Managing Power in CompactPCI Designs

Over the past five years, CompactPCI has gained tremendous acceptance as the architecture of choice for equipment manufacturers developing high availability embedded applications. Its powerful, standards-based, cost-effective computing solutions, as well as hot-swap and high-availability capabilities make it ideal for communication applications.

Packet-based advanced managed platforms, such as PICMG 2.16, take the CompactPCI architecture further by employing comprehensive management and a suite of highly compatible and feature-complete components to deliver high-availability, application-ready platforms. As companies consider developing their applications with advanced managed platforms, there are several power architecture issues to consider:

1. Calculating the power budget can be a complex issue, as high performance boards tend to draw large amounts of power. One must take into consideration all the elements that draw power, as well as the limited current for each of the four voltage rails: 5, 3.3, 12, and "12 V.
2. Advanced managed platforms support multiple power configurations to balance between costs, power, and redundancy.
3. A new breed of CPCI power supplies offers the standards-based intelligent platform management interface (IPMI). Power supplies can be a fickle component in high availability applications. By adding intelligent management, these supplies can be monitored closely, even to the point of predicting failures before they occur.
4. Finally, isolated power input architecture provides another level of high availability and reliability to the platforms.

Let's look at these topics in more detail.

Power Architecture
Before we look at the four power-management issues highlighted above, it is important to gain a good understanding of the advanced managed platform power architecture. Figure 1 show the typical power architecture for a 12U advanced management platform.

Figure 1: Power architecture for 12U advanced managed platform.

12U advanced managed platforms with DC inputs use two isolated -48-V power feeds to deliver power into the platform. For high reliability, the two feeds are not ORed together with diodes. This is important for telco facilities that provide redundant power plants, because a diode failure could allow faults in one power plant to affect the other power plant.

In the 12U architecture shown in Figure 1, the power architecture is N+N, where N=4. Power feed A delivers power to the odd numbered power supplies and Power feed B to the even numbered supplies. This provides two levels of redundancy: power feed failure and power supply failure.

Advanced managed platforms can handle any single failure at a given time. If a single power feed fails, the system continues to receive power from the four supplies powered from the healthy feed. If a single power supply fails, the system continues to receive adequate power from the healthy supplies. If the ambient temperature rises above 40C, the power supplies and the platform (connectors, cables harnesses, midplane, etc.) can continue to operate for up to 96 hours up to 55C ambient, as defined by NEBS.

In a 12U advanced management platform, four 325-W power supplies deliver 1300-W output power. It is important to note that the 325-W rating actually means 325 W of output power since the drop of power due to efficiency is already accounted for. Table 1 lists the effects of various types of power faults with eight 325-W power supplies.

Table 1: Power architecture with eight 325-W DC power supplies

The AC power architecture of a 12U design is similar, except that there are four power feeds. AC input power is provided via four IEC connectors labeled A1, B1, A2 and B2. The A1 and B1 connections provide redundant power feeds to power supplies 1-4. For full redundancy, these connections must have independent circuit breakers and come from different sources (or at least different phases). The A2 and B2 connections provide redundant power feeds to power supplies 5-8.

Unlike 12U architectures, 4U advanced managed platforms use N+1 power architecture. In this case, there are a total of three power supplies, where one supply is used as a back up should another one fail. Using the same 325-W DC power supplies, the platform can deliver 650-W of redundant output power. However, the cooling architecture in this chassis limits the amount of power that can be utilized in this chassis to 500 W of redundant power.

Power Budgeting
When dealing with a CPCI platform, it is important to develop a power budget analysis to ensure all components operate correctly. High performance applications will tax both the power and cooling architectures supported in the platform. In this section, we will focus on the power. This analysis can also provide a rough estimate of how much average power can be delivered to each platform slot.

Typically, poor power budgeting skills look something like this: "Well the platform delivers a total of 1300 W of redundant power, and I've got boards that draw a maximum of 50 W of power each, so I can install 26 boards in the chassis." Not only did we not account for all the infrastructure elements that draw power, such as fan trays, switches and shelf managers, but we didn't account for the fact that there are four voltage rails in the platform, and each have limited power.

To properly budget power for 12U platforms, it is best to develop a spreadsheet to account for all the components that draw power. Here are three steps that designers can tap to develop a spreadsheet for power budget analysis.

Step 1: Maximum Current Available for Each Platform
CPCI defines four independent and limited voltage rails that deliver power to the platform: 5, 3.3, 12, and -12 V. It is important to analyze power budgeting at each voltage rail in addition to overall power of the chassis (e.g. 1300 watts), because one specific configuration of boards in a chassis may not tax the 5-V rail but could significantly tax the 3.3-V rail.

Each voltage rail is independent of each other and does not share current. So as soon as one rail is maxed out, you have reached the limit of boards you can integrate into the chassis that use that specific voltage. Thus, designers must calculate the maximum current for each of the four voltage rails from the total set of power supplies in the platform. Remember to discount the redundant supply(s).

Step 2: Peak Power Rating Per Component
Step 2 is to list the maximum current rating for each component and board that will be integrated into the platform. For each unique board, write down its maximum current draw in amps (A) for each voltage rail.

It is important when calculating power budget to use the maximum power draw for each component to calculate worst-case scenario. If a component's datasheet does not provide current ratings for each rail, contact the manufacturer. Designers should also be sure to include the maximum draw on the V (I/O) pins in with the 3.3- or 5-V rail depending on the V (I/O) configuration of the system.

The maximum power rating should take high environment temperature conditions into consideration. Typically, 40 deg. C ambient temperature is used for each board and component. NEBS calls out that the shelf shall operate up to 40 deg. C continuously or 55 deg. C for 96 hours (where the one failure is the ambient temperature). The NEBS specification also assumes only one failure at a given time. Thus, if ambient is at 55 deg. C, all power supplies are still operational and can power the boards successfully, even if they draw more current.

It is important to also account for the resistance found in cables, such as wire harnesses, that are found inside the platform. Note: Cable resistance increases as the temperature increases.

Step 3: Power Budget Analysis
Now let's configure these boards into the platform and calculate total power budget. Create a table that lists each board, the quantity of unique boards in the platform and the total current consumed per voltage rail. A simple example is shown in Table 2.

In Table 2, each value in the total current draw row represents the sum of each voltage column. Total current available, on the other hand, is the maximum current available from the platform we calculated in step 1. The delta is the difference between total current available and the total current draw from components and boards for each voltage rail.

During power budget analysis, it's important to note that the resistance of pure metals, such as silver, copper, and aluminum, increases as the temperature increases. The amount of increase in the resistance of a conductor per degree rise in temperature above 20 deg. C is called the temperature coefficient of resistance. For copper, the value is approximately 0.004041. This applies to the entire length of wire and for each degree of temperature rise above 20 deg. C.

Intelligent Management
To achieve higher levels of high availability, equipment manufacturers are requiring all active components to have intelligent management information. Intelligent management allows the system (OA&M) managers to understand the current condition of a component and predict failures before they occur.

Power supplies that do not support the IPMI standard typically only provided the DEG# thermal warning and the FAL# fail signal. Intelligent power supplies monitor temperature, output voltages, output currents, thermal warning status and failed status. Additionally, intelligent management provides IPMI-based controller reset, geographic address and field replaceable unit (FRU) information, such as part number, version number, serial number and manufacturing date.

Since power supplies are typically one of the first components to experience failure, some system managers like to perform routine replacement of power supplies after a specific number of years, months or services. By reading the FRU information of each power supply through an intelligent shelf manager (ISM_, the system manager can remotely detect which power supplies need replacing and set the power supplies LED red via the ISM.

Isolated Power Feeds
Newer central office (CO) environments are based on redundant -48-VDC power sources, where power is supplied to equipment by two on-site power plants. AC power, delivered by the public utility, is rectified to -48-VDC power within these on-site power plants (although it can range from -40 to -60 V DC). These, in turn, supply the CO equipment.

In the event that local AC power is interrupted; a bank of batteries within these power plants provides DC power until the motor generators take over. Supplying power in this manner allows designers to achieve high availability in the CO power system.

In order to retain this level of redundancy through to the platform level, several platforms use an N+N power architecture. In the N+N architecture, one set of power supplies (N) receives power from one feed, and the second set of power supplies receives power from the other feed. This keeps the feeds completely isolated from each other. In contrast, diodes tying two feeds together can compromise the integrity of the redundant feeds.

Wrap Up
Fully understanding the power architecture of a CompactPCI platform is critical to the success of a system designs. During this analysis, designers must ensure that there is ample power to drive demanding applications. Additionally, they must ensure that enough power is available for future applications.

About the Author
Tony Romero is a senior product manager at Performance Technologies. Tony can be reached at tony.romero@pt.com.