This means that designers are being faced with the challenge of implementing digital power designs which can, initially, appear to be both complex and expensive. The reality, however, is that digital power can be simple to implement and can cost less than analogue solutions when considering the total cost of ownership.
In both analogue and digital power-supply architectures, the role of the power controller is to sense output load and drive a correction signal.
Some products use a core analogue controller and wrap digital controls around the edges, whilst others implement the design in an almost completely digital scheme. Both of these approaches share the following basic parameters:
• Set point: Output voltage target.
• Ripple: Switching noise imposed upon the output voltage, typically 5 - 50mV. It is filtered with ceramic capacitors, typically in the 10 - 100µF range.
• Bandwidth: Dynamic linear-mode response to error signals. There is never enough.
• Phase margin: A measure of circuit stability. Ideally 45 degrees or better.
• Transient response: The depth of a drop or rise in Vout due to a change in load.
• Load regulation: The amount of droop in output voltage per unit of load current.
• Line regulation: The amount of change in output voltage per unit change of input voltage.
• Switching frequency: How often the input voltage is switched on and off to the output filter. Faster switching gives lower ripple, higher bandwidth, with slightly lower efficiency. Some applications are sensitive to switching noise frequency and want a very high switching rate.
To support power system designers, power-supply manufacturers have developed elaborate tools to assist in the design of power systems. However, the manufacturer cannot account for variables that are unique to each solution, which means that the amount of engineering time can vary, depending on the solution.
For example, a discrete design using board-mounted components can take several hundred hours of costly development time.
While a design using digital point-of-load (POL) converters can take place in a few hours, as the POL can read the load and adjust the compensation automatically and therefore reduce development cost, significantly.
Some applications require sophisticated supply-voltage sequencing. An asic might require core VDD to always be brought up before I/O bus VDD. In other cases, supplies brought up in the wrong order can result in a SCR latch-up.
Then there are issues surrounding fault response: one bus can go into a fault, without all busses needing to be shut-down. These types of application problems require designers to either implement power management in their own logic devices or use a digital power manager IC.
The following examples show applications in which digital power can provide benefits.
Company A has a very short time to market and needs a reliable power system to be up and running within four weeks. In addition, the company has a very limited staff with no power-design engineer.
Its system is based on boards designed around an array of independent system-on-chip (SoC) parts, each powered off a single high-current, low-voltage source.
The asics involved all operate independently and require supervisory functions. Failure of any node does not affect the operation of any other node but affects overall performance.
A digital power architecture, using POL modules enables the faster development using proven modules, whilst a digital power manager is used to enable easy configuration.
Company B’s application is designed with several signal-array switches. Each switch has a set of asics and interface devices with multiple voltage rails and complex power-up and power-down sequences.
Each switch functions independently from each other. Using POL modules, grouped by switch, provides an independent supervisory system which can be used to manage the entire application. Each switch can use a digital power manager to connect to the supervisory system via an I2C bus.
In both of these applications, digital power provides key benefits. First, the design cycle was reduced significantly since implementing a digital architecture is more straightforward than designing a new power system from discrete components, as digital modules can be re-configured via software. Second, the monitoring capabilities of the digital power solution are able to alert the system of potential problems and take action before catastrophic failure.
Third, the digital power system can isolate faults and prevent them from spreading through the entire system.
Another benefit of a digital power architecture is the flexibility it provides in meeting power needs. When developing a product, estimates of power requirements can vary wildly in parameters such as accuracy, peak current demands and voltage levels.
It is common to find that, at the prototype stage, the actual power requirements vary from the initial calculations.
This can lead to significant delays in a project if the power system is not flexible, since PCB revisions take greater amounts of time and money than software updates. The digital power architecture is well suited to coping with such unexpected power requirement complexities.
Digital power can seem to be a complex proposition with factors such as z-transforms and zero-pole plots making it hard to understand and implement. However the design tools offered by power-supply manufacturers, such as Power-One, are intended to simplify power-system design and provide greater flexibility in dealing with real-world applications.
Furthermore, the built-in measurement and command capabilities provided by digital architectures open up possibilities for monitoring system health, selective power control, built-in fault injection for testing and adjusting for optimal performance, as well as reducing power consumption by shutting down unneeded rails on the fly.
Writers: Chris Leek is a power specialist with Charcroft and Charles Hotchkiss works for Power-One.