Open Access
Add to BibSonomyAbstract
We demonstrate a flexible RF photonics down-converter that enables spectral folding of a 2-18 GHz RF band into a common 2 GHz wide intermediate frequency (IF) band for identification of specific signals of interest. The system can then be reconfigured for selective down-conversion of a given sub-band to a common IF output. We present an analysis of the performance of the down-converter and experimentally demonstrate both the spectral folding and selective modes. A sensitivity of −42 dBm in an IF bandwidth of 2 GHz and a spurious free dynamic range of 103 dB.Hz2/3 is achieved in the spectral folding mode.
© 2017 Optical Society of America
1. Introduction
Radio frequency (RF) technology continues to play a key role in many commercial and military applications. These applications frequently require distribution, processing, and detection of analog RF signals. The application of photonics for these functions, where the analog RF signal is upconverted to an optical carrier, is beneficial as it presents the opportunity for ultra-low loss and ultra-wide bandwidth with low size, weight, and power (SWaP). RF photonics is particularly valuable for signal processing of wideband RF signals at high frequencies by relieving the bandwidth limitations of traditional RF electronic systems. With a growing number of applications looking to leverage the millimeter wave spectrum, photonic technologies could be transformative in enabling signal processing operations, where conventional RF technologies fall short or do not exist. Optical signal processing functionalities that have been considered include filtering [1], frequency channelization [2,3], frequency translation/down-conversion [4–6], and spectral monitoring [7].
In this paper, we present a novel photonically-enabled system for wideband spectral monitoring and down-conversion of RF signals using relatively low-bandwidth photodetection to support both operations. This is achieved in a two-step process via spectral folding followed by selective down-conversion. First, a spectral folding process enables down-conversion of multiple frequency sub-bands to a common intermediate frequency (IF) band, where, dithering of the optical local oscillator (LO) allows for clear identification of the sub-band where a particular RF signal may be present. Once identified, a specific sub-band of interest can be selectively down-converted to an IF and further digitally processed, if desired. Note that a single architecture supports both folding and selective modes of operation.
Previously, spectral folding of RF signals has been considered using a sampled analog optical link where the optical LO is modulated by the input RF signal [8,9]. Here, we employ a different approach that avoids the RF modulation of the optical LO, thus decoupling the LO from the RF signal. This not only allows for independent optimization of each path but also presents a less complex signal at the photodetector and its output. We present experimental results on down-conversion of RF signals with carrier frequencies in 2-18 GHz to a 2 GHz IF passband, some of which were previously presented in [10]. This paper extends those results and provides a comprehensive description that includes an analytical performance analysis and an experimental study of the down-converter system.
2. System concept
We first describe the fundamental working principles of our novel dual-mode RF down-converter. The conceptual system diagram is shown in Fig. 1(a). The output from a CW laser with a nominal center frequency of fo is split into two parallel paths. The upper path performs upconversion of the RF input signal band (f1 to f2) to double sidebands around the optical carrier.
Fig. 1 (a) Conceptual system diagram of dual-mode down-converter and (b) illustration of optical and IF spectra in folding and selective modes.
The lower path in Fig. 1(a) is used for optical LO generation, which is configured according to the desired down-conversion mode of operation: folding or selective. In the folding mode, the optical LO consists of a flat comb of carriers centered around the optical carrier and spaced by the LO drive frequency, fLO. In the selective mode, a single pair of optical LO carriers are generated at fo ± fLO. Depending upon the particular LO mode that is configured, different portions of the input RF signal band are detected as discussed below.
The two parallel optical paths are combined using an optical coupler followed by optical bandpass filtering. As illustrated in Fig. 1(b), the optical filters are used to extract the upper sideband components of both the optically-upconverted RF signal and the optical LO. The lower optical sideband is suppressed along with any optical carrier present at fo. The carrier is suppressed to minimize shot noise, which would otherwise result from its detection, and removing the lower sideband minimizes noisy beat terms which would otherwise appear during detection due to optical interferometer instability. It may be noted that the suppression of the lower sideband by the optical filter is more challenging and might limit the performance at the lower edge of sub-band 1 (2 – 6 GHz). The composite optical signal/optical LO spectrum is detected by a balanced receiver. The low-pass filter following this receiver determines the IF bandwidth, fIF, by selecting only the desired beat products. Nominally, fIF is equal to half the optical LO comb spacing to provide continuous spectral coverage in the folding mode of operation.
When the optical LO generator is configured for folding mode, the resulting optical LO comb divides the input RF band into sub-bands of width 2fIF (i.e. ± fIF from each LO comb line) as illustrated in the upper left of Fig. 1(b). Multiple signals of interest may appear in any particular sub-band, and after beating against their closest LO in the balanced detector, they will be aliased, or folded down to the common IF band, as illustrated in Fig. 1(b), lower left. To identify the sub-band a given RF signal lies in, we employ a frequency dither technique, described further below. The folding technique thus permits wideband spectral monitoring using relatively low bandwidth detection.
When the optical LO generator is configured for the selective mode, a single LO comb line (at fo + fLO) is situated within the sub-band of interest, as illustrated in the upper right of Fig. 1(b). Only RF signals that fall within + fIF of that optical LO will be detected at the receiver. Any coherent beat products produced by undesired signals at larger frequency offsets will be suppressed by the receiver’s electrical low-pass filters. Note that in the selective mode, RF signals that fall within -fIF of the optical LO will also be down-converted to the same IF frequency as RF signals that fall within + fIF. Such potential images can be an issue with passband architectures and could be resolved using a more complex receiver with separate I and Q detection and processing. However, the passband configuration presented here is best suited in applications where the input signal spectrum is sparsely populated, reducing the likelihood of images. Ideally, the optical LO is configured to generate a single LO at fo + fLO. However, the optical modulator employed for LO generation may produce higher-order terms that can result in LO lines in sub-bands other than the intended sub-band, which will lead to the down-conversion of any RF signals present in those sub-bands as well. The power of these higher-order terms is generally more than 20 dB lower than the desired LO. In addition, as shown in Fig. 1(a), the optical BPF will suppress the unwanted LOs, which generally fall outside its passband, further. Selectivity will thus depend on the suppression of the unwanted LO and the relative strength of the signals.
In order to identify in which carrier frequency sub-band an RF signal is present while operating in the folding mode, we apply a frequency dither technique. Similar frequency dither techniques have previously been applied to sampled analog optical links [9]. The concept as applied in our RF down-converter architecture is illustrated in Fig. 2. Example RF signals are shown in the first and third sub-band, which when the LO comb generator is driven at fLO, generate IF signals centered at fIF1 and fIF2, respectively. When the LO drive frequency is increased to fLO + δ, as shown at the bottom of Fig. 2, higher frequency LO comb lines experience larger frequency excursions such that the line offset frequency in each sub-band will scale as n x δ, where n is the sub-band index. This results in the IF of the two example spectra in Fig. 2 shifting by -δ and + 3δ, respectively. As a result, the sign of the IF shift can be used to unambiguously identify on which side of an LO comb line an RF signal may be present, while the magnitude identifies in which particular sub-band the signal lies.
Fig. 2 Illustration of the frequency dithering technique used to identify the location of RF signals in various frequency sub-bands.
When operating at baseband, the two interferometric paths would require stabilization or I/Q recovery is needed. In contrast, the interferometer in our passband down-conversion architecture does not require active stabilization at the wavelength scale. In the folding mode, where the intent is to measure spectral content and not recover any signals, the LO dithering ensures the relevant mixing products are always generated above baseband. In the selective mode, where the objective is to recover the down-converted signal, it is assumed the interferometer drift is sufficiently slow that it can be tracked by digital signal processing elements at the IF output, thereby suppressing the additional signal phase noise it would otherwise introduce.
3. Theoretical analysis
The link transfer function for a coherent, filtered analog link with a parallel architecture similar to that employed in Fig. 1 has been derived previously [5]. The results are applicable to the architecture presented here, both for the selective mode as well as the folding mode for an input RF signal within a sub-band, i.e., the analysis may be applied individually to each input signal spectrum. The derivation of the link transfer function is presented here briefly for completeness. It is then used to assess the down-converter system performance.
Consider a bandpass input RF signal centered at frequency ωRF, with amplitude ρ(t) and phase ϑ(t)
The optical field E1 after a dual-drive Mach-Zehnder modulator (MZM), driven by complementary RF signals (Z(t), –Z(t)) is given byWhere is the electric field of the laser, φ is the MZM bias, , and Jk(.) is the kth order Bessel function of the first kind. The signal after the MZM consists of sidebands centered at kωRF around the optical carrier. When the MZM is biased for carrier suppression (φ = π), E1(t) consists of only odd order sidebands, while biasing at the peak (φ = 0) results in only even order sidebands along with the modulated optical carrier. The optical field after bandpass filtering to select one sideband, for example the kth sideband, is given byFor and the optical field after the filter is given byThe signal after coherent downconversion is given by the beat term between the signal and the LOwhere is the electric field of the LO laser, ωIF is the difference between center frequencies of signal and LO lasers, η is the photodiode responsivity in A/W, and a = 2 for balanced detection (a = 1 for single-ended detection). Equation (5) is the link transfer function for a coherently-detected filtered RF link such as shown in Fig. 1. Next we analyze the sensitivity and dynamic range performance in such a link.The sensitivity is defined here as the minimum input RF power required at the signal MZM input in order to observe an output IF equal to the noise floor in the IF bandwidth. It can be derived based on the net RF link power gain/loss and the output RF noise power spectral density, Nout. Nout is given by the sum of noise power spectral density from various noise sources such as shot noise (Nshot), laser relative intensity noise (RIN) (NRIN), and detector trans-impedance amplifier (TIA) noise (NTIA).
These noise power spectral density terms in a bandwidth of 1 Hz are given by [11] where qe is the electron charge, RL is the load impedance, RIN is the laser RIN, ε is a RIN suppression factor obtained from balanced detection, IN is the TIA noise current, and Ipd is the dc photocurrent at one detector. With coherent detection, Ipd ≈ηPLO where PLO is the optical power in the LO at the detector with . Assuming a relatively low RIN laser combined with balanced detection, shot noise is generally the more dominant noise component compared to RIN noise. For example, lasers having a low RIN = −162 dBc/Hz are commercially available. In our experimental configuration employing balanced detection and assuming ε = 0.05, the laser RIN is lower than shot noise as long as Ipd < 100 mA.Dynamic range is typically characterized by intermodulation distortion (IMD) measured using a narrowband two-tone RF signal at frequencies f1 and f2. Nonlinearity results in generation of IMD components, of which third-order IMDs (IMD3) at frequencies 2f1-f2 and 2f2-f1 generally determine the usable dynamic range. The spurious free dynamic range (SFDR) is defined as where OIP3 is the output third-order intercept point, which is obtained from linear extrapolation of the fundamental and IMD3 responses [12]. Further, OIP3 = IIP3*Gp, where IIP3 is the input intercept point and Gp is the link RF power gain. Two-tone tests in optically filtered links result in an, where RS is the input impedance [5], while the RF power gain for the down-converter shown in Fig. 1 is calculated from Eq. (5).
We illustrate the impact of modulator Vπ and input laser power on the system sensitivity and SFDR for the downconverter operating in the folding and selective modes of operation in Figs. 3 and 4, respectively. It is assumed that the IF bandwidth is 2 GHz, TIA noise current is 19.8 pA/Hz1/2, a = 2, laser RIN = −162 dBc/Hz, ε = 0.05, Rs = 50 ohm, RL = 500 ohm, and η = 0.8. The total optical insertion loss due to modulator, bandpass filter and coupler is assumed to be 12.5 dB. These assumptions are based on the experimental parameters. Note that the shot noise contribution is increased in the folding mode due to the presence of multiple LO comb lines at the detector whereas in the selective mode, it arises only from a single LO line, albeit of higher power.
Fig. 3 Calculated sensitivity [dBm] for an IF bandwidth of 2 GHz (left) and SFDR [dB.Hz2/3] (right) of the downconverter as a function of modulator Vπ and input laser power in folding mode.
Fig. 4 Calculated sensitivity [dBm] for an IF bandwidth of 2 GHz (left) and SFDR [dB.Hz2/3] (right) of the downconverter as a function of modulator Vπ and input laser power in selective mode.
Figures 3 and 4 show that the system sensitivity (in an IF bandwidth of 2 GHz) and SFDR both depend on the input laser power and improve as it is increased. Improvements in performance will, however, depend on the limits of the operating optical power in the down-converter system, and might be constrained by the laser, the modulator, the detector, or optical losses in the link. Further, sensitivity is improved as modulator Vπ decreases, while the SFDR is independent of Vπ. Low Vπ modulators are key for achieving high system gain and thus high sensitivity. Finally, system performance is better in the selective mode compared to the folding mode due to the higher LO power in the selective mode, resulting in higher RF gain.
4. Experimental setup
The experimental setup for the photonic down-converter supporting both modes of operation is shown in Fig. 5. The output of a 22 dBm low-noise CW laser at 1550 nm (193.4 THz) and with linewidth less than 1 kHz is split by a 50/50 polarization maintaining splitter into two paths. The upper path consists of a dual-electrode Mach-Zehnder modulator (MZM), differentially driven with a 0/180° hybrid and biased for carrier suppression. The modulator upconverts the incoming RF signal having a carrier frequency in the range 2-18 GHz to optical frequencies. The MZM, which has a Vπ of 5 V (at 2 GHz) and a 3-dB bandwidth of 25 GHz is followed by a tunable optical bandpass filter (BPF) with a 3-dB bandwidth of 5 GHz. The BPF selects one optical sideband of the upconverted RF signal, while rejecting the other sideband along with the residual optical carrier.
Fig. 5 Experimental setup of dual-mode photonic down-converter. MZM: Mach-Zehnder modulator, PM: Phase modulator, BPF: Bandpass filter, HJ: Hybrid junction, RF: Radio frequency, LO: Local oscillator, IF: Intermediate frequency.
The lower path is the optical LO generation path. A serial cascade of a MZM followed by a phase modulator (PM) is employed for generation of the LO in both folding and selective modes. The MZM has a 3 dB bandwidth of 25 GHz while the PM has a bandwidth of 10 GHz. In the folding mode, both modulators are driven at 4 GHz to generate a comb of frequencies at a frequency spacing of 4 GHz [13]. Comb lines are generated at frequency offsets of 4-16 GHz from the optical carrier. Once optimized, the flatness of comb lines is ~2 dB. The gain, sensitivity and SFDR depend on the power of the LO. The flatness of the comb lines thus impacts these metrics in the folding mode. As described earlier, in the folding mode, a frequency dither of 100 MHz is applied to identify the particular sub-band that the incoming RF signal lies in. In the selective mode, the PM drive signal is disabled. Also, the drive frequency for the MZM is tuned in the range of 2-18 GHz to generate a single LO sideband tone so that the resulting IF after photodetection lies within a 2 GHz passband. Note that the RF signal and LO paths are decoupled and function independently in this architecture, thus allowing for greater control of signals on each path, enabling overall system performance optimization.
Next, the signal and LO paths are combined using a 50/50 polarization maintaining coupler. The outputs of the 50/50 coupler are sent to 10 GHz bandwidth photoreceivers, after which they are combined in a 180° hybrid junction (HJ) to enable balanced detection. The HJ has a 2 GHz upper operating frequency. The 2 GHz operating frequency of the HJ is well-matched to the 4 GHz optical comb spacing. In the experiment, the average optical LO power per tone at the receivers is −12 dBm in the folding mode and −5 dBm in the selective mode so that the total power level to each photoreceiver is limited to −1 dBm in order to stay below their damage threshold. The photoreceivers have a 500-ohm trans-impedance, resulting in a conversion gain of 400 V/W. The output of the hybrid junction is monitored on an RF spectrum analyzer. The photonic down-converter thus enables wideband continuous spectral coverage of input signals ranging from 2 to 18 GHz, while employing relatively lower bandwidth detection.
5. Experimental results
Figure 6 shows measured electrical and optical spectra for the photonic down-converter operation in both folding and selective modes. Figure 6(a) shows an example RF input, which consists of a two-tone signal centered at 17.5 GHz with ± 1 MHz offset. Figure 6(b) shows the optical spectrum after upconversion to an optical signal followed by bandpass filtering to select the lower sideband. Plots are shown for RF carrier frequencies centered at 9.5 and 17.5 GHz (note that the optical spectrum analyzer cannot resolve the individual ± 1 MHz tones). Figure 6(c) shows the optical LO spectrum in the folding mode (solid dark green). Comb lines with spacing of 4 GHz are seen as the modulators in the LO path are driven with a 4 GHz sine wave. Figure 6(c) also shows the optical LO spectrum in the selective mode (dashed light green). In this case, the drive frequency for the MZM in the LO path is tuned to 16 GHz while the PM is disabled. This results in sidebands at ± 16 GHz from the optical carrier, which is nominally suppressed. The electrical IF spectrum after balanced coherent detection is shown in Fig. 6(d). Beating of the RF signal optical sideband at 17.5 GHz and the optical LO sideband at 16 GHz results in an IF output centered at 1.5 GHz. Note, frequency components due to third-order intermodulation distortion (IMD3) offset ± 3 MHz from 1.5 GHz appear in the IF spectrum primarily due to the nonlinearity of MZM sinusoidal transfer response.
Fig. 6 Electrical and optical spectra at various measurement points of the photonic down-converter in both wideband and selective modes. (a) Two-tone RF input centered at fRF = 17.5 GHz with ± 1MHz offset, (b) 9.5 and 17.5 GHz two-tone input signal after upconversion to an optical carrier, (c) Optical LO in folding (4 GHz-spaced comb) and selective modes (16 GHz-offset sideband), (d) IF output in folding mode with down-conversion to fIF = 1.5 GHz.
Since the input RF signal is restricted to one sub-band, similar IF spectra are obtained in both the folding and selective modes. However, the IF output power in the selective mode is approximately 7 dB higher due to higher optical power in the LO tone, resulting in increased coherent gain. In the selective mode, the laser power used for optical LO generation is concentrated to only two dominant sideband tones (only one which is used in mixing with signal at the photodetector), whereas it gets divided amongst a larger number of comb lines in the folding mode.
5.1 Two-tone SFDR measurements
SFDR measurements were performed with the down-converter for a narrowband two-tone RF input signal centered at 17.5 GHz with ± 1 MHz offset and downconverted to an IF of 1.5 GHz. For a two-tone signal at frequencies f1 and f2, nonlinearity results in the generation of third-order distortion at frequencies 2f1-f2 and 2f2-f1 that generally determine the SFDR. To characterize the SFDR, the RF input power (per tone) was varied from −6.5 to + 3.5 dBm while the fundamental signal at f1 and f2 along with the third-order distortion products were monitored at the IF output on an RF spectrum analyzer. The results of the SFDR measurements for the folding and selective modes of operation are plotted in Figs. 7 and 8, respectively.
Fig. 7 Measured fundamental and IMD3 output power of the downconverter operating in folding mode using two-tone RF input centered at 17.5 GHz and down-converted to a 1.5 GHz IF.
Fig. 8 Measured fundamental and IMD3 output power of the downconverter operating in selective mode using two-tone RF input centered at 17.5 GHz and down-converted to a 1.5 GHz IF.
The measured data points were linearly fit to extrapolate the input third-order intercept point (IIP3), which was 24 dBm in both modes. The system noise floor, measured at −150 dBm in a 1 Hz bandwidth, includes laser shot noise, trans-impedance amplifier (TIA) noise, and the spectrum analyzer noise but is primarily due to TIA noise and spectrum analyzer noise. Laser RIN noise was negligible. As shown in the plots, the SFDR at 17.5 GHz is 103 dB⋅Hz2/3 and 106 dB⋅Hz2/3 for the folding and selective modes, respectively. The improved SFDR in the selective mode results from the higher coherent gain due to a higher optical LO power, as discussed earlier.
5.2 Signal carrier frequency identification using dithering technique
Sub-band identification of the RF signal was also demonstrated experimentally. As discussed in Section II, the optical LO comb was toggled by applying a dither (δ = 100 MHz) to the drive frequency. As the optical LO comb spacing dithered between 4.0 and 4.1 GHz, the resulting IF beat note toggled between 1.5 and 1.6 GHz, respectively, when the RF input signal was set to 2.5 GHz. The frequency changed by + δ, indicating that the RF signal was located in the lower half of the first RF sub-band. In contrast, when the RF input was set to 5.5 GHz, the IF beat note toggled between 1.5 and 1.4 GHz (frequency change of -δ), confirming that the signal was located in the upper half of the first RF sub-band. Further, when the RF input signal was increased to 14.5 GHz, the IF beat note toggled between 1.5 and 1.9 GHz (frequency change of + 4δ), confirming that the signal was located in the lower half of the fourth sub-band. Finally, when the RF input signal was set to 17.5 GHz, the IF beat note toggled between 1.5 and 1.1 GHz (frequency change of −4δ), confirming that the signal was located in the upper half of the fourth sub-band. The magnitude and sign of the frequency change thus clearly identify the folding sub-band of the incoming RF signal.
5.3 Sensitivity and system gain
Next, we measured the sensitivity and RF system gain of the photonic down-converter in both folding and selective modes using a single input tone. We measure a sensitivity at a carrier frequency of 2.5 GHz of −42 dBm and −48 dBm for an IF bandwidth of 2 GHz in the folding and selective modes, respectively. Note, these experimental results match well with the analysis presented in Figs. 3 and 4. At 17.5 GHz, the measured sensitivity is −38.7 dBm and −46.7 dBm in the folding and selective modes, respectively. The sensitivity at 17.5 GHz drops due to a combination of higher MZM Vπ and a slight roll-off of the LO tone power further from the carrier.
The RF system gain is defined as the ratio of IF output and RF input at the 0/180° hybrid connected to the signal MZM, for a single tone RF signal input. The measured system gain at 2.5 GHz is −15 dB and −9 dB in the folding and selective modes, respectively. At 17.5 GHz, the system gain is −18.3 dB and −10.3 dB in the folding and selective modes, respectively.
Table 1 summarizes the system sensitivity and gain measured for RF signal carrier frequencies from 2 to 18 GHz for both folding and selective modes of the downconverter.
Table 1. Measured system sensitivity and gain for photonic down-converter in folding and selective modes over a carrier frequency range of 2-18 GHz.
6. Summary
We have presented a novel dual-mode RF down-converter with large fractional bandwidths and wide spectral coverage enabled by the wideband properties of photonics and the flexibility of electronics. A single architecture supports two modes of operation. The folding mode enables simultaneous monitoring of multiple RF sub-bands. Further, application of a simple dither signature on each LO comb line allows for a particular RF sub-band to be identified in the folding mode. Reconfiguration of the local oscillator allows switching to the selective mode, which enables down-conversion of a single RF sub-band for subsequent processing in the IF band.
We have experimentally characterized key RF performance metrics, including dynamic range, sensitivity, and gain over a carrier frequency range of 2-18 GHz in both modes of operation. The down-converter achieves a sensitivity of −42 to −38.7 dBm and −48 to −46.7 dBm over 2-18 GHz in the folding and selective modes, respectively. The SFDR at 17.5 GHz is 103 dB.Hz2/3 and 106 dB.Hz2/3 in the folding and selective modes, respectively. This performance is possible due to optimization of the overall system architecture for maximum RF system gain while taking into account real world component constraints, such as realistic insertion losses and optical power limitations. A theoretical analysis was also presented and good agreement between theory and experiment was observed. Continued advances in electro-optic modulator technology will permit scaling to operating frequencies well beyond 18 GHz, thus enhancing the opportunity for photonics to be applied in ultra-wideband coherent signal processing applications.
References and Links
1. J. Capmany, J. Mora, I. Gasulla, J. Sancho, J. Lloret, and S. Sales, “Microwave photonic signal processing,” J. Lightwave Technol. 31(4), 571–586 (2013). [CrossRef]
2. A. Wiberg, D. J. Esman, L. Liu, J. R. Adleman, S. Zlatanovic, V. Ataie, E. Myslivets, B. P.-P. Kuo, N. Alic, E. W. Jacobs, and S. Radic, “Coherent filterless wideband microwave/millimeter-wave channelizer based on broadband parametric mixers,” J. Lightwave Technol. 32(20), 3609–3617 (2014). [CrossRef]
3. X. Xie, Y. Dai, K. Xu, J. Niu, R. Wang, L. Yan, and J. Lin, “Broadband photonic RF channelization based on coherent optical frequency combs and I/Q demodulators,” IEEE Photonics J. 4(4), 1196–1202 (2012). [CrossRef]
4. T. R. Clark and M. L. Dennis, “Coherent optical phase-modulation link,” IEEE Photonics Technol. Lett. 19(16), 1206–1208 (2007). [CrossRef]
5. A. Agarwal, T. Banwell, and T. K. Woodward, “Optically filtered microwave photonic links for RF signal processing applications,” J. Lightwave Technol. 29(16), 2394–2401 (2011). [CrossRef]
6. C. Middleton, S. Meredith, R. Peach, and R. DeSalvo, “Photonic-based low phase noise frequency synthesis for RF-to-millimeter wave carriers and wideband RF-to-IF down-conversion,” in Military Communications Conference (IEEE, 2011), pp. 51–54. [CrossRef]
7. S. Pan and J. Yao, “Photonics-based broadband microwave measurement,” J. Lightwave Technol.in press.
8. J. D. McKinney and V. J. Urick, “Radio-frequency down-conversion via sampled analog optical links,” Photonics Tech. Branch, Naval Research Laboratory, NRL/MR/5650–10–9275 (2010).
9. S. R. Harmon and J. D. McKinney, “Precision broadband RF signal recovery in subsampled analog optical links,” IEEE Photonics Technol. Lett. 27(6), 620–623 (2015). [CrossRef]
10. P. Toliver, A. Agarwal, J. M. Dailey, and T. C. Banwell, “Photonics-based spectral folding and selective down-conversion of RF Signals,” in Microwave Photonics (MWP), IEEE International Topical Meeting on (IEEE, 2016), pp. 69–72.
11. C. Cox, Analog Optical Links (Cambridge University, 2004).
12. W. K. Burns, G. K. Gopalakrishnan, and R. P. Moeller, “Multi-octave operation of low-biased modulators by balanced detection,” IEEE Photonics Technol. Lett. 8(1), 130–132 (1996). [CrossRef]
13. V. Torres-Company and A. M. Weiner, “Optical frequency comb technology for ultra-broadband radio-frequency photonics,” Laser Photonics Rev. 8(3), 368–393 (2014). [CrossRef]
