Featured Topic

Intelligently Heating Your Computer

Cold Kills as Badly as Heat

It doesn’t seem like it’s heard enough: frigid temperatures can be a serious problem for computers, in some respects an even bigger problem than heat. Extreme cold can cause display distortions like spotting and motion blur, or corrupt the output of onboard components in unpredictable ways, up to and including system failure. For outdoor industrial sites located in icy environments these problems are often the deciding factor when attempting to add digital control and automation.

As any petroleum industry line tech who’s had to manage a panel HMI on an arctic oil rig will tell you, protecting computers from freezing cold is at least as much of a problem as managing heat. Cold isn’t less of a problem, and thanks to industry bias, it’s a bigger one. Adding external heaters is inefficient and, often, ineffective. The solution is building systems with built-in heating controls. Unfortunately, many systems which are equipped with heating controls don’t go much beyond the average mirror defogger. You would think that engineering industrial-grade equipment in the 21st century would require a design more complicated than what you can find in any local hardware store, but sadly, for most systems that’s not the case.

Industrial Applications Need More than Corner-Store Tech

Hotel mirrors are relatively simple: plaster a low-grade heating element onto the back of a mirror and connect a switch. Turn it on, and it heats up until it hits the peak temperature. There are no adjustments involved beyond turning it off. For a computing system that’s going to be used in open, snowy-sleety spaces, though, something more than just an on-off heater is wanted. This kind of control loop is called a bang-bang control, and as one would suspect they are imprecise and, in systems requiring careful calibration within a specified temperature range (as is found on, say, a touch-screen panel), a critical liability. For sensitive applications, bang-bang control loops are too prone to premature system failures, hardware deterioration, and performance degradation to be of much use. A built-in heating system cannot simply set a target temperature and let the heater run. Advanced industrial computing hardware demands more sensitive controls, where the system temperature is adjusted relative to a target state that is constantly measured: an intelligent heating system.

Proportional Controls for Intelligent Heating

Proportional heating controls deliver a spectrum of wattage rather than just full on / full off. As shown in the diagram below, electricity supply is calibrated according to a sensor that must monitor heat output, managed by software systems that constantly compare the relationship.

Figure 1: Basic Proportional Control Feedback Loop

In this figure we see the basic components of a proportional feedback design. The temperature sensor feeds its output to a comparator, which then instructs the wattage controller (PWM) to either shut down the system (because things are too hot), or to increase or decrease electrical output—which in turn increases or decreases heat. Changes in heat output are registered by a thermistor, which then feeds that information back to the comparator. The control loop is a continuous cycle, and can only be said to end when it breaks out of the loop and turns itself off. For all of these reasons, even crudely precise proportional controls can deliver considerably greater system efficiency and safety in comparison to a bang-bang approach.

Engineering Proportional Controls

The tricky part in the design is the conservation of system resources, both in terms of software and hardware. This is a challenge when calibrating electrical output to target temperatures. Guaranteeing that the electrical supply delivers predictable heat output for each temperature in the target range is not as easy as it sounds. For one, each hardware platform carries a unique heat profile, and will exhibit unique behavior at different temperatures. For another, hardware components can sometimes behave erratically relative to the overall system: motherboard resistors, for example, change their performance at very low temperatures, and this can cause extremely erratic, inaccurate computational results that may even cause a system crash.

Consequently, considerable experimentation and measurement of each platform is required to calculate heat curves across the full range of electrical supply. In this way, the underlying software may be tailored to each platform to consistently deliver the target conditions.

Figure 2: The thermal curve of a computer heated from -40°C to 0°C over 40 minutes—notice its consistency.

X-axis is time (40 minutes) Y-axis is temperature.
Tc_HT1 = Temp. of heating panel A Tc_Front = Temp. of enclosure’s front panel (LCD panel)
Tc_HT2 = Temp. of heating panel B TA = Baseline environmental (ambient) temp.
Tc_Back = Temp. of enclosure’s back panel (computer)

An Overview of a Working Design

With proportional controls, when an automated heating system receives a power-on signal its comparator first checks the temperature, and finds that its thermistor is registering, say, -40°C. The heating elements are thus activated at full power. The software subsystem then continuously monitors wattage relative to heat output and, as changes in temperature are logged by the thermistor the values are passed to the comparator—which decides whether or not to continue powering the system—and then next, through the controller, which decides how much power to cut (or add).

Eventually, the system will heat to a temperature somewhere just beyond 0°C, whereupon the controller will either establish a state of equilibrium, or—if the heat generated by internal platform components is great enough to maintain the system above the target temperature—the comparator will turn it off. This is the sort of refinement and efficiency offered by proportional control; it is what puts the intelligence into an intelligent system.

Figure 3: A curve showing power output relative to heat output, as a platform is heated from -40°C to 30°C
Y-axis shows wattage output by the PWM controller as a percentage of capacity. X-axis shows heat output in Watts.

Wrapping It All Up

By taking the time and care to select the right components, build effective system failsafes, and perform extensive testing and full platform profiling, a safe, reliable heating system will intelligently support computer operations over an extremely wide range of cold temperatures that would render other platforms useless.

Figure 4:Temperature response times over a period of two heat cycles totaling nearly 40 hours of continuous operation.

This is a single test taken from late in the development of Moxa’s Intelligent Heating System. X-axis is time, Y-axis is temperature, in Celsius. Peaks are 30°C, lows are -40°C. Yellow line shows ambient temperature; the other lines represent benchmark locations within the platform. Rises in ambient temperature represent a period of about half an hour, decreases represent a period of about 2 hours.

Moxa’s latest line of panel computers, the EXPC-1319, now come with our Intelligent Heating System, Moxa IHS. For IHS, we have developed a number of patented technologies that provide a safe, reliable, and energy-efficient heating solution that guarantees these computers will rise up out of the freezing cold to securely and dependably deliver the performance you require for mission-critical industrial applications.

Come read up on the technical details of how Moxa Intelligent Heating Solution works in our latest white paper.

Back to index