Advertisement

Blog

Temperature Measurement, Part 4

I am finally getting to the topic of integrated temperature sensors. Before we get to actual measurements, for completeness I thought I should mention some devices that monitor temperature but act as limit detectors with digital outputs to indicate when the temperature has transgressed. Most have programmable set points that can be fixed by resistors or through communication with a host micro.

An example of the former would be the TMP01 from Analog Devices, or even the system management device from Linear Technology, the LT2995, which includes voltage monitors. System management devices like LTC2974 interface through an I2C bus and will monitor four external temperatures.

There is a unique device from Analog Devices, the TMP05, which has a ratiometric PWM output proportional to the temperature. The unique feature is that they can be chained together requiring only two microcomputer I/O pins for a large number of temperature sensors. As he points out, the distance between the sensors can be increased with the use of line driver/receivers and I would add that you could isolate as well.

You would think that when it came to semiconductor temperature sensors you only had five parameters to play with: package, accuracy, temperature range, internal peripherals, and serial interface. There are a host of products like the TC77 from Microchip with a serial number and SPI interface, the TMP175 from TI with I2C interface or DS18S20 from Maxim with a 1-wire interface and programmable trip points. But there are another two factors to consider.

The first is how you are going to thermally connect the sensor to the object whose temperature you are trying to read. True, this is something to consider with all sensors, but the IC can need additional connections and normally needs a PCB. Sure, flying leads are possible, but that has to be part of your manufacturing process and subject to frequent failure. The second is a sort of related issue. Where is the temperature actually being measured? The DS18S20 actually measures the temperature of the ground lead so that any conducted heat on the PCB will affect the temperature readings.

Figure 1:This temperature controller uses the DS18S20. The current limiting resistors of the LED displays you see on the front panel warmed the PCB and affected the temperature readings of the ambient air flow through the module. We were forced to mount the sensor off-board.

Figure 1:This temperature controller uses the DS18S20. The current limiting resistors of the LED displays you see on the front panel warmed the PCB and affected the temperature readings of the ambient air flow through the module. We were forced to mount the sensor off-board.

Some functions are built around their need to know the temperature in order to derive another measurement. Besides thermocouples, another such function is relative humidity (RH). When I had a project that needed to measure RH, specialized ICs were not available, but now there seem to be a few, all of which include a temperature sensor. Jon Titus did a review of the Honeywell offering. Measurement Specialities produces a similar device with a digital interface, the HTU21D, and a purely analog version, the HTG3500.

There are also sensor digitizers that take a sensor input and convert it to a digital (serial data) format. The MAX31865 is one such device for RTDs and the MAX31855 is the matching part for thermocouples, including cold junction compensation.

Temperature Measurement 5 will have some conclusions and “gotchas.”

49 comments on “Temperature Measurement, Part 4

  1. antedeluvian
    April 28, 2014

    This TI app note “AN-1952 Designing With Thermocouples” has just come to my attention

     http://www.ti.com/lit/an/snoa544a/snoa544a.pdf

     

    and this article in Electronic Design “Improve Noise Immunity On RTD Ratiometric Measurements”

    http://electronicdesign.com/analog/improve-noise-immunity-rtd-ratiometric-measurements

    There are some other interesting references in the article

  2. antedeluvian
    April 28, 2014

    An interesting tempearture sensor that can be calibrated is the LM335 range from TI

     

     

  3. antedeluvian
    April 28, 2014

    TI (and others) offer a temperature sensor in degrees C (LM35) or degrees F (LM34).

    One of the shortcomings with the LM35 is measuring temperatures below 2C. If you look at Figures 1 and 2 of the data sheet you can see that you need a negative voltage and resistor to go below this temperature. (http://www.ti.com/lit/ds/symlink/lm35.pdf)
     

    When I was faced with this problem, I solved it with the LM34 which is the same device only in Fahrenheit. Figure 1/2 show that the basic configuration will work from 5F (-15C) and has double the sensistivity 10 mV/F (http://www.ti.com/lit/ds/symlink/lm34.pdf)

  4. etnapowers
    April 28, 2014

    ”  LTC2974 interface through an I2C bus and will monitor four external temperatures.”

     

    The interface is a complex system that is very powerful as signal processor, but quite complex to be implemented and to detect a I2C codification protocol malfunctioning

  5. antedeluvian
    April 28, 2014

    etnapowers

    The interface is a complex system that is very powerful as signal processor, but quite complex to be implemented and to detect a I2C codification protocol malfunctioning

    Thanks for the heads up. Forewarned is forearmed.

     

  6. etnapowers
    April 28, 2014

    Nice post, thank you, it seems that LM34 is a updated LM35, good to know the advantages of utilizing LM35 in all the cases  LM34 has troubles.

  7. amrutah
    April 28, 2014

    Aubrey: Thanks for teh nice post on integrated Temperature sensors.

       Since most of these temp-sensors have ADC  and sensors, what about the parameters like dynamic response time and conversion time.  Usually the temperature does not change abruptly, but still how important are these parameters.

  8. antedeluvian
    April 28, 2014

    amrutah

    Since most of these temp-sensors have ADC  and sensors, what about the parameters like dynamic response time and conversion time.

    I may have given the impression that most of the temperature sensors have the ADC. I must admit I have no knowledge of the relative numbers, both in type and in units sold.

    But you ask a good question. If the device is measuring the airflow around it then normally its size determines how fast it will respond since the unit has to warm up and stabilize. In many, if not most applications, though, you are (at least  I am) measuring the temperature of something else, like a heat sink, a volume of water etc and the response is more likely to be determined by the target's thermal properties. In all of these I would bet that the conversion speed of the ADCs is several times faster that the rate of change of temperature and are not a significant consideration in a design.

     

    But you need to consider your application carefully. Every one has its own quirks.

  9. amrutah
    April 28, 2014

    Yes, the IC's that we design have a system around them and we monitor the temperature far off-the chip using thermistors, some need to respond quickly.

      Say the temperature at which we need to trigger is 85degC, then then system requirement is to shut off few parts with tens of micro seconds.  The conversion time, interrupt, processor acknoledgement and shut down has to be quick.

  10. amrutah
    April 28, 2014

    @Aubrey: “But you need to consider your application carefully. “

      Yes, I totally agree here.  If its a space application , then the temperature varies and the action has to be followed very quickly.

     

  11. samicksha
    April 29, 2014

    I agree you etnapowers, we can monitoring of four output voltages with external temperature and input voltage but where do you it complex during implementation.

  12. SunitaT
    April 29, 2014

    This is quite some piece of informative information on the parameters to play with when it comes to semiconductor sensors.  Semiconductor sensors are everywhere for example automotive systems are dependant on them for engine control, they have also been designed into heating and ventilation systems. Accuracy being one of the parameters to be considered ,most modern semiconductor sensors have been designed to be  able to give high accuracy but the question would be which among the different types of semiconductors offers high accuracy.

  13. amrutah
    April 29, 2014

    @SunitaT0: “which among the different types of semiconductors offers high accuracy.”

      This is a good question.  As I understand, for monitoring the temperature below 125deg C silicon is good enough, for higher temperatures (about 250deg C) we may have to look beyond silicon like GaN substrates.

       Any idea about GaAs substrate devices? MEMS transducers?

  14. SunitaT
    April 30, 2014

    This will be very important simply because we are talking about the question of integrating temperature. This will increase the scale of the temperature gauge as it will help us read the very low readings and the very high readings as well. Monitoring temperature has always been tricky especially when the temperature readings are on the extremes but this will definitely be another parameter to cub this problem. Semi conductor sensors should also be put as they help in the functioning of the temperature gauges.

  15. RedDerek
    April 30, 2014

    In general, I see that the semiconductor temp sensors have two main limitations.

    1. Limited temperature range – typically that of the silicon and packaging such as -55C to +150C.

    2. Long time constant – due to packaging methods.

    Thus, the semiconductor temp sensor would be applicable, to me, more for human environmental temperatures and not be concerned about a quick response. Beyond this realm, one would then revert back to the classic thermocouple, RTD, etc method; again watching for heat generation in the sensor due to measurement methods.

  16. antedeluvian
    April 30, 2014

    Derek

    In general, I see that the semiconductor temp sensors have two main limitations.

    1. Limited temperature range – typically that of the silicon and packaging such as -55C to +150C.

    2. Long time constant – due to packaging methods .

    I would like to add a third. Difficulty in attaching to the package to the object whose temperature you want to measure.

     

  17. RedDerek
    April 30, 2014

    @antedeluvian – Third point is well taken, yes.

  18. yalanand
    April 30, 2014

    This is a very nice piece of article, thank you Aubrey for this great information on temperature sensors. I am greatly impressed with TMP01 and LT2995; in that they can they have programmable set points which are possibly fixed by resistors. Yes, they can monitor temperature as well as detecting temperature; my question is, what are the limits of temperature which suitable for these two devices? I am certainly sure that they don't operate under all temperature conditions.

  19. antedeluvian
    May 1, 2014

    yalanand

     I am greatly impressed with TMP01 and LT2995; … what are the limits of temperature which suitable for these two devices?

    There is a bit of a difference between these two. The TMP01 is the temperature sensor itself and so it will operate from -55 to 125 degC.

    However your question as it pertains to the LT2995 has two parts. The LT2995 measures 2 temperatures, internal and external. The external measurement uses an external diode to sense the temperature and according to the spec this will work from about -40 to 140degC. Note that this is extyernal to the LT2995 and so there can be quite a difference in temperature between the didode an the IC itself.  The LT2995 itself will operate and measure its own internal temperature from -40 to 125 degC in the “H” version. The commercial “C” version is only spec'd to operate (and hence only measure temperature internally) from 0 to 70degC and the industrial “I” version stretches this to -40 to 85degC

     

  20. fasmicro
    May 3, 2014

    For temperature measurement, I have relied on Analog Devices especially the TMP series because for one reason they know how to write good datasheets. From formatting to what is right inside there, ADI is in another class.

  21. fasmicro
    May 3, 2014

    >> An interesting tempearture sensor that can be calibrated is the LM335 range from TI

    If you have to calibrate a temp sensor, it means you cannot use same in consumer electronics where calibration increases your cost model. Calibration in these commoditized components is never a good design strategy

  22. fasmicro
    May 3, 2014

    >> When I was faced with this problem, I solved it with the LM34 which is the same device only in Fahrenheit.

    It is very unfortunate to have a datasheet reporting in F as though we are talking about weather on CNN. Celsius is the default standard in the industry but America will never like to follow anything it does not invent. It is like whether, we report it in our own way. But engineers must do away with these datasheets that report in Fahrenheit. It makes no sense

  23. fasmicro
    May 3, 2014

    >> Since most of these temp-sensors have ADC  and sensors, what about the parameters like dynamic response time and conversion time.  Usually the temperature does not change abruptly, but still how important are these parameters.

    I think we need a very simple technique like bimetallic strip to measure temperature especially in consume electronics where you want to keep complexity low. For you to make a chip of 5cents, you need to innovate more than when you have to supply the automotive industry at $5. The point is that temp sensor could be weakened by the ADCs and other sub-units that make up the product.

  24. fasmicro
    May 3, 2014

    >> In all of these I would bet that the conversion speed of the ADCs is several times faster that the rate of change of temperature and are not a significant consideration in a design.

    That is actually how it works. The conversion rate of ADC is many times faster than the mechanical systems you employ to measure the temp. No question about that.

  25. fasmicro
    May 3, 2014

    >> The conversion time, interrupt, processor acknoledgement and shut down has to be quick.

    There is a big difference between the ADC and the processor. The ADC can still deliver the data on time and the processor may not be ready. Now you have the processor externally, many bad things can happen. Yet, I do not think that a well written code in the processor that has decent speed will be slow in any normal application. You do not have to grab data every msec for temp measurement. That is not necessary.

  26. fasmicro
    May 3, 2014

    >>  Yes, I totally agree here.  If its a space application , then the temperature varies and the action has to be followed very quickly.

    That would also mean you will use a different type of processors with specific types of memory that can be used in the space. In other words, it has to be rad-hard for space application. Even with that, a fairly decent processor can process temp measurement because new data acquisition does not have to be overly fast.

  27. goafrit2
    May 3, 2014

    Applications in space is another discussion. You cannot even use the same process to start with as space designs have unique process technologies. That is where the differences start. It will be wrong to shop on the web parts that will be used for space applications. I am not sure you will pass qualification for your design.

  28. etnapowers
    May 5, 2014

    @samicksha: the more complex a circuit is the more probable is a failure. Simplicity is the key.

  29. amrutah
    May 6, 2014

    @fasmicro: “The point is that temp sensor could be weakened by the ADCs and other sub-units that make up the product.”

        If we are to sense the temperature and want the to display the current temperature of the the device/product, then there is either a analog or digital way.  If the accruacy is not a matter of concern then we can do it with the analog temperature dial for display, but if we want to display the temperature, say to 2-decimal accuracy then we have to convert the sensed data to binary bits.

  30. amrutah
    May 6, 2014

    @fasmicro:

       I agree to most of your points here.

    1> The ADC and processor are different

    2> ADC might generate a data for the processor is sleep condition

    3> Processor need not monitor/measure the temperature every msec.

         But all that is important and necessary is if the temperature crosses a certain critical limit, the response time should be quick enough.  May be after reaching the critical limit the ADC pull down one of the pins of the processor instead of sending a code.

  31. fasmicro
    June 3, 2014

    >> @samicksha: the more complex a circuit is the more probable is a failure. Simplicity is the key.

    It is not just in circuits, I think in general technology product development. Simplicity is the new mantra

  32. fasmicro
    June 3, 2014

    >> If we are to sense the temperature and want the to display the current temperature of the the device/product, then there is either a analog or digital way. 

    It is always a problem when a device that measures a system is expected to be part of that system. The best is to sequest the two units.

  33. fasmicro
    June 3, 2014

    >>      But all that is important and necessary is if the temperature crosses a certain critical limit, the response time should be quick enough.

    You are fine then. Most times, the ADC is always faster than the system. Nevertheless, I try not to depend on ADCs for my temp measurement since it can also fail.

  34. etnapowers
    June 5, 2014

    I totally agree.

    Henry Ford often told his engineers that the only part of a car that cannot fail is the part that does not exist.

  35. etnapowers
    June 10, 2014
    @Aubrey, thanks for the link, I found interesting the meaning of the phrase that you cited: Advance warning provides an advantage.
    It's true, expecially when talking of technology.
  36. amrutah
    June 19, 2014

    Fasmicro: “The best is to sequest the two units.”

      Yes, and this is exactly most of the SOC now-a-days do.  They have the tempsense and ADC on the same device.  It is the microprocessor that decides when it want to read and display a temperture.

     

  37. fasmicro
    July 3, 2014

    >> Henry Ford often told his engineers that the only part of a car that cannot fail is the part that does not exist.

    That is the core of minimal design which companies like Nest demonstrated by making systems so easy that anyone can install them. It is creeping into the industry as I see fewer guys asking for IT guys to help when they receive new laptops at work.

  38. fasmicro
    July 3, 2014

    >>   Yes, and this is exactly most of the SOC now-a-days do.  They have the tempsense and ADC on the same device.  It is the microprocessor that decides when it want to read and display a temperture.

    Great strategy. I wish we can get to asynchronous ASIC so that we can do processing in “analog” without the need of clock. Hopefully, the production system can support such a radical design path.

  39. etnapowers
    July 7, 2014

    @fasmicro, you're right. The minimal design is an advantage in terms of easy usage and reliability on the long term. Another advantage is the possibility of a modular assembly of many simple parts to realize a complex function. 

  40. fasmicro
    August 3, 2014

    >> Another advantage is the possibility of a modular assembly of many simple parts to realize a complex function.

    That is a big driver is some of the consumer electronics. It is very amazing the weight of some of the modern cellphones. Ergonomics and lightweight packaging are competitive strategies in the industry.

     

  41. etnapowers
    August 4, 2014

    @fasmicro: I agree with you on the need of ergonomics and lightweight devices , but this implies the drawback of harder challenges for the engineers in terms of power dissipation and in terms of modeling of smaller size effects  for the simulation tools.

  42. goafrit2
    August 9, 2014

    >>  I agree with you on the need of ergonomics and lightweight devices , but this implies the drawback of harder challenges

    That is what makes grear firms – ability to overcome engineering challenges for great products. When you touch Samsung Galaxy and its lightweight, you will agree that hard engineering can make a difference.

  43. goafrit2
    August 9, 2014

    Minimal design is gaining momentum in the engineering world with frugal design. It not only offers better design paradigm but also gives better products that perform better over the metric of weight, size, power consumption etc. By learning to make products minimal, one is forced to make better designs.

  44. etnapowers
    September 1, 2014

    Agreed. The importance of the big companies is just this. Their Research & Development departments are pushed to develop advanced and lightweight devices by mean of new solutions: the competition on the market enhances the quality of the final products.

  45. etnapowers
    September 1, 2014

    The need of minimal design is increasing very quickly: as the components dimension is reduced and the ICs layout is shrinked, some issues related to the power dissipation and the secondary effects may appear, hence the engineers working on the new solutions have to face an hard challenge.

  46. fasmicro
    September 1, 2014

    The need of minimal design is increasing very quickly: as the components dimension is reduced and the ICs layout is shrinked

    The main reason is also that minimal design sells. It gives you lightweight, lower power, and better ergonomics that drive more trendy products. And finally, when you use minimal design, you have resources and that is creating value.

  47. fasmicro
    September 1, 2014

    Their Research & Development departments are pushed to develop advanced and lightweight devices by mean of new solutions: the competition on the market enhances the quality of the final products.

    That is the way markets work. When you introduce a product, you continue to refine it. Most times, feedbacks from customers and also what the competition is doing could help you improve your product. We have many generations of products simply because no product goes into maturity at first launch unless that product does not have market success meaning it does not have to be made anymore in future.

  48. etnapowers
    September 2, 2014

    @fasmicro: you're right, I would add that many times in the microelectronics industry a product does not reach the maturity because it does not complies with the customer requirements, so the new redesigning processes or some metal fixes are needed to reach the goal. Once this phase is completed and the IC works well inserted into the application board of the customer, its final product can be sold on the market and the final goal is achieved.

  49. etnapowers
    September 2, 2014

    @fasmicro: that's correct.

    The main challenge in minimal design is creating a complex device that has new and interesting functionalities, starting from elementary functions. The importance of systems engineers will grow in the next future , I guess.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.