Shop OBEX P1 Docs P2 Docs Learn Events
Thermistor Calibration — Parallax Forums

Thermistor Calibration

JLT7JLT7 Posts: 19
edited 2010-08-25 21:32 in Accessories
I want to use a Propeller board and NTC thermistor to measure air temperature in a greenhouse. The temperature range needs to be 32-120F and I need 0.5F accuracy. 0.25F accuracy would be even better.

How do I calibrate the thermistor to this level of accuracy? Is an ice/water bath at 32F sufficient, or do I need more data points, perhaps at 70F and 120F?

The thermistor I currently have is a 10k Ohm, 5% unit. Would replacing this with a 1% or 0.1% unit help me achieve my accuracy goals?

Comments

  • dredre Posts: 106
    edited 2010-07-27 23:24
    Hi,
    Might this site help?:

    http://www.emesystems.com/BS2math3.htm#thermistor

    cheers, David
  • JLT7JLT7 Posts: 19
    edited 2010-07-28 00:04
    Interesting, but that article pertains to interpolation of thermistor readings, not calibration.
  • ercoerco Posts: 20,256
    edited 2010-07-29 05:41
    With dozens of data points, you can use any thermistor. The 1% has more to do with part-to-part interchangeability and individual accuracy.

    Quarter-degree accuracy will be a challenge with any thermistor. Your wiring going to the thermistor will change resistance from 32-120 degrees F in the real world, which will affect your net resistance and calibration. So shoving the thermistor into an ice bath, then boiling water for calibration while the rest of your wiring & system is at room temp may yield quite different readings than you actually get in winter and summer conditions.
    ·
    Did you consider the DS1620 digital thermometer? No calibration required, 0.9 degree increments.·The clear winner in my mind.
    http://www.parallax.com/Store/Sensors/TemperatureHumidity/tabid/174/CategoryID/49/List/0/SortField/0/Level/a/ProductID/84/Default.aspx

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    ·"If you build it, they will come."
  • Tracy AllenTracy Allen Posts: 6,662
    edited 2010-07-29 06:10
    Thermistors are nonlinear, and they come in many quality levels. A measurement grade thermistor will come with a data sheet that gives either the 3 parameters for the Steinhart-Hart formula, or the B parameter of a simplified model, or else a table of T vs R. A thermistor that claims 5% is not one made for accurate measurement. Maybe you could calibrate it by doing a full scan next to a reference thermometer, or by measuring its resistance at three temperatures and trying to solve for the parameters in Steinhart-Hart. But it would probably be easier to purchase a precalibrated sensor or probe from a company like Omega.com or Meas-spec.com.

    But why use a thermistor? Are you measuring micro-habitats within the greenhouse? For ambient air temperature, a chip like the digital DS1620 or DS18B20 or analog LM34 can give outstanding results and they are linear and easy to interface with no tricky math. The resolution meets your requirements, but to get that accuracy of 0.25°F, you will have to do a one or two point calibration.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Tracy Allen
    www.emesystems.com
  • JLT7JLT7 Posts: 19
    edited 2010-07-29 14:48
    Tracy,

    The LM34 solution sounds like a better idea than a thermistor. What's the approach for doing a two point calibration? I assume that I measure the voltage of the LM34 at two widely spaced temperatures? Then what do I do with those data points?
  • FranklinFranklin Posts: 4,747
    edited 2010-07-29 15:52
    Read the data sheet www.national.com/mpf/LM/LM34.html no calibration needed.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    - Stephen
  • Tracy AllenTracy Allen Posts: 6,662
    edited 2010-07-29 16:01
    With the LM34, the first point can be an ice bath. (Of course, the sensor needs to be insulated!!) Use a thermos bottle and shaved or crushed ice for best results. The single point may be sufficient, because the offset error will have more import than the scale factor. For a second point you can use the steam point. Get a teakettle boiling and put the sensor inside the spout so that steam condenses on the sensor. Not into the boiling water.

    When I calibrate sensors, my transfer standard is a reference ASTM NIST traceable mercury in glass thermometer.

    Avoid self heating of the sensor by using a 5V power supply switched on only for the measurement. The sensor is low power, but in air there will be a small self-heating difference between using pulsed vs continuous power.

    The more important error will be representativeness. What is the true air temperature? It may be pretty consistent on a heavily overcast day, but when the sun is out the variation from place to place will be much much greater than 0.25 °F.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Tracy Allen
    www.emesystems.com
  • JLT7JLT7 Posts: 19
    edited 2010-07-29 16:21
    How does one account for the scale factor when converting a raw LM34 voltage to a temperature?
  • ElectricAyeElectricAye Posts: 4,561
    edited 2010-07-29 17:14
    Tracy Allen said...
    .... For ambient air temperature, a chip like the digital DS1620 or DS18B20 or analog LM34 can give outstanding results and they are linear and easy to interface with no tricky math......

    I second this opinion. If you're going to use a Propeller, why not take advantage of this sort of technology? You can place multiple DS18B20s on a single line using 1-wire technology since they each have their own built-in ROM addresses. They are somewhat expensive, but you eliminate all the hassles of calibration, etc. and they are easy to work with.

    datasheets.maxim-ic.com/en/ds/DS18B20.pdf

    Check it out.

    smile.gif
  • Tracy AllenTracy Allen Posts: 6,662
    edited 2010-07-29 17:14
    JLT7 said...
    How does one account for the scale factor when converting a raw LM34 voltage to a temperature?

    Using the Stamp ** operator.

    Ideal:
    ' degF = millivolts/10
dgF = mV  ' degrees time 10 = mV
DEBUG DEC dgF/10,".",DEC1 dgF, " degF", CR   ' shows xxx.x degrees



    With calibration, example,
    ' read 325 mV in ice bath, instead of ideal 320
' read 2125 mV in condensing steam, instead of ideal 2120
' degF = (mV -5) * (2120-320)/(2100-325) = (mv - 5) * 1.014084507042
'  ** multiplier is 0.0140845 * 65536 = 923
dgF = (mV - 5) ** 923 + (mV - 5)
DEBUG DEC dgF/10,".",DEC1 dgF, " degF", CR   ' shows xxx.x degrees



    Maybe you could beg borrow or steal (not) an accurate reference thermometer, in order to check the results and support your confidence level in the readings. The LM24 comes in several accuracy grades. The highest is the LM34CAH or LM34CAZ, with +/- 1 degree Fahrenheit. The limits of error are shown in the data sheet, including nonlinearity error, which is very low.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Tracy Allen
    www.emesystems.com
  • JLT7JLT7 Posts: 19
    edited 2010-08-02 21:43
    Tracy,

    Thanks for the detailed description of the LM34 calibration process.
  • SeariderSearider Posts: 290
    edited 2010-08-03 16:41
    JLT7,

    Don't forget to allow for any radiated heat. If you have have the sensor where sunlight could hit it or reflect on to it, then you will be measuing a combination of conducted heat (air temp in contact with the sensor) and radiated heat (infrared from any heat source).

    I once saw a rig where a person put the temp sensor inside a large (3") diameter PVC pipe in the shape of an upside down "U". Then added a slow speed fan to move the air through the tube. The Tube·protected the sensor from·radiated IR sources. It might have·been·wrapped in alluminum foil which is a very good IR reflector.·I seem to remember that there was a complex sequience where the fan would run for a few minutes to replace the air in the tube, then stop. a pause to wait for the sensor mass to come up temp with the air, then a reading.· The sequence was repeated for every temp reading.



    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

    Searider
  • marc4gmarc4g Posts: 1
    edited 2010-08-25 21:32
    Another possibility is the DS1624.

    They have a resolution of 5 bits to the left side of the decimal point. (0.031C)

    Up to 7 devices may be addressed.

    They are quite easy to use and interface with the basic stamp 2.

    I have some commented code in 'bs2' if you are interested.

    Marc
Sign In or Register to comment.