A solar panel produces direct current, the grid demands alternating current, and everything a string inverter does in between comes down to how well a handful of power transistors switch that current on and off thousands of times a second. For most of the utility scale and commercial solar market, that transistor is still the insulated gate bipolar transistor, better known as the IGBT, even as silicon carbide devices push into the smaller residential segment. Every fraction of a percentage point an IGBT gives up to heat is energy the plant owner never gets paid for, which is why inverter datasheets print efficiency numbers with two decimal places and why component selection inside the power stage still decides whether a string inverter design succeeds commercially or quietly loses money over a twenty five year warranty period.
String Inverters Turned IGBT Modules Into The Central Component Deciding How Much Solar Energy Reaches The Grid
A string inverter takes the combined output of several dozen panels wired in series, boosts and regulates that direct current through a maximum power point tracking stage, and then chops it into a sine wave using an H bridge or a multilevel bridge built from IGBTs paired with fast recovery diodes. The industry moved away from large central inverters toward this string based architecture specifically because a fleet of smaller units gives each string its own MPPT channel, so shading or mismatch on one string no longer drags down the output of the whole array. That architectural shift raised the bar for the power devices themselves, since a string inverter in the 25 to 250 kilowatt range needs devices that switch fast enough to keep the output filter small and cheap while still surviving the thermal cycling that comes from a plant running outdoors through fifteen or twenty years of daily heating and cooling. The IGBT earned its dominant position in this range because it combines a MOSFET style gate that needs almost no drive current with a bipolar style output stage that carries high current at low voltage drop, a combination neither a pure MOSFET nor a pure bipolar transistor could offer at the voltage and power levels string inverters operate in.
Conduction Losses And Switching Losses Split The Total Heat An IGBT Dumps Inside An Inverter Cabinet
Every watt an IGBT wastes shows up as one of two distinct loss mechanisms, and a designer has to understand both before picking a device. Conduction loss comes from the collector to emitter saturation voltage, the small voltage drop the device exhibits while fully on, multiplied by the current flowing through it at that moment. Switching loss comes from the brief but real overlap of voltage and current during every turn on and turn off transition, an overlap that happens tens of thousands of times per second at the pulse width modulation frequencies string inverters use. During low irradiance periods, early morning, late afternoon, or under cloud cover, the inverter runs at a fraction of its rated power, switching losses shrink because there is less current to interrupt, and conduction loss becomes the dominant term in the loss budget. At full sun and full rated output the balance shifts back toward switching loss, particularly at the higher end of the 20 to 40 kilohertz range many single phase and three phase string inverters use for their PWM carrier. This is why IGBT datasheets always publish both numbers separately, the saturation voltage curve for conduction and the turn on and turn off energy figures for switching, since a device optimized for one number often gives up ground on the other.
Choosing Between 1200 Volt And 1700 Volt IGBT Classes Depends On Whether The String Runs At 1000 Volts Or 1500 Volts
Voltage class selection is not a detail buried in the datasheet, it is one of the first decisions that shapes the entire inverter design. Systems built around a 1000 volt DC bus, still common in commercial and older utility installations, pair naturally with 1200 volt IGBTs, which leave enough margin to survive transient overvoltage without oversizing the device. The industry's move toward 1500 volt DC strings, driven by the fact that higher string voltage means lower current for the same power and therefore thinner, cheaper cabling and smaller connection losses across the array, pushed inverter designers toward 1700 volt IGBTs, since a 1200 volt device would have almost no safety margin left against transients on a 1500 volt bus. A split neutral point clamped three level topology built specifically for 1500 volt systems, using modules rated at 1200 volts blocking and 1200 amps on the chip itself inside a three level leg, demonstrated an output of 722 amps RMS at a 5 kilohertz switching frequency while holding 98.7 percent efficiency, delivering 750 kilowatts from a single leg pair without paralleling modules and scaling to 1.5 megawatts once two pairs run in parallel. Numbers like these explain why utility scale plants standardized on 1500 volt strings even though it forced a complete redesign of the power stage around a different IGBT voltage class and a multilevel topology instead of a simple two level bridge.
A Worked Loss Budget From A Five Kilowatt Inverter Shows How Conduction And Switching Losses Actually Add Up
A concrete example makes the trade-off between conduction and switching loss easier to hold in mind than any general description. Consider a single phase H bridge inverter rated at 5,000 watts into a 240 volt RMS output at unity power factor, giving a rated RMS output current close to 21 amps and, for a sinusoidal waveform, an average current near 19 amps through each device. If the design target calls for a total loss budget of 66 watts per device to hit the inverter's overall efficiency goal, and the conduction loss at that current works out to roughly 21.8 watts per device based on the IGBT's forward voltage drop characteristic, the remaining 44.2 watts per device becomes the switching loss budget that the PWM frequency has to be chosen against. Pushing the switching frequency higher would shrink the output filter and reduce audible noise and total harmonic distortion, but it eats directly into that 44.2 watt allowance, so the frequency gets set only after the switching loss budget is known, not before. Once the total dissipation is fixed, junction temperature follows directly from the device's thermal resistance from junction to case, and with a case temperature assumption around 100 degrees Celsius, a well chosen IGBT in this power class typically runs well below its 175 degree Celsius maximum rated junction temperature, leaving useful headroom for the ambient temperature swings a rooftop or ground mount enclosure sees over a full year.
Multilevel NPC Topologies Let IGBT Based Inverters Reach Megawatt Power Without Sacrificing Efficiency Numbers
A simple two level H bridge is not the only way to build an IGBT based inverter, and at higher power levels it is rarely the best one. Neutral point clamped three level topologies split the output voltage swing into smaller steps, which cuts the voltage each switching transition has to cross and therefore cuts the switching loss contribution directly, letting a megawatt scale inverter run at a lower effective switching frequency for the same output waveform quality a two level design would need a much higher frequency to match. Field stop IGBT generations built specifically for this kind of multilevel duty have pushed conduction loss down further still. One recent 1200 volt field stop generation reports a saturation voltage of 1.65 volts for its medium speed variant and a turn off energy as low as 57 microjoules per amp for its fast variant, figures the manufacturer credits to a narrow trench mesa cell structure and a proton implanted multiple buffer layer that together shrink chip size by roughly 30 percent compared to the prior generation while increasing current rating. The following mechanisms show up repeatedly across published multilevel and high power string inverter designs aimed at pushing efficiency upward without giving up power density:
- splitting the output voltage into three or five levels through NPC or flying capacitor topologies to shrink the voltage swing each switching event has to cross;
- selecting a field stop or trench gate IGBT generation specifically for its balance of saturation voltage and turn off energy rather than defaulting to the fastest or the cheapest part available;
- pairing every IGBT with a matched fast recovery diode so reverse recovery losses do not undo the gains made on the switching side;
- sizing the switching frequency against a calculated loss budget per device instead of picking a round number like 20 or 40 kilohertz by convention;
- keeping thermal resistance from junction to case low enough that the device runs with real margin below its rated maximum junction temperature across the full ambient range the site will see.
Peak Efficiency Numbers On A Datasheet Rarely Match What A String Inverter Delivers Across A Real Day
The efficiency figure printed largest on a datasheet is almost always peak efficiency, the single highest point on the curve, typically measured at a load between 40 and 60 percent of rated power, at the DC input voltage that happens to minimize both conduction and switching losses for that particular topology, and at an ambient temperature around 25 degrees Celsius. Published string inverter datasheets in the utility and commercial range commonly show peak efficiency figures around 98.6 to 99.1 percent, with European weighted efficiency, which averages performance across a standardized profile of load points meant to resemble a typical day rather than reporting only the best case, running slightly lower at roughly 98.3 to 98.7 percent for the same products. The gap between peak and weighted efficiency exists precisely because a real installation spends most of its daylight hours away from that 40 to 60 percent sweet spot, running instead at partial load in the morning and evening when conduction losses dominate the picture described earlier. A plant designer comparing two inverters on peak efficiency alone can end up choosing the device that loses more energy across an actual year, which is why weighted efficiency figures, whether reported under the European standard or the California Energy Commission standard used in North America, matter more for a bankable energy yield estimate than the bold number on the front page of a datasheet.
Silicon Carbide Devices Are Starting To Challenge IGBTs In The Segment They Once Owned Completely
IGBTs still dominate three phase and higher power single phase string inverters, but wide bandgap silicon carbide MOSFETs have already displaced them in a growing share of smaller single phase residential designs, and the reason comes straight back to the switching loss numbers discussed above. IGBTs carry a turn on energy roughly comparable to silicon carbide devices at a given switching frequency, with IGBTs only slightly worse in that specific comparison, but turn off energy tells a different story entirely, since the tail current that minority carriers create as an IGBT turns off produces a turn off loss that can run an order of magnitude higher than an equivalent silicon carbide MOSFET operating at the same voltage and current. That tail current is a direct consequence of the bipolar conduction mechanism that gives the IGBT its low saturation voltage in the first place, so the same physical trait that makes an IGBT efficient while conducting is what limits how far its switching frequency can climb before losses erase the advantage. Conduction loss behaves differently across temperature as well, since a superjunction MOSFET's conduction loss can roughly double from 25 to 125 degrees Celsius junction temperature as its drift layer resistance grows, while both silicon carbide and IGBTs avoid that severe thermal runaway, silicon carbide because its on-resistance rises only moderately with temperature and an IGBT because the negative temperature coefficient of its PN junction drop partly cancels the positive coefficient of its drift region resistance, so the two technologies reach similar thermal stability through different physical mechanisms even though silicon carbide still pays the switching penalty described above. For now this leaves IGBTs firmly in place across the power range where string inverters actually operate, with silicon carbide reserved for designs where the extra cost buys enough switching frequency headroom, filter size reduction, or thermal margin to justify it, and the choice between the two technologies is decided project by project rather than settled once for the whole industry.