3.1 Effects of the Electromagnetic Environment Outside the Human Body

The human body is composed of many tissues which have a high dielectric constant, in the low-frequency bands [6], especially bands under 100 MHz, which can be easily coupled with the human body due to their short effective wavelengths inside the body [18–20]. Outside the human body, various types of electronic devices emit electromagnetic waves in a low-frequency band [17]. The emitted electromagnetic waves are coupled with the human body due to its antenna function, and the coupled waves then generate an interference signal inside the body, so the receiver receives an interference signal caused by the electromagnetic environment outside the human body along with a data signal transmitted from the transmitter. To illustrate this effect, the interference signal received at the receiver was measured in various electromagnetic environments, as shown in Figure 4. A subject was exposed to a general electromagnetic environment like that shown in Figure 4(b), while the receiver’s signal-electrode was attached to the arm of the subject. The interference signal caused by the human body’s antenna function was then measured at the signal electrode using the measurement setup shown in Figure 4(a).

Figure 5 shows examples of interference signals measured inside buildings and on a subway, in which power of the interference signal is presented in the frequency domain. The interference signal does not have significant power at some measurement sites inside buildings, but otherwise it has high power over a wide-band region, as shown in Figure 5(a). Unlike the building environments, the interference signal has high power at most measurement sites inside the subway, and the high power is distributed under 30 MHz due to the high-voltage pantograph collector used on the subway system. The interference signal in HBC has a dynamic property such that its power and frequency change depending on the location; however, the interference signal is distributed mainly in the low-frequency band used by HBC. Any electronic device emitting an electromagnetic wave in the low-frequency band can be a source of interference. Therefore, the transmitter for HBC also is a significant source of interference; an electrical signal transmitted at one user’s transmitter for HBC is emitted outside the human body in the form of an electromagnetic wave due to the antenna function of the human body, and the emitted electromagnetic wave is then coupled with the human body of another user nearby. In this case, the power of the interference signal is dependent on the number of HBC users and the distances between each user. This is especially similar to that of the receiving signal for HBC when users are in close proximity to each other [18]. Therefore, the interference signal in HBC can have high power in the same frequency band occupied by the electrical signal transmitted through the human body, causing severe degradation of the bit error rate (BER) performance, a metric indicating the number of errors in the received data stream.

3.2 Channel Model for Human Body Communication

A channel model is required to design a transmission scheme and an analog front-end for HBC. The IEEE standard for HBC has been published [5], and its PHY structure is based on the channel model presented in an earlier study [21]. The channel model is composed of a channel filter and an interference signal generator¹, as shown in Figure 6. The channel filter represents the signal loss experienced by the electrical signal as it is transmitted through the human body. An electrical signal to be transmitted through the human body is filtered by the channel filter and the interference signal generated from the interference signal generator is then added to the filtered signal. After the thermal noise is added, the channel’s output signal is obtained.

Numerous researchers have studied signal loss by the human body [22–25]. For an accurate measurement of signal loss, it is important to isolate the grounds of the transmitter and the receiver from the earth ground, as the coupling between the ground planes strongly affects the signal loss. However, in previous research, the measurement equipment was connected to the earth ground [22,23], or abnormally large ground planes were used to measure the signal loss [24,25]. In order to measure the signal loss while maintaining an isolation condition between the ground planes, the measurement setup shown in Figure 7 is used [26]. The transmitter module, powered by a battery, generates the pulse signal, and the generated pulse signal is transmitted through the human body. The transmitted pulse is received at the receiver module and measured using the oscilloscope. The pulse signal has a very short width and hence has a high-frequency component to obtain the signal loss over a wide frequency band. The transmitter module simultaneously generates a signal synchronized with the pulse signal, and this synchronization signal is optically transmitted to the synchronization module using the optical cable. The synchronization module restores the synchronization signal and then triggers the oscilloscope whenever the pulse signal is generated at the transmitter module; hence, the receiving signal at the receiver module can be measured with the transmitter and receiver modules synchronized. Also, the isolation condition between the ground planes is maintained during the measurement, as the two modules are optically, not electrically, connected to each other. The pulse signal and the receiving signals are transformed into the frequency domain using the Fourier transform, and the signal loss is then computed by subtracting the receiving signal from the pulse signal in the frequency domain. The impulse response of the channel filter in Figure 6 is obtained like Eq. (1) after the transformation of the computed signal loss into the time domain [21]:

$h(t)=h_R(t)\cdot C_h$ (1)

(1)

Here, h_R(t) is a reference impulse response, and C_h is a coefficient related to sizes of the ground planes and distances between the transmitter and receiver, as follows:

$C_h=(0.0422G_T-0.184)\cdot (0.0078G_R+0.782)\cdot {\left(\frac{120.49}{d_{body}+d_{body}{\left(\frac{d_{air}}{d_{body}}\right)}^5}\right)}^2$ (2)

(2)

In the above equation, G_T and G_R are the ground plane’s size in cm² at the transmitter and receiver, respectively. Additionally, d_air and d_body are the distances related to the coupling between the transmitter and receiver through two mediums, as shown in Figure 1; d_air is the distance in cm between the transmitter and receiver through air; and d_body is that between the transmitter and receiver through the human body. C_h is valid only when 10 cm²≤G_T, G_R≤270 cm² and 10 cm≤d_air, d_body≤200 cm. The reference impulse response is an impulse response when the coefficient C_h is equal to 1; it is expressed as follows:

$h_R(t)=A_v\cdot A\cdot e^{\left(-\frac{(t-t_r)}{t_0}\right)}\cdot sin\left(\frac{\pi (t-t_r-x_c)}{w}\right)$ (3)

(3)

Each HBC user has different physical parameters related to the channel; these account for various body size and component ratios of body tissues, such as fat and muscle. The signal loss in HBC is affected by these physical parameters; hence, each user has a different amount of signal loss. A_v in Eq. (3) is a random variable used to represent this type of variation in the signal loss; it follows the Gaussian distribution with a mean of 1 and a variance of 0.16². A, t_r, t₀, x_c, and w are the constants related to a shape of the reference impulse response [21].

At the transmitter, the electrical signal before being transmitted through the human body is filtered by mask filtering to remove harmonic components and possible interference in other frequency bands [5], thus occupying a narrow band of about 6 MHz. The interference signal is modeled only over the occupying band and can therefore be approximated with the additive white Gaussian noise, which has a frequency-independent power. The measured interference signal has the Gaussian distribution; this is reasonable because multiple electronic devices, including HBC transmitters, emit electromagnetic waves, and these waves amount to an interference signal at the receiver. After the summation of independent random variables, it follows the Gaussian distribution according to the central limit theorem. The interference signal has a different variance according to a location where HBC is used; hence, the interference signal generator from [21] is modeled with maximum variance of the interference signal, for which the measured maximum variance was found to be 2.55×10⁻⁵.

4.1 Walsh Code

It is necessary to understand the characteristics of the spread code to understand the transmission scheme comprehensively.

The Walsh (–Hadamard) code is unique in that each non-zero code word has a Hamming weight of exactly $2^{n-1}$, which implies that the distance of the code is also $2^{n-1}$. In standard coding theory notation, this means that the Walsh–Hadamard code is a ${\left[2^n,n,2^n/2\right]}_2$ code. The Hadamard code can be seen as a slightly improved version of the Walsh–Hadamard code, as it achieves the same block length and minimum distance with a message length of n+1; i.e., it can transmit one more bit of information per code word, but this improvement comes at the expense of a slightly more complicated construction.

A Walsh (–Hadamard) code is obtained by selecting as code words the rows of a Hadamard matrix. A Hadamard matrix $M_n$ is an $n\times n$ matrix (n is an even integer) of 1s and 0s with the property that all rows differ from other rows by exactly $n/2$ positions. One row of the matrix contains all zeros. The other rows contain $n/2$ zeros and $n/2$ ones.

For n=2, the Hadamard matrix is

$M_2=\left[\begin{array}{cc}0& 0\\ 0& 1\end{array}\right]$ (4)

(4)

Furthermore, from $M_n$, the Haramard matrix $M_{2n}$ is generated according to the relationship

$M_{2n}=\left[\begin{array}{cc}{M}_n& M_n\\ M_n& {\overline{M}}_n\end{array}\right]$ (5)

(5)

where ${\overline{M}}_n$ denotes the complement (0s replated by 1s and vice versa) of $M_n$. For example, the $M_4$ matrix is created as follows:

$M_4=\left[\begin{array}{cc}\begin{array}{cc}0& 0\\ 0& 1\end{array}& \begin{array}{cc}0& 0\\ 0& 1\end{array}\\ \begin{array}{cc}0& 0\\ 0& 1\end{array}& \begin{array}{cc}1& 1\\ 1& 0\end{array}\end{array}\right]$ (6)

(6)

Each code has a fundamental frequency despite the fact that it has many frequency components. In the case of the $M_4$ matrix, the first row has no transitions, and the second row has three transitions and has the highest fundamental frequency component. The third and fourth rows have one and two transitions, respectively. If the matrix is rearranged by the number of transitions, the higher the index, the higher the number of fundamental frequency Walsh codes there will be. This means that the greatest power of each Walsh code is in its fundamental frequency.

4.2 Frequency-Selective Digital Transmission

Data transmitted in a digital waveform can be spread over the selected frequency domain using Walsh codes. A 64-chip Walsh code is used for HBC. The 64-chip Walsh code is rearranged according to the number of transitions; hence, the index of each code corresponds to the number of transitions. For example, a code with index number 48 of the 64-chip Walsh code, W48, has 48 transitions. It is divided into four groups with 16 codes using the index of each code. The fourth subgroup of the 64-chip Walsh code has 16 codes, the highest number of transitions and thus, the highest fundamental frequencies. If it is assumed that the system uses a clock of 32 MHz, the maximum fundamental frequency is 16 MHz. The occupied frequency band of subgroup 4 of the 64-chip Walsh code in such a case is depicted in Figure 9, which shows the simplified spectrum of the fourth subgroup, the received signal that passed through the human body and the normalized noise power from the interference signal caused by an antenna function of the human body. If the codes of this group are used in order to spread a signal, the highest signal power to be transmitted exists in the fourth subgroup mainly located in the 12 to 16 MHz frequency band.

Also, as shown in Figure 9, the normalized noise power with a 5 MHz bandwidth decreases in proportion to the frequency. An earlier study [27] also reported that there is a narrow window between 10 MHz and 20 MHz in which the interference signal’s power is relatively low. In the frequency response of the human body channel under study, the signal loss in the human body channel gradually increases in proportion to the frequency up to 40 MHz. Above 40 MHz, the signal loss increases more rapidly in proportion to the frequency.

Though a conventional spread system has a wider bandwidth that resembles white noise, the transmitted signal generated with the subgroup of the 64-chip Walsh code resembles the signal modulated with carrier frequency. The output spectrum appears as the baseband signal with the 4 MHz bandwidth but modulated with a carrier frequency of 14 MHz. With a group of Walsh codes, the information is spread into the selected frequency without continuous frequency modulation using the carrier frequency. The spread spectrum has the advantage of guaranteeing the quality of communication by increasing resistance to the interference of the human body coming from various appliances. The HBC modem in work [8] makes use of 16 codes out of the 64-chip Walsh codes that have the highest fundamental frequencies, i.e., the fourth subgroup (W48~W63) in Figure 9. Figure 10 shows a block diagram of the HBC modem and the operation of the frequency-selective digital transmission. The serial input data becomes a 4-bit symbol by a serial-to-parallel (S2P) block. This 4-bit symbol becomes the index of the 16 codes out of the 64-chip Walsh codes. If the source data rate is 2 Mbps, the symbol rate is 0.5 Msps (symbols per second) after the serial-to-parallel converter. The final chip rate after the frequency-selective spreader is then 32 Mcps (chips per second). The FSDT scheme has the advantage of a high data rate and a simple architecture. A previous system [28] makes use of Manchester coding, which always has a transition in the middle of each bit period; hence, an error in transition will directly result in a bit error. However, in the FSDT scheme, the minimum distance of the 64-chip Walsh code is 32; therefore, the receiver can still decode even if 15 bits are lost. The spread baseband signal has only two consecutive identical chips, 00 or 11. As a result, its fundamental frequency ranges from 12 to 16 MHz with a clock frequency of 32 MHz. Through the use of the 16 spread codes with the highest fundamental frequencies, the baseband appears to be modulated with a carrier frequency of 14 MHz. The FSDT scheme can be implemented without a digital-to-analog converter (DAC) at the transmitter, an analog-to-digital converter (ADC) and circuit block related to the radio frequency (RF). This ensures extremely low power consumption and implementation in a small size.

As shown in Figure 10, the HBC modem is composed of four main blocks, the interface block (HBC IF), the transmitter block (HBC TX), the receiver block (HBC RX), and the analog front-end block (HBC AFE). The HBC IF block is the interface block between a micro-controller unit (MCU) and the HBC modem. In an HBC IF, there are register files in which to store the control information for all sub-blocks, a buffer to store the traffic data to be transmitted or received, an interrupt controller, and a serial interface block. The HBC TX is the transmitter block, which includes a scrambler, a serial-to-parallel block, and a frequency-selective spreader. The HBC TX output is connected to a signal electrode directly. The HBC RX is the receiver block, and it includes a synchronization block to search the start of a frame, a frequency-selective de-spreader, a parallel-to-serial block, and a descrambler. The HBC AFE is the analog front-end block, which has a noise reduction filter, amplifier, a clock recovery, and data-retiming block.

To prevent a loss of synchronization due to clock drift, an optional “pilot” sequence can be inserted with data in the PSDU in the IEEE standard for HBC PHY [5]. The pilot signal is inserted periodically, interleaved with a block of split data, according to the value of a predetermined insertion period. There are three pilot insertion intervals according to the information data rate. Another approach is to use CDR in order to recover a clock synchronous to the input data stream; hence, the recovered clock retimes the incoming data. A receiver implemented using CDR does not need a pilot signal, which improves the data throughput performance. In addition, it can recover the transmitted data without an ADC requiring a fast sampling frequency, which reduces the power consumption. The IEEE standard for HBC PHY also has no pilot insertion mode.

The frame structure for the FSDT scheme implemented using CDR is composed of a downlink sub-frame (or packet) and an uplink sub-frame (or packet), as shown in Figure 11 [29]. The length of a frame is 10 ms, and three sub-frame ratios are supported. The sub-frame ratios of the downlink versus the uplink are 8:2, 5:5, and 2:8. A downlink sub-frame consists of a lock time of 2 μs, a preamble of 128 bits, and a header of 64 bits. The data is shown in Figure 11. The dummy signal, which has a repeated pattern of 0 and 1 with the highest chip rate of 32 M chip per second during 2 μs, is used for the lock time. The dummy signal is used to give a CDR circuit an approximate reference frequency so that it reduces the time to align the phase to within 2 μs. An uplink sub-frame consists of a lock time of 2 μs, a header of 64 bits, and the data. There is a guard time of 2 μs for every sub-frame, which is the transition time between a receive mode and a transmit mode. From 250 kbps to 2 Mbps, four variable data rates are provided.

5.1 Received Signal

Design of the analog front-end requires analysis of the signal propagation through the human body. The channel model presented in the previous section is usable for performance analysis, but not suitable for design of an analog front-end because performances of an analog front-end, such as a cut-off frequency of a filter or hysteresis of a comparator, cannot be accurately modeled in simulation using the channel model. As a result, designing an analog front-end based on signal measurement is more practical. The size of the ground plane affects the signal loss in the channel. A larger ground plane enhances the electric coupling such that the size of the ground plane for the receiver is determined depending on the application, e.g., a mobile phone. In this experiment, the receiver electrode was designed with an electrode size of 5 mm × 5 mm considering typical button size of a handheld phone, and a ground plane of 50 mm × 100 mm, which is a typical phone size. Adopting a single electrode is more user-friendly than employing two electrodes for the signal and the ground. A transmitter was located on one hand and powered using a USB cable by a battery-operated mobile terminal, i.e., an ultra-mobile PC. A transmitted signal generated by a field-programmable gate array (FPGA) passed through the human body and was transferred to the receiver electrode located on the other hand. The transmitted signal consisted of a dummy pattern of 2 μs and a Manchester-coded preamble of 4 μs. The dummy signal had a repeated pattern of 0 and 1 with a 32 M chip rate during 2 μs and hence provided the clock and data recovery (CDR) circuit an approximate reference frequency to reduce the time to align the phase within 2 μs. The Manchester-coded preamble had a 4 μs length with 128 bits.

The transmission distance was approximately 150 cm. The receiver electrode was connected to the active probe to maximize the isolation between the receiver and the earth ground. The measurement setup and measured signal, which has a strong low frequency mainly around 50 kHz for 110 ${mV}_{pp}$, are shown in Figure 12. Figure 12 also shows a zoomed-in graph of Point A, which is the start point of a packet.

The measured signal was band-pass filtered by combining a high-pass filter (cut-off frequency of 8 MHz) and a low-pass filter (cut-off frequency of 30 MHz) using the mathematical function of an oscilloscope. The band-passed signal had an amplitude of about 10 $m{V}_{pp}$.

5.2 Design of an Analog Front-End

An analog front-end is designed based on the received signal measured on the human body. The signal loss in the channel ($L_P$) is expressed in dB. It can be calculated as follows:

$L_P=20\log \left(\frac{V_{RX}}{V_{TX}}\right)$ (7)

(7)

Here, $V_{RX}$ is the received voltage and $V_{TX}$ is the output voltage of the transmitter. If the output voltage of the transmitter is 3.3 $V_{pp}$ and there is signal loss of approximately 60 dB through the human body [27,30], the receiver must have sensitivity of at least 3.3 $m{V}_{pp}$. As introduced in [28], the receiver separates the desired signal from all other signals induced on the channel, amplifies it to a level suitable for further processing, and determines its binary state using a comparator. A CDR circuit can be used to align the binary data to the clock. Figure 13 shows a block diagram of the analog front-end.

The first step is to define the gain of the amplifier. The gain of the receiver is determined based on the relationship between the minimum drive voltage of the comparator and the lowest amplitude of a signal that can be received. Figure 14 shows a simplified diagram of an HBC system that has a transmitter with a tri-state output buffer along with the channel (i.e., the human body) and a receiver front-end with an amplifier and comparator. High-pass filters implemented by passive and active components are omitted because they have little effect on the signal amplitude in the pass band. Because the human body is exposed to various electromagnetic environments, the received signal not only has strong low-frequency interference, but also high-frequency interference with relatively low amplitude. Positive feedback should be used to provide hysteresis and increase immunity to high-frequency noise. Maximum signal loss ($L_P$, dB), based on various body positions, of –70 dB [27] has been reported.

To define the amplifier gain ($G_A$), the channel and amplifier are regarded as a single block, as depicted in the dotted box in Figure 14. The gain of the block is related to the output voltage of the transmitter ($V_{TX}$) and the input drive voltage of the comparator ($V_{COMP}$) as follows:

$G_A=20\log \left(\frac{V_{COMP}}{V_{TX}}\right)-L_P$ (8)

(8)

If the signal experiences a maximum path loss ($L_P$) of –80 dB and the comparator can be driven by an input signal of 33 $m{V}_{pp}$, the amplifier should have a gain ($G_A$) of 40 dB. The receiver has a sensitivity of 330 $\mu V_{pp}$. The maximum signal loss, taken into account considering various body positions, is –70 dB [15], with loss variation of –5 dB and a receiver margin of –5 dB.

The next step is to define the active filter specifications (Figure 13). A Bessel filter with a maximally flat group delay is used as it preserves the wave shape in the pass-band. Because the spread baseband in the FSDT scheme is directly transmitted through a human body, its spectrum occupies a broad band. In order to find the characteristics of the filter type, a transient analysis was performed based on the measured data using the Agilent Advanced Design System. Based on the simulation, a fourth-order active high-pass filter was designed with an 8 MHz cut-off frequency. The utilization of hysteresis on the comparator, together with the 200 MHz gain bandwidth product of the operational amplifier IC, can remove the high-frequency noise in the received signal, thus allowing use of a high-pass filter instead of a band-pass filter that was used in previously described experiments.

The induced noise of the 50 Hz or 60 Hz alternating current, caused by electric power transmission and fluorescent lighting, could be tens of volts on the human body. This noise is likely to saturate the active circuits with the supply voltage of 3.3 V used in the analog front-end and thus should be reduced to a point as low as the signal as the receiver can accommodate without saturation. A passive second-order high-pass filter realized using a series capacitor and two resistors connected between the supply voltage and ground offers half of the supply voltage, which provides the maximum swing range of the received signal without saturation. Figure 15 shows a simulated signal from the electrode to the comparator based on measured data. The processed signal has 4.6 ns of root-mean-square (RMS) eye jitter. This originates from noise, the induced interference, inter-symbol interference, and a band-limited channel during signal transmission, and from the physical limitations of the slew rate of the circuit and the hysteresis of the comparator. The received signal is strongly amplified, filtered, and then switched rapidly between two levels, at which point the data are finally recovered from the processed signal with jitter using a CDR circuit. A transmitted signal consisting of a spread signal using a group of Walsh codes and a Manchester-coded preamble is represented by a unipolar non-return-to-zero (NRZ) level with 0 and 3.3 V and with only two consecutive identical chips. This enables the proposed scheme to be used in asynchronous communication. The timing information is extracted and the chip stream regenerated using a CDR circuit from the received data while receiving a signal within a 2-chip delay. A voltage-controlled oscillator (VCO) is a type of oscillator whose oscillator frequency is controlled by a voltage input. The applied input voltage, named as a control voltage ($V_C$), determines the instantaneous oscillation frequency. The CDR is used as a local oscillator during transmission to generate a nominal frequency or a center frequency of a VCO by setting the control voltage ($V_C$) to a specific level, generally at the center of the supply voltage ($V_{ext}$ in Figure 13). Within a frame, the master and slave should exchange data once with each other. This technique can periodically initialize a frequency offset between the clocks of the master and slave within a few ppm and prevent clock frequency drift. In HBC, data transfers cannot be deliberately started and stopped and thus a fast lock time is needed. The traditional frequency lock time of the phase lock loop used in a CDR is generally a few hundreds of microseconds. The slave operates as a receiver during a downlink and as a transmitter during an uplink time slot. In order to track the transmitted signal as receiving data at the downlink time slot, the output voltage of the loop filter (V_loop) is used as the control voltage (V_C) of VCXO by RXE in Figure 13. When no data is received at the downlink as well as at the uplink, the external voltage (V_ext) is used as VC to use CDR as a local oscillator. The clocks of the master and slave only exhibit a phase difference within a limited time, i.e., a frame length without buffer memory. In addition, dummy patterns of 2 μs composed of periodic data at 32 Mcps are used for training the CDR. The dummy pattern and oscillation mode during transmission can settle the control voltage to within 2 μs.

5.3 Operation of CDR

The receiver implemented using CDR has the advantage of higher data throughput performance compared to the scheme using pilot signal. The information in the frame can be used to control the CDR in order to synchronize between the HBC devices as well as to reduce the power consumption. Figure 16(a) shows the control signals related to the sub-frame ratio and the length of the received data. A control signal denoted as Rx_valid expresses the sub-frame ratio of the downlink versus the uplink and is set to a “logic high” setting during the downlink. A signal denoted as RXE expresses the length of the received data. When a master transmits data, a slave starts to receive the data, after it has passed through the human body. This point is marked as in Figure 16(a). The CDR of the slave in the receiving mode locks in the phase and frequency to the received signal that is represented as unipolar NRZ with the clock information of the transmitter. The control voltage of VCXO varies according to the frequency of the received signal, which is nearly identical to the frequency of the transmitter. This is the period marked as in Figure 16(a). If the received data is less than the allowable maximum size, the Rx data ends before the end of Rx_valid, as shown in Figure 16(a). This point is marked as in Figure 16(a). In general, at the end of the received data, a lack of level transitions in the received signal for a considerable number of clock cycles should be noted. As long as the received signal is not present, the CDR is kept idle by the loss-of-signal (LOS) control signal, which is generated by the input stage of the CDR. Most CDRs have a function that sets the control voltage of VCXO under the LOS condition. The CDR used for the FSDT scheme has a LOSIN terminal to control the control voltage. The LOS output signal is normally “logic low” and is set to “logic high” after 256 consecutive clock periods with no transition of the received signal. The LOS signal can be used to either flag external alarm circuits and/or drive the CDR’s LOSIN input. When LOSIN is set to “logic high,” the VCXO control voltage is switched to an internal voltage, usually half of the supply voltage, to generate a nominal frequency. The LOS signal is reset to “logic low” as soon as there are received signal transitions.

In HBC, however, random noise received from the human body is processed with the receiver chain, finally becoming binary digital data with an arbitrary pulse width. As a result, between the end of the received data and the end of the sub-frame of the downlink or the start of the sub-frame of the uplink (marked as in Figure 16(a)), the phase of the incoming signal has a random difference from the phase of the local oscillator of the CDR. The processed random noise causes the CDR to lose its lock.

When the slave transmits the data marked as in Figure 16(a), the transmitted signal leaks to the receive path of the slave. The feedback loop is formed: i) the CDR provides the clock frequency of the transmitted signal; ii) the transmitted signal leaks through an electrode to the receive path; iii) the leakage signal of the transmitted signal is processed and a comparator makes it a binary waveform with the frequency of the transmitter; iv) this signal is then connected to the CDR; and v) finally, the feedback loop is formed and the control voltage of VCXO converges to a constant value, generally half of the supply voltage, which generates the nominal frequency of VCXO. The same problem occurs when the transmitted signal is less than the allowable maximum period, marked as in Figure 16(a).

In HBC, the user rapidly establishes connections to networks by touching the devices. A fast lock-in time of CDR is essential to characterize HBC in comparison with general wireless communication. In order to reduce the lock-in time, the frequencies of the master and the slave should be synchronized or maintained at a nominal frequency, after which the acquisition of the phase lock and the tracking of the acquired phase timing are processed. A simple solution to synchronize or maintain the frequencies of the master and the slave is to maximize the size of the data for all sub-frames. This can be accomplished by adding a dummy signal to the transmitted data for both the master and the slave. The dummy signal has a waveform with the repeated pattern of 0 and 1, with the highest chip rate, as used in the lock time. By adding the dummy signal, there is no period without transitions of the received signal even when the size of the data is less than the maximum value. As a result, no random binary signal induced on the human body is injected into the CDR. However, this method consumes additional power due to the extra pattern.

Another approach is to manage the control voltage of VCXO, as shown at the bottom of Figure 16(b). The signal of RXE depicted in Figure 16 (b) is set to “logic high” while in the presence of received data, and the signal of RXE is the complementary signal of RXE. The RXE signal is a control signal that is used to switch the control voltage of VCXO between the output of the loop filter and the external voltage. While in the presence of received data, VCXO is controlled by the output voltage of the loop filter. The control voltage of VCXO is set to the external voltage throughout all periods of the frame except in the presence of the received signal. The external voltage is generally half of the supply to generate the nominal frequency, as shown in the middle of Figure 16(b). In addition, the signal of RXE or RXE can serve as a control signal to enable or disable a comparator. According to the polarity at which the comparator needs to be enabled, one of the two signals is used. If the comparator is shut down, a CDR that is positioned after the comparator receives no transition of the signal, as shown in Figure 15. The operation of the comparator by RXE is shown in Figure 16(b). As a result, the output signal, denoted as the CDR input in Figure 16(b), intuitively describes the condition of a loss-of-signal (LOS) for the CDR. Under the LOS condition, the control voltage of VCXO in the CDR generates the nominal frequency and the CDR then becomes ready to play the role of the local oscillator during the transmitting period.