An 800 V DC link is now routine in EV charger power modules, industrial high-voltage supplies, and PV/storage converters. When a team raises the bus voltage, the first thing that gets replaced is almost always the switch: 1200 V SiC MOSFET in, silicon IGBT out. What usually stays on the board are the 1200 V silicon fast-recovery diodes sitting in the PFC or Boost rectifier position, the freewheel position of the bridge leg, and the output rectifier position. The reasoning is usually casual — it is only a diode, pick another package. Then the machine lands half a percentage point below its efficiency target and will not move. Conducted EMI keeps failing around 150 kHz. After an hour at full load, the hottest part on the board is not the switch, it is that diode.

Those three symptoms are three faces of the same accounting error. Before you change the part, there are three calculations worth redoing.

HXY Electronics HC4D10120A is a 1200 V / 10 A silicon carbide Schottky diode with a 1.35 V forward drop, TO-220C-2L package, 250 µA reverse leakage, and a 71 A surge current rating. It has no reverse recovery current and no forward recovery voltage; its switching behavior is temperature-independent, and the positive temperature coefficient of VF supports passive current sharing in parallel without thermal runaway. It targets the PFC/Boost rectifier and freewheel positions of 800 V bus EV charger modules, industrial high-voltage supplies, and PV/storage converters. Same-class alternatives include HC3D10120E (TO-252-2L, 10 A, VF 1.40 V, IFSM 90 A), HC3D10120G (TO-263, 10 A, VF 1.40 V, IFSM 90 A), and HC4D20120H (TO-247-2L, 20 A, VF 1.50 V, IFSM 130 A).

1. Why the 1200 V Class Is Where the Math Goes Wrong

The problems with silicon fast-recovery diodes on a 400 V bus (with 650 V parts) have been documented for years. Move to an 800 V bus and nothing gets easier — three things get worse.

Voltage headroom is already spent. Pairing an 800 V bus with 1200 V parts is the industry default, roughly 67% voltage utilization. That pairing is not arbitrary. The standard derating practice for power devices holds the working voltage, turn-off overshoot included, to 80%–90% of the rating — for a 1200 V part that means the bus plus overshoot has to stay inside 960–1080 V. Once the bus goes to 800 V, only 150–200 V is left for commutation ringing. If the laminated busbar is weak and loop inductance climbs to tens of nanohenries, a commutation at several thousand A/µs on an 800 V bus will push the overshoot past 200 V and eat most of that margin.

Losses double. Reverse-recovery loss scales with bus voltage and with switching frequency. Take 300 nC as a mid-range figure for a 1200 V silicon fast-recovery diode (industry-class reference, not a specific part). At 400 V and 65 kHz that is roughly 7.8 W. Carry the same device to an 800 V bus and it becomes roughly 15.6 W. And most of that 15.6 W is not dissipated in the diode — it lands on the opposing switch as a turn-on current spike.

The thermal feedback loop is harder to break. Reverse recovery charge in a silicon fast-recovery diode is a strong function of junction temperature: hotter junction → larger Qrr → more loss → hotter junction. A heatsink with comfortable margin on a room-temperature bench enters that loop during full-load high-temperature burn-in, and efficiency and thermal data fail together.

All three trace back to one physical source: minority carrier storage in a bipolar device. The only real cure is a unipolar device with no stored charge. Silicon Schottky has no reverse recovery but stops being practical above roughly 200 V. The production device that delivers both no reverse recovery and 1200 V blocking is the silicon carbide Schottky diode.

2. Calculation One: Reverse Recovery Charge, Switching Loss, and the Current Spike

Start with where the diode sits. In a classic Boost rectifier position the inductor runs from the input to the switch node at the transistor drain, and the diode runs from that switch node to the positive terminal of the bus capacitor. When the switch is on, the inductor stores energy and the diode blocks the full bus voltage. When the switch turns off, inductor current flows through the diode into the bus. The trouble begins at the instant the switch turns on again and the diode is forced into reverse.

A silicon fast-recovery diode is bipolar. During forward conduction its drift region is flooded with stored minority carriers; when the voltage reverses, that stored charge must be swept out before the device can block again. The sweeping process is a reverse current spike that adds directly to the switch turn-on current. Three consequences follow: turn-on loss climbs sharply, switch junction temperature is pushed up, and the recovery current interacting with loop stray inductance produces ringing that dirties the EMI spectrum.

Put numbers on it. Take an 800 V bus, a 10 kW single-channel Boost/PFC, switching at 65 kHz. Diode average current is Io = 10000 W / 800 V = 12.5 A:

The difference at the same position is about 20 W, or roughly 0.2 percentage points on a 10 kW module. More importantly, most of the 15.6 W comes off the switch turn-on edge — switch junction temperature, heatsink size, and snubber component count all come down with it. And the remaining junction capacitance charge Qc on the SiC side does not move with temperature or di/dt, so this accounting does not turn against you at full load and high temperature. (Refer to the relevant datasheet for the actual Qc value.)

3. Calculation Two: Surge Current IFSM and Inrush Verification

Reverse recovery is a question of how good the design is. Surge is a question of whether the design survives — and it is the calculation most often skipped.

IFSM on a diode datasheet is the non-repetitive forward surge current. The industry convention is a 10 ms single half-sine (the half period of a 50 Hz line), specified at both 25 °C and an elevated case temperature. It describes transients — startup, faults, load steps — and has nothing to do with the steady-state average current rating. A 10 A diode may have a surge capability of tens or hundreds of amps depending on die area and structure.

To turn IFSM into something you can check against a real event, use the industry conversion:

i²t = 0.005 × IFSM² (10 ms half-sine convention)

Applying it to the listed IFSM values shows how capability is distributed across the 1200 V class:

Now check it against a real cold-start event. 800 V bus, 470 µF DC-link capacitance (a typical value class for 10 kW), charged through a precharge resistor Rpre:

• Rpre = 20 Ω: peak charging current 800 V / 20 Ω = 40 A, time constant τ = 20 Ω × 470 µF = 9.4 ms. Energy of the decaying exponential, ∫i²dt = Ipeak² × τ / 2 = 40² × 0.0047 ≈ 7.52 A²s. Against HC4D10120A‘s 25.2 A²s that is roughly 3.4× margin, and the 40 A peak is well under the 71 A IFSM. Pass.
• Rpre dropped to 5 Ω to shorten precharge time: peak becomes 160 A, τ = 2.35 ms, ∫i²dt = 160² × 0.001175 ≈ 30.1 A²s. That exceeds 25.2 A²s, and the 160 A peak is 2.3× the 71 A IFSM. Out of specification.

The conclusion is direct: the precharge resistor cannot be sized on precharge time alone — it must be checked back against the rectifier diode‘s IFSM and i²t. The more robust approach is to give the DC-link capacitor its own precharge path with a high i²t capability (with an NTC or a relay for current limiting), so the high-frequency rectifier never participates in cold-start charging. That is basic reliability practice in high-voltage supplies. One more caution: IFSM is specified at a defined pulse width and starting case temperature, and the high-temperature figure is meaningfully lower than the 25 °C headline number. Verify against the surge curve at the actual case temperature (per the datasheet).

4. Calculation Three: Parallel Current Sharing Under a Positive VF Temperature Coefficient

A SiC Schottky diode‘s forward voltage has a positive temperature coefficient. The part that runs hotter develops a higher forward drop, so current is naturally pushed toward its cooler parallel partner. That is passive negative feedback: parallel operation needs no ballast resistors, and the silicon PN junction failure mode — current crowding into the hottest die, which gets hotter, which pulls more current — has no path to develop.

Run the same 10 kW / 800 V machine. Diode average current is 12.5 A, and there are two ways to cover it:

Two in parallel save roughly 1.9 W over the single 20 A part, run each die at 62.5% utilization, and split the heat across two packages. The lower VF is where the 1.9 W comes from; the positive temperature coefficient is what makes the arrangement stable.

But paralleling is not unconditional. Four situations turn it against you:

Do not mix VF bins. In the table above HC4D10120A is 1.35 V and HC4D10120H is 1.50 V — a 150 mV spread. Paralleled, the 1.35 V part carries a persistently larger share. The positive temperature coefficient compensates some of that, but it cannot cancel a 150 mV static difference. The industry rule of thumb is that a 50 mV VF difference at the 20 A class can produce more than 20% current imbalance. If you parallel, parallel the same part from the same lot.

Layout and cooling must be symmetrical. The positive temperature coefficient corrects thermal imbalance; it cannot correct asymmetric copper resistance, lead length, or loop inductance. Parallel pairs want symmetrical copper, a shared heatsink, and equal-length connections so each die sees the same environmental impedance.

Do not count on the temperature coefficient during a surge. Thermal feedback operates on a millisecond-to-second time constant, while IFSM describes a 10 ms transient. In the first few milliseconds of a surge, sharing is set by resistive matching, not temperature. Surge capability does not simply double when you parallel two devices — taking 1.6× to 1.8× of the single-device IFSM and leaving margin is the safer engineering call.

Prefer one larger device when you can. Doubling the device count doubles BOM lines, solder joints, layout constraints, and risk surface. In most cases the larger single part (a 20 A HC4D20120H, for example) is the better answer, with paralleling reserved for cases a single device genuinely cannot cover, or where you need to spread heat across two regions of the board.

5. Reference Circuit: The Full Signal Chain at an 800 V Boost Rectifier Position

Put the device back into the system. Take an 800 V bus, 10 kW two-phase interleaved Boost/PFC — the front end used in EV charger power modules, industrial high-voltage supplies, and PV/storage converters.

Power path: input → EMI filter → boost inductors L1/L2 (designed on ripple current and saturation current) → switches Q1/Q2 (1200 V SiC MOSFET or a high-voltage superjunction part, depending on efficiency versus cost priorities) → boost diodes D1/D2 (the subject here, 1200 V SiC Schottky) → bus capacitor Cbus (800 V, sized on holdup time and ripple current) → isolated DC-DC stage or inverter bridge downstream.

Control loop: a DSP or dedicated PFC controller generates two PWM channels 180° apart, with an outer voltage loop regulating the bus and an inner current loop shaping the input current. Switching frequency lands in the 45–100 kHz range: the upper bound is set by the harmonic margin against the 150 kHz conducted-EMI start frequency and by switch turn-on loss. Once the diode is SiC, reverse recovery stops being the frequency limiter — switch loss and busbar overshoot are.

Drive loop: an isolated gate driver (drive voltage and negative turn-off set by the VGS requirements of the chosen SiC MOSFET) with 2–4 A class peak current and propagation delay on the order of 100 ns. Interleaving requires matched delay between the two channels, or the ripple cancellation benefit erodes. The diode has no drive stage of its own, but it does couple into one: its junction capacitance charges and discharges at every commutation, and that displacement current adds to the switch turn-on current. At nanocoulomb levels, that is what "clean commutation" physically means — and why the tall, wide spike on the switch turn-on waveform disappears after the swap.

Power loop: Q turns on, inductor current ramps, D blocks the bus. Q turns off, inductor current freewheels through D into the bus at the listed VF. Q turns on again, D commutates into reverse — no recovery current, only a small step set by Qc.

Layout constraints: minimize the commutation loop area formed by Q, D, and Cbus. At 800 V, commutation di/dt reaches several thousand A/µs; design the loop inductance to 30 nH or less (laminated busbar, or double-sided heavy copper with dense stitching vias), which corresponds to roughly 200 V of turn-off overshoot — add that to 800 V and the total sits inside the 80%–90% derating window for a 1200 V part. Note also that creepage and clearance requirements at 800 V are substantially tighter than at 400 V; size the spacing between high-voltage nodes against the system voltage class and overvoltage category (per the applicable clauses of IEC 62109 or equivalent).

6. Five Things to Get Right When Moving to 1200 V SiC Diodes

Keep inrush off the high-frequency rectifier. The bus capacitor charges from zero at cold start, and the inrush current far exceeds normal operating current. Route precharge through a separate high-i²t path (with an NTC or relay for current limiting), and verify IFSM and i²t using the method in Section 3. This is the calculation most often skipped in rectifier design and the one most likely to surface in production.

Read IFSM at temperature and pulse width, not the 25 °C headline. IFSM is defined at a specified pulse width and starting case temperature, and the elevated-temperature number is meaningfully lower. Verify against the surge curve at the worst-case case temperature (per the datasheet).

Parallel only same-part, same-lot devices with symmetrical layout and cooling, and do not multiply the surge rating. The positive temperature coefficient gives you passive sharing but cannot fix static VF spread or parasitic asymmetry. Derate the parallel pair by 20% or more, and take surge capability at 1.6×–1.8× the single-device figure.

Re-check snubbers and EMI after the swap. Commutation is cleaner, but the dv/dt of the switching edge changes. RC snubbers and EMI filter values tuned for silicon fast recovery may need adjustment — measure the spectrum and decide, do not carry values forward by assumption.

Account for high-temperature reverse leakage. Leakage in a Schottky device rises with junction temperature, and at high temperature under full reverse bias the leakage loss (leakage current × reverse voltage) belongs in the thermal balance equation. At the listed room-temperature IR, HC4D10120A dissipates about 0.3 W of leakage loss at a full 1200 V reverse bias — negligible at room temperature. But leakage climbs along the datasheet curve, so sealed enclosures at high ambient deserve a pass through the numbers rather than a thermal design built on VF alone.

7. Selection Summary

Close the three accounts: reverse recovery decides whether efficiency and EMI targets are reachable, surge decides whether startup and fault conditions are survivable, and current sharing decides whether a paralleled solution stays stable over time. Work all three and the part number converges on its own.

HXY Electronics is a power semiconductor brand with ten years in the industry and more than 5000 catalog part numbers. Its silicon carbide diode line covers 122 devices across 650 V, 1200 V, and 1700 V classes, with rated current from 2 A to 250 A, forward drop from 1.3 V to 1.7 V, and surge current from 18 A to 225 A, in more than ten package formats including TO-247-2L, TO-220C-2L, TO-263, TO-252-2L, DFN8X8B, and SOT-227. HXY staffs field application engineers for selection support, and parts are available through sample request, online ordering, volume quotation, lead-time inquiry, and scheduled (forward) ordering.

One sentence to close: if efficiency is stuck, EMI will not pass, and the diode runs hotter than the switch, do not reach for a bigger heatsink first. Put a current probe on the switch turn-on waveform. If that spike is tall and wide, the silicon fast-recovery diode is the next device whose numbers need redoing.

Disclaimer: Industry comparison figures in this article are drawn from public sources and are provided for selection reference only; they do not constitute a procurement or design commitment. Refer to the latest official HXY datasheet for electrical parameters. For selection and technical support, contact HXY Electronics at 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