Wireless Communication of Intraoral Devices and Its Optimal Frequency Selection

Abstract

This paper explores communication methods and frequencies for wireless intraoral electronic devices, by using an intraoral tongue drive system (iTDS) as a practical example. Because intraoral devices do not meet the operating conditions of the body channel communication, we chose radio frequency communication. We evaluated and compared three frequencies in industrial, scientific, and medical bands (27 MHz, 433.9 MHz, and 2.48 GHz) in terms of their data link performance based on path loss and radiation patterns over horizontal and vertical planes. To do so, we dynamically minimize the impedance mismatch caused by the varying oral environment by applying the adaptive impedance matching technique to 433.9 MHz and 2.48 GHz bands. Experimental results showed that 27 MHz has the smallest path loss in the near-field up to 39 cm separation between transmitter and receiver antennas. However, 433.9 MHz shows the best performance beyond 39 cm and offers a maximum operating distance of 123 cm with 0 dBm transmitter output power. These distances were obtained by a bit error rate test and verified by a link budget analysis and full functionality test of the iTDS with computer access.

I. Introduction

In recent years, considerable attention has been devoted by researchers to active intraoral devices that monitor medically relevant measures in the mouth, infer user intentions, or deliver information to the user through oral sensory modalities [1]-[3]. While some of the intraoral devices are still hardwired [4]-[7], to communicate with the outside world, the preferred mode of communication for the intraoral devices is wireless. However, wireless communication between intraoral devices and the outside world is challenging because the transmitter, often located inside the mouth, experiences additional losses that do not exist in typical wireless communications in the air. Although wireless communication has been successfully established in several intraoral devices [8]-[12], researchers focused on the system operation without the in-depth study of the wireless communication. There were a few approaches to characterize antennas working inside the mouth [13], [14]. However, they are still limited in analyzing specific antenna operating in a single frequency band and did not investigate overall factors necessary for the wireless communication. In this paper, we explore methods and frequencies to establish a robust and reliable wireless link for the intraoral devices, by using an intraoral tongue drive system (iTDS) as a practical example.

The explosive growth of wearable electronics and recent advancements in implantable medical devices (IMDs) has coincided with considerable research conducted on wireless communication inside and around the human body. Body channel communication (BCC) has been proposed as a power-efficient wireless communication method for devices that are in contact with the human body [15]. Fig. 1(a) and 1(b) shows simplified diagrams of two common BCC scenarios, capacitive and galvanic BCC working in 30–70 MHz and 0.01–1 MHz bands, respectively [15]-[17]. Utilizing the conductivity of the human body, capacitive BCC works according to the principle of quasistatic near-field coupling. That is, it must have a return path for the current, which is usually formed through capacitive coupling with common ground, e.g., the earth [15]. Galvanic BCC uses an alternating current initiated from a pair of differential transmitter (Tx) electrodes, distributed in the surrounding tissue, and picked up by the receiver (Rx) electrodes [17], [18]. Because this method does not need ground coupling, it can be used for communication inside the body. However, galvanic BCC is suitable for the implant-to-implant communication, in which both Tx and Rx electrodes are inside the body. It shows high attenuation when the signal meets skin because skin is highly resistive to alternating current [17], [18] and is completely ineffective when the Rx is detached from the body.

Fig. 1.

Simplified representation of the Tx and Rx in two types of body channel communication links: (a) capacitive BCC [15] and (b) galvanic BCC [17].

As intraoral devices are in contact with the gums, the tongue, or the palate, and their associated receivers may be in contact with the human body, BCC can be considered to be used for these devices. However, when applied to intraoral devices, neither of the BCC methods maintains their desired operating conditions. For capacitive BCC, coupling between the Tx and the common (earth) ground is severely attenuated because the Tx has to be located inside the mouth, as in the case of IMDs [18]. When galvanic BCC is applied to the intraoral device, high attenuation at a skin is expected because the Tx and the Rx is located on or outside the body. Moreover, both types of BCC require special electrodes to minimize attenuation at the boundary between the device and the body.

Unlike BCC, radio frequency (RF), which is widely used in wireless communication, does not require special operating conditions. It has been successfully adopted in IMDs such as the pacemaker and neurostimulator, and considerable ongoing research is modeling and characterizing the electromagnetic (EM) aspects of IMDs [19]-[28]. Because intraoral devices are located inside the body, they have EM characteristics similar to those of IMDs. However, designing a wireless link for an intraoral device could be more challenging than it is for the IMD link because, depending on its anatomical location, the latter is often surrounded by a distinct and fairly stable tissue environment. By contrast, the intraoral device is located in a constantly changing environment depending on the relative positions of the jaws and movements of the tongue, which continuously changes shape when one swallows, breathes, or speaks. The lower jaw position also changes during speech and ingestion.

To employ RF communication for intraoral devices, we must choose a carrier frequency because of performance degradation from three main sources that are highly frequency dependent. One source is the size of an antenna, which is limited so that it comfortably fits inside the mouth and does not interfere with natural functions such as breathing, ingestion, and speech. The size reduction of the antenna lowers its radiation or coupling efficiency, particularly at lower frequencies with longer wavelengths. Another source is the human body, including the gums, the teeth, and the lips, which surrounds the electronics and attenuates the RF signal, more at higher frequencies. The third source is the impedance mismatch between the power amplifier (PA) and the antenna. A change of geometry, a thickness, and proximity of the tissue around the electronics can distort the impedance matching, particularly at higher frequencies [29].

To determine the optimal carrier frequency for the intraoral device, we selected three candidates: 433.9 MHz, 2.48 GHz, and 27 MHz. All are in the ISM-band, which are available for unrestricted usage. Among the ISM frequencies that behave as propagation waves at distances below 1 m, 433.9 MHz is the lowest with relatively small attenuation in the human body. This frequency is also close to the medical implant communication service (MICS) 402–405 MHz band; therefore, it is expected to show similar characteristics. We chose 2.48 GHz because of its good antenna radiation efficiency in small sized antennas, considering that its quarter wavelength is ~3 cm. We also chose 27 MHz to employ a near-field communication by inductive coupling, which can be more efficient than EM wave propagation at short distances. In addition, the human body is almost transparent to the near-field communication at 27 MHz because its wavelength is >10 m and it mainly depends on the magnetic field [30].

To evaluate the wireless communication of intraoral devices at three selected carrier frequencies, we employ the iTDS as an exemplar verification tool. The iTDS is a new assistive technology (AT) that utilizes user-defined voluntary tongue gestures as commands [11], [12]. Since the iTDS operates inside the mouth and transmits magnetic sensor data to the outside, it uses wireless communication. Using commercially available off-the-shelf (COTS) components, we have developed three versions of the iTDS, each of which operates at one of the three selected frequencies. We compare the performance of the three versions to determine the best choice for the wireless communication of an intraoral device and to identify their advantages and disadvantages. Section II presents system block diagrams and actual implementations of the iTDS for the Tx and the Rx. Section III depicts adaptive impedance matching, which minimizes the power loss between the PA output and the antenna. Section IV shows the test setup and measurement results of the path loss experiment involving human subjects. Section V describes the link budget analysis that identifies the communication distance for each frequency, and concluding remarks follow.

II. ITDS Implementation at Designated Bands

Fig. 2(a) and (b) shows the functional block diagram of the iTDS. All three versions of the iTDS have the same sensing circuitry. Four 3-axial digital magnetic sensors (HMC5983, Honeywell, Morristown, NJ), located at four corners of the iTDS, measure the field strength generated by a magnetic tracer attached to the user’s tongue. This information is delivered to the microcontroller (MCU) through a serial peripheral interface (SPI). Power management is commonly employed in all three versions of the iTDS. A 60 mAh Li-ion battery and a regulator provide 3 V to all electronics, and secure an 8-hour lifetime with 7.5 mA average current consumption [12]. A charging coil connected to a bridge rectifier delivers inductively coupled power from a 13.56 MHz carrier to a battery charging circuitry.

Fig. 2.

Functional block diagram of (a) iTDS operating at 27 MHz, (b) iTDS operating at 433.9 MHz/2.48 GHz, and (c) triple-band iTDS-Rx operating at 27 MHz/433.9 MHz/2.48 GHz.

The Tx architecture of the iTDS operating at 27 MHz differs from that operating at 433.9 MHz or 2.48 GHz. As shown in Fig. 2(a), the Tx for 27 MHz includes an oscillator, a PA, a balun, and a coil. The coil is used not only for charging circuitry but also for 27 MHz communication. A ceramic capacitor is connected in parallel with the coil and tuned to maximize the output signal at 27 MHz while maintaining moderate efficiency for receiving the 13.56 MHz power carrier. The charging circuitry is engaged only when the 13.56 MHz carrier is detected by the built-in analog-to-digital converter (ADC) of the MCU (MSP430) in order not to leak the 27-MHz output signal through the 13.56 MHz charging path. The transmitters for 433.9 MHz and 2.48 GHz utilize the CC1110 and CC2510 RF transceiver system-on-chips (SOCs), respectively. They also include baluns and chip antennas, as described in Fig. 2(b). To minimize the impedance variations of the chip antennas in the dynamic intraoral environment, the iTDS also employs an adaptive matching network.

Fig. 3 shows actual implementations of the iTDS for three operating frequencies. We select the smallest packages for every electronics with 10–12 mm chip antennas to integrate them into a small trapezoidal PCB, and four small sensor boards are connected via a 6-wire flat-cable. A 12 × 23 × 4 mm³ Li-ion battery is located on the PCB, satisfying the clearance requirements for chip antennas. The charging coil is implemented as a 3-turn enamel-coated copper with a maximum circumference for strong inductive coupling without violating clearance requirements for chip antennas. To minimize the parasitic effect caused by the copper line, the electronics for the adaptive matching network are located close to the antenna feed point.

Fig. 3.

Implementation of the iTDS in three ISM bands using COTS components: (a) 27 MHz iTDS, (b) 433.9 MHz iTDS, and (c) 2.48 GHz iTDS.

Fig. 4 shows a transparent Essix-type dental retainer that positions the iTDS electronics and the Li-ion battery below the palate while firmly clipping onto the upper teeth [31]. The Li-ion battery is carefully placed to avoid the required clearance around the antenna. The Essix-type dental retainer is made up of two 0.5-mm-thick transparent polypropylene sheets that have minimal effect on the EM performance of the iTDS, because of their thickness, nonmagnetic properties, a loss tangent of <0.001, and a dielectric constant of 2.2 [32], [33]. A vacuum heat press machine shapes the iTDS retainer with two sheets of polypropylene and the iTDS in between, which creates a hermetically sealed environment that protects the iTDS electronics from saliva.

Fig. 4.

Photos of the iTDS embedded in the Essix-type transparent dental retainer: (a) top side of 433.9 MHz iTDS and (b) bottom side of 2.48 GHz iTDS.

The iTDS-Rx not only receives data from the iTDS but also interfaces with external devices such as a PC, a smartphone, or a power wheelchair [12]. Fig. 2(c) shows the functional block diagram of the iTDS-Rx. The 27 MHz data are received from the same coil used for charging, and a mechanical switch controls the operating mode of the coil for charging and reception. A ceramic capacitor is connected in parallel with the coil and tuned to maximize the sensitivity at 27 MHz while maintaining moderate efficiency for transmitting the 13.56 MHz power carrier. Connected to the coil, a super heterodyne architecture recovers data bits through a low-noise amplifier, a 27-MHz passive bandpass filter, an active mixer, a 10.7-MHz high-Q ceramic bandpass filter, a power detector, and a comparator. The receivers designed for 433.9 MHz and 2.48 GHz employ CC1110 and CC2510 RF transceiver SOCs, respectively, along with the baluns and monopole antennas. For all three frequencies, the UART data recovered by the transceivers are fed to the external USB interface through a UART-to-USB converter chip.

The iTDS-Rx is implemented inside a 3-D printed enclosure, shown in Fig. 5(a), which can be mounted on a power wheel-chair or placed on a flat surface [12]. Because of size constraints, either the 433.9 MHz or 2.48 GHz Rx is housed in the enclosure along with the 27 MHz Rx. A 13.56-MHz charging circuit is also embedded in the box using a TRF7960 (Texas instrument, Dallas, TX, USA) RFID reader, a PA, and a planar spiral coil designed on an FR4 62-mil thick PCB. The PCB coil is geometrically optimized for 13.56 MHz in terms of the line thickness, the line spacing, the inner/outer radius, and the number of turns [34]. A charging cup located in the upper side of the box, a hollow structure, holds the iTDS next to the PCB coil in good alignment. Because of size limitations and the geometry of the enclosure, we select an ANT-433-HE (Linx technology, Merlin, OR, USA) helical monopole wire antenna for the 433.9-MHz frequency and an 88 mm straight monopole copper wire antenna for the 2.48-GHz frequency, as shown in Fig. 5(b) and (c), respectively. The USB interface is implemented with a USB mini-A connector on the side of the box, an iPhone interface with an Apple 30-pin dock connector, and a power wheelchair interface through a standard DB-9 connector from the bottom of the box, as described in [12].

Fig. 5.

Implementation of the iTDS-Rx [12]: (a) Exterior design of the iTDS-Rx allowing room for a smartphone and charging cup for the iTDS; (b) 27 MHz + 433.9 MHz setup; (c) 27 MHz + 2.48 GHz setup.

III. Adaptive Impedance Matching

To minimize the signal reflection between the balun and the antenna in the time-varying environment such as an intraoral space, we employ an adaptive impedance matching technique for the iTDS operating at 433.9 MHz and 2.48 GHz. We do not include 27 MHz here because the human body, including varying intraoral environments, has a minimal effect on near-field inductive coupling at 27 MHz [35]. To minimize reflection at the interconnect, impedances toward the balun (Z₁) and the antenna (Z₂), as shown in Fig. 2(b), need to be conjugate matched with each other [29]. We choose 0433BM15A0001 (Johanson, Camarillo, CA, USA) and BD2425N50ATI (Anaren, East Syracuse, NY) as the baluns and ANT1204F002R0433 (Yageo, San Jose, CA, USA) and 2450AT45A100 (Johanson Technology, Camarillo, CA, USA) as the antennas for 433.9 MHz and 2.48 GHz, respectively. In the air, Z₁ and Z₂ are well matched because the selected baluns have a 50Ω unbalanced port impedance and the selected antennas are also matched to 50Ω at the target frequency. Fig. 6 shows the measured reflection coefficient toward both the balun and the antenna in the air, which indicates negligible reflection between the balun and the antenna. The bandwidths of 433.9 MHz and 2.48 GHz antennas are found as 28 MHz and 100 MHz, respectively, in their datasheets and closely match with the measurement results.

Fig. 6.

Measurement results of the S₁₁ in the air: (a) toward 433.9 MHz Tx balun side; (b) toward 2.48 GHz Tx balun side;(c) toward 433.9 MHz Tx antenna side; (d) toward 2.48 GHz Tx antenna side.

However, in an intraoral space, Z₁ and Z₂ become mismatched because Z₂ changes considerably in dynamic intraoral environments in case of small-sized chip antennas while Z₁ remains almost the same. To address this problem, we employed an adaptive matching network that dynamically adjusts the Z₂ in response to changes of the intraoral environment [36], [37]. The adaptive matching network consists of a switched CLC pi matching network and a power detector. A power detector generates voltage according to the power at the antenna port and the MCU finds the optimum setting of the CLC pi matching network by sweeping the switch settings and monitoring the power detector output. To deliver a small portion of the power at the antenna port to the power detector with minimal loading effect, a directional coupler is connected between the matching network and the antenna, as in Fig. 2(b). We chose MDC201 (MACOM, Lowell, MA) and DC25-73 (Skyworks, Woburn, MA, USA) as directional couplers for 433.9 MHz and 2.48 GHz, respectively.

The CLC pi-network components, shown in Fig. 7(a), are carefully selected to cover the matching range for the varying environments inside the mouth. Though we cannot cover every variation of the intraoral environment because of the limited resolution of the CLC pi-network, we try to cover two extreme cases of intraoral environments in terms of the dielectric loading to the antenna, that is, when the tongue rests on the bottom of the mouth and when the tongue approaches the antenna, as shown in Fig. 8(a) and (b), respectively. For each case, we find the optimum CLC combination by using the Smith chart with some iterations. To cover both combinations, both sides of the CLC network consist of two high-Q ceramic capacitors with two FET switches, as shown in Fig. 7(a). Note that we used a fixed inductor value by compromising each CLC combination because the switch at the inductor could generate an undesirable parasitic effect and totally change the CLC network characteristics.

Fig. 7.

Adaptive impedance matching inside the mouth: (a) CLC pi-network; (b) output voltage of the power detector during impedance matching at 433.9 MHz; (c) output voltage during matching at 2.48 GHz.

Fig. 8.

Experimental setup to measure S₁₁ inside the mouth, with two different tongue positions: (a) resting; (b) folded backward to approach the chip antenna.

While the MCU controls four FET switches in all 16 possible combinations, a power detector feeds the voltage output into the built-in ADC of the MCU. The power detector connected to a coupling port of the directional coupler monitors output power with minimal effect onto the original output. The MCU digitizes the power detector output for each of the 16 switch settings, finds out the one that generates the highest value, and fixes four switches at that configuration. Fig. 7(b) and (c) shows the output voltage of the power detector during the adaptive matching. On the left side of the waveforms, the Tx sends data with unmatched setting when the matching condition is distorted by the change of intraoral environment. During the optimization period shown in the middle of the waveforms, the adaptive matching monitors the power detector output with 16 switch settings and updates an optimal switch setting, which maximizes the output power at the antenna port, as shown on the right side of the waveform. The optimization period takes ~30 ms and is repeated in every 1 s to update the optimum configuration according to changes in the intraoral environment. Because the iTDS sends data in every 20 ms and cannot send data normally during the optimization period, the adaptive matching may degrade the real-time operation of the iTDS [38].

Fig. 8 shows the S₁₁ measurement setup toward the balun and the antenna with one of the subjects wearing the iTDS. A coaxial cable was soldered to the PCB on one side and exposed on the other side, which is connected to a network analyzer via an SMA cable. S₁₁ was measured with and without the adaptive impedance matching while the tongue position changed from its resting position in Fig. 8(a) to being folded backwards so that it approaches the antenna, shown in Fig. 8(b), to view the effect of environmental variations inside the mouth.

Fig. 9 shows S₁₁ measurements toward the antenna side. Fig. 9(a) shows at 433.9 MHz, shifted by ~20 MHz toward lower frequencies inside the mouth as shown by the dotted line. However, when the adaptive impedance matching is applied, it readjusted the S₁₁ to the original frequency as shown by the solid line in Fig. 9(a). The dashed line shows that S₁₁ still maintains a good range when the tongue tip moves toward the antenna, which suggests that the adaptive matching works well. At 2.48 GHz, the S₁₁ antenna loses its Q-factor and, as expected, the shifted toward lower frequencies, shown by the dotted line in Fig. 9(b). The solid line and dashed line in Fig. 9(b) and (c) shows that S₁₁ shifted towards 2.48 GHz after the adaptive impedance matching was applied. However, the selectivity was not recovered and the matching also deviated from 2.48 GHz. We can conclude that unlike at 433.9 MHz, the changed characteristics of the COTS antenna could not be fully compensated by the adaptive impedance matching at 2.48 GHz.

Fig. 9.

Measurement results of the S₁₁ inside the mouth: (a) toward 433.9 MHz Tx antenna side and (b) toward 2.48 GHz Tx antenna side.

IV. Path Loss Experiments

To compare the performance at each frequency band, we measured the path loss from the Tx output to the Rx input for nine distances from 4 cm to 388 cm equally-spaced in a log scale, covering the operating range of interest in the application of the iTDS. We conducted the path loss measurement not only in the actual operating condition of the iTDS but also in two more experimental setups in the air to show the tendency of the path loss at each frequency according to the change of operating conditions (e.g., size of the antenna and operating environment). Also, these additional path loss measurements help us to break down the sources of power loss in detail, for the following link budget analysis. The Tx output power was set to 0 dBm for all frequencies and the Rx input was measured as long as the received power was larger than the input noise floor of the spectrum analyzer. To avoid multipath fading from reflections, we performed all experiment in an open space. To evaluate the path loss when the Tx was placed inside the mouth, two healthy human subjects (a 33-year-old male and a 30-year-old female) participated in this experiment, after receiving approval from the institutional review board (IRB) at the Georgia Institute of Technology.

The objective of the first experiment is to compare the performance of each frequency when both the Tx and the Rx are located in the air and antennas/coils are used without size constraint. The path loss between two Rx antennas/coils (Fig. 5) was measured in the air at three selected bands, as shown in Fig. 10(a). The transmitting Rx coil was connected to the signal generator output through a 27-MHz balun, while 433.9-MHz and 2.48-GHz Rx antennas were directly connected to a signal generator. Similarly, the 433.9-MHz and 2.48-GHz Rx antennas on the receiver side were directly connected to a spectrum analyzer, while the 27-MHz Rx coil was connected to an oscilloscope through a power detector. Fig. 11(a) shows the average path loss in ten measurements with error bars indicating the standard deviation. These results show that, when antennas are used without any size constraint, 433.9 MHz has the smallest path loss for any distance and is the most efficient frequency for the wireless communication in the air.

Fig. 10.

Experimental setup to measure the path loss: (a) Between a pair of Rx antennas/coils in the air and (b) Between a Tx antenna/coil transmitting from inside the mouth and an Rx antenna/coil.

Fig. 11.

Measured path loss between: (a) two Rx antennas/coils (Fig. 5) in the air; (b) Tx antenna/coil (Fig. 3) and Rx antenna/coil both in the air; (c) Tx antenna/coil transmitting from inside the mouth (closed) and Rx antenna/coil in the air.

We conducted the second experiment to compare the performance of each frequency when both the Tx and the Rx are located in the air and Tx antenna/coil is used under strict size constraint of the iTDS. The path loss between the Tx antenna/coil (Fig. 3) and the Rx antenna/coil (Fig. 5) was measured in the air at three selected bands. The measurement results in Fig. 11(b) show that the path loss at 433.9 MHz is much increased and exceeds that at 2.48 GHz, mainly by changing the monopole antenna to the chip antenna. This is an expected result because the radiation efficiency decreases as the electrical length of the antenna decreases, especially in small chip antennas [39], [40]. Although the detailed structure and the radiation efficiency of the chip antennas are not included in their datasheets, we expect the electrical length of the 433.9-MHz chip antenna to be much smaller than that of the 2.48-GHz chip antenna because their sizes are similar while the wavelength of 433.9 MHz is ~5 times longer than that of 2.48 GHz in the intraoral environment. In case of the 27-MHz band, the embedded Tx coil in the iTDS maintains good coupling with the Rx coil and shows very modest ~5 dB degradation compared to the coupling between Rx coils. Thus, 27 MHz is the most efficient band for distances <10 cm, and 2.48 GHz is the most efficient band for distances >10 cm.

The goal of the third experiment is to compare the performance of each frequency in the actual operating condition of the iTDS. The path loss between the Tx antenna/coil (Fig. 3) and the Rx antenna/coil (Fig. 5) was measured when the Tx was located inside the mouth and the Rx was located in the air, at three selected bands, as shown in Fig. 10(b). Considering that the open/closed status of the mouth changes the quality of the wireless communication [14], we instructed subjects to close their mouths when the Tx (iTDS) was inside their mouth, to create the worst case path loss. Then, the adaptive impedance matching was activated, as described in Section III, to minimize the power loss between the balun and the antenna. The path loss measurement results and the experimental setup are shown in Figs. 11(c) and 12, respectively. Because of the high attenuation coefficient of a high-frequency electromagnetic field in the tissue environment, the configuration at 2.48-GHz band degrades more than the other two bands, as summarized in Table I [41]. It should be also noted that impedance matching at 2.48 GHz is suboptimal inside the mouth even with the adaptive matching, as shown in S₁₁ graph in Fig. 9(b). As a result, the 27-MHz band shows the smallest attenuation for distances <39 cm, and the 433.9-MHz band shows the smallest attenuation at distances >39 cm.

Fig. 12.

Experimental setup for measuring the path loss and radiation/coupling patterns at: (a) 27 MHz with subject #1 and (b) 433.9 MHz/2.48 GHz with subject #2.

Table I. Frequency-Dependent Attenuation Characteristics of Tissue [41].

Carrier Frequency (MHz)	Conductivity (σ, S/m)	Relative permittivity (εr)	Attenuation coefficient (α, Np/m)	Attenuation per cm (dB)
27	0.70*/0.03**	120/10	7.60/1.07	0.33/0.06
433.9	1.00/0.05	60/5	23.1/3.73	1.00/0.16
2480	2.50/0.09	50/4.2	65.5/8.07	2.85/0.35

^*Muscle

^**Fat

We also characterized the Tx antenna radiation pattern and the coil coupling pattern for three selected bands in both horizontal (Φ) and vertical (θ) planes, using the path loss measurement setup. The horizontal plane pattern was measured by rotating the subject on a rotating disk for every 15°. The vertical plane pattern was measured by rotating the Rx antenna/coil 15° at a time. Both horizontal and vertical patterns were measured at a Tx-Rx distance of 22 cm, the maximum separation at which sufficient power larger than the noise level can be received at all different angles in all three bands.

Measurement results in Fig. 13 indicate that 2.48 GHz shows the largest variations in the horizontal plane. This finding corresponds to higher attenuation characteristics of this band in the human body as in Table I. In the vertical plane, 27 MHz shows the largest variation of ~20 dB while both 433.9 MHz and 2.48 GHz show variations of <10 dB. This is because the coupling coefficient between the Tx and Rx coils is highly directional, and roughly proportional to cos(θ) [42]. The 27-MHz band also shows a symmetrical pattern with the smallest attenuations at 0° or 180° when the coils are perfectly aligned, which indicates that attenuation at 27 MHz is not dependent on the anatomy of the human body, but on the coil alignment. In the 433.9-MHz and 2.48-GHz bands, the largest attenuation is observed within 90°–180° from above to the back of the head, where the electromagnetic field faces considerable attenuation as it penetrates the head from inside the mouth to reach the Rx antenna.

Fig. 13.

Measured antenna radiation/coil coupling patterns in horizontal (Φ) and vertical (θ) planes at 22-cm Tx-Rx separation with two human subjects (solid line for subject #1 and dashed line for subject #2) at (a) 27 MHz, (b) 433.9 MHz, and (c) 2.48 GHz.

Measurement results from path loss experiments, antenna radiation patterns, and coil coupling patterns show that the 433.9-MHz band is the best choice among the three selected frequencies for the iTDS communication from inside the mouth. Even though the 27-MHz band shows the best performance for distances <39 cm because of its negligible loss in the tissue and strong inductive coupling, it is highly sensitive to distance (slope of the curve in Fig. 11) and angular alignment between the Tx and the Rx coils due to their near-field interactions. The 2.48-GHz band, showing the best performance when the iTDS is operating in the air, is severely degraded when the iTDS is worn inside the mouth and results in inferior performance compared to the 433.9-MHz band.

V. Link Budget Analysis

At marginal operational distances obtained from a bit error rate (BER) test with 0-dBm output power, we conducted a link budget analysis to check if the result of the BER test matches with the result of path loss experiments and to investigate the composition of power loss at each band. For each band, we measured BER at nine logarithmic distances used in the path loss experiments, and determined a marginal operational distance as the maximum distance satisfying BER < 10⁻⁷. With 0-dBm output power at each band, the iTDS continued sending a random packet that exceeds 10⁷ bits (8 bits/byte × 50 bytes/packet × packets/s × 600 s = 1.2 × 10⁷ bits). Then we compared the Tx and Rx data in Excel. We repeated the test for 10 times to gain a reasonable level of confidence of the bit error rate. The Tx and the Rx antennas/coils were positioned close to the worst case angle, according to Fig. 13, and their distances increased until the BER was less than 10⁻⁷ in the lab environment. Marginal operational distances were found as 22 cm, 123 cm, and 39 cm for the 27 MHz, 433.9 MHz, and 2.48 GHz bands, respectively. At the marginal operational distance for each band, we divide the path loss, obtained in the experiments described in Section IV, into each source of power loss for the link budget analysis. We first identified sources of power loss in the wireless communication of intraoral devices, as described in Fig. 14. We included details before the Tx and after the Rx antennas/coils as well as those described in [43] and [44]. We then obtained values of each source of power loss from the experimental results in Sections III and IV, theoretical calculations, and information in datasheets, as in Table II.

Fig. 14.

Sources of power loss in wireless communication from an intraoral device, used for the link budget analysis.

Table II. Link Budget Analysis for the ITDS.

^*Calculated from the Friis’ transmission formula [29]

^**Calculated/derived from datasheets and [45]–[47]

Attenuation in a feeding line and a balun (L_Feed/Balun) was calculated from the output power measured at the unbalanced port of the balun by subtracting it from the Tx output power. We obtained the sum of L_Feed/Balun and insertion loss of the Tx input matching network (L_Tx.IMN) from the difference between the power detector output and the output power at the coupling port of the directional coupler, and got values of L_Tx.IMN by subtracting L_Feed/Balun. The Tx antenna/coil loss (L_Tx.ANT(Coil)) was calculated by subtracting the sum of L_Feed/Balun, L_Tx.IMN, L_AIR, and L_Rx.ANT(Coil) from the path loss between Tx and Rx antennas when both are located in the air. Increases in both L_Tx.ANT(Coil) and L_Tx.IMN (ΔL_Tx.ANT(Coil) + ΔL_Tx.IMN) along with L_HB, which represents the additional loss inside the mouth, was obtained by calculating the degradation in the path loss when the Tx was placed inside the mouth compared to the path loss when the Tx was in the air. Attenuation in the air (L_AIR) was calculated using the Friis’ transmission formula for 433.9 MHz and 2.48 GHz at a distance longer than 0.1 λ, where a far-field assumption is applicable [29], [45]. In case of the 27-MHz band with a wavelength of 11 m, L_AIR was calculated from measurement by using the difference between the maximum received power and the received power at a specified distance. The Rx antenna/coil loss (L_Rx.ANT(Coil)) for each band was calculated by subtracting L_AIR from the path loss between the two Rx antennas and dividing it by two. To ensure robust communication, we also included the potential loss caused by an angular misalignment (L_Angle) based on its worst-case value. Fig. 15 depicts the composition of the loss sources for each band at 22 cm Tx-Rx separation. This figure shows that the major source of power loss vary in each band, and 433.9 MHz shows the smallest sum of all losses at 22 cm, in the worst-case condition considering angular misalignment of coils or attenuation through the head.

Fig. 15.

Composition of power loss for 27-MHz, 433.9-MHz, and 2.48-GHz bands, measured at 22 cm from the iTDS inside the mouth.

We used CC1110 and CC2510 wireless MCUs in the Rx, for 433.9-MHz and 2.48-GHz bands, respectively, and designed a customized super heterodyne Rx for the 27-MHz band (see Section II) [45]. We set the data rate as 250 kbps for both 433.9-MHz and 2.48-GHz bands and adopted the recommended modulation schemes in each MCU datasheet: Gaussian frequency shift keying (GFSK) and minimum shift keying (MSK) for 433.9 MHz and 2.48 GHz, respectively. For the 27-MHz band, P_Rx was calculated from

$$P_{\mathit{Rx}}=N_0+10\log \left({\mathit{BW}}_{\mathit{Ch}}\right)+{\mathit{NF}}_{\mathit{Rx}}+{\mathit{SNR}}_{\mathit{Rx}}$$ (1)

where thermal noise floor N₀ = −174 dBm, BW_Ch is the channel filter bandwidth, NF_Rx is the cascaded noise figure for Rx, and SNR_Rx is the required signal-to-noise-ratio [45], [46]. The SNR_Rx at 27 MHz was derived from

$${\mathit{SNR}}_{\mathit{Rx}}=\left(\frac{E_b}{N_0}\right)\cdot \left(\frac{R}{B_T}\right)$$ (2)

where E_b is the required energy per bit, R is the data rate, and B_T is the system bandwidth, which is 2R for the selected non-coherent on–off keying (OOK) modulation [45], [47]. The required E_b/N_o for the 27-MHz band that satisfies BER < 10⁻⁷ was obtained from the BER waterfall graph in [45]. The received power at the Rx antenna/coil was measured at each band, after adjusting the orientation of the Tx antenna/coil to be close to the best angle at the marginal operating distance. We then calculated the received power at the Rx transceiver in the worst case condition by subtracting the received power at the Rx antenna/coil by L_Feed/Balun and worst case L_Angle. Accordingly, we obtained the fading margin which indicates the headroom between the received power in the worst case condition and the Rx sensitivity that satisfies the BER < 10⁻⁷. As shown in Table II, the fading margin is 7 ~ 13 dB at marginal distances for all three frequencies, which indicates that the link budget analysis based on path loss measurement results and radiation/coupling patterns corresponds to BER test results.

To verify the BER test results and the link budget analysis in actual operating condition, we tested the functionality of the iTDS with computer access at nine logarithmic distances used in the previous experiments. Fig. 16 shows the test setup, in which subjects controlled the mouse cursor on a 21-inch monitor using their tongue motion. Details and outcomes of a similar experiment using an earlier version of the iTDS were presented in [11]. Three iTDS dental retainers shown in Figs. 3 and 4, each tuned at one of the designated bands, transmitted the raw magnetic sensor data from inside the mouth to the iTDS Rx on a desk in front of the subject. An iPod embedded in the iTDS Rx processed the data and delivered the resulting seven tongue commands (left, right, up, down, single-click, double-click, and neutral) to the computer via USB to substitute the computer mouse input [12]. The iTDS operated properly in each band up to its worst case marginal distance, shown in Table II.

Fig. 16.

Test setup for assessing the iTDS operation by accessing a computer and controlling the mouse cursor on the monitor screen [11].

VI. Conclusion

We presented a comparative study of the RF wireless communication for the intraoral devices, using the iTDS operating at three suitable ISM-band frequencies (27 MHz, 433.9 MHz, and 2.48 GHz), to find the best operating frequency among them and to understand the advantages and disadvantages of each frequency. Using three prototypes of the iTDS made of COTS components, we measured the path loss at nine distances between 4 and 388 cm and the radiation/coupling patterns on the horizontal and vertical planes. The measurement results showed that the 2.48-GHz band is the best choice when the iTDS is operating in the air because of its decent chip antenna performance as good as a λ/4 monopole antenna. However, despite the adaptive impedance matching of this band, its performance significantly declined when the iTDS was located inside the mouth as its actual operating condition. The 27-MHz band showed the best performance at short distances of <39 cm with negligible interference from the dynamic intraoral environment. However, it was highly sensitive to distance and alignment between the Tx and Rx coils resulting from their near-field interaction. The 433.9-MHz band, assisted by the adaptive impedance matching that dynamically minimizes the effect of oral kinematics, showed the best overall performance by demonstrating the furthest operating range of 123 cm with 0-dBm Tx output power. It also shows sufficient robustness against head orientation and relative angle between the Tx and Rx antennas.

In future work on the iTDS operating at 433.9 MHz and 2.48 GHz bands, a custom-designed antenna will replace the chip antenna to increase the radiation efficiency and decrease the dielectric loading effect of the antenna caused by intraoral environments such as the tongue. For the adaptive impedance matching network to find the optimum setting for various intraoral environments, we need to increase the resolution of the capacitor bank in the CLC pi-matching network. In addition, to successfully apply adaptive impedance matching without any adverse effect to the real-time operation of the iTDS, we need to reduce both the duration and the period of the optimization for the adaptive impedance matching.

Biographies

Hangue Park (S’11) was born in 1980. He received the B.S. and M.S. degrees from Seoul National University, Seoul, Korea, in 2006 and 2008, respectively. He is currently pursuing the Ph.D. degree at the Georgia Institute of Technology, Atlanta, GA, USA.

From 2001 to 2004, he was with Bluebird-soft, where he designed circuits and systems for industrial personal digital assistants (PDAs). From 2008 to 2010, he was with Samsung-Electronics and designed SAW-less transceivers and PLLs for cell-phone applications. His current research interests lies in closed-loop biomedical system design and fundamentals of neuroscience.

Mr. Park is a recipient of the Best Demonstration Award of the 2012 IEEE Biomedical Circuits and Systems Conference.

Maysam Ghovanloo (S’00–M’04–SM’10) received the B.S. degree in electrical engineering from the University of Tehran, Tehran, Iran, in 1994, the M.S. degree in biomedical engineering from the Amirkabir University of Technology, Tehran, Iran, in 1997, and the M.S. and Ph.D. degrees in electrical engineering from the University of Michigan, Ann Arbor, MI, USA, in 2003 and 2004, respectively.

From 2004 to 2007, he was an Assistant Professor at the Department of Electrical and Computer Engineering, NC-State University, Raleigh, NC, USA. He joined the faculty of Georgia Institute of Technology, Atlanta, GA, in 2007 where he is currently an Associate Professor and the Founding Director of the GT-Bionics Lab in the School of Electrical and Computer Engineering. He has authored or coauthored more than 150 peer-reviewed publications.

Dr. Ghovanloo is an Associate Editor of the IEEE Transactions on Biomedical Engineering and IEEE Transactions on Biomedical Circuits and Systems. He served on the Imagers, MEMS, Medical, and Displays (IMMD) subcommittee of the International Solid-State Circuits Conference (ISSCC) from 2010 to 2014. He is the 2010 recipient of a CAREER award from the National Science Foundation. He has organized several special sessions and was a member of the Technical Program Committees for major conferences in the areas of circuits, systems, sensors, and biomedical engineering. He is a member of the Tau Beta Pi, AAAS, Sigma Xi, and the IEEE Solid-State Circuits Society, IEEE Circuits and Systems Society, and IEEE Engineering in Medicine and Biology Society.