Digital Control of Power Conversion and Management

This article will discuss extensively the technical aspects of applying digital technology to power conversion and management to meet market trends and requirements of different market segments. We will also touch upon the applications and challenges of this technology compared to analog control .

Power conversion belongs to the operation of the power system (feedback loop), while power management covers the operation modes, synchronization tracking and margining of start/stop delays, and phase locking (interleaving) functions for parallel operation and system communication.

Over the years, the definitions of "analog" and "digital" have become a little blurred. To avoid confusion, in this article, "analog" refers to "continuously variable physical quantities" and "digital" refers to "discrete variables." It is because of the characteristics of digital technology that we can store data, perform calculations, and communicate efficiently.

Until now, the actual processing of voltage and current in power conversion has always been in the realm of analog rather than digital, but control can be either analog or digital. Since the control is not completely digital, an analog-to-digital converter is required in the feedback loop. So what benefits can digital technology bring?

technology

Digital technology is ubiquitous in our daily lives, but it was only about four years ago that this technology began to be fully applied in the field of power conversion and management.

Features

The most prominent advantage of digital technology lies in the application of memory, which includes three basic levels of access: factory-level access (not accessible to users) specifically for registers containing internal calibration data and lookup tables of the controller; controller configuration (accessible by users through passwords) used to select management topology and control mode (voltage, current and hybrid) and provide different fault protection function settings; and monitoring and control (accessible freely through the PMBus protocol). The perfect data storage capability allows designers to optimize the design and even reuse it on different projects.

Digital technology's strength in storage leads to another advantage - communication. Communication via I2C gives the controller the ability to calibrate and program, and can also perform different functions in real time, including control, monitoring, status monitoring, remote identification and diagnosis. Other features include different intensity adjustments during transmission, including resolution (digital quantity), calibration (analog-to-digital conversion, external sensors), output voltage/current settings and protection limits (voltage, current, temperature).

Timing functions include frequency conversion, delay and phase adjustment. Control/management functions include operation mode conversion (start/shutdown, pulse train, pulse skipping, pulse frequency modulation, pulse width modulation, phase number), self-test and output voltage conversion (margin). Timing and control flexibility both help reduce electromagnetic interference, a feature that is completely unattainable in analog applications.

The magnitude of the benefits of digital technology may vary depending on the market segment, but some logical elements are required in the implementation process.

Types of logic

Flexibility, speed/bandwidth and cost constraints are decisive factors when choosing the right logic for an application. These include digital hardwired logic (state machine) PID control and digital pulse width modulation (PWM); digital hardwired logic PID control and digital PWM + non-volatile memory; hybrid = analog PWM + digital interface (commonly called "digital wrapper"); microcontroller (mC); digital signal processing (DSP) and digital control processing (DCP), including the best combination of DSP and mC.

Most digital ICs include power conversion control and power management functions. The power conversion control (feedback loop) can work in continuous real-time analog and near real-time (some response time required) digital states. Other functions are triggered, programmed and dormant (memory) based on events. Hard-wired logic can be used for high-volume low-power applications (<200W). They generally operate at higher frequencies (200kHz~2MHz) and close to analog control speeds. The cost is also relatively low. This can reflect the most robust structure, and does not require customers to make a large degree of programming, or even completely save (pin programming, or implement a graphical user interface through I2C); and it can speed up the time to market. Adding NVM can reflect higher flexibility in IC design, but it will increase the inspection and verification work.

If you take an analog controller and combine it with a digital interface that supports I2C communication and sometimes VID control, you have hybrid logic. It shares the same space as the hardwired structure, but is less flexible and more expensive. Both are used primarily in the DC-DC space. mC, DSP, and DCP all use coding (assembly language or C) to achieve greater flexibility and speed, but at a higher cost and longer time to market. However, the increased flexibility also makes the circuit structure more complex, so the cost of the verification and validation procedures will also be higher.

Silicon Process

The technology choices made when implementing these logic types are generally driven by cost; and as lower submicron technology (0.15/0.18mm) becomes more affordable, digital technology is undoubtedly more dominant, thus accelerating the transition from analog to digital technology. At a certain stage, "more functions at lower cost" will become the new value, replacing "more functions at the same cost". In the range of 0.25mm, the chip cost of analog and digital structures has long been the same, but as the chip cost in the range of 0.18mm has dropped, the R&D cost has more than doubled. (See Figure 1, source ISSCC 2007 / SESSION 1 / PLENARY / 1.1)

At TSMC, the production volume of 0.15/0.18mm products is much higher than that of 0.25mm products (Figure 1, source ISSCC 2007/SESSION 1/PLENARY/1.1). It is expected that economies of scale will further promote the relevant conversion process.

Segmentation and packaging

“Partitioning” refers to combining the functionality of the power conversion section in one or more integrated circuits . The choice should be based on the hierarchy of the control loop (protection, current, voltage, thermal), the power level, and the trade-offs between efficiency and space. When discussing DC-DC technology, we encounter terms like “discrete solutions”, “integrated controller and driver ”, “integrated power stage”, and IR’s iPOWIR functional module is a good example.

Discrete solutions are discrete controller ICs with 1 to 6 phases, where the drivers and FETs are separate (see Figure 3). This solution is open to alternative sources, offers the greatest flexibility and performance, and is lower cost, but takes up the most board space. Module manufacturers can offer more space-saving discrete solution modules, but at the expense of other sources. However, pin-to-pin compatibility can be retained.

Integrated controller and driver means that 1 to 3 phase outputs and related drivers are set in the same integrated circuit (Figure 4). It needs to be matched with external field effect transistors, has no alternative supply source, and has limited driving capabilities, but the advantage is that it saves less space than discrete solutions. In addition, due to the limited heat dissipation function of the integrated circuit, only 3 drivers can be set, and the pull/sink current can only be maintained at their maximum level.

An integrated power stage is a single phase driver and FET in the same integrated circuit. It takes up less space than discrete solutions, so it is suitable for space-constrained applications above 500kHz. The driver is closer to the FET, so it can operate efficiently at high frequencies. With a suitable controller, it can be used for one or more phases, but there is generally no alternative source of supply. The integrated circuits are usually optimized for a specific output current range and duty cycle, so the cost of the entire solution is often higher than a discrete solution.

This device is generally a fully integrated multi-chip module or MCM (controller, driver and field effect transistor are all installed in the same integrated circuit). Although it takes up the least board space, it is only suitable for relatively low power levels due to the thermal energy limitations of the package, and there is generally no alternative supply source. The choice of package depends on the chip area, pin count and heat dissipation requirements. In most cases, people will overlook the fact that the integration of control functions and power stages usually results in the failure to fully utilize the power stage, resulting in operation only at the maximum available temperature of the controller, or the reliability of the parts is compromised due to the need to force the controller to operate at the maximum power stage temperature.

A package with lower top-to-bottom thermal resistance is more suitable for all structures, but the part price is higher. When the FET operates at a lower temperature, the on-state resistance will be lower, and only a smaller FET will be needed to complete the work, or the efficiency can be improved, saving energy , which is better than the high cost of using a better package and adding additional heat dissipation design. Another advantage is that reliability will be greatly improved.

Because of the additional functions such as communication and pin function assignment, the system solution including digital controller requires fewer components and controller parts. Whether in terms of functional combination or packaging selection, digital structure is superior to analog structure.

Market Distribution

In terms of market distribution, the consumer market accounts for more than 50%, while the information technology industry has dropped to less than 45%, and the remaining 5% belongs to government users. A large number of business opportunities have emerged in the consumer market. Although high-end systems have begun to adopt digital technology (see Figure 1), the low production volume and profit margin make it difficult for the market to maintain. In fact, the consumer market that can have a relatively high output is still dominated by analog products, whether in the field of power conversion or management. To promote digital technology, it is necessary to take a two-pronged approach in both the high-end and low-end markets: the high-end market focuses on perfect functions and excellent performance, but with low cost as the goal; the low-end market focuses on bare metal dominated by analog technology, but focuses on lower costs to absorb market share. Ironically, the largest share belongs to power and not controllers. Therefore, the market expects that if the two can be combined, they can complement each other and increase the marginal profit and turnover of products. The market also expects that a large number of analog components will be replaced by smaller configurable digital components. Achieving a more sophisticated structure through "renovation" (addition/reduction) will bring greater advantages.

application

There are several different levels of high-volume power converters on the market, from lowest to highest cost per watt: DC-DC non-isolated 3~15W (cell phones, PDAs and handheld devices), DC-DC non-isolated 15~250W (point-of-load and VRD/VRM) embedded devices and modules, isolated AC-DC up to 2000W, and isolated DC-DC board-mount power modules (full regulation, semi-regulation and DC-DC transformers).

Low-power handheld devices mainly use analog power control and digital power management.

Digital power control and management technology was first used in high-end computers and image processing (VRM/VRD), followed by network communications (datacom and telecom) and storage systems. Point of load (PoL), embedded devices or modules and isolated DC-DC modules are the preferred solutions in network communications, while storage systems use both VRDs/VRM and PoL. These applications all use digital hard-wired or hybrid logic structures. Different types of AC-DC systems use μC, DSP or DCP control and power management technology, and there are more and more isolated DC-AC inverters and isolated boost DC-DC converters to follow suit.

Regardless of the application, improvements in operating modes help improve efficiency in low-load environments. This includes phase reduction, pulse skipping, lower gate drive voltage, and lower switching frequency. As for full-load environments, the improvement in operating efficiency comes from lower on-state resistance (better FETs) and revolutionary topology technology.

To add value to digital structures, generate greater profits and increase productivity, companies must go beyond Moore's Law.

challenge

External challenges include changes in market inertia, including engineers and hardware. The hardware challenge is based on the large number of users but the lack of digital interfaces. In addition, market segmentation makes it difficult for marketers to define products and requires everyone to adopt new technologies; and the market's resistance to new technologies (such as high prices and proprietary communication buses based on the "Z series" ) or delayed adoption due to fear of IP litigation all pose challenges. At the same time, inconvenient graphical user interfaces and incomplete operating guide documents are likely to increase the operating training time for customers and distributors. On the other hand, competition in existing analog technologies is quite fierce. Manufacturers have cut prices and launched new products to compete for the market, and even created all kinds of fear, uncertainty, and doubt to attract users. In addition, small purchases and low profits often create obstacles for new users.

Internal challenges are no less numerous and serious than external challenges. Long learning times, design limitations, and resource requirements for verification and validation procedures slow time to market and even cause products to lose market share. The design verification procedures for configurable digital controllers are theoretically infinite, and insufficient verification and validation methods and tools exacerbate the problem. In addition, analog and digital engineering resources need to be coordinated and coordinated.

Finally, we should not ignore another major challenge, which is that corporate culture often affects the attitude and thinking of adopting new technologies. The idea that "our products, resources, time and market are not enough" should never be a limiting factor.