Fullscreen HTML5Med Res (??MB)HTML4

Circuit Surgery Regular clinic by Ian Bell The LM35 temperature sensor I n this month’s article we look at the LM35 – a widely used temperature sensor integrated circuit (IC) –see this month’s Flowcode article on p.49. Temperature is one of the most frequently measured physical quantities and is commonly used as an input to microcontroller unit (MCU) based systems (and embedded systems in general). Typically, a (contact) temperature sensor produces an analogue signal (voltage or current) which varies in a well-defined way with the temperature of the sensor. This can be converted to a digital representation by an analogue-to-digital converter (ADC), which may be built into the MCU, be part of a temperature-sensing IC, which communicates with the MCU digitally; or it may be a separate converter chip. The MCU reads the ADC (directly or indirectly) to obtain a binary number related to the temperature of the sensor. Usually, this number must be processed in some way to obtain the temperature value on a standard scale, such as degrees Celsius (°C). The complexity or processing required depends on the type of sensor used and the accuracy of measurement needed – one advantage of the LM35 is the straightforward relationship between its output and temperature being measured. Although processing is more complex if code is used to compensate for accuracy variation with temperature. Of course, analogue-output sensors can also be used in circuits which do not require an MCU or any code to be written – for example, analogue temperature switches and controllers. In an MCU system using an analogue sensor IC, such as the LM35, the ADC, which may be on the MCU chip or external, must be set up correctly. This includes setting the reference voltages which determine the ADC’s conversion range. The output voltage range from a typical temperature sensor may be smaller than the default input range of the ADC, so if the ADC range is not adjusted then the full ADC resolution will not be used. Correct ADC setup is also important when using other analogue temperature sensors. 46 Sensor choices In general, when designing a temperature measurement system, a decision has to be made on what type of sensor is to be used – there are several to choose from. A broad category is contact and non-contact sensors. The most common approach to non-contact temperature sensing measures infrared radiation and includes pyrometers and thermal imaging cameras, both of which measure the temperature of the surface of the object they are ‘looking at’. Contact temperature sensors include thermistors, thermocouples, resistance temperature detectors (RTDs) and IC temperature sensors. These devices are all fundamentally based on electrical properties of materials which vary with temperature in predictable ways that can be calibrated to provide a useful measurement. The devices vary in terms of their basic physics as well as the materials and fabrication processes used to build them. The advantage of ICs over the other devices listed is in their enhanced functionality and/or simplicity of use, but they cover a narrower range of temperatures than sensors such as thermocouples and cannot achieve the accuracy of the best RTD circuits. Temperature sensor ICs fall into two main categories – those that output an analogue signal (usually voltage) directly related to temperature and those with The LM35 sensor digital interfaces. Thetemperature latter are convenient in microcontroller systems with relatively low accuracy requirements, but will require the user to manage the digital interface (such as an SPI or I2C bus). developing a sensor (like the LM35) is to obtain a linear, and well-controlled dependence between temperature and device output over a wide temperature range. An ordinary diode (PN junction) can be used as a temperature sensor, since its forward voltage changes by approximately –2mV/°C. Improved accuracy can be obtained by using two diodes (or transistor base-emitter junctions) – the voltage difference between two PN junctions, operated at different current densities, varies linearly with absolute temperature. This temperature sensitivity has been exploited for many temperature sensor ICs, such as the LM34, LM35, LM50, LM60, LM61, MCP9700, MCP9701, TMP35, TMP36, TMP37 and TSIC301. Other approaches are also used. For example, the AD22100 and AD22103 use a temperature-dependent resistor. The LMx35 series (x is 1,2,3) are like zener diodes with breakdown voltages directly proportional to absolute temperature at 10mV/K. The LM135’s range is from −55°C to 150°C. There is also the AD950, which is a two-terminal device. This has a linear current output of 1µA/K (again, a kelvin scale) with a supply of 4V to 30V over a measurement range of −55°C to +150°C. How the LM35 Works We can relate the emitter current (IE) of a bipolar transistor to its base-emitter voltage (VBE) using the basic form of the Ebers-Moll equation: 𝐼𝐼! = 𝐼𝐼" exp & 𝑞𝑞𝑞𝑞#! + 𝑘𝑘𝑘𝑘 Here, IS is the reverse saturation current of the base-emitter 𝑘𝑘𝑘𝑘 𝐼𝐼! PN junction, which is ln & +individual transistor; q #! = a 𝑞𝑞 property Many IC temperature sensors, like the 𝑞𝑞 of the 𝐼𝐼" is the electronic charge (the charge on one LM35, are based on the temperature-deelectron, which is a physical constant); k pendent behaviour of semiconductors, is𝑘𝑘𝑘𝑘 Boltzmann’s constant specifically theLM35 PN-junction found in The temperature sensor 𝐼𝐼!$ 𝑘𝑘𝑘𝑘 𝐼𝐼!%(another physical 𝛥𝛥𝑞𝑞#! = ln & and + − T islnthe & absolute + constant); temperadiodes and bipolar transistors. Circuit de𝑞𝑞 𝐼𝐼" 𝑞𝑞 𝐼𝐼" ture (in kelvin). The equation applies for signers often spend a lot of effort trying 𝑞𝑞𝑞𝑞#! relatively large currents (IE >> IS). to overcome the effects of temperature on 𝐼𝐼! = 𝐼𝐼" exp & emitter + 𝑘𝑘𝑘𝑘 𝐼𝐼 𝐼𝐼 𝑘𝑘𝑘𝑘 be rearranged 𝐼𝐼!$ transistor circuits (see the December 2021 This!$ equation to give !% 𝑘𝑘𝑘𝑘 can 𝛥𝛥𝑞𝑞#! = 𝑙𝑙𝑙𝑙 & 2 + = 𝑙𝑙𝑙𝑙 & + and January 2022 Circuitry Surgery articles the base-emitter voltage: 𝑞𝑞 𝐼𝐼" 𝐼𝐼" 𝑞𝑞 𝐼𝐼!% on logarithmic and exponential amplifi𝑘𝑘𝑘𝑘 𝐼𝐼! ers) – temperature-dependent signals are 𝑞𝑞#! = ln & + 𝑞𝑞𝑘𝑘𝑘𝑘 𝐼𝐼" not hard to obtain, but the challenge when 𝛥𝛥𝑞𝑞#! = 𝑙𝑙𝑙𝑙(𝑁𝑁) 𝑞𝑞 Practical Electronics | April | 2022 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝑘𝑘𝑘𝑘 𝐼𝐼!% 𝛥𝛥𝑞𝑞#! = ln & + − ln & + 𝑞𝑞𝛥𝛥𝑞𝑞 = 𝐼𝐼" 𝑞𝑞 𝑞𝑞 𝐼𝐼" #! &'(' Principles of operation 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝐼𝐼!% 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝛥𝛥𝑞𝑞#! = 𝛥𝛥𝑞𝑞#! = I I I/2 I/2 Q1a 2E E Q1b Q2 ∆VBE IE 0V Fig.1. Concept circuit for obtaining VBE from two transistors with scaled sizes. In which ln() is the natural logarithm 35 temperature sensor function, the inverse of the exponential function. If we have two transistors 𝑞𝑞𝑞𝑞#!currents (IE1 and with different emitter 𝐼𝐼! = 𝐼𝐼" exp & + I ) and we assume the transistors are 𝑘𝑘𝑘𝑘 E2 M35 temperature sensor at the same temperature (T) and closely matched in terms of their physical 𝑘𝑘𝑘𝑘 𝐼𝐼!= I = I ) then the characteristics (solnIS1 S2 S 𝑞𝑞#! = &&𝑞𝑞𝑞𝑞 +#! = 𝐼𝐼𝑞𝑞" exp 𝐼𝐼"𝑘𝑘𝑘𝑘 +base-emitter difference𝐼𝐼!between their voltages ( VBE) is: 𝛥𝛥𝑞𝑞#! = 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝑘𝑘𝑘𝑘 𝐼𝐼!% ln &= 𝑘𝑘𝑘𝑘 + −ln &𝐼𝐼!ln+ & + 𝑞𝑞 #! 𝑞𝑞 𝐼𝐼" 𝑞𝑞 𝑞𝑞 𝐼𝐼 𝐼𝐼" " We are subtracting two logarithms, which 𝑘𝑘𝑘𝑘 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝐼𝐼𝐼𝐼!% 𝑘𝑘𝑘𝑘 !$ 𝐼𝐼𝐼𝐼!% !$+ = 𝑘𝑘𝑘𝑘 𝑙𝑙𝑙𝑙 is#!equivalent 𝛥𝛥𝑞𝑞 = #! = 𝑙𝑙𝑙𝑙 & lnto 2& dividing 𝛥𝛥𝑞𝑞 + − 𝑞𝑞 lnby &&𝐼𝐼 the ++ sub𝑞𝑞 𝐼𝐼 𝐼𝐼 𝑞𝑞 "inside 𝐼𝐼"" the𝑞𝑞logarithm. 𝐼𝐼!% tracted value So, " we can write: 𝑘𝑘𝑘𝑘 𝐼𝐼!% 𝑘𝑘𝑘𝑘 =𝐼𝐼!$ 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝛥𝛥𝑞𝑞#! = 𝛥𝛥𝑞𝑞#! 𝑙𝑙𝑙𝑙 & 𝑞𝑞 2𝑙𝑙𝑙𝑙(𝑁𝑁) += 𝑙𝑙𝑙𝑙 & + 𝐼𝐼" 𝐼𝐼" 𝐼𝐼!% 𝑞𝑞 𝑞𝑞 The transistor matching can be achieved 𝛥𝛥𝑞𝑞#! = 𝑞𝑞𝑘𝑘𝑘𝑘 &'(' with correct𝛥𝛥𝑞𝑞 layout and fabrication of tran𝑙𝑙𝑙𝑙(𝑁𝑁) #! = sistors on the same𝑞𝑞IC (it is much more difficult with discrete transistors). The result is the difference in VBE only depends 𝛥𝛥𝑞𝑞 = 𝑞𝑞&'(' on temperature#!and the emitter current, not on the transistor characteristic (IS). 35 temperature sensor This is important because although two transistors can be matched on a single IC, 𝑞𝑞𝑞𝑞#!will have transisdifferent𝐼𝐼individual ICs + ! = 𝐼𝐼" exp & 𝑘𝑘𝑘𝑘 tors with different IS. Furthermore, IS varies significantly with temperature, resulting in complex and variable temperature de𝑘𝑘𝑘𝑘 𝐼𝐼! pendence𝑞𝑞#! for=transistors ln & +across multiple 𝑞𝑞 𝐼𝐼" possible to make ICs. This means it is not an accurate temperature sensor based on a single transistor, but using 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝑘𝑘𝑘𝑘 the 𝐼𝐼!%difference between IC is 𝛥𝛥𝑞𝑞#! = twolntransistors & + − on ln &a single + 𝑞𝑞 𝐼𝐼" 𝑞𝑞 𝐼𝐼" a viable approach. PTAT𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝐼𝐼!% 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝛥𝛥𝑞𝑞 & 2 currents += 𝑙𝑙𝑙𝑙 If#! the= two𝑙𝑙𝑙𝑙emitter are& set+ to be 𝑞𝑞 𝐼𝐼" 𝐼𝐼" 𝑞𝑞 𝐼𝐼 in a fixed ratio (N), that is I E2!%= NI E1, then we get: 𝛥𝛥𝑞𝑞#! = 𝑘𝑘𝑘𝑘 𝑙𝑙𝑙𝑙(𝑁𝑁) 𝑞𝑞 In which ln(N), k and q are constant, so VBE is proportional to absolute temper𝛥𝛥𝑞𝑞#! = 𝑞𝑞&'(' ature with a positive relationship ( VBE Practical Electronics | April | 2022 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝑘𝑘𝑘𝑘 𝐼𝐼!% ln & + − ln & + 𝑞𝑞 𝐼𝐼" 𝑞𝑞 𝐼𝐼" 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝐼𝐼!% 𝑘𝑘𝑘𝑘 𝐼𝐼!$ 𝑙𝑙𝑙𝑙 & 2 + = 𝑙𝑙𝑙𝑙 & + 𝑞𝑞 𝐼𝐼" 𝐼𝐼" 𝑞𝑞 𝐼𝐼!% increases with increasing temperature). This is 𝑘𝑘𝑘𝑘 referred to as a ‘Voltage 𝛥𝛥𝑞𝑞 𝑙𝑙𝑙𝑙(𝑁𝑁) #! = 𝑞𝑞 Proportional to Absolute Temperature’, or VPTAT: 𝛥𝛥𝑞𝑞#! = 𝑞𝑞&'(' +VS A1 + R1 Q1 Q2 A2 – VO UT = 10mV/°C V PTAT If we have N = 10 (as 0.125R2 used in the LM35) VPTAT 10E E nR1 in the above equation is 8.8mV/°C 198µV/°C , which can be R2 scaled to give 10mV/°C or I other convenient ‘round number’ values (eg, for the Fahrenheit scale). 0V Setting up two indepenFig.2. LM35 simplified device schematic (based on Texas dent currents of ratio N (as Instruments data sheet). might be implied by the above discussion) is not the most practithe voltage across it will be n VBE. As an cal way to create a temperature sensor. It end of nR1 is grounded, the voltage at the is easier to scale the effective size of the base of Q1 will be a PTAT voltage related two transistors and ensure exactly the to the VBE difference. The output of the same current passes through both of them. LM35 is derived from the voltage at the Scaled transistors can be implemented current source, which is two diode voltage on ICs using transistors with multiple drops below Q1’s base voltage (these also emitters, which is equivalent to multiple contribute to the final output temperature parallel (ideally) identical transistors. coefficient). This voltage is scaled by the voltage amplifier A2, whose gain is set to provide a 10mV/°C variation with temSensor circuits perature at the output. An example of this principle is shown in The circuit in Fig.2 can be calibrated Fig.1. Q1 has two emitters and is representafter fabrication by adjusting nR1 using ed as two identical parallel devices (Q1a and Q1b) in Fig.1. Q2 is a single transistor various trimming techniques, for examidentical to Q1a and Q1b. The circuit is ple, fuse links across the resistors in a set up so that the same current (I) flows series string can be blown to increase through Q1 and Q2, which means that the resistance. The calibration sets the the currents in Q1a and Q1b are both I/2 correct temperature coefficient of the (the currents split equally between Q1a output voltage. and Q1b because they are identical). The For a more detailed discussion of the total current is kept constant and well operation of these devices, including controlled by the current source IE – the calibration and nonlinearity compensation (not covered here) refer to LM34/ equality of currents in Q1 and Q2 is enLM35 Precision Monolithic Temperature sured by circuitry not shown in Fig.1. Sensors, National Semiconductor AppliThe VBE voltages of Q1a and Q1b must cation Note 460, October 1986. be equal due to their parallel connection and each transistor has a current of I/2 with respect to Q2. Thus, the difference The LM35 IC and variants in base-emitter voltages is given by the The LM35 is a three-terminal device which VBE equation above, with N = 2, and has two power pins and an output pin producing a voltage which, as described so has a linear dependence on temperaabove, varies linearly with temperature ture (PTAT). at 10mV/°C. There are a large range of The circuit in Fig.2 is a simplified similar devices, some with similar part schematic of the LM35. The core of the numbers such as those listed above. These circuit, Q1 and Q2 and the current source, devices vary in various ways, such as the is configured as in Fig.1, expect that there temperature scale, supply voltage range, is a ten-times rather than two-times relaand presence, or otherwise, of a powertionship between the transistors (10E and saving shutdown pin. Most devices are E on Fig.2). The differential current-in, aimed at the Celsius scale, but the LM34 voltage-out amplifier A1 is in a feedback provides a 10mV/°F output. loop controlling the difference between Even with a given basic part number, the base-emitter voltages ( VBE) so that such as ‘LM35’, there are quite a few the collector currents of Q1 and Q2 (A1’s variants with different packaging (such inputs) are forced to be equal. as TO92, TO220, TO46, SOT and SOIC), The VBE voltage is across R1, so it will different accuracy ratings and different cause a current of VBE/R1 to flow through temperature ranges – not all devices proit. Assuming this is much larger than Q1’s vide the full datasheet headline range. This base current (which can be ignored) the diversity means that care must be taken same current flows in the resistor nR1, so 47 which can measure negative temperatures on a single VO UT VO UT + LM35 LM35 LM35 supply. Examples VO UT include the TMP36 0V – 0V from Analog Devices R1 1N914 R1 and the LM50 from –VS Texas Instruments. These devices have 0V a 10mV/°C output, like the LM35, but Left to right (all based on Texas Instruments datasheet): with a +0.5V offset Fig.3. Basic LM35 circuit; Fig.4. Full range LM35 circuit; and (so output 0.5V at Fig.5. LM35 range extension on a single supply. 0°C, 0.75V at 25°C and 0.25V at –25°C) which facilitates when ordering parts. Some variants are output of voltages representing negasignificantly more expensive than others. tive temperatures. An example circuit Using the LM35 and similar devices for the TMP36 is shown in Fig.6. Note the shutdown pin which is available on The most basic LM35 circuit is shown in some package options. Fig.3. This provides a 10 mV/°C output over the range 2°C to 150°C (20mV to 1.5V out) – notice that the lower end of the Long leads LM35’s temperature range is not covered It is not uncommon to need to locate a tembecause the output of the LM35 cannot perature sensor away from the main circuit go negative in this circuit. The solution board. This may make the sensor wiring is to use a negative supply in the system susceptible to noise pick-up from various and wire the LM35 as shown in Fig.4. The sources. Fortunately, in most situations, LM35 can only source current (current temperatures change relatively slowly and flows out, not in), but with the resistor appropriate low-pass filtering can reduce (R1) connected to a negative supply, a noise without excessively impacting the measurement process. As usual, shielded sourced current can result in an output and/or twisted pair cable should also be voltage of either polarity with respect to used for lengthy sensor connections to ground. The value of R1 is specified in reduce the amount of noise pickup. the LM35 datasheet as VS/50µA. Another issue with long sensor connecIf a negative supply is not available, the tions, which should not be overlooked, is circuit shown in Fig.5 can be used. The the capacitance of the cable. A common diodes raise the voltage at the ground pin solution is to use an isolation resistor beof the LM35 above 0V (system ground) tween the output and capacitive load. The so the system ground is like a negative capacitive drive capability varies signifisupply from the perspective of the LM35. cantly for different integrated analogue The approach used in Fig.4 can then be temperature sensors, but in general, the applied. It may be tempting to assume series resistor approach is applicable (see the diodes drop about 0.6V or 0.7V (the Fig.7) and sensor datasheets may provide typical assumption), but the LM35 may advice on appropriate values (this is the only be consuming 60µA, and at these case for the LM35 and TMP36). Alternacurrent levels the 1N914 forward drop is tively, an RC damper may be used (see in the range 0.45V to 0.5V. If we assume Fig.8). The series resistor, plus cable ca0.9V total, then the rule for R1 in Fig.4 pacitance, or a damper circuit, also forms gives the datasheet value for R1 of 18kΩ. a low-pass filter assisting reduction noise. The circuit in Fig.5 differs from the one As with many integrated circuits, it is in Fig.4 in that the output voltage is not often a good idea, or even essential, to referenced to ground. Various approaches connect one or more supply decoupling to dealing with this are possible. Just the capacitors as close as possible to the sensor normal output pin of the LM35 could be IC. A typical value is 0.1µF, but the dataused, on the assumption that the diode sheet should be consulted for the most drop is constant, subtracting this to get the appropriate values. The circuits in Fig.7 temperature voltage, but this is very likely and 8 show typical circuits for using temto be inaccurate as a diode drop value will perature sensors on long cables. In more change (eg, with temperature and LM35 extreme situations (very long cables, high supply current). Better approaches are electrical noise) converting the sensor voltto use the differential output (as shown) age to a current for transmission down the with a differential input amplifier, a difcable may be a better approach. ferential input ADC, or to measure both voltages separately (eg, using multiple ADC channels) and subtract in software. Layout considerations The circuit in Fig.5 is not very conThere is more to temperature sensing than venient, so a better approach with a the circuit design – the physical strucsingle supply is to use a different chip, ture is also important to ensure that the +VS 48 +VS 0.1µF +VS +VS SHUTDOWN 0V +VS VO UT TMP36 GND 0V Fig.6. TMP36 circuit with −40°C to +125°C range (based on Analog Devices datasheet). right temperature is being measured and device operation is not undermined. The leads of the device conduct heat to the chip inside the package, so in situations where a device is glued to the surface of an item whose surface temperature is being measured, the air temperature surrounding the leads may have an unwanted influence on the reading – particularly for plastic package devices. The LM35 datasheet advises covering the leads with epoxy if this might be a problem (if the surface and air temperatures may be different). Metal package devices are less problematical in this respect and can be soldered to metal surfaces being measured, but only if the metal to which it is attached can be at the device negative supply lead voltage. If the device is used to measure liquid temperature, then arrangements must be made to keep it and its wiring dry and electrically insulated from the liquid. Moisture may also be an issue where cold temperatures are being measured and condensation may occur on the device or surrounding circuitry – suitable coatings can help in these situations. +VS C1 Sensor VO UT R1 Long cable 0V Fig.7. Analogue integrated temperature sensor on long cable with decoupling capacitor C1 and load capacitance isolation resistor R1. +VS C1 Sensor VO UT R1 Long cable C2 0V Fig.8. Analogue integrated temperature sensor on long cable with decoupling capacitor C1 and RC damper (R1 and C2). Practical Electronics | April | 2022