Monolithic Vernier tuned super-structure grating distributed Bragg reflector (SSG-DBR) lasers are expected to become one of the most promising sources for swept source optical coherence tomography (SS-OCT) with a long coherence length, reduced sensitivity roll-off, and potential capability for a very fast A-scan rate. However, previous implementations of the lasers suffer from four main problems: 1) frequencies deviate from the targeted values when scanned, 2) large amounts of noise appear associated with abrupt changes in injection currents, 3) optically aliased noise appears due to a long coherence length, and 4) the narrow wavelength coverage of a single chip limits resolution. We have developed a method of dynamical frequency tuning, a method of selective data sampling to eliminate current switching noise, an interferometer to reduce aliased noise, and an excess-noise-free connection of two serially scanned lasers to enhance resolution to solve these problems. An optical frequency comb SS-OCT system was achieved with a sensitivity of 124 dB and a dynamic range of 55-72 dB that depended on the depth at an A-scan rate of 3.1 kHz with a resolution of 15 μm by discretely scanning two SSG-DBR lasers, i.e., L-band (1.560-1.599 μm) and UL-band (1.598-1.640 μm). A few OCT images with excellent image penetration depth were obtained.
© 2013 Optical Society of America
Optical coherence tomography (OCT) is a non-contact, non-invasive, high-resolution imaging technique  that is widely used for various biomedical applications . Technological advances in OCT have continued toward higher-speed, higher-resolution, deeper imaging depth, and real-time 3D imaging at video rates to further extend its applications. OCT can be carried out with three different methods: time domain (TD) OCT , spectral domain (SD) OCT, and swept source (SS) OCT. The latter two methods utilize Fourier domain (FD) detection schemes, where an interferometric signal is acquired as a function of frequency and then it is Fourier transformed to generate an axial scan (A-scan). Since the entire signal is measured simultaneously in FD-OCT, significant advantages in detection sensitivity can be achieved over TD-OCT [3–6], and the improvements to sensitivity allow a dramatic increase in imaging speed with FD-OCT methods. With two FD-OCT methods, a deeper imaging depth can be achieved with SS-OCT than with SD-OCT, because the coherence length in SS-OCT can be much longer than in SD-OCT, enabling extended depth range imaging with significantly reduced sensitivity roll-off. Therefore, SS-OCT is becoming the major OCT method for various applications.
The specifications of an SS-OCT system are mainly determined by the swept source it adopts and many developments have evolved to improve swept sources, such as external cavity tunable lasers (ECTLs) employing rotating polygons and diffraction gratings [7,8], scanning filters [9,10], resonant mirrors combined with a diffraction grating , a polygon-based ECTL with a short-length unidirectional ring resonator  and Fourier domain mode locked lasers (FDMLs) with faster scan rates and longer coherence lengths [13,14]. An ECTL has been miniaturized using micro-electric-mechanical systems (MEMS) technology and a reflective Fabry-Perot filter has been adopted as a fast scanning filter . A tuning range of 140 nm, a fast scan rate of 100 kHz, and a coherence length longer than 13 mm were achieved. An axial scan rate up to 400 kHz was demonstrated  with this laser and buffering technique. However, the longitudinal multi-mode operation of the laser generated a cavity length mode, which limited the imaging depth . The reduction of laser cavity length to achieve single longitudinal mode operation has recently been demonstrated using MEMS-tunable vertical-cavity surface emitting lasers (VCSELs) in which the cavity length mode is absent and an extremely long imaging depth range of ~50 mm has been achieved. OCT imaging using the lasers was reported at 1300 nm  and 1060 nm , where variable A-scan rates of 50 kHz to 1 MHz and adjustable operating parameters were also demonstrated. The introduction of VCSELs has significantly improved SS-OCT.
Rapid scanning with other types of single longitudinal mode lasers has been reported [20–23]. O’Connor et al. demonstrated the possibility of an A-scan rate up to 1 MHz . They used sampled grating distributed Bragg reflector (SG-DBR) lasers. They are a kind of Vernier tuned (VT) DBR lasers and at least five kinds of VT-DBR lasers have been reported thus far: an SG-DBR laser , super-structured grating distributed Bragg reflector (SSG-DBR) lasers [25–27], a vertical grating assisted codirectional coupler laser with a rear sampled grating reflector , a digital supermode distributed Bragg reflector laser , and a modulated grating Y-branch (MG-Y) laser . Of these five, the SG-DBR laser (JDSUniphase, Milpitas, USA) and MG-Y laser (Finisar, Sunnyvale, USA) are commercially available. A VT-DBR laser system is commercially available from a company (Insight Photonic Solutions, Lafayette, USA). They can potentially be used by OCT researchers. These VT-DBR lasers have promising characteristics of a long coherence length , fast switching speeds of a few nanoseconds [32,33], a wide spectral coverage of 140 nm with multiple chips , small chip sizes of less than about a few millimeters , the possibility of low cost mass production because of monolithic structures, and scanning parameters that can be tailored to suit particular applications. Despite such potential expectations, insufficient functionality by VT-DBR lasers for OCT applications has been reported, i.e., poor quality of the point spread function (PSF) [21,23] and large stitching noise at wavelengths where large changes in injection currents abruptly occur .
A VT-DBR laser emits single frequency light of very narrow width. If we scan the frequency of the VT-DBR laser stepwise in equal frequency interval, we can use it as a source for optical comb SS-OCT . We have had a few reports published on OCT imaging with SSG-DBR lasers [31,37–42] for sources of such optical comb SS-OCT systems, in which the dynamic range of imaging was relatively poor due to poor PSF dynamic range and insufficiently reduced optically aliased noise. Optically aliased noise is due to reflected or stray light, which has an optical path length difference (OPD) longer than the principal OCT depth range. The frequency of interference signal of such light is higher than that observed in the principal depth range in the case of conventional continuously swept SS-OCT. A low-pass filter at the output of a photodetector can reduce aliased noise considerably in the case. On the other hand, the frequency is constant and interference signal is DC for optical comb SS-OCT during a frequency step. The low-pass filter does not effectively work in the case. Care must be taken to reduce optically aliased noise in frequency comb FD-OCT [36,43]. The reason to develop optical comb SS-OCT, nevertheless, is that it is superior to conventional continuously swept SS-OCT in sensitivity roll-off as explained in Subsection 3.1. Here, we report on OCT imaging with significantly improved dynamic range. Improvements to the PSF dynamic range have been made since our preliminary report on partial results in the present work . We found that deterioration in the PSF dynamic range was mainly caused by deviations in frequencies from the target value. We propose and describe a new method of dynamic frequency tuning in Subsections 2.1 and 2.2 to improve the dynamic range of PSF. Selective sampling of redundant oversampled data to reduce switching noise, including stitching noise, is explained in Subsection 2.3. The method of selective data sampling, also applied to continuation of the interference signal obtained with two lasers to enhance resolution, is explained in Subsection 2.3. The OCT system and reduced optically aliased noise that is inherent in practice in optical frequency comb SS-OCT is explained in Subsection 2.4. The results for PSF and its dependence on the axial depth are discussed in Subsection 3.1 and compared with numerical calculations taking into account the deviations in frequencies from the target values. A few good OCT images are presented in Subsection 3.2.
2. Experimental system and procedure
2.1 Initial setting of injection currents
There is a schematic of the SSG-DBR laser (prototype, NTT Electronics (NEL), Yokohama, Japan) used in this experiment in Fig. 1 . The principles of operation of the laser are described in the developers’ original work [25–27]. The laser cavity is a Fabry-Perot interferometer (FPI); one mirror is the front SSG segment (F-SSG) and the other mirror is the rear SSG segment (R-SSG). Both mirrors reflect light at frequencies determined by the injection currents: If to the F-SSG and Ir to the R-SSG. When the reflecting frequencies of both mirrors coincide, the laser lases at that frequency. The free spectral range (FSR) of the FPI is about 50 GHz. Injection current Ip to the phase segment (P) changes the optical path length between the two mirrors by changing the refractive index. This enables the lasing frequency to be continuously fine tuned. The power within the laser is mainly determined by current Ia injected to the active segment (Act). Injection current Is to the semiconductor optical amplifier segment (SOA) enhances the output power and enables it to be finely adjusted. The absorber (Ab) reduces unwanted reflections from the rear segment. The injection current sources supply currents to respective segments following look-up tables. Both static (CW) and scanning operations are made possible with the scan control circuit. The scan control circuit also enables each numerical datum to be refreshed in the look-up tables with an external computer via the RS232C interface. A laser in our system can be refreshed even when it is scanning. This is an important capability used in this work. Two signals to synchronize acquisition of OCT data are output: the scan-start trigger (Trig) and scanning clock (Scan clock). We only used the Trig signal and did not use the Scan clock signal. Data were acquired synchronously with the internal clock of the data acquisition system.
The frequency of reflection by a DBR can be tuned with the injection current as previously mentioned, but it also varies sensitively with temperature. This temperature sensitivity of DBR introduces complexity into tuning an SSG-DBR laser. The temperature of the whole laser chip is controlled to a constant value to minimize the effects of temperature with a thermoelectric controller (TEC). However, small temperature changes occur locally at segments within a chip associated with changes in the injection current to the segments. This small change in temperature alters the characteristic of the segments. This is a serious problem in applying VT-DBR lasers to SS-OCT systems, where injection currents are varied to scan the output frequency and Joule heating always changes from one output frequency to another. Fujiwara et al. introduced a thermal drift compensator (TDC) section composed of the four heater segments in Fig. 1  to minimize such local changes in temperature. Four temperature-compensation currents Irt, Ipt, Ift, and Ist are respectively supplied to the TDC segments of the R-SSG, P, F-SSG, and SOA. The purpose of these TDC currents is to equalize thermal inputs to individual segments while the laser is scanned. Fujiwara et al. confirmed that stitching noise became negligible with the TDC current strategy, and deviations in output optical frequency from respective target values were suppressed within a standard deviation of 5 GHz. The scheme worked up to a scanning speed of 500 ns per frequency step. It was fine to remove stitching noise, but the frequency deviations were too large to obtain a good PSF dynamic range in an SS-OCT system, as will be discussed in Subsection 3.1. We needed to develop a better method of tuning for OCT.
We must determine the initial look-up tables for injection currents to start tuning a VT-DBR laser. Determining the look-up tables in previous conventional work both on telecommunications and OCT started by measuring optical spectra for many values of If and Ir, and then plotting a frequency contour map on the If - Ir plane; e.g., see Fig. 6 in Derickson et al.  or Fig. 3 in Sarlet et al. . The contour map can be sectioned to more than ten regions, where each region corresponds to individual single mode lasing. The best values of If, Ir, and Ip for a target frequency were determined from the side-mode-suppression-ratio (SMSR) and/or the symmetry of side peaks; e.g., see Fig. 3 in O’Connor et al. . Such procedures to determine the look-up tables were too time consuming. We found an alternative procedure, which was simpler and suited to precisely set the frequency for SS-OCT.
We set Ift, Irt, Ipt, and Ist to a constant value of 10 mA each to determine the initial look-up tables for an SSG-DBR laser. These TDC currents were kept constant and the previously mentioned procedure to compensate for temperature was not used in our method. The particular finite value of 10 mA was not significant and other values including zero could be chosen. We thought that the slight positive offset of Joule heating might stabilize temperature at small values for control injection currents. We used two SSG-DBR lasers: a long (L) band and an ultra-long (UL) band laser. The Is was set to 65 mA for the L-band laser and 80 mA for the UL-band laser. The Ia was set to 100 mA for the L-band laser and 130 mA for the UL-band laser. Larger injection currents Is and Ia were required for the UL-band laser to obtain the required output power. The initial value of Ip was set to 6 mA for both lasers. The If and Ir were varied in 0.5 mA intervals from 0.5 to 50 mA. Optical spectra were measured with the experimental set up in Fig. 2(a) for all 10000 pairs of If and Ir, where the output of the laser was attenuated using a 90:10 coupler C and directed to an optical spectrum analyzer (AQ6370, Yokogawa, Yokosuka, Japan). The laser was set to CW output operation at each If and Ir pair by the controller via the RS232C interface using an embedded controller (PXI-8105, National Instruments (NI), Austin, USA) inserted into the chassis (PXI-1031, NI), and the optical spectra were measured with the optical spectrum analyzer, which was then transferred to the controller via a General Purpose Interface Bus (GPIB) interface. The procedure to acquire the spectra was automatically repeated for 10000 pairs of If and Ir. Examples of the spectra are given for the L-band laser in Fig. 2(b) for Ir = 4 mA and in 2(c) for Ir = 3 mA. The If was set to 26.5 mA for both cases. The SMSR was defined as the ratio of the strongest peak height to the second strongest peak height. Significant changes in SMSR were observed with the small change in Ir for the same If value.
We simply plotted If (blue) and Ir (red) values as functions of the frequency of the strongest peak in Fig. 3(a) for the L-band laser instead of the contour mapping of observed peak frequencies on an (If, Ir) plane like that in previous work. We found that the plots had well defined curves with narrow widths, which evidently allowed numerical estimates of best fit curves. Evident exotic data points that deviate from the main curves can be noted in the plot of all 10000 data in Fig. 3(a). A spectrum with a large SMSR value like that in Fig. 2(b) indicates a better lasing state, while a spectrum with a small SMSR value like that in Fig. 2(c) corresponds to a worse lasing state. The plot in Fig. 3(b) was obtained, where 7673 data were left by selecting data points having SMSR values larger than 35 dB. A considerable number of exotic data points were removed. Only five exotic points indicated by the arrows were observed, which we did not include in estimating the best fitting curves because we did not use such lasing states. Setting the SMSR threshold at 35 dB was arbitrary. Threshold values such as 40 dB or 30 dB may work as well. Although a higher threshold can choose data at better lasing states, it loses a number of data points to estimate the best fits. Therefore, there is a trade-off between them. We used the data in Fig. 3(b) to determine the best fitted curves to create the approximate starting look-up tables for the L-band laser. Least square fitted curves to third order polynomials are plotted in Fig. 3(b) with the thin solid curves for If (blue) and Ir (red). The fitted curves are not distinct in some parts because the data points are dense. A similar data set was also obtained for the UL-band SSG-DBR laser, which is not shown here. Lasing frequency regions were confirmed by processing from 187.2815 to 192.2041 THz for the L-band laser and from 182.8383 to 187.6648 THz for the UL-band laser.
The seven curves for If and Ir in Fig. 3 result from the existence of seven modulated parts in the front and rear SSG segments [25–27]. The SSG-DBRs were designed so that If would cover a wider frequency range than Ir. The slope of Ir with frequency is steeper than that of If if we compare nearby If and Ir. The values of the two currents define frequency with different slopes so that a pair of If and Ir values determines a unique single frequency like that obtained with a Vernier scale on calipers.
Figure 3 shows how we created look-up tables of If (blue) and Ir (red) for the L-band (Fig. 3(c)) and UL-band (Fig. 3(d)) SSG-DBR lasers. The thin solid curves plot the least-square-fitted curves. A current pair for a required output frequency can be obtained by drawing a vertical line at the frequency and choosing intercepts with the least-square-fitted curves. The choice is not fixed. We established two criteria for the choice of If and Ir after gaining experience from many trials. The first criterion was that these injection currents should not be smaller than 3 mA because lasing often became unstable and fluctuations in output frequency (Δf) increased. We regarded this to be due to fluctuations (ΔΙ) in the injection currents caused by electrical noise in the scan control circuit including the injection current sources in Fig. 1. The slope (df/dI) is steeper in Fig. 3 and the frequency fluctuations Δf = (df/dI)ΔΙ are larger for smaller values of I. The second criterion was that excessively large values for If and Ir are not preferred because the output intensity decreases with increases in these injection currents because of increased absorption in DBRs. Following these criteria, the look-up tables for If and Ir were determined by choosing current values from the thick solid curves in Figs. 3(c) and 3(d): If (blue) and Ir (red). The frequency is usually scanned from lower to higher values. We chose the starting frequency, f1 = f0 + δf, near the lowest possible value for the UL-band laser, where δf was the frequency interval. The i-th injection currents, Ifi and Iri, were determined at the intercept of the vertical line, fi = f0 + iδf, with the thick solid curves. They were numerically determined. The frequency coverage of the L-band laser overlapped that of the UL-band laser, as seen in Figs. 3(c) and 3(d). The frequencies for the L-band laser were chosen at the same values as the corresponding frequencies for the UL-band laser in the overlapping region. The two lasers were scanned serially and frequency values were redundant in the overlapping region. This redundancy was eliminated after data were acquired by selective sampling as is explained below.
Either Ifi or Iri, or both, must be changed from larger to smaller values abruptly at nine values of frequencies both for the L-band and UL-band lasers in the look-up tables determined above. These frequencies are called stitching frequencies (points) and have been indicated with the label, s, in Figs. 3(c) and 3(d). The response time for current switching was about 80 ns in our electrical circuit in Fig. 1. Due to this finite response time, the injection current required durations longer than the response time to reach the target values within the required accuracy at stitching frequencies. We repeated the same value for injection currents four times at stitching frequencies to acquire data after the laser reached the required frequency. Therefore, the look-up tables included redundant insertions of the same repeated currents. This redundancy was eliminated after data were acquired by selectively subtracting redundant data.
We scanned the lasers stepwise. The fastest scanning speed of the scan control circuit was 100 ns per step. When we used this stepping time, the interference signal indicated transient variations during most of this time interval due to the 80 ns response time of the electronic circuit. We chose a step time of 200 ns to obtain data after transient relaxation and carried out multiple sampling during the time interval. Selective sampling was done for the multiple data acquisition, as explained below.
The frequency interval was chosen to be δf = 6.25 GHz. The look-up tables for the UL-band laser covered a frequency range from 182.85 to 187.61875 THz with 800 outputs including 4x9 = 36 redundant outputs at stitching frequencies. The look-up tables for the L-band laser covered a frequency range from 187.43125 to 192.2 THz with 800 outputs including 36 redundant outputs at stitching frequencies. Other injection currents were made constant as the starting look-up tables, as was previously mentioned. Both lasers were continuously scanned over 1600 outputs to eliminate transient thermal relaxation related to turning the lasers on and off. The Is and Ia were set to zero while the lasers were not in use: from 801 to 1600 outputs for the UL-band laser and from 1 to 800 outputs for the L-band laser. Other injection currents were set to individual starting values. As a result, the UL-band and L-band lasers were scanned simultaneously and continuously. However, they alternately output light power. Data were acquired at 1600 acquisition points to obtain an interference signal.
2.2 Fine tuning of SSG-DBR lasers
Two lasers were used in this work to enhance resolution by expanding the wavelength range more than that from a single laser. The optical system to combine the two lasers is outlined in Fig. 4. Single mode fiber was used for the optical system. The two lasers were optically combined with a 50:50 coupler (CP) (Wanshing, New Taipei, Taiwan). One output monitor from the CP was used to monitor the combined light with the frequency tuning system outlined in Fig. 5. The other output was amplified with a semiconductor optical amplifier SOA (Covega, Jessup, USA) to enhance the intensity used for OCT measurements. The output of both lasers was linearly polarized at the output of the laser chip and the state of polarization changed while light traveled through the single mode fiber. A polarizer (POL) (Optoquest, Ageo, Japan) and two polarization controllers (PC-UL and PC-L) (Fiberpro, Daejeon, Korea) were used to align the polarization of light from the two lasers. Both polarization controllers were adjusted so that a maximum intensity was obtained for both lasers at the output of polarizer POL, which is indicated as an OCT source in Fig. 4. The POL also worked to filter ASE light from the SOA, which had a different state of polarization from the POL.
The observed average power was 4.5 mW at the monitor and 52 mW at the OCT source. The intensities as functions of the acquisition point number are plotted in (a) and (b) in Fig. 4. We adjusted the SOA injection currents of the lasers so that the power at the monitor output would be constant as plot (b). Due to the amplification characteristics of the SOA in Fig. 4, the power of OCT source output had variations depending on frequency as indicated in plot (a). The variations were corrected in OCT measurements as is explained below. The reason we amplified the intensity was because we used a coupler instead of a circulator to direct light to the sample. The purpose was to reduce optically aliased noise, as is explained below.
The frequency tuning system is outlined in Fig. 5. Light from the monitor output of the combined SSG-DBR laser system in Fig. 4 was attenuated with an attenuator (Att) (Dicon Fiberoptics, Richmond, USA) and divided by a 1x4 coupler (NEL). We used four 80-MHz balanced photoreceivers BPR1~4 (1817 type, New Focus, San Jose, USA) to detect the intensity with BPR1 and three outputs of Mach-Zehnder interferometers MZI1~3 with BPR2~4. The MZI1 was home made by connecting two 50:50 couplers (Wanshing, New Taipei, Taiwan), and a frequency spacing of 351 GHz was obtained. It was used to monitor the sinusoidal waveforms of the interference signal. The frequency spacing of MZI2 (planar light-wave circuit (PLC) type custom product, NEL) was 25 GHz and was used for monitoring step-like variations in the interferometer signal that changed in four steps while the SSG-DBR laser was scanned over 25 GHz frequency intervals with a frequency step of 6.25 GHz. The frequency spacing of MZI3 (PLC type custom product, NEL) was 12.5 GHz and used to precisely detect the zero crossing of the targeted frequency spacing of 6.25 GHz. The PLC type MZIs were temperature tunable and NEL’s MZIs could be temperature controlled so that the interference signal peaks matched the International Telecommunication Union (ITU) frequency grid. The outputs from balanced photoreceivers were A/D converted with a multi-channel 60-MS/s DAQ (PXI-5105, NI) and processed with an embedded controller (PXI-8105, NI) inserted into a chassis (PXI-1031, NI). Acquired data were displayed on an LCD monitor.
There is an example of a set of three signals from the three MZIs on the right of Fig. 5 from 1154th to 1169th acquisition point numbers, when adjustment was almost completed. An average of 25 scans was done to decrease noise and these are shown here. The data acquisition rate was 60 MHz for all the signals and 12 samplings were made at each output frequency step of 200 ns duration. This is the reason that signals were observed stepwise, demonstrating a frequency comb sweep. Precise tuning to the 6.25 GHz frequency interval step was achieved by confirming zero crossing of the interference signal output from the MZI3 (12.5 GHz) as indicated by plot (c) in Fig. 5. The spike noise observed at every boundary was due to the fact that the interference signal passed a sinusoidal maximum or minimum as indicated by the red dotted curve. As is explained below, because only Ip was adjusted to fine tune the output frequencies, fine tuning was accomplished with a single parameter. The completion of adjustment was confirmed by the zero of the signal from the MZI3. However, two intentionally misadjusted examples confirmed adjustment was not completed only from the signal. Although the zero of MZI3 output was confirmed for the 1158th acquisition point, the frequency was the same as that for the 1157th acquisition point as detected by the output of the MZI1 (plot (a)). This mistake can easily be detected from anomalies in the signal from the MZI2 (plot (b)). Neither of the signals from the MZI2 nor MZI3 revealed anomalies for the 1168th acquisition point, although the frequency was the same as that for the 1164th acquisition point. Such mistakes can easily be detected from the signal from the MZI1 as indicated by plot (a). We found the use of the three MZI signals was sufficient to accurately adjust the output frequencies. The noise at the start of 1163rd acquisition point was an example of switching (stitching) noise due to the abrupt change in Ip from a large to a small value.
After we had set the starting look-up tables as previously mentioned, we scanned both lasers simultaneously over 1600 steps with a time interval of 200 ns per step. Thermal relaxation occurred following the change in injection currents, as was discussed above. A typical thermal relaxation time is on the order of milliseconds . The time duration of 200 ns per frequency step was much shorter than this and allowed us to regard lasing to have occurred under an “adiabatic” condition before thermal relaxation started. When the injection currents were switched, the laser lased under the thermal condition just before switching. The absolute value of lasing frequency slightly deviated from that calibrated under static thermal equilibrium under this condition. The precise equality of frequency intervals between all adjacent outputs is important in OCT. If the laser is repeatedly scanned continuously, it soon reaches a certain thermal equilibrium and the same interference signal is repeatedly observed, because the laser chip is controlled by the TEC. We adjusted the frequency of the laser while it was scanned using the tuning system shown in Fig. 5 in this stationary scanned state. While we monitored the interference signals from the MZIs, we used phase injection current Ip as only one parameter to finely adjust output frequencies, as is explained below.
Before fine frequency tuning with Ip, we needed to adjust the output power of the lasers to a constant. The output power detected by the BPR1 in Fig. 5 had saw-tooth like variations as seen in Fig. 6(c) for the UL-band and 6(d) for the L-band with the initial look-up tables. With an increase in If or Ir, the output power of the lasers decreased due to increased absorption by injected carriers. The intensity was adjusted to be constant by changing SOA current Is as indicated by the plots in Fig. 6(a) for the UL-band laser and 6(b) for the L-band laser. The adjustment was made automatically with a computer program. The values of Is required to make the adjustment are given in Fig. 6(e) for the UL-band laser and 6(f) for the L-band laser.
Although the output of BPR1 was constant as seen in Fig. 6, the actual optical power was stronger at longer wavelengths (smaller output numbers). The responsivity of the photoreceiver decreased as the wavelength increased. Our comparison of the power reading with a calibrated power meter (AQ2160-02, Yokogawa) and that with the photoreceiver (BPR1) indicated about a 15% nearly linear decrease in responsivity as the acquisition point number decreased from 1600 to 1. The catalogue for the photoreceiver indicates a steep decrease in responsivity as the wavelength exceeds 1600 nm. However, the particular detectors we used did not demonstrate such behavior. The intensities in Figs. 4(a) and 4(b) were found from photoreceiver outputs and the scale units on the vertical axis are approximately in milliwatts.
After the output power was adjusted to a constant value, the output frequency at each output number was finely adjusted by observing the three interference signals with the system outlined in Fig. 5. The phase injection current, Ip, was set to 6 mA for all the outputs in the starting look-up tables. The interference signal from the MZI1 scanned with these look-up tables was strongly deformed from the sinusoidal function and exhibited many discontinuities, as shown in Fig. 7(a), for the L-band laser. The deformed interference signal can be explained by the dependence of output frequency on phase injection current Ip. Such dependencies were measured for selected output numbers under CW operation using the system presented in Fig. 2(a). The results are given in Fig. 7(b). The results for eleven even acquisition point numbers are presented from the 1160th to 1180th acquisition points that prevented the curves from overlapping. The laser cavity of the SSG-DBR is an FPI as previously explained. The SSG-DBR lases in a single longitudinal mode at the frequency tuned by the FPI. The effective spacing of the FPI can be varied by the injection current into the phase section shown in Fig. 1. Therefore, we can sweep the output frequency by Ip, but the range is restricted within the FSR of the FPI. As the plot for the 1160th acquisition point (black curve) shows, the output frequency repeats sweeps within the limited range indicated by the two thin dotted horizontal lines as Ip increases. This limited frequency range is the FSR indicated in Fig. 7(b), which was about 47 GHz for the plot. The lasing frequency shifts to higher values as the acquisition point number increases with increases in the If and Ir injection currents. The straight vertical green line indicates Ip = 6 mA. Lasing occurs at intercepts on this line with characteristic curves. The continuous lasing frequency ranges are indicated by the thick green vertical lines. There are gaps in output frequency where there are no outputs under the condition of Ip = 6 mA. This explains the presence of many jumps in the interference curve shown in Fig. 7(a).
Fine tuning can be carried out by adjusting Ip. Part of the interference curve within the red rectangle in Fig. 7(a) is depicted in Fig. 7(c), which has been expanded on the horizontal scale. The black curve is the interference curve observed for the constant Ip value of 6 mA. The red curve is the target sinusoidal function to be achieved by adjusting Ip. The red curve is there for explanatory purposes and is not exactly the same as that actually attained after adjustment. The target frequencies are indicated by the horizontal red dashed lines in Fig. 7(b) with a spacing of 12.5 GHz for even numbers of acquisition point (6.25 GHz x 2). The intercepts on these lines with the characteristic curves, which are indicated by the closed circles, are Ip values to achieve the target frequencies at corresponding acquisition point numbers. For examples, target frequencies are obtained for the 1160th (black curve) and 1162nd (red curve) acquisition points by increasing Ip, while a target frequency is obtained for the 1164th (orange curve) acquisition point by decreasing Ip. The Ip values at the closed circles correspond to required values to accomplish target frequencies. However, Ip was not actually tuned using the plots in Fig. 7(b). The characteristic curves in the figure were obtained through “static” measurements. When the lasers are scanned, the thermal conditions at each output number differ from those under static CW lasing. Figure 7(b) illustrates the principles underlying tuning with Ip for explanatory purposes.
Actual tuning was conducted using the system outlined in Fig. 5 while the lasers were continuously scanned. Although the phase current could be varied from zero to the maximum rating value of 30 mA, we set criteria for the lower and upper values of Ip settings. Electrical noise was inevitable in the scan control circuit in Fig. 1 and caused fluctuations ΔIp in the phase current. These fluctuations caused frequency fluctuations Δf proportional to (Δf/ΔΙp) ΔIp. The slope (Δf/ΔΙp) increased for the smaller Ip values plotted in Fig. 7(b), leading to larger frequency fluctuations. Therefore, we chose Ip values that were larger than about 3.5 mA. As seen in Fig. 7(b), Ip must be abruptly changed from a large to a smaller value; e.g., see from the 1162nd (red curve) to the 1164th acquisition points (orange curve). The abrupt change generated switching noise (stitching noise) due to the change in Ip. The noise increased with an increase in the current jump. Therefore, we set the Ip value so that it was smaller than ~12 mA. Considering these lower and upper bound values, fine tuning with Ip could only be done by observing the display in the frequency tuning system shown in Fig. 5. As a result of tuning, increasing (I) and decreasing (D) of Ip from the initial 6 mA were repeated as shown in Fig. 7(c) and the discontinuous interference signal (black) was corrected to be a continuous sinusoidal function (red). As a result of tuning, the dependence of Ip on the acquisition point number was obtained as plotted in Fig. 7(d). Phase stitching points are labeled ps. By adjusting Ip, interference signal output from MZI1 was obtained as shown in Fig. 7(e) for the L-band laser. Acquisition points, where stitching of If and/or Ir occurred, are labeled s. Similar adjustments were also done for the UL-band laser. After fine tuning had been adjusted, intensity to the constant value was re-adjusted because the change in Ip caused small variations in output intensity.
Because fine tuning with Ip was done while the lasers were scanned, after all outputs were adjusted, accurate scanning with equal frequency intervals was achieved. The results from adjustments were saved as improved look-up tables in the system shown in Fig. 1. The resulting look-up tables could be read by a computer via the RS232C interface and saved to outer memory.
After fine tuning had been completed, spectra at selected outputs under CW conditions were measured using the system in Fig. 2(a), and these are plotted as functions of wavelength in Figs. 8(a) for the L-band laser and in 8(b) for the UL-band laser. The SMSRs are better than about 45 dB at all outputs. The lasers in our system were scanned stepwise at a time interval of 200 ns, which is much longer than the mode stabilization time of a few nanoseconds for a VT-DBR laser . Therefore, these SMSR values observed under CW operation were expected to be obtained even when the lasers were scanned for SS-OCT measurements. The floor with undulation has broad spectral spread and very short coherence length. Therefore it contributes to DC component in PSF and is negligible.
Our method of stepwise scanning explained so far differs from that used by Derickson’s group [20–23]. They used electrical filters to smooth injection currents, which were initially generated by D/A converters stepwise. The filtering allowed analog injection currents and the frequency could be scanned continuously and very rapidly, e.g. an A-scan rate of 100 kHz . However, they reported poor quality in the PSF dynamic range [21,23].
The A-scan rate for the present system is 3.1 kHz and it is definitely slow. The speed is limited by the scan control circuit in Fig. 1. Once a higher speed control circuit is developed, the frequency step time will be able to be reduced to a few nanoseconds of the switching time for the SSG-DBR laser. We intend to develop such a system in future work.
2.3 Selective data sampling
The interference signal from the MZI1 in Fig. 5 is shown in Fig. 9(a) as a function of the acquisition point number. The stitching points in Fig. 9(a) are labeled s. Large noise is observed at each stitching point. The same injection currents were repeated five times after stitching, as was previously explained. Transient variations in the injection currents are relaxed during this time interval and the observed interference signal is stabilized. The interference signal is shown in Fig. 9(d) for acquisition point numbers from 100 to 112 near the stitching point indicated by the red label s in Fig. 9(a). Twelve samplings were carried out at a sampling rate of 60 MHz during a 200 ns time duration at each output frequency and step-wise change in the interference signal was explicitly observed. The stitching occurred at the 104th acquisition point. It stabilized after five repetitions. Four initial output data indicated by the red plot were eliminated from the five outputs. Then, large stitching noise was eliminated as shown in Fig. 9(e). This selective sampling was performed for all the stitching points and the interference signal in Fig. 9(b) was obtained with all the stitching noise being eliminated. The horizontal axis indicates the selected point number. The selected point number was reduced from 1600 to 1528 through the eliminations.
There is a gap in the interference signal at selected point number labeled c in Fig. 9(b). The gap is the transient point from the UL-band to the L-band laser. The frequency regions of the two lasers partially overlap and the green region is redundant. We could continue the UL-band data and L-band data by eliminating this region of 30 selected points, as shown in Fig. 9(c). The resulting number of sampled points was reduced to 1498. The sampled point number, where the two interference signals were connected, is labeled c in green. The interference signal near the sampled point number is shown in Fig. 9(f). The noiseless connection of the two interference signals was accomplished by selective data sampling. Only five scanning clocks with a total duration of 500 ns were required between successive scans of the combined lasers. The duty of effective data acquisition was 93%.
We did not insert redundant data or undertake selective sampling for stitching points of phase injection current Ip. The phase stitching noise was small as shown in Fig. 5(a) for the 1163rd acquisition point.
A method of connecting two SG-DBR lasers was reported by George and Derickson . They discussed concatenation by only observing intensity. They did not mention polarization and demonstrated no frequency continuity by observing an interference signal, which are both important for OCT.
The frequency sweep range of a connected laser is Δf = 9.36 THz. The resolution expected from it is 2.78/Δk = 2.78c/(2πΔf) = 14 μm with a rectangular window . Here, k is the wavenumber and c is the speed of light. If a Hanning window is used, the resolution is expected to be 22 μm. A higher resolution is possible if we connect the four lasers developed by Fujiwara et al. . The frequency span range is 17.5 THz and a resolution of 7.6 μm is expected for a rectangular window. The purpose of this work was to demonstrate an excess noiseless method of connecting two lasers. It can be extended to connecting four lasers, although the system will become more complicated.
The connection system in Fig. 4 should become simpler if multiple lasers are integrated with polarization maintaining light waveguide. SSG-DBR lasers are linearly polarized at their output and their direction of polarization can be controlled. Therefore, polarization controllers and polarizers are unnecessary in built-in integrated light waveguide systems together with lasers. We should be able to control intensity, as shown in Fig. 4(b), with a built-in SOA. Such a chip is preferred for an OCT system. VT-DBR lasers are so far available with relatively narrow wavelength coverage of less than about 40 nm with a single chip. VT-DBR lasers with wider wavelength coverage are required for OCT purposes. Efforts to develop such lasers have been made for telecommunications. Kano et al. demonstrated a broad tuning range of more than 100 nm with stable single mode operation  (with some gaps) for SSG-DBR lasers, and Ishii et al. reported achieving a 62.4 nm quasi-continuous single mode tuning range , which can be applied to an OCT source. The possibility of tuning an SG-DBR laser over a wavelength range of 40-80 nm has been claimed and an example of tuning of 72 nm has been reported . Once such wide range lasers become available for OCT, the present method can be effectively used for ultra-high resolution connecting multiple wide range lasers to extend the ranges of wavelengths.
2.4 OCT system
Figure 10 is a schematic of our SS-OCT system. The output (OCT source) of the combined SSG-DBR laser system in Fig. 4 was divided by a coupler (CP1) (Wanshing). The intensity of 10% output of CP1 was further attenuated by a coupler (CP3) (Wanshing) and was detected with a balanced photoreceiver (BPR5) (2117 type, New Focus). The signal was used to monitor the intensity of the OCT source at every sampling. The 90% output of the CP1 was divided by a coupler (CP2) (Wanshing) to a reference arm (20% side) and sample arm (80% side). A circulator (CR) (Wanshing), a collimator (CLR) (FH10-IR-APC type, Newport, Irvine, USA) and a lens (OLR) (Thorlabs Japan, Tokyo, Japan) were used to direct light to the reference mirror (RM).
Care had to be taken with the sample arm to reduce optically aliased noise. Because the instantaneous coherence length of an SSG-DBR laser is very long and an antialiasing electrical filter at the output of the photoreceiver does not work for an optical comb source, reflections and crosstalk leakage light in the sample arm must be minimized. When we used a circulator in the sample arm to direct light to the sample, the crosstalk between input port 1 and output port 3 was about −55 dB. It generated a significant white noise floor in the OCT image. The reason optically aliased noise was observed as a white spectrum is explained in Subsection 3.1. We used a coupler (CP4) (Custom product, Adamant Kogyo, Tokyo, Japan) to eliminate such crosstalk leakage light. Although direct crosstalk is negligible with the CP4, the reflected light at the end surface (angled physical contact without anti-reflection coating) connected to the CLS generated stray light of about −60 ~−65 dB (depended on product). The surface of the end of the fiber connected to a collimator (CLS) (custom product, Canon, Tokyo, Japan) was antireflection coated to reduce it. Reflected light from the CLS was reduced by using Canon’s custom product rather than Newport’s product (FH10-IR-APC type). Light from the 20% output port of the CP4 was directed to the sample using the CLS and an objective lens (OLS) of 100 mm focal length (Custom product, Canon). The OLS was slightly tilted from the beam direction to eliminate surface reflections. From a beam diameter of about 8 mm from the CLS and the focal length of OLS, the lateral resolutions were 25 and 480 μm respectively at the focal point and 6 mm (half of the axial depth range) away from the focal point. The 80% output port from the CP4 was connected to an optical terminator (Adamant Kogyo) to reduce reflections from the end surface. The reason we used the SOA in Fig. 4 to enhance source intensity was that only 20% of the power was effectively directed onto the sample by the CP4. Without a sample, the reflected light intensity from the sample arm, measured at the output of the CP4, was about −76 dB of the light intensity illuminated onto the sample. Even this weak leak light generated an optically aliased white noise floor. The distance between the CLS and galvano mirror (GM) (6220H type, Cambridge Technology, Lexington, USA) to reduce the noise floor was longer than 700 mm. The reason for this rather long separation is explained in Subsection 3.1. The GM was used to scan the beam laterally on the sample.
Polarization of light in the sample arm was adjusted with polarization controller PCS (Fiberpro) and that in the reference arm was adjusted with PCR (Fiberpro). Light from the reference arm and sample arm interfered at coupler CP5 (Wanshing) with a 50:50 split ratio. The output signal from CP5 was detected with a balanced photoreceiver (BPR6) (PDB420C type, Thorlabs Japan). The bandwidth of the BPR6 was 75 MHz. A low pass filter (F) of 22 MHz (BLP 21.4 + type, Mini-Circuits, New York, USA) was used at the output of the BPR6 to cut the high frequency component near the Nyquist frequency of the 50 MHz sampling frequency. The signal output from the BPR4 in Fig. 5 was also sampled to rescale the frequency. The signals from the photoreceivers (BPR4, BPR5, and BPR6) were A/D converted using two 14-bit 50 MHz 2-channel DAQ boards (PXI-5122 type, NI) and acquired by the controller (PXI-1031 type, NI). The GM was controlled using a D/A converter (PXI-6229 type, NI). The interference signal was divided by the monitored intensity for each sampled data to compensate for the power variations in Fig. 4(a).
3. Experimental results and discussion
3.1 Point spread function measurements
The principal depth range, Δz, of SS-OCT is determined by sampling frequency interval δf as Δz = c/(4δf), where c is the speed of light. The principal depth range was Δz = 12 mm by choosing δf = 6.25 GHz. The PSFs observed in the principal OCT imaging depth are shown in Fig. 11. The PSF at a depth of z = 8.99 mm has been shown in Fig. 11(a) as an example. In addition to the noise floor of about −55 dB, peaked noise N1-N6 is observed. Peaked noise is expected either due to some optical aliasing or due to some structured deviations in the frequencies of the scanned lasers from target values. We used the interference signal of MZI3 (12.5 GHz) to estimate frequency deviations from target values to check for the latter possibility. If a frequency deviates from the target value by Δf (GHz), the output signal is expressed as Asin(2πΔf/12.5). Amplitude A was determined by temporarily shifting the frequency of MZI3 by 3.125 GHz, which is possible because MZI3 is temperature tunable. Using this relation, we calculated Δf from the observed MZI3 signal. We estimated corrected interference signals at target frequencies by interpolation from the value of Δf and the observed interference signal. The PSF obtained from the corrected interference signal is presented in Fig. 11(b), in which peaked noise is considerably reduced compared with that in Fig. 11(a), while the noise floor is nearly the same. These results verify that the noise peaks N1-N6 were generated by some structured deviations in frequency from respective target values. A dynamic range of about 55 dB was achieved.
Observed PSFs at selected OCT imaging depths are given in Fig. 11(c). Corresponding PSFs after the correction are shown in Fig. 11(d). No appreciable sensitivity roll-off was observed within the principal depth range, as seen in Fig. 11(c). The PSFs at shallow OCT depth z indicate a small noise floor and have relatively large dynamic ranges of about −72 dB at z = 0.05 mm (red curve) and about −70 dB at z = 1 mm (black curve). However, as the depth position increases, the noise floor increases and the dynamic range decreases.
We extended PSF measurements at depths greater than the principal OCT imaging depths to determine the sensitivity roll-off and variations in the noise floor. The results are plotted in Fig. 12. Peak intensities are plotted with blue dots and noise floor intensities are plotted with red dots. A decrease in the peak intensity by 6 dB is indicated by the dotted line (a). An intensity roll-off by 6 dB is observed at a depth of 66 mm. The 6 dB roll-off occurs at a great depth as expected from the long instantaneous coherence length of SSG-DBR lasers. This value is larger than the 40 mm range reported for a VCSEL-OCT . The roll-off in peak intensity, which is usually used as a measure of coherence length in OCT measurements, is not determined by the instantaneous coherence length of the laser which is longer than 25 m . It is determined by the reduced visibility of the interference fringe due to deviation in the laser frequency from the respective target value. The reduced fringe visibility is caused by frequency fluctuations within a single frequency step (intra-step fluctuations) and fluctuations in frequency intervals between adjacent frequencies when the lasers are scanned (inter-step fluctuations).
An interference signal at the i-th frequency for a reflector placed at depth z is proportional toFigs. 9(d), 9(e), and 9(f) exhibits transient relaxation after an abrupt step change. We discarded four initial data during relaxation and only used six data for processing. This was selective sampling within a single frequency step to reduce frequency setting errors. Therefore, data during ts = 120 ns were used as one acquired datum. The data at frequency fi were proportional to the integration of Eq. (1) over ts due to the finite bandwidth of the detection system and summation of the sampled data. The only time dependent factor during time interval ts in Eq. (1) is Δf(t). If Δf(t) fluctuates or drifts during this time interval, fringe visibility is reduced; this reduction is larger for greater depths as was reported for an SD-OCT system . Reduced fringe visibility due to phase fluctuations during measurements has also been reported .
Reduction of the fringe visibility is measured by reduction of the amplitude of the interference signal, which takes maximum at 4πfiz/c = 2πm (m: integer) without deviation Δf. The maximum decreases as Δf increases. The reduction can be estimated putting 4πfiz/c = 2πm in Eq. (1), which gives cos[4πΔf(t)z/c]. The PSF is proportional to the power spectrum obtained from the fast Fourier transform (FFT) for Eq. (1). The reduction of the peak in PSF due to frequency fluctuations Δf(t) is given byFig. 5. However, if the frequency deviated from the target value by Δf(t) (GHz) at time t, the MZI3 output a signal proportional to Asin[2πΔf(t)/12.5], as previously mentioned. The frequency deviation, Δf(t), was estimated from the observed signal amplitude. The sampling rate enabled M = 7500 data points to be acquired within 120 ns. Two examples of such samplings carried out for the selected targeted output frequency of 185.2625 THz at 402nd acquisition point number from the UL-band laser are plotted in Figs. 13(a) and 13(b). These figures indicate the Δf,j to use in the summation of Eq. (2). The deviation does not exhibit fast fluctuations but indicates gradual drift-like variations around zero during this time interval. The results for integration vary depending on the data set for a given z value, as can be seen in the differences between Figs. 13(a) and 13(b). Therefore, an ensemble of such measurements was required and an average had to be obtained. The measurements were repeated 166 times, calculations were done for them, and an average was obtained for each value of z. The results are plotted on curve (b) in Fig. 12, which gives the reduced fringe visibility due to intra-step frequency fluctuations.
We regarded the frequency variations in Figs. 13(a) and 13(b) to have been caused by variations in injection currents. If injection current I varies, the output frequency of the laser varies according to the characteristic curves in Fig. 3 for If and Ir and in Fig. 7(b) for Ip, as was discussed in Subsection 2.1. Larger frequency variations were expected for smaller injection currents for the same current fluctuations ΔI from the slope of frequency versus current (df/dI). We measured standard deviations in frequency at the 352nd acquisition point (184.950 THz, If = 4.613 mA, Ir = 5.153 mA, and Ip = 6.830 mA), the 402nd acquisition point (185.2625 THz, If = 11.792 mA, Ir = 14.396 mA, and Ip = 6.556 mA), and the 452nd acquisition point (185.575 THz, If = 25.845 mA, Ir = 31.793 mA, and Ip = 6.227 mA) to respectively obtain 0.39 GHz, 0.28 GHz, and 0.17 GHz by choosing the fourth branch from the left in the characteristic curves for the UL-band laser shown in Fig. 3(d). As expected, the largest frequency variations were observed for the smallest injection current, which corresponded well with variations in the slope (df/dI). We used the data for the 402nd acquisition point, which was a middle value, to obtain curve (b) in Fig. 12. The curve should be considered an approximate estimation by considering the dependence of variations on the acquisition point number.
We estimated the standard deviation of injection current assuming that the frequency variations were due to injection current fluctuations and that its standard deviation was the same for all three injection currents of If, Ir, and Ip. The results were 4.9 μA for the 352nd acquisition point, 7.7 μA for the 402nd, and 6.0 μA for the 452nd. These values are not rigorous and an average of ~6 μA is an estimate of injection current fluctuations in the control circuit in Fig. 1. We could improve fringe visibility with the depth plotted in Fig. 12(b) by reducing current noise.
The reduction of the fringe visibility discussed above is related to a superiority of optical comb SS-OCT to conventional continuously swept SS-OCT in sensitivity roll-off. In the latter, the frequency of the source varies as during an effective sampling duration τ at time t and integration of an interference signal leads to fringe visibility reduction as follows,Eq. (2) can be made practically negligible by suppressing frequency fluctuations Δf to be sufficiently small.
We estimated the reduced fringe visibility due to inter-step fluctuations by considering Δf in Eq. (1) as the deviation in frequency from the target value for the sampled data. The interference signal introduces phase deviations by 4πΔf z/c, which cause reduced fringe visibility as has been pointed out in a previous report . The greater depth z is, the larger phase deviation is. If FFT is performed on such a disarrayed sinusoidal function, the peak of the power spectrum decreases and the noise floor increases as z increases. We observed the output of MZI3 (12.5 GHz) to estimate deviations Δf. The procedure to obtain Δf from the output was the same as that previously explained. Variations associated with frequency scanning are expected to superpose on the fluctuations in Figs. 13(a) and 13(b). An example of results is presented in Fig. 13(d) for the 10th data point of 10 samplings within a single frequency step. The measured standard deviation of the frequency is 0.32 GHz. The standard deviation is slightly larger than the mean of 0.28 GHz of the standard deviations for CW operation previously mentioned. The increase is due to additional variations associated with scanning. There is a histogram of the frequency variations in Fig. 13(c), where the red dashed curve is a Gaussian distribution with a standard deviation of 0.32 GHz. Although the variations seem random, the power spectrum obtained with the FFT of the data in Fig. 13(d) exhibits a peak labeled P in Fig. 13(e). The horizontal axis is the depth and the peak is observed at 12 mm, which is the higher boundary of the principal OCT depth range. The peak indicates that a sinusoidal structured variation proportional to cos(πi + ϕ) is included in the variations in Fig. 13(d). Here, i is the sampled point number and ϕ is an arbitrary phase. The variation modulates the interference signal and it gives rise to peaks in PSF shifted from the primary peak by 12 mm. The primary peak labeled P in Fig. 11(a) is observed at 8.99 mm. The noise peak labeled N1 is observed at a depth of 3.01 mm, which agrees with the absolute value of the difference, 8.99−12 = −3.01. This fact and the fact that noise labeled N1-N6 are considerably reduced by frequency rescaling as shown in Fig. 11(b) indicate that these are not optically aliased noise. This noise would be reduced by improving the tuning procedure in future work. For example, noise N1 may be due to the fact that we used MZI3 with a 12.5 GHz period. The tuned frequency interval was δf = 6.25 GHz. If MZI3 and its detection system had an imbalance of variations or some DC offset, tuned intervals would alternatively have deviated from the target of 6.25 GHz to larger or to smaller values, which would have given peak P in the power spectrum in Fig. 13(e). This could be improved by using an MZI of the 6.25 GHz period instead of the 12.5 GHz period.
We numerically simulated the peak and noise floor of PSFs as a function of depth by taking into account the standard deviation of the inter-step fluctuations. Function cos[4π(fi + Δi)z/c] was generated with fi = f0 + iδf for i = 1-1498 to simulate the interference signal with random phase. The fluctuations Δis were generated randomly so that the distribution was Gaussian with a standard deviation of 0.32 GHz. A set of six functions was generated using independent Δis and their average was used for the simulation. The power spectrum was obtained from the average at given depth z. The peak and noise floor were read from the power spectrum at each z value. The results are plotted as functions of depth z in Fig. 12; the blue dashed curve (c) is the peak and the red dashed curve (d) is the noise floor. The asymptote of both curves is the horizontal line at −35 dB. The signal to noise ratio of PSF for the number of sampled point Ns = 1498 is expected to be Ns/2 . The addition of six data increases the signal to noise ratio by a factor of six. Then, the asymptote is expected to be −10log(3Ns) = −37 dB, and the numerically obtained −35 dB is close to this value.
The numerically calculated noise floor (red dashed curve) is −92dB at z = 0.1 mm, which is smaller than the experimentally observed noise floor of −72 dB by 20 dB. One possible source for the limit of dynamic range is accuracy of the 14-bit A/D converter used in the DAQ and controller in Fig. 10. The least significant bit is not strictly defined and introduces an error of 1 to 214. We performed numerical simulation taking into account this error and found that this gives rise to a noise floor of −125 dB, which is much smaller than −72 dB. We consider that fluctuations in the OCT source intensity (Fig. 4(a)) set a lower limit for the noise floor. From the variation exhibited in Fig. 4(a), a large relative intensity noise (RIN) is expected for the source. However, the RIN was considerably compensated by data processing. The intensity in the OCT system in Fig. 10 was monitored simultaneously with the interference signal and the signal was divided by the monitored intensity for each data to compensate for the intensity variations in Fig. 4(a). This compensation procedure should have a limit for accuracy. Numerically calculated PSF with intensity fluctuations of 1.3, 1.4, and 1.5% respectively give depth-independent noise floors of −72.3, −72.1, and −71.5 dB. Comparing these values with the experimental noise floor, we estimated effective compensation for intensity to be about 1.4% in standard deviation giving the noise floor a lower limit of −72 dB, which is indicated by the orange dashed horizontal line (e) in Fig. 12.
The peak was calculated, indicated by the blue solid curve (f) in Fig. 12, and the noise floor was calculated, indicated by the red solid curve (g), taking into account the effects of intra-step frequency variations indicated by curve (b), inter-step frequency fluctuations indicated by curves (c) and (d), and the lower limit line for intensity fluctuations (e). The fits of curves to experimental data are poor, indicating inaccuracy in the numerical calculations. However, the numerical simulation explains the overall behavior of the peak and noise floor as functions of the depth. It tells us that we can improve sensitivity roll-off and the dynamic range of the peak-to-noise floor by reducing the frequency fluctuations that were caused by fluctuations in injection currents in our particular system. The injection current fluctuations could be reduced by reducing the electrical noise in the scan control circuit in Fig. 1.
We used the PSF characteristics in Fig. 12 to reduce noise optically aliased from outside the principal OCT imaging depth. An anti-aliasing electrical filter at the output of the photodetector was not effective in optical frequency comb FD-OCT. Therefore, optically aliased light outside the principal OCT imaging depth were observed as noise [36,43]. Figure 12 indicates that this noise was observed as peaks with a noise floor when the OPD was shorter than about 270 mm, which was observed as white noise without a peak at OPDs longer than it, and the white noise level gradually decreased with further increases in OPD. The optically aliased noise in our system outlined in Fig. 10 was mainly generated by the assembly of the collimator CLS and the fiber end to it. We made the OPD longer than 700 mm by setting the distance between the CLS and GM slightly longer than the length. The aliased noise floor under this condition was expected to be about −50 dB lower than it was observed as a peak at OPD = 0. The reflected light intensity without a sample from the sample arm was about −76 dB of the light intensity illuminated to the sample in our system. Therefore, the sensitivity limit set by the noise floor due to optically aliased noise was estimated to be about 126 dB. The actually measured sensitivity of the system in Fig. 10 was 124 dB with an illumination power onto the sample of 9.4 mW, which is within the American National Standards Institute (ANSI) safety limit . The sensitivity was limited by the noise floor due to optically aliased noise, which was 4 dB higher than the noise floor observed by cutting the sample arm completely (disconnecting a connector between the PCS and CP5 in Fig. 10). Therefore, further reducing optically aliased noise would improve sensitivity. The theoretically expected sensitivity using the formula 10log(ηP0/hνfA)  is 133 dB with a detector sensitivity of η = 0.8, illumination power of P0 = 9.4 mW, and A-scan rate of fA = 3.1 kHz. The hν is the single photon energy.
The resolution measured with the FWHM of PSF peaks was 15 μm at all depths within the principal OCT imaging depth without using a window. The resolution was reduced to 23 μm using a Hanning window. It was slightly worse than the theoretically expected resolution of 14 μm for a rectangular window and 22 μm with the Hanning window.
3.2 Tissue imaging
Tissue was imaged with an illumination power of 9.4 mW onto the sample using a Hanning window in FFT processing. The 1498 data we obtained was zero padded to a size of 4096 and FFT processing was performed. Therefore, there were 2048 data points per 12 mm in depth. We carried out 1024 A-scans per image for Figs. 14(a), 14(c), and 14(d) and 512 for Fig. 14(b). The transverse length of scans was adjusted by changing the amplitude of swing of the GM. The re-scaling process discussed in relation to Fig. 11 was not performed for OCT imaging to save time, because no serious ghosts were observed without processing.
There is an OCT image of the whole of the anterior segment of a human eye in Fig. 14(a). OPD = 0 was set at the top of the image and the focal point of the objective lens was set at a depth of 10 mm. The cornea (C), sclera (S), iris (I), ciliary body (CB), and crystalline lens (CL) are imaged. Despite the strong absorption of light by water in this wavelength region, the posterior capsule (PC) of the CL was imaged as well as the anterior capsule (AC). As far as we know, this is the first report on imaging the PC in a wavelength region of 1600 nm. Light absorption in water has a local minimum at about 1700 nm. A clearer imaging of the PC might be possible in the 1700 nm range. Better quality of OCT imaging has been reported in the 1700 nm wavelength region compared with the 1300 nm wavelength region due to reduced multiple scattering in some tissues [52,53] despite stronger absorption of light by water. The sclera is a strong scatterer of light and the wavelength region of 1700 nm may be better for imaging the CB and angle compared with the conventionally used wavelength of 1300 nm. The wavelength region of 1700 nm may be a promising option to measure the whole anterior segment of the eye.
Part of the surface of the crystalline lens (SC) could sometimes be observed beneath the iris. There is an example in Fig. 14(b). The subject was an Asian man with a pigmented black iris. The nail plate (NP) of a little finger is imaged in Fig. 14(c). The dorsal nail plate (DNP), layered structure of the intermediate nail plate (INP), eponychium (E), epidermis (EP), and dermis (D) are imaged. The dermis beneath the nail plate is clearly imaged. The blood vessel (BV) under the skin could be imaged for skin at the back of the hand as seen in Fig. 14(d). We considered the imaging depths to be good in all the images shown in Fig. 14.
Although the exhibited image quality is good, we notice tail-like background above and beneath the sclera in Fig. 14(a). And we can also detect very faint ghost of the iris beneath the left hand iris image. These are due to insufficient reduction of the noise floor and peaks in PSFs shown in Fig. 11. The image quality would be improved by further improvement of PSF dynamic range and purity, which could be attained by developments of better tuning methods and electrical control circuits with reduced noise than this work.
4. Summary and conclusions
Details on the dynamical tuning method of SSG-DBR lasers were explained, which could be applied to other VT-DBR lasers. The frequency of the lasers was scanned stepwise in the tuning to achieve an optical comb SS-OCT. A simple procedure to create start-up look-up tables for injection currents was introduced. Elements were redundantly inserted into the look-up tables, which were eliminated by selectively sampling the data processing, to eliminate stitching noise associated with abrupt large changes in injection currents. Frequencies were fine tuned to respective target values by only adjusting the phase injection current while monitoring signals of three MZIs of different periods for lasers scanned in the same mode as that used for SS-OCT measurements. Monitoring interference signals from MZIs allowed frequencies to be accurately tuned. A method of concatenating two lasers to enhance resolution was explained, where polarization, intensity, and frequency were continuously connected. The scan regions of the two lasers were overlapped for frequency continuation, and data were selectively sampled with processing to remove redundant data. Single mode lasing with SMSR that was better than 45 dB was confirmed over the entire frequency range of the two lasers.
An SS-OCT system to reduce optically aliased noise that is inherent in optical comb SS-OCT and to compensate for source power fluctuations was designed. Leak and cross-talk light was minimized in the system so that it was as small as possible in the sample arm. Ten samplings were created in each frequency step during data acquisition and four initial samples were discarded to reduce the effect of transient variations. PSFs measured at the principal OCT imaging depth demonstrated no sensitivity roll-off. The PSF dynamic range varied from 72 to 55 dB as the depth increased at the principal OCT imaging depth. The dependence of the PSF peak and noise floor on depth was measured to a depth of 340 mm. Sensitivity roll-off by 6 dB was observed at a depth of 66 mm. The dependence of the peak and noise floor of PSF on depth was simulated taking into account the observed deviations in frequencies from target values. The simulation could explain the overall behavior of the PSF and it suggested that it is important to improve the quality of SS-OCT using VT-DBR lasers by reducing electrical noise in injection currents, which leads to reduced frequency variations from target values.
OCT images were obtained for the anterior segments of human eyes and human skin. Good imaging depths were confirmed for all the images. Confirmation of imaging of the posterior capsule of the crystalline lens in the 1600 nm wavelength range suggests a practical anterior segment OCT system near this wavelength, preferably with a center wavelength of 1700 nm.
We intend to improve the speed and range of wavelength coverage in future work as well reduce electrical noise in the control circuit.
This work was supported by the Japan Science and Technology Agency’s (JST) “Development of Systems and Technology for Advanced Measurement and Analysis” program. The authors wish to extend their thanks to Professors Kimiya Shimizu and Yuzo Yoshikuni of Kitasato University for their continuous support and the valuable discussions they had with us. Thanks are also extended to Dr. Hiromi Oohashi and the members of her group at NTT Photonics Laboratories for their helpful advice and for setting up the prototype SSG-DBR laser system for us.
References and links
1. D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W. Chang, M. R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, and J. G. Fujimoto, “Optical Coherence Tomography,” Science 254(5035), 1178–1181 (1991). [CrossRef] [PubMed]
2. W. Drexler and J. G. Fujimoto eds., Optical Coherence Tomography: Technology and Applications (Springer-Verlag, Berlin, 2008).
3. J. F. de Boer, B. Cense, B. H. Park, M. C. Pierce, G. J. Tearney, and B. E. Bouma, “Improved signal-to-noise ratio in spectral-domain compared with time-domain optical coherence tomography,” Opt. Lett. 28(21), 2067–2069 (2003). [CrossRef] [PubMed]
5. M. A. Choma, M. V. Sarunic, C. H. Yang, and J. A. Izatt, “Sensitivity advantage of swept source and Fourier domain optical coherence tomography,” Opt. Express 11(18), 2183–2189 (2003). [CrossRef] [PubMed]
6. B. E. Bouma, G. J. Tearney, B. J. Vakoc, and S. H. Yun, “Optical frequency domain imaging,” in Optical Coherence Tomography: Technology and Applications, W. Drexler and J. G. Fujimoto, eds. (Springer-Verlag, Berlin, 2008), pp. 209–237.
7. S. H. Yun, C. Boudoux, G. J. Tearney, and B. E. Bouma, “High-speed wavelength-swept semiconductor laser with a polygon-scanner-based wavelength filter,” Opt. Lett. 28(20), 1981–1983 (2003). [CrossRef] [PubMed]
10. R. Huber, M. Wojtkowski, K. Taira, J. G. Fujimoto, and K. Hsu, “Amplified, frequency swept lasers for frequency domain reflectometry and OCT imaging: design and scaling principles,” Opt. Express 13(9), 3513–3528 (2005). [CrossRef] [PubMed]
11. B. D. Goldberg, S. M. R. Motaghian Nezam, P. Jillella, B. E. Bouma, and G. J. Tearney, “Miniature swept source for point of care optical frequency domain imaging,” Opt. Express 17(5), 3619–3629 (2009). [CrossRef] [PubMed]
12. W. Y. Oh, B. J. Vakoc, M. Shishkov, G. J. Tearney, and B. E. Bouma, “>400 kHz repetition rate wavelength-swept laser and application to high-speed optical frequency domain imaging,” Opt. Lett. 35(17), 2919–2921 (2010). [CrossRef] [PubMed]
13. R. Huber, M. Wojtkowski, and J. G. Fujimoto, “Fourier Domain Mode Locking (FDML): A new laser operating regime and applications for optical coherence tomography,” Opt. Express 14(8), 3225–3237 (2006). [CrossRef] [PubMed]
14. W. Wieser, B. R. Biedermann, T. Klein, C. M. Eigenwillig, and R. Huber, “Multi-megahertz OCT: High quality 3D imaging at 20 million A-scans and 4.5 GVoxels per second,” Opt. Express 18(14), 14685–14704 (2010). [CrossRef] [PubMed]
15. M. Kuznetsov, W. Atia, B. Johnson, and D. Flanders, “Compact ultrafast reflective Fabry-Perot tunable lasers for OCT imaging applications,” Proc. SPIE 7554(75541F), 75541F (2010). [CrossRef]
16. B. Potsaid, B. Baumann, D. Huang, S. Barry, A. E. Cable, J. S. Schuman, J. S. Duker, and J. G. Fujimoto, “Ultrahigh speed 1050nm swept source/Fourier domain OCT retinal and anterior segment imaging at 100,000 to 400,000 axial scans per second,” Opt. Express 18(19), 20029–20048 (2010). [CrossRef] [PubMed]
17. A.-H. Dhalla, D. Nankivil, and J. A. Izatt, “Complex conjugate resolved heterodyne swept source optical coherence tomography using coherence revival,” Biomed. Opt. Express 3(3), 633–649 (2012). [CrossRef] [PubMed]
18. B. Potsaid, V. Jayaraman, J. G. Fujimoto, J. Jiang, P. J. S. Heim, and A. E. Cable, “MEMS tunable VCSEL light source for ultrahigh speed 60 kHz-1 MHz axial scan rate and long range centimeter class OCT imaging,” Proc. SPIE 8213(82130M), 82130M (2012). [CrossRef]
19. I. Grulkowski, J. J. Liu, B. Potsaid, V. Jayaraman, C. D. Lu, J. Jiang, A. E. Cable, J. S. Duker, and J. G. Fujimoto, “Retinal, anterior segment and full eye imaging using ultrahigh speed swept source OCT with vertical-cavity surface emitting lasers,” Biomed. Opt. Express 3(11), 2733–2751 (2012). [CrossRef] [PubMed]
20. D. Derickson, M. Bernacil, A. DeKelaita, B. Maher, S. O’Connor, M. N. Sysak, and L. Johanssen, “SGDBR single-chip wavelength tunable lasers for swept source OCT,” Proc. SPIE 6847(68472P), 68472P(2008). [CrossRef]
21. S. O’Connor, M. A. Bernacil, A. DeKelaita, B. Maher, and D. Derickson, “100 kHz axial scan rate swept-wavelength OCT using sampled grating distributed Bragg reflector lasers,” Proc. SPIE 7168(716825), 716825(2009). [CrossRef]
22. B. George and D. Derickson, “High-speed concatenation of frequency ramps using sampled grating distributed Bragg reflector laser diode source for OCT resolution enhancement,” Proc. SPIE 7554(75542O), 75542O (2010). [CrossRef]
23. J. Ensher, P. Boschert, K. Featherston, J. Huber, M. Crawford, M. Minneman, C. Chiccone, and D. Derickson, “Long coherence length and linear sweep without an external optical k-clock in a monolithic semiconductor laser for inexpensive optical coherence tomography,” Proc. SPIE 8213(82130T), 82130T(2012). [CrossRef]
24. V. Jayaraman, Z.-M. Chuang, and L. A. Coldren, “Theory, design, and performance of extended tuning range semiconductor lasers with sampled gratings,” IEEE J. Quantum Electron. 29(6), 1824–1834 (1993). [CrossRef]
25. Y. Tohmori, Y. Yoshikuni, T. Tamamura, H. Ishii, Y. Kondo, and M. Yamamoto, “Broad-range wavelength tuning in DBR lasers with super structure grating (SSG),” IEEE Photon. Technol. Lett. 5(2), 126–129 (1993). [CrossRef]
26. H. Ishii, H. Tanobe, F. Kano, Y. Tohmori, Y. Kondo, and Y. Yoshikuni, “Broad-range wavelength coverage (62.4 nm) with superstructure-grating DBR laser,” Electron. Lett. 32(5), 454–455 (1996). [CrossRef]
27. H. Ishii, H. Tanobe, F. Kano, Y. Tohmori, Y. Kondo, and Y. Yoshikuni, “Quasicontinuous wavelength tuning in super- structure-grating (SSG) DBR lasers,” IEEE J. Quantum Electron. 32(3), 433–441 (1996). [CrossRef]
28. M. Öberg, S. Nilsson, K. Streubel, J. Wallin, L. Bäckbom, and T. Klinga, “74 nm wavelength tuning range of an InGaAsP-InP vertical grating assisted codirectional coupler laser with rear sampled grating reflector,” IEEE Photon. Technol. Lett. 5(7), 735–737 (1993). [CrossRef]
29. A. J. Ward, D. J. Robbins, G. Busico, E. Barton, L. Ponnampalam, J. P. Duck, N. D. Whitbread, P. J. Williams, D. C. J. Reid, A. C. Carter, and M. J. Wale, “Widely tunable DS-DBR laser with monolithically integrated SOA: design and performance,” IEEE J. Sel. Top. Quantum Electron. 11(1), 149–156 (2005). [CrossRef]
30. R. Laroy, G. Morthier, T. Mullane, M. Todd, and R. Baets, “Stabilisation and control of widely tunable MG-Y lasers with integrated photodetectors,” IET Optoelectron. 1(1), 35–38 (2007). [CrossRef]
31. T. Amano, H. Hiro-Oka, D. H. Choi, H. Furukawa, F. Kano, M. Takeda, M. Nakanishi, K. Shimizu, and K. Ohbayashi, “Optical frequency-domain reflectometry with a rapid wavelength-scanning superstructure-grating distributed Bragg reflector laser,” Appl. Opt. 44(5), 808–816 (2005). [CrossRef] [PubMed]
32. O. Ishida, Y. Tada, N. Shibata, and H. Ishii, “Fast and stable frequency switching employing a delayed self-duplex (DSD) light source,” IEEE Photon. Technol. Lett. 6(1), 13–16 (1994). [CrossRef]
33. J. E. Simsarian, M. C. Larson, H. E. Garrett, H. Xu, and T. A. Strand, “Less than 5-ns wavelength switching with an SG-DBR laser,” IEEE Photon. Technol. Lett. 18(4), 565–567 (2006). [CrossRef]
34. N. Fujiwara, R. Yoshimura, K. Kato, H. Ishii, F. Kano, Y. Kawaguchi, Y. Kondo, K. Ohbayashi, and H. Oohashi, “140-nm quasi-continuous fast sweep using SSG-DBR lasers,” IEEE Photon. Technol. Lett. 20(12), 1015–1017 (2008). [CrossRef]
35. N. Fujiwara, H. Ishii, H. Okamoto, Y. Kawaguchi, Y. Kondo, and H. Oohashi, “Suppression of Thermal Wavelength Drift in Super-Structure Grating Distributed Bragg Reflector (SSG-DBR) Laser with Thermal Drift Compensator,” IEEE J. Sel. Top. Quantum Electron. 13(5), 1164–1169 (2007). [CrossRef]
37. T. Amano, H. Hiro-Oka, D. Choi, H. Furukawa, F. Kano, M. Takeda, M. Nakanishi, K. Shimizu, and K. Obayashi, “OFDR with an SSG-DBR laser,” Proc. SPIE 5531, 375–382 (2004). [CrossRef]
38. D. Choi, T. Amano, H. Hiro-Oka, H. Furukawa, T. Miyazawa, R. Yoshimura, M. Nakanishi, K. Shimizu, and K. Ohbayashi, “Tissue imaging by OFDR-OCT using an SSG-DBR laser,” Proc. SPIE 5690, 101–113 (2005). [CrossRef]
39. D. Choi, H. Hiro-oka, T. Amano, H. Furukawa, N. Fujiwara, H. Ishii, and K. Ohbayashi, “A method of improving scanning speed and resolution of OFDR-OCT using multiple SSG-DBR lasers simultaneously,” Proc. SPIE 6429(64292E), 64292E (2007). [CrossRef]
40. K. Ohbayashi, T. Amano, H. Hiro-Oka, H. Furukawa, D. Choi, P. Jayavel, R. Yoshimura, K. Asaka, N. Fujiwara, H. Ishii, M. Suzuki, M. Nakanishi, and K. Shimizu, “Discretely swept optical-frequency domain imaging toward high-resolution, high-speed, high-sensitivity, and long-depth-range,” Proc. SPIE 6429(64291G), 64291G (2007). [CrossRef]
41. H. Kakuma, K. Ohbayashi, and Y. Arakawa, “Optical imaging of hard and soft dental tissues using discretely swept optical frequency domain reflectometry optical coherence tomography at wavelengths from 1560 to 1600 nm,” J. Biomed. Opt. 13(1), 014012 (2008). [CrossRef] [PubMed]
42. H. Kakuma, D. Choi, H. Furukawa, H. Hiro-oka, and K. Ohbayashi, “24 mm depth range discretely swept optical frequency domain imaging in dentistry,” Proc. SPIE 7162(717208), 717208 (2009).
43. T. Bajraszewski, M. Wojtkowski, M. Szkulmowski, A. Szkulmowska, R. Huber, and A. Kowalczyk, “Improved spectral optical coherence tomography using optical frequency comb,” Opt. Express 16(6), 4163–4176 (2008). [CrossRef] [PubMed]
44. D. Choi, R. Yoshimura, H. Hiro-oka, H. Furukawa, A. Goto, N. Satoh, A. Igarashi, M. Nakanishi, K. Shimizu, and K. Ohbayashi, “Discretly swept optical coherence tomography system using super-structure grating distributed Bragg reflector lasers at 1561-1639 nm,” Proc. SPIE 8213(82132F), 82132F (2012). [CrossRef]
45. G. Sarlet, G. Morthier, and R. Baets, “Control of widely tunable SSG-DBR lasers for dense wavelength division multiplexing,” J. Lightwave Technol. 18(8), 1128–1138 (2000). [CrossRef]
46. F. Kano, H. Ishii, Y. Tohmori, M. Yamamoto, and Y. Yoshikuni, “Broad range wavelength switching in superstructure grating distributed Bragg reflector lasers,” Electron. Lett. 29(12), 1091–1092 (1993). [CrossRef]
47. L. A. Coldren, “Monolithic tunable diode lasers,” IEEE J. Sel. Top. Quantum Electron. 6(6), 988–999 (2000). [CrossRef]
48. S. H. Yun, G. T. Tearney, B. E. Bouma, B. H. Park, and J. F. de Boer, “High-speed spectral-domain optical coherence tomography at 1.3 mum wavelength,” Opt. Express 11(26), 3598–3604 (2003). [CrossRef] [PubMed]
51. American National Standards Institute, “American national standard for safe use of lasers,” ANSI Z136.1–200 (ANSI, 2000).
53. S. Ishida and N. Nishizawa, “Quantitative comparison of contrast and imaging depth of ultrahigh-resolution optical coherence tomography images in 800-1700 nm wavelength region,” Biomed. Opt. Express 3(2), 282–294 (2012). [CrossRef] [PubMed]