For any AC-DC power supply running from a 220V mains input — server PSUs, telecom rectifiers, LED drivers, high-power chargers — the power factor correction (PFC) stage is almost universally built on a 400V DC bus, and the high-frequency switch on that bus has been, for the past decade, the domain of 600V/650V silicon superjunction MOSFETs, with CoolMOS as the archetype. Today, as efficiency red lines keep moving up (80 PLUS Titanium for server PSUs, half-load efficiency mandates for telecom rectifiers), superjunction devices are hitting physical walls on three separate parameters, and 650V SiC MOSFETs have moved to the front line of the replacement discussion. Yet what holds many engineers back is not performance — it is the gate drive: early-generation SiC required a negative bias at turn-off, meaning the driver circuitry had to be rebuilt from scratch. This article works through both the efficiency math and the gate drive math.

HXY Electronics HXYS52N65MPI is a 650V / 52A / 40mΩ silicon carbide N-channel MOSFET in a TO-247-4L four-pin Kelvin-source package, supporting 0V turn-off (turn-on +15V to +18V, turn-off 0V, compatible with unipolar silicon MOSFET gate drive architectures). It targets 400V DC-bus power supplies: totem-pole PFC high-frequency half-bridges, CCM Boost PFC stages, and LLC resonant converters, as a like-for-like replacement for 650V superjunction CoolMOS devices. Its 40mΩ on-resistance benchmarks directly against the mainstream superjunction 41mΩ class, its high-temperature on-resistance increase is roughly 60% of the superjunction figure, and the four-pin Kelvin source significantly reduces high-frequency switching ringing and switching loss.

1. The Efficiency Race on the 400V Bus: Three Walls That Superjunction CoolMOS Has Hit

Superjunction technology used charge balancing to bring the on-resistance of 650V devices down to tens of milliohms — that is its achievement. But from the perspective of a high-frequency power supply designer working on a 400V bus, three physical properties are turning into design liabilities.

The first wall is body-diode reverse recovery. When the superjunction body diode conducts, the drift region is flooded with minority carriers; turning it off takes time to extract that charge. Take the industry benchmark — the 650V CoolMOS CFD7 series, specifically engineered for a fast body diode — in its 41mΩ class: the public datasheet lists a reverse recovery charge of approximately 1.2µC typical and a reverse recovery time of roughly 177ns. Note that these figures are measured under the gentle commutation condition of 100A/µs. As frequency and di/dt climb, this tail current converts directly into turn-on loss, ringing, and EMI. This is precisely why the high-frequency leg of a totem-pole PFC cannot use superjunction devices, and why designers fall back to conventional Boost or interleaved CrM schemes.

The second wall is high-temperature on-resistance. By industry measurement conventions, a 650V superjunction device typically sees its on-resistance rise to more than 2.5x its 25°C value at a 175°C junction temperature, whereas a SiC MOSFET rises by roughly 1.5x. Inside a sealed power supply enclosure, full-load junction temperatures sit above 100°C continuously — the 40mΩ you picked from the front page of the datasheet becomes 100mΩ in real operation, and several watts of extra conduction loss eat the fan budget.

The third wall is output capacitance nonlinearity. Superjunction Coss inflates by hundreds of times in the low-voltage region; at light load the resonant current in an LLC converter cannot fully discharge the Coss charge within the dead time, ZVS fails to establish, and the stored Eoss dissipates as heat. Light-load efficiency stalls — and Titanium certification grades precisely at the 10%–20% load points.

The engineer who wants to switch to SiC usually has exactly one reservation: older SiC MOSFETs have low threshold voltages, and under high-frequency, high-dv/dt operation the Miller effect invites parasitic turn-on. The classic drive scheme is therefore +18V on / -5V off, requiring a negative rail or a dual-output isolated supply — added BOM and added design complexity on the driver side. That gate drive math is the focus of this article.

2. HXYS52N65MPI: A Direct Benchmark Against CoolMOS — Without Rewriting the Gate Drive

The HXYS52N65MPI from HXY Electronics is a 650V / 52A / 40mΩ N-channel SiC MOSFET in a TO-247-4L four-pin Kelvin-source package. It earns the leading role in a superjunction replacement story for two reasons: its 40mΩ benchmarks directly against the superjunction 41mΩ class, giving loss calculations a like-for-like reference; and it supports 0V turn-off, preserving the unipolar gate drive architecture inherited from silicon MOSFET design.

The table below compares it against the mainstream 650V superjunction CoolMOS (CFD7 41mΩ class, industry-public data):

Running the numbers. Conduction loss: for a 2kW-class PFC with an inductor RMS current of about 10A, 40mΩ yields roughly 4W at 25°C. At 175°C, the superjunction at 2.5x becomes 100mΩ — about 10W at 10A — while the SiC at 1.5x becomes 60mΩ — about 6W. Roughly 4W of conduction loss separates them in hot, full-load operation, and inside a sealed enclosure 4W means a smaller heatsink and a slower fan.

The switching and frequency math: SiC body-diode reverse recovery charge is more than an order of magnitude below the benchmark fast-recovery superjunction (by industry-public comparisons, Infineon‘s CoolSiC 650V reduces Qrr by about 80% versus CoolMOS CFD7). Turn-on loss drops accordingly, and switching frequency can move from the superjunction range of 65–100kHz up to 150–250kHz — the PFC inductor and the LLC resonant inductor shrink with it, stepping the whole unit up a power-density class. As frequency climbs, copper and core losses in the magnetics take a larger share of the loss budget; pair the frequency move with litz or flat-wire windings and low-loss core materials to capture the full dividend.

The gate drive math is the highlight this time. The HXYS52N65MPI supports 0V turn-off: +15V to +18V recommended for turn-on to reach full conduction performance, and a straight 0V for turn-off — no negative rail. This preserves the unipolar drive architecture of the silicon MOSFET era: one isolated gate driver plus one positive supply. Bootstrap supplies and single-auxiliary-winding schemes remain usable, and the PCB does not need a re-layout for a negative rail. The pin-1 Kelvin source routes the drive loop separately, keeping power-loop stray inductance out of the gate loop, and significantly reducing gate ringing and parasitic turn-on risk during high-dv/dt switching.

A note on scope: dynamic parameters such as VGS(th), Qg, Eoss, and switching-loss curves should be taken from the device datasheet; the figures in this article are industry-class references.

3. Typical Application: The Complete Signal Chain of a 2kW Totem-Pole PFC High-Frequency Half-Bridge

Consider the front end of a 2kW telecom rectifier (220V mains input, 54V/37A output) and the position of the HXYS52N65MPI within it. The totem-pole PFC consists of four power switches forming two legs: the slow (line-frequency) leg commutates at 50Hz following the grid polarity, while the fast (high-frequency) leg does the switching; the boost inductor connects between the AC input and the midpoint of the fast leg, charging the 400V bus capacitor.

The complete signal chain, in four links.

Control link: a digital controller or a dedicated totem-pole PFC controller (UCC28056, NCP1681-class, or a DSP-based digital scheme) runs CCM average current-mode control with a switching frequency set between 100kHz and 150kHz. The current-loop bandwidth stays above 5kHz to meet THD and power factor targets, while the controller manages the synchronized commutation logic between the line-frequency and high-frequency legs.

Drive link: each of the two high-frequency-leg switches receives one isolated gate driver channel (UCC21520-class dual-channel isolated drivers), driven at +15V or +18V on and 0V off. Because the design is 0V turn-off, the driver side needs only a single positive supply — bootstrap or independent auxiliary winding — and propagation delays in the 100ns class to match dead-time accuracy. Choose drivers with 2A-class peak source/sink current to match SiC‘s 100ns-class switching edges, and prefer Miller-clamp driver variants for the high-dv/dt totem-pole half-bridge. Gate resistance starts in the 5–20Ω range, balancing switching loss against EMI.

Power link: Q1 and Q2, two HXYS52N65MPI devices, form the high-frequency half-bridge. In the positive half-cycle, Q1 switches at high frequency while the inductor stores energy; when Q1 turns off, current freewheels through the Q2 body diode into the 400V bus capacitor. The negative half-cycle is symmetric. At 2kW, input RMS current is about 9A and peak inductor current about 13–15A; the 52A rating leaves roughly 3.5x headroom. At 40mΩ, conduction loss at peak current is on the order of 7W, which the TO-247-4L thermal characteristics handle comfortably on a baseplate. The reason to choose SiC over superjunction for the fast leg is the body-diode Qrr: under CCM hard commutation, the superjunction‘s microcoulomb-scale Qrr produces a recovery peak several times the load current, whereas SiC‘s nanocoulomb-scale Qrr lets the topology operate stably in continuous conduction mode — that is the dividing line on which a totem-pole PFC stands or falls. The line-frequency leg, switching only at 50Hz, may use superjunction silicon or SiC; cost-first designs can stay with superjunction there.

Load link: downstream of the 400V bus capacitor, the LLC resonant converter‘s primary half-bridge can likewise use the same 650V SiC family. SiC‘s more linear Coss characteristic lowers the ZVS establishment threshold at light load, and the smaller Qg keeps drive losses in check at high frequency — efficiency benefits at both half load and light load.

4. Design Pitfall Guide: 0V Turn-Off Works — But Do Not Overlook These Five Things

First, understand the boundary of 0V turn-off. It simplifies the drive, and its margin rests on a sensible threshold design. In mid-to-low frequency (below 80kHz), cost-sensitive Boost or single-switch topologies, unipolar 0V turn-off is a BOM-saving choice. But in hard-commutating, high-dv/dt half-bridges such as totem-pole legs, prefer gate drivers with Miller clamp capability, or reserve the option of a -2V to -5V turn-off as a safety net — the negative bias is not a requirement, it is a fuse; industry application notes share this consensus. In one sentence: the topology dictates the drive margin; do not trade reliability for driver cost savings.

Second, respect the gate voltage limit. SiC MOSFET gate oxide is thinner than silicon‘s; industry-class devices carry VGS absolute maximums in the -10V to +22V range (per datasheet). On drive voltage, err on the low side rather than accept overshoot: +15V to +18V is the recommended window, and layout ringing stacked on drive spikes can transiently breach the limit. A nearby zener clamp plus an RC snubber between gate and source is cheap insurance. Gate-oxide breakdown is a common and irreversible SiC failure mode, and there is no field repair for it.

Third, keep Kelvin-source layout discipline. The fourth pin of the TO-247-4L exists for the drive loop: gate resistor, driver output ground, and the Kelvin pin form one small independent loop, joining the high-current power-source path at a single point, so power-loop inductance does not induce voltage into the gate loop. Keep drive traces short, route them in parallel pairs, and avoid crossing plane splits.

Fourth, switch tracks on voltage-margin thinking. Motor-drive designs favor the conservative rule of keeping the bus below half the device rating — that exists for stall back-EMF and doubled spikes. An AC-DC PFC‘s 400V bus is clamped by the bus capacitor; the stress picture is bus voltage plus switching ringing, and the actual 450–550V stress with a 650V device at 1.2–1.4x margin is mainstream industry practice. Manage ringing with layout and snubbers, not by stacking voltage ratings.

Fifth, always run loss calculations hot. Do not stop at the 25°C RDS(on) headline: inside sealed enclosures, 110–125°C junctions are normal. Compute conduction resistance at the 175°C convention (superjunction ~2.5x, SiC ~1.5x), then size the thermal resistance and heatsink — that is what passes full-load burn-in testing. Meanwhile, the LLC dead time can leverage the SiC body diode‘s short recovery time to compress to the 100ns class, but do not push it to zero: leave the body diode its physical commutation time, and keep margin between circulating-current loss and shoot-through risk.

5. Vendor and Summary

In the 400V-bus power supply arena — the most contested 650V device segment — the logic of replacing CoolMOS with SiC is now clear: not sentiment about "new replacing old," but the dividends of three physical parameters (Qrr, high-temperature conduction behavior, Coss linearity), plus 0V turn-off driving the gate-drive conversion cost to nearly zero. HXY Electronics, as a power semiconductor supplier, offers a 650V SiC MOSFET portfolio spanning mainstream packages (TO-247, TO-247-4L, TOLL, DFN) and multiple on-resistance classes; the HXYS52N65MPI (650V / 52A / 40mΩ / TO-247-4L, 0V turn-off supported) is the direct option in that lineup for kilowatt-class PFC/LLC superjunction replacement. Product parameters and volume availability are available via sales@hxymos.com.

In one sentence: first run the efficiency math (high-temperature conduction loss, switching loss, the frequency dividend), then run the drive math (0V turn-off eliminates the negative rail). When both ledgers clear, the replacement window for CoolMOS is open.

Disclaimer: Industry comparison data cited in this article is drawn from public sources and serves only as a technical selection reference; specific electrical parameters are subject to HXY‘s official datasheets. For selection and technical support, contact sales@hxymos.com.

声明 / Disclaimer
本文部分参数引自华轩阳产品手册或第三方规格书，数据有更新可能，仅供选型参考，最终设计请以最新版规格书与实测为准。
联系 Contact：sales@hxymos.com