Stretched-Pulse Mode-Locked Laser Using Post-Cavity Stretching for Ultrahigh-Speed OCT

Abstract

We demonstrate a chirped fiber Bragg grating (CFBG) based stretched-pulse mode-locked (SPML) wavelength-swept laser that uses both intra-cavity and post-cavity stretching. The SPML laser with complementary post-cavity stretching offers versatility in OCT imaging speed and eliminates the need for optical buffering. Using an SPML laser with a 5-m intra-cavity CFBG and a 5-m post-cavity CFBG, providing a wavelength sweep rate of 1075 nm/μs, we achieved OCT imaging with a sensitivity of 97.4 dB and a single-sided 6-dB coherence length of 5.6 mm. This demonstrated more than two-fold increase in coherence length with a slight sensitivity drop of 0.6 dB compared to an SPML laser with the same wavelength sweep rate using intra-cavity stretching only. The CFBG-SPML laser with post-cavity stretching offers a compelling option for OCT systems, particularly when long imaging range is required.

I. Introduction

STRETCHED-PULSE mode-locked (SPML) lasers have emerged as one of the most promising light sources for ultrahigh-speed optical coherence tomography [1], [2], [3], [4], [5], [6]. In an SPML laser, active mode-locking provides a short optical pulse and intra-cavity chromatic dispersion is used to first time-stretch and then time-compress this pulse. By locating the output coupler after time-stretching, the SPML laser provides wavelength-swept light at repetition rates from several MHz to tens of MHz, which is determined by the amount of the chromatic dispersion used for the time-stretching. These speeds are beyond what can be achieved using OCT light sources based on mechanically driven filters. In addition, the wavelength sweeps that do not require mechanical filters ensure intrinsic phase stability between the laser and the imaging system [6].

The first SPML lasers used in OCT employed kilometers of optical fiber as the positive and negative dispersive media used to time-stretch and time-compress the mode-locked pulse (or vice versa). These sources were designed to operate in the 1550 nm wavelength region [1], [2], where positive and negative dispersive fibers are readily available. Later, multimeter continuously chirped fiber Bragg gratings (CFBGs), which reflects different wavelengths at different longitudinal locations, therefore resulting in chromatic dispersion, replaced the long fiber spools. Typically, a single CFBG was used in a theta cavity configuration [7]. The CFBG design greatly simplified the laser design [3] while also allowing operation at a 1300 nm [4], [5] and 1050 nm [6]. Recently, an all-polarization-maintaining fiber (PMF)-based CFBG-SPML laser was demonstrated, further simplifying the SPML operation with excellent stability [5], [6].

Time-stretching can also be used outside of a laser cavity to generate wavelength-swept light sources from a short, broadband optical pulse, generated, for example, by a mode-locked or supercontinuum light source [8], [9]. This post-cavity stretching (PCS) approach requires only a single-sign of dispersion (e.g., positive dispersion only), whereas the SPML approach based intra-cavity stretching (ICS) requires matched dispersive media of opposing signs. Both ICS and PCS can be used to generate OCT sources, each having its own set of properties and challenges.

In this letter, we present an SPML design that uses both ICS and PCS. There are three reasons why this can be advantageous. First, PCS allows one to reduce the speed of the SPML laser for the applications where ultrahigh-speed imaging is not necessary. Because CFBGs have a maximum length (currently 10 meters, limited by fabrication methods), the sweep duration of an SPML employing ICS laser cannot be lower than the round-trip travel time of the CFBG. With a 10-m long CFBG, this sets a sweep duration of approximately 100 ns at maximum. PCS using a second CFBG can allow the sweep duration to be further expanded. This is especially valuable as a means to reduce the cost and complexity of high-speed electronics required to capture the imaging data. While the second CFBG could be included within the cavity, this leads to a complex and lossy cavity architecture. Using the second CFBG for PCS may be preferrable. The second reason to implement PCS is that this approach can be an alternative to optical buffering. In a common CFBG-SPML laser architecture, the intra-cavity semiconductor optical amplifier (SOA) is modulated to suppress laser oscillation within the sub-cavity defined by transmission through the CFBG. However, this limits the duty cycle of the wavelength-swept laser output to be less than 50%. To achieve a duty cycle close to 100%, a post-cavity 2x optical buffer, which optically copied, delayed, and interleaved the laser output [4], can be utilized. A PCS can double the sweep duration of the laser output, and generate 100% duty cycle (at half the speed) without the need for optical buffering. Third, as we show in this work, the ICS plus PCS design can yield improved coherence length performance at the expense of an increased relative intensity noise (RIN), although this increased RIN remains small enough to be effectively removed by a balanced receiver.

II. Experimental Methods and Results

We experimentally built and compared two CFBG-SPML lasers, one based on ICS and one based on ICS plus PCS. Both all PM cavities, and both were provided an output wavelength sweep rate of 1075 nm/μs, as depicted in Fig. 1(a). The ICS design is similar to our prior work [5]. It used 10-m-long CFBG (Proximion) as an intra-cavity stretcher (10-m ICS) with a post-cavity 2x buffer. We achieved a wavelength sweep from 1240 nm to 1340 nm at a 9.8 MHz repetition with a 94% duty cycle, which corresponds to a sweep rate of 1075 nm/μs, as shown in Fig. 1(b, c). The ICS/PCS design used a 5-m-long CFBG for ICS and a second 5-m-long CFBG for PCS, achieving a wavelength sweep from 1240 nm to 1340 nm at a 9.1 MHz repetition with an 88% duty cycle, corresponding to the same 1075 nm/μs sweep rate (Fig. 1(b, c)).

Fig. 1.

(a) Schematics of CFBG-SPML lasers. The blue dashed boxes correspond to the 10-m ICS SPML laser and the red dashed boxes correspond to the 5-m ICS/5-m PCS SPML laser. The polarization-maintaining fiber and single-mode fiber segments are presented in black and gray colors, respectively. EOM, electro-optic intensity modulator; SOA, semiconductor optical amplifier; BOA, booster semiconductor optical amplifier; BPG, bit pattern generator; DL, delay line; PG, pulse generator; FRM, faraday rotating mirror; PC, polarization controller. (b) Integrated optical spectrums of the laser outputs. (c) Time traces of the laser outputs. Blue traces, 10-m ICS SPML laser; Red traces, 5-m ICS/5-m PCS SPML laser.

Figure 2(a) and (b) show the point spread functions acquired at varying depths for the ICS and ICS/PCS designs. A 35 GHz photoreceiver (New Focus 1474-A) and a 25 GHz sampling oscilloscope (Pico Technology 9301-25) were used for this measurement. The single-sided 6-dB coherence length of the 10-m ICS SPML laser was measured at 2.4 mm, whereas the coherence length of the 5-m ICS/5-m PCS SPML laser was measured at 5.6 mm, more than double that of the ICS design. Interestingly, the temporal profiles of the mode-locked pulses measured after the electro-optic intensity modulator (EOM) reveal that the 3-dB pulse width of the 5-m ICS/5-m PCS SPML laser (69.7 ps) is almost half that of the 10-m ICS SPML laser (126.1 ps), as shown in Fig. 2(c). The electrical pulse width from a bit pattern generator (Sympuls PAT5000) driving the EOM in resonance with the cavity round trip time was 200 ps for both cases, and the temporal profile of the mode-locked pulse was measured using the same high-speed photoreceiver and sampling oscilloscope through a 1% tap coupler temporarily connected for the measurement. The instantaneous linewidth of the wavelength-swept output, which determines the coherence length, should scale in proportion to the width of the mode-locked pulse, all other factors being controlled [4]. Thus, we suspect that the enhanced coherence length of the ICS/PCS design relative to the ICS design is substantially due to the ability of the 5-m ICS cavity to generate shorter mode-locked pulses. One possible reason why the 5-m CFBG generates shorter mode-locked pulses is that this grating showed a smaller variation in cavity round-trip relative group delay (RGD) across most of the wavelength sweep range (Fig. 2(d)). Note that the RGD variation was also measured to be very small when the CFBG was excluded (black dots in Fig. 2(d)), indicating that the RGD variation is dominated by the CFBG. It would seem reasonable that a smaller cavity RGD variation would lead to shorter mode-locked pulses.

Fig. 2.

Point spread functions measured varying a depth location of a partial reflector for (a) 5-m ICS/5-m PCS SPML laser and (b) 10-m ICS SPML laser. (c) Mode-locked pulse temporal profiles measured after the EOM. Blue line, 10-m ICS SPML laser; Red line, 5-m ICS/5-m PCS SPML laser. (d) RGDs over a single cavity round-trip. Blue dots, 10-m ICS SPML laser; Red dots, 5-m ICS/5-m PCS SPML laser; Black dots, 13-m-long PMF in the laser cavity.

Figure 3(a) shows the measured RIN of the two SPML designs. The RIN was measured using the 35 GHz photoreceiver and a 4 GS digitizer (Alazartech ATS9373, Bandwidth: 1.9 GHz). The mean RIN (measured after the BOA) of the 10-m ICS SPML laser across a frequency range of up to 1.9 GHz was measured to be −135.3 dB/Hz, whereas the mean RIN of the 5-m ICS/5-m PCS SPML laser was −129.3 dB/Hz, which is 6.0 dB/Hz higher than that of the 10-m ICS SPML laser. Both configurations had the same peak power level in the SPML laser trace, with a measurement of 126 mW following the end-stage booster amplifier (32 mW directly from the laser cavity), which ensured a fair comparison of the RIN performance. Interestingly, the RIN measured immediately following the 5-m ICS SPML cavity (without PCS) was −128.4 dB/Hz, which was 3.6 dB/Hz lower than that measured immediately after the 10-m ICS SPML cavity (−124.8 dB/Hz). When the output of the 5-m ICS SPML laser was amplified by the BOA without any PCS (the input power to the BOA was attenuated to generate the same amplified peak power level of 126 mW), the RIN was measured to be −137.5 dB/Hz (the first (leftmost) point of the red plot in Fig. 3(b)). To further investigate the impact of PCS on the RIN performance, we conducted the RIN measurements (after BOA) using various combinations of ICS and PCS CFBG lengths, as shown in Fig. 3(b). Using two ICS SPML lasers with 5-m and 10-m long CFBGs, respectively, we investigated the RIN performance when CFBGs of different lengths, specifically 1 meter, 2.5 meters, 5 meters, and 10 meters were employed for PCS. Again, we adjusted the peak power level of the SPML trace to be 126 mW after the BOA in every case, using additional attenuation before the BOA if necessary. Figure 3(b) shows that for both SPML lasers, higher RIN was measured when using the longer PCS CFBG. Comparing different ICS/PCS combinations but with the identical amount of the total pulse stretching, which corresponds to identical wavelength sweep rate, as denoted by * and ** in Fig. 3(b), the SPML laser with a longer PCS-CFBG consistently exhibited higher RIN than the SPML laser with a shorter PCS-CFBG. For example, the 5-m ICS/5-m PCS SPML laser had 6.0 dB/Hz higher RIN than the 10-m ICS SPML laser, both providing 1075 nm/μs sweep rate (* in Fig. 3(b)). In another example that compares the two SPML laser configurations generating 717 nm/μs sweep rate (** in Fig. 3(b)), the 5-m ICS/10-m PCS SPML laser showed 3.5 dB/Hz higher RIN than the 10-m ICS/5-PCS SPML laser. We attribute the noise generated by the PCS to laser mode partition noise. In the laser, the optical power is distributed across many longitudinal cavity modes. Gain saturation sets the amplitude and phase of each cavity mode such that the power (summed over all cavity modes) is stabilized. The PCS CFBG modifies the relative amplitude and phase of each cavity mode, which causes a variation in the total power [10], [11].

Fig. 3.

(a) Measured RINs. (b) Mean RIN across a frequency range of up to 1.9 GHz. Blue, 10-m ICS and additional PCS; Red, 5-m ICS and additional PCS. *, wavelength sweep rate of 1075 nm/μs; **, wavelength sweep rate of 717 nm/μs.

The sensitivities of the 10-m ICS SPML laser and the 5-m ICS/5-m PCS SPML laser were measured using a balanced receiver (Thorlabs PDB480C-AC, Bandwidth: 1.6 GHz) to be 98.0 dB and 97.4 dB, respectively, with a power of 53 mW on the sample, which indicate that the OCT systems using these SPML lasers effectively removed most of the RIN, and suggesting that either configuration achieves an acceptable level of RIN when used with balanced detection. Figure 4(a) shows the sensitivities of both SPML lasers as a function of depth measured using the 35 GHz photoreceiver and the 25 GHz sampling oscilloscope. The sensitivity of the 10-m ICS SPML laser was slightly higher than that of the 5-m ICS/5-m PCS SPML laser up to a depth of 500 μm, but for depths exceeding 500 μm, the 5-m ICS/5-m PCS SPML laser demonstrated higher sensitivity. The red-shaded area in Fig. 4(a) represents the imaging range that corresponds to the 3-dB bandwidth of 1.6 GHz of the balanced photoreceiver used for imaging. Figure 4(b) shows the cross-sectional OCT intensity images of polydimethylsiloxane (PDMS) block containing TiO₂ as scattering particles acquired using both SPML WSLs with the 1.6 GHz balanced receiver and the 4 GS/s digitizer. While the depth-wise intensity profile (Fig. 4(c)), obtained by averaging all A-lines in the cross-sectional intensity images in Fig. 4(b), showed a similar trend to the sensitivity vs. depth plot in Fig. 4(a), the difference between two lasers was not significant over the depth range corresponding to the bandwidth of the balanced receiver. Figure 4(d) show the cross-sectional OCT intensity images of a human finger acquired using both SPML lasers. The transverse resolution is 18.7 μm and the axial resolution is 9.6 μm in air. Consistent with the previous results, images obtained using both lasers show similar qualities.

Fig. 4.

(a) Measured OCT system sensitivity as a function of depth. Dotted vertical lines, 6-dB roll-off depths. (b) Cross-sectional OCT intensity images of PDMS block with TiO2 scattering particles. Scale bars, 200 m. (c) Depth-wise OCT intensity profiles obtained by averaging all A-lines in the cross-sectional images in (b). (d) Cross-sectional OCT intensity images of a human finger (ten frames averaged). Scale bars, 500 m. Blue, 10-m ICS SPML laser; Red, 5-m ICS/5-m PCS SPML laser.

III. Conclusion

In conclusion, we demonstrate a CFBG-SPML laser using PCS. Through a variety of ICS and PCS combinations, the SPML laser with additional PCS offers versatility in OCT imaging speed and eliminates the need for optical buffering. For two SPML lasers with the identical wavelength sweep rate of 1075 nm/μs, the laser employing both ICS and PCS in equal amounts showed a coherence length that was more than twice as long as the laser using ICS only. The PCS increased the RIN of the SPML laser output, likely due to mode partition noise. The OCT system with the laser using both ICS and PCS in equal amounts exhibits a slight sensitivity drop of 0.6 dB near zero depth. However, the SPML laser using both intra- and post-cavity stretching showed higher sensitivity at depths exceeding 500 μm due to the improved coherence length. The CFBG-SPML laser with PCS offers a compelling option for OCT systems, particularly when long imaging range is required. We note that it is possible that some of the observed behavior can be attributed to individual properties of the utilized CFBGs, and more studies should be done to confirm that these behaviors are seen across a broader number of fabricate gratings.