Open Access
Add to BibSonomyAbstract
A photonic preprocessor for analog to digital conversion is demonstrated and characterized using a cavity-less optical pulse source. The pulse source generates high fidelity pulses at 2 GHz repetition rate with temporal width of 3 ps. Chirped pulses are formed by cascaded amplitude and phase modulators, and subsequently compressed in dispersion compensating fiber. Sampling operation is performed with a dual-output Mach-Zehnder modulator, where the complimentary output enables a reduction of noise by 3 dB. Phase noise characterization shows that the phase noise of the generated pulses is fully dictated by the RF source. The high quality of the pulse source used in a sampling preprocessor experiment was verified by measuring 8 effective number of bits at 10 GHz and 7.0 effective number of bits at 40 GHz.
©2012 Optical Society of America
1. Introduction
A wide range of applications rely on digital signal processing (DSP) which have had a significant speed improvement following development of faster computational power resulting in an increased demand on analog-to-digital-conversion (ADC). The availability of an appropriate sampling pre-processor can reduce the timing and dynamic requirements of the following quantization circuitry in ADCs, resulting in improved ADC performance for high speed operation.
The sampling-preprocessor can also have higher bandwidth than the sampling rate, which allows sub-rate or subsampling, i.e. sampling a high frequency signal with sampling speed lower than the Nyquist condition (i.e. at half the sampling rate). This operation can also be used in order to directly sample a signal on a carrier to avoid frequency down-conversion complexity and impairments, as well as to allow the ADC to perform quantization at a lower rate [1]. For subsampling applications, the dynamic nonlinearity and jitter are handled by the sampling pre-processor, which in electronics is performed by a track-and-hold (TAH) circuit [1,2]. Although, the bandwidth of electronics has constantly been improving, other technologies, such as photonics, have been proposed to overcome electronic sampled bottlenecks [3]. Photonic technology has proven to be able to generate high quality short temporal optical pulses and be able to use them in gating or sampling operation. Therefore photonic technology is a serious candidate for implementing a broadband and high accuracy sampling pre-processor which is followed by an electronically quantizing ADC. Sub-rate sampled ADC demonstrations using photonic sampling have gained considerable attention [3–5]. The predominant technique of generating high quality pulses having temporal width in extent of the picosecond regime has been based on mode-locked lasers (MLL) in these applications, and high frequency narrow signal sampling with notable performance has recently been shown [5].
Although mode-locked pulse sources can be built with excellent amplitude and phase jitter characteristics, the stringent phase locking condition results in non-trivial operation and sensitivity to external perturbation when operated outside of the laboratory environment. As a result, mode-locked sources require complex and rigorous phase locking arrangements to maintain stability. Moreover, the cavity also dictates the operation frequency and hence the ADC sampling frequency has to strictly coincide with that of the pulse source.
In an attractive alternative approach, a cavity-less pulse source can be used to generate high quality optical pulses to circumvent the stringent limitation imposed by a cavity. For instance, chirped optical pulses can be generated by concatenating amplitude and phase modulators (PMs), which subsequently are temporally compressed by propagation in a dispersive medium, providing short optical pulses [6–8]. Very high performance has been demonstrated previously in optical time division multiplexed (OTDM) transmission experiment [9]. A major advantage of a cavity-less pulse generation is that of obviating the need of a feedback control system (in the absence of mode-locking), as well as the timing jitter being solely determined by the underlying RF-source. Furthermore, an important property of the cavity-less pulse source is its frequency tunability (determined by the RF-source) allowing flexible operation at a frequency dictated by the ADC itself, thus allowing a continued development of faster and more accurate electronic ADCs.
In this paper we demonstrate a high fidelity cavity-less pulse source operating at 2 GHz, introduced in [10]. The performance of the pulse source is quantified by using the pulse source in a sampling pre-processor before an ADC for a Nyquist and sub-rate- sampling demonstration. The ADC preprocessor is characterized using the IEEE ADC standard [11]. We quantify the system performance in terms of the system signal-to-noise and distortion ratio (SINAD) and effective number of bits (ENOB) of signals input up to 40 GHz, and the results show a very high performing cavity-less based pulse source architecture.
2. Cavity-less pulse source principle
The principle of generating optical pulses without an optical cavity, i.e. no optical feedback, relies on a single-pass structure which can be implemented by using concatenated optical modulators driven by a common RF source. Chirped optical pulses are formed from a continuous-wave (CW) light source followed by optical modulators, as depicted in Fig. 1 . The chirped pulses are subsequently compressed in a dispersive medium, e.g. standard single mode fiber (SMF) or dispersive compensating fiber (DCF). In the absence of a cavity, the cavity-less pulse source is not constrained in its operation in a manner mode-locked sources are, which enables unique features for the former source. Indeed, the cavity-less pulse source can be operated with seamless wavelength tunability and nearly-arbitrary repetition rate, set solely by the driving RF source. Moreover, the phase noise performance of the generated pulses, and therefore the timing-jitter, is dictated only by the employed seeding RF source.
The temporal duration of the generated pulses is dependent on the amount of chirp that can be induced in the pulse and subsequently temporally compressed in a dispersive medium. The chirp of the pulses is imposed by one or more concatenated PMs, whilst the frequency deviation is proportional to the modulation amplitude as well as to the modulation frequency. Finally, as can be inferred from the above description, the amount of frequency deviation (chirp) that is generated, will determine the overall achievable extent of compression.
3. Cavity-less pulse source design
The sampling rate in this work was dictated by the available ADC which was operating at 2 GS/s, and consequently the cavity-less pulse source was designed to generate pulses with 2 GHz repetition rate. Outgoing from the aforementioned principle of pulse generation, four electro-optical modulators were used in the pulse generation. In order to generate pulses with picosecond duration and at the same time maintain as high optical signal to noise ratio (OSNR) as possible through the modulators, the underlying order of the modulators was optimized as well as the fundamental modulation frequency. The experimental setup of the pulse source is depicted in Fig. 2 . A 100-mW distributed feedback (DFB) laser was used as a CW seed and initially the light was modulated by two concatenated PMs. As mentioned above, the frequency deviation is proportional to modulation amplitude and modulation frequency, and consequently by increasing modulation amplitude and the frequency deviation (chirp) is enhanced, as well as by adding additional PMs. In this experiment two PMs were driven by the second and fourth RF harmonics, respectively. In between and after the two phase modulators, two EDFAs were used in order to compensate for the insertion loss of the modulators. The additive noise from the amplifier was minimized by having the EDFAs operated in saturation. The subsequent amplitude modulator was used to carve out the seeding pulses. The modulator was driven by an 8 GHz RF tone by a three-stage amplifier that was set in such a way that the first two stages were saturating the output stage, creating a drive signal with a three-fold temporal compression, which was measured with an electrical oscilloscope. After further amplification, a second MZM modulator was used to select one out of four pulses in order to reduce the pulse repetition rate from 8 GHz to the target 2 GHz. The pulses were subsequently propagated through a DCF in order to achieve linear compression, whereas the 64 m DCF length provided the optimal compression in the actual setup and resulted in the shortest pulse duration.
Fig. 2 Experimental setup. LD: lasers diode, EDFA: Erbium doped fiber amplifier, PPG: Pulse pattern generator PM: phase modulator, PS: Phase shifter, MZM: Mach-Zehnder modulator
3. Sampling experiment
The generated high quality optical pulses can be used in the gating function of an optical sampling preprocessor, followed by quantization using electronic ADC. Using short temporal optical pulses the bandwidth operation can be extended beyond the capability of the electronic ADC and to band limited signals with frequency higher than the Nyquist bandwidth (half the sampling rate) can be captured. However, if the signal frequency is higher than the sampling rate, the sampling theorem is not fulfilled and as a result the signal is under-sampled or sub-rate sampled. Consequently, the detected signal will be an aliased version of the real signal, which can be described by
where frec is the detected signal frequency, fsig is the signal frequency, fs is the sampling frequency, and N is an integer defined byMoreover, this results in an ambiguous frec as many input signal frequencies can satisfy (1) and thereby uncertainty in the actual received signal frequency. This is illustrated in Fig. 3 , where several sampling conditions are exemplified. However, by knowing N, the aliasing effect can be used to directly down convert a signal within the frequency band [Nfs/2,(N + 1)fs/2]. Note that this operation is equivalent to the basic function of an electronic mixer down-converting a signal frequency to an intermediate frequency (IF) by mixing it with a local oscillator (LO).Fig. 3 Example of the received signal samples under different sampling conditions: Nyquist sampling fs/2>fsig (blue line and green squares), and sub-rate sampling fs<<fsig (red dots).
An optical sampling pre-processor was constructed to demonstrate Nyquist sampling and sub-rate sampling by sending the generated pulses though a Mach-Zehnder modulator with dual outputs (DO-MZM) driven with the signal under test, as depicted in Fig. 4 . The sample information was imprinted on the incoming pulses and inherently the two outputs of DO-MZM had complimentary information. The two optical sample outputs were sent in two parallel ADCs with matched path lengths, which were synchronized with the pulse source. The captured channels were subsequently subtracted in order to reduce the common mode noise [4,12,13] e.g. laser relative intensity noise (RIN) and amplified spontaneous emission (ASE), which is analogous to using a balanced detector, in order to improve the SNR of the digitized signal. However, noise added after the modulator e.g. thermal and quantum (shot) noise, remained uncompensated.
Fig. 4 Setup of sampling experiment. DO-MZM: Dual-output Mach-Zehnder modulator, PD: Photo diode, LP: low pass filter.
4. Results
The seeding CW-laser centered at 1564.8 nm was phase modulated (chirped) by two succeeding PMs modulated with 16 and 32 GHz tones, respectively, which were harmonics derived from the fundamental driving frequency. Subsequently, 20-ps pulses were carved by a MZM driven with the saturated modulator driver amplifier that created a chirped pulse train with 8 GHz repetition rate, as is seen in Fig. 5(a) . The target pulse rate of 2 GHz was achieved by the second MZM operated with pattern of “…1000…” to reduce the pulse repetition rate by gating (i.e. transmitting) one out of four pulses. After propagation through the DCF, the pulses were compressed to a temporal width of 3 ps. Optical sampling oscilloscope captures of the pulses are shown in Fig. 5, with part 5(a) showing the uncompressed 8 GHz pulses and 5(b) and 5(c) parts showing the compressed pulses at 8 and 2 GHz, respectively. The optical path was carefully optimized to maintain a high SNR by keeping optical power at as high level as possible while the EDFAs were used to compensate for insertion and modulation losses. The noise degradation due to loss and subsequent amplification was kept to minimum by ensuring that the optical amplifiers were operated in deep saturation. The SNR measured by the optical sampling oscilloscope surpassed the 40 dB level, however it ought to be stressed that this measurement was limited by the internal noise of the oscilloscope. Furthermore, the phase noise of the RF-source and pulse source was characterized using the direct spectrum technique using an electrical spectrum analyzer [14]. The results of the phase noise measurements are shown in Fig. 6 . As predicted, the phase noise of the pulse source at 8 GHz repetition rate is almost identical to the phase noise of the driving RF source. The phase noise at 2 GHz repetition followed the same trend at 8 GHz, albeit with a 12 dB lower value. The observed behavior is predicted by theory and is a consequence of the lower repetition rate, which is characterized by an equal amount of timing jitter. Due to limitations of the available phase noise measurement setup, no measurements of the phase noise at frequency over 1 MHz could be measured as the observed phase noise was below the noise floor of the instrument which is limited by the internal oscillator performance and instrument internal noise. Therefore, no proper estimation of the pulse source jitter could be estimated following the procedure in [3], which requires that the phase noise be considered from the inverse of the sample time (record length) and up to precisely half the sampling repetition rate for a satisfactory jitter information measurement, taking all spectral phase noise contribution into account.
Fig. 5 Optical sampling oscilloscope captures of signals generated by pulse source (a) 8 GHz chirped pulses before compression, (b) 8 GHz compressed pulses and (c) 2 GHz compressed pulses.
Fig. 6 Phase noise measurement of the RF-source at 8 GHz and the pulse source at 8 and 2 GHz, respectively.
The input electrical test signal was varied between 0 and 40 GHz and was sent to the DO-MZM, where it was sampled by optical pulses sent through the modulator. The modulator was biased at the quadrature point in order to minimize the second order distortion. The pulses from the two complimentary outputs of the DO-MZM were received by two 20-GHz bandwidth photo diodes. To match the input peak-to-peak voltage of the electrical ADC, an electrical ADC front end was used consisting of a low pass filter (LP), high dynamic range amplifier, DC lock and an additional low pass filter. The first filter bandwidth (LP1) was optimized for optimum operation (as seen in Fig. 7 ) at 5.2 GHz. The role of the first filter was to reduce the noise, as well as to limit the pulse bandwidth, and reduce the pulse peak power to avoid saturation effects in the subsequent amplifier. The second low pass filter had a bandwidth of 1.9 GHz. The ADCs used were commercial high performance two-channel 2 GS/s 12-bit ADCs with 9 ENOBs and a SINAD of 56 dB. The ADCs were synchronized with the pulse source, and a phase shifter was used to ensure that the peak of the sampling pulses was captured. Potential improvement of the receiver can be realized by using an integrate and dump (IAD) and TAH circuits that convert the pulse to a low bandwidth square-like pulse in order to significantly reduces the jitter requirements on the subsequent ADC [15].
The recorded data from the complementary ADC channels were processed off-line on a personal computer. The inherent MZM modulator nonlinearity, giving rise to strong tones at odd harmonics, was mitigated employing an equalization method relying on a two dimensional look-up Table (2D-LUT) correction [16]. Look-up table based correction circumvents any computational operations and involves only limited latency associated with a random access memory (RAM) readout and is readily implementable in the existing FPGA technology.
The two channels were subtracted and the amplitude mismatch was corrected for by optimization, resulting in the common mode noise reduction, which is predicted to give a 3 dB improvement in SNR. A single sine-wave was used as the system test signal, which was characterized using the IEEE standard metric [11]. The time domain signal and fast Fourier transform (FFT) of the two DO-MZM outputs are shown in Figs. 8(a) and 8(b) at 10.202 GHz frequency, demonstrating the response of the two channels to be almost indistinguishable in performance. The FFT of the combined signal of the two channels is shown in Fig. 8(c), and as predicted, the common mode noise cancellation reduces the noise floor by about 3 dB, giving rise to an equal amount of the SNR improvement. The individual channels were characterized by SINADs of 47.5 dB and 47.4 dB, respectively, whereas the combined channels had a SINAD of 50.4 dB (i.e. the afore-mentioned 3-dB improvement).
Fig. 8 (a) Time domain signal of the two DO-MZM outputs, (b) FFT of the two DO-MZM outputs, and (c) FFT of the two channels combined of 10.202 GHz input.
The FFT spectra of the tested signals varying from 0.2 to 40 GHz are shown in Fig. 9(a) - 9(e), captured with 65000 sample record length. We note that the sampling aliasing effect of using sub-rate sampling makes the frequency appear at 202 MHz, as predicted by Eqs. (1) and (2). Moreover, as seen in Figs. 8 and 9, the 2D-LUT effectively reduces the static nonlinearities dominated by the MZM, with spurious tones remaining after correction due to the channel calibration limitation [17]. The calculated ENOB and ENOB limited by SNR (ENOBSNR) are shown in Fig. 10 , where the ENOB performance spans from 8.2 to 7.0 between 0 and 40 GHz. The uniform performance deterioration is recognized to render from the corresponding drop in the frequency response of the DO-MZM, which had a 3 dB bandwidth of 22 GHz, as well as from increasing sampling gate requirements with frequency [3]. It is important to note, however, that the performance of a 40 GHz signal is 7.0 ENOB, similar to the results reported for sub-rate sampling using a high performing mode-locked laser as the pulse source [5]. Moreover, a demonstration of the signal capture capability of the system is shown in Fig. 11 , where a 1-ms-pulse modulated carrier at 40.202 GHz is captured in real time.
Fig. 9 FFT of the two channels combined with signal having a frequency at (a) 0.202 GHz, (b) 1.798 GHz, (c) 14.202 GHz, (d) 20.202 GHz, (e) 30.202 GHz, and (f) 40.202 GHz.
Fig. 10 ENOB and SNR limited ENOB as a function of signal frequency for the photonic preprocessor sub-rate sampled ADC operating at 2 GS/s.
Fig. 11 Capture of 1-ms-pulse modulated 40.201 GHz carrier using the photonic preprocessor at 2 GS/s.
5. Conclusion
We have designed and constructed a high fidelity optically gated sub-rate sampled ADC preprocessor based on a cavity-less pulse source and a dual-output Mach-Zehnder modulator (DO-MZM). The pulse source was operating at 2 GHz with 3 ps pulsewidth, with an SNR of over 40 dB. The phase noise of the pulse source was dictated only by the driving RF source, which was fully confirmed by the performed phase noise measurements. Furthermore, rigorous characterization of the sampling ADC pre-processor system was performed using the IEEE standard, and the effective number of bit (ENOB) levels of 8.2 and 7.0 were achieved at input signals of 0.2 and 40 GHz, respectively. The observed performance degradation with frequency was mainly due to the frequency bandwidth of the DO-MZM. The presented result constitutes the highest to date reported performance based on a cavity-less based pulse source in photonic sampled ADC and fully validates the utilization of cavity-less pulse sources as a strong alternative to mode-locked pulse sources.
Acknowledgment
This material is based in part on research sponsored by the Office of Naval Research (ONR).
References and links
1. Y. Bouvier, A. Ouslimani, A. Konczykowska, and J. Godin, “A 1-gsample/s, 15-GHz input bandwidth master–slave track-and-hold amplifier in inp dhbt technology,” IEEE Trans. Microw. Theory Tech. 57(12), 3181–3187 (2009). [CrossRef]
2. S. L. Morton, A. E. Cosand, D. A. Hitko, C. Baringer, L. Luh, C.-M. Lin, J. F. Jensen, C.-M. Li, and D. Crampton, “Ku-band subsampling track-and-hold amplifier with 8 enob resolution,” Electron. Lett. 42(8), 459–460 (2006). [CrossRef]
3. G. C. Valley, “Photonic analog-to-digital converters,” Opt. Express 15(5), 1955–1982 (2007). [CrossRef] [PubMed]
4. P. W. Juodawlkis, J. C. Twichell, G. E. Betts, J. J. Hargreaves, R. D. Younger, J. L. Wasserman, F. J. O'Donnell, K. G. Ray, and R. C. Williamson, “Optically sampled analog-to-digital converters,” IEEE Trans. Microw. Theory Tech. 49(10), 1840–1853 (2001). [CrossRef]
5. A. Khilo, S. J. Spector, M. E. Grein, A. H. Nejadmalayeri, C. W. Holzwarth, M. Y. Sander, M. S. Dahlem, M. Y. Peng, M. W. Geis, N. A. DiLello, J. U. Yoon, A. Motamedi, J. S. Orcutt, J. P. Wang, C. M. Sorace-Agaskar, M. A. Popović, J. Sun, G.-R. Zhou, H. Byun, J. Chen, J. L. Hoyt, H. I. Smith, R. J. Ram, M. Perrott, T. M. Lyszczarz, E. P. Ippen, and F. X. Kärtner, “Photonic ADC: overcoming the bottleneck of electronic jitter,” Opt. Express 20(4), 4454–4469 (2012). [CrossRef] [PubMed]
6. T. L. Koch and R. C. Alferness, “Dispersion compensation by active predistorted signal synthesis,” J. Lightwave Technol. 3(4), 800–805 (1985). [CrossRef]
7. J. J. Veselka and S. K. Korotky, “Pulse generation for soliton systems using lithium niobate modulators,” IEEE J. Sel. Top. Quantum Electron. 2(2), 300–310 (1996). [CrossRef]
8. A. O. J. Wiberg, C.-S. Brès, B. P.-P. Kuo, E. Myslivets, and S. Radic, “Cavity-less 40 GHz pulse source tunable over 95 nm,” in Proc. ECOC’09, paper 5.2.3 (2009).
9. A. O. J. Wiberg, B. P.-P. Kuo, C.-S. Brès, N. Alic, and S. Radic, “640-Gb/s Transmitter and Self-Tracked Demultiplexing Receiver Using Single Parametric Gate,” IEEE Photon. Technol. Lett. 23(8), 507–509 (2011). [CrossRef]
10. A. O. J. Wiberg, L. Liu, Z. Tong, E. Myslivets, V. Ataie, N. Alic, and S. Radic, “Cavity-less pulse source based optical sampled ADC,” in European Conference and Exhibition on Optical Communication (ECOC) (Optical Society of America, Washington, DC, 2012), Mo.2.A.3.
11. I. E. E. E. Std, 1241–2000, (IEEE, Inc., New York, NY, USA, 2001).
12. E. Ackerman, S. Wanuga, J. MacDonald, and J. Prince, “Balanced receiver external modulation fiber-optic link architecture with reduced noise figure,” IEEE MTT-S Int. Microw. Symp. Dig. 2, 723–726 (1993). [CrossRef]
13. K. J. Williams, L. T. Nichols, and R. D. Esman, “Photodetector nonlinearity limitations on a high-dynamic range 3 GHz fiber optic link,” J. Lightwave Technol. 16(2), 192–199 (1998). [CrossRef]
14. Agilent X-series signal analyzer manual, “N9068A & W9068A Phase Noise Measurement Application Measurement Guide,” (Agilent Technologies, 2012). http://cp.literature.agilent.com/litweb/pdf/N9068-90011.pdf.
15. T. D. Gathman and J. F. Buckwalter, “An integrate-and-dump receiver for high dynamic range photonic analog-to-digital conversion,” in Proc. IEEE 12th Topical Meeting Silicon Monolithic Integrated Circuits in RF Systems (SiRF), TU1C–4 (2012).
16. F. H. Irons, D. M. Hummels, and S. P. Kennedy, “Improved compensation for analog-to-digital converters,” IEEE Trans. Circ. Syst. 38(8), 958–961 (1991). [CrossRef]
17. V. Ataie, E. Myslivets, A. O. J. Wiberg, B. P. P. Kuo, N. Alic, and S. Radic, “Linearization of copy-and-sample photonics analog to digital converter based on a multidimensional look-up table,” in Optical Fiber Communication Conference, OSA Technical Digest (Optical Society of America, 2012), paper OM2B.6.
