1. Introduction

Ultra wideband (UWB) technology is generally defined as any transmission scheme featuring a 10-dB spectral bandwidth greater than 500 MHz or a fractional bandwidth greater than 20% of its central frequency [1]. UWB transmission is attracting a greater interest as compared to traditional narrowband systems since it brings different advantages, such as: lower power consumption, immunity to multipath fading, the possibility of interference mitigation by means of exploiting spread spectrum techniques, carrier free, high data bit rate and capability to penetrate through obstacles. Furthermore, UWB systems can be displayed in coexistence with other conventional radio systems [1–3]. These advantages have fuelled the interest in UWB systems for wireless communications, networking, location systems and radar imaging [4]. For wireless communications, the US Federal Communications Commission (FCC) approved, in 2002, the use without license of the UWB spectrum from 3.1 to 10.6 GHz with a restriction in the power spectral density of -41.3dBm/MHz [1]. In practical terms this means that the spectra of UWB pulsed systems must fit in a mask to agree with spectral requirements.

During the last three years, there has been a considerable interest in the development of diverse photonic domain techniques to optically generate UWB pulses [6–21] because these approaches can benefit from the well known advantages brought by microwave photonics devices and subsystems: light weight, small size, large tunability, reconfigurability and the immunity to electromagnetic interference [5]. The possibility of integrating the UWB generation directly in the optical domain is very relevant for broadband indoor wireless access since UWB comunication systems can only operate within a limited distance of meters [6]. In principle, UWB over fiber can be implemented with electrical passband filters available commercially to generate UWB pulses by means of a pure electrooptical conversion [7]. However, these electrical filters are designed to satisfy the specifications and requirements in the electrical domain. Therefore, the nonlinear processes of the electro/optical (E/O) conversion have to be considered additionally when E/O and O/E converters are employed through external or internal optical modulation and photodetection, respectively. In this context, a specific design would be needed for transmission link with a given modulator and detector and other additional components such as amplifier. Therefore, it would be always necessary a filtering adaptation to satisfy the FCC mask requirements. Moreover, electrical filters do not permit reconfigurability. In this context, direct UWB pulse generation in the optical domain can become a promising solution, and, furthermore, frequency dependence of all optical components can be taken into account together with electrical UWB pulses.

Apart from satisfying the FCC-specific spectral mask, UWB optical pulse generators have other recent key challenge which is related to the possibility of pulse codification using different modulation techniques such as Pulse Position Modulation (PPM), Pulse Polarity Modulation or Bi-Phase Modulation (BPM), Pulse Amplitude Modulation (PAM), On-Off Keying modulation (OOK) and Orthogonal Pulse Modulation (OPM) [3] but using photonic procedures [8–13].

In this context, several techniques have been already reported for classical UWB pulses, monocycle and doublet generation [14–21]. For example, we can find schemes based on optical spectral shaping [14] and dispersion-induced frequency-to-time mapping such as [15, 16]. Other set of approaches are based on phase-modulation-to-intensity-modulation conversion (PM-IM) by using a dispersive device [17] and an optical frequency discriminator [18]. Both methods for UWB pulse generation offer a low capacity to arbitrary reconfigure the pulse shape. The experimental conditions to obtain a given pulse are fixed and reconfigurability is not easy to achieve. Therefore, UWB pulse generation using microwave photonic filtering is a promising technique from a practical point of view [19–21]. Indeed, most of switchable UWB approaches are based, more directly or indirectly on photonic microwave filtering.

filter [22] where one is interested in the possibility of synthesizing a given impulse response. Therefore recent advances in the field of microwave photonic filters [23, 24] including the possibility of implementing negative coefficients through different techniques, the possibility of tap windowing and transfer function reconfiguration can be exploited in connection with suitable time domain filter synthesis methods previously developed [25]. For instance, several of the above, including cross-gain modulation for negative coefficients [19], cross-polarization modulation [21] and dependence of the half-wave voltage of a Mach-Zehnder modulator [20] have been already proposed as auxiliary techniques in the photonic generation of monocycle and doublet UWB signals with promising results which however do not yet totally satisfy with the above referenced FCC mask.

In this paper we propose an optically switchable architecture for UWB pulse generation which has the potential to overcome the above limitation generating UWB signals that fulfill the FCC mask spectral requirements. It is based on a fully reconfigurable and tunable N-tap photonic microwave filter that features the possibility of both positive and negative coefficients through a proper biasing of an electrooptic modulator as we reported elsewhere [26]. Although this microwave photonic filter approach was proposed as an UWB pulse generation [27], this manuscript provides a detailed theoretical analysis including an extension to generate higher-order UWB pulses. In addition, we show how the proposed UWB generator can be adapted to different modulation formats with a reconfiguration time up to tens of nanoseconds by using optical devices available commercially. The paper is organized as follows: In section 2 we provide the theoretical background necessary to understand the operation principles of the filter and its basic design rules to achieve the desired operation in the time domain for the generation of different UWB pulse shapes. These results are then applied in section 3 to design and experimentally demonstrate the operation of a four-tap microwave photonic filter optimized to comply with the FCC spectral requirements. In addition, we describe how to achieve different pulse modulation techiques by using our approach in a flexible procedure. Finally in section 4 we summarize some conclusions and future directions of our work.

2. Theoretical fundamentals

Figure 1 shows the experimental scheme to generate high-order UWB pulses. It corresponds to an N-Tap tunable and reconfigurable microwave photonic filter whose spectral properties have been considered in [26]. The principle of the proposed system is based on the use of two electro-optic modulators (EOM1 and EOM2) which are biased in regions with opposite slopes through the applied bias voltage V ^DC ₁ and V ^DC ₂, respectively. Both, EOMs are modulated by the same electrical pulse ϕ_RF(t) by using an RF splitter. Therefore, the EOM response m_k(t)of each modulator (k=1,2) in the small signal approximation is given by the following expression:

where ϕ_RF(t) is a gaussian pulse. For simplicity, identical EOMs properties are considered.

 

Fig. 1. Experimental layout of photonic filter. Inset: electrical spectrum of the input Gaussian pulse train.

Download Full Size | PDF

An array of N optical lasers are launched into one of the two EOMs through the cross or bar state of an optical switch corresponding to k = 1 or 2 respectively by using an Nx1 optical coupler. The spectral density of each laser is given by s(ω):

where P_n is the total optical power and δω is the optical linewidth which is assumed identical for each one, and ω_n,k is the central optical emission frequency (n=1,2..N) where the index k just indicates the input EOM for each laser source. The modulated optical signals coming from each EOM are combined and launched into a dispersive element that introduces different time delays for each optical laser. The propagation constant β(ω) for the dispersive element is approximated by a second-order expansion around the angular frequency ω_o:

In order to theoretically analyze the proposed system, we follow the formulation developed by Marcuse for pulse propagation in dispersive elements [28]. Accordingly, the ensemble average of the optical power travelling a distance z in the fiber with a propagation constant β is:

In our case, the term R_k (ω′,t)depends on the index k since each laser can be modulated by either of the two EOMs and M_k(ω) is the spectrum of the modulated signal m_k(t) as shown in Eq. (1).

Introducing Eqs. (1)–(3) into Eq. (4), we can obtain optical power 〈P(z,t)〉 at the output of the dispersive element which is detected by the photodiode (PD). Furthermore, by using the Wiener-Khintchine theorem, the power spectral density of 〈P(z, t)〉 can be obtained and from it the electrical transfer function H_RF (Ω) of the system which is given by:

As Eq. (5) shows, the representation of 〈P(z,t)〉 in the frequency domain is useful since we can identify several terms corresponding to the different effects that conform the input electrical pulse ϕ_RF(t). Note that the frequency responses of the EOM and PD are considered in Eq. (5) through the terms H_EOM(Ω) and H_PD(Ω), respectively. Therefore, this approach is interesting since the E/O and O/E conversions are taken into account in order to avoid their non desired effects over the generated UWB pulse [7].

For instance, the first term in Eq. (5) corresponds to the input electrical pulse. The next term corresponds to the effect of the N-Tap microwave photonic filter. When it is combined with the former term it gives the spectrum of N optical pulses delayed by the dispersive element a time delay given by τ_n = β ₂ L(ω_n,t - ω_o). In addition, two extra terms appear in Eq. (5). The third term is the conventional carrier suppression effect (CSE) due to the beating between the upper and lower sideband frequency components with the optical carrier at the receiver. Finally, the fourth term takes into account the time spreading due to the finite linewidth of the laser. These two last terms introduce a bandpass effect and since the typical bandwidth of the electrical pulses used to implement UWB signals is around 10 GHz, we need them to be negligible in the UWB frequency region. Both last terms can be neglected provided the following conditions that relate the propagation distance, pulse width and the laser linewidths are satisfied:

In practical terms, the previous conditions are met if the propagation distance z is small enough to neglect the dispersive effects over each optical pulse.

Under these conditions, we can appreciate the flexibility of the proposed pulse generation system since it permits a full reconfigurability of the microwave photonic filter with a potential large number of positive and negative coefficients. N lasers can be used to generate different coefficients of the equivalent electrical filter which weights are controlled by tuning the output power of the lasers. For the implementation of positive and negative coefficients, the state of the optical switch (k=1, 2) determines the sign of the corresponding coefficient.

3. Experimental measurements

In Fig. 1, the experimental layout to generate UWB pulses is shown. It consists of four tunable lasers centered at different optical wavelengths given by λ₁=1548.52, λ₂= 1549.32, λ₃=1550.12 and λ₄= 1550.92 nm respectively. These laser sources have a tuning range of ±1 nm around the central wavelength in 0.01nm steps and a linewidth of 100 MHz. Moreover, their optical output powers can be set and modified independently. Each laser is connected to an optical switch to select either the EOM1 or EOM2 which are biased with different voltages in order to operate with opposite slopes in the linear region. Each EOM is fed with an electrical signal coming from a electrical pulse generator with a fixed pattern of one “1” and sixty-three “0” (total 64 bits) and 12.5 Gb/s bit rate. Inset of Fig. 1 shows the electrical spectrum of the input signal Gaussian pulse train. The modulated signals coming from both EOMs are coupled and launched into a 5.43 km standard SMF-28 fibre link with a total dispersion of 93 psec/nm around 1550 nm. The sign and value of coefficients of the corresponding microwave filter are controlled by selecting the input EOM (k=1,2) and the optical power of each laser.

In order to test the performance of the proposed generator, we first analyze the generated pulse by selecting the EOM1 or EOM2 when a laser centered at 1550.12 nm is employed. Figure 2(a) shows the positive optical pulse and Fig. 2(b) plots the negative optical pulse corresponding to select either EOM1 or EOM2, respectively, before (black lines) and after (red lines) propagating through the SMF link. Apart from the effect of the fibre attenuation, we observe that the electrical pulse width is practically the same before and after transmission with a value of 51.7 psec and 51.1 psec for the positive and the negative pulse, respectively. Therefore, the fiber length is suitable to satisfy the conditions imposed by Eq. (6).

As we have explained in section 2, each generated pulse is delayed by controlling the laser wavelength combined with the fibre dispersion. The insets of Fig. 2(a) and Fig. 2(b) show the relationship between the time delay and the wavelength increment when a positive and negative optical pulse, respectively, is generated by tuning a laser initially centered at 1550.12 nm. A linear dependence is observed with a slope of 93.63 psec/nm which corresponds to the fibre dispersion. Since laser wavelength can be tuned in 0.01 nm steps, a minimum delay time difference around 1 psec can be obtained between consecutive optical pulses.

 

Fig. 2. Optical pulses normalized before (black line) and after (red line) SMF, (a) positive pulse and (b) negative pulse. Inset graph relationship between time-delay and wavelength.

Download Full Size | PDF

In order to show the flexibility of the system, different optical UWB pulses have been implemented. In all the cases, the time delay between optical pulses is set to 68 psec to fit the maximum of the electrical transfer function in the middle of FCC mask. Therefore, the wavelength separation is set to 0.74 nm around 1550.12 nm.

Firstly, monocycle pulse is implemented using two lasers with identical optical output power. One of the optical switches works in bar state while the other is set to the cross state to select the EOM1 and EOM2 respectively for each optical wavelength. Under these conditions, the equivalent coefficients of the microwave filter are [1,-1] as shown in the electrical transfer function plotted in Fig. 3(a), where DC component is deleted successfully. Figures 3(b) and 3(c) plot the temporal monocycle pulse and the corresponding measurement of the electrical power spectral (black line) with the FCC mask (blue line).

 

Fig. 3. Experimental (black line) and theoretical (blue line) (a) electrical transfer function and (b) pulse shape for monocyle pulse. In (c), blue line represents the FCC mask for the corresponding experimental normalized electrical power (black line).

Download Full Size | PDF

Next, a doublet pulse is implemented using three optical lasers with equivalent coefficients of [0.5,-1,0.5]. Figure 4(a) shows the electrical transfer function for doublet pulse. Therefore, the first and third wavelengths are launched into EOM1 and the second wavelength is launched into EOM2. The temporal and frequency responses of the doublet pulse are plotted in Figs. 4(b) and 4(c), respectively.

 

Fig. 4. Experimental (black line) and theoretical (blue line) (a) electrical transfer function and (b) pulse shape for doublet pulse. In (c), blue line represents the FCC mask for the corresponding experimental normalized electrical power (black line).

Download Full Size | PDF

A good agreement is found between the theoretical results and experimental measurements. However, Figs. 3(c) and 4(c) show that monocycle and doublet pulses do not fit to the FCC mask requirements, especially in the 0.96-1.61 GHz band as it is well known. Nevertheless, these UWB pulses are usually implemented by other approaches due to the difficulty to add and control more than two coefficients. Our proposed system allows nevertheless to incorporate a large number of positive and negative coefficients and therefore the former limitation can be overcome.

In order to prove the flexibility of our approach, a third-order UWB pulse was implemented after an optimization of four coefficients. In fact, the full versatility of the proposed system allows to reconfigure the electrical transfer function with a high performance fitting to the FCC mask requirements. Four lasers spectrally separated by 0.74 nm are employed with a optimized coefficient vector given by [-0.35, 1, -1, 0.35]. Therefore, the odd optical pulses have negative polarity while the even ones are positive. As shown in Fig. 5(a), the electrical transfer function has a higher amplitude extinction ratio from DC to 2 GHz. Figure 5(b) plots the corresponding UWB pulse and Fig.5(c) represents the electrical power which complies with the FCC mask requirements.

All waveforms exhibit peaks amplitudes of 500 mV which can be modified by the optical power of each laser source which is set around 5 dBm.

 

Fig. 5. Experimental (black line) and theoretical (blue line) (a) electrical transfer function and (b) pulse shape for 4-coefficients pulse. In (c), blue line represents the FCC mask for the corresponding experimental normalized electrical power (black line).

Download Full Size | PDF

As it has been previously mentioned, the possibility of employing optical signal generators is very relevant since modulation can be achieved in optical domain. Furthermore a flexible architecture should be capable of producing several modulation formats upon request. In this sense, we want to emphasize the flexibility of our architecture to provide a large number of pulse modulation formats as compared with other previously reported [8–13]. Firstly, PAM and OOK modulation can be implemented easily since the system permits to control the pulse amplitude through the optical power of the laser sources. Moreover, fastly variable optical attenuators can be added to the system in order to achieve higher communication rates. In addition, the control of each laser output power in combination with the states of the optical switches allows the implementation of an OPM where each pulse shape corresponds to one of the modulation states. For example, Fig. 4(b) and Fig. 5(b) represent an OPM between monocycle and doublet pulses.

Since PAM, OOK and OPM are the less common modulation schemes employed in UWB systems [3], our interest is focused on conventional techniques such as BPM and PPM. In the proposed architecture BPM can be easily achieved by two different methods. On the one hand, the modulator bias can be changed to work in an opposite linear slope region. On the other hand, pulse polarity inversion can be obtained by setting the state of the each optical switch to the opposite state for a given UWB pulse implementation. Figure 6 shows the original pulses implemented previously (black line) and the corresponding inverted pulse (red lines) for different pulse shapes.

Pulse position modulation (PPM) is the most common pulse modulation format for UWB communication systems [3] but there are not many reported optical techniques based on PPM as far as we know [13]. However, our architecture allows to be adapted to this modulation format easily. As we represent in Fig. 2, the time delay can be controlled by changing the optical wavelength of each optical laser. Figure 7 plots the monocycle pulse when different optical wavelengths are selected for each optical laser. In the three cases, the wavelength separation is 0.74 nm as previously proposed in the implementation corresponding to Fig. 3. However, both optical wavelengths can be tuned simultaneuosly. Comparing with Fig. 7(a) which optical wavelengths are 1549.75 and 1550.49 nm, Fig. 7(b) shows the corresponding monocycle pulse when both wavelenghts are decreased 1.07 nm and Fig. 7(c) plots the monocyle pulse when a opposite tuning of 1.07 nm is applied. According to the linear dependence between the time delay and wavelength detuning as shown in Fig. 2, the variation of 1.07 nm corresponds to a time delay around 100 ps. Note that our resolution time is around 1 ps as shown in Fig. 2. Comparing with previous configurations [13], we increase by one order of magnitude the resolution time. Indeed, up to two orders of magnitude could be achieved by using commercial tunable lasers with a wavelength selection resolution of 1 pm.

Regarding to the reconfiguration time of UWB pulses that determines the maximum data rate, our approach can achieve a reconfiguration time up to tens of nanoseconds. A large number of optical attenuators, optical switches and tunable lasers can be found in the market with a fast response time from milliseconds to nanoseconds according to the nature of the optical device which is generally based on mechanical or electrooptical effects, respectively.

 

Fig. 6. Original pulses (black line) and inverted pulses (red line) for (a) Monocycle, (b) Doublet and (c) Four coefficients pulse.

Download Full Size | PDF

 

Fig. 7. Reference monocycle pulse (a) and monocycle pulse for a (b) negative and (c) positive wavelength detuning around 1.07 nm.

Download Full Size | PDF

4. Conclusion

In this paper, we have proposed and experimentally demonstrated an UWB pulse generator based on an N tap microwave photonic filter fed by a laser array and featuring positive and negative sample polarity by using phase inversion in a Mach-Zenhder modulator. The system allows the easy implementation of a large number of positive and negative reconfigurable coefficients, a feature which is essential to generate UWB pulses complying with the requirements of the FCC spectral mask. The generation of simple, monocycle, and doublet pulses has been successfully shown obtaining an excellent agreement between theoretical and experimental results. The high capability and flexibility of the architecture has been demonstrated by generating an UWB pulse using a four tap filter configuration that successfully satisfies the FCC mask requirements. Finally, we have shown that the approach can be easily adapted to produce other modulation formats such as PAM, OOK, OPM and common formats such as PPM and BPM have been demostrated experimentally. As far as we know, this is the first proposal that permits to achieve all these pulse modulation methods simultaneously. Reconfiguration times for all modulation format techniques up to tens of nanoseconds can be achieved by using optical attenuators, switches and tunable lasers which are commercially available.