Digital control of power conversion and management

To date, power conversion—the actual processing of voltage and current—has been analog rather than digital, but the control has been analog or digital, although not purely digital, as an A/D is still needed in the feedback loop. So what does digital have to offer?

Digital technology has made its mark on many aspects of our lives. But it is only in the past four years or so that the digital age was introduced into the control of power conversion and management.

FEATURES
Digital technology’s most salient feature is memory, with three basic levels of access: factory only (not accessible to the user) for registers containing controller internal calibration data and lookup tables; controller configuration (accessible to the user with password) where the governing topology and the control mode (voltage, current, hybrid) are selected, along with various protection/faults settings; and lastly monitoring and control (free access via PMBus protocol). The ability to store data allows the designer to optimize the design and reuse it from project to project.

Next to memory—and enabled by it—is the key feature of communication. Communications, via I2C, allows calibration and programming of the controllers and real time access for control, monitoring, status and remote identification and diagnostics. Additional features encompass magnitude adjustments “on-the-fly” that include resolution (more or less digits), calibration (A/D, external sensors), output voltage/current setting and protection limits (voltage, current, temperature).

Time adjustments include frequency change, delays and phase adjustments. Among the control/management features are operating mode change (start up/shut down, burst, pulse skipping, PFM, PWM, number of phases), self test, and output voltage change (margining). Time adjustments and control flexibility can facilitate (undreamed of in analog implementations) in EMI mitigation.

The list of features, whether shorter or longer according to the granularity necessary to address the market, requires a certain type of logic for implementation.

LOGIC TYPE
The “logic type” selected for a certain application is dictated by its flexibility, speed/bandwidth and cost tolerance, and includes digital “hard wired” logic (state machine) PID control and digital PWM, digital “hard wired” logic PID control and digital PWM + Non Volatile Memory (NVM), hybrid (analog PWM + digital interface, known as “digital wrapper” ), microcontroller (µC), digital signal processing (DSP), and digital control processing (DCP), as a merging of the best in DSP and µC.

Most digital ICs include power conversion control and power management. The power conversion control (feedback loop) operates continuously, real time for the analog, and almost real time (with some latency) for the digital. The rest of the functions are event triggered, scheduled or dormant (memory). Hardwired logic is used in high volume low power applications (<200W), which tend to operate at relatively high frequency (200kHz to 2MHz) with speed close to analog implementations and low cost. This flavor is the most robust and needs no or little programming by the customer (pin programming, or GUI via I2C). It also allows the shortest time to market. Adding NVM provides more flexibility to the IC design but increases the verification and validation task.

Hybrid logic occurs when using an analog controller with a digital interface for I2C communication and sometimes VID control. It shares the same space with the hardwired implementation, with less flexibility and slightly higher cost. Both address mainly the DC/DC space. The µCs, DSPs and DCPs use code (Assembly or C) and offer increasing flexibility and speed at increased cost and time to market. Increased flexibility leads to higher complexity and more costly verification and validation.

SILICON TECHNOLOGY NODE
The technology node selected for the implementation of these logic types is cost driven and as lower submicron (0.15/0.18µm) becomes more affordable the balance will irreversibly tilt on the digital implementation side and speed up the transition from analog to digital. At that point “lower cost and extra features” will be the new value proposition to replace the present “same cost and extra features.” Already at 0.25µm, the chip cost is the same for analog or digital implementation, but while the chip cost drops (at 0.18µm), the R&D cost more than doubles (Figure 1; Source ISSCC 2007/SESSION 1/PLENARY/1.1).

Production volumes of 0.15/0.18µm at TSMC are dramatically higher than those of 0.25µm (Figure 2; Source ISSCC 2007/SESSION 1/PLENARY/1.1) and it is expected that the economy of scale will further stimulate the transition.

PARTITIONING AND PACKAGING
Partitioning refers to the grouping of functions in the power conversion train in one or several ICs. The choice should be based on the control loop hierarchy (protection, current, voltage, thermal), power level, and efficiency/space trade off. In the DC/DC arena, for example, we encounter terms like: “discrete solution”, “integrated controller and drivers”, and “integrated power stage” such as International Rectifier’s iPOWIR building blocks.

Discrete solutions refer to a separate controller IC with one to six phases, separate drivers and separate FETs (Figure 3; Source [3]). This solution is open to second sources, offers maximum flexibility and performance, costs less, but takes the most board space. Module manufacturers offer space saving discrete solution modules, but the second source benefit is often void, though pin-to-pin compatibility may exist.

The integrated controller and driver refers to one to three output phases and associated drivers in one IC (Figure 4; Source [3]). It needs external FETs, has no second source and has limited driving capability, but takes less space than the discrete. The limited heat dissipation capability of the IC limits the number of drivers to three and their maximum source/sink currents.

Integrated power stages refer to drivers and FETs for one phase in one IC. These take less space than the discrete solution and are used in space limited application, at >500kHz. The proximity of the drivers to the FETs allows efficient operation at high frequencies. It can be used for one or several phases with a suitable controller. A second source is typically not available.

The IC is typically optimized for a given output current range and duty cycle and offers a higher solution cost than the discrete.

Fully integrated is typically a multi-chip-module or MCM, where the controller, drivers and FETs are in one IC, take the least board space but are confined to lower power levels due to the thermal limitation of the package, and typically don’t have second sourcing. Packaging selection is dictated by the die area, the number of pins and the thermal dissipation need. Often it is overlooked that integrating control and power stages leads most of the time to underutilizing the power stage, forcing it to operate at the controller maximum allowable temperature or, reducing the part reliability by forcing the controller to operate at fully utilized power stage temperature. A low top and bottom thermal resistance package is desirable for all implementations, even at a part price premium. Running the FETs at lower temperature means lower Rds(on), with either a smaller FET being able to do the job or, improving the efficiency and saving energy that would far exceed the premium paid for better package and even extra heat sink. Improved reliability comes as a bonus.

Because of extra features, including communication and pin function assignment versatility, the system solution implementation with digital controller requires fewer components and fewer controller part numbers. So even here, in the grouping of functions and the choice of package, the digital implementation is superior to analog.

MARKET MIX
On the market mix side, the consumer market share passed the 50 percent mark, while IT dropped below 45 percent, with the balance of 5 percent going to the government. The high volume opportunities are to be found in the consumer market. While the digital adoption started in high end systems, the volumes and profit margins are too low to survive. The higher volumes promised by the consumer market are still dominated by analog implementations at both power conversion and management. The digital push has to be pursued at both high-end, for features rich/top performance, but low cost, and at the low end for bare bones like analog, but lower cost to gain market share.

Ironically, the lion’s share in the power conversion belongs to the power stage, not the controller. So it is expected that integrating the two would boost both, the product’s profit margin and the revenue. It is also expected that the multitude of analog parts will be substituted by significantly fewer digital configurable parts. Fine granularity, by features “trimming” (inclusion/elimination), becomes a significant benefit.

APPLICATIONS
There are a few classes of high volume power converters, listed here in the order of increasing cost ($/W): DC/DC non-Isolated 3W-15W (cell phones, PDA, portables); DC/DC non-Isolated 15W-250W (POLs and VRD/VRMs) embedded and modules; and isolated AC/DC up to 2kW and isolated DC/DC board mounted power modules (regulated, semi-regulated and DC-DC transformers). Low power portables use mainly analog power control and digital power management.

The early adopters of digital power control and management were in the high end computer and graphics segment (VRM/VRD), followed by netcom (datacom, telecom) and storage. The POLs, embedded or modules and isolated DC/DC modules are the solution of choice in netcom, while the storage employs both VRDs/VRMs and POLs. These applications use digital “hard wired”, or hybrid type logic implementations. The wide range of AC/DC are using µC, DSP or DCP control and power management, as do the increased number of isolated DC/AC inverters and isolate boost DC/DC converters.

Regardless of the application, efficiency improvements will come at low loads from improvements in the operating mode (phase dropping, pulse skipping, lower gate drive voltage, lower switching frequency) while at full load mainly from lower Rds(on) (better FETs), and revolutionary topologies.

For the digital implementation to create more value and command higher profit margins and/or volumes, the companies have to go beyond Moore’s law.

CHALLENGES
External challenges include the changing market inertia and incorporate both the attitudes of engineers and the hardware. Hardware refers to the large installed base and the lack of digital interface. Market fragmentation makes it hard for marketers to create a product definition and demands that new technology be all things to all people. Challenges also include market rejection (e.g. high price and proprietary communication bus of “Z family”), or delayed adoption for fear of IP litigations. User unfriendly GUIs and poor documentation meanwhile can lead to longer learning curves for customers and distribution channels. In addition, competition from incumbent analog technology is fierce, fighting back by offering lowering prices, introducing new products and spreading FUD, while there is a high barrier of entry for new comers and low sales volume and low profit margins prospects.

Internal challenges are as numerous and critical as the external ones. Learning curve and limited design, verification and validation resources leading to delayed time-to-market and loss of market share, are just to name a few. Likewise, design verification testing for configurable digital controllers is theoretically infinite, and the lack of verification and validation methods and tools exacerbate the problem. There is also a need to harmonize analog with digital engineering resources.

And finally, we should not omit attitude and imagination in the corporate culture as major challenges. “We don’t have enough: products, resources, time, market, etc.,” should not be limiting factors.
References
[1]Ed Herbert, “Converter Achieves High Efficiency through Digital Control” http://powerelectronics.com/digital_power/news/edward-herbert-patent-power-converter/

[2]Lou Petchi “Are There Too Many Power Management ICs?” http://powerelectronics.com/power_management/news/power-management-ics-0212/

[3] Patrick Le Fèvre, “From Digital Confusion to Digital Conversion”, Power Electronics Europe, Issue 7 2006] http://www.ericsson.com/solutions/operators/news/2007/q1/20070131_digital.shtml

[4] S.Lupan, “Digital Power Management: Changing the Value Ecosystem”
http://www.powermanagementdesignline.com/howto/191600998

[5] Lewis Counts, “Analog and Mixed-Signal Innovation: The Process-Circuit-System-Application Interaction.” ISSCC 2007/SESSION 1/PLENARY/1.2

Click here for the illustrations:

Figure 1, Figure 2, Figure 3, Figure 4

Caption
Figure 1: Total R&D costs versus technology node, relative to costs at 0.25µm.

Figure 2: Actual and projected wafer volume as a function of time following introduction for foundry technologies at TSMC.

Figure 3: Discrete solution.

Figure 4: Integrated controller and driver.