scenario and hence adds extra pessimism in the timing analysis.
Case II: To avoid extra pessimism in timing analysis, for the same cell in the common clock path, one should take either late or early derates for both the launch and capture paths.
Two possible cases:
Assuming Late derates for both launch and capture:
Worst case for setup would correspond to the one with minimum clock skew. In the above example, early case presents are more critical scenario with lesser skew.
Setup Slack would now be: -0.9 + 1.8 = 0.9.
Hence Case II has lesser pessimism than Case I.
|
|
|
|
|
Navigate to related information