Staying Cool About Thermal Management

Any discussion surrounding heat dissipation within industrial PCs is relevant because the temperature in the housing has a significant influence on the service life of the complete system. This means it is important in practice to have the support of the IPC manufacturer and create a suitable cooling concept for every device. One of the central questions is whether the industrial computer is equipped with a high-quality fan or whether passive cooling is the appropriate solution.

In an industrial environment, high-performance PC systems and displays normally operate at the upper end of the permitted temperature limit when subjected to maximum loading. This limit is determined by the installed electronic components that, depending on the specification, should run continually at an internal temperature that does not exceed 55° - 65°C. For example, a temperature difference ∆ T of only 5 – 15K remains at a 50°C ambient temperature.

In general, heat is dissipated in three ways; conduction, radiation and convection, and all three physically different cooling methods are often combined. In all cases, the thermal resistance from the interior to the exterior should be as small as possible as it exhibits similar behaviour to electrotechnology; the temperature difference corresponds to the voltage drop, while the thermal resistance corresponds to the electrical resistance.

Maximising Conduction

Heat conduction is a molecular heat transport in a material as result of a temperature difference. And as the Second Law of Thermodynamics states that heat always flows from the hotter body to the colder body, this means all components in a system must have the same temperature if they are in thermodynamic equilibrium. The calculation basis of the heat conduction is:

Rth: thermal resistance (K/W)
l: length of the body (m)
A: cross-section area (m²)
λ: specific thermal conductance (W/m/K)

The specific thermal conductance for dry air at 20°C, for example, is 0.0256 W/m/K. Water has a λ of 0.598 W/m/K (20°C). With 380 W/m/K, copper obviously has a very high specific thermal conductance. The thermal resistance Rth is the reciprocal of the thermal conductance.

The thermal resistance multiplied by the heat flow Q (in watts), for example the heat loss of a processor, produces the temperature difference caused by the thermal resistance.

∆T = Q (W) * Rth (K/W)

For the series connection of thermal resistances, the Rth values are added; for a parallel connection, the thermal conductances are added.

Some heat conduction paths have a number of resistances; CPU cooling, for example. Here, the Rth values of the semiconductor, the chip housing, the heat-conducting paste, the heat sink and the transfer resistances from one material to another or from the heat sink to the surroundings, such as the temperature in the interior of the housing are added, see Figure 1.


Figure 1: Heat is dissipated from the processor die over the heat spreader and the heat sink

Air spaces must always be avoided during the installation of a heat sink, so flexible heat-conducting materials must be used to compensate for any unevennesses and tolerances. Too much, however, is also not good. The temperature increase caused by an excessively thick layer of heat-conducting paste or heat-conducting pad can be calculated easily. The specific thermal conductance for a heat-conducting paste is approximately 1.2 W/m/K. For a surface area of 25mm x 25mm and a thickness of 0.2mm, the above equation yields an Rth of 0.26 K/W.

A state-of-the-art mobile-class CPU, e.g. an Intel Core i5 520M, has a TDP (Thermal Dissipation Power) of 35W and in the initial approximation (the heat is completely dissipated at the top) produces a temperature difference of 9.1 K, even for this thin layer. The temperature difference would double for twice the layer thickness or half the surface area:

∆T = 35 W * 0.26 K/W = 9.1 K

It is not difficult to appreciate how important the optimisation of the thermal resistance is for the heat dissipation of a device.

Heat Radiation

Radiation is a form of heat transfer through air via electromagnetic waves, where the heat radiation depends only on the temperature and the nature of the radiating surface. A body that has a large absorption capacity also radiates a large amount of heat (Kirchhoff thermal radiation law). The maximum possible heat radiation of a ‘black box’ at 25°C is 447 W/m2, and 617 W/m2 for 50°C. The deviation for the surface properties from the ‘black box’ is described by the emission coefficient ε and corresponds to the absorption ratio.

The emission ratio of a bare heat sink is approximately 0.03, whereas that of a black anodized heat sink (depending on the surface quality) is approximately 0.6 to 0.7. What does this now mean for the design of a heat sink, for example, in the housing interior? In this case, the heat sink is often directly surrounded by other components and housings that also radiate heat.

If the associated heat sink is made of a black anodized material it has a high emission ratio, but because of its high absorption ratio it also accepts heat radiated from the surroundings very well. The temperature difference between the heat sink and surrounding components can be so small that the heat radiations almost cancel themselves out, so the emission ratios are no longer significant. Indeed, when heat sinks are mounted externally on the housing, the irradiated energy for system-internal or outdoor applications can be greater than the radiated heat energy.

Experience has shown that the radiation proportion for the cooling of electronic systems is very small (maximum 20%). Numeric calculation methods have proved to be best for calculating and optimising the system configuration.

Fan-Supported Convection

Free, natural convection depends on the flow length of the body (for example, the height of the heat sink) and the flow speed that in turn results from the temperature difference. The heat transmission is determined by the thermal conductance of the air and the thermal boundary-layer thickness. The thermal resistance at the transition from a heat sink or a housing to air depends on the heat transfer coefficient and the size of the associated surface.

For passive cooled systems an upright system is much better than a flat device with the same dimensions, as the flow length is optimised. This alone can reduce the internal temperature by more than 5K.

In addition to free convection, forced convection (active cooling) is deployed in many cases. The housing fan transports cool ambient air into the housing interior or extracts the heated interior air to the exterior. A side-effect is it also reduces the thermal resistance of the internal air at the external housing.

DSM Computer deploys various fans for actively cooling its wide range of industrial PCs and embedded systems. Figure 2 shows an external infrared image and an internal infrared image of the compact NanoServer E4-GM45 embedded system with fan, immediately after opening the cover. The high-performance industrial computer is based on the Intel GM45 chipset with a 12W TDP and the CPU Intel Core 2 Duo P8400 with two CPU cores and a 25W TDP. The maximum heat loss of the system with all components, such as main memory, hard disk and power-pack losses, is approximately 50 - 55W.


Figure 2: The NanoServer E4-QM45 embedded system is equipped with a fan; IR image

The selected fan should operate in the ideal range of the characteristic curve (Figure 3). To make the device characteristic curve as flat as possible and so keep the pressure loss small, the inlet/outlet openings should be at least 1.5 times the size of the fan cross-section. For device characteristic curve B, the fan operates in a much more favourable range than for characteristic curve A. This means that the volume flow is significantly higher. To minimise the fan noise, no sharp edges may be located directly in front of the inlet opening.


Figure 3: Fan characteristic curve. The operating point is the point where the fan and the device characteristic curves intersect

The widely-held view that a fan negatively affects the service life of a system is not correct. Nowadays, state-of-the-art fans have a service life of more than 80,000 hours and are no longer the weakest PC system component.

In practice, actively ventilated industrial PC housings, even with high-performance Intel Core i5-based systems, attain exterior-interior temperature differences of only 8 – 15K under full loading. However, passive cooled devices, even when the energy-saving Intel ATOM CPUs with 2 – 2.5W power loss and the appropriately small housing dimensions are used, require special measures in order to achieve a ∆T smaller than 20K. The passive cooling in this case depends only on the external surface, the internal surface and the geometry (height, edges, etc.) that affect the free convection internally and externally at the housing.

DSM's new H1-A DIN rail-mounted personal computer is characterised by passive cooling without forced convection from the interior to the exterior. Thanks to its standardised installation capability, the industrial computer with a width of only 7 depth units (122mm) can be easily installed in a switchgear cabinet or in a standard electrical cabinet. The central component of the computer is a small 70mm x 70mm Qseven module based on the energy-saving Intel Atom Z510 processor, with the Intel US15W SCH (System Controller Hub) and integrated Intel GMA 500 (Graphics Media Accelerator) chipset. To achieve optimum heat dissipation, the Qseven board is connected directly with the computer housing via a heat rail. The CPU and the System Controller Hub are also connected directly to the aluminium housing (Figure 4). The cooling fins are arranged so they also act optimally in an electrical cabinet. The maximum nominal system heat loss is approximately 8W. This means that the DIN rail-mounted PC does not require a fan.


Figure 4: The compact H1-A DIN rail-mounted PC integrates a small 70mm x 70mm Qseven module that for cooling is directly connected with the computer housing

DSM Computer uses temperature measurement to ensure all deployed components operate within their specified values. The temperature is determined using sensors (ambient, surface), with an infrared camera or by simulation. Furthermore, the function, the optics, etc. of the device can be tested in a temperature cabinet at the maximum temperature limit values in accordance with various DIN EN standards over a specific time period (Figure 5).


Figure 5: In the climate cabinet, the device function can be tested at the maximum temperature limit values over a specific time period

Serviceable Life

Why does the heat dissipation of a PC system or display play such an important role? The device temperature has a direct effect on its service life. The failure rate specifies how many objects fail on average during a time unit. If the failure rate is constant, the reciprocal is the mean service life MTTF (MTBF for systems that can be repaired). The MTTF (Mean Time To Failure) is the expected value for the time before failure. The MTBF (Mean Time Between Failures) is the expected value for the operating time between two successive failures. The expected service life specifies after which time the device fails because of wear.

Certain components are significantly more susceptible to temperature changes than others. In addition to hard disks these are, for example, the electrolytic capacitors of the power pack or the power supply for the CPU on the mainboard. Nowadays, to achieve a long system life time high-quality electrolytic capacitors are deployed in industrial PCs. These components are designed for an operating temperature of at least 105°C and are characterised by a particularly long service life. A reduction of the ambient temperature by 10K doubles the service life of the electrolytic capacitor. This rule-of-thumb serves as an average for all components deployed in the PC.