SiC Diodes, SiC MOSFETs, IGBTs & Power Modules Manufacturer | Hypersemi
TechForum - application notes and selection guides

TECH FORUM

Dropping a Domestic EconoDUAL3 Half-Bridge Module Into a PCS: Can the Heatsink Stay?

Dropping a Domestic EconoDUAL3 Half-Bridge Module Into a PCS: Can the Heatsink Stay?

In the 100 kW–500 kW range of energy-storage PCS, EconoDUAL3 is the packaging benchmark that cannot be bypassed. Infineon's FF series has built a large installed base in this footprint, which makes "pin-to-pin domestic replacement" an easy narrative to accept: pin definitions match, mechanical dimensions match, so it should just drop in. But package compatibility only solves the "it fits" problem. What actually determines whether the module survives a PCS load profile is junction temperature.

Two modules in the same package, with opposite loss distributions

Hypersemi offers two 1200 V / 200 A half-bridge modules in EconoDUAL3, positioned against Infineon FF225R12ME4_B11 and FF200R12KT4 respectively. Placed side by side, they lead to a conclusion that is not particularly "comparison-table friendly."

HGF200MP120X100: VCE(sat) 1.81 V, VF 2.03 V, Eon 11.7 mJ, Eoff 24.8 mJ, Tjmax 175°C. Turn-off loss is more than twice the turn-on loss.

HGF200MA120X100: VCE(sat) 1.84 V, VF 2.42 V, Eon 19.97 mJ, Eoff 13.08 mJ, Tjmax 150°C. Turn-on loss clearly exceeds turn-off loss.

The VCE(sat) values differ by only 0.03 V—on a quick read, "conduction characteristics are almost identical." But the switching-loss distribution is completely inverted. A PCS does not operate at a single full-load point: charge/discharge transitions, light-load operation, and short-time overload all shift the weighting of Eon and Eoff in total loss.

If the target operating range is turn-on-loss dominated, the low Eon of HGF200MP120X100 is an advantage; if it is turn-off dominated, the low Eoff of HGF200MA120X100 becomes the better fit. Making a replacement decision from a VCE(sat) comparison table alone hides the variable that matters most.

Rth(j-c) differences get amplified when the heatsink stays fixed

The typical Rth(j-c) for the Infineon FF225R12ME4 family is on the order of 0.14 K/W, and about 0.135 K/W for FF200R12KT4. Hypersemi's HGF200MA120X100 is specified at 0.15 K/W for the IGBT portion of module thermal resistance. On paper, a 0.01–0.015 K/W increment produces only 1–1.5°C of junction-temperature rise at 100 W of loss.

But thermal design is never a single static loss point. A PCS load profile includes minute-scale power ramps, charge/discharge direction reversals, and short-time overload. During these transients, instantaneous chip loss can be several times the steady-state value, while the heatsink's thermal time constant is far larger than the chip's. A "fixed" heatsink means the case-to-ambient path is locked; the peak and swing of junction temperature are then governed almost entirely by the module's internal Rth(j-c) and chip thermal capacitance. A 0.01 K/W larger Rth(j-c) does not stay at the 1°C level under overload pulses—the transient component of thermal impedance stretches the difference.

The more insidious difference is Tjmax. HGF200MP120X100 is rated at 175°C, HGF200MA120X100 at 150°C. If the heatsink was dimensioned against the former's thermal margin, dropping in the latter consumes 25°C of safety boundary before any load is applied. Same package, same pins—but a different thermal boundary.

What a replacement review should actually do: put the loss profile and thermal impedance into the same model

The correct replacement check is not a VCE(sat) comparison. It is re-computing junction temperature with the target Eon/Eoff weighting, Rth(j-c), and heatsink thermal resistance in one thermal network.

In engineering practice, this is typically done with a Foster thermal network: junction-to-case is represented by a multi-stage RC network, case-to-heatsink and heatsink-to-ambient by first-order RC stages, and junction temperature is the superposition of the temperature differences across the nodes. Foster parameters can be extracted directly from the module datasheet without needing internal chip geometry, making them suitable for a fast electro-thermal calculation during replacement evaluation.

Three inputs are essential: the IGBT and diode loss waveforms corresponding to the target load profile, the module's Rth(j-c) data, and the existing heatsink's Rth(h-a) (or a measured case-temperature boundary). Substituting the domestic module's Rth(j-c) and Eon/Eoff into the same model yields the junction-temperature trajectory after replacement—not two parameter tables viewed side by side.

The point of having both HGF200MP120X100 and HGF200MA120X100 is precisely to force that check to happen. Two modules in the same package and current class with opposite loss distributions make it clear that EconoDUAL3 domestic replacement is not a one-to-one mapping of module to module—it is a thermal-design action that requires selection by operating profile.

Package compatibility opens the electrical and mechanical entry points. Thermal-design compatibility has to be earned separately.

Back to top