Skip to content

ideal-diode.md

Latest commit

 

History

History
1023 lines (757 loc) · 32.3 KB

ideal-diode.md

File metadata and controls

1023 lines (757 loc) · 32.3 KB

LTspice-Style Ideal Diode

SpiceSharpParser.CustomComponents contains an opt-in ideal diode component for LTspice-style diode models that use parameters such as Ron, Roff, and Vfwd.

This component is separate from the built-in SpiceSharp diode. The built-in diode uses the normal semiconductor exponential model. The custom ideal diode uses a piecewise linear current law. It is useful for power electronics, protection clamps, rectifiers, and behavioral-level circuits where a simple on/off diode is more useful than a detailed PN junction model.

Compatibility Snapshot

Area Status
Model detection Selected when a .MODEL D(...) contains Ron, Roff, Vfwd, Vrev, Rrev, Ilimit, RevIlimit, Epsilon, or RevEpsilon.
.OP / .DC Uses an LTspice-style region-wise-linear current law with forward, off, reverse-breakdown, smoothing, and current-limiting regions.
.AC Uses the operating-point derivative, including derivatives from Ilimit, RevIlimit, Epsilon, RevEpsilon, finite Roff, and reverse breakdown.
.TRAN Memoryless device behavior; transient operating points use the same current law at each time point.
Scaling Supports LTspice-style M parallel scaling and N series scaling.
Shared diode syntax Accepts area, OFF, L, W, and Rs; area and Rs are ignored electrically for LTspice ideal-diode parity.
External parity tests Optional LTspice-backed goldens cover .DC, .AC, and .TRAN behavior.

When To Use It

Use this component when a netlist contains LTspice ideal diode parameters:

.model did D(Ron=0.1 Roff=1e9 Vfwd=0.7)
D1 out 0 did

Use the built-in diode when the model is a semiconductor model:

.model d4148 D(Is=2.52e-9 Rs=0.568 N=1.752 Cjo=4e-12 M=0.4 Tt=20e-9)
D1 out 0 d4148

The ideal diode is faster and simpler, but it does not include junction capacitance, charge storage, temperature-dependent semiconductor equations, or noise.

Installation

Reference the custom component assembly in the application that reads the netlist:

<ProjectReference Include="..\SpiceSharpParser.CustomComponents\SpiceSharpParser.CustomComponents.csproj" />

When the package is published, use the package reference instead:

dotnet add package SpiceSharpParser.CustomComponents

Enable Parser Mappings

The parser does not enable custom components by default. Enable them on the reader settings before calling Read():

using System;
using SpiceSharpParser;
using SpiceSharpParser.CustomComponents;

var netlist = string.Join(Environment.NewLine,
    "Ideal diode example",
    "V1 in 0 3",
    "D1 in 0 did",
    ".model did D(Ron=2 Roff=1e9 Vfwd=1)",
    ".op",
    ".end");

var parser = new SpiceNetlistParser();
parser.Settings.Lexing.HasTitle = true;
parser.Settings.Parsing.IsEndRequired = true;
var parseResult = parser.ParseNetlist(netlist);

var reader = new SpiceSharpReader();
reader.Settings.UseCustomComponents();

var spiceModel = reader.Read(parseResult.FinalModel);

UseCustomComponents() replaces the parser mappings for diode models and diode instances with custom-aware generators. Ordinary diode models still fall back to the built-in SpiceSharp diode.

Without UseCustomComponents(), the core parser does not know how to map LTspice ideal diode parameters to a SpiceSharp entity.

Netlist Syntax

The instance syntax is the normal diode syntax:

D<name> <anode> <cathode> <model_name> [<area>] [ON|OFF] [M=<m>] [N=<n>] [L=<length>] [W=<width>] [<parameter>=<value> ...]

The custom component is selected when the referenced .MODEL D(...) contains at least one ideal diode model parameter:

.model did D(Ron=2 Roff=1e9 Vfwd=1)
D1 in 0 did

A classic model remains a classic diode:

.model regular D(Is=1e-12 N=1)
D1 in 0 regular

Rs alone does not select the ideal diode, because Rs is also a classic diode model parameter. At least one of Ron, Roff, Vfwd, Vrev, Rrev, Ilimit, RevIlimit, Epsilon, or RevEpsilon must be present on the model.

If multiple suffixed models share a base name, instance L= and W= values are used with model Lmin, Lmax, Wmin, and Wmax selection parameters.

D1 in 0 did L=0.5 W=5
.model did.fast D(Ron=2 Roff=1e9 Vfwd=1 Lmin=0.1 Lmax=1 Wmin=1 Wmax=10)
.model did.slow D(Ron=4 Roff=1e9 Vfwd=1 Lmin=1 Lmax=10 Wmin=10 Wmax=100)

In this example, D1 selects did.fast.

Parameters

Model Parameters

Parameter Description Default
Ron Forward on resistance 1 ohm
Roff Off resistance simulation Gmin if omitted
Vfwd Forward threshold voltage 0 V
Vrev Reverse breakdown voltage magnitude not enabled
Rrev Reverse breakdown resistance Ron if omitted
Ilimit Forward current limit not enabled
RevIlimit Reverse current limit not enabled
Epsilon Forward transition smoothing width 0 V
RevEpsilon Reverse transition smoothing width 0 V

Instance Parameters

Parameter Description Default
area Accepted by diode syntax, ignored by the idealized model 1
M Positive parallel multiplier 1
N Positive series multiplier 1
L Length used only for binned model selection not set
W Width used only for binned model selection not set
ON / OFF Initial state hint on
Ron, Roff, Vfwd, Vrev, Rrev, Ilimit, RevIlimit, Epsilon, RevEpsilon SpiceSharpParser instance override for the same model parameter model value
Rs Accepted by shared diode parameter handling, ignored by the idealized model 0 ohm

For the ideal diode, N is a series multiplier. This differs from the classic diode model, where model parameter N is the emission coefficient.

L and W are not electrical parameters of the ideal diode. They are evaluated during simulation setup to select a matching suffixed model.

For strict LTspice netlist parity, put the ideal-diode electrical parameters on the .model line. Instance-level ideal-parameter overrides are accepted by this custom component as a SpiceSharpParser extension and are useful when a generated netlist needs per-instance tuning.

Parameter Scaling

The model is evaluated as one ideal diode cell. LTspice's idealized diode uses M and N for electrical scaling:

Parameter Effect
M Represents parallel diode cells. Current and conductance scale by M.
N Represents series diode cells. Total voltage is divided by N before evaluating one cell.
area Accepted by the shared diode syntax but not used electrically by the idealized model.
Rs Accepted by shared model/instance handling but not used electrically by the idealized model.

area, M, and N are still validated as positive diode instance values.

For a forward-biased diode without smoothing or current limiting:

$$ \begin{aligned} v_{\text{local}} &= \frac{V(\text{anode}, \text{cathode})}{N} \\ i_{\text{local}} &= \frac{v_{\text{local}} - V_{\text{fwd}}}{R_{\text{on}}} \\ I_{\text{total}} &= M \cdot i_{\text{local}} \end{aligned} $$

So the approximate total forward threshold is:

$$ V_{\text{total}} \approx N \cdot V_{\text{fwd}} $$

and the approximate effective forward resistance is:

$$ R_{\text{effective}} \approx \frac{R_{\text{on}} \cdot N}{M} $$

For LTspice parity, area and Rs do not change that effective resistance for the idealized Ron/Roff/Vfwd model. Use M for parallel scaling, N for series scaling, or add an explicit resistor if a series resistance is required.

This matches the intended interpretation: more parallel cells carry more current, and more series cells need more voltage.

Model Parameter Sweeps

Ideal diode model parameters can be stepped with normal LTspice-style bracket syntax:

Ideal diode Ron sweep
V1 in 0 5
D1 in 0 did
.model did D(Ron=2 Roff=1e9 Vfwd=1)
.op
.step D did(Ron) LIST 2 4
.end

The first operating point uses Ron=2, so the current is approximately:

$$ \frac{5 - 1}{2} = 2,\text{A} $$

The second operating point uses Ron=4, so the current is approximately:

$$ \frac{5 - 1}{4} = 1,\text{A} $$

Stepped model values are applied as simulation-local ideal-diode parameter overlays. After a stepped simulation finishes, the shared model parameter set is restored to its original value.

For binned models, the sweep target can use either the requested base model name or the concrete selected model name:

Binned ideal diode Ron sweep
V1 in 0 5
D1 in 0 did L=0.5 W=5
.model did.fast D(Ron=2 Roff=1e9 Vfwd=1 Lmin=0.1 Lmax=1 Wmin=1 Wmax=10)
.model did.slow D(Ron=8 Roff=1e9 Vfwd=1 Lmin=1 Lmax=10 Wmin=10 Wmax=100)
.op
.step D did.fast(Ron) LIST 2 4
.end

Here D1 selects did.fast, and the sweep steps that selected model's Ron.

Current And Voltage Sign

The anode is the first diode node and the cathode is the second diode node:

D1 anode cathode did

Positive diode voltage is:

$$ V(\text{anode}, \text{cathode}) $$

Positive diode current flows from anode to cathode. Property exports follow that sign convention:

.save @D1[v] @D1[vj] @D1[i] @D1[gd] @D1[p]
Export Meaning
@D1[v] or @D1[vd] Terminal diode voltage
@D1[vj] or @D1[vdiode] Internal ideal-diode voltage; equal to terminal voltage for the LTspice-parity idealized model
@D1[i], @D1[id], or @D1[c] Diode current
@D1[gd] Terminal small-signal conductance at the operating point
@D1[p] or @D1[pd] Terminal diode branch power

When the diode is reverse biased, @D1[i] is usually negative.

Current Law

The ideal diode current law is evaluated for one diode cell first. Instance scaling is applied after that.

Local Evaluation Algorithm

The implementation for this section lives in IdealDiodeEquation. The method Evaluate(...) does not stamp the circuit matrix. It only evaluates the local current law for one ideal-diode cell and returns two values:

Output Meaning
current Local diode current at the present voltage
conductance Local small-signal derivative di/dv used by Newton iteration

The voltage v is positive from anode to cathode, and positive current flows in that same direction. The algorithm uses a line representation for each operating region:

i(v) = slope * v + intercept

This representation matters because the solver needs both the current and its derivative. The line slope is already the small-signal conductance for that region.

The local evaluation is:

gon  = 1 / Ron
goff = Roff was given ? 1 / Roff : max(simulation Gmin, 0)
vf   = Vfwd was given ? Vfwd : 0

current     = goff * v
conductance = goff

if Vrev was given:
    vrev = abs(Vrev)
    rrev = Rrev was given ? Rrev : Ron
    grev = 1 / rrev

    reverse line = grev * v + (grev - goff) * vrev
    off line     = goff * v
    boundary     = -vrev

    current, conductance =
        transition(v, boundary, RevEpsilon, reverse line, off line)

forward line = gon * v + (goff - gon) * vf
off line     = goff * v
boundary     = vf

if v is past the forward boundary, or inside the forward smoothing window:
    current, conductance =
        transition(v, boundary, Epsilon, off line, forward line)

if current > 0 and Ilimit was given:
    apply tanh current limiter
else if current < 0 and RevIlimit was given:
    apply tanh current limiter

The order is intentional. The off line is the default because most voltages are between reverse breakdown and forward conduction. Reverse breakdown is evaluated before forward conduction because it only applies on the negative-voltage side. The forward transition is evaluated last so positive forward conduction replaces the off result when the voltage reaches the forward knee.

The transition helper is shared by both knees. It receives the line on the low-voltage side, the line on the high-voltage side, the nominal LTspice knee, and the optional smoothing width. For reverse breakdown, the low-voltage side is the reverse line and the high-voltage side is the off line. For forward conduction, the low-voltage side is the off line and the high-voltage side is the forward line.

With finite Roff, the conducting lines are anchored to the off-line current at the nominal threshold. Forward conduction is continuous with the off line at Vfwd, and reverse breakdown is continuous with the off line at -Vrev. LTspice keeps those nominal knees rather than moving the transition to the mathematical intersection of the two unsmoothed lines.

With no smoothing, the transition helper simply chooses one line. With smoothing, the epsilon value is treated as a one-sided LTspice-style width. Forward smoothing starts at the forward boundary and extends into the forward-conduction region. Reverse smoothing starts in reverse breakdown and ends at the reverse boundary. The conductance ramps linearly from the left slope to the right slope, and the current is the integral of that ramp. This keeps both current and conductance continuous at the start and end of the smoothing window. It also shifts the fully conducting side by half the epsilon width; for example, deep in reverse breakdown with RevEpsilon=e, the approximate line becomes:

$$ i \approx \frac{v + V_{\text{rev}} + e / 2}{R_{\text{rev}}} $$

Current limiting happens after the piecewise line and optional smoothing have selected a raw current. The limiter is based on tanh, so it approaches the limit smoothly and also scales the conductance by the derivative of the limiter. The sign of the raw current selects the limiter: positive current uses Ilimit, negative current uses RevIlimit, and zero current is left unchanged.

The caller, Biasing, handles everything outside the local cell. It divides the internal diode voltage by N before calling Evaluate(...), multiplies the equivalent branch current by M, and scales the conductance by M / N. The accepted area and Rs values are ignored electrically for LTspice parity.

The formulas below describe the same algorithm. Define:

$$ \begin{aligned} v &= \text{voltage across one series cell} \\ V_f &= \begin{cases} V_{\text{fwd}}, & \text{if } V_{\text{fwd}} \text{ is given} \\ 0, & \text{otherwise} \end{cases} \\ g_{\text{on}} &= \frac{1}{R_{\text{on}}} \\ g_{\text{off}} &= \begin{cases} \frac{1}{R_{\text{off}}}, & \text{if } R_{\text{off}} \text{ is given} \\ G_{\text{min}}, & \text{otherwise} \end{cases} \end{aligned} $$

The forward, off, and reverse-breakdown lines are:

$$ \begin{aligned} i_{\text{on}}(v) &= g_{\text{on}} \cdot (v - V_f) + g_{\text{off}} \cdot V_f \\ i_{\text{off}}(v) &= g_{\text{off}} \cdot v \\ i_{\text{rev}}(v) &= g_{\text{rev}} \cdot (v + V_{\text{rev}}) - g_{\text{off}} \cdot V_{\text{rev}} \end{aligned} $$

where i_rev(v) is used only when Vrev is given, and:

$$ g_{\text{rev}} = \frac{1}{R_{\text{rev}}} $$

If Rrev is omitted, Ron is used for Rrev.

Without smoothing or current limiting, the raw current is the line selected by the present voltage:

$$ i_{\text{raw}}(v) = \begin{cases} i_{\text{rev}}(v), & \text{if } V_{\text{rev}} \text{ is enabled and } v < -V_{\text{rev}} \\ i_{\text{off}}(v), & \text{if } v < V_f \\ i_{\text{on}}(v), & \text{otherwise} \end{cases} $$

The small-signal conductance is the derivative of the selected line:

$$ g_{d,\text{raw}} = \frac{d i_{\text{raw}}}{d v} $$

So the unsmoothed conductance is usually one of:

$$ g_{d,\text{raw}} \in {g_{\text{rev}}, g_{\text{off}}, g_{\text{on}}} $$

The forward transition point is the nominal LTspice threshold:

$$ v_{\text{fwd,boundary}} = V_{\text{fwd}} $$

The forward line is shifted by the off-line current at Vfwd, so finite Roff adds a small term:

$$ i_{\text{on}}(v) = \frac{v - V_{\text{fwd}}}{R_{\text{on}}} + \frac{V_{\text{fwd}}}{R_{\text{off}}} $$

Reverse Breakdown

Reverse breakdown is enabled by Vrev. Vrev is a magnitude, so this model starts reverse breakdown near:

$$ v = -V_{\text{rev}} $$

The reverse breakdown region is approximately:

$$ i \approx \frac{v + V_{\text{rev}}}{R_{\text{rev}}} $$

The reverse transition point is the nominal LTspice threshold:

$$ v_{\text{rev,boundary}} = -V_{\text{rev}} $$

The reverse line is shifted by the off-line current at -Vrev, so finite Roff adds a small term:

$$ i_{\text{rev}}(v) = \frac{v + V_{\text{rev}}}{R_{\text{rev}}} - \frac{V_{\text{rev}}}{R_{\text{off}}} $$

Example:

.model clamp D(Ron=2 Roff=1e9 Vfwd=1 Vrev=2 Rrev=4)

At v = -6 V, the approximate reverse current is:

$$ i \approx \frac{-6 + 2}{4} = -1,\text{A} $$

Current Limiting

Ilimit and RevIlimit smoothly limit forward and reverse current. The limiter uses a tanh shape, so it approaches the limit gradually instead of clipping with a hard corner.

Forward current limit:

.model did D(Ron=0.1 Roff=1e9 Vfwd=0.7 Ilimit=10)

Reverse current limit:

.model did D(Ron=0.1 Roff=1e9 Vfwd=0.7 Vrev=20 RevIlimit=2)

The smooth limiter also reduces the small-signal conductance as the current approaches the limit.

Let i0 and gd0 be the current and conductance after the piecewise line and optional smoothing, but before current limiting. The forward limiter uses this shape:

$$ \begin{aligned} i_{\text{limited}} &= I_{\text{limit}} \cdot \tanh\left(\frac{i_0}{I_{\text{limit}}}\right) \\ g_{d,\text{limited}} &= g_{d0} \cdot \left(1 - \tanh^2\left(\frac{i_0}{I_{\text{limit}}}\right)\right) \end{aligned} $$

The reverse limiter uses the same equation with RevIlimit. Because reverse current is negative, $\tanh(i_0 / I_{\text{rev-limit}})$ is also negative.

Transition Smoothing

Epsilon and RevEpsilon smooth the forward and reverse transitions. A value of zero gives a sharp piecewise transition. A positive value ramps the slope over a one-sided LTspice-style voltage window next to the transition.

.model did D(Ron=0.1 Roff=1e9 Vfwd=0.7 Epsilon=10m)

For any smoothed transition, let the left line be:

$$ i_{\text{left}}(v) = a_{\text{left}} \cdot v + b_{\text{left}} $$

and the right line be:

$$ i_{\text{right}}(v) = a_{\text{right}} \cdot v + b_{\text{right}} $$

For the forward transition, the left line is i_off(v) and the right line is i_on(v). For the reverse transition, the left line is i_rev(v) and the right line is i_off(v).

With smoothing width e, the forward smoothing window is:

$$ \begin{aligned} v_{\text{start}} &= v_{\text{boundary}} \\ v_{\text{end}} &= v_{\text{boundary}} + e \end{aligned} $$

The reverse smoothing window is:

$$ \begin{aligned} v_{\text{start}} &= v_{\text{boundary}} - e \\ v_{\text{end}} &= v_{\text{boundary}} \end{aligned} $$

Inside that window:

$$ \begin{aligned} d &= v - v_{\text{start}} \\ i(v) &= i_{\text{left}}(v_{\text{start}}) + a_{\text{left}} \cdot d + \frac{(a_{\text{right}} - a_{\text{left}}) \cdot d^2}{2e} \\ g_d(v) &= a_{\text{left}} + \frac{(a_{\text{right}} - a_{\text{left}}) \cdot d}{e} \end{aligned} $$

Outside the window, the model uses the normal left or right line.

For the fully conducting side of a smoothed transition, the straight line is shifted by half the epsilon width so it joins the ramp continuously. With tiny Roff, this means the deep forward region is approximately:

$$ i \approx \frac{v - V_{\text{fwd}} - Epsilon / 2}{R_{\text{on}}} $$

and the deep reverse region is approximately:

$$ i \approx \frac{v + V_{\text{rev}} + RevEpsilon / 2}{R_{\text{rev}}} $$

Use smoothing when a hard corner causes convergence problems in DC or transient operating-point iterations.

How The Parser Selects The Component

The parser bridge has two pieces:

Class Role
IdealDiodeModelGenerator Reads .MODEL ... D(...). If the model has ideal diode parameters, it creates an IdealDiodeModel; otherwise it delegates to the built-in diode model generator.
IdealDiodeGenerator Reads D... instances. If the referenced model is an IdealDiodeModel, it creates an IdealDiode; otherwise it delegates to the built-in diode generator.

The flow is:

reader.Settings.UseCustomComponents()
  -> replace D model generator
  -> replace D component generator

.model did D(Ron=...)
  -> IdealDiodeModelGenerator detects Ron
  -> creates IdealDiodeModel
  -> stores model parameters

D1 in 0 did
  -> IdealDiodeGenerator resolves model did
  -> sees IdealDiodeModel
  -> creates IdealDiode
  -> binds live model parameters for each simulation
  -> applies simulation-local model sweep overlays, if any
  -> applies instance overrides

Classic diode models are delegated back to the existing parser behavior:

.model regular D(Is=...)
  -> no ideal diode parameter
  -> built-in DiodeModel
  -> built-in Diode

The selected model is resolved again before simulation setup. This lets L and W expressions participate in binned model selection and lets stochastic or stepped simulations use their simulation-specific model data.

How The SpiceSharp Component Works

IdealDiode is a normal two-pin SpiceSharp component:

pin 0 = anode
pin 1 = cathode

During simulation setup, it creates behavior objects:

Behavior Used by Purpose
Biasing .OP, .DC, and operating-point parts of transient Evaluates diode current and conductance, then stamps the real-valued matrix.
Frequency .AC Uses the operating-point conductance for small-signal AC.

The component is memoryless. It does not add dynamic state for charge or capacitance.

Biasing Stamp

At each biasing load, the behavior:

  1. Reads the present internal diode voltage.
  2. Divides it by N to get the voltage across one series cell.
  3. Evaluates current i and local conductance gd.
  4. Scales current by M and conductance by M / N.
  5. Stamps the internal diode conductance into the matrix.
  6. Stamps the equivalent current into the right-hand side.
  7. Stamps the zero-volt branch equation between the external and internal anodes.

The local linear form is:

$$ \begin{aligned} i &\approx g_d \cdot v + i_{\text{eq}} \\ i_{\text{eq}} &= i - g_d \cdot v \end{aligned} $$

The conductance stamp is resistor-like:

Matrix entry Added value
Y[internal_anode, internal_anode] +gd
Y[cathode, cathode] +gd
Y[internal_anode, cathode] -gd
Y[cathode, internal_anode] -gd

The right-hand side receives the equivalent current source contribution.

Series Resistance

The component always creates a private internal anode and a branch:

external anode -- 0 V constraint -- internal anode -- ideal diode -- cathode

That fixed topology lets Rs be parsed or stepped without rebinding the component, but for LTspice parity Rs is ignored electrically by the idealized Ron/Roff/Vfwd model. Add an explicit resistor in series with the diode when series resistance is required.

The standard v, vj, gd, and p exports are terminal-equivalent quantities.

Convergence Check

The biasing behavior implements a convergence check using the diode current predicted from the previous local linearization:

$$ i_{\text{predicted}} = i_{\text{old}} + g_{d,\text{old}} \cdot \Delta v $$

If the predicted current differs from the actual loaded current by more than the simulation tolerances, the iteration is marked as not converged and SpiceSharp continues iterating.

AC Behavior

AC analysis uses the operating-point conductance. The conductance is the actual local derivative from the selected operating region: Ron, Roff, Rrev, the linear Epsilon or RevEpsilon ramp, or the derivative after Ilimit / RevIlimit compression. The AC stamp uses the internal diode conductance, while property exports report terminal voltage, terminal current, and terminal complex power:

$$ y_{\text{ac}} = g_d $$

There is no frequency-dependent capacitance in this model. That means the AC response of the diode itself is resistive and frequency independent. Any frequency response must come from surrounding capacitors, inductors, sources, or other dynamic components.

Transient Behavior

The ideal diode has no charge storage, capacitance, or dynamic state of its own. During transient analysis, SpiceSharp solves the circuit operating point at each time step using the same memoryless current law used by .OP and .DC. Source waveforms, capacitors, inductors, and other dynamic elements in the surrounding circuit determine the time-dependent behavior.

Direct Code Usage

You can also create the component directly without parsing a netlist:

using SpiceSharp;
using SpiceSharp.Components;
using SpiceSharp.Simulations;
using SpiceSharpParser.CustomComponents;

var diode = new IdealDiode("D1", "in", "0");
diode.Parameters.OnResistance = 2.0;
diode.Parameters.OffResistance = 1e9;
diode.Parameters.ForwardVoltage = 1.0;

var circuit = new Circuit(
    new VoltageSource("V1", "in", "0", 3.0),
    diode);

var op = new OP("op");
var current = new RealPropertyExport(op, "D1", "i");

foreach (int _ in op.Run(circuit))
{
    Console.WriteLine(current.Value);
}

This prints approximately 1 A, because:

$$ \frac{3,\text{V} - 1,\text{V}}{2,\Omega} = 1,\text{A} $$

Worked Examples

Forward Conduction

Forward ideal diode
V1 in 0 3
D1 in 0 did
.model did D(Ron=2 Roff=1e9 Vfwd=1)
.op
.save @D1[i] @D1[v]
.end

The expected current is approximately:

$$ i \approx \frac{3 - 1}{2} = 1,\text{A} $$

Off Leakage

Off leakage
V1 in 0 0.5
D1 in 0 did
.model did D(Ron=2 Roff=1e9 Vfwd=1)
.op
.save @D1[i]
.end

The diode is below its forward region. The expected current is approximately:

$$ i \approx \frac{0.5}{10^9} = 0.5,\text{nA} $$

Reverse Clamp

Reverse clamp
V1 in 0 -6
D1 in 0 clamp
.model clamp D(Ron=2 Roff=1e9 Vfwd=1 Vrev=2 Rrev=4)
.op
.save @D1[i]
.end

The expected reverse current is approximately:

$$ i \approx \frac{-6 + 2}{4} = -1,\text{A} $$

Series And Parallel Multipliers

Series and parallel ideal diode
V1 in 0 3
D1 in 0 did M=2 N=2
.model did D(Ron=2 Roff=1e9 Vfwd=1)
.op
.save @D1[i]
.end

The local diode voltage is:

$$ v_{\text{local}} = \frac{3}{2} = 1.5,\text{V} $$

The local current through one diode cell is:

$$ i_{\text{local}} = \frac{1.5 - 1}{2} = 0.25,\text{A} $$

With M=2, total current is:

$$ I_{\text{total}} = 2 \cdot 0.25 = 0.5,\text{A} $$

Current-Limited Clamp

Current-limited clamp
V1 in 0 5
R1 in out 10
D1 out 0 did
.model did D(Ron=0.1 Roff=1e9 Vfwd=0.7 Ilimit=10 Epsilon=10m)
.op
.save V(out) @D1[i] @D1[gd]
.end

The 10 ohm resistor keeps the operating current far below the 10 A limit, so the limiter is almost inactive in this operating point. Once the diode is above the smoothed transition region, the approximate equations are:

$$ \begin{aligned} I &\approx \frac{5 - V(\text{out})}{10} \\ I &\approx \frac{V(\text{out}) - 0.7}{0.1} \end{aligned} $$

Solving them gives:

$$ \begin{aligned} I &\approx 0.426,\text{A} \\ V(\text{out}) &\approx 0.743,\text{V} \end{aligned} $$

If the surrounding circuit tried to drive much more current, the forward limiter would use:

$$ I = 10 \cdot \tanh\left(\frac{(v - 0.7) / 0.1}{10}\right) $$

LTspice Golden Comparison

The test project includes an optional LTspice-backed golden comparison for the ideal diode. It is skipped by default so normal test runs do not require LTspice to be installed.

To enable it, set LTSPICE_EXE to the LTspice executable path and run the focused ideal-diode tests:

$env:LTSPICE_EXE = "C:\Program Files\ADI\LTspice\LTspice.exe"
dotnet test .\src\SpiceSharpParser.Tests\SpiceSharpParser.Tests.csproj --filter FullyQualifiedName~IdealDiode

The golden test generates temporary LTspice .cir files, runs LTspice in batch ASCII mode, parses the LTspice .raw output, and compares the results against the custom IdealDiode implementation.

The external golden suite covers:

Analysis Comparison
.DC Sweeps diode voltage and compares I(D1) for each sampled point.
.AC Compares complex small-signal voltages, including nonlinear derivative cases around Ilimit, Epsilon, RevEpsilon, reverse breakdown, and finite Roff smoothing.
.TRAN Runs a sinusoidal bridge rectifier and compares the rectified waveform against equivalent operating-point samples.

The .DC case matrix covers:

Case Parameters exercised
Forward and off regions Ron, Roff, Vfwd
Reverse breakdown Vrev, Rrev
Current limiting Ilimit, RevIlimit
Transition smoothing Epsilon, RevEpsilon, including finite-Roff ramps
Scaling and ignored shared syntax area, M, N, Rs, off

The .AC derivative matrix covers:

Case What it checks
Forward biased Small-signal conductance from Ron.
Forward current limit Reduced derivative after Ilimit tanh compression.
Forward epsilon ramp Conductance inside the Epsilon transition window.
Finite Roff forward epsilon Ramp derivative when the off line has visible slope.
Reverse epsilon ramp Conductance inside the RevEpsilon transition window.
Finite Roff reverse epsilon Reverse ramp derivative when the off line has visible slope.
Reverse breakdown Small-signal conductance from Rrev.

Unsupported Or Ignored Parameters

When an ideal diode model is selected, the custom parser keeps the ideal-diode parameters and ignores classic diode model parameters such as:

Is, Tnom, N, Tt, Cjo, Cj0, Vj, M, Eg, Xti, Fc, BV, IBV, Kf, Af

The instance parameters temp and ic are ignored for the ideal diode. Unsupported ideal diode instance parameters produce validation errors.

Metadata-style LTspice parameters such as mfg, pn, description, and ratings such as irms or ipk are ignored.

Limitations

  • The component is memoryless: it does not model junction capacitance or charge storage.
  • It does not provide noise behavior.
  • It does not model temperature-dependent semiconductor behavior.
  • Classic diode parameters such as Is, Cjo, Tt, and BV are ignored when an ideal diode model is selected.
  • UseCustomComponents() is required. Without it, the core parser treats LTspice ideal diode parameters as unsupported in LTspice compatibility mode.

Troubleshooting

Symptom Likely cause Fix
Ron or Vfwd is reported as unsupported Custom mappings are not enabled Call reader.Settings.UseCustomComponents() before Read().
A classic diode unexpectedly behaves like an ideal diode The model contains at least one ideal diode parameter Remove Ron, Roff, Vfwd, Vrev, Rrev, Ilimit, RevIlimit, Epsilon, or RevEpsilon from the model.
N does not behave like emission coefficient In the ideal diode, N is the series multiplier Use a classic diode model for emission-coefficient behavior.
A binned model sweep does not affect the selected diode The sweep target does not match the requested or selected model name Use the base model name, such as did(Ron), or the selected suffixed name, such as did.fast(Ron).
AC result has no diode capacitance effect The ideal diode has no capacitance model Add explicit capacitors or use a classic diode model.
DC or transient convergence is rough near switching The transition is too sharp Add Epsilon or RevEpsilon, or reduce extreme Ron/Roff ratios.

Complete Example

Ideal diode clamp
V1 in 0 5
R1 in out 10
D1 out 0 did
.model did D(Ron=0.1 Roff=1e9 Vfwd=0.7 Ilimit=10)
.op
.save V(out) @D1[i]
.end

This uses the same current law:

$$ \begin{aligned} g_{\text{on}} &= \frac{1}{0.1} = 10,\text{S} \\ g_{\text{off}} &= \frac{1}{10^9} = 1,\text{nS} \\ i_{\text{on}}(v) &= 10 \cdot (v - 0.7) \\ i_{\text{off}}(v) &= 10^{-9} \cdot v \end{aligned} $$

Because Roff is so large, the forward boundary is approximately:

$$ v_{\text{fwd,boundary}} \approx 0.7,\text{V} $$

With the 10 ohm source resistor, the operating point is approximately:

$$ \begin{aligned} I &\approx \frac{5 - V(\text{out})}{10} \\ I &\approx \frac{V(\text{out}) - 0.7}{0.1} \\ I &\approx 0.426,\text{A} \\ V(\text{out}) &\approx 0.743,\text{V} \end{aligned} $$

The Ilimit=10 parameter is still part of the model. It matters when the surrounding circuit tries to push the pre-limit forward current toward 10 A:

$$ i_{\text{limited}} = 10 \cdot \tanh\left(\frac{i_0}{10}\right) $$