All electronics generate heat. No matter how low the power specification is on a device, it will generate some heat. But heat generation becomes particularly significant when it reaches high levels, because excessive heat is not good for electronic circuits. At overly high temperatures, circuit performance deteriorates, systems shut down and, in extreme cases, the equipment creates a fire hazard. On the other hand, extreme cold presents challenges, too, since a critical component or even an entire circuit may not operate as temperatures approach -40°C.
The temperature profile of every electronic application depends on the cumulative effect of heat generated by each device and on the ambient environment. So when faced with harsh environmental conditions and/or high levels of power dissipation in circuit, equipment designers often must measure the ambient and/or component temperatures and adjust the operating parameters accordingly.
Once the designer has established that temperature sensing is required, the next step is to design an accurate temperature-measuring system. This is not as straightforward as it may seem. If done without any consideration given to basic temperature-measurement principles, the design could suffer from inaccurate temperature readings or the wrong temperature zone being monitored.
While the focus here will be on applying digital temperature sensors, which constitute the bulk of the silicon temperature sensor market, the principles also apply to more mature discrete temperature sensors such as thermistors, thermocouples and resistive temperature detectors.
As a first step in developing a temperature-measurement system, the designer must determine what thermal zones need to be monitored, as well as the temperature ranges in which the application could safely work and the ranges in which it would perform best. The aim is to maintain the equipment temperature within an optimum range by such methods as controlling current consumption or airflow through the thermal zone affected.
|Material||Thermal conductivity (W/mK)|
|Diamond||1000 to 2600|
Accuracy of the temperature measurement is the next factor a designer must consider. If the operating temperature range is small, then a digital temperature sensor with high accuracy, perhaps ±0.5°C, may be necessary. On the other hand, the designer may have a wide operating temperature range with which to work. In the latter case, a temperature sensor with accuracy on the order of ±2°C would suffice and such a sensor would obviously be less expensive.
Once these considerations have been addressed, the designer may have a good idea of the performance specifications the temperature sensor must have and be able to select a sensor from half a dozen options available. The challenges then become selecting the right sensor package, placing the sensor in the correct position on the pc board to measure the appropriate temperature and, finally, designing the pc-board layout to get the best and most accurate temperature measurement. But to accomplish these goals, designers must understand the basic theory of how temperature sensors work, or they risk making system-design errors.
The naturally occurring direction of heat transfer is normally from a high-temperature object to a lower-temperature object. Heat transfer from a cold region to a hot region (as in refrigerators, for example) is only possible with the addition of external energy to the heat-transfer system. Apart from these conditions, there are three basic modes of heat transfer: conduction, convection and radiation.
Conduction is the most common means of heat transfer in a solid. On a microscopic scale, conduction occurs as rapidly moving or vibrating atoms and molecules collide with neighboring atoms and molecules, transferring some of their energy (heat) to these neighboring atoms.
Convection is usually the dominant form of heat transfer in liquids and gases. This is a term used to characterize the combined effects of conduction and fluid flow. In convection, heat transfer occurs partly through conduction and partly through transport by movement of hot and cold portions of the fluid.
Radiation is the only form of heat transfer that can occur in the absence of any form of medium. As such, it is the only means of heat transfer through a vacuum. Thermal radiation is a direct result of the movement of atoms and molecules in a material. Because the amount of radiated energy increases with rising temperature, a net transfer of energy from higher temperatures to lower temperatures results.
Thermal Conduction in PC Boards
If one end of a pc board is at a higher temperature, energy is transferred through the pc board toward the colder end. The higher-speed particles collide with the slower ones, resulting in a net transfer of energy to the slower ones. This is shown in Fig. 1, and the rate of heat transfer for conduction is given by:
where H equals energy conducted in time (J/sec), K equals thermal conductivity of the copper (385 W/mK) at room temperature, A equals the area of copper on the pc board (m2), T equals temperature (°C) and L equals the distance between hot and cold bodies (m).
Heat travels faster from a hot body to a cold body if the area of the medium it is conducting through (copper, for example) is increased. Likewise, if the area of the medium is reduced, the heat-transfer rate is reduced. Common sense deduces that the more distance between the two bodies, the longer it takes for the cold body to heat up. As an excellent conductor of heat, copper is used in many pc-board designs to dissipate heat from a heat source. Silver and diamond are the only other materials to have better thermal conductivity (Table 1).
PC-Board, Component Temperature Sensing
A designer uses a digital temperature sensor to measure two primary thermal zones. One zone is the temperature of the pc board and mounted components. The other is the temperature of the ambient air before it is affected by any heat-generating components on the pc board. To accomplish either goal, the designer must be aware of one important fact: The pins of the temperature sensor transfer 60% to 65% of the heat to the thermal sensor die (for silicon sensors).
The ground pin is connected to the substrate. Therefore, the ground pin has the least thermal resistance between the temperature sensor and heat source. Yet, many designers are under the misconception that heat is mainly transferred through the plastic package of a digital temperature sensor.
Though there are many succinct descriptions available in numerous textbooks and temperature-sensor datasheets, the intricacies of a digital temperature-sensor circuit are beyond the scope of this article. However, a designer may apply three pc-board tips to ensure the temperature sensor tracks and accurately measures the temperature of the correct pc-board region that will accurately reflect the temperature of the target heat source or device. The first tip is to use a common ground plane between the temperature sensor and heat source. Second, ensure that all ground pins of the temperature sensor are connected to the heat-source ground plane. Third, keep the temperature sensor and heat source as close as possible to each other on the pc board.
Fig. 2 illustrates how these techniques are applied to a pc-board layout in which the temperature of a heat-producing component is to be monitored. Fig. 3 reveals that this arrangement is not only accurate, but also allows for temperature changes to be tracked in real time.
Measuring Ambient Temperature
Some customers want to monitor air temperature, but still take advantage of the the accuracy, linearity, speedy response and convenience of an IC temperature sensor. However, in doing so, they must prevent the heat dissipated by the main heat source on the pc board from affecting the temperature measurement.
There are five guidelines that can be applied to prevent the heat-dissipating component from affecting the temperature sensor and to accurately monitor ambient temperature. First, use a hash ground plane that will reduce the ground plane area to increase thermal resistance. Second, keep the temperature sensor as far away from heat sources as possible. Third, use a separate ground plane for the temperature sensor and keep the number of connections to the main ground plane as low as possible. Fourth, use narrow ground connections to increase thermal resistance. Fifth, use a solid ground plane under the main heat source and expose the green solder mask. Following this last guideline gives the minimum thermal resistance for the main heat source to dissipate heat, diverting the heat flow away from the sensor.
The pc-board layouts shown in Figs. 2 and 4 are used to implement the same circuit topology but are physically different, because the layout in Fig. 4 is optimized for measuring ambient air temperature, according to the previous guidelines. The graph in Fig. 5 clearly shows that the temperature sensor implemented in the pc board in Fig. 4 is not affected by the heat from the main heat source and is accurately measuring the ambient air temperature.
Package Thermal Characteristics
Apart from size and pin count, there are several other package considerations such as package thermal resistance, power dissipated in the device, soldering temperatures and response to thermal shock. Two package performance metrics usually indicated in datasheets are junction-to-air thermal resistance (θJA) and junction-to-case thermal resistance (θJC).
The ease of heat flow between the die surface and air is indicated by θJA. It reflects how heat flows from junction-to-ambient temperature via all package paths. The primary temperature-transfer path is from the package leads to the board, so the thermal resistance of the leads is more important to digital temperature sensors than package resistance. The value of θJA is especially relevant for packages used without external heatsinks (Fig. 6). The lower the value — measured in units of degrees Celsius per watt — the more effective the package at heat transfer. For example, for an 8-lead MSOP, θJA equals 205.9°C/W. However, for an 8-lead SOIC, θJA equals 157°C/W. Therefore, the latter package is more effective at heat transfer between the internal die and the ambient environment.
The ease of heat flow between the chip surface and package surface is indicated by θJC. It therefore reflects how heat flows into the external heatsink (Fig. 7). The lower the value, the more easily the heat flows into a heatsink. This parameter is also directly determined by the package design. For example, for an 8-lead MSOP, θJC = 43.74°C/W, while θJC = 56°C/W for an 8-lead SOIC.
An equation is available in the absolute maximum ratings section of most datasheets for temperature-sensing components. It is defined as a maximum power rating. It is a general equation for any temperature-sensing device, but not a major factor in temperature sensing. Temperature sensors are designed to draw as little power as possible because self-heating can result in a rise in temperature readings. Thus, the power-dissipation number is intended to warn a designer that as heat rises, the component's ability to dissipate power is diminished, and this affects temperature-measurement accuracy as well as potentially damaging the device. This ability is determined by the relationship defined by the equation for maximum power:
where PMAX equals the maximum power dissipated in the device, TAMBMAX equals the maximum ambient temperature specified in the datasheet, TJ equals the junction temperature, and θJA equals the junction-to-ambient thermal resistance in degrees Celsius per watt.
Again, the basic guideline for applying the equation for temperature sensors is to warn the user that extreme heat exposure has a destructive effect on the temperature sensor. IC plastic packages are designed to withstand heat between 100°C and 150°C for a limited period of time. Prolonged exposure will shorten the lifetime of the device. Fig. 8 shows how this equation determines the relationship between PMAX and TAMBMAX, and how a MSOP package has lower thermal margins for heat and temperature than an SOIC due to the latter's lower θJA value. The graph also reveals that package components are in danger of being destroyed once the temperature rises above 100°C.
Package Soldering and Thermal Shock
Analog Devices now produces two types of package leads, tin-lead leads and lead-free leads (as of 2006, all new parts released from Analog Devices contain lead-free materials). There are different time and temperature parameters when soldering both types of leads. The most important difference between the two lead types is the peak soldering temperature. The peak soldering temperature specifications for tin-lead and lead-free assembly are highlighted in Table 2.
Reducing the thermal resistance between the die and heat source reduces the thermal time constant and improves the thermal response of the die. One thermal time constant is the time it takes for a temperature step applied to the sensing region of the sensor to produce a reading of 63.2% of the final temperature value, which should be equal to the final value of the applied temperature step. In Fig. 9, the ADT7301 experiences a thermal shock from 25°C to 125°C; it typically takes 2 seconds for the ADT7301 to reach 88.2°C, so this is the value of the thermal time constant.
The same thermal time constant is used for the ADT7301 for both the SOT-23 and MSOP packages. Evaluation data has shown that package type only has a small effect on the thermal time-constant value. This indicates that most of the heat flows through the package leads. Therefore, the values of θJA and θJC have little effect on the thermal response of surface-mount digital temperature sensors. In general, effective ground-pin contact to the ground plane of the heat source is far more important than the package type. Most modern temperature sensors draw very little current, on the order of microamps. As a result, power dissipation and, consequently, self-heating are not significant design factors.
In the case of current-output temperature sensors (for example, the AD590, AD592 and TMP17), package types TO-52, T0-92 CQFP and SOIC rely on a low θJC + θJA for fast thermal response. Note that there are no ground pins on these parts.
The LFCSP (Fig. 10) has a metal stub at the base that is directly connected to the ground of the die. Connecting this stub to the pc-board ground plane gives the LFCSP a lower thermal resistance than most packages and, consequently, the device has a lower thermal time constant.
|Average ramp-up rate||Tin-lead assembly(3°C/sec maximum)||Lead-free assembly(3°C/sec maximum)|
|• Temperature minimum (TSMIN)||100°C||100°C|
|• Temperature maximum (TSMAX)||150°C||150°C|
|• Time (TSMIN to TSMAX)||60 sec to 120 sec||60 sec to 120 sec|
|Time Maintained Above|
|• Time||60 sec to 150 sec||60 sec to 150 sec|
|• Peak temperature||220°C||260°C|
|• Time within 5°C of actual peak temperature||10 sec to 30 sec||20 sec to 40 sec|
|• Ramp-down rate||6°C/sec maximum||6°C/sec maximum|
|• Time from 25°C to peak temperature||6-minute maximum||6-minute maximum|