In this section we will only cover the simple case of how to specify tolerances on devices at the schematic level for use with Monte Carlo, sensitivity and worst-case analyses. SIMetrix has much more comprehensive features for specifying tolerances. For complete documentation on tolerance specification please refer to the Simulator Reference Manual/Monte Carlo, Sensitivity and Worst-case/Specifying Tolerance.
In this topic:
To select individual device tolerances proceed as follows:
Device tolerance will be ignored for other devices.
.MODEL NPN1 NPN IS=1.5e-15 BF={180*GAUSS(0.5)}
The GAUSS function above returns a random number base around 1.0 with a 3???MATH???\sigma???MATH??? Gaussian distribution of 0.5. (Gaussian distribution with standarad deviation of 0.5/3).
There are many other distribution functions including a user defined function which can be used for any distribution. In addition you can use parameter definitions to define random variables to create matched devices and parameters. Full details may be found in the Simulator Reference Manual/Monte Carlo, Sensitivity and Worst-case/Specifying Tolerances.
Some devices such as resistor networks are constructed in a manner that their tolerances track. Such devices often have two specifications one is an absolute tolerance and the other a matching tolerance. A thin film resistor network might have an absolute tolerance of 1% but a matching tolerance of 0.05%. This means that the resistors will vary over a +/-1% range but will always be within +/-0.05% of each other.
To specify matched devices for Monte Carlo analysis two pieces of information are required. Firstly, the parts that are matched to each other must be identified and secondly their matching tolerances need to be specified.
To specify device match tolerances, proceed as follows:
The default distribution for device tolerances is Gaussian with the tolerance representing a ???MATH???3\sigma???MATH??? spread. This can be changed to rectangular using two simulator options. These are:
MC_ABSOLUTE_RECT | If set absolute tolerances will have a rectangular distribution. |
MC_MATCH_RECT | If set matching tolerances will have a rectangular distribution. |
Distributions can be specified on a per part basis or even a per parameter basis by using distribution functions in an expression. See Simulator Reference Manual/Monte Carlo, Sensitivity and Worst-case/Specifying Tolerances for details.
◄ Sensitivity/Worst-case Example | Running Monte Carlo ▶ |