March 1999

The Enemy Within: Heat Build-Up In Computer Telephony Systems

BY JIM SATHER

The boom in the telephony industry has been accompanied by escalating processing demands and ever-shrinking form factor requirements. Rack space is expensive, so industrial computer systems must be strategically configured in the smallest possible chassis. The objective of "maximum power in minimum space" is further complicated by the requirement for redundant systems to assure reliability. While manufacturers and OEMs configure systems to reconcile this mixed bag of concerns, they can unknowingly introduce a threat to the application, one that will lurk transparently and strike without warning: heat build-up due to excessive power draw.

As processing modules are layered into industrial chassis as small as 7" high, they must share prime real estate with a hodgepodge of demanding neighbors. The cramped quarters must also play host to power supplies, hard drives, CD-ROM, tape and floppy drives, and possibly a passive backplane that facilitates multiple systems and their peripherals in a single box. Further, the OEM addresses specific user requirements by adding a customized mix of devices for functions such as networking capability, Internet data transfer, telephone dialing, and telephony management. After addressing these design issues, the OEM cannot breathe a sigh of relief, however, until the pervasive cloud of heat dissipation is addressed via a well-conceived thermal management strategy.

The build-up of heat inside the chassis can damage components or create a power supply thermal overload, ultimately resulting in costly down time for the entire system. Thermal management can be defined in simple terms as getting heat out of the box and into the temperature-controlled environment where it can be dissipated. More than simply "cooling" a box, thermal management suggests that the industrial computer manufacturer must take a comprehensive view of routing heat out of a chassis that is congested with traditional and specialized components, as well as a maze of cabling.

REDUNDANCY: RELIABILITY OR RISK?
The mission-critical nature of telephony applications calls for dual redundancy of essential components. Multiple processors, for example, can be tailored for a combination of parallel processing and redundancy. Should a processor malfunction, its counterparts pick up the slack and assure uninterrupted operation. Similarly, modular, pluggable power supplies are often duplicated to assure that redundant processors can perform without disruption. This combination of multiple processors and power supplies creates a formidable thermal management challenge for the industrial computer supplier.

A common misconception when configuring industrial systems is that specifications for individual components can be extrapolated to complex, multi-layered systems. In addition, a specification sheet for a 7" chassis might "document" its tolerance of a 400-watt power supply. Historically, however, such systems have not been taxed to this maximum level of power draw. For, example, traditional data processing applications may only require dissipation of 150 watts in the 400-watt chassis.

The space and performance demands of the telephony industry, however, have transformed the application landscape, and consequentially have tested the limits of heat tolerance in the smaller chassis. By extending the power requirements for a single processor across an entire dual redundant system, the telephony system buyer might conclude that the 8.75" or 10.5" chassis requires a 600-watt power supply. What the customer might not understand, however, is that such a system literally becomes a heater. Crowded with components, the chassis simply cannot dissipate that much wattage, and the significance of thermal management increases dramatically.

ALTERNATIVES: MINIMUM COST, MINIMUM COMPLEXITY
The spectrum of solutions for effective thermal management starts on one end with a water cooling approach. However, cooling the racks with water and subsequently routing the water through a chiller or a heat exchanger is an expensive approach that introduces its own set of maintenance concerns that might not appeal to a cost-conscious industry.

At the other end of the cost and complexity spectrum is simple, economical, effective convection cooling. The incorporation of convection cooling into industrial computers, while not highly complex, calls for the teaming of a savvy buyer with a knowledgeable manufacturer. This thermal management solution starts with "plain Jane" fans. The objective is to move air - properly filtered to avoid the introduction and accumulation of dust - through the box. The stream of air moves through the chassis, capturing heat, which is then strategically and uniformly routed out of the box.

One of the few certainties in the design of industrial computer systems for the telephony industry is that requirements vary greatly, and each application must be considered on its own merit. By connecting with the engineering and technical support staff within the manufacturing company, the buyer can be assured that the total system requirements will be met, not just from the data processing side, but from the thermal management side as well. Once the total picture of processing power, heat dissipation, and reliability is clear, the customer should not be surprised if the manufacturer suggests a slightly larger box. The suggestion doesn't mean that the manufacturer cannot configure all of the requirements into a smaller box, but rather than such a design might not be prudent.

COMPROMISES AND TRADE-OFFS FOR LONG-TERM SUCCESS
Certainly, the perfect "minimum size, maximum power" solution might be a 7" or 8.75" chassis that is completely maintainable from the front. However, the system reliability offered by proper thermal management calls for the consideration of compromises, such as rear-pluggable power supplies and a taller box, 10.5" or perhaps even 12.5" high.

The most logical approach to filtration is "positive pressurization," where air is blown into the box and through the filters. In doing this, the entire box is afforded the benefits of positive pressurization, meaning it does not have to be airtight. Further, the majority of the heat in the system is generated by the combined load of the computer cards and peripherals.

However, up to a third as much heat again is generated by the power supply itself. By mounting the power supplies in the front of a shorter chassis - and incorporating positive pressurization - the heat from the power supply is blown directly onto the card cage, compounding the problem by raising the temperature of the card cage. This temperature increase is not significant in moderately loaded systems, but in a 20-slot computer populated with 15 or so power-hungry intelligent processors, the temperature in the card cage area might reach levels of marginal data processing integrity.

Consequently, when the power supply loading becomes significant, usually due to multiple parallel or redundant processor cards, the most logical approach is to move the power supplies to the rear of a taller chassis. In this manner, the power supply (and the heat that it generates) is safely isolated from the card cage. The sacrifice of front loading power supplies and a shorter box pay off in the long run with more effective thermal management.

Further advantages of the 10.5" chassis make it a viable alternative for power-intensive telephony applications. The benefits start with the free path for air flow across the top of the card cage. In addition, the management of cabling is more efficient. The components and peripherals configured in the chassis require a maze of ribbon cables - from the card cage to hard drives, CD-ROM drives, and streaming tape drives - that creates an obstacle to free air flow. The slightly taller box facilitates a more sensible routing of cables - with the power supply underneath the card cage and sufficient space to route the cables down (while maintaining space for air to flow across the top of the card cage). Any hot air that is not dissipated in the lower regions of the box will rise freely to the top and be evacuated by the air flow across the card cage.

THE ONLY RULE: BE AWARE
The highly-customized nature of telephony applications preclude any steadfast "formulas for success." The industry has challenged the industrial computer market by setting new standards for performance and reliability. Each application must be considered independently, and customized solutions must be developed. The one rule that makes sense is to team with the technical staff of an experienced industrial computer manufacturer to properly evaluate the implications of the system requirements. Maximum power in minimum space is the starting point.

From there, consider thermal management issues as they pertain to the size of the box and heat dissipation. Build a prototype and test it in the application field. Keep in mind that, historically, most "500-watt," 20-slot boxes have rarely had to draw that much power in the field. A rule of thumb for the industry might be "the 300-watt threshold." Beyond this point, thermal issues move to the forefront of system configuration. Downtime is expensive, and "maximum power in minimum space, without reliability" doesn't make sense.

Jim Sather is the engineering manager of SBS Micro Alliance. SBS Micro Alliance specializes in the design and manufacture of rugged, COTS, or special-purpose PC, VME, and CompactPCI industrial and military computers, enclosures, and turnkey systems. They offer a variety of CPU boards and system enclosures, including rackmount, benchtop, workstation, and portable systems. For more information, please visit their Web site at www.sbs-microalliance.com.