Average from Sweep Simulation Spectre

How are you doing these sweeps? How are you "keeping track of the various delays"? Not enough information here to really guide you - I'd have to guess or try to cover every possible way you might be running the simulations and doing your calculations.

Andrew.

How are you doing these sweeps? How are you "keeping track of the various delays"? Not enough information here to really guide you - I'd have to guess or try to cover every possible way you might be running the simulations and doing your calculations.

Andrew.

Hello, this is basically my file:

simulator lang=spectre

swp_v1 sweep paramset=data_input {

swp_v2 sweep paramset=data_input2 {

         tran_x tran start=0ns stop=4.1ns

}

}

simulator lang=spice

.MEASURE TRAN A1_S1 TRIG v(102) VAL=0.5 TARG v(120) VAL=0.5

.MEASURE TRAN A0_S1 TRIG v(105) VAL=0.5 TARG v(120) VAL=0.5

I'm not that familiar with the SPICE .MEASURE syntax, but I don't think you can find the average across a sweep. Similarly I don't think this can be done with Spectre MDL either (I tried, but got the same result as doing it with a .MEASURE):

Either:

alias measurement finddelay {
  export real a1_s1, a0_s1

  run swp_v1

  a1_s1=deltax(sig1=v(102),sig2=v(120),thresh1=0.5,thresh2=0.5)
  a0_s1=deltax(sig1=v(105),sig2=v(120),thresh1=0.5,thresh2=0.5)
}

run finddelay

or

alias measurement finddelay {
  export real a1_s1, a0_s1

  run tran(stop=4.1n)

  a1_s1=deltax(sig1=v(102),sig2=v(120),thresh1=0.5,thresh2=0.5)
  a0_s1=deltax(sig1=v(105),sig2=v(120),thresh1=0.5,thresh2=0.5)
}

foreach data_input {
  foreach data_input2 {
    run finddelay
  }
}

Both do the same thing (and adding an avg() in the measurement doesn't help). Similarly with .MEASURE adding:

.MEASURE TRAN A1_S1_AVG AVG A1_S1
.MEASURE TRAN A0_S1_AVG AVG A0_S1

doesn't help as it gives the average per sweep point, which is the same as the measurement itself.

So I can't see a way of doing it. I'd probably just post-process the ".mt0" file with the tabular measurement results and compute the average outside. Or contact customer support to see if we can add an enhancement to support this (or maybe there's a way that I missed from a very quick look into this)?

Regards,

Andrew

Ok I did the "postcomputation" using a Matlab script importing the .mt0 file.
However, now there is another problem: the number of bits (and so the possible input combinations) has increased, and it takes forever to try to analyse everything.
At the same time, I know that Genus does not provide an accurate timing report in case of clock absence (like in my case of asynchronous adders).

Do you maybe know whether it is possible to have precise results with the synthesizer?

Regards,

Simone.

If you have a question about Genus, please post a new post in the Logic Design (or maybe Design Implementation) forum. Custom IC Design is not the right forum for such a question.

Regards,

Andrew.