TEMPERATURE-DEPENDENCY
OF CAPACITIVE POLYMER HUMIDITY PROBES
Simone Griesel, Horst Niemand, Manfred Theel, Eckhard Lanzinger
Deutscher Wetterdienst (DWD)
Frahmredder 95, D- 22393 Hamburg, Germany
Tel.:++49 406690-2457, Fax: ++49 40 6690-2499, E-mail: simone.griesel@dwd.de
ABSTRACT
To determine relative humidities (RH) during times of rapid humidity changes sensors
with low response times are needed.
Because saturation vapour pressure changes with temperature and air pressure the
relative humidity measurement may also be affected by temperature. The CIMO Guide
only states significant temperature and humidity dependence of solid state sensors but
without quantifying the influence on measurement uncertainty and sensor time constant.
The aim of this work was to determine the extent to which the response times of
humidity sensors are affected by temperature. Eight different capacitive polymer
sensors (heated and non-heated) versions were tested in a climate chamber at Deutscher
Wetterdienst (DWD). Some results of these laboratory tests are presented and the set up
for a field test is described.
Introduction
The accurate measurement of relative humidity (RH) is important for forecasting the visibility and
the calculation of the dewpoint. Capacitive polymer humidity sensors are widely used for these
measurements due to their easy handling, good cost performance ratio, wide measurement range,
good accuracy and fast response. The CIMO Guide (2008)[1] reports achievable uncertainties of
3%RH to 5%RH (Table 4.1) and time-constants of 1s to 10s (Table 4.2). While the accurracy
requirements are not separated neither by temperature nor by humidity the 1/e time constant values
for electrical capacitive sensors are given for 85%RH but without mentioning the temperature
influence.
So far most manufacturers differentiate the accuracy of their capacitive humidity sensor between
temperature and humidity, while the definition of response time is limited to a single reference
temperature, typically 20°C. Furthermore the specification of the 1/e time constant or t63 for
capacitive humidity probes suggests that the sensors respond as a first-order LTI system (linear
time-invariant).
1
Dooley&O´Neal [2] have reported that the transient response for capacitive thin film sensors could
be characterized by an exponential curve with two time constants, a fast initial response is followed
by a much slower drift toward equilibrium at the final RH value.
This paper describes a laboratory method for determining the response time and behaviour of
different humidity sensors. Examples for results of the lab and outdoor measurements are shown.
The knowledge of the response time of the RH sensors as a function of temperature is needed for
the proper specification of these sensors.
Test sensors
Eight different capacitive polymer sensors were tested at DWD. Three sensors were heated
versions. All of them were used with the original filters recommended by the manufacturers and
with their factory calibration. Table 1 gives an overview of the tested sensors. Manufacturers
specifications are shown together with WMO/CIMO recommendations and DWD requirements.
Table 1:
Tested sensors with manufacturers specifications for the accuracy and response time in comparison to
WMO/CIMO recommendations and DWD requirements.
Relative
humidity
Range
measurement uncertainty
sensor time constant
WMO CIMO
0 - 100 %RH
1%RH reqiured [3%RH to 5%RH achievable]
1s to 10 s
DWD
3 %RH [0 - 100%RH, 45°C < TT < 0°C ]
5 %RH [0 - 95%RH, 0°C < TT < -10°C ]
8 %RH [0 - 85%RH, -20°C < TT < -10°C ]
15s
mela CPC 1/9-ME 0 - 100 %RH
2 %RH [5 -95 %RH, 10 - 40 °C]
< + 0.1 %/K [for <10°C, >40°C]
not specified
EE 08
0 - 100 %RH
not specified
EE 33*
0 - 100 %RH
2 %RH [0 - 90 %],
3 %RH [90 - 100 %]
0.03 %RH/°C [at 20°C]
1.3%RH+0.3*reading %RH [-15-40°C, <90%RH]
t 90 =<15s [20°C]
2.3%RH [-15-40°C, >90%RH]
Rotronic HC2-S3
0 - 100 %RH
1%
t 63 = 12 - 15 s [at 23°C]
Vaisala HMP45D
0.8 - 100 %RH
1 %RH [Laboratory, 20 °C],
2 %RH [Field 0 – 90 %RH],
3 %RH [Field 90 -100 %RH]
15s [at 20°C]
Vaisala
HMP155A*
HMP155D
0 - 100 %RH
1 % RH [0-90%RH, +15 - +25°C]
1.7 % RH [90-100%RH, +15 - +25°C]
1.0+0.008*reading %RH [-20 - +40°C]
20s [for 63% at 20°C]
60s [for 90 % at 20°C]
testo 6337 9742*
0 - 100 %RH
2,5 % RH [0-100%RH, 25°C]
not specified
* heated versions
2
Laboratory test
1. Accuracy at different constant temperature conditions
To investigate the temperature dependency of the measurement uncertainty the senors are tested in a
climate chamber at DWD. A benchtop two-pressure humidity generator (Thunder Scientific Model
2500 ST-LT [RH: 10 % RH to 95 % RH, T: -10°C to70°C]) is used as reference system. It is
humidity and temperature controlled and the expanded uncertainties are U(RH)=± 0.5%RH, U(T)=
± 0.06°C.
Three sensors can be tested at one time in a manifold placed in the centre of the chamber. The
temperatures investigated were -5°C, 5°C, and 20°C. The uncertainty of the temperature
measurement was determined to be ±0.2°C over the whole measurement range. RH was set
stepwise to 10%RH, 30%RH, 50%RH, 70%RH, 90%RH and 95%RH. The sensors are allowed to
stabilize down to ±0.3%RH within half an hour of measurement. Then the mean uncertainty for at
least 20 minutes was calculated.
Results
Figure 1 shows the results for the accuracy in stationary conditions at 20°C. The sensors show good
results below 90%RH with deviations between -1%RH and +2%RH. For humidities above 90%RH
only the HMP45D shows larger deviations up to -3%RH. The corresponding results for stationary
conditions at -5°C are different. For lower humidities of 30%RH and 50%RH the deviations to the
reference RH of the chamber are within the range of ±3%RH. For higher humidity values
considerable negative deviations appear for some sensors (see Figure 2). Two sensors showed
deviations of more than -5%RH at humidities above 70%RH. The deviation for the heated sensors
EE33 and testo 6337 and the non heated CPC1/9, HMP45D and HMP155D were within the
acceptable range of ±5%RH.
5
T chamber = 20°C
CPC 1/9
EE 08
4
EE 33
3
HC2S3
Difference to chamber [%RH]
2
HMP45D
HMP155D
1
testo 6337
0
-1
-2
DWD requirement ±3% RH
-3
-4
-5
20
30
40
50
60
70
80
90
RH chamber [%RH]
Figure 1: Accuracy of the humidity sensors at isothermal conditions [20°C]
3
100
5
CPC 1/9
T chamber = -5°C
4
EE 08
EE 33
3
HC2S3
Difference to chamber [%RH]
2
HMP45D
HMP155D
1
testo 6337
0
-1
-2
-3
-4
-5
20
30
40
50
60
70
80
90
100
RH chamber [%RH]
Figure 2: Accuracy of the humidity sensors at isothermal conditions [-5°C]
2. Temperature dependency of response time
The impact of temperature on the response time (transient performance) was tested with a special
temperature and humidity controlled chamber (see Figure 3). The components of the humidity
generator are shown on the left hand side of Figure 3, the temperature chamber is on the right side.
The humidity generator constists of a mixing unit to combine two gas flows, one with 100%RH the
other one with 0%RH, in a selectable ratio. A fast polymer humidity sensor in a small temperature
stabilized chamber (T > room temperature) is monitoring changes in humidity. Because of the
constant temperature and restricted humidity range the response time of this sensor is nearly
constant. After the mixing unit the gas tubes are heated to avoid condensation. The gas flow is then
transmitted into a sensor manifold inside a temperature chamber to vary temperature in a range of 30°C to 60°C. As the humidity generator is providing a constant dew point the relative humidity is
adjusted depending on the temperature.
Humidity Generator
Compressor
Flow
Controller
1
Pressure
controller
Flow
Controller
2
Gasdrying
apparatus
TT
TP
RH
RH, TT
0% RH
100%
RH
heated
Stabilized T
Sensor-manifould
with test-sensor
Dewpointhygrometer
Temperature Chamber
Water vapour
saturator
Figure 3: Configuration of the climate chamber at DWD for investigating the temperature dependency of response times
4
For response time tests the sensors were exposed to two "forward" steps in RH: 30-70-95%RH and
one "reverse" step 95-30%RH. The air flow rate and the temperature were kept constant during the
measurements. These tests were conducted at five different temperatures, at -20°C, -10°C, 0°C,
10°C, and 20°C to investigate the temperature influence on the response time. For calculating the
sensor response time we specified the tolerance to ±0.3% RH within one minute for reaching its
final RH value.
Results
2.1 Characteristic of response
While the specification of the time constant for capacitive polymer humidity probes implies that the
sensor has a first-order LTI system response with 1τ =1/e =63.2%, measurements of response times
of several sensors in our laboratory revealed a different behaviour. For some sensors it takes longer
to reach the 5τ =99.3% level of the final RH. Some sensors are faster. Therefore the sensor step
response seems to depend on the sensors polymer and the used filters. It was mentioned before that
non appropriate filters may influence the measurements by a time delay (see [3]).
For clarification Figure 4 shows the sensor step response of the EE33 and the HMP45D at
isothermal condition of 10°C and a humidity step from 30%RH to 70%RH. While the sensor step
response of the EE33 is close to the exponential function (see pink line) the sensor step response of
the HMP45D is faster at the beginning until around 80% of the final RH and slower at the end (see
green line in comparison with the exponential function blue line in Figure 4). Thus the declaration
of 1 τ indicates a faster response for the HMP45D. But the EE33 is faster in reaching the final RH.
To be able to compare different sensors it is proposed to define another time constant. It has to be
independent of the characteristics of the response and should indicate the time the sensor reaches
the desired humidity. For practical reasons the measurement time should be as short as possible.
Therefore a T97 constant could be an agreement. Table 1 gives an example for the sensors EE 33
and HMP45D.
Figure 4: Different sensor step response of two sensors: nonheated (HMP45D) and heated (EE33) version. Isothermal
condition at -10°C and a humidity step 30%RH to 70%RH compared to the exponential function.
5
Table 1:
Response time of the sensors EE33 and HMP45D for isothermal condition at -10°C and humidity step from
30%RH to 70%RH
sensor
EE 33
HMP 45D
1τ
54s
30s
5τ
411s
431s
T97
258s
274s
2.2 Dependency on humidity step
For some sensors the measurement uncertainty is differentiated by humidity. Therefore we also
investigated the response time for different humidity steps. The first step was chosen for dryer
conditions (30%RH to 70%RH), the second step for meteorologically more relevant higher
humidity conditions (70%RH to 95%RH) when condensation may be a problem.
As an example the results for the sensor HC2S3 are presented in Figure 5 for chamber temperatures
of 0°C. The response time is shorter for the step from 30%RH to 70%RH as for the step to higher
humidity from 70%RH to 95%RH. Our measurements revealed the same characteristic for all other
investigated sensors. The definition of response times for the sensors therefore also should denote
the valid humidity range. Additionally the reverse step from 95%RH to 30%RH is also shown in
Figure 5. As it can be seen it takes longer until the sensor starts to react but it reaches the final RH
earlier compared to the forward step from 70%RH to 90%RH (for results see also Table 2).
120%
Tcham ber = 0°C
% of final RH
100%
80%
60%
40%
HC2S3: step 30%RH to 70%RH
HC2S3: step 70%RH to 95%RH
20%
HC2S3: step 95%RH to 30%RH
0%
0
100
200
300
400
500
600
700
Response time [s]
Figure 5: Differences in sensor step response for different humidity steps: 30%RH to 70%RH compared to 70%RH to
95%RH an the reverse step from 95%RH to 30%RH. Results are shown here for the sensor HC2S3 at 0°C.
Table 2:
Sensor step response [s] of the sensor HC2S3 at 0°C.
time [s]
starting point
1τ
2τ
T90
T95
T97
5τ
6τ
sensor step
30%RH to
70%RH
7
66
135
160
222
270
367
392
70%RH to
95%RH
10
148
331
386
493
549
623
642
95%RH to
30%RH
38
261
336
358
414
461
585
613
6
2.3 Dependency on temperature
In Figure 6 and 7 the sensor step response time T97 for four sensors is presented over the whole
investigated temperature range (-20°C to +20°C) and for both humidity steps (30%RH to 70%RH
and 70%RH to 90%RH). The results for the non heated sensors (HMP45D, CPC 19) and heated
versions (testo 6337, EE 33) are shown. The curves T97 (T) look very similar for the different
humidity steps. Generally the step from 70%RH to 95%RH is taking longer than the step from
30%RH to 70%RH. It could be pointed out that all tested sensors showed longer response times for
colder temperatures. The T97 increases from around 90s (T97 [20°C, 30%RH to 70%RH]) to around
500s (T97 [-20°C, 70%RH to 95%RH]).
900
testo6337:
70%RH to 95%RH
800
HMP45D:
70%RH to 95%RH
Response time [s]
700
600
HMP45D:
30%RH to 70%RH
500
testo6337:
30%RH to 70%RH
400
300
200
100
0
-35
-30
-25
-20
-15
-10
-5
0
5
10
15
20
Temperature [°C]
Figure 6: Results from laboratory tests: Sensor step response T97 over temperature range. Non heated (HMP45D)
compared to heated version (testo6337)
1000
900
800
Response time [s]
700
EE 33:
70%RH to 95%RH
600
CPC 19:
70%RH to 95%RH
500
400
CPC 19:
30%RH to 70%RH
300
EE 33:
30%RH to 70%RH
200
100
0
-35
-30
-25
-20
-15
-10
-5
0
5
10
15
20
Temperature [°C]
Figure 7: Results from laboratory tests: Sensor step response T97 over temperature range. Non heated (CPC 19)
compared to heated version (EE 33)
7
Field test
During the field tests the response of the sensors could be monitored under outdoor environmental
conditions. The same sensors as in the laboratory were used. Before installation the sensors were
first checked for accuracy.
Set up
Two different locations were used for field tests:
- the test field at our institute in Hamburg
- a meteorological station in the mountains (Wasserkuppe 950m altitude, 220 fog days/pa on
average)
In the winter (2009/2010) seven sensors were installed in two small multiplate screens with
additional forced ventilation (LAM630, Eigenbrodt). Because of icing conditions at the screens
during a long period of low temperatures (2009/2010), the screens were blocked with ice and the
results of the humidity sensors could not be used.
To avoid the creation of different local microclimates a single louvred wooden screen with natural
ventilation and artificial aspiration (“Giessener Hütte”) was built up and equipped with all sensors
(see Figure 7). The sensors are mounted in a circular set up with the sensor heads close to each
other, generating good comparability of the results. Additionally a Pt100 was installed close to the
humidity probes to measure the temperature inside the shelter. The characteristics of the screens
were eliminated and only the differences of the sensors could be investigated.
As reference system the Thygan (Meteolabor), a chilled mirror dew point sensor, is used. It is
installed near to the screen. To compare the results with the humidity sensors RH was calculated by
using the dew point of the Thygan and the Pt100 temperature inside the wooden sreen.
Figure 7: Set up of the field test at Wasserkuppe. All sensors were installed in one louvred wooden screen
Results
Because of icing conditions during the winter time measurement at temperatures below 3°C could
not be evaluated with our first test set up. In order to get relevant statistics for various situations,
8
more measurements are necessary especially at lower temperatures. However, the outdoor work is
in progress and will be continued during the winter 2010/2011.
As an example of the consequences of different response times a comparison of four different
sensors to the Thygan RH is presented in Figure 8. The visibility is also shown in reverse scale. The
average temperature was 6°C and fog was present until around 9:40. As it can be seen the sensors
demonstrate different response times. The fastest sensors react after 10 min, the slowest after 40
min. For a final conclusion more measurements in the field with extensive changes in temperature
and humidity are necessary.
0 visibility [m]
RH [% RH]
100
10000
95
90
20000
85
30000
80
EE 33
EE 08
40000
HC2S3
75
HMP45D
Thygan
visibility
70
9:30
9:50
10:10
10:30
time [h:min]
10:50
11:11
50000
11:31
Figure 8: Results from the test field Wasserkuppe. The RH from four sensors are shown compared to the Thygan RH.
The visibility is presented in reverse scale. The average temperatur was 6°C and fog was present until 9:40.
Discussion and Conclusion
For comparison and specification of capacitive polymer humidity sensors the exact knowledge of
accuracy and sensor step response is needed. In this study we investigated eight sensors, heated and
nonheated versions. The accuracy was tested at different temperatures and a laboratory set up was
developed to study the different response curves.
For all investigated sensors the deviation and the response time increase at lower temperatures and
higher humidities. Bloemink [4] showed this beaviour already for other sensors. In our study the
response time extent by decreasing temperature from around 90s (T97 [20°C, 30%RH to 70%RH])
to around 500s (T97 [-20°C, 70%RH to 95%RH]).
To characterize the response time three issues are important:
- the sensor step response do not behave like a normal exponential function for all sensors
- the response time depends on the RH step
- the response time is different for different temperatures
For these reasons the declaration of response time should include specifications for regarding the
RH and temperature range. As recommendation of response time a T97 value could be an
agreement to get better comparision between the sensors.
The results of this study will be used to develop an acceptance test for capacitive humidity sensors.
9
Acknowledgement
The authors would like to thank Martin Mair (ZAMG, Austria) for his valuable contribution to the
field set up.
References
[1]
WMO (2008) :Guide to Meteorologocal Instruments and Methods of Observation,
WMO-No. 8, 7th edition
[2]
Dooley, J.B; O´Neal, D.L. (2008) The transient response of capacitive thin film
polymer humidity sensors
[3]
van der Meulen, J.P. (1988) On the need of appropriate filter techniques to be
considered using electrical humidity sensors
[4]
Bloemink, H. (2008) Humidity sensors for land and maritime stations
10