A 16-Electrode Biomimetic Electrostatic ImagingSystem for Ocean Use

Jonathan Friedman, Henry Herman, Newton Truong, Mani B. SrivastavaNetworked and Embedded Systems Laboratory

University of California, Los Angelesjf8, hherman, newtontruong, mbs@ucla.edu

Abstract—A compilation of our latest design, simulation, and eval-uation efforts regarding the construction of a biomimetic electrostaticimaging platform capable of visualizing submerged objects in a salt-water environment. In our previous work, the position accuracy was nobetter than 10cm. The imager described here achieves 2.5cm accuracy onthe center line and degrades to 5cm accuracy at the peripheral channels.

I. INTRODUCTION

Nocturnal ocean animals and those that live at depth are unableto employ light-dependent forms of perception to navigate theirsurroundings, hunt, or avoid predators [3]. Rather, the primarymeans of perception involves the passive detection of electrical fieldsnecessarily created by muscle activity in their prey (a process similarto the human electrocardiogram). In some species, the process isan active one in which the generation of a Voltage gradient in theintermediate environment enables the discernment of non-emissivenavigational hazards (such as rocks) by the manner in which theydisturb the self-generated field [8].

Electroreceptivity is achieved through a dense grid of electric field(Voltage) sensors which are concentrated heavily around the headof the fish and tightly arrayed along either side of the body [4].Anatomically, each sensor is constructed of a glycoprotein gel-filledtube with nerve endings at either end [10][3]: altogether this systemof sensors is labeled the Electro-Sensory Organ (ESO) and is usedin the detection of active, e.g. emissive, targets. The ESO, by itself,is unable to perceive non-emissive objects [1]. The inclusion of anElectric Organ (EO) rectifies this problem. The EO, through thedirect conduction of current generated in the EO, establishes theaforementioned Voltage gradient in its surroundings, an event whichis known as an Electric Organ Discharge [11].

To comprehend the flow of current around the fish in the ocean,one can picture two electrodes placed at either end of a wide flatconductive plate. Once the plate is energized, the majority of thecurrent will flow from cathode to anode, with little or no deviationfrom the most direct path between the electrodes [15]. However,as current in any direction is comprised of like-polarity charges, arepulsive force exists between them causing some charges to take anindirect, more circuitous route. The current is following the isolinesof the charge concentration gradient created by the electric fieldestablished by the electrodes [7].

If an object less conductive than the ocean water (rocks, plastics,etc) enters the field, the current, following the path of least resistance,will curve around it. The redistribution of current necessarily spreadsout the field lines which changes the location of the isovoltaic lines’intersection with the fish’s body: it effectively casts an electrical”shadow” by creating a region where the Voltage is more constant perunit distance along the body [8] [2] [11]. Objects more conductivethan the surrounding water have the opposite effect; they concentratethe field lines, establishing an electrical bright spot – a region of

1 0.5 0 0.5 11

0.8

0.6

0.4

0.2

0

0.2

0.4

0.6

0.8

1

Fig. 1. An example field analysis showing the transmit electrodes as blackX’s, the receiving electrode array as red squares, and the poles of the induceddipole as blue circles. A dotted line is drawn horizontally through the figure tohelp visualize the observations made by a linear array parallel to the transmitaxis.

rapid Voltage change per unit distance along the body. As such, whenelectroreception is an active sensor, this mimetic system can not onlydetect objects and their location, but may classify them as well.

In this work, we summarize our latest experimental findings increating an engineered sensor which mirrors this biological systemand demonstrate its ability to visualize targets with different con-ductivities from the background ocean environment – a process weentitle Biomimetic Electrostatic Imaging (BEI) [5]. Our most recentpublished work with BEI involved a two-channel prototype. It wasable to construct images of the disturbance to a generated electric fieldin salt-water in a manner consistent with its modeled physiology. Itwas demonstrated to be capable of detecting a conductive steel pipetarget in an ocean simulating salt-water tank at a range of more than20cm with an overall initial accuracy of at least 10cm. Here, wediscuss the design and development of our 16-channel prototype anddemonstrate its performance improvements over the prior 2-channeldesign [5].

II. OPERATIVE THEORY

Electrostatics, as the name implies, involves the use of electric(E) fields which are, traditionally, invariant with time [6]. For ourpurposes it is illustrative to consider a sequence of time-invariantfields with each successive field having more (then less, then more,etc) strength than its predecessor. When this approximation is valid,the field is said to be quasi-static [13].

II.A. Validity of the Electrostatic Assumption

In order to prove the validity of this assumption, consider thatwhen the electric field, and hence the current flowing in the field,

978-1-4244-9289-3/11/$26.00 ©2011 IEEE

Fig. 2. A model of two sensor channels (cyan and blue) being disturbed byan induced dipole. The red X’s are the the transmitting electrodes.

Fig. 3. A detail view of the receiver channel located at the midpoint betweenthe two transmitting electrodes. Only the center part of the response is shown.

changes with time two currents must be considered – the conductioncurrent and the displacement current.

The conduction current is Ohmic [7] resulting from the movementof charges between atoms1, while the displacement current resultsfrom the movement of bound charges within an atom (typicallycaused by the application of some external field) [15]. It is thisdisplacement current, which is responsible for electromagnetic (EM)radiation.

The attenuation of EM waves in any conducting medium increasesboth with an increase in conductivity and an increase in frequency[16]. Within water, increasing the salt content (salinity) increasesconductivity and its attendant EM attenuation. It can be approximatedfrom (1) where α is attenuation in dB per meter, f is frequency inHertz, and σ is conductivity in Siemens per meter [14]:

α = 0.0173√fσ (1)

As is evident from (1) (given that σ ≈ 4.8 S/m at 20oC [17]),attenuation in sea water is very high. Even a very low frequency(10kHz) signal (α ≈ 3.8) will lose more than half of its originalpower each meter! For this reason, it is the Ohmic conduction currentwhich plays the key role in electroreception. For a more in depthdiscussion see [6] [13].

II.B. Modeling Electrostatics

Having established the operating bounds for the validity of theelectrostatic assumption, we may now introduce what is effectively

1...or the movement of charged atoms (ions) in space, as is common inbiological systems

a static (DC) field into the ocean environment. An electric field ina conductive environment will necessarily result in current flow. Werefer to the electrode pair utilized to establish this field as the transmitelectrodes. Detection is performed differentially through adjacentequally-spaced electrodes arranged collinearly with the transmitters(the receive electrodes).

In [6], we introduced our purpose-built physical modeling engine,built in Matlab, used to visualize the electric fields and disturbancethat occurs with environmental objects. The work demonstrated theeffect of a conductive object (which presents as a dipole) on a univer-sally referenced field (a single-ended measurement). However, whenthe sensor was actually constructed [5], differential measurementswere used. In this work we have extended our model to evaluatefield disturbances differentially.

The model works by ignoring the homogenous background en-vironment and restricting analysis to only the relevant net charges(with respect to the background ion concentration). Direct analysisthen becomes computationally practical via2:

V oltage =

Nc∑n=0

1

4πε0

qnrn

(2)

Where Nc is the number of net charges in the environment to considerand qn expresses the net magnitude of each charge. rn is the distancebetween the spatial location under consideration and each specific netcharge. ε0 is the permittivity of a vacuum. The equation may be scaledby the relative permittivity of other media for consideration there, butas this has no spatial consequences (only influences magnitude), hereit is ignored.

Figure 1, is an example field analysis showing the transmit elec-trodes as black X’s, the receiving electrode array as red squares, andthe poles of the induced dipole as blue circles. The actual units andmagnitudes are unimportant, the purpose is to illustrate the isovoltaicfeatures at scale. A dotted line is drawn horizontally through thefigure to help visualize the observations made by a linear array inthe field which is parallel to the transmit axis.

II.C. Basic Assumptions of BEI

In the specific case of making our electric field disturbancedetections, some additional constraints are required:

The diameter of the dipole must be much smaller than the transmitbasis (the minimum distance between two transmit electrodes). Theend effect is that the solution space must be sparse. It may containmultiple small-diameter objects so long as they are unclustered. Ifthe dipole were modeled as a substantial percentage of the transmit-basis then the receiver array would experience the dipole as a set ofsmaller dipoles all interacting (as we are in the near-field).

The transmit-basis must remain much larger than the separationbetween receive electrodes used in the differential measurement.Increasing the receiver separation reduces the spatial resolution of themeasurement and in the limiting case (electrodes very far apart whencompared with the transmit electrodes), the differential measurementdevolves to the single-ended case.

II.D. BEI Expectations

With these restrictions in place, a series of analyses were performedto identify the expected behavior of a differential field disturbanceprobe (a single channel composed of two individual electrodes over

2This model is only partially correct as the presence of the induced dipolewill, in turn, affect the other charges in the vicinity, but we may ignore thesehigher-order effects for now as we concentrate on establishing the basic BEIimage formation theory.

Fig. 4. The experimental configuration consisting of (from left) an instrumentrack (sitting on the table), a custom 2-axis motion control platform with thewater tank installed below, and a control station. (Inset) The stainless-steelpipe used as the target.

which a potential difference is recorded). In each of the figures, thered X’s indicate the location of the transmitting electrodes, the coloredcircles indicate the center location of the receiving electrodes (whichwere located equidistant to either side of the marker), and the coloredcurve families represent the observations of the correspondinglycolored receive channel. The X and Y axes form a plane in real spacewhere the dipole is physically located at the time the Z value for fieldpotential is observed. The spatial coordinates are units of relativedistance normalized to: transmit-basis = 0.4 units. The amplitudevalues are relative to their far-field zero, but lack significance as theactual magnitude will depend on the strength of the induced dipole.

Figure 2, shows the results from two receive channels, one locatedat the origin and one to its left. For each channel, three curvesare plotted with the decrease in amplitude corresponding to theincreasing distance of the dipole away from the transmit-axis (intothe page). Several observations should be readily apparent from thefigure: (1) The disturbance field is limited in extent and will notinfluence all of the receive channels if the array is large. (2) Thedisturbance field is centered on the receive channel which observedit. Put more simply, the peak sensitivity of the channel is to dipoleslocated directly in front of it. (3) This is a near-field phenomenon.The disturbance quickly returns to baseline with increasing distancebetween the dipole and the sensor. (4) There is a baseline value uniqueto the channel which is determined by the transmit electrode spacing,receiving electrode spacing and position3, output power, and medium.

Figure 3 explores the sensing range in greater detail. The parabolicnear-field behavior results from the dipole’s diameter being equalto the receiver electrodes’ separation and vanishes with increasingdistance (the dipole diameter appearing smaller to the receivers). Thesignal strength vanishes inversely proportional to the separation (e.g.∆V α 1/d).

III. EXPERIMENTAL SETUP

To establish and verify oceanic verisimilitude, experiments wereconducted in a 55-gallon (208 liter) capacity tank filled with 50gallons (189 liters) of drinking water (0ppt salinity). Salt in the formof Sodium Chloride was added in measured amounts to vary the

3relative to the transmitting electrodes

salinity of the test vessel. A Marineland Labs, inc. Instant OceanHydrometer [12] was used to record the salinity. The actual salinityof the ocean varies with position, depth, temperature, and seasonover a range of 28ppt4 to 35ppt [12] in open ocean and as high asan average of 39ppt in the Mediterranean and Persian Gulf [9]. Afinal value of 28ppt was achieved representing the extreme low end(worst-case) of Pacific Ocean conditions.

III.A. The Connections

The experimental configuration is shown in figure 4 which consistsof (from left) an instrument rack (sitting on the table), a custom 2-axismotion control platform (1 meter x 1 meter of travel area maximum)with the water tank installed below, and a control station. Thetransmit electrodes were connected to an Agilent 33220A ArbitraryWaveform Generator (AWG), which was used to deliver a 2Vpp5

sine wave excitation at 1,085 Hz through a 50.0Ω source resistance.The receive electrodes were connected through an Analog Front End(AFE) to an Agilent 33401A Digital Multi-Meter (DMM), which waspowered from an Agilent E3631A Power Supply delivering ±2.5V .The electrodes were submerged to a depth of 14cm (≈ 5.5”). Duringthe trials the laboratory air temperature was 26.1oC

Inset in figure 4, in the lower right corner, is the stainless-steelpipe used as the target. The pipe is mounted to the motion controlplatform and moved about the tank as recordings are taken from thereceive electrode pairs. The X-Y planes of figures 2 and 3 representthe value recorded when the pipe was in the indicated position.

III.B. The Analog Signal Path

Powerline, power supply, nearby processors, airborne radio signals,and other sources of electrical noise can severely hamper the high-impedance, high-gain measurements required for artificial electrore-ception. A six stage Analog Front End (AFE) was designed for ourBiomimetic Imaging System to allow it to disinterr the differentialdisturbance field signal. In order, they are: (a) platinum electrodesexposed to the tank’s salt water, (b) a bridge bias balance network torecenter the signal after AC-coupling, (c) an instrumentation amplifierto provide gain, bandwidth limiting, and conversion from differentialto single-ended signaling6, (d) an extremely high-quality (Q > 30)active Band-Pass Filter (BPF), (e) a passive Low-Pass Filter (LPF),and, ultimately, (f) a final gain stage provided as part of the dataacquisition unit – an Agilent 33401A.

A complete AFE handles two electrodes as one differential pair.For the experiments 8 differential pairs (16 electrodes) were used andtherefore 8 AFE’s were required. Only two physical AFE’s were built.To achieve the eight required the two AFE’s were repositioned againsta static environment emulating the existence of multiple simultaneousAFE’s.

III.C. The Electrodes

The material design of the electrodes is critical as it establishesa practical upper-bound on achievable performance. Prior work withsteel electrodes [5] revealed a number of operational problems dueto its iron content [6]. Reactions with the ocean water can resultin electrical noise (unstable capacitance, long-term ionization, etc),sensor degradation (ex. oxidation, producing a substantial increase

4Parts Per Thousand (ppt) by mass ratio with respect to water.5Volts peak-to-peak6Since the AFE is duplicated for each channel, minimizing the amount

of hardware is important to help control costs. Processing analog signalsrequires about half the hardware when the signals are single-ended, ratherthan differential.

Fig. 5. Two representative channels of the sixteen electrode (8 channel)imager recording a metal pipe target swept through a salt-water tank.

in resistance at the interface), and even device failure (the electrodedissolving in its entirety).

In this work the steel content of [5] was replaced with 99.9% pureplatinum. The reduction potential for platinum is much lower thanthat for iron and, thus, can endure a much higher voltage withoutreacting with the salt water.

The electrodes are cylindrical in shape and made as small as possi-ble to best emulate a point source in the water. For practical reasonsof handling, durability, and ease of construction, the electrodes werefashioned from 30 gauge (AWG) 99.9% pure platinum wire withfinal assembled dimensions of 267µM in diameter and 3600µM inlength. The top surface was sealed to the feed system resulting in anelectrode working surface area of 3mm2.

An acrylic frame was fabricated to fix the relative positions of twoelectrodes. Each was placed collinearly ±2.54cm from the centerpoint to form a differential receiver basis of 5.08cm. The framewas placed flush against an acrylic plate to eliminate the back lobe(directivity is covered in more detail in [6]). The transmit-basis was10.16cm.

IV. RESULTS

Two representative channels of the sixteen appear in figure 5. Theupper trace is from the center channel (indicated by a lighter/yellowdot on the x-axis) and the lower trace is from one of the left-sidechannels (it’s true position indicated by the darker/blue dot on the x-axis). The plot compares quite favorably with the expected featuresof figure 2. Most notably, the centers are in the correct locations,the pattern has the peaked oscillatory shape, the pattern featuresremain recognizable across channels as the center moves, the patternis limited in extent and returns to baseline, and the baselines areunique per channel.

V. CONCLUSION

In this work, we have presented our latest design, simulation,and evaluation efforts regarding the construction of a biomimeticelectrostatic imaging platform capable of visualizing submerged

objects in a salt-water environment. In our previous work, theposition accuracy was no better than 10cm. The imager describedhere achieves 2.5cm accuracy on the center line and degrades to5cm accuracy at the peripheral channels. The output power remainsunchanged between design iterations. Curiously, this imager showslittle signal degradation over the 20cm of range tested. This is likelya result of the limited volume of the water tank. The walls of thecontainer structure serve to focus and enhance the field – much like afish’s body would. Our next investigation will involve a more infinitefar-field to better determine range performance.

ACKNOWLEDGEMENTS

The authors would like to express their sincere gratitude to JessicaHorvath for her editing and revision of this manuscript and to TannerMiller for his assistance during data collection.

This material is supported in part by the U.S. Office of NavalResearch under MURI Award CR-19097-430345 and the UCLACenter for Embedded Networked Sensing. Any opinions, findings,conclusions, or recommendations expressed in this material are thoseof the authors and do not necessarily reflect the views of the listedfunding agencies.

REFERENCES

[1] E. G. Barham, W. B. Huckabay, R. Gowdy, and B. Burns, “Microvoltelectric signals from fishes and the environment.” Science, vol. 164, pp.965–968, May 1969.

[2] T. H. Bullock, “The future of research on electroreception andelectrocommunication.” J Exp Biol, vol. 202, pp. 1455–1458, 1999.[Online]. Available: http://cogprints.org/106/

[3] D. H. Evans, The physiology of fishes. ISBN 0849384273, 1997.[4] L. Fishelson and A. Baranes, “Distribution, morphology, and cytology

of ampullae of lorenzini in the oman shark, iago omanensis (triakidae),from the gulf of aqaba, red sea.” The Anatomical Record, vol. 251, pp.417–430, Jan. 6th 1999.

[5] J. Friedman, D. Torres, Y. H. Cho, and M. B. Srivastava, “Submergedbiomimetic electrostatic imaging in salt water,” The 9th Annual IEEEConference on Sensors, Nov 2010.

[6] J. Friedman, D. Torres, T. Schmid, J. Dong, and M. B. Srivastava, “Abiomimetic quasi-static electric field physical channel for underwaterocean networks,” ACM Workshop on Underwater Networks (WUWNET),2010.

[7] D. Halliday, R. Resnick, and J. Walker, Fundamentals of Physics, 5thed. John Wiley and Sons, Inc., 1997.

[8] B. Kramer, “Communication behavior and sensory mechanisms inweakly electric fishes,” Advances in the Study of Behavior, vol. 23, pp.233–270, 1994. [Online]. Available: http://epub.uni-regensburg.de/2655/

[9] O. C. Laboratory, World Ocean Atlas. U.S. National OceanographicData Center, 2005.

[10] S. Lorenzini, The curious and accurate observations of Mr. StephenLorenzini of Florence. Printed for Jeffery Wale, 1705.

[11] M. A. Maciver, N. M. Sharabash, Mark, and M. E. Nelson, “Prey-capturebehaviour in gymnotid electric fish: motion analysis and effects of waterconductivity,” J. Exp. Biol, vol. 204, pp. 534–557, 2001.

[12] I. Marineland Labs, “Instant ocean hydrometer,” Product Catalog, 2007.[13] H. Momma and T. Tsuchiya, “Underwater communication by electric

current,” Oceans, IEEE Journal of, pp. 631–636, Sept 1976.[14] R. R. Moore, “Radio communications in the sea,” IEEE Spectrum, vol. 4,

pp. 42–51, Nov 1967.[15] S. B. Palmer and M. S. Rogalski, Advanced University Physics. Taylor

and Francis, 1996.[16] M. Siegel and R. King, “Electromagnetic propagation between antennas

submerged in the ocean,” Antennas and Propagation, IEEE Transactionson, vol. 21, no. 4, pp. 507–513, Jul 1973.

[17] K. K. Turekian, Oceans. Prentice-Hall, 1968.