Improve board power system reliability

最近几年,典型设备板上的电压显著地降低,很多情况下降到1V或更低,而总的板功率继续急增。增加的不同轨电压也使电源系统中轨之间定序和跟踪变复杂。同时,为了降低设备发生故障时间,对可靠性和可用性的要求不断增加。

There are several ways to meet the increased design requirements of power systems without sacrificing reliability. High-reliability power converters are a key part of these solutions, but they need to be supported by a well-chosen overall device architecture. Attention must also be paid to power system integration.

Figure 1. A typical 48V board power system where a single isolated DC-DC converter (brick) generates a 5V intermediate bus voltage that feeds a large number of non-isolated POL power converters.

On-board power system

Products no longer rely on simple 5V power distribution systems. Today, it is not uncommon to have 6 or more voltages on a single board. Some high-end systems have up to 20 or more separate power rails, most of which are less than 2V. These very low voltages must be efficiently provided at high currents and must meet increasingly stringent regulation, ripple and transient performance specifications. Therefore, distributed power systems now typically use multiple DC-DC converters on each board to generate low voltages close to the load.

In addition to requiring very low voltage rails, many ICs also require sequencing and tracking between power rails during startup and shutdown. The power rails must be controlled so that the difference between them does not exceed specified voltage and time limits, even under short transient conditions. Combine these requirements with the need to monitor all rails for overvoltage (OV) and undervoltage (UV) protection, and it is clear that the board power system is beyond the scope of simple construction.

Power system implementation

Figure 1 shows an example of a board power system. In this example, a typical product, such as a communications system or high-end computing server, is powered by 48Vdc. DC-DC converters provide the required voltage rails for the board and provide the required isolation between the 48V input and the logic outputs. In this example, a single isolated DC-DC converter (often called a brick converter) generates a 5V intermediate bus voltage that is supplied to a large number of non-isolated point-of-load (POL) power converters.

Many manufacturers offer brick and POL converters as standard products in many output voltage and current combinations. These standard products can easily function as building blocks in a board power system. As can be seen from the figure, a single brick, two or more brick converters are often used to generate the rails required for the highest power supply, while POLs are required for the lower power rails. Multiple combinations of power converters can meet the specific requirements of any particular board.

To coordinate the work of the DC-DC converter, the board power system requires an overall management function. The isolated primary and secondary sides also need to be managed. The power management function usually includes some or all of the following requirements:

Turn the power system on and off at the specified input voltage;

Control the opening and closing of all outputs at the required timing;

Monitor all outputs for OV and UV faults;

Control shut down when a fault occurs;

If necessary, adjust the output voltage;

Inform the system controller of the power status;

Power reliability

Power reliability can be understood from two different perspectives.

Component level: A bottom-up component level approach is used. Its reliability is usually expressed in terms of mean time to failure (MTBF) or failure in time (FIT). Since 1 FIT is 1 failure in 109 device hours, 1000 FIT is 1 million hours of MTBF. The two most common determination methods are MIL-HDBK217 and Telcordia TR-332. This type of determination only considers component failures, without considering design errors or inappropriate performance indicators.

System level: The system must be designed based on its ability to perform the required functions, using a top-down system approach. This approach allows for worst-case design, simulation, and testing of the complete system. Testing must be sufficient to ensure that the design meets all required functions under all operating conditions, a process called constraints. Good design strategies must be followed. Testing cannot alone guarantee correct performance under all conditions. It is most important to consider the above two aspects in the design.

Improve system reliability

Most power reliability issues are caused by system-level reliability (component application and system constraints) rather than the basic MTBF of the component. This includes:

The peak current drawn by the on-board product is higher than desired, resulting in voltage drop under extreme conditions.

In the field, noise releases cause power systems to shut down unexpectedly.

At the user's site, the board failed, but when repairs were restored, there was no failure found (NFF)

Rail-to-rail sequencing relies on component tolerances that do not always meet IC requirements.

Sequencing during shutdown was not considered during design.

Under extreme input voltage and temperature conditions, the power system cannot provide full load.

When the board was installed in the device, the power module overheated due to restricted airflow.

Table 1 lists power system problems and solutions.

Clearly, good power system design is a complex, multifaceted subject that involves the entire product and its environment. The complexity of the task cannot be underestimated. Furthermore, although the initial focus is on efficient power conversion, remember that power management functions are equally important in achieving good power system performance.

Improved MTBF

The following three basic methods can improve the MTBF of any system: use fewer components, make components more reliable, and produce system functionality even if components fail. Each method, together with comprehensive limited condition testing, plays a role in improving the reliability of power systems.

Fewer components

The number of components in a power management system can often be reduced.

A dedicated power management IC can replace a large number of discrete components (comparators, op amps, optocouplers, RC time delays) used for monitoring and control. At the same time, the performance of the power management IC is much better than that of discrete solutions, improving system reliability by accurately informing margin performance and avoiding noise release.

A typical POL contains fewer internal components than an isolated brick converter, and the failure rate is much lower. For a typical POL, the manufacturer's failure rate is about 5 million hours MTBF, while the typical brick converter is 2 million hours MTBF. In addition, POL output power is usually lower than brick converters, so more POLs can be used to meet the total power requirement. Of course, reliability is only one of many factors in selecting a power converter. Considering reliability early in the design may make the application have the best compromise considerations.

More reliable components

Component reliability is primarily affected by the conditions and quality control processes in manufacturing and the emphasis in the application. Using a modular approach, power management designs using gate arrays or microcontrollers require extensive testing under rated operating and failure conditions. This is to ensure that the logic in the programming does not cause incorrect behavior. Obviously, the performance of dedicated power management devices has been fully tested and qualified by the manufacturer.

Failure tolerance

为了显著地改善系统可靠性,所设计的系统是具有失效容限。在理想情况下,一个有效的备用元件在任何元件失效时,能立即取代,使系统性能不受影响。在实际系统中,对所达到的备份度是有限制的,而利用率不可能达到100%。通过仔细地设计,备份可能提供任何单个失效的完全保护,可以达到99.999%有效性或更好。

Most backup systems achieve this by backing up the entire board. For example, two identical control processor boards may be used in a rack, and if one fails, the other can take over control. 48V distribution systems are also backup systems, with dual 48V feeds from separate circuit breakers to each board. If any individual circuit breaker releases, the board still receives uninterrupted power through the second feed. In most cases, there is no need to consider the benefits of backing up the power system itself on the board, as any board failure (power or otherwise) means simply replacing the board.

For effective backup, it is important that all component failures are immediately communicated to the board before the backup fails. In a power system, this means not only fully monitoring all output voltage rails, but also monitoring fuses and power feeds to detect backup failures. Additional monitoring, such as input current measurement and thermal sensing can provide warnings of overload conditions and further improve reliability.

Although today's power systems are becoming more complex, high reliability is achievable. Minimizing the number of components improves failure rates and produces an appropriate MTBF. Overall device reliability can also be improved with effective power management. Note that reliability is more important than a reasonable MTBF. Ensure that device requirements are met under all conditions by performing power system qualification testing.