High-sensitivity optical to microwave comparison with dual-output Mach-Zehnder modulators

www.nature.com/scientificreports OPEN Received: 29 December 2017 Accepted: 26 February 2018 Published: xx xx xxxx High-sensitivity optical to microwave comparison with dualoutput Mach-Zehnder modulators Mamoru Endo1, Tyko D. Shoji1 & Thomas R. Schibli1,2 We demonstrate the use of two dual-output Mach-Zehnder modulators (DO-MZMs) in a direct comparison between a femtosecond (fs) pulse train and a microwave signal. Through balanced detection, the amplitude-to-phase modulation (AM-PM) conversion effect is suppressed by more than 40 dB. A cross-spectrum technique enables us to achieve a high-sensitivity phase noise measurement (−186 dBc/Hz above 10-kHz offset), which corresponds to the thermal noise of a +9 dBm carrier. This method is applied to compare a 1-GHz fs monolithic laser to a 1-GHz microwave signal generated from photodetection of a free-running 500 MHz mode-locked laser. The measured phase noise is −160 dBc/ Hz at 4-kHz, −167 dBc/Hz at 10-kHz, and −180 dBc/Hz at offset frequencies above 100-kHz. The measurement is limited by the free-running 500-MHz laser’s noise, the flicker noise of the modified unitraveling carrier photodiode and the thermal noise floor, not by the method itself. This method also has the potential to achieve a similar noise floor even at higher carrier frequencies. Great progress in the generation of ultra-low phase noise optical pulse trains and associated optical frequency-division (OFD)-based microwave generation1–5 has enabled not only novel approaches to radar applications6, but also very-long baseline interferometry7, high-speed analog-to-digital conversion8 and timing distribution at large-scale scientific facilities such as next-generation X-ray free-electron lasers9–11. A recent study showed the phase noise of an OFD-generated 12-GHz carrier reaching −173 dBc/Hz at a 10-kHz offset from the carrier3. Also, fiber-based optoelectronic microwave oscillators have the potential to produce a 10-GHz signal with −160 dBc/Hz at a 10 kHz offset, while only requiring several hundred meters of optical fiber without any reference12. Characterizing such low phase noise microwaves requires a complex apparatus and long acquisition times because typical mixer-based phase detectors and local oscillators lack sufficient sensitivity and a cross-spectrum method with two phase detectors is essential. While some phase noise analyzers have a cross-spectrum option, the acquisition times required for measuring the state-of-the-art ultra-low noise oscillators is of the order of days in order to remove the phase noise of the local oscillators or the electrical noise from the detection circuit. Therefore, in order for the cross-spectrum method to eliminate the residual detection noise, it is still important to have high sensitivity even with a single-channel phase detector. Due to the physics in oscillators, there is potential for optical femtosecond (fs)-pulse trains to generate microwave signals with much lower jitter than signals produced by electronic means13. This suggests that a direct comparison between an optical reference pulse train and a microwave signal could greatly improve the achievable sensitivity and the dynamic range of microwave phase noise measurements. Recently, optical delay lines14–17, Sagnac loop fiber interferometers9,18, and dual-output Mach-Zehnder modulator (DO-MZM)-based phase detectors19 have been demonstrated. When used as phase detectors, DO-MZMs are simple to calibrate, provide a direct link between optical and microwave oscillators and offer a high sensitivity. The DO-MZM phase detectors have been applied to timing synchronization between a microwave and a fs optical pulse19. In such applications, the bias voltage drift of the DO-MZM may spoil its long-term stability because this voltage directly introduces timing error between the signals, so a special bias control technique is required. However, in contrast to timing synchronization, the effect of the bias drift is not critical for phase noise characterization. In fact, the sensitivity error was less than 0.01 dB in our experiment for a typical measurement over a million averages (Supplement 1). Another concern regarding the DO-MZM phase detector is the shot-noise of the balanced detector that is used to extract the phase difference between the optical and the microwave signals, and the intrinsic flicker noise of the DO-MZM. Here, we overcome this limitation with cross-spectrum techniques (Methods). 1 Department of Physics, The University of Colorado, Boulder, Colorado, 80309–0390, USA. 2JILA, NIST, and the University of Colorado, Boulder, Colorado, 80309-0440, USA. Correspondence and requests for materials should be addressed to T.R.S. (email: ) Scientific REPOrTS | (2018) 8:4388 | DOI:10.1038/s41598-018-22621-1 1 www.nature.com/scientificreports/ Figure 1. (a) Concept apparatus of a DO-MZM phase detector. (b) Phase error and discrimination signal. (i) (ii) and (iii) illustrate phase errors of < 0, = 0 and > 0, respectively. (iv) Shows the discrimination signal. Each blue circle corresponds to (i) to (iii). After a low-bandwidth phase lock, the residual voltage noise corresponds to the phase noise. Although fiber delay line methods have the ability to measure over a wide frequency range, they also require temperature-stabilized km-long optical fibers, which must be well-isolated from acoustic or mechanical vibrations. The use of telecom-band DO-MZMs can eliminate the difficulties associated with fiber delay lines. The nonlinear effects in the fiber can also be a concern for fs-sources because they typically limit the achievable dynamic range of fiber-based phase noise measurements. In contrast, DO-MZMs enable robust, high-sensitivity measurements. Furthermore, this method is largely insensitive to amplitude noise in both the microwave signal and the fs light source. In this regard, the proposed dual DO-MZM phase measurement setup is like an optoelectronic equivalent to a triple-balanced microwave mixer. In this paper, we compare a fs pulse train from a free-running 1-GHz monolithic laser20 to a 1-GHz microwave signal generated by OFD of a free-running laser. DO-MZMs and balanced photodetectors suppress the amplitude-to-phase modulation (AM-PM) conversion effect by more than 40 dB in the phase detection, and cross-spectrum techniques suppress the associated shot-noise. Even without any external optical or electronic references, the noise floor of this method is −186 dBc/Hz above 10-kHz offset, which exceeds state-of-the-art cross-spectrum based fiber delay analyzer (OEWaves, microwave PNTS) by 10 dB21. With the free-running OFD, the measured phase noise of the 1-GHz microwave reaches −160 dBc/Hz at 4 kHz and −167 dBc/Hz at 10 kHz. These values are limited by the noise of the 500-MHz laser, the flicker noise of the modified uni-traveling carrier photodiode (MUTC)22 (−130 dBc/Hz at 1 Hz offset) and the thermal noise floor of the 50-Ohm system. Notably, the last two nois (...truncated)