RTC Magazine

Over the past five years, CompactPCI has gained wide acceptance as the architecture of choice for equipment manufacturers developing high-availability embedded systems. Its powerful, standards-based computing solutions, hot-swap functionality and high-availability capabilities make it ideal for applications such as communication. AdvancedTCA (ATCA) is also emerging as a standards-based solution in the communications market. For both form-factors, intelligent power management and the power distribution architecture are critical to the success of high service availability for the complete system. System designers considering the use of advanced managed platforms in their applications must consider several power architecture issues.

Intelligent Power Management

CompactPCI power supplies can be fickle components in high-availability systems. Since today’s embedded systems demand higher performance and compute power, power supplies have been stepped up to deliver twice as much power as they did just a few years ago. However, they must deliver that increased power within the same form-factor. These denser power supplies require ample cooling and management, hence the need for intelligent power management.

In the past, power supplies offered simple management information. The two most typical management signals are the Degrade signal, which provides warning when power supply temperature is within 20°C of derating point, and the Power Fail signal, which indicates any output below 90% and/or a low input <36 VDC.

Recently, power supplies have begun implementing the industry standard Intelligent Platform Management Interface (IPMI). IPMI offers a comprehensive set of information to fully manage and predict failures for each power supply. It allows subcomponents from multiple vendors to be monitored by a system’s single shelf manager or redundant shelf managers.

The Intelligent Platform Management Bus (IPMB) is a logical bus that was specified to use the I2C control bus as the physical interface. Power supplies implementing IPMI contain a satellite IPMI/(IPMB) controller. This controller provides an interface to sensors and attaches to the IPMB/I2C bus as the physical interface.

It processes IPMI commands communicated via the IPMB bus protocol. IPMI is a request-response protocol: the shelf manager, also considered the master or baseboard management controller, issues a request message to an intelligent power supply, also considered the slave or satellite management controller. The supply then responds with a separate response message. Request messages and response messages are transmitted on the bus using I2C master write transfers. The sensor data record (SDR) is the sensor information stored on the power supply.

Many shelf management modules provide an in-band and/or out-of-band Ethernet interface so that the shelf can be remotely accessed with a common interface by operations, administration and maintenance (OA&M) managers. The shelf manager then communicates to all intelligent components via IPMI.

A power supply that supports IPMI provides five distinct classes of commands (Figure 1). The main commands are: (1) application commands, (2) sensor/event commands, (3) field replaceable unit (FRU) commands, (4) firmware commands and (5) OEM commands.

Application commands initiate operation with the controller, including controller resets and enabling self-tests. The device ID field allows controller-specific software to identify the unique functionality provided by a particular controller. Since multiple power supplies with the same device ID can operate in a system, power supplies can be standardized, yet, at the same time, maintain different geographic locations and FRU information. The Cold Reset command makes the controller reset, while the Warm Reset command makes it start at the beginning of the program, causing the initialization and startup function to be executed.

Sensor/event commands provide the information read from the power supply’s sensor data record. Typical SDR information for a power supply includes temperature, voltage and current. The Set Sensor Threshold command lets the user establish different threshold levels for each sensor: minor, major and critical. Each threshold on each sensor can be enabled to generate event messages if that threshold is crossed. For controllers that support the generation of event messages, the Set Event Receiver and Get Event Receiver commands must be implemented. When a significant event occurs, such as a power supply failure, that event must be communicated to the shelf manager as soon as possible, rather than the shelf manager being required to wait to respond to a command. In this case, an event message is initiated by the satellite controller.

The FRU commands provide asset information about the power supply. Asset information is critical to high-availability applications in many ways. The power supply FRU stores the version of firmware it is running. This firmware can be upgraded remotely if required. The FRU information about the power supply—such as part number, serial number, asset tag and date of manufacture—ensures that the technician removes the correct power supply and replaces it appropriately. In addition, knowing which specific power supply is installed can also drive routine FRU replacements, since some OA&M managers proactively replace a component before it fails based on its Failure in Time analysis.

Firmware commands can update the power supply effectively and remotely. These commands can also be used to erase flash memory space occupied by the program code being updated and to program the flash.

OEM commands allow users to change the state of the power supply either locally or remotely. This includes commands that monitor and override the state of the Power OK LED and the power supply shutdown output, a device status command, device-implemented FRU-type commands and device set-up commands.

The ATCA specification eliminated independent power supplies from the platform, but ATCA power entry modules (PEMs) also include IPMI controllers. PEMs take the -48V input from telecommunication facilities and distribute it to various slots and supporting components in the platform, such as fans. Each board must convert the -48V into the voltage levels needed. Typical ATCA platforms support two redundant PEMs. The shelf manager can monitor the voltage, current and thermal sensors on the PEMs. If one PEM fails, the shelf manager will be notified so the failed PEM can be replaced.

Some power supplies combine high power capacity and standards-based IPMI management capability with monitoring and reporting capability (Figure 2). These critical elements, along with redundancy and hot swap, give embedded applications the performance and management capabilities in power supplies they need to achieve high availability.

Power Budgeting

When dealing with high-performance applications in a CompactPCI platform, it is important to develop a power budget analysis to ensure that there is sufficient power for the complete system configuration. High-performance applications tax both the power and cooling architectures supported in the platform. Power budgeting can also provide a rough estimate of how much average power can be delivered to each platform slot.

To properly budget power for 12U platforms, it is best to develop a spreadsheet to account for all components that draw power. CompactPCI defines four independent and limited voltage rails that deliver power to the platform: 5V, 3.3V, 12V and -12V. It is important to analyze power budgeting at each voltage rail, because one specific configuration of boards in a chassis may not tax the 5V rail but could significantly tax the 3.3V rail, for example. Each voltage rail is independent and does not share current. Thus, when one rail’s limit has been reached, the limit of the number of boards that can be integrated into the chassis using that specific voltage has also been reached.

The power budget analysis begins with a list of the maximum current rating for each component and board to be integrated into the platform. For each unique board, the designer should write down its maximum current draw in amps per voltage rail. Using the maximum power draw for each component is necessary to calculate a worst-case scenario.

The next step is to list all boards and components to be configured in the platform, including quantity. Figure 3 shows a sample power budget analysis spreadsheet in table form. Each row represents the maximum current rating for the total number of units configured in the platform. This is broken out into the four voltage rails. At the bottom of the table, the total current for each voltage rail is added up to determine if the power supplies provide ample power. If the analysis shows a deficit in the current available for one of the power rails, the designer has several options. These may include removing one payload from the system, using alternate boards that consume less power, or looking for power supplies that can deliver more power.

Another question to keep in mind in a CompactPCI or AdvancedTCA deployment is whether the facility provides enough power to the rack or frame. The power budget analysis shown in Figure 3 assumes that the facility can provide sufficient power to the power supplies for CompactPCI and to the PEMs for AdvancedTCA.

Increasingly, designers are realizing the benefits of higher levels of availability. Not only does this provide them with a leg up on the competition, but it also reduces total cost of ownership during the life of deployment. System management is an important element in this equation, and power supplies are no exception. They need to be managed intelligently like any other high-availability board or component in the platform. With never-ending performance enhancements, it is important to keep in mind the analysis necessary to ensure power supplies can power the complete configuration.

Performance Technologies.
Rochester, NY.
(585) 256-0200.
[www.pt.com].