Advertisement

CHRIS YOUNGDigital power helps reduce risk in new designs by allowing for rapid changes as new information is encountered concerning the load behavior, noise environment, fault conditions, or simply last minute changes.  

A few years ago, power supply engineers had to re-spin the board, change the bill of materials, add filtering, or change controllers. With digital power, users can facilitate changes without having to re-spin the board, without additional component additions or changes and do it within minutes. Thus, digital power reduces both the schedule risk and the performance risk in new designs.

For example, many power controllers now are able to retain data in non-volatile memory. This capability can be used to save data about historical operating conditions even after power has been lost to the controller, so users can get details about the forensics of system shutdowns much faster and more accurately, leading to better designs on the front end and reduced warranty costs on the back end.

While power supply designers like the flexibility, they also want devices to be easy to use.  Digital power provides a rich feature set that for many can make digital power look complex. The challenge for digital controller suppliers is to provide the same level of flexibility but make the parts easier to use. There are two trends at work here simultaneously - an even better set of features along with a simplification of the design process for digital power.

New generations of power controllers also are becoming more “self aware.”  Historically, the power supply designer had to know a great detail about the load behavior in order to design a power supply. Many times, the level of detail needed was lacking, because it did not include a basic of how the load would behave. However, with digital power, any power supply is capable of characterizing the load behavior and adapting its own configuration to the load. This decreases the workload of the design engineer and leads to a more robust power conversion process.

Perhaps the most recent and impressive potential of digital power is that compensation has become easier. The design process has gone from selecting a half dozen to a dozen components in the traditional analog process to selection of a few coefficients in the modern digital controllers. And while designers in the analog process typically have used spreadsheets of one form or another, digital power designers have enjoyed the benefits of GUI based tools. These kinds of tools automatically optimize, given the circuit parameters.

Up until now, the challenge for both camps has been compensating in situations where the circuit parameters are not known, may be different from nominal component values, or are not constant. In these cases, traditional compensation techniques lead to a lack of or, at best, unsatisfactory results. The result is a less than robust power supply that may not regulate properly with transient loads or may not regulate properly when a particular variation in manufacturing occurs or may not regulate adequately in the presence of ambient temperature variations.

Figure 1 shows the schematic for a buck converter with a high Q filter circuit. The Q of this circuit is around 11. Although the bode plot shows this circuit to be nicely compensated (78 degrees of phase margin, 19 dB of gain margin), Figure 2 shows how the circuit can go unstable with only a 10 percent variation in the inductance and capacitance using the same compensation. Such a variation in this example results in a negative 18 degrees of phase margin and less than five dB of gain margin. This variation is not unreasonable in a production environment, given the tolerances of typical inductors and capacitors.

Figure 1. Schematic and Bode Plot for a high Q design. The white line is the magnitude of the gain, and the red line is the phase of the gain.

In practice, designers have been able to de-tune the compensation so that it is less sensitive but at the same time it is not optimal over the entire operating environment. The variations illustrated above can arise not only because of part-to-part tolerance variations but also due to thermal drift, aging, and, in some cases, parasitic elements.

Figure 2. Schematic and Bode Plot with a 10 percent increase in inductance and capacitance.

So, the challenge is how best to compensate a power supply if the circuit parameters are not known to sufficient accuracy. 

The answer: digital power controllers where the compensation is automatically determined by the controller itself. In short, the controller characterizes the circuit parameters and then determines the best compensation for those parameters. This characterization can be done on command, or it can be “free running” so that as parts age, as the temperature changes, or as parasitics come and go, the circuit is optimally compensated.

Today’s embedded power systems are typically designed for optimal efficiency at maximum load, reducing the peak thermal stress by limiting the total thermal dissipation inside the system. Unfortunately, many of these systems are often operated at load levels far below the peak where the power system has been optimized, resulting in reduced efficiency. While this may not cause thermal stress to occur, it does contribute to higher electricity usage and results in higher overall system operating costs.

To restrain these costs, a DC/DC controller should be able to enable the power converter to automatically change the operating state, which will increase efficiency and overall performance with very little or no user interaction. Auto compensation is the solution, because it eliminates the need for manual compensation of the PID filter.

In practice, auto compensation is “hands-free” because, by enabling the power supply, the controller does all of the work. Figure 3 shows the transient response of a typical power supply before (the upper section) the supply is adequately compensated.  The second trace (lower) shows the transient response after auto compensation.

Automatic compensation works by adjusting the compensation coefficients in a systematic way while observing the response of the system. While this does produce a slight perturbation on the output it is almost imperceptible and well within the allowed transient envelope. One of the benefits of this method is that it has a limited and controlled perturbation on the output unlike other methods, which may have a non-deterministic or non-controlled perturbation on the output.

The bottom line is that this automatic compensation eliminates a substantial burden from the power supply design and results in a more robust power supply for the life of the power supply. Auto compensation eliminates the need for manual compensation design work. Adaptive performance optimization algorithms improve power conversion efficiency. 

One of the newest ICs providing auto compensation and all of the associated benefits of digital power is the ISL6105 from Intersil-Zilker Labs. For more details about the features of the ISL6105, visit http://www.intersil.com/zilkerlabs/

Advertisement
Advertisement