need answers for this pd interview questions

[cyrax747]

how does tools do placement optimizations during placement.

why clock period is not required in hold time calculation

what is crosstalk window concepts

 

[gstekboy]

why clock period is not required in hold time calculation

Hold time depends on current clock edge (depends whether data can be sampled correctly by the current flipflop), so no need to consider extra one clock period.

 

[itsmeteja]

For hold analysis in general for the flops which are same phase we check the hold timing on the same edge i.e. zeroth edge and hence even if you increase or decrease the frequency your hold will not impact.

If you are having half cycle paths in the design rise to fall or fall to rise paths which means those launch and capture happens at different phases in such scenarios changing frequency may lead to hold violation. Hold failures not always frequency independent.

- - - Updated - - -

In general tools first try to global placement whose main objective is reduce wire length and incremental optimisation using global placement solution achieved in the first step for reducing/minimising either primary cost functions like trans/cap/timing or secondary cost functions like power/area/leakage. and then the detail placement to legalise the cells to valid legal cell locations.

- - - Updated - - -

In general if you have multiple aggressors then timing analysis must be done including the correlation of the multiple aggressors. STA obtains this information from the timing windows. Timing windows will have every net and their latest and earlier switching times will be present. These times represent the timing windows during which the nets might switch with in a clock period. timing windows provide necessary information on whether aggressor nets switch together or not.

To calculate the cumulative effect we might need to plot the switching time of the aggressors. Lets sat A1,A2,A3 are aggressor nets. We need to draw vertical lines whenever switching window starts or ends. Lets say A1 is active for 0-1ns and A2 for 0.5-2ns and A3 from 1.5 to 2ns. Crosstalk compositions for A1 is 0.100, A2- 0.200 and A3 - 300.

We need to draw lines at 0(A1 start) 0.5(A2 start) 1(A1 end) 1.5(A3 start) 2ns(A3,A2 end) So we have got 4 bins 0-0.5, 0.5-1, 1-1.5, 1.5-2. In all the bins calculate the total noise by simply adding the noise compositions. In bin1 A1 is only active 0.1
In bin2 A1,A2 are active 0.1+0.2=0.3
In bin3 A2 is only active 0.2
In bin4 A2,A3 -> total noise impact = 0.2+0.3= 0.5

Worst noise is at bin4 0.5

If the timing windows are not used in the design we call it as infinite window approach where all the aggressors are cumulative. So total noise at the victim would be 0.1+0.2+0.3=0.6 which is very conservative and pessimistic approach.