For today’s makers of high-reliability equipment, increased performance requirements have meant packing more and more hardware onto boards and stuffing higher-density boards into shrinking chassis.
As a result, the power density (power dissipated per unit area) of electronic products, measured by the ability to dissipate heat, has skyrocketed, increasing by a factor of 20 to 50 in the last few decades. More and more, hot new products mean hotter chips and higher internal temperatures. However, the maximum temperatures that components can withstand haven’t changed significantly recently.
Managing this heat is essential because more heat means lower performance, premature shutdown, and even system failure. High power levels and high packaging density require fans with higher pressure and airflow performance to push air through the product. This also means increasing power consumption for fans. How can an equipment maker provide the performance its market demands and still keep things cool? One approach is to use an airflow sensor to monitor hot spots.
TOO LITTLE, TOO LATE
Sensors used on circuit boards typically measure only the board or ambient temperature in a product. But by the time a component overheats and the sensor detects a high temperature, there is a delay due to the thermal inertia of the board and chassis. High heat density and slow response due to high thermal inertia can be a fatal combination.
Because airflow does the cooling, impending temperature rise can be much better predicted, much earlier, by monitoring airflow across the component or board area. A good solution would be to use a small sensor capable of airflow measurement at the board/component level. Such a device can monitor airflow and warn the system of a future temperature rise as the flow drops. One such device is the Accusense F600 pulse-airflow sensor (Fig. 1). This board-mounted sensor is just 0.5 by 1 in. In addition to measuring air temperature from 20°C to 60°C, it directly measures airflow. Velocity measurement range is 0.15 to 5 m/s, or 30 to 1000 fpm, with an absolute accuracy of ±20% of the reading with a repeatability of better than 5%.
With several sensors placed at critical locations on a board, these additional airflow data points help complete a picture of temperature rise before it becomes a serious problem, allowing the machine’s hardware and software to respond appropriately. For electronic products with both high power density and high expectations of availability, real-time air temperature and air velocity measurement can spell the difference between simply promising reliability and consistently delivering it.
PROTECT THE EQUIPMENT, PROTECT THE USER
For many applications, effective early warning may be enough to keep the product working at an acceptable level of service. By reducing power to certain applications long enough to cool their chips, machines can stay in service. In addition, temperature rises caught early can clue operators in on some predictable trouble spots, such as a fan failure, clogged air filters, or clogged inlets. When such preventive measures fail, knowing when critical levels are approaching means an opportunity for graceful shutdown, preventing data loss, lengthy reboot procedures, and unhappy users. But providers of critical services such as communications and IT systems don’t have the option of shutting down.
For these situations, the early warning via the addition of airflow information does not mean early termination, but more effective thermal management. Maintenance programs can rely on receiving additional data points, such as when air filter replacement or other maintenance is due. Airflow blockages can generate alarms and receive immediate service, changing the situation from an emergency repair to a preventive maintenance call.
And, conserving cooling power doesn’t mean just saving electricity. It means keeping additional power on tap at all times, staying up and running to provide critical services. Most importantly, for effective thermal management, componentlevel airflow sensors can provide the data necessary to manage the array of fans and fan boards found in this type of highpower equipment.
POWER, TEMPERATURE, AND FANS
More than any other single factor, knowing when to increase fan power and when it can be safely decreased is critical for managing both temperature and power consumption. Hot chips rely on fans for cooling, and fans rely on accurate temperature readings.
But a fan’s ability to cool a device isn’t linear. From heat transfer equations, the temperature rise of a device is inversely proportional to the airflow over it. At low airflow levels, temperature increases can be significant. As airflow increases beyond approximately 600 fpm, though, the reduction in temperature rise is minimal in relation to the rise in airflow.
The energy consumption by a fan is proportional to the cube of fan speed (N3). Therefore, as the airflow is increased for lower device temperature rise, the power consumed by fans can be significantly high, leading to negative net returns.
For example, in a sample case, an increase in airflow from 400 to 800 fpm results in a temperature drop of only 5°C with a required power increase of approximately 85 W (Fig. 2). The clear challenge is to operate at an optimum airflow range while keeping the necessary fan power within budget. This requires measurement of airflow. Air velocity becomes equally critical as air temperature in the efficient cooling of circuit boards and in controlling fan power.
Continue on Page 2
Designing a sensor into an existing product can often be challenging. In applications such as server motherboards where several 12-V fan connectors are available, the F600 may be used in an innovative mode called Frequency Out. Here, the sensor generates pulses of varying frequency similar to a fan.
The sensor may be mounted at an existing but unused three- or four-pin fan connector. The sensor emulates a fan by generating higher signal frequency at higher airflow levels. The system interprets the signal exactly as a fan’s tachometer signal and sends it through standard fan interface circuitry and software.
Many motherboards support BIOS that can measure the speed of fans, and many operating systems such as Linux, Windows, and Mac OS can read the fan speeds through firmware. A thermal-management application program can be developed to monitor and manage the thermal environment with predictive servicing features.
The F600 also supports other interfaces such as addressable I2C, TTL/RS-232, pulse-width modulation, and alarm interface. Components like the F600 are deployable as a single sensor or part of a network of sensors on a board. When using a dedicated I2C bus, up to four individually addressable sensors can operate simultaneously.
SIZE AND BUDGET CONCERNS
When dealing with heavily populated circuit boards, the airflow sensor, measuring 0.5 in. high and 1 in. long, is small enough to easily fit alongside chips and resistors. Keeping a low profile, the component minimally distorts the true airflow picture while recording air velocity and temperature across the board’s surface. The flow guide at the top of the package defines the airflow acceptance angle and helps protect its sensing element (Fig. 3). All of these features allow placement on the board in optimal locations, where airflow is relatively uniform, such as inline with heatsink fins or rows of memory modules.
Addressing cost concerns, the early warning, board-level airflow sensor is cost-effective when compared to much larger and less flexible airflow switches. Without exceeding budget lines, manufacturers can use more sensors and locate them in more places on each board, helping to make airflow measurement a mainstream application in board design.
Moreover, better thermal monitoring reduces power requirements and allows for tighter real-estate design for critical components, increasing overall benefits. In short order, improved thermal design may not be just a requirement to stay competitive, it may also point the way to new design strategies for reliability, supportability, and increased market differentiation.
This flexibility should free up manufacturers to design more compact, more powerful, and better-cooled products. The keys to the strategy are simple: small, flexible, inexpensive, and cool.