Ditch the NTC thermistor: use an analog temp sensor instead

Thermistors and analog temperature sensors are two commonly used temperature sensing solutions that can be used for most electronic applications.  Deciding which technology is the best fit for your application can be a difficult task.  However, I’m going to show you a few reasons why you should ditch the NTC thermistor and design in an analog temperature sensor instead.

Figure 1 shows an output voltage vs. temperature comparison.  Notice that NTC thermistors requires to be used in a voltage divider circuit. Furthermore, other design techniques adding resistors in parallel to the NTC or proceeding to a polyfit calculation in the MCU help linearize their output.  This is because their resistance vs. temperature characteristic has an exponential shape.  Unlike NTC thermistors, analog temperature sensors do not require any additional circuitry as they have a virtually linear output voltage.  For example, Texas Instrument’s TMP235 analog temperature sensor provides a very linear and accurate output voltage across the devices entire operating temperature range of -50°C to +150°C.

As you can see from the three NTC thermistor curves in figure 1, you can change the value of the bias resistor to adjust the location of the linear portion of the curve.  Notice that this is a limited range and that the curves start to saturate at low and high temperatures.  When interfacing with an ADC, this saturation will cause temperature errors if the resolution of the ADC is not high enough to detect a change in output voltage per degree Celsius.  As a result, NTC thermistors tend to be less accurate across the entire operating temperature range and often require a higher resolution ADC in these cases.

Otherwise, if you want to keep the very small footprint of your thermistor and have as little disruption in your actual design as possible, TI released a series of pin to pin compatible thermistors for 0402 and 0603 NTCs, the TMP61. The TMP61 has the immense advantage of being linear all over its temperature range (-40°C to +125°C), and this at the same price point, This linearity simplifies the design of your system thanks to easy calibration, a smaller lookup table and no polyfit or external circuitry to linearize, as well as allowing you to reach better temperature accuracy over wide temperature ranges.

Figure 1: Output Voltage (V) vs. Temperature (°C)

Figure 2 shows a supply current vs. temperature comparison.  The TMP235 has a typical value of 9µA and a max value of 14.5µA.  NTC thermistor networks tend to dissipate more power as their supply currents are much higher and vary greatly over temperature.  Notice that if you increase the resistance of the bias resistor, the supply current for the NTC thermistor network will decrease.  But remember, the bias resistor is also chosen to ensure that the output voltage vs. temperature curve is linear for the desired temperature range.  This is a tradeoff that can be ignored by using an analog temperature sensor as they have both a fairly constant low supply current and a virtually linear output voltage.  Another disadvantage to NTC thermistors is that engineers must account for the self-heating effect as this will cause additional errors.

Figure 2: Supply Current (µA) vs. Temperature (°C)

You can avoid these NTC thermistor design issues by ditching them in favor of TI’s easy-to-use analog temperature sensors and linear thermistors. Check out my Engineer It video on the TI E2E community and learn more about TI’s analog temperature sensors.  

  • Brian,

    Very informative article. I find that there are very few people that understand that they need a voltage divder for the NTC thermistor instead of a current source. Your arguements are excellent. I find that the NTC thermistor is better suited for temperature test circuits over a limited temperature range, say +/- 25 C or smaller. Is this a more realistic temperature range for the temp sensor user? Also, can you give me an estimate as to the cost differential between the two circuits is?  

  • Hi Bonnie,

    Appreciate you taking the time to read through my article - thanks!  I agree - NTC thermistors are a useful temperature sensing solution for temperature test circuits over a limited temperature range.  There are more complex resistive networks that can be used in order to extend that range, but with that comes an increase in complexity, cost, footprint, etc...  Analog temp sensors, on the other hand, can be used for temperature test circuits across the device's entire operating temperature range.  For example, TI's LMT series of analog temp sensors work well across their operating temperature range of -50C to 150C.  Also, you can see from figure 2 that analog temp sensors have a clear supply current advantage across temperature.  Lastly, TI's LMT series of analog temp sensors are available at a comparable system cost.

    Best Regards,

    Brian Gosselin Jr.

  • Thanks for the info... I went and looked and the price of the LMT87 is very nice.  At DigiKey ( not known for low prices ) theya re 73 cents for one and below 20 cents in quantity.  I am trying to use 48AWG copper wire for more linearity ( PTC ) and very low price.    99guspuppet

  • Oh yeah   I forget to add.... i wish the LMT87 had a frequency output or a PWM output so no ADC was required.    99guspuppet

  • Hi Puppetboy,

    You mentioned that you wished that the LMT87 did not require the use of an ADC - are you familiar with our digital temp sensors?  Our digital temp sensors do not require the use of an ADC as they provide a digital output.

    Digital temp sensor parametric search: www.ti.com/.../digital-output-products.page

    Best regards,

    Brian Gosselin Jr.

  • Hi Brian,

    I see the errors in temperature in the data sheet as high at 2.7C. Do you have any estimates of the relative, rather than absolute, error for these devices.

    It sounds like you know a lot about the resistor divider circuits, but, in an act of shameless self promotion, and because I think there is good information for your readers on how to use a thermistor for accurate measurements, I point you to the following article. In it, we show that thermistors can be used over wide ranges (though wide is a subjective term) and with three sigma errors in absolute temperature of 0.35 C or less, and with relative errors on the order of 0.04 C.




  • Hi Darren,

    Yes, if you take a look at figure 2 of the LMT87 datasheet, you can see the temperature error vs. temperature plot across the devices entire operating temperature range of -50C to 150C.  Another graph that may be of interest to you is the graph of LMT84 output accuracy vs. temperature which can be found on the bottom right of the analog temp sensor homepage.  "From 96 randomly-selected LMT84 devices, and without calibration, this data shows the very tight accuracy and linear performance of LMT84."

    Analog temp sensor homepage:  www.ti.com/analogtempsensors

    I took a look at your article - great read!  The data corresponded to a temperature range of -32C to 32C - I'm interested in seeing results across a wider temperature range like -50C to 150C.

    Best regards,

    Brian Gosselin Jr.

  • Brian, Excellent article. I will be changing my design approach to temperature monitoring applications. Thanks for the valuable input!

  • I like something like an LM335 because you can connect it with only 2 leads, just as the thermistor. (I know, the article isn't about simplicity, but I wanted to note that :)

  • Hello Guys,

    I have used thermistor in my few projects, where I needed to place that thermistor in a drilled hole of 3mm diameter of Aluminimum assembly or on top of Aluminium part of optical chamber with area of 4 cm sq. In both cases wire length was approximately 1 meter. I don't think I can use this LMT87 in drilled hole, but may be I can use it for optical chamber application. Let me know about wire length contraints.

    My temperature range is just 25 to 50 Deg C with 40 Deg C as set point.

  • Now I'm confused.  I always thought that an NTC thermistor is an analog sensor (despite being nonlinear).  Is the world changing again?

  • Yuvraj,

    You would be okay with using an analog temp sensor, such as the LMT87, for applications that have a wire length of approximately 1 meter.  Analog temp sensors have lower output impedance than NTC thermistors and as a result are less susceptible to noise.  You may find that some filtering may be required for applications with longer wire lengths, say 100 meters.


    'Analog temp sensors' is what we are calling our analog output local temperature sensors.

    Analog Temp Sensors: www.ti.com/analogtempsensors

    Parametric Search: www.ti.com/.../analog-output-products.page

    Best Regards,

    Brian Gosselin Jr.

  • Hello Brian,

    Thanks for your reply.

    As I said in my earlier post, I want to use it for two applications, first in drilled hole in aluminium chamber and anothe ron flat Aluminum chamber.

    In any way, is it possible to use this sensor inside 3mm diamater hole? Any other similar sensor comes as direct replacement to standard NTC replacement?

    In second case, on flat Aluminium surface, what should be the glue material used to glue LMT87 on Aluminium surface?

    Do you have any reference circuit for filter if wire length exceeds 1 meter?

    Best regards


  • Hi Yuvraj,

    If you could make your hole bigger you could use a TO92 package leaded package such as the LM19 and LM35. The TO92 is slightly bigger than 5mm. Also in the near future the LMT87 will also be available in this package. The LMT87 is small enough to fit but you would need to come up with some mechanical way to fit it in the 3mm hole. We have made some long thin PCBs and mounted the sensors at the tip perhaps something like that would work for the SC70 package for you. The minimum width we have done is about 0.25”. I haven’t done such a small mechanical design and would have to do one to know for sure. The package height is 1mm and the width is a little over 2mm so you may need to make the hole slightly bigger. Also it will depend on the pcb fab house you use and their capabilities. I suppose you could also just mount the pcb such that the sensor only is in the hole and the PCB sits flat on top of the aluminum. Again though, for this to work and for the whole package to fit into the hole, the hole will probably need to be slightly bigger.

    You can use any thermally conductive electrically isolating epoxy to glue a sensor to an aluminum surface. There are many available in the market from 3M, Monsanto and many more. It's best to contact these manufacturers for a recommendation that best suits your needs. Depending on the mechanical connection method an adhesive tape could also be used.

    You ask for a reference circuit if the wire exceeds 1 meter. That will depend on the environment surrounding the sensor so I will be giving you general guidance that you can hopefully use to apply to your specific needs. For the analog output you can refer to a specific analog temperature sensor datasheet as there is a recommended minimum series resistance requirement specified for a given capacitance that can be found in the applications section. The longer the wire the more capacitive loading on the output the higher the resistance required. For example on the LMT87 datasheet we list on page 9 a table for minimum R_S for a given C_LOAD, use this as guidance. The filter cutoff frequency depends on the thermal response time of the system because an output RC filter response time (if too low) will combine with the physical thermal response time of the device to provide the overall response on the output of the analog temp sensor. Place the filter as close as possible to the ADC or other device that is connected to the temp sensor analog output.

    Take care,

    Emmy Denton

    Temperature Sensor Applications

    Texas Instruments