|Home | Articles | Forum | Glossary | Books|
Sensors convert information about the environment such as temperature, pressure, force, or acceleration into an electrical signal. With the development of microelectronic technology with silicon as the base material in the 1970s, sensors using the properties of silicon entered the component market. Silicon's physical properties make it an ideal building material for mechanical devices. Silicon has the hardness of steel, the thermal conductivity of diamond, piezoresistive properties, a light weight, and low thermal expansion; also, it’s relatively inert.
It’s free of hysteresis and its crystalline structure is well suited to the fabrication of miniature precision products. Silicon micromechanical products have several advantages over their conventionally manufactured counterparts: They generally are much smaller, their performance is higher because of the precise dimensional control in the fabrication, and costs are lower due to the possibility of mass-scale production.
Silicon micromachining is a powerful outgrowth of semiconductor process technology, whereby integrated circuit manufacturing techniques are supplemented by the silicon etching process to create very precise micromechanical structures. These silicon microstructures can have electronic features that allow conversion of physical input into electrical signals. Similarly, electrical signals can be applied to these devices to provide control functions. Initially developed in the 1950s and 1960s at leading semiconductor pioneers including Fairchild and National Semiconductor, the technology was further advanced in the 1970s at universities throughout the world. Commercial activities picked up in the early 1980s, with a number of startups located in the Silicon Valley area.
By the beginning of the 1980s, designers were able to incorporate integrated circuits on a single die with the sensor elements. Although this complicates the fabrication process and can limit the operating-temperature range for the sensor, it often leads to superior performance at an acceptable cost. These integrated microsensors can provide a more linear output than that of the sensor itself or an output having a digital format that can readily be handled by associated data-logging or display systems. By the late 1980s, microsensors for measuring pressure, temperature, and the like were readily available, while silicon accelerometers and so forth were entering the market. FIG. 1 shows the comparative scale of microsensors.
Nowadays, miniaturization is the aim of many research laboratories and companies. As a part of microsystem technology, sensors also will play a major role in the future and sensor interfaces and related standards are getting ready for this developing component sector. Many producers in Japan, Europe, and the United States forecast growth rates for sensors above 10% beyond the year 2000. This section is a summary of modem semiconductor sensors, their characteristics, and applications with some representative devices.
2. The Properties of Silicon and Their Effects on Sensors
TABLE 1 The effect of silicon used in sensors
Physical Dimension Effect Application
Radiation Photoresistive Photoresistor Photo-interface Photodiode, phototransistor Ionization Nuclear radiation sensor Photocapacitive Photocapacitance Mechanical Piezoresistive, Piezoresistive power and pressure piezo junction, and sensors, piezoelectric diode and piezo-tunnel transistor; Thermal; Thermal resistance; Resistance temperature sensors; Thermo-junction Temperature sensors (diode, transistor)
Thermoelectric Thermopile Pyroelectric; Pyroelectric sensor Magnetic signals Magnetoresistive; Magnetoresistive sensors Hall; Hall generator; Magnetic interface; Magnetic diode and transistor Chemical signals; Charge-sensitive field ISFET
Silicon is a suitable material for sensor technology as it manifests sufficient physical and chemical effects of an acceptable strength to use in uncomplicated structures across a wide range of temperatures. Table 1 presents the most important effects and their applications for sensor technology.
The use of silicon has a number of implications for sensors. First, the physical properties of silicon can be used directly to measure the desired dimension, as indicated in Table 1. However, the range of possibilities is limited. Beyond this, For example, silicon can be extremely useful when used as the substrate for thin-film sensors, even when information processing electronics are integrated.
The term micromechanics, with its obvious similarity to the term microelectronics, is used to describe a completely new discipline. Its objective is the construction of complex microsystems consisting largely of integrated sensors, a logical signal processing stage, and actuators. In this connection, micromechanics refers to the fabrication of mechanical structures whose geometrical size, at least in one dimension, is so small that it no longer is sensible to use the methods of fine mechanics. Depending on the boundary conditions imposed by the desired function or the properties of the material, this limit may be located anywhere between the millimeter and the sub-micrometer range (see FIG. 2). In contrast to microelectronics, micromechanics is concerned with the production of three-dimensional structures.
Modern micromechanics make it possible to produce micro-pumps, micro-valves, micro-loudspeakers, and microphones; therefore, it’s of interest to disciplines other than sensor technology (Hauptmann, 1991; Guckel, 1992).
4. Temperature Sensors
The most common electronic temperature measurement devices currently available include the thermocouple, the resistance temperature detector, the thermistor, and the integrated circuit temperature transducer. All have associated application benefits and limitations, which are delineated in Table 2 (Microswitch, 1997a).
4. 1 Resistance Temperature Detectors
Resistance temperature detectors (RTD) are wire windings or thin-film serpentines that exhibit changes in resistance with changes in temperature. While metals such as copper, nickel, and nickel-iron often are used, the most linear, repeatable, and stable RTDs are constructed from platinum.
4.2 Negative Temperature Coefficient Thermistors
Negative temperature coefficient (NTC) thermistors are composed of metal oxide ceramics, low in cost, and the most sensitive temperature sensors. They also, however, are the most nonlinear and have a negative temperature coefficient.
Thermistors are offered in a huge variety of sizes, base resistance values, and R-T curves to facilitate both packaging and output linearization schemes.
Thermocouples consist of two dissimilar metal wires welded together at both ends to form two junctions. Temperature differences between the junctions cause a thermoelectric potential (i.e., a voltage) between the two wires. By holding the reference junction at a known temperature and measuring this voltage, the temperature of the sensing junction can be deduced.
Thermocouples have very large operating temperature ranges despite a very small size. However, they have low output voltages, susceptibility to noise pickup by the wire loop, and relatively high drift. Silicon integrated circuits are available for interface with the thermocouples. Some examples are AD-594 and AD-595 from Analog Devices (Le Fort and Ries, AN-274; Marcin, AN-369).
TABLE 2 (coming soon) A comparison of thermal sensors. (by Microswitch Honeywell Inc.)
4.4 Silicon Temperature Sensors
Temperature sensors that utilize the temperature-dependent properties of silicon are appearing in the market in a wide variety of types, and their prices are reasonably low. Practical integrated circuits available in the market basically are either voltage output or current output temperature sensors.
FIG. 3(a) shows the use of the temperature dependence of the PN junction voltage to provide a temperature-dependent voltage output, Vbe. The voltage is related to the temperature and other parameters by the equation
Vbe = In (1)
q where IF --forward current of transistor; Is = saturation current of transistor; q --elementary charge; k = Boltzmann's constant.
If the ratio IF ~Is is kept constant, then the result would be a sensor exhibiting ideal linear temperature-dependence of the forward voltage.
A similar relationship is found in transistors. If the collector and base are held at the same potential ( FIG. 3(a)), then the relationship of the base-emitter voltage, Vbe, to the collector current, Ic, is given by
Vbe = ~ In (2)
Here again the saturation current, Is, is influenced by the temperature dependence of a number of parameters, similar to the case of diodes. Despite this, if the collector current, Ic, is held constant and the components are carefully selected, it’s possible to obtain approximately linear behavior for temperatures between -50 and 150°C.
Motorola's MTS 10X series silicon temperature sensors are .a classic example of this technique. The device family allows temperature measurement precisely in the range -40 to 150°C. Modern temperature management ICs range from purely analog voltage vs. temperature devices to mixed-signal VLSI chips containing logic and ADCs.
Most ICs rely on a bandgap reference with a known temperature coefficient to provide temperature information.
4.4.1 Simple Current Output Temperature Transducers
Most simple current output temperature transducers use more practical forms of the circuit in FIG. 3. The AD-590 from Analog Devices is an example of this. Referring to FIG. 3(b), the difference in base-emitter voltages of transistors Q1 and Q2 is given by ....
The temperature dependence of Vbe therefore is solely dependent on the ratio, r, of the two collector current densities:
Provided that this ratio can be kept constant, A Vbe is directly proportional to the absolute temperature. There are two ways of keeping the ratio constant.
First, it’s possible to operate two transistors with the same geometric dimensions on a single chip using two collector currents (Ic1 7/= Ic2). The alternative is for a constant collector current to flow through two transistors with different emitter areas (AI ::7(: A2). The second variant has been of greater practical relevance because of the simpler circuitry involved. An example of this type of integrated sensor is presented in the basic circuit diagram in FIG. 3(b). Transistors Q1 and Q2 perform the detection function. The identical transistors, Q3 and Q4, act as current mirrors. This causes a splitting of current I into two equal collector currents, Ic1 and Ic2. The emitter area of Q2 should be r times that of Q1. Its collector current density therefore is only 1/r that of T1. The difference of A Vbe causes a current, Ic2, that is proportional to the temperature to flow across a resistor, R. Because of the current mirroring, the value of I also must be proportional to the absolute temperature. Laser alignment of the resistance R makes it possible to adjust the constant of proportionality in equation (4) to 1 gAK -l. If the circuit is changed to allow for a voltage output signal, then temperature coefficients of a few millivolts per Kelvin can be achieved.
In the AD-590, this A Vbe, directly proportional to absolute temperature (PTAT), is converted to a PTAT current by low-temperature-coefficient thin-film resistors. The total current of the device is then forced to be a multiple of this PTAT current.
FIG. 4(a) is a schematic diagram of the AD-590. Q8 and Q11 are the transistors that produce the PTAT voltage. R5 and R6 convert the voltage to current. Q 10, whose collector current tracks the collector currents in Q9 and Q11, supplies all the bias and substrate leakage current for the rest of the circuit, forcing the total current to be PTAT. R5 and R6 are laser trimmed on the wafer to calibrate the device at +25°C. FIG. 4(b) shows the typical V -I characteristic of the circuit at +25°C and the temperature extremes.
The device features a 1 gA/K linear current output over the -50 to 150°C temperature range. Some applications and accuracy are discussed by Analog Devices.
Current output temperature transducers have a number of advantages:
• They are based on a linear relationship and are highly repeatable.
• The current is independent of voltage drops, voltage noise, common-mode voltage, and practically independent of excitation voltage.
• The current can be translated to a voltage at a remote destination via an appropriate value of resistance (V = IR); simple offsetting circuitry may be used when necessary.
They are easy to use; they require no linearization circuitry, high-precision voltage amplifiers, resistance-measuring circuitry, or cold-junction compensation.
Current output temperature sensors are widely used for cold-junction compensation of thermocouple circuitry.
When voltage drops and noise are not an important consideration, it may be more convenient to work with a voltage output temperature transducer. These provide a direct output to an analog-to-digital converter or a comparator set point.
Many practical components provide a voltage output as well as other functions.
4.4.2 AD-22100: A Ratiometric Voltage Output Temperature Sensor
The AD-22100 is a ratiometric temperature sensor IC whose output voltage is proportional to the power supply voltage. The heart of the sensor is a proprietary temperature-dependent resistor, similar to an RTD, built into the IC. FIG. 5(a) is a simplified block diagram of the AD-22100. The temperature-dependent resistor, RT, exhibits a change in resistance that is nearly linearly proportional to temperature. This resistor is excited with a current source proportional to the power supply voltage (V+). The resulting voltage across Rr therefore is both supply voltage proportional and linearly varying with the temperature (TA). The remainder of the AD-22100 consists of an opamp signal conditioning block that takes the voltage across Rr and supplies the proper gain and offset to achieve the following output voltage function:
Due to its ratiometric nature, the device offers a cost-effective solution when used as an interface to an analog-to-digital converter. This is accomplished by using the ADC's -t-5 V power supply as a reference to both the ADC and the AD-22100 (see FIG. 5(b)), eliminating the need for a precision reference.
The devices such as AD-22100 provide low-cost temperature measurement for microprocessor-and microcontroller-based systems. Many inexpensive 8-bit microprocessors now offer an onboard 8-bit ADC capability at a modest cost.
Total "cost of ownership" becomes a function of the voltage reference and analog signal conditioning necessary to mate the analog sensor with the microprocessor ADC. Such devices can provide a low-cost system by eliminating the need for a precision voltage reference and any additional active components. The ratiometric nature of the device allows the microprocessor to use the same power supply as its ADC reference. Variations in the supply voltage have little effect, as the sensor and the ADC use the supply as their reference.
4.5 Temperature Management ICs
Silicon temperature sensors easily can be combined with other circuit blocks for temperature control (Travis, 1996; Freeman, 1993). Overtemperature alarms, faulty circuitry shutdown, or initiation of corrective actions in a thermal feedback loop are ways in which temperature management ICs can prevent catastrophic failures. Devices commercially available include temperature controllers, airflow and temperature sensors, serial digital output, thermostat ICs, and programmable thermostat ICs. These devices are produced on a mass scale, using the standard IC production processes and prices vary from $0.50 to $4. Most of these ICs rely on a bandgap reference with a known temperature coefficient, coupled with other analog and digital circuitry, which may include the logic and ADCs as well. Two modern trends in temperature management ICs are increasing incorporation of digital circuitry and incorporation of more management functions (Travis, 1996). Temperature control ICs include a temperature sensor that generates a voltage output proportional to the absolute temperature and a control signal from one or two outputs when the device is below or above a specified temperature range. An example of these devices is TMP01 from Analog Devices.
The TMP01 consists of a bandgap voltage reference combined with a pair of matched comparators. The reference provides both a constant 2.5 V output and a voltage proportional to absolute temperature (VPTAT), which has a precise temperature coefficient of 5 mV/K and is 1.49 V (nominal) at +25°C. The comparators compare the VPTAT with the externally set temperature trip points and generate an open-collector output signal when one of these thresholds has been exceeded. FIG. 6(a) is a functional block diagram of the TMP01.
Hysteresis also is programmed by the external resistor chain and determined by the total current drawn out of the 2.5 V reference. This current is mirrored ( FIG. 6(b)) and used to generate a hysteresis offset voltage of the appropriate polarity after a comparator has been tripped. The comparators are connected in parallel, which guarantees no hysteresis overlap and eliminates erratic transitions between adjacent trip zones.
The device utilizes proprietary thin-film resistors in conjunction with production laser trimming to maintain a typical temperature accuracy of +2°C over the rated temperature range, with excellent linearity. The open-collector outputs are capable of sinking 20 mA, enabling the TMP01 to drive control relays directly.
The TMP01 is a very linear voltage-output temperature sensor, with a window comparator that can be programmed by the user to activate one of two open-collector outputs when a predetermined temperature set point voltage has been exceeded. A low drift voltage reference is available for set point programming (see FIG. 7). In many temperature sensing and control applications, some type of switching is required. The open collector outputs (over and under) of TMP01 can be used to turn on a heater or switch off a motor. In such applications, the switches need to handle large currents, usually much more than 20 mA, which is the rated current of the output. In such cases, external switching devices such as relays, power MOSFETs, thyristors, IGBTs, or Darlington transistors can be used, as shown in FIG. 8. For further details, see Analog Devices (1994).
4.6 Serial Digital Output Thermometers
Several manufacturers offer basic sensor devices coupled with analog-to-digital converters. These ICs allow a series digital output from the ADC proportional to the temperature. Examples are TMP03/04 from Analog Devices, LM 75 from National, and DS 1621 from Dallas Semiconductor.
The TMP03/TMP04 is a monolithic temperature detector generating a modulated serial digital output that varies in direct proportion to the temperature of the device. An onboard sensor generates a voltage precisely proportional to absolute temperature, which is compared to an internal voltage reference and entered into a precision digital modulator.
The sensor output is digitized by a first-order E-A modulator (see FIG. 9(a)). This type of converter utilizes time-domain oversampling and a high-accuracy comparator to deliver 12 bits of effective accuracy in an extremely compact circuit. FIG. 9(a) is a basic functional block diagram, and FIG. 9(b) describes the first-order modulator interacting with the VPTAT and the voltage reference source.
The modulated output of the comparator is encoded using a circuit technique that results in a serial digital signal with a mark-space ratio format easily decoded by any microprocessor into either Centigrade or Fahrenheit degrees and readily transmitted or modulated over a single wire. Most important, the encoding method neatly avoids major error sources common to other modulation techniques, as it’s clock independent.
184.108.40.206 Output Encoding
The TMP03/04 is designed as a low-cost three-terminal device with the output format shown in FIG. 9(c). This patented design avoids an accurate external clock or high accuracy, low-drift types of internal clock systems within the IC. The modulation and encoding techniques within the TMP03/04 achieve this by using a simple, compact onboard clock and an oversampling digitizer that are insensitive to sampling rate variations. The digitized signal is encoded into a ratiometric format in which the exact frequency of the clock is irrelevant, and the effects of clock variations are effectively canceled on decoding by the digital filter.
The output of the TMP03/TMP04 is a square wave with a nominal frequency of 35 Hz (-t-20%) at +25°C. The output format is readily decoded by the user as per FIG. 9(c):
Temperature (°C)= 235-400 x Tl T2 (6)
The time periods TI (high period) and T2 (low period) are values easily read by a microprocessor timer/counter port, with the preceding calculations performed in software. Since both periods are obtained consecutively using the same clock, performing the division indicated in these formulas results in a ratiometric value independent of the exact frequency of, or drift in, either the originating clock of the TMP03/TMP04 or the user's counting clock. FIG. 10 shows the output frequency and T1/T2 values versus temperature.
220.127.116.11 Application Considerations
These types of components are quite useful in applications such as isolated sensors, environmental control systems, computer thermal monitoring, thermal protection, and industrial process control and power system monitors. The low-voltage power supply (4.5-7 V) of these devices, low-cost three-pin package, low-power consumption, and the flexible open-collector output (TMP03) or CMOS/TTL-compatible output (TMP04) are the useful features of these devices for the wide variety of applications proposed.
Precision analog products such as the TMP series require a well-filtered power source. Since the TMP03/04 devices operate from a single +5 V supply, it’s convenient to use the logic supply. Unfortunately, the logic supply often is a switchmode design, which generates noise in the 20 kHz-1 MHz range. In addition, fast logic gates can generate glitches of hundreds to a few millivolts in amplitude due to wiring resistance and inductance.
To minimize the noise affecting the operation, the circuit arrangement in FIG. 1 l(a) is proposed. Even if a separate power supply trace is not available, however, generous supply bypassing will reduce supply-line-induced errors.
Local supply bypassing consisting of a 10 uF tantalum electrolytic in parallel with a 0.1 uF ceramic capacitor ( FIG. 11 (b)) is recommended. As the quiescent power supply current of the device typically is 900 uA, a simple RC filter network as in FIG. 11 (c) could be used when the device drives a light load, such as a CMOS gate.
The TMP03 ( FIG. 12(a)) has an open-collector NPN output suitable for driving a high-current load, such as an optoisolator. Since the output source current is set by the pull-up resistor, output capacitance should be minimized in TMP03 applications. Otherwise, unequal rise and fall times will skew the pulse width and introduce measurement errors.
The TMP04 has a "totem pole" CMOS output ( FIG. 12(b)) and provides a rail-to-rail output drive for logic interfaces. The rise and fall times of the TMP04 output are closely matched, to minimize errors caused by capacitive loading. If load capacitance is large ( for example, when driving a long cable), an external buffer may improve accuracy. For more details on output configurations and interfaces to low-voltage logic and the like, see Analog Devices (1995a).
18.104.22.168 Microcontroller and DSP Interfaces
Here is an example of an 80C51 interface. The TMP03/TMP04 output easily is decoded with a microcomputer. The microcomputer simply measures the T1 and T2 periods in software or hardware and calculates the temperature using equation (6). Since the TMP03/TMP04's output is ratiometric, precise control of the counting frequency is not required. The only timing requirements are that the clock frequency be high enough to provide the required measurement resolution and that the clock source be stable. The ratiometric output of the TMP03/TMP04 is an advantage because the microcomputer's crystal clock frequency often is dictated by the serial baud rate or other timing considerations.
Pulse width timing usually is done with the microcomputer's on-chip timer.
A typical example, using the 80C51, is shown in FIG. 13. This circuit requires only one input pin on the microcomputer, which highlights the efficiency of the TMP04's pulse width output format. Traditional serial input protocols, with data line, clock, and chip select, usually require three or more I/O pins.
The 80C51 has two 16-bit timers. The clock source for the timers is the crystal oscillator frequency divided by 12. Therefore, a crystal frequency of 12 MHz or greater will provide resolution of 1 us or less. The 80C51 timers are controlled by two dedicated registers. The TMOD register controls the timer mode of operation, while TCON controls the start and stop times. Both the TMOD and TCON registers must be set to start the timer.
Software for the interface is shown in Listing 1. The program monitors the TMP04 output and turns the counters on and off to measure the duty cycle.
The time that the output is high is measured by Timer 0, and the time that the output is low is measured by Timer 1. When the routine finishes, the results are available in special function registers 08AH through 08DH.
Listing 1. An 80C51 Software Routine for the TMP04. [Analog Devices, Inc.]
Test of a TMP04 interface to the 80C51, using timer 0 and timer 1 to measure the duty cycle This program has three steps:
1. Clear the timer registers, then wait for a low-to-high transition on input PI.0 (which is connected to the output of the TMP04).
2. When PI.0 goes high, timer 0 starts. The program then loops, testing PI.0.
3. When PI.0 goes low, timer 0 stops & timer 1 starts. The program loops until PI.0 goes low, when timer 1 stops and the TMP04's T1 and T2 values are stored in Special Function registers 8AH through 8DH (TL0 through THI).
When the READ_TMP04 routine is called, the counter registers are cleared.
The program sets the counters to their 16-bit mode, and then waits for the TMP04 output to go high. When the input port returns a logic high level, Timer 0 starts.
The timer continues to run while the program monitors the input port. When the TMP04 output goes low, Timer 0 stops and Timer 1 starts. Timer 1 runs until the TMP04 output goes high, at which time the TMP04 interface is complete.
When the subroutine ends, the timer values are stored in their respective SFRs and the TMP04's temperature can be calculated in software.
Since the 80C51 operates asynchronously to the TMP04, there is a delay between the TMP04 output transition and the start of the timer. This delay can vary between 0 Its and the execution time of the instruction that recognized the transition. The 80C5 l's "jump on port-bit" instructions (JB and JNB) require 24 clock cycles for execution. With a 12 MHz clock, this produces an uncertainty of 2 us (24 clock cycles/12 MHz) at each transition of the TMP04 output. The worst-case condition occurs when T1 is 4 us shorter than the actual value and T2 is 4 us longer. For a 25°C reading ("room temperature"), the nominal error caused by the 2 its delay is only about +0.5°C. The TMP04 also easily interacts with digital signal processors, such as the ADSP-210x series. Again, only a single I/O pin is required for the interface ( FIG. 14). The ADSP-2101 only has one counter, so the interface software differs somewhat from the 80C51 example. The lack of two counters is no limitation, however, because the DSP architecture provides very high execution speed. The ADSP-2101 executes one instruction for each clock cycle, versus one instruction for 12 clock cycles in the 80C51, so the ADSP-2101 actually produces a more-accurate conversion while using a lower oscillator frequency.
The timer of the ADSP-2101 is implemented as a down counter. When enabled by a software instruction, the counter is decremented at the clock rate divided by a programmable prescaler. Loading the value n-1 into the prescaler register will divide the crystal oscillator frequency by n.
For the circuit of FIG. 14, therefore, loading 4 into the prescaler will divide the 10 MHz crystal oscillator by 5 and thereby decrement the counter at a 2 MHz rate. The TMP04 output is ratiometric, of course, so the exact clock frequency is not important.
A typical software routine for an interface between the TMP04 and the ADSP-2101 is shown in Listing 2. The program begins by initializing the prescaler and loading the counter with 0FFF. The ADSP-2101 monitors the FI flag input to establish the falling edge of the TMP04 output and starts the counter.
When the TMP04 output goes high, the counter is stopped. The counter value is subtracted from 0FFFh to obtain the actual number of counts, and the count is saved. Then, the counter is reloaded and runs until the TMP04 output goes low.
Finally, the TMP04 pulse widths are converted to a temperature using the scale factor of equation (6).
22.214.171.124 Miscellaneous Other Applications
Sensors similar to TMP03/04 can be used for many other useful applications.
One such use is to monitor the temperature of a high-power microprocessor.
The TMP04 interface depicted in FIG. 15 could be used to measure the output pulse widths with a resolution of +1 us. The TMP04 sensors T1 and T2 periods are measured with two cascaded 74HC520 counters. The counters, accumulating clock pulses from a 1 MHz external oscillator, have a maximum period of 65 ms.
The circuit shown in FIG. 15 can be an ASIC application (as part of the system ASIC) so that the microprocessor would not be burdened with the overhead of timing the output pulse width. For details, see Analog Devices (1995a). Another example of using such an IC to monitor a high power dissipation ULSI is shown in Figure 16. The device, in a surface mounted package, is mounted directly beneath the device's pin grid array (PGA) package. In a typical application, the device's output could be connected to an ASIC, where the pulse width could be measured (FIG. 15 is a suitable interface.) The TMP04 pulse output provides a significant advantage in this application because it produces a linear temperature output while needing only one I/O pin and no A/D converter.
4.7 Precision Temperature Sensors and Airflow Temperature Sensors
4.7.1 Low-Voltage Precision Temperature Sensors Low-voltage temperature sensors which provide a voltage output directly proportional to the temperature for temperature monitoring and thermal control systems; For example, the TMP35, TMP36, and TMP37 from Analog Devices ( FIG. 17) or the LM35 and LM36 from National Semiconductor. These devices require no external calibration to provide a typical accuracy of +1°C at 25°C and +20°C over the -40 to 125°C temperature range. For application details, see Analog Devices (1996a).
4.7.2 Airflow Temperature Sensors
Modern electronic systems and products need be incorporated with suitable airflow temperature control systems. For such systems, such as low-cost fan controllers and overtemperature protection, commercial silicon sensors measure the airflow temperature. These devices consist of a bandgap element (with a voltage reference source and a VPTAT) and a heating element plus the associated circuitry. An example is the TMP12, an airflow and temperature sensor from Analog Devices ( FIG. 18).
The TMP12 incorporates a heating element, temperature sensor, and two user-selectable set point comparators on a single substrate. By generating a known amount of heat and using the set point comparators to monitor the resulting temperature rise, the TMP12 can indirectly monitor the performance of a system's cooling fan. The TMP12 temperature sensor section consists of a bandgap voltage reference that provides both a constant 2.5 V output and a voltage proportional to absolute temperature. The VPTAT has a precise temperature coefficient of 5 mV/K and is 1.49 V (nominal) at +25°C. The comparators compare the VPTAT with the externally set temperature trip points and generate an open-collector output signal when one of the respective thresholds has been exceeded. The heat source for the TMP12 is an on-chip 100 f2 thin-film resistor with a low temperature coefficient. When connected to a 5 V source, this resistor dissipates:
P_D= V^2 / R = 5^2 /100= --= 0.25 W (7)
… which generates a temperature rise of about 32°C in still air for the small outline SO packaged device. With an airflow of 450 feet per minute (FPM), the temperature rise is about 22°C. By selecting a temperature set point between these two values, the TMP 12 can provide a logic-level indication of problems in the cooling system.
A typical application for devices similar to TMP12 is shown in FIG. 19(a). The airflow sensor is placed in the same cooling airflow as a high-power dissipation IC. The sensor's internal resistor produces a temperature rise proportional to the airflow, as shown in FIG. 19(b). Any interruption in the airflow will produce an additional temperature rise. When the sensor's chip temperature exceeds a user-defined set point limit, the system controller can take corrective action, such as reducing clock frequency, shutting down unused peripherals, or turning on an additional fan. These devices have hysteresis profiles similar to discussions in previous sections. For further details, see Analog Devices (1995b).
4.8 Sensors with Built-in Memories
To store set points for temperature monitoring systems, some manufacturers have developed processes to embed memory devices inside the sensor ICs; For example, the DS 1621 from Dallas Semiconductor and LM75 from National Semiconductor. The DS 1621 provides 9 bit (serial) temperature data and user-settable thermostatic set points. As the user settings are nonvolatile, the devices can be programmed before insertion into the system. For details, see Dallas Semiconductor (1994-1995). One of the more unusual temperature management ICs is Dallas Semiconductor's DS 1820 with its digital thermometer ( FIG. 20). This device is a multidrop temperature sensor with 9 bit serial digital output. Information is sent to and from the DS 1820 over a single-wire interface, so only one wire (and ground) needs to be connected from a central microprocessor to a DS 1820. Power for reading, writing, and performing temperature conversions can be derived from the data line itself with no need for an external power source.
Because each DS 1820 contains a unique silicon serial number, multiple DS 1820s can exist on the same single-wire bus. This allows placing temperature sensors in many different places. Applications where this feature is useful include HVAC environmental controls; sensing temperatures inside buildings, equipment, or machinery; and process monitoring and control. The block diagram of FIG. 20(b) shows the major components of the DS 1820. The DS 1820 has three main data components: a 64 bit ROM, a temperature sensor, and nonvolatile temperature alarm triggers, TH and TL. The device derives its power from the one-wire communication line by storing energy on an internal capacitor during periods of time when the signal line is high and continues to operate off this power source during the low times of the one-wire line until it returns high to replenish the parasite (capacitor) supply. As an alternative, the DS 1820 can be powered from an external 5 V supply.
4.9 Thermal Response Time of Sensors
The time required for a temperature sensor to settle to a specified accuracy is a function of the thermal mass of and thermal conductivity between the sensor and the object being sensed. Thermal mass often is considered equivalent to capacitance. Thermal conductivity, commonly represented by the symbol 0, can be thought of as thermal resistance. It commonly is specified in units of degrees per watt of power transferred across the thermal joint.
The time required for the sensor IC to settle to the desired accuracy depends on the package selected, the thermal contact established in the particular application, and the equivalent power of the heat source. In most applications, the settling time probably is best determined empirically. Thermal time constants for thermal sensors can vary from a few seconds to over 100 seconds, depending on the package and socket used, air velocity, and other factors. Practical techniques for maximum accuracy from sensors are discussed in Steele (1996).
(cont. >> )