you have to think in terms of how the aocv table has been generated. The aocv is generated using back to back one inverter, two inverter and three inverter. So when the delay says 1st stage then variation is 0.9 and the delay gets multiplied by 0.9 ....when two are there then the variation of delay is multiplied by 0.95 ....and so on..... now your AOCV table has only depth but there is another parameter which is distance also included in some characterization tables..... Now these have been bundled in the LVF format ( Library Variation Format)....
Hi artmalik, thanks a lot.
I know what you say, and I want to check(re-generate) aocv table by hspice.
But I find a question here. Because the aocv table is generated by what you say.
If we want to calculate the real timing ,it should be
(liberty timing *stage number) * (the aocv data of that stage number)
by my example it should be
if 2 stage :
(20 *2) * 0.95
if 3 stage :
(20 *3) * 0.97
But by the timing report of PT, the data is (liberty timing * the aocv data of that stage 1) + (liberty timing * the aocv data of that stage 2) + (liberty timing * the aocv data of that stage 3)
by my example it is (20 * 0.9) + (20 + 0.95) + (20 * 0.97)
PT: (20 * 0.9) + (20 + 0.95) + (20 * 0.97)
reverse aocv table : (20 * 3)*0.97
Because it is not equal , I have this question. Thank you!!