Lake Shore Model 321 Autotuning Temperature Controller User’s Manual
D-14
Application Notes
MEASUREMENT SYSTEM INDUCED ERRORS IN DIODE THERMOMETRY
by John K. Krause and Brad C. Dodrill
Diode temperature sensors are capable of being used at the accuracy level of a few hundredths of a kelvin. However, in
order to achieve this performance, proper measurement techniques must be used. Poorly shielded or improperly
grounded measurement systems can introduce ac noise which will create an apparent shift in the dc voltage reading
across a diode sensor. This results in a temperature measurement error which may approach several tenths of a kelvin.
The presence of the ac noise in question is not obvious during normal usage and several quick tests are outlined to verify
whether or not a noise problem exists. Experimental data and derivations from theoretical
p-n
junction characteristics are
given which correlate the ac noise level with possible voltage/temperature measurement errors. These results can be
used in estimating the accuracy and performance of a temperature measurement system. Several of the more common
problems which introduce noise into diode circuitry are described.
INTRODUCTION
Current technological uses of temperature sensors require better calibration accuracies and better device performance
than ever before. However, the assurance of an accurate temperature measurement does not stop with simply the sensor
specifications. Just as critical is the instrumentation used with the sensor and the manner in which the instrumentation is
used. This paper concentrates on identifying, verifying, and eliminating an often overlooked instrumentation or system-
induced error in the use of diode temperature sensors.
I. PROBLEM DEFINITION
Semiconductor diode temperature sensors have been in use for over
20 years and, with the advantages they offer over resistance sensors
or thermocouples for many applications, their popularity continues to
increase. Diodes are operated at a constant current, typically 1, 10, or
100 µA, while the voltage variation with temperature (V[T]) is
monitored. The diode sensor has a useful temperature range from
above room temperature to as low as 1 K, with reproducibilities to
better than ±50 mK. Figure 1 shows the voltage variation with
temperature for a typical silicon diode temperature sensor.
An error arises in diode thermometry if the excitation current is not a
true dc current but has an ac component superimposed on the dc.
Although the ac component can be due to a poorly designed current
supply, a more common source of the ac is noise induced in the
measurement circuit. This noise can be introduced through improper
shielding, improper electrical grounds, or ground loops. Currently
available voltmeters have sufficient normal-mode rejection
capabilities in their dc measurement modes that these noise effects
can go completely unnoticed if they are not explicitly checked. The
equivalent temperature error which may be caused by this problem is
typically a few tenths of a kelvin, although an extreme case with a 4 K
error has been observed.
The effect of the ac noise appears as a shift in the dc voltage
measurement due to the nonlinear current/voltage characteristics of
the diode. An illustration of this effect is shown in Fig. 2 where an
exaggerated IV curve is given. An induced ac noise current
superimposed on the dc operating current (I
dc
) is shown along the
current axis. The resulting voltage seen by the voltmeter is shown
along the voltage axis. The nonlinear IV characteristics of the diode
have caused a distortion in the ac voltage signal making it
asymmetrical with respect to the voltage reading corresponding to I
dc
.
When a voltmeter operating in a dc voltage mode reads this signal,
the signal is processed (by integrating, filtering, etc.) to give an
average dc voltage reading which will be lower than expected. The
apparent temperature measurement will then be too high. Note that
this voltage offset is due to induced currents in the total measuring
system and is not simply a voltage pickup by the diode itself. An ac
voltage superimposed symmetrically about the dc operating voltage
of the diode would not cause a dc voltage offset.
FIGURE 1.
Voltage-temperature curve for a typical
silicon diode temperature sensor at a constant current
of 10 µA.
FIGURE 2.
IV curve for a silicon diode sensor showing
effect of an induced ac current superimposed on the dc
operating current I
dc
. The expected dc operating voltage
is V
dc
, which is shifted from the average voltage V
ave
indicated by the voltmeter in a dc measurement mode.
