Designer’s Guide Community

nus_lin

Community Member

Offline

Posts: 48

uncertain settling time with analog artist
Dec 01^st, 2005, 6:35pm

I found that the setup of transient analysis is tricky for the simulation of settling time. the settling time extracted is strongly dependent on the setup, say, for the default setup, the 0.1% settling time is 28ns, but if i manually set the maxstep to be 5ps, the settling time became 36.6ns, if i change other setup such as soving method, the value also change.

now my question is, which value is correct, which kind of setup has the best accuracy in terms of settling time extraction?

Back to top

IP Logged

Sezi

Junior Member

Offline

I LoVe ICs!

Posts: 12
Sweden

Re: uncertain settling time with analog artist
Reply #1 - Dec 2^nd, 2005, 3:38am

The settling time depends on the simulation step: the smaller the simulation step, the higher the accuracy. In real life, the simulation step corresponds to infinitely small time duration so I think if you set the simulation step smaller, your result will be more accurate, but will take longer time to simulate.

Back to top

IP Logged

sheldon

Community Fellow

Offline

Posts: 751

Re: uncertain settling time with analog artist
Reply #2 - Dec 2^nd, 2005, 3:58am

nus_lin,

It would help if you provided more information, for
example, the log file. The amount of variation in the
settling time that you are describing seems rather
large. How stable is the amplifier, is it ringing as it
settles? Is the simulation just the settling time of an
amplifier or ar other things events occuring in the
simulation? Are there any special devices in the
simulation, for example, an n-port? Also would it
be possible to list up the simulation conditions that
you tried and the results?

Best Regards,

Sheldon

Back to top

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: uncertain settling time with analog artist
Reply #3 - Dec 3^rd, 2005, 12:19am

Hi nus_lin,

The choice of the integration method and the accuracy setting have a significant impact on the calculation of the settling time or other transient timeconstants.

Ideally, the trapezoidal integration method is the best as it is lossless, and will also show you any potential overshoot and ringing. However, there is no reason why the gear2 method will not work effectively provided your accuracy settings are tight enough.

How tight they ought to be is something which depends on the target accuracy of the circuit you are simulating.

Please look up the chapter on Transient Analysis in the book "The Designer's Guide to SPICE and Spectre" by Ken Kundert for more details.

Regards
Vivek

Back to top

IP Logged