Convergence

Multisim Help

Edition Date: February 2017
Part Number: 375482B-01
View Product Info

DOWNLOAD (Windows Only)

	Multisim 14.0 and 14.0.1 Help
	Multisim 14.1 Help
	Multisim 14.2 Help

The simulator uses the iterative Newton-Raphson method to solve non-linear equations which arise in practically all electronic circuits. The Newton-Raphson method involves successively linearizing and solving the circuit until the differences in the solutions between successive iterations become sufficiently small, a condition referred to as convergence. The Newton-Raphson method is used to find the solution to the DC operating point and to every single time point of a time domain (or transient) simulation.

Every analysis in Multisim runs either a DC operating point or a time-domain simulation, or both. For example, AC Analysis calculates the DC operating point around which it linearizes the circuit and performs a phasor analysis.

To learn more about how the Newton-Raphson method works, refer to the Linearization section.

Resolving Non-Convergence

In general, almost every aspect of the simulation, including models, analysis settings, and simulation options, affects the simulator's ability to achieve convergence. This is why resolving non-convergence is not always trivial. However, application of good modeling practices and an awareness of which options to tweak in which scenarios can significantly increase the success rate of convergence.

When encountering convergence problems, many users immediately make adjustments to the simulator rather than the circuit. Although this is a viable approach, it is not always the best. In terms of performance, accuracy, and overall stability of the fix, it is often more effective to make adjustments to the circuit.

Circuit Adjustment Recommendations

Remove unused or unnecessary components.

Unless you disable components using Variants, Multisim includes their models in the simulation. Unused or unnecessary components put unnecessary stress on the simulator, especially because they are often highly isolated from the ground node (0).

Use realistic values.

Do not use resistances with values less than 1µW. Also avoid using unrealistically small inductors and unrealistically large capacitors. These lead to very large conductances in the simulation matrix and may cause numerical difficulties which negatively impact convergence.

Use simpler models when possible.

Using simplified models not only aids convergence but also improves simulation speed. For instance, if you are studying the general topological behavior of a switching power circuit, consider using components with idealized models (found in the database in the Power group/SWITCHES family) to model the switching function of the MOSFET, IGBT, or diode elements.

Avoid isolation.

Circuits that are more isolated (that is, have a larger impedance to the ground node (0)) are more difficult to solve than those that are less isolated. Therefore, unless you are specifically simulating a function provided by isolation (for example, common-mode rejection), it is recommended that you avoid isolating circuits. For example, if you are simulating a fly-back converter, consider grounding the secondary side of the transformer as shown below, on the right.

Beware of P-N junctions.

Many components in the database contain semiconductor device models, which contain P-N junctions. Bias or drive these models as you would in real life. For example, the electrical model in some of the 7-segment display components is a diode. As you would in real-life, avoid driving this diode with a hard voltage source—use current-limiting resistors.

Simulator Adjustment Recommendations

The simulator provides options to configure how it solves the circuit. There are many options and some are much more effective than others.

Below is a list of options which are generally most effective at resolving convergence issues. Suggested values are also provided, although the values should take into account the ranges of values in the actual circuit. For instance, low-power sensitive analog circuits require tighter absolute convergence tolerances than high-power switching circuits.

Apply the suggested settings sequentially in the order shown, re-simulating as you apply each individual one.


Option Name	Suggested Setting	Comment
ABSTOL	1e-8	This loosens the absolute current tolerance. The default tolerance of 1e-12 is not required for many designs, particularly power circuits. Do not increase beyond 1e-6A.
VNTOL	1e-4	This loosens the absolute voltage tolerance. The default tolerance of 1e-6V is not required for many designs, particularly power circuits. Do not increase beyond 1e-3V.
RELTOL	0.01	This loosens the relative voltage and current tolerances. Do not increase beyond 0.01.
METHOD	Gear	This is the method used for numerical integration during a time-domain simulation. While the default Trapezoidal method is generally faster and more accurate, it can introduce instabilities which can ultimately degrade convergence. Switching the simulation method to Gear may help.
ITL4	100-500	This is the maximum number of iterations the Newton-Raphon method takes during the calculation of each time-domain time point before non-convergence for the time-point is declared. If the simulator cannot converge with 500 iterations, increasing the number of iterations further is unlikely to help.
ITL1	100-500	This is the maximum number of iterations the Newton-Raphon method takes during the calculation the DC operating point before non-convergence is declared. If the simulator cannot converge with 500 iterations, increasing the number of iterations further is unlikely to help
CONVLIMIT	Off	This option limits the change in value between Newton-Raphon iterations for certain models. While it is effective in keeping values within range, it can force an unnecessarily large number of iterations, which may exceed the iteration limits.

Refer to the Custom Analysis Options section to learn about other options and where simulation options are located.

WAS THIS ARTICLE HELPFUL?

Helpful

Not Helpful