The world of power delivery is currently going through a very exciting time now that several major semiconductor companies are offering integrated power modules. Some of these devices combine the PWM controllers, gate drivers and switching MOSFETs in one package. With these modules, you only need to design the input and output filters, and you have got an optimized dc-dc converter. What’s more, these devices usually operate at high switching frequencies, which results in smaller footprints when compared to discrete solutions.
These integrated dc-dc modules are usually offered for load currents less than 10 A. Among these components, almost all of the lower-current devices (<2 A) are monolithic. Another category of integrated modules — the DrMOS class of devices — targets higher current dc-dc applications.
Designed for use in synchronous buck converters, DrMOS devices comprise the gate drivers, the control MOSFET and the synchronous rectifier, and can deliver currents ³30 A per module. These modules are intended for use in servers and PCs, where multi-phase synchronous buck converters are the topology of choice.
With all of these types of modules, the main advantages are high performance, small footprints and reduced development efforts for the design engineer. Using these modules saves design, component selection and prototyping time. However, all of these advantages must come to the market with a high degree of reliability guaranteed by the supplier.
Though the electrical design of these modular subsystems is inherently complex, this complexity is compounded by requirements for high-speed current switching, power dissipation management, package thermal resistance and thermal performance in elevated ambient temperatures. These are just a few of the factors that affect a module’s reliability.
All of this demands a rigorous design approach, design verification and final testing of the device. In multi-chip modules (MCMs), the design of each individual component must be optimized as a part of a subsystem not as simply a stand-alone device. In other words, each component’s performance and interface parameters must be a perfect fit for every other component it interacts and interfaces with. This clearly means that even the best-in-class, mass-produced general-purpose devices may not be thrown together to make a module, but rather the module specifications must dictate the individual specification of its parts.
The first step in design verification is to make sure that the actual design meets the specifications. However, this process must be expanded to verify that these specs are met under the most demanding conditions in the module’s intended application. Those who user new technologies are very sophisticated engineers and are likely to find ways to use the devices that never occurred to the module makers. In many cases, engineers will place the modules in extremely demanding applications and environments.
The trick here, for the module makers, is to walk this tight balancing act of having sufficient safety margins to adequately meet these unknown demands without pushing the price beyond what is acceptable to users. This brings us to the last point, the final test.
Testing philosophies and methodologies vary significantly from one semiconductor company to the next. But despite the differences, they all agree on one thing: make sure that the products come out of the factory meeting all the datasheet specifications with some safety margin.
In the case of the MCM class, several tests should be performed beyond simple datasheet compliance. For instance, testing of random samples in an actual application board to assure proper performance is mandatory.
One also must mention performance characterization using corner samples. This is done by sorting out devices that have the individual components’ parameters at one end or the other of the statistical distribution of the particular parameters. These samples are to be collected for all the major parameters and fully tested to verify that the MCM modules fully meet the specs.
Integrated modules will continue to grow in numbers and complexity, placing the responsibility of detailed subsystem design in the hands of the module maker. This trend will allow the design engineer to shift his/her attention away from the dc-dc converter and concentrate on solving the design problems of the smart devices that use these modules.
