Digital power techniques threaten analog power-supply obsolescence

Patrick Le Fevre, Ericsson Power Modules

January 1, 2011

Patrick Le Fevre, Ericsson Power Modules

Following decades of development, top-specification dc/dc converters have arrived at a performance plateau that requires power-supply engineers to reconsider their design approaches. In effect, converters that use familiar analogue control schemes are now incapable of making anything but very small improvements in terms of the conversion efficiency and power density metrics that traditionally drive designers’ choices. As a result, digital power control techniques are rapidly gaining market share as designers increasingly appreciate the advantages that the technology offers over its analogue counterpart.

In reality, any digital power converter employs a combination of analogue and digital circuitry that mixed-signal silicon processes make possible. In this context, “digital power control” refers to implementing the inner control loop of a power converter with digital circuitry rather than using analogue schemes.

For the simplest example of a buck converter, this means substituting an analogue-to-digital converter for the conventional error-signal feedback amplifier and controlling the pulse-width modulator that drives the power switches using digital-signal-processing techniques in place of a voltage reference, ramp generator, and comparator, Figure 1:

 

Figure 1. A digital buck-converter substitutes digital-signal-processing techniques for the familiar analogue control loop.

Given a mixed-signal core, it’s hugely attractive to add on-chip “digital power management” functionality that consists of supervisory and control circuitry that can communicate with external logic to facilitate power management schemes. At negligible additional silicon cost, integrating such functionality slashes PCB real estate requirements while optimising the coupling between the power converter’s core and its measurement and control subsystem.

 The board-level I/O system of choice is the PMBus™ power-industry standard that’s based on SMBus hardware, which is also easy and cheap to integrate. Several silicon vendors now offer chips that embody all of the elements that are necessary to build digital power converters with minimal support in terms of external component count. Fewer components help shrink the digital converter’s footprint, improve reliability, and reduce cost—and the digital values that control the converter’s operation do not drift with time or temperature.

Fundamental requirements for a digital converter

A basic requirement for any digital power converter is that it must at least equal the efficiency of the best-available analogue solutions without sacrificing any aspect of electrical performance—including regulation accuracy, transient response, or output noise levels. Virtually all power converters are least efficient at light loads, with most analogue dc/dc converters beginning to move towards efficient operation at 15 - 20% of their output power capability.

Typically, such converters achieve maximum efficiency at around 50 – 70% of full load, which is the load area that their designers expect users to exploit. Until quite recently, this characteristic operating envelope has suited most systems with their relatively stable loads, but today’s systems are increasingly designed to power down as many functions as possible to save power whenever it is feasible to do so.

This situation requires far more of the power supply in terms of its ability to perform efficiently from very light loads upwards, which can have a knock-on or secondary effect in terms of the converter’s input voltage levels that may not be at all well regulated. As Figure 2 shows for a representative pair of quarter-brick intermediate-bus dc/dc converters, a well-designed digital converter can achieve similar or better efficiency than its analogue counterpart from around 10% of full load and then maintain a performance advantage right up to 100% while offering superior tolerance towards varying input voltages:

Figure 2. A well-designed digital dc/dc converter easily outperforms its best-available analogue counterpart. Graphs show typical efficiency versus load current and input voltage at 25°C, for 12 V/33 A supplies. Top graph is for analogue PKM 43048 PI supply; lower graph is for digital BMR453 supply.

A key technique that digital converters employ to extend and flatten the operational efficiency envelope is to vary the dead-time between the power switches conducting. For Figure 1’s buck-converter example, this “shoot-through” prevention measure ensures that neither MOSFET is on at the same time, which would almost certainly result in their mutual destruction.

For greatest efficiency there would be zero time between the devices switching, but most converters use a fixed period that guarantees safe operation across the line-and-load range. Varying this period to reflect the input line conditions can improve the converter’s efficiency by several percent at its operational extremes. While one silicon vendor has patented a technique that accomplishes this task within an analogue controller chip, a digital converter more easily and flexibly meets the requirement.

Any digital power converter must also offer competitive power density—which is relatively straightforward to ensure—and be as easy to apply as an analogue converter. The difference between design and application then becomes highly significant, as for most engineers the major downside of digital power conversion is the learning curve that the technology demands. While analogue converters rely upon resistors and capacitors to set the poles-and-zeros that balance the control loop’s dynamic response versus its stability—and sometimes to set the dead-time period—digital converters use sets of constants for these purposes.

Assuming appropriate control algorithms, it is the ability to swap between alternative values of PID (proportional-integral-derivative) constants in response to line and load conditions in real time that allows a digital converter to consistently outperform an analogue-based design. Despite the best efforts of silicon vendors to produce development environments that help simplify tuning a digital converter’s control loop, developing robust firmware remains a major undertaking.

Accordingly, many engineers prefer to specify pre-qualified digital power converter modules that make the technology transition seamless. In developing the first of its 3E family of digital power products, Ericsson proved that it’s possible to improve upon the high power of a conventional loosely-regulated intermediate bus converter by around 5% to squeeze 396 W from the quarter-brick format.

At the same time, this digital converter offers the tight ±2% voltage regulation of a fully-regulated analogue dc/dc converter that manages only 204 W from the same footprint. Conversion efficiency improves to better than 96% from around 10% of full output power upwards and while the digital converter integrates the PMBus interface, you can ignore its facilities and apply the device just as easily as any analogue component.

Similar benefits apply to varying degrees across a growing range of digital power products that are now appearing in the marketplace, with the result that even highly experienced power-supply designers increasingly accept modular solutions over design-it-yourself challenges.

Digital configurability delivers life-cycle benefits

But digital power has much more to offer than bettering the electrical performance and power-density requirements that previously dominated the mindset of power-supply designers. Integrating the power management hardware and PMBus interface alongside the converter’s core confers a raft of benefits that can apply throughout the lifetime of the end-user’s application.

Crucially, it’s now possible to configure the digital converter when it is initially made, during the development phase of the power-system designer’s application, at the distributor’s depot, when the equipment is manufactured, and/or when it is operating within the end-user’s equipment. This degree of flexibility extends the programmable logic model to the power conversion industry for the first time.

For instance, each member of Ericsson’s 3E family of digital power converters offers an array of programmable parameters that includes output voltage selection; turn-on/off delay times to implement power sequencing for multi-rail loads; slew rate control that provides inrush current protection; voltage margining for system testing; and multiple thresholds for warning and fault conditions for overcurrent, overtemperature, and under- and overvoltage. It’s even possible to adjust the response of a 3E digital converter’s control loop to optimise its performance for a particular set of load and bulk output capacitance conditions. Figure 3 shows the result of fine-tuning the constants that set the responses of a 3E point-of-load regulator’s control loop to optimise its transient response for a given environment:

Designers can program any of these parameters at any point during a 3E product’s lifecycle using the converter’s PMBus interface. The PMBus protocols that underpin such operations include a standard command set that is optionally extensible to accommodate custom operations. Crucially, PMBus mandates a “set-and-forget” mode that permits designers to program a compliant device once, after which operation the device retains its settings for life or until it is reprogrammed.

This capability opens a range of capabilities that span rationalising inventory holdings by replacing multiple fixed-voltage analogue converters with a single programmable digital counterpart that will run in stand-alone mode—that is, without any need for PMBus on the target board—to developing a fully PMBus-enabled system that can help ensure maximum system up-time by flagging any warnings and error conditions that may signify impending failure.

Moreover, a fully PMBus-enabled system might minimise its energy consumption by intelligently managing its power-rail voltages to optimise efficiency across the system’s load power range. For instance, reprogramming the intermediate-bus voltage that supplies multiple point-of-load converters from say 12 VDC to 9 VDC under light load conditions reduces the burden across those converters and minimises their losses. This dynamic bus-voltage adaptation technique is well known and particularly suits systems that spend significant periods under widely different load conditions, but it is challenging to implement using analogue converter technology.

PMBus™ underpins rapid prototyping

The key to realising the flexibility that digital power conversion offers is to make it as easy as possible for designers to access. Figure 4 shows the structure of a typical board within a conventional backplane-based system that includes a communications link to enable power-management opportunities.

Figure 4.  PMBus makes it easy to monitor and control compatible power-system devices such as the 3E family

In this scenario, on-board power-management logic that may optionally include local intelligence links the board’s PMBus-compatible devices with the system host, which—depending upon the sophistication of the end-user’s application—might be a PC running application software or an embedded controller with LAN/WAN links.

Because the physical layer of PMBus relies upon SMBus—which is electrically very similar to I2C—PMBus is generally limited to the board domain, leaving designers free to implement their choice of backplane connectivity. The board power control logic requirement is generally undemanding and can comprise a low-cost microcontroller or even some spare gates in an FPGA:

For development purposes, the system host and user interface might comprise a PC that runs dedicated application development software and connects to the prototype board via a USB-to-PMBus adapter. This approach provides an extremely fast method for experimenting with parameters such as output voltage settings, power sequencing routines, voltage margining, and fault handling without any need for hardware changes on the board-under-test. When the designer is satisfied with a set-up, the application software might save a configuration file for each programmable power converter. Figure 5 shows some of the options that the 3E evaluation kit’s software presents within a device configuration screen:

Figure 5.  The 3E evaluation kit’s graphical user interface software greatly simplifies device configuration.

Planning for the future

It’s very often the reality that a system’s power requirements will evolve over the lifetime of a particular family of equipment. Designers who wish to incorporate some degree of future-proofing are able to do so by taking advantage of power converters that offer scalable footprints.

It’s also often desirable to have a choice between through-hole-plate and surface-mount assemblies. As well as addressing these needs, the 3E family’s mechanical design eases the issues that comparatively large components can create with pick-and-place equipment. The family currently comprises eighth- and quarter-brick intermediate-bus converters with isolated outputs and power ratings of up to 240 W and 396 W, together with 20 A and 40 A point-of-load converters. In the near future, more devices will augment the point-of-load range to address power levels down to 12 A.

About the author

Patrick Le Fevre is Director of Marketing and Communication at Ericsson Power Modules division. With more than 25 years in the business, he's actively involved in association affairs, and leads a number of standardization forums for the power community.