Calibration with confidence -
the assurance of temperature accuracy

R.D. Collier

Taylor Instrument / Consumer-Industrial Products / Sybron Corporation
Arden, North Carolina 28704


Highly sensitive temperature devices, particularly those with multi-digit electronic display, give the illusion of accuracy. However, knowledge of true temperature -- the real concern of measurement accuracy -- is only indirectly related to sensitivity or precision. To assure temperature accuracy, it is necessary to maintain a temperature reference standards capability. this must include equipment and procedures that permit calibration of operating devices with temperature standards in a way that insures minimum uncertainty. For most requirements the creation and maintenance of such capability is neither expensive or difficult, but lack of understanding often results in expense and inaccuracy. Equipment and procedures are discussed that permit calibration with confidence at three levels of accuracy; an uncertainty level of +/- 1.0 degrees Celsius, +/- 0.1 degrees Celsius, and +/- 0.01 degrees Celsius, respectively.

Subject index: Calibration methods, general.


This paper outlines temperature instrument calibration fundamentals that apply to "daily use" conditions in laboratory and industry. In style, language, and content, therefore, it differs from the majority of papers on temperature measurement.

Most technical papers are written to advance knowledge in a given field, and are written primarily to benefit the few working actively in, and are most familiar with, that field.

In contrast, this paper is written to restore to general understanding a knowledge of long-standing calibration fundamentals that are familiar to experienced professionals in the field, but are not generally understood by many who have a "need to know."

The task of assuring accuracy in temperature measurement is critically important. Safety or health would be compromised, equipment damaged or product wasted in many processes if the temperature were incorrect. And no matter how precise the measurement or careful the operator, if the device is not calibrated correctly, the result is wrong.


The assurance of temperature accuracy begins with an understanding of four key concepts: "Accuracy," "Precision," "Reference" and "Standards," and the relationship between these terms:



Most temperature measurement involves use of a measuring instrument of some type, usually a thermometer. Assurance of accuracy of that instrument involves basically a two step process:
  1. Compare -- under conditions as close as possible to actual operation, the indication of an operating instrument with a "working standard" -- a master reference thermometer whose accuracy is known with very small uncertainty.

  3. Periodically check the accuracy of the master reference thermometer in accordance with the manufacturer's instructions -- either by reference to a primary standard of temperature such as an ice bath, or by having the instrument re-calibrated at NIST or a respected testing laboratory.
Details of the methods and equipment needed to accomplish these two procedures with confidence, depend on the level of uncertainty required. We will consider in this paper the equipment and procedures needed to calibrate to three uncertainty levels; an uncertainty (or maximum expected error) of +/- 1 degrees Celsius, of +/- 0.1 degrees Celsius, and +/- 0.01 degrees Celsius. However, calibration at all three levels involves this same basic two-step approach.


There are four important fundamental considerations that are most important in assuring good calibration procedure:
  1. Insure that conditions of installation of the sensing element approximate actual use conditions as closely as possible. Degree of immersion, ambient temperature, shielding and housing (protective shield or other installation accessory) all may affect the heat flux around the sensing element and thus influence its calibration. Much calibration work is done by using a rapidly-agitated liquid bath as an approximation of actual use conditions. Such baths are the least expensive way to provide a stable, uniform, easily regulated temperature transfer medium. They may or may not closely simulate actual sensing element heat flux conditions. For instance -- consider a sensing element that is in a metal-shielded housing (with a substantial heat flux through the housing to cooler surroundings and hence a higher-than-normal reading. On the other hand, oil has much poorer heat transfer capability than water or steam due to its insulating properties, and hence may supply less heat to maintain the heat flux, resulting in a lower-than-normal reading.

  3. Insure that the equipment used for calibration, and the surroundings and procedures, contribute the smallest error that is possible. This usually includes having a relatively large mass of liquid medium, agitated vigorously to insure good heat transfer and minimum temperature gradient; insulation to aid in temperature stability; and a sensitive proportioning temperature control system to minimize fluctuation. Depending on temperature range and conditions, the equipment need not be sophisticated or expensive. For example -- for calibration between ambient and, say, 140 degrees F, a large, insulated food/beverage container provided with a kitchen food mixer and simple paddle for agitation, and with the temperature controlled by manually opening and closing a hot water faucet -- can become, in the hands of a skilled operator, a precision calibration bath useful for calibration at uncertainty levels less than 0.1 degrees C.

  5. The master reference thermometer must have an accuracy such that its level of uncertainty is a small fraction of the allowable calibration error desired; preferably on the order of one to two tenths. This means that for calibration of thermometers or temperature control devices to within a maximum expected error of +/- 1 degree C, the reference thermometer should have a maximum error of no more than a few tenths of one degree; for the calibration error to be less than +/- 0.1 degree C, the reference thermometer must have a maximum error of no more than a few hundredths of a degree, and so forth. Equally important is the long-term stability of the master reference standard. It must be able to be used with confidence for a practical period of time between its own calibration checks, and with reasonable certainty that it is not subject to short-term variations in calibration.

  7. The above three fundamental considerations are all involved with the process of comparing a temperature sensing instrument to a master reference standard thermometer, to do with the second step, that of insuring that the master reference instrument is itself continuing to be accurate. This assurance of the accuracy of the standard themometer is again done by comparison. In this instance, however, the comparison is usually done by referencing its indications to a primary standard or near equivalent. The most commonly-used of these are the triple point of water, or for most laboratory and industrial use, its near equivalent, the ice point. It is commonly understood that an ice point may have uncertainties on the order of 0.01 deg C. However, James L Thomas of NIST, in 1939 performed an exhaustive test that indicated that with care, the ice point could be realized with an uncertainty of little more than the triple point of water. (ref 2) Moreover, an ice bath is so much easier to prepare and use than any of the standard fixed points that it has become the common choice for reference standard calibration. However, as with any procedure in high-accuracy work, care must be taken. It is therefore appropriate to describe procedures that will insure minimum error.


The basic steps required to insure ice-point accuracy are:

  1. Insuring water purity

  2. For most purposes, ice made from ordinary culinary water is sufficient. However, since most dissolved minerals affect the freezing point, it is common to use only ice and water that has been demineralized. For an ice point with less than 0.01 deg C uncertainty, only distilled water, and ice made from distilled water should be used and the container should be of carefully-cleaned glass or stainless steel. As little as 12PPM of some salts can cause a 0.01 deg C reduction in the ice point.
  3. Insuring minimum heat flux

  4. To insure that the sensor being tested is unaffected by ambient conditions, it should be placed in the center of a relatively large mass of ice and water (normally two liters or more), well away from the walls of the container, and the container should be insulated to minimize melting of ice. The sensing element being calibrated should be immersed adequately to minimize heat transfer through its housing (remembering the rule that calibration conditions should approximate use conditions).
  5. Insuring equilibirium

  6. To guard against temperature rise due to insufficient ice, and to insure against poor heat transfer due to air in the bath, the following procedure is recommended:
    1. Fill the container with crushed or chipped ice.
    2. Fill the container with water to an overflow condition.
    3. Add more ice until ice is tightly paced to bottom of container, allowing water to overflow.
    4. Insert sensor to be calibrated and allow temperature to reach equilibrium (normally 5 minutes or more).
    5. If test continues more than a few minutes, add more ice periodically, as before, insuring that ice is packed tightly to bottom of container each time. The goal is to insure that at all times the sensor is in contact with an ice/water mixture over its entire surface.


Application of fundamentals discussed above to the calibration of specific temperature sending elements will vary somewhat, depending on the level of accuracy required. It is uneconomical and unnecessary to take the time and care needed for extremely precise calibration, when not required by the needs of the process being monitored, or when the sensor has substantial built-in inaccuracy. The important consideration is the amount of inaccuracy (or, more properly, the level of uncertainty) that is permissible. For convenience, we will discuss procedures for three levels of uncertainty: +/- 1.0 deg C, +/- 0.1 deg C, and +/- 0.01 deg C.
  1. Calibration within +/- 1.0 deg C uncertainty:

  2. For many uses where an uncertainty of the order of +/- 1 deg C is acceptable, thermometers and controllers are purchased having specifications that claim inaccuracies no greater than that amount. The instruments are then used for extended periods of time without calibration -- often, in fact, until breakage or major malfunction occurs. If in fact, and accuracy of +/- 1 deg C is important, this is a dangerous practice, since few instruments will remain in calibration for extended periods unless specifically made for long-term stability. Even many glass thermometers, generally accepted as "correct unless broken," are no longer regularly made with the expensive glass annealing and aging steps that insure the necessary stability.

    The simplest calibration procedure for such instruments is to make a periodic ice point check, if 0 deg C is included in the instrument range, and/or to compare desired readings with that of a high-quality mercury-in-glass thermometer such as the ASTM precision series, ASTM 62C through 70C (or F). These reference thermometers have scale graduations, in the moderate ranges, of +/- 0.1 deg C or +/- 0.2 deg F and hence are within the accuracy range (an order of magnitude more accureat than the instrument to be calibrated) needed for such service.

  3. Calibration within +/- 0. 1 deg C uncertainty:

  4. In order to insure that routine temperature measurements with operating instruments are accurate to within +/-0. 5 deg C to +/-1. 0 deg C, it is necessary for the instrument itself to be calibrated to an uncertainty of no more than +/-0. 1 deg C. Since this is the accuracy range most commonly needed in industrial use, the calibration procedures will be described in more detail than those above.

    1. Equipment: Care must be taken in selecting and using equipment for calibration at this level of uncertainty, since the reference thermometer, temperature controller and other items must introduce errors of no more than a few hundredths of a degree.

    2. The following items are recommended:

      1. Reference Standard Thermometer: One of two types of instrument is commonly used; a high-accuracy mercury/glass thermometer accompanied by a signed certificate of calibration with corrections to the nearest 1/5 of a graduation division, or a precision platinum resistance probe with high-accuracy indication system, also accompanied by a NIST-traceable calibration record. Since there is a cost difference of between 10: 1 and 50: 1 between the two instruments, the mercury/glass thermometer is most commonly used.
      2. Ice bath: The same ice bath can be used as described above, as long as care is taken to avoid contamination of the water or ice. One additional piece of equipment is needed, a 10X microscope and stand, to allow reading of the mercury/glass thermometer without parallax and to permit careful interpolation to at least the nearest 1/5 of a graduation division.
      3. Temperature bath: There are three important criteria in good bath construction: First, that the heating/cooling elements be isolated from the test area; second, that the bath be well insulated to minimize heat transfer load and controller stabilization needs; and third, adequate agitation. As a rule of thumb, on all baths except those at low temperatures where the medium is highly viscous, adequate agitation is insured when the liquid surface has the appearance of water at a "rolling boil" condition. Also, in order to insure stability, most well-designed baths have a minimum exposed surface area. If this is not possible, a well-insulated cover should be made to cover all but the minimum exposed surface area.
      4. Temperature Controller: Since the advent of solid-state electronics, vast improvement in proportioning controllers has come about. The best for calibration bath purposes have a visual indicator--a flickering lamp that indicates control status (off when temperature is above control point, on when below, and flickering intermittently when at control point) . For control temperature below ambient, it is common to install a throttle-able refrigeration system for gross control (continuous operation) and an electric heater with sensitive controller to override for fine control. As noted under "A" ." above, manual control can also be used if calibration is infrequently done and the cost of an adequate proportioning action must be simulated by a variable resistance unit that allows a varying heat input rates rather than "on-off" control.

      Procedures: Actual calibration procedure for achievement of less than +/-0. 1 deg C uncertainty is quite simple--still following the "BASIC APPROACH" described at the beginning of this paper. The major effort centers around extra precautions taken to insure that each error and uncertainty is less than a few hundredths of a degree, so that the sum of all uncertainties is less than one tenth. The degree of difficulty in achieving this result depends on the temperature. It is not difficult--with proper equipment and training--in the range from 0 deg C through 90 deg C, more difficult in the range 0 deg C to -40 deg C and 90 deg C to 200 deg C, and extremely difficult outside those ranges due to equipment limitations.

      Greatest attention will be given to procedures using the most dependable and economical components; (a) a rapidly-agitated liquid bath or baths for temperature comparison, and (b) a master reference standard thermometer or set of thermometers that are mercury-in-glass units built to ASTM Precision-series standards but calibrated and certified accurate to the nearest 1/5 of the smallest scale division--with certification directly traceable to the NIST. Comments are in two groups corresponding to the two steps of the Basic Approach; comparison of thermometer to be tested with the reading of the master reference thermometer, and when calibration check of the master reference thermometer:

      (a) Comparison of thermometer to be calibrated with master reference thermometer in agitated liquid bath. This involves primarily attention to details that could influence the accuracy of results, including:

      (b) Calibration check of master reference thermometer: If the master thermometer is a high-accuracy mercury-in-glass unit that has been properly made and certified, this calibration check is primarily a matter of making a periodic ice point check under carefully- controlled conditions (described below); and recalculating calibration corrections if necessary. Normally, such a thermometer can be used for decades without needing to be returned to the factory or laboratory for recalibration. If a platinum resistance thermometer is used as a master reference, it should be completely recalibrated (at least at all temperatures needed for use) once per year or oftener.

      The continued use of mercury-in-glass thermometers for the majority of applications as master reference standards is due to this unique feature--the face that if proper records are maintained and procedures followed, the accuracy of the thermometer can be known with confidence for several decades without the need for a full recalibration. This is true of few other temperature devices. The following explanation might help understand this unique feature:

      a. All measurement devices are subject to change with time and usage. This includes the resistance elements of platinum resistance thermometers and bridges as well as the glass of mercury-in-glass thermometers. The important criterion is to be able to measure and know the magnitude of these changes.

      b. The ideal way to know how much change has occurred in a device is to compare it periodically with something that does not change--a "primary standard."

      c. This brings us to a pair of interesting phenomena that combine to provide the unique capability of the high-accuracy mercury-in-glass thermometer as a master reference standard:

      (1) High-accuracy glass thermometers have been made for over a hundred years, and during that time manufacturing techniques have been developed and tested that have been time-proved to assure a remarkable capability: That is, that essentially all measureable change that will affect the temperature indication will occur in the bulb of the thermometer. It is possible, then, if the temperature representing the freezing point of water (the ice point), is included within the thermometer scale, that a careful calibration check at that one temperature will, in effect, provide a calibration check of the entire scale, since there will be no relative change of indication of one part of the scale over another.

      (2) It is relatively easy and inexpensive to realize the temperature of freezing water to a level of uncertainty of a few thousandths of a degree in any laboratory or office. Thus, a temperature instrument that needs only an ice point check to assure its accuracy over its entire temperature scale can be recalibrated indefinitely by simply making ice point checks and applying any correction needed to all other temperatures indicated by the instrument.

      d. To permit this simple calibration check, mercury-in-glass thermometers made for use as master reference standards include the following features:
      (1) An auxiliary "ice point" scale if 0 deg is not included in the range.

      (2) Unusual care in manufacturing--up to 75 or more manufacturing steps including aging and annealing operations compared with 20 or less steps in making "laboratory" glass thermometers.

      (3) An individually-graduated scale etched into the glass surface. Each individual graduation may be spaced slightly differently than the adjacent graduation to exactly match variations in the glass bore diameter.

      (4) A signed certificate of calibration resulting from a retest of the thermometer at a number of points throughout the scale range, under extremely carefully controlled conditions, using a reference thermometer kept in calibration through a high-level recalibration and Measurement Assurance Program as described under "Calibration within O .010 uncertainty" in the main body of this paper. Any corrections noted on the certificate should then be applied to appropriate readings of the thermometer, with interpolation between certification points.

    For a greater understanding of thermometry practice using mercury-in-glass thermometers, refer to NBS Monograph 150 (ref 3) and for greater understanding of high-accuracy thermometry using platinum resistance elements, see NBS Monograph 126. (ref 4)

    For a high accuracy ice point check of a mercury/glass master reference thermometer, the following should be observed:

    -- Insuring that only demineralized water and ice are used, that the bath is kept full of ice and water, that precautions are taken to insure minimum heat flux and complete stability, as described above under "Realization of Ice Point."

    -- Use a 10X microscope, carefully aligned to insure that the microscope axis is perpendicular to the axis of the thermometer. This insures against parallax error, and allows accurate interpolation of mercury column height to 1/10 of the smallest graduation division.

    -- Also using a 10S or 20X microscope, examine the bulb and bore of the thermometer to insure that there is no evidence of "air" in the bulb or mercury column, and no droplets of mercury separated from the column.

    -- Keep adequate records to gradually gain confidence in the stability of the master reference thermometer. A good plan is to check the ice point at least every four months until a shift of less than 0.2 of the smallest division occurs between checks, then extend to an annual check. However, if annual checks show a change of 0.2 division or more, return to four-month checks until again stabilized.

    -- Whenever a careful ice point check shows shift in calibration of more than 0.2 of a division, the calibration certificate should be amended to add the correction to all calibration values. (For mercury-in-glass thermometers only.) This can be done as a result of over 100 years of experience that verifies that essentially all change occurs in the glass bulb of the thermometer, and its magnitude is determined by the ice point check. Readings at all other scale points will have, therefore, shifted the same amount as the ice point.

  5. Calibration within +/-0.01 deg C uncertainty:

  6. This is the level of accuracy required to perform initial calibration and recalibration of the master reference thermometers described under "B ." immediately above. Since this level of accuracy requires a calibration uncertainty of no more than a few thousandths of one degree, unusual care must be taken. Mercury/glass thermometers cannot be used, due to their lack of resolution as well as mechanical variations. The thermometric standard commonly used is a precision platinum-resistance element used with a precision potentiometric bridge. Newer systems such as quartz thermometers and electronic digital indicating devices are available, but do not have the long-term performance record of the platinum resistance element and bridge combination.

    Actual calibration procedures are similar to those described under "B." above except that greater care is taken at each step, and long-term experience in calibration techniques is required to minimize errors. However, to insure the continued accuracy of the master reference thermometer used for such calibrations requires sophisticated equipment and procedures. Basically, the resistance element as well as the precision bridge are trouble-free, extremely stable instruments. They are both, however, subject to small changes with time, and these changes can affect the output value at one portion of the range while not affecting it in other areas. This requires a regular recalibration schedule for both the bridge and resistance elements. At this accuracy level, interlaboratory correlation becomes important as part of a Measurement Assurance Program. (ref 5)

    Such a program is planned to assure confidence that uncertainty levels of no more than a few thousandths of a degree are maintained. Beyond that basic element, however, the program includes a system of checks and double checks to virtually eliminate the possibility of error due to equipment failure or operator mistake. This program includes most, or all, of the following steps:

    -- Periodic (oftener than once per year) checks of working bridges and resistance elements against a master bridge and element.

    -- Calibration check of master bridge by use of a standard resistor on an annual or more frequent basis.

    -- Calibration check of standard resistor by independent testing agency--annually until fully stabilized, then every three to five years.

    -- Round-robin interlaboratory comparison tests of resistance elements.

    -- Periodic check of both working systems and the master calibration standard system against primary standards--not only the triple point of water, but other according to need, such as:



In summary, it is possible to have confidence that temperature-measuring instruments are accurate by following a simple two-step process: First, comparison under controlled conditions of an operating temperature device with a master reference standard thermometer; and second, periodically checking the accuracy of the master thermometer by appropriate means.

A calibration program offering assurance of accuracy to a level of uncertainty of less than 0. 1 deg C can be developed at low cost, based on the use of carefully-made and calibrated mercury-in-glass thermometers as master reference standards. This accuracy and economy is possible because of the simplicity of high-accuracy calibration check at the temperature of freezing water (the "ice point") , and the property of a mercury-in-glass thermometer that an ice point check insures that the magnitude of calibration change is known throughout the entire temperature range of the thermometer.


  1. For a complete discussion of IPTS-68, see the authorized text in Metrologica 5, 35 (1969). Return to text
  2. Thomas, James L., "Reproducibility of the ice point," in 1941 edition of Temperature, Its Measurement and Industry, New York, Reinhold Publishing Co., 1941. Return to text
  3. J. Wise, Monograph 150, U.S. Department of Commerce, National Bureau of Standards, January 1976. Return to text
  4. J. Riddle, G. Furukawa and H. Plumb, Monodgraph 126, U.S. Department of Commerce, National Bureau of Standards, April 1973. Return to text
  5. For a description of considerations in an effective Measurement Assurance Program, see Furukawa, G.T., "A Measurement Assurance Program - Thermometer Calibration," unpublished ASTM Technical Talk, June 25, 1980. Available from Dr. Furukawa, U.S. Department of Commerce, National Bureau of Standards. Return to text