Open Access
Add to BibSonomyAbstract
The additive RF phase noise of a microwave photonics link is measured using, as the optical source, a semiconductor laser operating in the class-A regime. The relative intensity noise of this laser being below the shot noise relative level, the phase noise floor of the link is shown to be shot noise limited, −152 dBc/Hz in our experimental conditions. As a result, the phase noise floor evolves as the inverse of the detected photocurrent, pushing the limits of performance to the availability of high power photo-detectors. Below 6 kHz from the carrier frequency at 3GHz, some noise, in excess with respect to the shot noise limit, is observed but remains lower than −110 dBc/Hz at 100 Hz offset frequency. This residual noise originates mainly from environmental noise and can be reduced by isolating the laser from acoustic/electromagnetic perturbations.
©2008 Optical Society of America
1. Introduction
The implementation of optical links in microwave antenna systems offers several advantages, such as: immunity to electro-magnetic radiations, low losses, wideband frequency capability, parallel processing, and possible reduction of weight [1, 2]. Among the considered applications, remote transmission and distribution of microwave local oscillators (LO) has very stringent requirements in terms of phase noise [3, 4]. With this application in view, microwave phase noise of directly (laser modulation) and externally (Mach-Zehnder modulation) modulated optical links has been the subject of numerous studies [5–8]. In particular, it has been shown that the laser relative intensity noise (RIN) is in most cases the main contribution to the noise level of the link and must consequently be as low as possible in order to minimize the additive phase noise brought by the optoelectronic conversion [7]. Recently, we have shown that the RIN of semiconductor (SC) lasers can reach extremely low levels, provided that the laser operates in the so-called class-A regime [9,10]. This regime is achieved when the photon lifetime in the laser cavity becomes much longer than the carrier lifetime in the active medium and presents the strong advantage of a flat spectral noise density, i.e., without any increase of the noise level around the relaxation oscillation frequency. Shot noise limited operation is then obtained over bandwidths as large as 100 MHz–18 GHz, namely in the frequency window of interest for microwave photonics applications. Among the two architectures we have proposed, high-Q cavity semiconductor laser including a half-VCSEL active medium has been shown to be the most attractive one, owing to its compactness and simple design [10]. The intensity noise characteristics of this laser have been studied theoretically and experimentally [11]. But their impact on the microwave phase noise of the optical link remains unexplored. In this letter, a microwave photonics (MP) link optimized for the transmission of RF local oscillators (LO) is implemented including a low noise laser operating in the class-A regime. The additive phase noise of this link is measured and discussed with respect to conventional MP links including class-B lasers.
2. Experimental arrangement
The shot noise limited laser we use in the experiment is a high cavity finesse semiconductor laser operating in class-A regime. It consists of a ½-VCSEL in a 45 mm long external cavity. Single frequency operation is obtained thanks to an intracavity étalon. This laser, described in detail in Ref. 10, is optically pumped at 0.8 µm and provides 70 mW output power at 1.06 µm. This wavelength is not the targeted one to be implemented in a realistic link since 1.5 µm is preferred. However, very good optically pumped lasers at 1.06 µm are available [12] and the choice of this wavelength does not alter the generality of the concept demonstration. The cavity length and the output mirror reflectivity (99%) have been properly chosen to achieve a photon lifetime sufficiently long to exceed the carrier lifetime in the SC active medium. In these conditions, the relaxation oscillations, which are responsible for the excess intensity noise in conventional SC lasers (class-B regime), vanish, leading to a shot noise limited operation over a large frequency bandwidth ranging from 100 MHz up to 18 GHz. At lower frequencies, i.e., below 100 MHz, we have shown in a previous work [11], both theoretically and experimentally, that the relative intensity noise is mainly induced by the pump noise that is not filtered out by the laser dynamics. Nevertheless this noise is still very low and can potentially reach -173 dB/Hz at 10 kHz. Consequently, this laser is a good candidate to improve the phase noise characteristics of MP links, in particular those dedicated to microwave LO transmission and distribution. To test this idea, we have implemented this laser in a generic microwave photonics link devoted to LO distribution. As shown in Fig. 1, the optical link under consideration consists of our laser, followed by a Mach-Zehnder modulator (MZM), a fiber, and a high speed low noise photodiode. The microwave signal of interest feeds the MZM. After the photodiode, a low noise RF amplifier followed by a RF attenuator permit to adjust the level of the microwave signal before it enters the phase noise measurement set-up.
Fig. 1. The experimental set-up used to assess the impact of the laser on the RF phase noise of a microwave photonics link. (1) Reference arm including coarse (Δτ) and fine (φ) time delay adjustment and a RF attenuator (Att.). (2) Arm containing the optical link under test followed by a low noise RF amplifier (G) and a RF attenuator (Att.) The optical link is composed of the low noise class-A laser, a Mach-Zehnder modulator (MZM), an optical fiber and a high frequency low noise photodiode (D).
The key feature of the optical distribution of a RF local oscillator, is the preservation of the spectral purity of the initial microwave signal. This latter must not be degraded along its propagation through the optical link. A common way to assess this degradation is to measure the additive phase noise, namely the microwave phase noise added by the link with respect to a very stable RF reference. The experimental arrangement designed to reach this goal is schematized in Fig. 1. A high spectral purity microwave synthesizer, Agilent E8251A, is used as microwave source. Its output signal is split, with a power divider, into two arms, in order to form a microwave interferometer. The first arm, labeled 1, carries the reference signal whose amplitude and phase can be continuously adjusted, whereas the second arm, labeled 2, contains the MP link under test. The microwave signals at the output of these two arms are mixed and processed in a phase noise test-set (Europtest PN9000). The length of the reference arm (labeled 1) is adjusted in order to get almost equal delays. This allows the absolute phase noise contribution of the synthesizer to be cancelled. Then, for a given modulation frequency, the two arms are put in quadrature, leading to the cancellation of the RF amplitude noise. As a result, the signal measured after an RF mixer and a low pass filter is proportional to the microwave phase noise originating from the MP link only, i.e., to the additive phase noise. It is worthwhile to mention that the additive phase noise measured here is related to the entire MP link. It can originate from the laser, the MZM, the photodiode and/or the RF amplifier. As we expect the contribution of our laser to the phase noise to be very low, much care must be taken in minimizing the contribution of the other devices. The MZM, photodiode, and RF amplifier are battery biased and carefully isolated from ambient electromagnetic radiations. In addition, the MZM is thermally stabilized. Finally, the RF amplifier is operated in saturated regime in order to screen its noise.
Let us remind that in the peculiar case of radar applications which we address here, the spectral range of interest is 100Hz-1MHz. This spectral region covers the Doppler shift of classical moving objects (aircrafts or ground vehicles). For instance, if one considers an aircraft flying at 1000 ms-1 along the radar line-of-sight, then the observed Doppler shift is 20 kHz only, at 3 GHz. Thus, unlike in telecom applications where the spectral range of interest can extend up to several GHz, the spectral region above 1 MHz does not contain any useful information.
3. Results and discussion
After careful calibration of the test bench, the additive phase noise of the MP link is measured for different photocurrents and different modulation depths of the optical beam as illustrated in Fig. 2. The average photocurrent is adjusted thanks to an optical attenuator located in front of the photodiode. Actually, 10 mW of average optical power are available at the output of the MZM. However, a maximum of 4 mW power is sent on the photodiode in order to avoid saturation effects which might lead to underestimate the actual phase noise. The optical modulation index is adjusted by tuning the RF power at the input of the MZM. The curves labeled 1 and 2 in Fig. 2(a) have been obtained at 3 GHz modulation frequency for two optical modulation depths m of respectively 26% and 59%. The associated photocurrents are respectively 1.2 and 1.59 mA. One can notice that, at the frequencies above 6 kHz from the microwave carrier (at 3 GHz), the measured phase noise levels off to reach respectively -144 dBc/Hz (m=26%) and -152 dBc/Hz (m=59%). Our laser being shot noise limited, the phase noise floor level must be also shot noise limited. Here, we call phase noise floor the spectral region labeled (II) in Fig. 2(a) where the noise level becomes almost constant. As shown in Ref. 7, the phase noise power at the floor level for a modulation frequency fc reads:
where R is the photodiode load impedance (50 Ω), kB Boltzman’s constant, e the elementary charge, Iph the average detected photocurrent, and RIN(fc) the relative intensity noise of the laser at the modulation frequency fc. In this expression, the laser relative intensity noise RIN(fc) accounts for the optical excess noise independent from the shot noise. Consequently, if the laser intensity is shot noise limited, i.e., RIN(fc)≪2eIph, its contribution can be neglected. Since, in our case, the thermal noise is also negligible as compared to the shot noise level, Eq. (1) then simplifies to:
Finally, the additive phase noise Lfloor(fc), being given by the ratio between the phase noise and the signal power, we end up with the following simple expression :
where m is the modulation index of the optical beam at fc. Injecting the experimental values of m and Iph into Eq. (3), one ends up with phase noise floors of respectively -144 dBc/Hz and -152.4 dBc/Hz, in very good agreement with the experimental results of Fig. 2. As expected, above 6 kHz [region II in Fig. 2(a)], the phase noise floor is actually limited by the detection shot noise. Furthermore, it is seen from Eq. (3) that m 2 Lfloor evolves as the inverse of the detected photocurrent. We remind that Eq. (3) applies only if the laser is shot noise limited. Otherwise, m 2 Lfloor is independent of Iph as in conventional MP links including a DFB laser where the dominant contribution is the laser RIN. Consequently, increasing the detected photocurrent in conventional MP links does not improve the RF phase noise characteristics of the link. In our case, we show in Fig. 2(b) that m 2 Lfloor evolves as the inverse of the detected photocurrent, as expected from Eq. (3).
This result proves that the laser does not contribute to the phase noise of the MP link under these conditions. The phase noise floor is only restricted by the amount of optical power available on the photodiode. Thus, if we were able to increase the detected optical power without saturating the photodiode, the additive phase noise floor level would be even lower than that obtained. However, it must be noted that this is actually true to some extent. Indeed, at high power levels the pump-to-laser noise conversion as well as the residual amplified spontaneous emission in the vicinity of the laser free spectral range (FSR) cannot be neglected anymore. For instance, in our experimental conditions, we evaluate these effects to lead to a relative intensity noise of -172 dB/Hz at 3 GHz. The RIN limit of the laser should then be reached for a photocurrent level of about 50 mA, namely for at least 60 mW optical power. Obviously, reducing further the RIN limit of the laser for a given RF frequency is still possible by (i) implementing a single mode low noise pump source, (ii) adjusting the cavity length in order to center the RF modulation frequency with respect to two adjacent cavity resonances (iii) increasing the finesse of the intracavity étalon, and (iv) increasing the laser cavity finesse.
Fig. 2. Additive phase noise spectra. (a) Plot 1 : optical modulation index and detected photocurrent are respectively 26% and 1.2 mA. Plot 2 : optical modulation index and detected photocurrent are respectively 59% and 1.59 mA. Plot 3 : Measured noise floor when the optical link is replaced by an RF cable. (b) The Product m 2 Lfloor (m : optical modulation index, Lfloor : additive phase noise floor) evolves as the inverse of the detected photocurrent (Iph), proving that the phase noise is actually shot noise limited.
Let us now consider the low frequency part of phase noise spectra, namely below 6 kHz [region (I) of Fig. 2(b)]. In this spectral region, the RF phase noise is no longer shot noise limited. It behaves like a plateau whose level is independent of the modulation index m (see Fig. 2). This is the signature of multiplicative noise which usually has a technical origin. In order to test this hypothesis we have covered the laser and its pump source with a set of thick metallic panels. This very basic acoustic/electromagnetic isolation scheme leads to a significant reduction of the phase noise close to the carrier (for instance 15 dB at 1 kHz), as shown in Fig. 3. We believe that this residual low frequency phase noise could be further reduced by designing an efficient acoustic and electromagnetic insulation housing, and could even be cancelled with a monolithic cavity design.
Fig. 3. Additive phase noise spectra. Plot 1 : same as plot 2 of Fig.2, The optical modulation index and the detected photocurrent are respectively 59% and 1.59 mA. Plot 2 : the laser is covered with basic acoustic/electromagnetic insulation panels. The optical modulation index and the detected photocurrent are respectively 30% and 1.8 mA. Plot 3 : Measured noise floor when the optical link is replaced by an RF cable.
At this stage, it is important to verify that the MZM is not involved in the observed phase noise degradation even though much care is taken to stabilize it. Indeed, the residual fluctuations of its bias voltage could lead to the observed excess noise below 6 kHz. To this aim, we replace our laser by a Mephisto solid-state Nd:YAG laser from Innolight [13]. Owing to its monolithic design, this solid-state laser, oscillating at 1.06 µm, is one of the most stable and noiseless commercially available lasers. Phase noise measurements are then conducted in experimental conditions close to the previous ones, namely 1.73 mA average photocurrent and 53% modulation index. In these conditions, one obtains the phase noise spectrum reported in Fig. 4 (plot 2). It is seen, in particular, that below 6 kHz the phase noise remains almost constant, proving that the presence of the MZM does not bring additional noise.
Plot 2 also evidences a noise plateau, in excess with respect to the shot noise level, followed by a peak at the laser oscillation relaxation frequency, 430 kHz. Above 430 kHz the noise of the solid-state laser drops rapidly and reaches the shot noise floor. The noise plateau and the resonant noise, are typical of class-B solid-state lasers, which makes them in practice not easy to handle in MP links, unless a noise reduction loop is implemented. Let us recall that conventional SC lasers belong also to class-B. However, their relaxation oscillations are in the GHz range. As a result, the excess noise, induced around the microwave carrier, appears as white noise in the MHz scale. This is no longer valid in SC class-A lasers where the relaxation oscillations are suppressed, thus limiting the RIN to the shot noise level without resort to electronic noise control loop. As evidenced in Fig. 4, the class-A laser leads to an average phase noise reduction of 10 dB from 2 kHz to 200 kHz and more than 40 dB reduction at 430 kHz as compared with the solid-state laser. Above 1 MHz, the RF phase noise of the MP link based on the solid-state laser reaches the shot noise level.
Finally, we have checked that the phase noise spectral shape of the class-A laser MP link remains similar to that presented in Fig. 4 for any choice of the RF modulation frequency, between 1 GHz and 4 GHz, except at 3.4 GHz±5 MHz corresponding to the FSR of the laser cavity. At this specific frequency the noise amplitude of the laser is no more shot noise limited [11] leading to a significant increase of the phase noise. From a practical point of view, if it happens that the RF modulation frequency coincide with the cavity FSR, a slight detuning of the cavity length enables to shift the frequency at which the noise is in excess and finally to recover a situation where the phase noise is shot noise limited.
Fig. 4. Comparison between additive phase noise spectra when the semiconductor class-A laser (plot 1) and the solid-state Mephisto laser (plot 2) are implemented in the microwave photonics link. The optical modulation index and the detected photocurrent are respectively 30% and 1.8 mA for the class-A laser, and respectively 53% and 1.73 mA for the solid-state Mephisto laser. Plot 3 is the measurement noise floor. The dashed line displays the shot noise level.
4. Conclusion
In conclusion, we show in this letter that the peculiar intensity noise properties of class-A semi-conductor lasers make them suitable for low phase noise transmission of RF signals in MP links. The measurements carried out with our class-A VECSEL evidence the fact that the additive phase noise floor of a MP link is shot noise limited. The phase noise spectrum levels off to -152 dBc/Hz for 1.59 mA detected photocurrent and 59% optical modulation index. At frequencies closer to the RF carrier (below 6 kHz in our case) the phase noise grows up to -110 dBc/Hz at 100 Hz frequency offset. A comparison with the phase noise obtained with a solid-state laser confirms that the observed excess noise does not originate from the Mach-Zehnder modulator used in the MP link. It is mainly attributed to environmental noise. Preliminary implementation of a basic acoustic/electromagnetic insulation set-up enabled noise reduction of 15 dB at 1 kHz for instance. Above 6 kHz, the phase noise floor, which is shot noise limited, is shown to evolve as the inverse of the detected photocurrent, contrary to conventional MP links (including class-B lasers) where the phase noise floor remains constant since it is limited by the laser RIN. The lowest level achievable is then ruled, up to a certain extent, by the amount of optical power that can be detected. This remarkable property, related to the use of a class-A laser, opens the way towards the realization of ultra-low phase noise MP links. Possible improvements of this links rely on the availability of high power photodiodes with low noise and high frequency response. Moreover, a similar laser, but operating at 1.5 µm, could be used in order to increase the detection efficiency and consequently decrease the phase noise level. This wavelength is also more adapted for long range LO distribution. Finally, a monolithic design of the laser cavity should lead to a drastic reduction of the phase noise observed below few kHz offset frequency. In its simplest configuration such a monolithic laser could benefit from the electrical pumping of the active medium.
Acknowledgments
Authors would like to acknowledge Emmanuel Poitiers, Christophe Feuillet, and Morgan Quéguiner for their kind help.
References and links
1. S. Blanc, S. Formont, D. Dolfi, S. Tonda-Goldstein, N. Vodjdani, G. Auvray, S. Blanc, C. Fourdin, Y. Canal, and J. Chazelas, “Photonics for RF signal processing in radar systems,” in 2004 International Topical Meeting on Microwave Photonics (IEEE/LEOS, Piscataway, NJ, 2004), 305–308 (2004).
2. C. H. Cox III, E. I. Ackerman, G. E. Betts, and J. L. Prince, “Limits on the performance of RF-over-fiber links and their impact on device design,” IEEE Trans. Microwave Theory Tech. 54, 906–920 (2006). [CrossRef]
3. S. Blanc, M. Alouini, K. Garenaux, M. Queguiner, and T. Merlet, “Optical Multibeamforming Network based on WDM and Dispersion Fiber in Receive Mode,” IEEE. Trans. Microwave Theory Tech. 54, 402–411 (2006). [CrossRef]
4. A. J. Seeds and K. J. Williams, “Microwave photonics,” J. of Lightwave Technol. 24, 4628–4641 (2006). [CrossRef]
5. G. Qi, J. Yao, J. Seregelyi, S. Paquet, C. Bésisle, X. Zhang, K. Wu, and R. Kashyap, “Phase-noise analysis of optically generated millimeter-wave signals with external optical modulation techniques,” J. Lightwave Technol. 24, 4861–4875 (2006). [CrossRef]
6. A. S. Daryoush, “Phase coherency of generated millimeter wave signals using fiber optic distribution of a reference,” in 1996 International Topical Meeting on Microwave Photonics (IEEE/LEOS, Kyoto Japan, 1996), Technical DigestWE4-3, 225–228 (1996).
7. M. Bibey, F. Debrogies, M. Krakowski, and D. Mongardien, “Very low phase-noise optical links-experiments and theory,” IEEE Trans. Microwave Theory Tech 47, 2257–2261 (1999). [CrossRef]
8. P. J. Matthews, P. D. Biernacki, and R. D. Esman, “RF Phase-noise performance of a two-channel optical downconverting link for microwave phase detection,” IEEE Photon. Technol. Lett. 10, 594–596 (1998). [CrossRef]
9. G. Baili, M. Alouini, C. Moronvalle, D. Dolfi, and F. Bretenaker, “Broad-bandwidth shot-noise-limited class-A operation of a monomode semiconductor fiber-based ring laser,” Opt. Lett. 31, 62–64 (2006). [CrossRef] [PubMed]
10. G. Baili, M. Alouini, D. Dolfi, F. Bretenaker, I. Sagnes, and A. Garnache, “Shot-noise-limited operation of a monomode high-cavity-finesse semiconductor laser for microwave photonics applications,” Opt. Lett. 32, 650–652 (2007). [CrossRef] [PubMed]
11. G. Baili, F. Bretenaker, M. Alouini, L. Morvan, D. Dolfi, and I. Sagnes, “Experimental investigation and analytical modeling of excess intensity noise in semiconductor class-A lasers,” J. Lightwave Technol. 26, 952–961 (2008). [CrossRef]
12. M. Jacquemet, M. Domenech, G. Lucas-Leclin, J. Dion, M. Strassner, I. Sagnes, and A. Garnache, “Singlefrequency cw certical external cavity surface emitting semiconductor laser at 1003 nm and 501 nm by intracavity frequency doubling,” Appl. Phys. B 86, 503–510 (2007). [CrossRef]
