Page 9 - PEN eBook May 2023
P. 9

COVER STORY – DESIGN                                                                          COVER STORY – DESIGN

 plate, which was isolated from the gate electrode and   low- and medium-voltage power MOSFETs. In order to
 instead electrically connected to the source potential   also further reduce the FOM  = R DS(on)  × Q  and
 g
 g
 (Figure 1d). While the charge compensation principle   FOM  = R DS(on)  × Q  values, the gate trench underwent
 gd
 gd
 operates as before, the buried field plate does not   a complete redesign to minimize its lateral extension.
 introduce any additional contributions to the   However, the substantially smaller dimensions of the
 gate-drain capacitance. Instead, the field plate shields   gate impose a new challenge, as the use of polysilicon
 the gate electrode from the drain potential, which   as gate material would result in unacceptably large
 reduces the gate-drain capacitance C  and related   internal gate resistances. The introduction of gate
 gd
 charges. These devices, at the time of their introduction   fingers usually solves this issue, but these reduce the
 to the market, showed best-in-class performance with   active area available for current conduction.
 low gate charge and gate-drain charge characteristics,
 high switching speeds and good avalanche ruggedness. 3  Instead, a new metal gate system has been introduced
 to avoid any loss of active area, which otherwise would
 INFINEON’S INNOVATIVE APPROACH   be rather significant. This system not only reduces the
 TO RAISING POWER MOSFET DESIGN   internal gate resistance but also enhances the gate
 TO THE NEXT LEVEL  resistance uniformity across the chip.  Furthermore,
 4
 To reach the next level in power MOSFET evolution,   the field plates are directly connected to the source
 new MOSFET devices are required to provide   metal, ensuring a rapid and homogeneous transition at
 improvements across all FOMs. This is needed to enable   turn-on and turn-off. This minimizes switching losses   Figure 3: System diagram of the IBA
 high-frequency switched-mode power supply (SMPS)   and mitigates the risk of an undesired dv/dt-induced
 operation, whereby losses are associated with charges   parasitic turn-on of the MOSFET.  digital loads (xPUs and ASICs).  Depending on the end   in these applications, showing typical peak efficiencies
                                    5
 (switching) and on-state resistance (conduction). To   application, PoL converters can be optimized to operate   that exceed 98%, much higher than their regulated
 meet these more demanding requirements, a novel   BOOSTING END-TO-END CONVERSION   either with a narrow or wide input voltage range.   counterparts.
 cell-design approach has been developed and   EFFICIENCY IN TELECOM AND DATA   In telecom systems, where the –48-V bus shows wide
 implemented, which explores for the first time a true   CENTER SYSTEMS  tolerance, it has been common for the PoL regulators   DEVICE BEHAVIOR AND EFFICIENCY
 3D charge compensation.  Intermediate bus converters (IBCs) are considered a   to operate with a narrow input voltage range (e.g., 12 V),   MEASUREMENTS UNDER
 demanding application for power MOSFETs. As part of   thus requiring a regulated IBC. Taking the burden of the   HARD-SWITCHING CONDITIONS:
 First, a direct connection of the field-plate electrodes   the intermediate bus architecture (IBA), the IBC is a   regulation from a wide input range   TESTING A 600-W IBC FOR TELECOM
 to the top-side source metal is required, as illustrated   DC/DC converter that performs an intermediate   (–36 V to –75 V), the IBC tends to be rather inefficient   APPLICATIONS
 in Figure 1e. Second, the device layout must move away   conversion to supply the downstream point-of-load   and plays an important role in defining the   The IBC in this application operates as an isolated
 from the common stripe layout to a grid-like layout   (PoL) step-down converters, as shown in Figure 3.  end-to-end conversion efficiency. Improving the IBC   DC/DC IBC with a nominal –48-V input (overall range
 structure, as depicted in Figure 2. This increases the   efficiency is thus paramount to boosting the overall   from –36 V to –75 V) and a 12-V output voltage bus.
 silicon area for current conduction compared with a   This architecture is prevalent in telecom and data   conversion efficiency.  The fully regulated converter in the industry’s standard
 structure with stripes, allowing a further reduction of   centers and aims to achieve the best conversion   quarter-brick form factor operates at a switching
 the overall on-resistance in the new OptiMOS™ 6   efficiency from the AC/DC power supply unit to the   In modern data center systems or advanced AI   frequency of 250 kHz and can deliver an output current
          hardware accelerators, extremely high currents have   of a maximum of 50 A. The IBC is based on a
          to be supplied to sub-1-V digital loads. The efficiency   hard-switching full-bridge (FB) topology with a
          of the two stages can be maximized by playing with the   center-tapped (CT) synchronous rectifier (SR) on the
          down-conversion ratio of the IBC and with the burden   secondary side, schematically shown in Figure 4.
          of regulation being transferred to the multiphase
          PoL/voltage regulator module. Indeed, unregulated IBCs
          in the form of DC transformers (DCX) are employed




















          Figure 4: Simplified schematic of the 600-W IBC board in   Figure 5: Efficiency in the 600-W IBC comparing the new and
 Figure 2: Comparison of the commonly used stripe layout with the new grid-like layout approach  FB-CT configuration  predecessor technology

 8  MAY 2023 | www.powerelectronicsnews.com                                 MAY 2023 | www.powerelectronicsnews.com  9
   4   5   6   7   8   9   10   11   12   13   14