IDept of Electrical & Computer Engineering, Texas Tech University, Lubbock, ... 2Dept of Civil & Coa;tal Engineering, University of Florida, Gainesville, FL 32611, ...
Highly Accurate Noncontact Water Level Monitoring using Continuous-Wave Doppler Radar l l 2 l 3 Guochao Wang , Changzhan Gu , Jennifer Rice , Takao Inoue and Changzhi Li I
Dept of Electrical & Computer Engineering, Texas Tech University, Lubbock, TX 79409, USA 2 Dept of Civil & Coa tal Engineering, University of Florida, Gainesville, FL 32611, USA
;
National Instruments, Austin, TX 78759, USA Abstract
-
A novel water level monitoring technique using
continuous-wave (CW) Doppler radar is presented. The CW radar
works
with
a
DC-coupled
baseband
that
allows
accurate monitoring of the slow-varying water level. Ray tracing model has been used to simulate and validate the feasibility of water level monitoring in the presence of ripple. The proposed technique is robust against clutter interference and the fluctuation of water ripples by properly choosing the carrier frequency. Experiment was carried out to monitor the water level in a rain barrel when the water was draining out. Both simulation and experiment shows that the proposed technique can accurately monitor the water level with a high accuracy in millimeter-scale.
Index Terms
-
Fig.
Water level monitoring, continuous-wave
1.
Radar ray tracing model for water level measurement.
Doppler radar, ripple, non-contact.
I.
INTRODUCTION
II.
The setup is shown in Fig. 1. The CW Doppler radar is placed over the center of water surface and transmits unmodulated single tone carrier towards the water. The variation of the water level, which results in the change of distance to the CW Doppler radar, would modulate the phase of the reflected signal. The displacement could be recovered if the reflected signal is captured and properly demodulated [3]. A model that combines water level simulation and microwave ray-tracing simulation was established for the study. The model mimicking the water behavior employs shallow water differential equations, which are derived from the conservation of mass and conversion of momentum. As shown in Fig. I, the square shape water surface is divided into equal grids. Each grid's current height is decided by its own and adjacent grids' mass velocity, potential vorticity at last discrete time and gravity. One falling object serves as excitation to create ripple that will spread out to the whole area. Its height, diameter and position (indicated by the offset on the X and Y-axis from the center of the surface) decide the ripple's mode. Ripple will be bounced back and forth within the boundaries. By carefully setting the boundary condition, the strength of ripple can be set to increase, decrease, or stay the same over time. Given the same excitation and boundary, the ripple mode can be precisely reproduced. The average water level is controlled by a separate factor, and the ripple is added onto the average water level. This
Precise water level monitoring is desirable in many applications, e.g. gauging the depth of liquid in a storage tank or a vessel, and monitoring the level of large water bodies in nature such as ocean, river or dam. Microwave radar offers an attractive solution for noncontact distance and displacement measurement. The existing radar techniques for water level monitoring mainly include FMCW radar [1] and pulse radar [2]. However, FMCW radar suffers from system complexity and is vulnerable to the clutter interference from surrounding stationary subjects, e.g. wall of the liquid tank or reef, which may lead to false beat frequency. Pulse radar, on the other hand, requires short pulse width and high-speed switches, which hinders the system integration and increases the system complexity and cost. CW Doppler radar sensor, owing to simple structure, low cost and high level of system integration, has been used for displacement measurement when operating in an interferometry mode [3, 4]. It is free from the stationary clutter interference problem of FMCW radar since it is based on coherent detection of phase change, and the reflection from stationary subjects only contributes to a DC offset. However, conventional AC-coupled Doppler radar does not work well when monitoring slow movement such as water level change, due to the signal distortion problem [5]. In this paper, a DC-coupled CW radar is used to accurately monitor the slowly varying water level.
978-1-4673-3105-0/13/$31.00
©
2013 IEEE
THEORY AND MODELING
19
WiSNet 2013
o �
oS
�
:----,---. ---r ,-.;::----------
-2
-3 iii - 4
Qj
-l
�
Fig.
r
'"
_
-5
:.
__ .....r
-1
2.
/
� :\ � / ,
I ' ,
' ..... ,
: \IV 1 0'00
,
-
_
_
�
.. 4 3
-
,
2
,
------±--
- water --- error
200 ' 0
....
�a-�, -.. ,,, .. : : : ,: level
I
0-5r--;----:---;---;:====;]
e
�
o
> Q) -l
-0_5
.$ �
W
oJ
�
:
40'00 50 ---'- 0 - 0-----' 1
30'00
Time IS]
Fig.
Simulation of water level monitoring and measurement
3.
No Ripple
- +-
No Wale level
-1 Ilrcnl'dmv
: - -
, ,
I I
- - - - -I - - - --
Ripple + N,o water lelfel lirmv{/oJVIl
1- - - - - -J - - - -
e l'' _----' 5 _ ---"D"--'ro"':''' -=--1. 0� 200 400
L-__-'---_ - _---'-_ --- -l
___
600 Time [sl
800
1000
Simulation of water level monitoring under different
ripple strengths.
error.
significant, leading to a smaller accumulated measurement error. Despite this error, the CW Doppler radar shows a satisfactory performance in water level monitoring by providing millimeter-scale accuracy.
model facilitates the analysis of the interaction between electromagnetic wave and realistic water behavior, which has not been thoroughly studied before. In the ray tracing simulation, by tracking and combining the modulated signal reflected from each grid, the total baseband output can be obtained as:
B(t)
E
Q)
E oS
=
� exp { J'. [-4lrdo 4lrd (t) - -L.,
A
n=1
4lrrn (t)
A
A
B.
The existence of ripples acts as clutter noise to the demodulated radar signal. Since ripple is hard to avoid, it is necessary to investigate the effect of ripples on the water level measurement. Fig. 3 compares two simulations with the same conditions except for different ripple strength. The operating frequency was 1 GHz. At the 150th second, a falling object with a diameter of 7 cm would hit the water surface at the position with 10 cm offset on both axis, as shown in Fig. 1. The water level began to drop from the 600th second at a speed of 0.2 mmls. For the small ripple case, the falling object fell from 1 meter but for the large ripple case, it fell from 3 meters. When the ripple was small, (time average RMS was 0.01 m), it causes less error and the slope of the measured water level trace agrees well with the ground truth of dropping rate. However, when the ripple was large (time average RMS was 0.04 m), the ripple introduced significant error to the measurement.
2 do +rp(t--)]} c
(1) where d(t) is the water level movement shared by every grid, rn(t) is the ripple movement of the nIh grid, A is the wavelength, and do is the initial distance between the radar and the water surface. To avoid the null detection point, the total received signal is converted into IIQ channels for arctangent demodulation [6]. Phase unwrapping is applied to reconstruct the demodulated phase signal when the water level motion exceeds half of a wavelength [3]. III.
Ripples in water level monitoring
SIMULATION RESULTS
In simulation, the water surface was set as 50 cm X 50 cm. The initial distance between the antenna and the water surface was 1 m, and the projection of the radar was at the center of the water surface. The boundary condition was set so that the ripple will stay constant once it is generated.
C.
Operating frequency
According to (1), the radar detection resolution improves as the operating frequency increases. However, higher frequency causes a potential problem: the random ripple motion may cause the phase modulation to go beyond 360 degree when the ripple's amplitude exceeds half of a wavelength. Such a situation would result in false signal demodulation due to the difficulty in frequent phase unwrapping compensation. One straightforward solution is to decrease the operating frequency to ensure that the phase modulation caused by water ripple does not exceed 360 degree. To better illustrate this issue, Fig. 4 shows three simulation scenarios with the same water level motion. First, the simulation was carried out without any ripple, serving as a reference. For the other two scenarios, ripples were introduced by the same falling object from 4 m with a diameter of 6 cm. For the simulation with a 2
A. Accuracy of water level detection The accuracy of the CW Doppler radar was evaluated by monitoring the tranquil water surface. Without any ripple, a comparison between the detected water level and the ground truth is used to evaluate the radar accuracy. The operating frequency was set to 3 GHz and the pre-set water drawdown speed was 0.7 mmls. Fig. 2 shows the detected motion of water level along with the detection error. The slope of the detection trace equals to the pre-set dropping rate while the error is within several mm . The increasing error was mainly caused by the amplitude modulation: as the distance between radar and water surface increases, the antenna gain distribution at each grid changes as well. As the distance becomes larger, the changes of antenna gain distribution become less
20
0.5 I
E Qi
> OJ --l
�
-0.5
�
C\l
�
-1 -1.5
Fig.
I
0 ����������---�-
4.
r
;
•
•• ··,-to.· ···]···"·· •• Droplet
0
. . .
200
firm down
400
600
. .
800
Time [s]
1 000
1200
'
1400
1600 Fig.
Simulation of water level monitoring under different
5.
operating frequency
Experiment Setup with simplified block diagram 20 .-----------.--------- --,--===��==
GHz carrier frequency, the measured result deviated a lot from the reference, because the ripple severely distorted the phase unwrapping process. However, for the case of 500 MHz carrier, the effect of ripple was suppressed and the detected motion was close to the reference. Therefore, in practice, it is necessary to balance the tradeoff between the detect resolution and water ripple suppression.
Time [s] Fig.
6.
Experiment result: (a) time domain IIQ signals and
(b) constellation graph.
IV. EXPERIMENTAL RESULT
system robust to water ripple problem. The well agreement between the experiment and simulation proves the practical value of this technique.
To verify the feasibility of using CW radar for water level monitoring, an experiment was carried out in the outdoor environment by using a rain barrel with a radius of 30 cm. In the experiment, a DC-coupled CW Doppler radar working at 3 GHz was placed 50 cm above the water surface as shown in Fig. 5. No external excitation but the environmental wind introduced a small amount of ripple. After the faucet at the bottom of the barrel was turned on, the water level began to drop, which was monitored by the radar. The experiment result is shown in Fig. 6. The reason that the measured water level trace was not straight is that the water drawdown speed decreased as the water level dropped, due to the decreasing water pressure. Insets (a) and (b) are the time domain IIQ signals, and the constellation of the phase trajectory respectively. As the water level dropped, the reflected signal strength decreased as well, which resulted in the shrinking amplitudes for both IIQ channel signals and caused the helical trajectory in the constellation graph. The positions of the water level were also manually recorded to provide the reference. The experimental result shows that the water level measured by the radar dropped by 65.86 cm, which matches well with the reference water level drop of 66 cm.
ACKNOWLEDGEMENT
The authors would like to acknowledge NSF for grant support (CMMI 1131506), and acknowledge National Instruments for providing funding support as well as RF instruments and AWR design software. REFERENCES
[I]
G. S. Woods, D. L. Maskell, M. V. Mahoney, "A high accuracy
microwave
application",
[2]
Measurement, vol.
R.
M.
ranging
IEEE
system
Trans.
for
industrial
Instrumentation
and
42,no. 4,pp. 812-816,1993.
Heitsenrether,
M.
H.
Bushnell,
J.
D.
Boon,
"Understanding the impact of surface waves on microwave Water
[3]
Level
Measurement,"
MTSIIEEE
S.
Kim,
technique
C.
Nguyen,
using
"A
displacement
millimeter-wave
1724-1728,2003.
L.
Lu,
C.
multifunctional
Li,
J.
radar
A.
Rice,
sensor
for
1-8.
"A
08
measurement
interferometry",
Trans. Microwave Theory and Tech., vol.
[4]
Oceans'
International Conference, Quebec, Canada, pp.
IEEE
51, no. 6, pp.
software-defined
linear
and
reciprocal
displacement measurement," IEEE Topical conference on
VII. CONCLUSION
Wireless
The simulation based on the ray-tracing and shallow water model verifies that continuous-wave Doppler radar with DC-coupled baseband is capable of monitoring the motion of water level. This technique is free from the static clutter interference and can achieve mm-scale accuracy. The adaptive operating frequency makes the
[5]
January
Sensors
2011.
and Sensor Networks,
Phoenix, AZ,
C. Gu, C. Li, "DC coupled CW radar sensor using fine tuning adaptive feedback loop," Electronics Letters, vol.
[6]
no.
48,
6,pp. 344-345,March 2012.
B. K. Park, O. Boric-Lubecke, V. M. Lubecke, "Arctangent demodulation with DC offset compensation in quadrature Doppler radar receiver systems," IEEE Trans. Microwave Theory and Tech. vol.
21
55,no. 5,pp. 1073-1079,2007.
