The IGBT is the power semiconductor that most engineers meet last, and it is also the one whose datasheet reads most like a compromise: a device that switches like a transistor and conducts like a diode, with a voltage rating that demands respect and a switching speed that has a firm ceiling. Understanding the IGBT starts with one mental model — the gate-controlled valve that holds high voltage — and then a ratings map, the frequency corner, a calm datasheet reading order, and a short list of selection questions. This article is the beginner hub that makes the rest of the IGBT cluster usable.
One Mental Model: The Valve That Holds High Voltage
The IGBT is best modeled as a hybrid: a gate that controls conduction like a MOSFET, and a collector-emitter structure that blocks high voltage like a diode with a working voltage. The gate opens with a voltage, the part conducts, and when the gate is off, the collector-emitter junction holds the blocking voltage across the structure. The descriptor simplified: it is a valve that holds high voltage.
The consequence of the model is the two-number identity of the device: it is rated by the voltage it can block (the collector-emitter rating) and the current it can carry when open (the collector rating), and the two are separate limits that must both be respected. An IGBT is not a bigger transistor; it is a different compromise built around holding high voltage while switching on command. The IGBT guide frames the market and sourcing context, and the IGBT vs MOSFET boundary shows where the compromise earns its place; this article builds the reading order the other pieces assume.
A structural contrast makes the hybrid nature concrete. A MOSFET conducts through a straight resistance when its gate is open, so its on-state loss at low current is tiny and grows as the current squared. An IGBT conducts through a diode-like collector-emitter junction, so its on-state has a roughly fixed drop across a range of current — better than a MOSFET at very high current where the resistive term explodes, worse at low current where a MOSFET’s tiny resistance wins. The crossover is how an engineer decides where an IGBT beats a MOSFET: at high voltage and high current where the fixed drop is a smaller fraction of the total loss, against low-voltage low-current rails where the MOSFET owns the efficiency.
That same crossover is why the IGBT is the workhorse of high-voltage drives and grid power while the MOSFET keeps the low-voltage switching world. The datasheet reading that follows — the fixed drop as the loss number rather than a temperature-scaled resistance — is the direct consequence of the structure.
A worked reading of the ratings makes the datasheet order concrete. A 1200 V class IGBT is evaluated for a 700 V DC rail: VCES of 1200 V clears the rail with the usual margin. The collector current is quoted at 25 °C, and the 55 °C application de-rates it by the datasheet’s temperature factor, so the usable current is lower than the headline. The saturation drop of roughly 2–3 V at rated current sets the conduction loss, and the frequency corner — around 4–16 kHz for this class — decides whether the motor-drive duty is inside the envelope. Reading the four numbers in order produces a verdict: survive the rail, carry the current, shed the loss, stay inside the corner. The thermal design guide closes the loop with the mounting term, and the package comparison tells which physical package fits the heat.
Ratings That Matter First: VCES, IC, and Saturation
The first ratings to read on an IGBT datasheet are the collector-emitter voltage VCES, the collector current IC, and the saturation behavior. VCES is the blocking voltage the part holds when off — the rating that decides whether the device survives the rail it commands. IC is the current it conducts when open, and like any power device it is quoted at a temperature and must be de-rated for the real ambient.
The saturation behavior is the surprise for engineers raised on MOSFETs: an IGBT has an on-state forward drop that does not vanish with more drive, because the collector-emitter junction is diode-like. The datasheet’s voltage drop at rated current is the number that feeds loss and thermal calculations, and it behaves like a fixed drop more than a resistance. The power MOSFET comparison is the boundary that makes this difference explicit; the reading here is that the IGBT’s loss at a given current is set by the drop, not by an RDS(on) that scales with temperature the same way.
The rating map is the order to read: voltage first (survival), current second (capability), saturation drop third (loss).
The thermal closure is worth working once so it is not a mystery. The IGBT’s saturation drop at the operating current gives the conduction loss, the switching-loss energy times the frequency gives the switching loss, and the two sum to the total the junction must shed. That total through the junction-to-ambient thermal resistance lands at a junction temperature, and the design is sound only if that temperature stays under the datasheet limit at the worst ambient. The loop is identical to the rectifier method and the MOSFET worksheet; the only difference is that the IGBT’s conduction term is a fixed drop rather than a resistance, which makes the conduction loss scale with current linearly, not by the square.
The linear-versus-squared scaling is the practical consequence that shapes the selection. Because the IGBT’s conduction loss grows with current rather than current-squared, it is the better part at high current where the MOSFET’s squared term runs away. The six questions in the next section use exactly this to route a design: a high-current, moderate-frequency, high-voltage duty points at the IGBT, and a low-current or high-frequency duty points back at the MOSFET or toward SiC. The IGBT family spans the voltage classes, and the SiC transition article is where the flow continues when frequency or temperature pushes past silicon.
Switching Losses and the Frequency Corner
The IGBT’s switching losses grow with frequency, and the growth has a firm corner that decides where the device is usable. Each switch event carries a defined energy — the tail current and the recovery behavior of the collector-emitter structure — and at high repetition the per-event energy becomes a continuous power that heats the device. The datasheet’s switching-loss figures and the frequency-dependent curves are the numbers that set the corner.
The frequency corner is typically in the kilohertz to tens-of-kilohertz range for power IGBTs, far below where a small-signal MOSFET is comfortable. A design that tries to switch an IGBT at meghertz rates spends more energy per second in the switching events than it saves anywhere else, and the part overheats. The 1200 V IGBT motor-drive article works the 4–16 kHz corner for a real application, and the high-dv/dt driver article shows the drive requirements that make the corner reachable.
The practical read: match the switching frequency to the corner before sizing the thermal path, because a device at its frequency limit looks fine cold and overheats in service.
Reading an IGBT Datasheet Without Fear
The IGBT datasheet loses its intimidation when read in order. First, the voltage rating VCES — block it against the rail and the transients. Second, the collector current at the operating temperature. Third, the on-state saturation drop — this is the loss number. Fourth, the switching-loss figures and the frequency-dependent curves. Fifth, the thermal resistance and the junction limit that close the loop with the loss.
The disciplined order prevents the classic IGBT errors: sizing on the current headline without de-rating for temperature, or trusting the part to switch fast when the frequency corner says otherwise. The thermal closure method used for rectifiers applies unchanged, because the IGBT’s loss must leave through the same junction-to-ambient chain. The TO-247 versus D2PAK mounting article fills in the package-specific thermal reading the datasheet page cannot.
First-Pass Selection: Six Questions for a New Application
Six questions make the first pass fast. Question one, what blocking voltage does the rail demand — set VCES against it with margin. Question two, what current at what ambient — de-rate the collector current. Question three, what switching frequency — is it inside the corner? Question four, what is the load — resistive, inductive, or a drive profile with high inrush. Question five, what thermal path is available — the heatsink and mounting decide how the saturation loss leaves. Question six, what is the drive — the gate-drive capability decides whether the device can switch at the needed speed.
The six answers produce either a go, a voltage family change, or a different device tier. The IGBT families and the 650 V IGBT application article cover the parts and the lower-voltage applications; the IGBT welding case and the SiC transition are where the flow moves when voltage or frequency pushes past what silicon IGBT can deliver. The six questions keep the first pass honest, and the deeper articles in the cluster follow the answers.