SIMULATOR OPTIONS REFERENCE

Overview

This topic lists every option on the simulator options dialogue form, tab by tab, with the property name it sets and the value it takes if you leave it alone. The final section covers the properties that have no control on the form and can only be set on the design.

For how to reach the form and which simulations each route affects, see Setting Simulator Options.

Tolerances

This group of parameters determines how accurately SPICE computes the solution. Higher accuracy is generally achieved at the expense of longer simulation times, and in some circumstances the circuit may fail to converge at all if you ask for too high a set of tolerances.

Field Property Default Effect
Absolute current error tolerance (Amps):

ABSTOL

1e-12

The absolute accuracy to which branch currents must agree between successive iterations before the solution is accepted.
Absolute voltage error tolerance (Volts):

VNTOL

1e-6

The same test applied to node voltages.
Charge error tolerance (Coulombs):

CHGTOL

1e-14

The same test applied to the charge on capacitive elements.
Relative error tolerance:

RELTOL

0.001

The proportional accuracy demanded of every solved quantity, applied alongside the absolute tolerances above. This is the single most useful tolerance to relax when a circuit will not converge.
Minimum acceptable pivot value:

PIVTOL

1e-13

The smallest matrix element the solver will accept as a pivot. Raising it can get a matrix that reports as singular to solve.
Minimum acceptable ratio of pivot:

PIVREL

1e-3

The smallest acceptable ratio between a chosen pivot and the largest element available in the same column.
Minimum conductance (Siemens):

GMIN

1e-12

The conductance placed across reverse biased semiconductor junctions and other theoretical points of infinite impedance.
Minimum transient conductance:

TRANGMIN

1e-9

The same quantity as applied during a transient analysis, and it moves in the same direction.
Shunt Resistance (Ohms):

RSHUNT

1e12

A resistance connected from every voltage node to ground. This is the mechanism that gives a DC path to nodes which would otherwise have none, such as the far side of a coupling capacitor.

GMIN is the option most often reached for when a circuit will not solve, and it is easy to move it the wrong way. Increasing it may help achieve convergence for circuits which fail to simulate, at the expense of accuracy. Values above 1E-9 give meaningless results, so that is the ceiling worth trying. Reducing it below the default is the accuracy direction, not the convergence direction.

See Also:

Convergence Problems

Ensuring a DC Path to Ground

MOSFET

SPICE simulation of MOSFETs is based on the assumption that you are doing IC design, and consequently implements a scheme for scaleable geometries. In practice this means that there are a number of parameters which determine the default physical sizes for elements of the MOSFET devices, used where the model itself does not state them. Those values are defined here, along with two compatibility switches for models written against the behaviour of older versions of SPICE.

Field Property Default Effect
MOS drain diffusion area (Metres²):

DEFAD

0

Default drain diffusion area.
MOS source diffusion area (Metres²):

DEFAS

0

Default source diffusion area.
MOS channel length (Metres):

DEFL

100u

Default channel length.
MOS channel width (Metres):

DEFW

100u

Default channel width.
Use older MOS3 model?

BADMOS3

off

Restores the earlier level 3 MOSFET equations for models that were characterised against them.
Use SPICE2 MOSFET limiting?

OLDLIMIT

off

Restores the SPICE2 voltage limiting algorithm used while iterating MOSFETs.

Iteration

The properties on the Iteration tab determine how SPICE deals with circuits that are hard to converge, and how hard it tries before giving up.

Field Property Default Effect
Integration method:

METHOD

GEAR

Either GEAR or TRAPEZOIDAL. Gear integration generally gives more accurate results for a given number of timesteps; trapezoidal is provided mainly for compatibility with earlier versions of SPICE.
Maximum integration order:

MAXORD

2

How much of the past history of a timepoint SPICE may use in predicting the next one. With Gear integration the value is held between 2 and 6; with trapezoidal integration it is forced to 2.
Number of source steps:

SRCSTEPS

120

The number of steps used when the sources are ramped up from zero to find the operating point.
Number of GMIN steps:

GMINSTEPS

120

The number of steps used when GMIN is stepped down from a large value to find the operating point.
DC iteration limit:

ITL1

100

Maximum iterations allowed at the operating point.
DC transfer curve iteration limit:

ITL2

50

Maximum iterations allowed at each step of a transfer analysis.
Upper transient iteration limit:

ITL4

10

Maximum iterations allowed at each timepoint of a transient analysis. Raising the three limits may get a result out of a hard or near unstable circuit.
Go directly to GMIN stepping?

NOOPITER

off

Skips the direct attempt at the operating point and starts with GMIN stepping, which saves time on a circuit you already know will need it.
Try compaction for LTRA lines?

TRYTOCOMPACT

off

Applies only to circuits incorporating lossy transmission lines (the LOSSYLINE model). Near identical values in the data pipeline are discarded so that fewer data points have to be processed.
Allow bypass on unchanging elements?

BYPASS

on

Saves SPICE from re-computing the values of semiconductor devices whose node voltages have not changed since the last evaluation.

Temperature

Field Property Default Effect
Operating temperature (°C):

TEMP

27

The actual temperature of the circuit. Individual components can override it with a TEMP property of their own.
Parameter measurement temperature (°C):

TNOM

27

The temperature at which temperature dependent device parameters are taken to have been measured.

See Also:

Temperature Modelling

Transient

These four settings govern the timestep control used by transient analysis, and by the transient engine underneath the other time domain analyses.

Field Property Default Effect
Number of Steps:

NUMSTEPS

50

Divides the analysis interval to give a ceiling on the timestep, so that a long analysis cannot take steps so large that it misses the detail.
Truncation error over-estimation factor:

TRTOL

7.0

How much SPICE is allowed to over-estimate the error it makes in taking a step. If you have results which appear overly spiky, or suffer from overshoots, try a lower value here.
Mixed Mode Timing Tolerance:

TTOL

1e-9

The tolerance to which analogue timepoints and digital events are synchronised in a mixed mode simulation. It is consumed by the interface models, so it is the value to look at when analogue and digital halves of a circuit appear to disagree about when something happened.
Minimum Analogue Timestep:

TMIN

1e-18

The floor on the analogue timestep. A simulation that tries to step below it aborts rather than grinding to a halt.

DSIM

The DSIM tab carries two groups of radio buttons rather than a list of fields, because the digital settings are modes rather than values. Each mode corresponds to a combination of properties, and knowing the mapping lets you set the same thing from the design instead.

Propagation Delay Scaling

The TDSCALE variable controls the scaling of all timing properties used by models in the simulation run that have not been explicitly referenced by the model as non-scaleable. It can be assigned either a constant floating-point value or the keyword RANDOM, and the four radio buttons in this group select between those cases.

  • Scale all values by constant amount: sets TDSCALE to the number in the box beside it.
  • Fully random scaling. sets TDSCALE=RANDOM with no seed, so the run differs every time.
  • Pseudo-random scaling based on seed: sets TDSCALE=RANDOM together with TDSEED set to the number you give.
  • The random scaling is limited to a range: sets TDLOWER and TDUPPER.

If a constant value is specified, then all timing properties specified in models used in the simulation are multiplied by the value; a value less than 1.0 reduces the timing properties whilst a value greater than 1.0 increases them. For example, the variable assignment:

TDSCALE = 1.1

has the effect of extending all timing properties used in the simulation by 10%. The default value for TDSCALE is the constant value 1.0; this causes all timing properties to be used unmodified.

If the TDSCALE variable is assigned the keyword RANDOM, the DSIM engine will randomly scale each timing property by multiplying it by a random floating-point time-scaling value. The range of time-scaling values chosen by the simulation engine can be limited by assigning the TDLOWER and TDUPPER variables; these define the lowest and highest random floating-point scale values allowed respectively. The default values for TDLOWER and TDUPPER are 0.9 and 1.1 respectively, which has the effect of limiting random scaling to within 10% either way.

The sequence of random values generated is said to be pseudo-random, that is, each successive value generated, whilst appearing to be random, is in fact determined by the previous value and a complex formula. It follows that any sequence of random values is determined by a seed value and that for a given seed value the sequence of random values generated is always the same. The TDSEED property allows you to specify a seed value for the generation of random time-scaling values and thus guarantee that the sequence chosen from one simulation run to another is always the same. This avoids the problem of a timing design error based on random propagation delays appearing and then disappearing on successive simulation runs.

The TDSEED variable should only be assigned positive integer values in the range 1 to 32767. The default value for TDSEED is itself a random value based on the date and time, and this provides a means for randomising the sequence of values generated from one simulation run to the next.

For example, the assignments:

TDSCALE = RANDOM

TDLOWER = 1.50

TDUPPER = 2.00

TDSEED  = 723

have the effect of randomly increasing all timing properties by between 50% and 100% with a random sequence of scaling values. The seed value of 723 causes the same sequence of pseudo-random values to be generated for each simulation run.

The dialogue form's scaling limits stop at 2, so a scale factor above that has to be set as a design property rather than through the form. TDLOWER must not be negative and TDUPPER must be greater than it, or the simulation stops with "Bad TDLOWER/TDUPPER values."

Random Initialisation Values

The INITSEED property is used to seed the random initialisation value generator, used by DSIM primitive models whose initialisation properties have been assigned the RANDOM keyword. The two radio buttons in this group choose between:

  • Fully random values:, which leaves INITSEED unset. The simulator then seeds itself from the date and time, so every run starts differently.
  • Pseudo-random values based on seed:, which sets INITSEED to the number you give, in the range 1 to 32767, so that every run starts identically.

As with the TDSCALE and TDSEED properties described above, whilst the sequence of values generated by the random initialisation value generator is random, the sequence as a whole is finite and determinate. The INITSEED property provides a means of selecting which sequence of random values is used by the DSIM simulator.

Properties without Dialogue Controls

A number of live simulator properties have no control on the dialogue form. Set them by giving the design a property of the name shown.

Property Default Effect

MAXBOOTEVTS

128

The maximum number of iterations DSIM will make at the boot timepoint while settling the circuit into its initial state.

MAXRACEEVTS

32

The maximum number of iterations allowed at any one timepoint after boot. Exceeding it is what DSIM reports as a logic race condition.

MAXRUNEVTS

1000000

The maximum number of digital events in a single simulation run. A design that trips it is usually oscillating rather than genuinely busy, but raise it if you know otherwise.

MINBREAK

computed

The minimum time SPICE will allow between breakpoints. It defaults to a small fraction of the maximum timestep and is also held below the mixed mode timing tolerance.

LOGSTART

0

Simulated time at which messages start being written to the simulation log. Use it to skip a noisy start-up.

LOGTIME

unlimited

How much simulated time is logged from LOGSTART onwards. Together the two bound a window of interest in a long run.

VTHERM

derived

The thermal voltage, derived from TEMP for the use of schematic models. See Temperature Modelling.

GEARTH

1E-8

The conductance of the earth resistor synthesised for a net with no DC path to ground. Zero disables the repair. See Ground and Power Rails.

DCPATH

none

A per-component override of the DC conduction paths built into the device type. See Ground and Power Rails.

GLEAK

1E-12

Capacitor leakage admittance, given on one capacitor or on the design. The default is only supplied to a design that carries neither RSHUNT nor a GLEAK of its own.

RLEAK

none

Capacitor leakage as a resistance. Takes precedence over GLEAK where both are given.

See Also:

Setting Simulator Options

Ensuring a DC Path to Ground

The Thermal Voltage

Diagnostic Messages