Combined Pulse Data Transmission and Indoor Localization Using 60-GHz-UWB MMIC Technology

This article presents a pulse radio transmission system operating in the millimeter-wave regime for simultaneous binary information transmission and location in an intrabuilding environment. In particular, the location solution mutualizes the hardware architectures of three sensors of known position to extract the location information of a sensor entering their vicinity and, thus, uses the radio communication signals to extract this location information. Experimental validation considering the 60-GHz transceiver designed and realized on pseudomorphic high electron mobility transistor (pHEMT) technology is exemplary shown with localization error quantified to 16 cm.

information signal is used to power or not the entire transmitter device, including the oscillator. This gives a very high On/Off ratio compared to transmitters using external modulation. In addition, in conventional architectures, all three sensors are connected using external wiring. In the proposed architecture, two of the sensors received the source signal and repeat it through a second emitting antenna: their specific task is to record the input signal and transmit it to the third sensor. This latter receives both source signal and those emitted by each of the other sensor. This versatile architecture allows to consider a wide range of experimental scenarios with different X and Y positions of the sensors. It has to be mentioned that this technique requires the exact spatial knowledge of the three sensors.
The synthesis of the results of this study is divided in three sections. Section II describes the pulse data transmission system that has been realized as well as its performance and characteristics. Section III details how the localization functionality is implemented in the architecture of the radio communication system. Section IV deals with the experimentally established results and makes the assessment on the system constraints presented in the previous sections.

II. MILLIMETRE-WAVE UWB TRANSCEIVER
This millimeter-wave transceiver is designed to transmit ultrawide band (UWB) impulse signals with pulse width in the range of 300-400 ps. The novelty of the proposed approach concerns the energy that is consumed only during the transmission of frequency up-converted pulses at 60 GHz. The 60-GHz frequency band is chosen because of its vast amount of spectral resources which are not regulated to −41 dBm/MHz such as for the conventional UWB frequency band [15]. Another advantage of the 60-GHz frequency band is the size reduction of the antennas and of the overall system [16]. The monolithic microwaveintegrated circuit (MMIC) technology employed is based on pseudomorphic high electron mobility transistors (pHEMTs) with transition frequencies f T up to 120 GHz. Transmission lines and stubs are designed in microstrip technology. The MMIC chip is connected to an external alumina microstrip line ended by a coaxial V-band patch antenna having a 60 • beamwidth. The antenna has not been integrated to limit the MMIC size.

A. MILLIMETRE-WAVE RF TRANSMITTER
The architecture of the Tx transmitter consists essentially of a voltage-controlled oscillator (VCO) associated to a medium power amplifier (MPA) as described in Fig. 1. The oscillator is designed on the basis of a negative differential resistance (NDR). This NDR is achieved by means of a normally-on transistor (labeled N1 on Fig. 1). To make it unstable in the frequency range around 60 GHz, this 4 × 65 μm 2 transistor is fed back through the gate using a 600-μm stub ended by a short circuit. In this configuration, the impedance seen by the source access of N1 is close to an open circuit which fulfils the oscillation conditions when the voltage VS1 equals 0.65 V. The average signal power delivered by this oscillator is slightly above 0 dBm. Following the NDR, a two-stage MPA is cascaded: the first stage has two roles. First, it acts as a buffer that reduces the pulling effect that the NDR could suffer. This buffer consists of the N2 transistor mounted as a common source. Its role is to optimize the power of the signal delivered by the NDR under load conditions close to 50 at the frequency of 60 GHz. The matching network consists of a transmission line backed by a stub ended by an open circuit. The design of these matching networks also aims to boost the power gain of the overall.
The actuation of the RDN oscillator is performed by impulse signals applied to the gate access of the N1 transistor. The impulse signals are provided by an integrated generator: the main principle of this pulse generator is based on a fast XOR gate associated with a capacitive element inducing the expected delay. As depicted in Fig. 1, the clock signal is connected directly to one of the inputs of the XOR gate. The delayed clock signal is applied to the second port: the delay results from the capacitance C0 of a varactor diode whose capacitance value is tunable by a supply voltage. By the way, this topology enables to tune the pulse width. The pulse repetition depends on the clock frequency and can be adapted in the typical range 330 ps-1 μs.
In the use case presented, the pulses perform a direct amplitude modulation of the millimeter waveform by inhibiting or not the NDR oscillator operation depending on whether or not a pulse is generated. Because it is based on an amplitude modulation technique, this transmitter can perfectly well make do with a local oscillator with degraded phase noise properties. Thus, this amplitude modulation causes a frequency upconversion of the pulse signals delivered by the integrated generator. The clock signal feeds the oscillator only when a "1" symbol is transmitted: depending on the clock frequency, a greater or lesser number of pulses will be transmitted, depending on the symbol duration. No pulse is transmitted in the case of a "0" symbol because the clock signal is inhibited in this case. Thus, the clock signal triggers pulses only when a 1 symbol is transmitted. The operation which consists in activating a clock signal only when a symbol 1 contained in a pseudo-random sequence is transmitted is carried out with the help of the data timing generator (Tektronix DTG 5334). The maximum power at the output of the transmitter is equal to 10 dBm.   The measured power added efficiency of the Tx is 3.5% during the continuous emission of a sinusoidal signal with a 60-GHz center frequency. Fig. 2(a) shows a measured waveform of the 60-GHz impulse signals generated by the Tx and Fig. 2(b) shows the related spectrum.
The frequency bandwidth of the transmitter is equal to 3 GHz so that the pulse width is optimal for the value of 330 ps. The test data rate reaches up to 100 Mbits/s. A micrograph of this transmitter is shown in Fig. 3 and the TX specifications are summarized in Tables 1 and 2.

B. 60-GHZ RECEIVER
The 60-GHz reception chain described Fig. 4 is based on the quadratic detection of the propagated signals. Thus, the   phase noise of the local oscillator does not affect the integrity of the communication This also minimizes the number of components involved in the receiver. Indeed, the front-end of the receiver only includes a chain of low noise amplifiers (LNAs) followed by an envelope detector. The architecture of the LNA is based on a cascading of transistors mounted in a common source. The matching networks between the different stages are optimized to extract an optimal compromise between the gain, noise, and stability of the amplifier and its footprint. Indeed, this last point remains crucial if one wishes to cascade several of these amplifiers without causing parasitic oscillations in the reception chain. This is all the more difficult to achieve, as the receiving antenna does not have an impedance of 50 Ohm in the whole frequency band for which the amplifier has a gain greater than 1. In order to be able to cascade these amplifiers during a transfer to a ceramic box, a fine analysis of the stability was carried out on the basis of the normalized determinant function (NDF) method [17].
The five stages of this LNA are composed by a pHEMT mounted as a common source with a 4 × 15 um gate width. It provides a power gain of 27 dB over the 57-60-GHz frequency band. The noise figure of this amplifier was measured at 60 GHz, for which it has a value of 5.2 dB. After this low noise amplification, the envelope of the amplified signal is obtained by detection. This envelope separation operation involves a RF detector [9] with a sensitivity of 1.6 mV per μW at 60 GHz.
The global architecture of the receiver is depicted in Fig. 5.

C. EXPERIMENTAL RESULTS IN COMMUNICATION RELIABILITY
The downside of such a simplified architecture is that it only allows the use of weak energy-efficient modulation VOLUME 2, 2023 8000307 Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. schemes such as OOK. However, it remains compatible with radio transmissions over distances of a few meters, typical of communications in indoor environments. In order to validate the link reliability, a data transmission is achieved using an OOK modulation scheme with a data rate of 100 Mbit/s. The test bench is composed by the transmitter and receiver described in the previous section. For a range of 5 m, the eye diagram is recorded using a Keysight Infinium scope DSO1204A and shown in Fig. 6. Using a logic analysis system (Keysight 16903A), the bit error rate (BER) of the transmitted signals is measured offline as a function of the signal-to-noise ratio (SNR).
By tuning the Tx power, the BER of the transmitted signal is evaluated as a function of SNR as described in Fig. 7. Although the OOK modulation scheme has a lower energy efficiency, the hardware configuration of the developed solution enables to reach BER of the order of 10 −4 for a SNR of 13 dB that relates to a transmission range of approximately 10 m with respect to the system parameters.

III. DISTRIBUTED LOCALIZATION DEVICES A. MEASUREMENT PRINCIPLE
The various communicating objects equipped with this transceiver can thus communicate between them different types of digital data, whether they are dedicated to the payload or to the routing protocols, including information about their location.
The context of our study consists of a mesh network of cooperative wireless sensors. Among them, three sensors called anchors mutualize their hardware architectures in order to estimate the location of the transmitter (Tx) arriving in the localization area. These three anchors, labeled A1, A2, and A3, work in pairs. By organizing the three sensors into pairs, it is possible to extract two time difference-of-arrival (TDOA) values. The main anchor (labeled A1 in Fig. 8) is common to both pairs (A1;A2) and (A1;A3). Each of the pairs is used to determine a (TDOA 1 and TDOA 2 , respectively). The set of possible source locations for each TDOA value corresponds geometrically to a hyperbola. The Tx location is estimated by the intersection of two hyperbolas, each related to one of the measured TDOA values.
One of the anchors (A2 or A3 in turn) acts as a repeater, relaying the pulse signal from the source to be located to the master anchor A1. Then, the master receives both source signal and those emitted by each of the sensor A2 and A3. As the locations of the anchors are known, the travel time between each anchor can be determined and subtracted from the TDOA measured using a digital scope. As the pulse signal retransmitted by a repeater has a power in the order of 10 dBm, it has been found that reflected signals (parasitic signals) remain at much lower amplitudes.
The overall topology is described in Fig. 8: the transmitter, whose location is to be estimated, enters the direct path of the other sensors by emitting pulsed signals at 60 GHz as described above. These signals are then received by A1, A2, and A3. As mentioned previously, only one anchor A2 then A3 repeats the signals to the master anchor A1.
Potentially, the location accuracy depends closely on the pulse width but also on the distance between the two antennas collecting the signals involved in the TDOA measurement. At a minimum, this distance called the baseline must be greater than the product c.T (i.e., 10 cm) where T corresponds to the pulse width (T=330 ps). In order to prevent the phenomenon of geometric dilution of the localization precision (Fig. 9), it is recommended to consider a baseline value as large as possible, usually at least ten times the minimum value (i.e., 1 m). The possibility of relaying the pulse signal via a repeater is an advantage for fulfilling this requirement without using cables.

B. THEORY
The system of equations modeling the possible solutions of the Tx location as a function of the TDOA 1 and TDOA 2 values can be solved using the Chan approach [18]. If we consider R1, R2, and R3 the respective distances between the location A (X, Y) of the Tx to be estimated and the anchors A1 (X 1 , Y 1 ), A2 (X 2 , Y 2 ), and A3 (X 3 , Y 3 ) involved in the location process, the estimated coordinates (X,Y) of Tx are obtained by solving (1) with where and R 2,1 = TDOA 1.c , R 3,1 = TDOA 2.c .

C. EXPERIMENTAL RESULTS
The location process described above is submitted to experimental validation in an indoor environment consisting of a room of dimensions 7 × 5 m 2 , furnished with metal and wood cabinets and furniture. The three anchors forming an orthogonal frame of reference are placed in the room in such a way as to direct their antennas toward the whole room in which the location of the Tx source will evolve. In a first configuration, the three sensors are arranged equidistantly on two perpendicular axes as shown in Fig. 8. Thus, we performed the first measurements considering baselines of values equal to 3 m for the different pairs. As can be seen from the results shown in Fig. 10, the location error tends to deteriorate with the increase of the radial distance.
In such an experimental setup, the location error is mainly driven by the pulse width (330 ps) which is translated by a minimum spatial error of 10 cm. The sampling frequency of the analyzed signal is also a key parameter: at 10 gigasamples per second, the sampling step is 100 ps, which corresponds to an additional step difference of more than 3 cm that weight the localization error. These results remain unchanged whether the OOK modulation is active or not. Indeed, the principle of TDOA is transparent whether the pulses are periodic or not. These results validate the simultaneous multifunctionality of the solution developed in this study since the location process is performed from the impulse signal carrying the binary data.
The measurement environment is a furnished room (5 × 7 × 2.5 m 3 ) comprises mainly five tables (wood) with 12 chairs (metals) with conventional computers (screens, keyboards, . . . ). The walls are in plywood. In order to examine the performance of the developed device in a more realistic way, the locations of both the source and the anchors VOLUME 2, 2023 8000307 are then modified to successively cover the different areas of the room. Fig. 11 shows the configuration of the room used for the measurements with the locations taken into consideration for the different sensors. A photograph of an anchor including the modules Tx/Rx is presented Fig. 12.
For each configuration tested, each anchor occupies one of 26 possible locations. The source to be located then occupies successively the 23 remaining locations. Once the 23 estimates are made, the locations of the three anchors are then modified. This leads to a large number of measurements since 26 locations are listed homogeneously in this room. As mentioned earlier, patch antennas with a 3-dB aperture of the order of +/− 60 • are used. During TDOA measurements, the antennas of the transmitter to be located and those of the sensors are facing each other in order to ensure a radio link under line of sight conditions. More than 400 configurations are thus considered for TDOA measurements: for each of them, the localization error is experimentally determined. To synthesize these results, their distribution function is shown in Fig. 13. This distribution can be approximated by a Gaussian function with a mean value of 16 cm and a standard deviation of 5.5 cm These experimental results demonstrate that the location error is mostly close to the spatial resolution associated with the 330-ps pulse width, i.e., 10 cm. The sampling frequency of the analyzed signals is a parameter that also affects the location performance. Typically, the location error remains within a bound corresponding to twice this minimum observed value. The disadvantage remains the need to sample these pulses after detection, which requires fast samplers.

IV. CONCLUSION
To address the growing need of compact devices for communication and location in an indoor environment, an integrated dedicated pulse radio transmission system has been designed and fabricated in monolithic integrated technology. To achieve a high compact solution, UWB signals are transposed to the millimeter-wave regime at 60 GHz. The chip can address simultaneous binary information transmission and location in intrabuilding environment with measurement error of 16 cm, quantified through more than 400 experimental configurations considered for TDOA measurements. The impact of the Z-position of the Tx on the location error is relatively small if the Z-position varies within one meter. This has been verified qualitatively, but quantitative data has not been extracted as it is beyond the scope of this procedure. In particular, we verified that the error is greatly reduced when the anchors are spaced further apart.
Such a system can work with more fix anchors. Even if the scalability has not been the main objective of this preliminary work, one perspective will be to study the scalability in regard with the additional complexity of the overall solution and related performance (location accuracy, full spatial 2-D or 3-D coverage, . . . ).
Regarding the frequency, around 60 GHz, line of sight is indeed the predominant signal for which no specific problems have been encountered. Indeed, if there is an obstruction, more sensors can be definitively useful. Regarding the algorithm modification, the analytical modeling considered two pairs of TDOA and, consequently, the direct and simplest idea should be directed toward the use of the same algorithm but with different anchors.