Dieser Leitfaden ist derzeit auf Englisch verfügbar.
Technology / Technical guide
Inverter Dead Time: Commutation Interval & Waveform Error
Real semiconductor transitions take time. In a complementary inverter leg, a deliberate interval between opposing turn-on requests can separate those transitions. That timing interval also changes how current reaches a conductive path and can make the delivered voltage differ from an ideal switching waveform.
A command gap and the resulting voltage are different quantities.
Dead time describes switching timing; the current, devices and parasitic paths determine what happens during that interval.
01 / Complementary leg
Ideal non-overlap omits real transition behaviour.
In an ideal two-state leg model, a pole occupies one rail state at a time.
In an ideal two-state leg model, a pole occupies one rail state at a time. Real devices require finite turn-on and turn-off transitions, and command propagation is not instantaneous. Opposite commands issued at the same logical instant need not create non-overlapping physical conduction.
Dead time describes an intended gap between the relevant complementary requests. It is related to timing uncertainty and transitions, but it is not identical to a measured device turn-off time. The distinction matters before a gap is described as an electrical behaviour.
02 / Command versus path
Both off commands do not establish zero current.
Current through an inductive circuit cannot generally vanish merely because two gate requests are off.
Current through an inductive circuit cannot generally vanish merely because two gate requests are off. Depending on the device and circuit, it can continue through another conductive or capacitive path. The pole’s actual voltage follows that interaction.
The Infineon-hosted original dead-time research describes current-commutation effects for its tested inverter arrangements. Those arrangements illustrate a mechanism only. Their device types, timing values, voltage levels and numerical loss results are not transferred to MARSTON equipment.
03 / Current dependence
The same timing gap can cause different voltage errors.
The direction of existing current influences which paths become active during the gap.
The direction of existing current influences which paths become active during the gap. Consequently, identical command timing need not give the same pole-voltage error for both current directions. Behaviour near a current zero crossing can differ from behaviour with a clear established current.
The discrepancy also depends on real device drops and charging effects. It should not be represented universally as an interval of exactly zero pole voltage. No current-polarity rule is used here as an operational correction or as a claimed controller function.
04 / Average error
Waveform area explains one consequence of a short interval.
For a defined averaging period T, voltage-average error is the integral of actual minus ideal voltage over that period, divided by T.
For a defined averaging period T, voltage-average error is the integral of actual minus ideal voltage over that period, divided by T. In a restricted illustration with constant error Δv over one interval td and no error elsewhere, that contribution is Δv td/T.
This area relationship explains why a short timing change can affect an average waveform. Its sign follows the actual error waveform. It is not a formula for choosing dead time, and multiple transitions or nonconstant errors require their own complete integral.
05 / Different meanings
Switching gaps are not yaw deadbands or mechanical clearance.
A control deadband concerns a range in which a decision rule does not request change.
A control deadband concerns a range in which a decision rule does not request change. Gear backlash concerns a mechanical relationship during reversal. Inverter dead time concerns timing between electrical switching requests. These quantities can all create differences between requests and responses without sharing a cause.
The backlash and deadband guide treats those separate boundaries. A distorted electrical waveform likewise does not uniquely prove dead-time error: several source, load, device and measurement effects can contribute.
06 / Model scope
Timing terminology does not provide an accepted setting.
The ideal vector guide deliberately excludes transitions, while the loss guide distinguishes energy in a transition from average voltage error.
The ideal vector guide deliberately excludes transitions, while the loss guide distinguishes energy in a transition from average voltage error. Each model explains a different observable quantity.
No MARSTON dead-time interval, compensation method, transistor type or switching quality is assigned. This page provides no timing adjustment, gate-command sequence, oscilloscope procedure or protection setting; it explains the causal gap between ideal requests and real voltage.
Related technical guides: Three-Phase Space Vectors & Ideal Inverter States · Diode Reverse Recovery: Stored Charge & Commutation · Miller Capacitance, Gate Charge & Switching-Node Coupling.
General background: Reiter et al.: original dead-time and current-commutation research. These references explain industry concepts; they are separate from the applicable MARSTON product and project documents.
Key concepts and conditions.
Read each quantity with its reference, units and stated conditions. A general relation does not establish a model-specific value.
| Concept / quantity | Meaning / relation | Conditions / limits |
|---|---|---|
| Ideal state | One selected pole state | Finite transitions are excluded. |
| Dead time | Gap between complementary requests | Not automatically the actual device transition time. |
| Current continuity | Existing circuit current | Off commands do not necessarily make it zero. |
| Average error | Integral of voltage error divided by T | Complete waveform and signs are defined. |
| Other dead zones | Control deadband or gear clearance | Different physical and logical mechanisms. |
Engineering questions.
Is dead time the same as a device turn-off time?
No. It is a command-timing interval with its own relationship to real transitions.
Does both-off mean zero pole voltage?
Not generally. Current and available circuit paths affect the pole.
Can current direction change the voltage error?
Yes. It changes the relevant commutation-path behaviour.
Is Δv td/T a universal setting equation?
No. It is a restricted waveform-area illustration.
Is yaw deadband another name for inverter dead time?
No. A decision range and an electrical timing gap are different quantities.
Are actual MARSTON timing or compensation values provided?
No. No implementation or operational adjustment is assigned.
Technical enquiry
Discuss the technical question.
Real semiconductor transitions take time. Share the model and applicable operating or design information for an engineering discussion.