Browse

• Abstract

SECTION 1

INTRODUCTION

Ultrawideband (UWB) radio has recently been considered as a promising candidate to meet the ever-growing demand for wide bandwidth and high speed in the future wireless personal-area networks (WPANs). UWB technology is attractive due to its many abilities such as low complexity, low cost, low power consumption, high data capacity, multipath fading immunity and sharing the radio spectrum with the existing wireless systems [1]. In 2002, the Federal Communications Commission (FCC) has allocated 7500 MHz of spectrum for unlicensed use of UWB devices in the 3.1–10.6 GHz band, with a transmitted power spectral density (PSD) less than −41.3 dBm/MHz for indoor wireless communications [2]. Two commonly used UWB transmission approaches are carrier-free impulse radio UWB (IR-UWB) and multiband orthogonal frequency division multiplexing UWB (MB OFDM-UWB) schemes. In IR-UWB systems, data are modulated on series of very short duration and wideband pulses, and due to its carrier-free modulation and avoiding use of frequency mixer [3], IR-UWB technology has become a topic of interest recently.

However, one fundamental limitation of the IR-UWB technology is its extremely low power density, which restricts its coverage distance to only a few or tens of meters. In order to extend the transmission distance and take advantage of optical fiber, such as low loss, large bandwidth and immunity to electromagnetic interference, a technique called UWB over fiber has been proposed as a promising solution [4], with which the generation, modulation and transmission of UWB signals are implemented in the optical domain. In the past few years, a number of photonic schemes for the generation of Gaussian-based monocycle and doublet pulses, which are considered as conventional pulses with mathematical simplicity, lower bit error rate (BER) and robust multipath resilience [5], have been widely reported, such as phase modulation to intensity modulation (PM-IM) conversion by using a frequency discriminator or a chromatic dispersive element after phase-modulated signals [6], optical cross-gain or cross-phase modulation effect in a semiconductor optical amplifier (SOA) [7], optical cross-polarization modulation for monocycle or doublet pulses generation [8], and UWB pulses creation based on photonic microwave delay-line filter [9]. However, the spectra of either monocycle or doublet pulses have strong frequency components in low-frequency band, which will result in interference with other wireless communication systems, especially in the global positioning system (GPS) band (0.96–1.61 GHz). Hence, the transmitted power of these conventional pulses has to be attenuated to respect the FCC spectral mask, leading into reduction of the power efficiency. In order to enhance the power efficiency and reduce the system complexity, many approaches have been proposed and focus on generation of power-efficient UWB waveforms. Optical arbitrary UWB pulse generation based on spectral shaping and frequency-to-time conversion has been reported [10]. Power-efficient FCC compliant UWB waveforms is also realized utilizing apodized fiber Bragg grating (FBG) [11]. Besides, FCC-compliant UWB pulses are also achieved by using incoherent summation of two asymmetric monocycles [12] and linear combination of modified doublet pulses [13].

Additionally, for a practical UWB over fiber system, data should be modulated on stream of IR-UWB pulses using on–off keying (OOK) modulation, binary phase shift keying (BPSK) modulation, pulse position modulation (PPM), pulse amplitude modulation (PAM) and pulse shape modulation (PSM). UWB signals with different modulation formats will have different PSDs, which is significant for the performance of UWB over fiber systems [14]. To implement OOK, BPSK, PPM, PAM, and PSM in the optical domain, the amplitude, polarity, position and shape of the generated UWB pulses should be switchable at a high speed. However, most of the proposed techniques are mainly for the monocycle and doublet pulses applied to two or three modulation formats with the same structure [15], [16], [17], or power-efficient UWB pulses with OOK modulation scheme for the simplicity reason [10], [11], [12], [13], and there appears less experimental demonstration on the power-efficient UWB pulses employing multiple modulation formats, which would be better compliant with the FCC spectral mask and more flexible for future UWB systems.

In this paper, a reconfigurable IR-UWB over fiber system for photonic generation of Gigabit/s polarity-, shape-, and position-switchable UWB pulses, which is based on the incoherent summation of monocycle pulses with inverse polarities, proper time delays and adjustable absolute amplitude utilizing a symmetric PM-IM conversion architecture [17], is proposed and demonstrated for the first time. A good agreement between the simulation and experimental results is observed. Different from those approaches mentioned above, both the conventional UWB monocycle and doublet pulses and the nonconventional power-efficient UWB pulses with OOK, BPSK, PSM and PPM modulation formats can be achieved flexibly and independently in our proposed scheme with good tuning ability. In addition, transmission performance of power-efficient UWB signals employing multiple modulation formats over a combined 20 km single-mode fiber (SMF) and wireless link is evaluated by measuring the electrical spectra, eye diagrams and BERs, which is performed using a digital signal processing (DSP) method. For all the modulation formats a forward-error-correction (FEC) limit error-free operation is achieved and the power penalties of transmission are less than 1 dB. Hence, the proposed approach is potential to meet future requirements for multishape, multimodulation and high-speed transmission of UWB over fiber systems.

SECTION 2

PRINCIPLE OF OPERATION

It is well known that Gaussian pulse is a typical signal which has been widely adopted in IR-UWB radar and communication systems, and it can be expressed as TeX Source $$g(t) = A\exp \left(-{2\ln 2t^{2} \over T_{FWHM}^{2}}\right)\eqno{\hbox{(1)}}$$ where $A$ is a scaling constant and $T_{\rm FWHM}$ is the Gaussian pulsewidth [full-width at half-maximum (FWHM)]. As mentioned earlier, the UWB antenna is not efficient for direct current (DC), and so it is preferable to use derivatives of Gaussian pulse that can reduce the low-frequency components and comply with the FCC mask. The Fourier transforms (FTs) of the $n$th-order derivatives of the Gaussian pulse can be expressed as TeX Source $$G_{n}(f) = A\left({\pi T_{\rm FWHM}^{2} \over 2\ln 2} \right)^{1/2}(j2\pi f)^{n}\exp\left[-{(\pi T_{\rm FWHM}f)^{2} \over 2\ln 2}\right].\eqno{\hbox{(2)}}$$ The power spectra of the Gaussian pulse, monocycle or doublet can be obtained when substituting $n = 0$, 1 or 2 to (2), respectively. However, the power spectra emitted by monocycle and doublet pulses will violate the FCC regulation when trying to maximize transmission power. Although they can fit the mask by simply reducing the transmitted power to a sufficiently low level, the spectrum assigned by FCC will not be exploited efficiently [12].

As mentioned previously, IR-UWB systems are highly power limited, and attenuation of the transmitted power will greatly degrade the signal-to-noise ratio (SNR), which determines the UWB system performance. Hence, UWB pulse shape design becomes a key issue and a pulse waveform that fully complies with FCC mask and maximizes the bandwidth as well should be created. In this sense, UWB waveforms can be optimally designed as described in [11], [18] TeX Source $$P(f) = \sum_{n = 0}^{L - 1}\omega[n]G_{1}(f) \exp(-j2\pi fnT_{0}) \eqno{\hbox{(3)}}$$ where $P(f)$ is the spectra of the newly created UWB pulse, $\omega[n]$ are $L$ weight coefficients, $T_{0}$ is the pulse spacing and $G_{1}(f)$ is the power spectra of monocycle pulse defined by (2). In order to utilize the entire UWB band in 3.1–10.6 GHz, $T_{\rm FWHM} = 54\ \hbox{ps}$ is selected with the peak frequency located at 6.85 GHz which is the center of the allocated UWB bandwidth. Then, weight coefficients can be optimized numerically by respecting the FCC mask requirement and maximizing the power efficiency $\eta$, which is defined as the average power within the useful UWB band normalized by the total admissible power under FCC spectral mask, and expressed by TeX Source $$\eta = \left[\left.\int\limits_{f_{l}}^{f_{h}} \left \vert P(f)\right\vert^{2}df\right/\int\limits_{f_{l}}^{f_{h}}S_{\rm FCC}(f)df\right]\times 100\%\eqno{\hbox{(4)}}$$ where $S_{\rm FCC}(f)$ denotes the FCC spectral mask and $f_{l} = 3.1\ \hbox{GHz}$, $f_{h} = 10.6\ \hbox{GHz}$.

To produce power-efficient UWB pulses in the optical domain, a flexible photonic UWB pulse generator based on symmetric PM-IM conversion architecture is applied and the operation principle of which is described in Fig. 1. Two optical carriers with the same output power are phase modulated by two electrical Gaussian pulse trains, whose pulsewidth and pulse spacing is about 50 ps and 112 ps, respectively. The wavelengths ($\lambda_{1}$ and $\lambda_{2}$) of the optical carriers are located at the left and right linear slope of the frequency discriminator, as shown in Fig. 1(a). It is obvious that both the negative and positive monocycle $(L = 1)$ pulses can be generated at the output port of the frequency discriminator due to the PM-IM conversion [6], as shown in Fig. 1(b). The conventional doublet pulses $(L = 2)$ with inverted polarities are created by incoherent summation of two polarity-inverted monocycle pulses with only one Gaussian bit delay using an optical coupler as illustrated in Fig. 1(c). Moreover, UWB pulses with more complex waveforms can also be produced by combining multiple basic monocycle pulses with proper polarities, time delays and optimized weight coefficients. For the generation of UWB pulses with $L = 3$, as shown in Fig. 1(d), two monocycle pulses with the same polarity are combined with a polarity-inverted monocycle pulse, corresponding to an input Gaussian bit pattern of “1010” and “0100,” respectively. The weight coefficients after an optimization for the two monocycle pulses with the same polarity are both 0.67 and that of the polarity-inverted pulse is 1. Fig. 1(e) shows the generation of UWB pulses for $L = 5$, where three monocycle pulses are combined with two polarity-inverted ones, corresponding to a Gaussian bit pattern of “1010001” and “0100100,” respectively. The three monocycles with identical polarity have optimized coefficients of 0.67, 0.67, and 0.16, while those of the two polarity-inverted ones are 1 and 0.47, respectively.

Fig. 1. (a) Schematic diagram of the proposal for UWB pulse generation. (b) $L = 1$. (c) $L = 2$. (d) $L = 3$. (e) $L = 5$.

As described above, we carry out simulations about generation of different UWB waveforms with both positive and negative polarities. In simulation, two optical carriers, with the same output power, are tuned at 1543.04 nm and 1543.25 nm, respectively. The frequency discriminator is represented by a Gaussian-like optical bandpass filter (OBPF). The simulation results are shown in Fig. 2. As can be seen from Fig. 2(a) and (b), both the positive and negative UWB waveforms including monocycle, doublet and nonconventional pulses are created. Their corresponding spectra compared with FCC spectral mask are also plotted in Fig. 2(c). According to the simulation results, we observe that the newly created UWB pulses for $L = 3\ \hbox{and}\ 5$ fit the FCC mask much better than the monocycle and doublet pulses even in the severely power-restricted GPS band. Furthermore, since the designed UWB pulses are generated by combination of multiple basic monocycle pulses, they have more zero crossings compared with the conventional monocycle and doublet pulses, which moves the energy of the pulses to a higher frequency band but with longer time duration, about 350 ps and 680 ps, respectively. Accordingly, the resulted pulse for $L = 3$ has a central frequency of about 6.6 GHz and a 10 dB bandwidth of 6.8 GHz with fractional bandwidth of about 103%, and the pulse for $L = 5$ has a central frequency of about 6.6 GHz and a 10 dB bandwidth of 7.2 GHz with fractional bandwidth of about 109%. The spectral power efficiency of the newly designed UWB pulses for $L = 3\ \hbox{and}\ 5$ can reach as high as 33.9% and 47.4%, respectively. By attenuating the transmitted power of conventional monocycle and doublet pulses with about 26 dB and 14.8 dB, respectively, their spectra power efficiency can also respect the FCC mask, but the power efficiency decreases to 0.21% and 2.21%. In addition, in this paper, the UWB pulses with higher power efficiency are designed by assuming an antenna with flat gain and a linear phase response in our interesting frequency band. A commercially available antenna with flat response in the UWB band has also been reported [16]. Ones can see that, based on our proposed scheme, the generated UWB pulse shape and polarity can be switched flexibly, which indicates that this UWB pulse generator has the potential to implement OOK, BPSK and PSM modulation formats. Moreover, the PPM can also be realized by changing the position of input Gaussian pulse electrically, although this method only provides discrete pulse positions for PPM. Therefore, as the modulation formats of BPSK, PSM, and PPM can be achieved independently with the same structure, which means three degrees of freedom for modulation, the hybrid modulation schemes can potentially be obtained, such as PPM-PSM, PPM-BPSK, BPSK-PSM or implementation of the three modulation formats together, which may increase the transmission bit rate to two or three times of the original [19].

Fig. 2. (a) Positive-polarity UWB waveforms (b) negative-polarity UWB waveforms for ${\rm L} = 1, 2, 3, \hbox{and}\ 5$. (c) Corresponding spectra.
SECTION 3

EXPERIMENTAL SETUP AND RESULTS

To verify the photonic generation of designed UWB pulses and multiple modulation schemes, an experiment is performed based on the experimental setup, shown in Fig. 3(a). Laser1 and Laser2 with the same output power are tuned at 1543.25 nm and 1543.45 nm, respectively, and a tunable OBPF with a central wavelength of 1543.35 nm and 3 dB bandwidth of 0.25 nm performs as a frequency discriminator. Two streams of predefined input pulse patterns generated by an arbitrary waveform generator (Tektronix AWG 7122B) with a bandwidth of 10 GHz are applied to the $\hbox{LiNbO}_{3}$ phase modulators (PM) to phase modulate the light wave. The input pulse has a shape close to a Gaussian with a full width half maximum of about 70 ps. The two arms of phase modulated optical signals are combined using a polarization beam combiner (PBC) to avoid the polarization-dependent beat-frequency interference and amplified by an erbium-doped fiber amplifier (EDFA). After passing the OBPF, the optical signal is detected by a 45 GHz photodiode (PD) to convert the UWB signals into the electrical domain. The generated UWB temporal waveforms are observed by a digital sampling oscilloscope (Tektronix DSO TDS8200) and their frequency spectra are measured by an electrical spectrum analyzer (ESA) with a resolution bandwidth (RBW) of 1 MHz. It should be noted that, as a proof of concept, the Gaussian pulse patterns are constructed offline by a computer in MATLAB and then sent to the AWG, running at a sampling rate of 12 GSamples/s, which acts as a pulse pattern generator. However, for a practical realization, an electrical pulse generator which is dedicated to create the designed Gaussian pulse patterns can replace the AWG using a pulse synthesizer or digital CMOS circuitry.

Fig. 3. (a) Experimental setup for 1-Gb/s IR-UWB over fiber system reconfigurable for different modulation schemes. (b) The frequency responses of the UWB antennas pair with a distance of 1, 5, 15, and 20 cm.

In order to evaluate the transmission performance of the power-efficient IR-UWB pulses for $L = 3$ and 5, the modulated UWB signals following a data pattern of $2^{11} - 1$ pseudorandom bit sequence (PRBS) are transmitted over combined 20-km SMF and wireless link without any dispersion compensation. The obtained UWB signals are then amplified by a broadband electrical amplifier, emitted to free space and received through an UWB antennas pair. The frequency responses of the antennas pair with a distance of 1, 5, 15, and 20 cm are shown in Fig. 3(b). As can be seen, the antennas pair has a flat gain response and a larger distance leads to a higher insertion loss. In the experimental demonstration, the distance between the antennas pair is set to be about 5 cm, mainly due to the limited gain provided by the electrical amplifier and the 10-dB bandwidth of the antennas pair is about 9 GHz. At the receiver side, a 20 GHz real-time digital phosphor oscilloscope (Tektronix DPO 72004) running at a sampling rate of 50 GSamples/s is used to collect data for demodulation and BER measurements. Additionally, the electrical spectra of the power-efficient UWB signals before and after transmission are also measured and compared to the FCC spectral mask.

At first, two arms of the optical lightwave are phase modulated by two 12 bits-length Gaussian pulse sequences launched from AWG at a bit rate of 12 Gb/s, which is equivalent to a UWB bit rate of 1 Gb/s. The bit sequence is set with a fixed UWB bit pattern “1000 0000 00” (one “1” per 10 UWB bits) indicating that the repetition rate is 100 MHz. The temporal UWB waveforms for $L = 1, 2, 3,\ \hbox{and}\ 5$ with positive and negative polarities and their corresponding electrical spectra without fiber transmission, are shown in Fig. 4. The generated monocycle pulses and their PSDs are shown in Fig. 4(a). The upper and lower FWHMs of the positive monocycle pulse $(L = 1)$ are 60 ps and 63 ps, respectively, and the time duration of the pulse is about 200 ps. The central frequency [CF in Fig. 4(a)] and 10 dB bandwidth [BW in Fig. 4(a)] of the positive monocycle are about 4.9 GHz and 8.7 GHz, respectively, with a fractional bandwidth of 177%. The upper and lower FWHMs of the negative monocycle are 61 ps and 62 ps, respectively, and the time duration is about 205 ps. The central frequency and 10 dB bandwidth of the negative monocycle are about 4.7 GHz and 8.2 GHz, respectively, with a fractional bandwidth of 174%. Notice that the pulse asymmetry of the measured monocycles and deviation from the simulation results is caused by the nonideal input Gaussian pulse shape and response of the OBPF, which also result in slight performance difference between the positive monocycle and negative one. In addition, since the UWB pulses for $L = 2$, 3, and 5 are all generated based on the incoherent summation of multiple positive and negative monocycles, they also endure slight deviations from the simulated high-order UWB pulses due to the asymmetry of the monocycles, shown in Fig. 4(b)(d).

Fig. 4. Generated UWB temporal waveforms with positive and negative polarities and their corresponding electrical spectra. (a) $L = 1$ (b) $L = 2$ (c) $L = 3$ (d) $L = 5$.

As depicted in Fig. 4(b), the pulse width of positive and negative doublet pulses $(L = 2)$ are approximately 59 ps and 58 ps, respectively, and their time duration are about 280 ps and 283 ps, respectively. The corresponding electrical spectrum of the positive doublet has a central frequency of 5.3 GHz, a 10 dB bandwidth of 8.8 GHz and a fractional bandwidth of 166%, while the central frequency and 10 dB bandwidth of the negative doublet is about 5.2 GHz and 8.7 GHz, respectively, with a fractional bandwidth of 167%. The positive UWB pulse for $L = 3$ has the upper and lower FWHMs of about 58 ps and 60 ps, respectively with the time duration of 280 ps, while the negative UWB pulse for $L = 3$ is generated with the upper and lower FWHMs of about 58 ps and 61 ps with the time duration of 283 ps, shown in Fig. 4(c). The central frequency of the positive and negative UWB pulses for $L = 3$ is about 5.4 GHz, and their 10 dB bandwidth is 5.8 GHz and 5.3 GHz, respectively, corresponding to a fractional bandwidth of 107% and 98%. As shown in Fig. 4(d), the upper and lower FWHMs of the positive UWB pulse for $L = 5$ are 58 ps and 56 ps, respectively with the time duration of about 283 ps, meanwhile the negative one has the upper and lower FWHMs of approximately 57 ps and 60 ps with the time duration of 702 ps. The central frequency of the positive and negative UWB pulses for $L = 5$ is about 6.0 GHz and 5.9 GHz, and their 10 dB bandwidth is 6.8 GHz and 6.7 GHz, respectively, with a fractional bandwidth of about 113%.

For all the generated UWB pulses, a good agreement between the measured PSDs and calculated spectra envelopes is obtained despite of slight deviations caused by the same reason described above. Considering that the central frequency of the generated UWB pulses in the experiment increases from 4.7 GHz to 6.8 GHz and their 10 dB bandwidth decreases from 8.7 GHz to 6.7 GHz as shown in Fig. 4, the spectral components in the low-frequency band are significantly suppressed and decreases faster than the higher spectral components as the number of $L$ becomes larger, indicating that the generated UWB pulses with larger $L$ ($L = 3$ and 5) are potential for better fitting the FCC spectral mask. Moreover, electrical spectra of the UWB pulses for $L = 3$ and 5 mainly focus in the 3–9 GHz band, which is beneficial for UWB over fiber system. It is worth noting that neither the central frequency nor the 10 dB bandwidth of the generated UWB pulses for $L = 3$ and 5 is the optimal one due to the limited bandwidth of the Gaussian pulse generator in this experiment, while by narrowing the pulsewidth of the electrical Gaussian pulse applied to the phase modulator, the central frequency and 10 dB bandwidth will increase with the PSDs complying with the FCC mask much better.

According to the analysis of reconfigurable methods in principle and the experimental results demonstrated above, one can see that the proposed scheme is efficient for polarity and shape switchable UWB pulse generation, and the UWB pulse modulation formats of OOK, BPSK, PPM, and PSM can also be implemented with the same structure. In this case, 1 Gbit/s UWB signals following a code pattern of “01101001” with different modulation formats is presented. Fig. 5 shows the 8-bit temporal waveforms of the UWB signals with positive and negative OOK, BPSK, nonorthogonal pulse position modulation (NPPM), Orthogonal pulse position modulation (OPPM) and PSM formats. As a simple and low-cost modulation scheme, externally modulated OOK for all the generated UWB pulses are realized by switching the electrical drive Gaussian pattern applied to the PM on and off. The corresponding modulated UWB signals with both positive and negative polarities are clearly shown in Fig. 5(a) and (b), respectively. To implement BPSK, we benefit from the two-branch structure of our UWB pulse generator, as shown in Fig. 3(a). As mentioned earlier, the designed UWB waveforms are produced by combination of the upper and lower arms of the phase modulated lightwaves, whose center frequencies located at the left and right linear slope of the OBPF. By swapping the input electrical Gaussian pulses to the PM in the upper and lower arms, the polarity of the UWB pulses can be changed. Fig. 5(c) shows that BPSK scheme for all the generated UWB waveforms are achieved very well with waveforms inversed based on the input data, where positive UWB pulses represent data “1” and negative ones represent data “0.” For NPPM, an intuition method is changing the position of “1” in the generated UWB pulses electronically with time delaying of the input Gaussian patterns to the PM. The modulated signal is shown in Fig. 5(d), and the time shift for data “0” and “1” bits is about 170 ps. However, this method can only provide discrete pulse positions for PPM, which may deteriorate the flexibility of a UWB system. Methods for achieving continuous change of the pulse positions are under investigation. OPPM are also accomplished with the same hardware as NPPM, and in the OPPM format, a UWB pulse is transmitted in the first or the second half of the bit time based on the input data, as shown in Fig. 5(d). Due to the longer time duration of UWB pulse for $L = 5$, the repetition rate for OPPM becomes 500 MHz only half of the OOK format. To realize OPPM with the same data transmission rate as OOK for $L = 5$, the hybrid modulation scheme of OPPM-BPSK may be used. An example of PSM is also realized with a positive monocycle pulse representing a “0” bit and a positive doublet pulse standing for a “1” bit, as shown in Fig. 5(e). Obviously, for all the generated UWB pulses, high-speed OOK, BPSK, PPM, and PSM can be independently achieved without influence on the other three modulation formats, and so our proposed UWB over fiber system proves the feasibility and potential for future multicasting applications.

Fig. 5. UWB signals with a pattern of “01101001” for (a) positive OOK (b) negative OOK (c) BPSK (d) NPPM (e) OPPM and (f) PSM.

The electrical spectra and eye diagrams of the UWB signals for $L = 3$ and 5 with different modulation schemes are shown in Figs. 6 and 7, and the spectral spikes of these UWB signals are manually attenuated to comply with the FCC mask between 3.1–10.6 GHz. It is worthy to note that UWB signals with different modulation formats would have different PSDs, which is very important for the design and implementation of a practical UWB system. As can be seen in Figs. 6 and 7, the PSDs of all the UWB signals consist of discrete and continuous components. However, the strong discrete spikes are undesirable because they would present great interferences to other narrowband systems and limit the total transmitted power from the antennas regarding the FCC regulation. For the BPSK modulated signals, displayed in Figs. 6(c) and 7(c), they have the largest continuous part and no strong discrete components are observed, which indicates that much higher power efficiency can be realized as compared with OOK or PPM modulation schemes. For the OOK signals shown in Figs. 6(a) and (b) and 7(a) and (b), the discrete spikes appearing at multiples of 1 GHz which is exactly equal to the repetition rate of the UWB pulses, are about 30 dB larger than the continuous part and so the powers of OOK signals have to be significantly reduced to satisfy the FCC spectral mask. Just as the case in OOK, the NPPM and OPPM modulated signals also have large discrete spikes, and in addition, there would be many notches in the continuous part with spacing of $1/T_{\rm shift}$, which is the time shift for “0” and “1” bits. As shown in Figs. 6(d) and 7(d), the spacing of the notches in the continuous part is about 5.88 GHz, since the time shift of $T_{\rm shift}$ is about 170 ps. For the OPPM UWB signals with $L = 3$, the spacing of the notches is about 2 GHz with $T_{\rm shift}$ of 500 ps, shown in Fig. 6(e). Fig. 7(e) shows the OPPM UWB signals with $L = 5$ and the spacing of the notches becomes about 1 GHz with $T_{\rm shift}$ of 1 ns. The experimental results of the power efficient UWB signals for $L = 3$ and 5 have a good agreement with the theoretical predictions in [14].

Fig. 6. Electrical spectra and eye diagrams of the UWB signals for $L = 3$, with positive OOK (P_OOK), negative OOK (N_OOK), BPSK, NPPM, and OPPM schemes (a)–(e) with wireless transmission; (f)–(j) with combined 20-km SMF and wireless transmission.
Fig. 7. Electrical spectra and eye diagrams of the UWB signals for $L = 5$, with positive OOK (P_OOK), negative OOK (N_OOK), BPSK, NPPM and OPPM schemes (a)–(e) with wireless transmission; (f)–(j) with combined 20-km SMF and wireless transmission.

Since the continuous spectral components contain the data information in the UWB signals, we compare the measured continuous part of the UWB signals for $L = 3$ and 5 after 20-km SMF and wireless transmission shown in Figs. 6(f)(j) and 7(f)(j) with those only after wireless transmission. As can be seen, the total transmitted powers in the 3.1–10.6 GHz band are enhanced, with lower frequency components decreasing faster than the upper ones after fiber transmission. This phenomenon is beneficial for the generated UWB pulses to have better electrical spectra matching the FCC mask and is potential to give positive impact on the UWB signals transmission over fiber, which is mostly due to the slightly frequency chirp of the UWB pulses produced by the PM to IM conversion. In addition, the widely opened eye diagrams after combined SMF and wireless link as shown in Figs. 6 and 7 indicate that no significant distortion is induced for all the modulation formats after propagating over the fiber link, promising a good tolerance to the chromatic dispersion over fiber. Notice that the wireless transmission experiment is carried out under open-air conditions, therefore, the UWB antennas would receive radiated signals lower than 3 GHz from other wireless systems, such as Global System for Mobile communication (GSM 900 and GSM 1800) at 0.9 GHz and 1.8 GHz, respectively and Code Division Multiple Access systems (CDMA 2000) at 2.1 GHz, which would add low frequency components and deteriorate the transmission performance of the UWB signals. However, it is believed that the system performance will be further improved by applying a high-pass filter (HPF) to eliminate noise component in the low-frequency band or using the commercial UWB antennas.

In order to demodulate the received UWB signals and evaluate the BER performance of the proposed IR-UWB over fiber system with different modulation schemes, a correlation technique of UWB receiver is applied, which has been reported for demodulating different modulation formats [20]. Our UWB signal receiver architecture is based on the DSP method and the data recorded by the digital oscilloscope is then demodulated in the digital domain, which is expected to be more robust and convenient for the variety of the wireless link [21]. The BER curves versus received optical power of the UWB signals for $L = 5$ with OOK, BPSK, NPPM, and OPPM modulation formats is shown in Fig. 8(a). Limited by the memory depth of our oscilloscope, for each BER measurement point, 100000 UWB bits corresponding to more than 48 replicas of the UWB PRBS of length $2^{11} - 1$ are transmitted and recorded. The BER is subsequently computed offline using a DSP algorithm in a bit-for-bit comparison between the transmitted and received data. As an example, the DSP processing of 4-bit IR-UWB signals (“0101”) with OOK modulation scheme is shown in Fig. 8(b). First, the received UWB signal is normalized and correlated with a normalized mask bit, and then, the modulated UWB signal can be decoded easily by employing an optimum decision threshold and clock recovery timing which are chosen appropriately to minimize the BER. Finally, the logical value is determined by comparing the sum of the processed signal points for each UWB bit to the determined decision threshold, and then if the sum is larger than the decision threshold, then the bit is assigned a logical “1,” and otherwise for a logical “0.” According to the results of the BER measurement, UWB signals with BPSK scheme have the best BER performance in our proposed system since they have larger continuous part in the PSDs and much higher power efficiency than the other modulation schemes, as described above. The NPPM modulated UWB signals have the poorest BER performance due to its lower modulation efficiency, which could be further improved by employing a technique called time hopping (TH) [14]. Generally, for all the modulation formats, clean and open eye diagrams of the UWB signals with combined 20-km SMF and wireless transmission are obtained and the power penalties are less than 1 dB introduced by the fiber dispersion. Especially, for the BPSK and OPPM modulated UWB signals, no errors are detected when the received optical power levels above −1 dBm. Considering the FEC limit $(2 \times 10^{-3})$, the experimental results demonstrate a successful transmission of IR-UWB over a 20-km SMF and wireless link for access applications.

Fig. 8. (a) BER measurements of the IR-UWB over fiber system with OOK, BPSK, NPPM and OPPM. Insets: Eye diagrams of the UWB signals with combined 20-km SMF and wireless transmission. (b) DSP processing of 4-bit IR-UWB signals.
SECTION 4

CONCLUSION

A reconfigurable IR-UWB over fiber system for photonic generation of Gigabit/s polarity-, shape-, and position-switchable UWB pulses, has been investigated theoretically and demonstrated experimentally based on the incoherent summation of multiple monocycle pulses with inverse polarities and proper time delays. Two kinds of FCC-compliant IR-UWB pulses with high power efficiency for both positive and negative polarities are achieved and compared to the conventional monocycle and doublet pulses. Furthermore, 1 Gbit/s excellent photonic OOK, BPSK, PPM, and PSM modulation schemes are also realized in our proposed system. Transmission performance of the generated IR-UWB signals with different modulation formats over a combined 20-km SMF and wireless link is experimentally evaluated by electrical spectra, eye diagrams and BER measurements. For all the modulation formats, the power penalties of 20-km SMF transmission are less than 1 dB and a FEC-limit error-free transmission can be achieved. We believe that our newly proposed system has a potential application for future high-speed IR-UWB over fiber access networks.

Footnotes

This work was partially supported by National Program on Key Basic Research Project (973) under Contract 2012CB315703, NSFC under Contract 60736002, 60807026, 61111130188 and the Program for New Century Excellent Talents in University (NCET-10-0520). Corresponding author: H. Chen (e-mail: chenhw@tsinghua.edu.cn).

References

No Data Available

Cited By

No Data Available

None

Multimedia

No Data Available
This paper appears in:
No Data Available
Issue Date:
No Data Available
On page(s):
No Data Available
ISSN:
None
INSPEC Accession Number:
None
Digital Object Identifier:
None
Date of Current Version:
No Data Available
Date of Original Publication:
No Data Available