Table 2b. Statistical summaries: All RC network
measure the dielectric constant of a known liquid.
tests, spring and autumn data, both probe circuits
This set of tests required the addition of the probes
for each box.
to the test system; it demonstrates the use of a
statistical approach for this method for convert-
Standard deviation as
ing the capacitance as measured by the system to
Box
Percent error
a percent of the mean
Nulled?
a dielectric constant measurement. (Previous re-
2
3.0
5.4
Yes
ports describing the use of this measurement tech-
2
39.6
14.8 (outlier ?)
Yes
nique reported capacitance as the output data but
2
20.0
0.3
Yes
never reported dielectric constants or suggested a
3
36.7
2.5
Yes
3
53.8
4.4
Yes
method for converting the capacitance measure-
3
25.6
2.4
Yes
ments to a dielectric constant.) In addition, as part
4
1.1
0.3
Yes
of the test procedure, the center tine of two of the
4
50.0
3.8
Yes
test probes was covered with heat-shrink tubing
4
12.1
0.3
Yes
to eliminate ionic conduction between the
2&4
11.1
2.7
Yes
2&4
28.8
4.8
Yes
capacitor's plates. This modification to the other-
2&4
62.1
4.6
Yes
wise bare tines was a test program to determine if
3
40.1
1.8
No
the salinity problem (see the discussion on page
3
38.4
0.8
No
3) could be solved. The concept was to prevent
3
68.3
1.5
No
highly conductive saline soils from shorting out
Average error of all nulled circuits: 28.6 %
the capacitance measurement. The error in the
Average error of all unnulled circuits: 48.9 %
capacitance measurement could then be corrected
Average standard deviations as a percent of the mean:
as a part of the calibration process.
all nulled circuits: 2.9 %
all unnulled circuits: 1.3 %
It should be noted that the probes initially sup-
plied for these tests had BNC connectors mounted
An examination of these data shows that while
directly in their base so the connecting cable was
the electronic circuits themselves are not very ac-
attached at that point. These probes were designed
curate for the measurement of capacitance, they
to be inserted into the soil for a surface measure-
are nevertheless highly repeatable and reliable.
ment, so the BNC cable connection was perfectly
This conclusion indicates that the circuits will work
adequate. However, the BNC connector was not
very well for soil moisture measurements once a
watertight, so for probes that were to be buried
calibration has been performed. Another tenta-
and left in the soil for an extended period it be-
tive conclusion that might be drawn from the
came necessary to replace the BNC connectors
statistical data is that the best accuracy is achieved
with a permanent, sealed, waterproof connection
when the cable's capacitance is nulled (note that
between the probe and cable. The sealing com-
the tabulated summaries include all the unnulled
pound used for this probe modification was a
portion so that the actual sample population for
two-part encapsulating epoxy. The requirement
the unnulled tests is 24 points, not the three val-
to follow this procedure further complicated the
ues listed in Table 2b). Better accuracy might be
measurement process since it dedicated each probe
expected from a nulled circuit since the correction
to a given cable length and also introduced an
for cable capacitance in the unnulled circuit can
additional dielectric constant, that of the epoxy
be as large as the measured value itself. However,
into the measurement probe circuit.
note also that there is apparently no significantly
The method for converting a capacitance mea-
valid reason to assume that the unnulled test data
surement to a dielectric constant as described in
is less repeatable than the nulled circuit data.
the Capacitance Measurements section is summar-
ized below:
1. Determine the capacitance of the probe in
DIELECTRIC CONSTANT
air (eq 1 and 3).
MEASUREMENTS
2. Take the soil measurement and calculate
the capacitance (eq 1).
Using uninsulated probes
3. Divide the soil capacitance by the air ca-
After the initial all-electronic measurements
pacitance; the resultant quotient is the di-
were completed, a series of tests was conducted
electric constant of the soil (eq 2).
to determine how well the technique described in
the Capacitance Measurements section above would
Note that the air capacitance is most easily mea-
7
Previous Page
