Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Embedded

Science and art of telecom thermal design

Posted: 02 May 2005 ?? ?Print Version ?Bookmark and Share

Keywords:psb? pci? picmg? compactpci? pstn?

Open platforms like PICMG 2.16, also known as the Compact PCI packet-switching backplane (PSB), provide an ideal framework for quickly configuring high-density, high-availability telecom infrastructure systems that bridge the PSTN with emerging packet networks. To take advantage of the high functional density afforded by the CompactPCI PSB and the Advanced Telecom Computing Architecture (Advanced-TCA), however, designers must adopt power-management and cooling methodologies that can efficiently allocate power and dissipate heat. By optimizing thermal design, designers can make do with fewer and smaller fans and blowers, reducing acoustic pressure levels.

Thermal basics

CompactPCI PSB blades have a practical maximum power limit of 50W. Attached PCI mezzanine cards, the most popular option for CompactPCI PSB blades, have a maximum power limit of 7.5W (about 20W for processor PCI mezzanine cards). Advanced-TCA blades have a maximum power limit of 200W, while attached advanced mezzanine card (amc) modules range from 20W for the smallest module (half-height single-width) to 60W for the largest module (doublewide full-height).

In a CompactPCI PSB or Advanced-TCA shelf, blades are mounted vertically, and blowers force air across the boards from bottom to top. The air warms as it passes from component to component over the board, and board to board as it moves vertically through each shelf in the frame. High-density mezzanine cards like PCI and AMC create localized hotspots and obstruct airflow. Advanced-TCA blades are particularly troublesome because of their large size (8U) and ability to carry up to eight AMC modules. Compact PCI PSB blades are 6U and carry a maximum of two processor PCI modules.

Because of the impact of localized hotspots, preheating and airflow obstruction on overall heat dissipation, board designers must consider more than the aggregate power dissipation of the components they use on their boards. They must consider heat production and airflow impedance throughout the span of the board along critical paths. They should also be prepared to supply blade-level simulation data to telecom equipment makers, which can use it in their own simulations to optimize shelf- and frame-level thermal performance.

The PICMG 3.0 thermal specification provides a framework for thermal analysis of Advanced-TCA blades, but the same principles can be applied to develop cooling strategies for CompactPCI PSB blades.

Thermal analysis starts with the calculation of power dissipation and airflow impedance for various airflow paths and airflow rates. Next, the designer determines which portion of the airflow provided by the fan or blower assembly is distributed to each of the airflow pathways. Knowing the volumetric flow rate for each pathway and the power dissipated across that pathway, designers can calculate the temperature rise along that path. If the temperature rise across a given path is out of limits, designers can make a number of adjustments, including increasing the blower capacity, reducing the power dissipation in the highest dissipation paths or reducing airflow impedance in the most resistive paths.

Good thermal design is equal parts science and art. It is an iterative process requiring constant adjustments based on real-world feedback. It is also subject to real-world constraints such as the need to achieve a routable design with proper signal quality. So throughout the process, designers use empirical thermal-imaging and airflow gradient data collected from anemometers and thermocouples to verify that the original design parameters are realized.

A number of layout options were simulated when the Katana 3752, a PICMG 2.16 blade with three PowerPC processors, was designed. First, the designers placed the three processors along one side of the board (parallel to the airflow) and the memory devices on the opposite side, also parallel to the airflow. The effect was to create a low-impedance air path down the center of the board, effectively robbing the processor and memory components of airflow. This, together with preheating that occurred as air flowed from one processor to the next, caused the processor at the top of the board to run hot. Using larger heat sinks to restrict airflow down the middle of the board reduced processor junction temperature by making airflow across the board more uniform, but the heat sinks required to achieve an acceptable result were deemed to be too large.

Ultimately, the designers opted for a layout that positioned the three processors, equipped with moderate heat sinks, across the top of the board, perpendicular to the airflow. This enhanced airflow and, by placing the processors in parallel with respect to the airflow, minimized preheating. The result was a routable design with good signal quality that achieved a significant reduction in CPU and memory operating temperature.

System-level thermal management further requires a distributed monitoring and control infrastructure that gives shelf managers continuous access to such key blade-level parameters as voltage, temperature and fan speed. Both the CompactPCI PSB and Advanced-TCA provide such an infrastructure, known as the intelligent peripheral management interface (IPMI). Using an onboard I2C-based intelligent peripheral management bus, IPMI lets each blade communicate with shelf management and provide data derived from onboard sensors.

Active power negotiation

The ability to monitor blade-level voltage and temperature gives shelf management a leg up in ensuring that individual blades function as advertised. But shelf management also handles systemwide power, cooling and acoustic budgets.

One way that Advanced-TCA facilitates systemwide power, cooling and acoustic management is by supporting negotiated power management, which enables shelf managers to negotiate with individual blades for power allocation prior to powering on and then renegotiate that allocation while active.

Integrated IPMI management and negotiated power management give shelf management the visibility into and control over blade operation needed to optimize system power allocation and cooling.

- Robert Bourne

Global Technical Sales Manager

Artesyn Communication Products

Article Comments - Science and art of telecom thermal d...
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top