2.1.

Optical Design

Assembly of the module was done using the NanoGlue active alignment workstation by Nanosystec.²¹ The NanoGlue is composed of two independent stages, which use a customized gripper appendage to pick and place optical components onto the substrate. Each stage has six degrees of freedom ($x$, $y$, $z$, pitch, roll, and yaw) with a linear and rotary axis resolution of 20 nm and 0.0005 deg, respectively. A vertically mounted camera provides machine vision capability and visual feedback to the operator. A scanning slit beam profiler (Thorlabs, BP209-IR/M) is mounted on a translation stage, which provides beam profiling capabilities as well as simple off-board optical power detection of NIR wavelengths. Each component was glued to the substrate using a low shrinkage UV cured adhesive (Dymax, 6-621-VT).

The light source used in this device is a Denselight (DL-CS3184A) superluminescent diode (SLD). This source has a central wavelength, ${\lambda }_0$ of 1310 nm, with a bandwidth, $\mathrm{\Delta }\lambda$ of 40 nm, giving an axial resolution $\mathrm{\Delta }z\sim 19\mu \mathrm{m}$. The light from the SLD is collimated using a fiber collimator, which is mounted directly on the module giving a beam diameter of $350\mu \mathrm{m}$. A 5-mm beam splitter is used to split the input beam into the sample and reference paths. The sample and reference paths contain an achromat focusing lens (Ross Optical, L-MAC000) with a 3-mm diameter, resulting in a spot size of $30\mu \mathrm{m}$ and a Rayleigh range of $570\mu \mathrm{m}$. On the reference arm, a partial mirror having a reflectivity 80% is placed a distance of $100\mu \mathrm{m}$ in front of the reference mirror. The reference mirror is mounted on a voice coil (Alps, ATMC1Z301A). The detection side of the design consists of a focusing lens and an InGaAs photodiode detector (ThorLabs, FDGA05) with a $500\mu \mathrm{m}$ active area. The photo detector is connected to a custom designed transimpedance amplifier, which sits underneath the optical bench. All surfaces of the optical components (except for the reference mirror and partial mirror) are AR coated.

During assembly of this photonic module, it was observed that the overall performance is heavily dependent on the movement of the voice coil. The pointing stability of a particular voice coil affects the quality of the interference fringes observed at the detector. This sensitivity was alleviated somewhat through the use of focusing lenses on the sample and reference arms. Operation of the voice coil is based on a sinusoidal input voltage at 152 Hz with a scan range of $\sim 35\mu \mathrm{m}$. Tuning the DC offset is of vital importance since this has an effect on the spacing between the partial mirror and reference mirror and hence the separation between consecutive orders of reflection from the reference mirror.²² It should be highlighted that the voice coil used in this work was being used outside of its design parameters. This particular voice coil was designed for focusing a lens in the camera of a mobile phone and does not normally operate in a bipolar mode. For this reason, tuning is required. In future iterations, a custom voice coil with improved pointing stability will be used.

The substrate on which all the optical components were placed is made of aluminum with dimensions of $40\times 57\mathrm{mm}$. The design of the substrate included a raised central “rail” on which the main optical components were placed. Mechanical mounts for the optical components were custom designed and manufactured.

2.2.

Alignment Procedure

Extensive planning went into the design of the alignment procedure. Since it was not possible to remove or alter the orientation of individual components after gluing, the sequence of adding components had to be planned within the constraint of allowing spatial clearance of the arms of the NanoGlue system at each step. In terms of the alignment of each of the individual optical components, a number of methods were used to optimize the position and tip/tilt of each component.

The NanoGlue system has an onboard beam profiler and photodiode, both of which were used during alignment. The detector could be used by monitoring a DC signal or an AC signal in cases where interference fringes were being optimized. When using a DC signal, a feedback control system could be employed to perform “active” alignment. During assembly, a circulator was used on the input side of the module. This proved to be extremely useful in eliminating alignment variables of other components by maximizing power reflected back into the collimator. During alignment, a mirror (the same as the reference mirror) was used on the sample side. When optimizing alignment using interference patterns (or interferograms), an oscilloscope was used to maximize the interference amplitude. Specially designed alignment tools for fine-tuning the direction and position of the beams were manufactured. These were made from precision-machined aluminum and consisted of a $450\text{-}\mu \mathrm{m}$ diameter hole with a datum surface that was in contact with the corresponding datum surface on the substrate. This served to locate the position of the beam center consistently (within the machined tolerance of $\pm 10\mu \mathrm{m}$) when the tool was in contact with the substrate. The alignment of optical components was based on achieving maximum power transmitted through the tool. Two alignment tools were utilized when necessary.

The choice of optical components was aided by the use of Zemax to model the beam characteristics into and out of each component. The working distances of the lenses were modeled, and the position of each of the components could be approximated. These measurements were imported into SolidWorks 3-D CAD software, where the prototype was developed for assembly.

The steps of alignment were as follows. The first component placed on the bench was the collimator. This was first glued to a mount, and the mount was then aligned with respect to the substrate. Two alignment tools were used to ensure that the collimated beam was both horizontal and not deviating from the optical axis. The on-board beam profiler was used to provide a feedback signal to optimize alignment. Next, the voice coil and reference mirror were aligned. The voice coil was fixed into its mount, and the reference mirror was glued to the barrel of the voice coil. The voice coil was actuated during alignment, and the input collimated beam was reflected straight back into the collimator by the reference mirror. The angle of the voice coil mount was optimized by maximizing the reflected signal observed at port 3 on the circulator signal. Next, the beam splitter was aligned. An alignment tool was placed on the turned path (sample arm), and by maximizing the signal observed by the beam profiler, the position and angle of the beam splitter were optimized. Next, a sample mirror was positioned off-board. This was mounted on a tip/tilt stage and a translation stage. It was placed at the zero-path length position, thereby allowing interference to be observed at the circulator signal using an oscilloscope. At this stage, the photodiode detector was aligned at the output side of the interferometer. Its output was connected to the transimpendance amplifier, and the interference pattern could then be monitored using this detector and an oscilloscope. The detector was held in place with one of the arms of the NanoGlue. A removable aluminum mount was screwed onto the sample side of the interferometer, and an optical blank was glued to it. This blank acts as dispersion compensation for the partial mirror. Next, the partial mirror was put in place. This was first roughly aligned using the circulator feedback and was then fine-tuned by maximizing the amplitude of the interference fringes and, most importantly, the number of orders seen. Typically, 20 orders could easily be seen. Moving the sample mirror through the orders of reflection and measuring the displacement tuned the position of the partial mirror relative to the position of the reference mirror. Next, the sample and reference lenses were aligned. Maximizing the DC signal seen on the circulator optimized the position of both of these. The detector lens was aligned to achieve maximum signal intensity on the photodiode detector. The lens, and then the detector, were glued into place.

2.3.

System Control and Experimental Set Up

System control for the photonic module was done via a custom-built FPGA board. The FPGA board interfaces with a PC to provide a user interface and display the acquired images. As with all time-domain OCT systems, the depth-information is achieved by mechanical movement of the reference mirror, which is mounted on the voice coil. Lateral scanning is achieved by mounting the module on a translation stage. The oscillation of the voice coil and synchronization with the lateral translation stages is controlled via the FPGA board. During data acquisition, the photonic module is mounted in the vertical orientation, so the voice coil is symmetrically loaded. The analog data are acquired and conditioned by the transimpedance amplifier. The data are then digitized by an analogue to digital converter. The FPGA is used to stream the data from the A to D to a laptop, via a USB 3 interface, where the data are processed. This is shown diagrammatically in Fig. 2.

Fig. 2

(a) Schematic of the system control design. (b) Picture of the overview of the experimental set up. The critical system components are highlighted and correspond to the colors in (a). Note that the photonic module is highlighted in yellow. The inset image labeled (c) shows a close-up of the photonic module.

4.1.

System Characterization

The spectrum of the SLD used with this device was measured using a spectrometer and is shown in Fig. 4(a). A Gaussian function was fit to this profile, and a central wavelength, ${\lambda }_0$ of 1311 nm and a bandwidth, $\mathrm{\Delta }\lambda$ of 40.9 nm were measured. Using these values, the “round trip” coherence length, $l_{\mathrm{c}}\sim 18.5\mu \mathrm{m}$ (which corresponds to the axial resolution of the system) was approximated. This was verified experimentally by analyzing the axial point-spread function of a single reflector by using a plane mirror as the sample [Fig. 4(b)]. The definition of the round-trip coherence length is the full-width half-max of the point-spread function,²³ which was measured to be $\sim 19.4\mu \mathrm{m}$. The lateral resolution of the system is equal to the spot size of the scanning beam, which is $\sim 30\mu \mathrm{m}$.

Fig. 4

(a) Input spectrum of the SLD. The central wavelength ${\lambda }_0=1311\mathrm{nm}$ and the full-width half-max bandwidth $\mathrm{\Delta }\lambda =40.9\mathrm{nm}$. This results in a “round-trip” coherence length and axial resolution $l_{\mathrm{c}}$ of $18.5\mu \mathrm{m}$. (b) The coherence length was verified experimentally by measuring the full-width half-maximum of the point-spread function of a single A-line taken from a mirror.

To characterize the signal-to-noise ratio (SNR) for each order, a mirror was used as the sample and was moved in depth through the scanning range. A-lines were acquired for each position of the sample, all of which are shown in Fig. 5(a).

Fig. 5

(a) A-lines of a mirror were acquired at multiple depths. (b) An example of a single A-line with a Gaussian fit. The maximum amplitude is used to calculate the SNR. (c) Normalized SNR for each position of the sample mirror.

For each A-line, a Gaussian function was fit to the data points and the maximum amplitude $A_{\max }$ was recorded in each case. An example of a single A-line with its Gaussian fit are shown in Fig. 5(b). A “noise” A-line was also acquired when no sample was present and the standard deviation $\sigma$ was calculated. The SNR for each mirror position was calculated in decibel according to the following equation:

4.2.

Sample Images

Examples of the imaging capabilities of the miniature MR-OCT device can be seen in Fig. 6. Figure 6(a) shows an image of the cross section of a roll of highly scattering translucent tape. The individual layers of the tape can clearly be seen. Figure 6(b) shows the cross section of the surface of a metallic screw. This is relevant to applications of MR-OCT such as nondestructive testing where cost-effective means of detecting scratches and defects on objects produced in high-volume are necessary. Figures 6(c) and 6(d) show in vivo images of the human finger. [All experiments were approved by the ethics committee of the Galway University Hospitals (Ref: C.A. 1053).] Figure 6(c) shows the palmer region of the finger-tip. The epidermis and dermis layers of the tissue can clearly be distinguished (labeled “ED” and “D”). The undulations of the surface of the skin are seen replicated in the deeper dermis–epidermis boundary. This is the subsurface primary fingerprint to which reference was made in the introduction of this paper. Furthermore, the presence of eccrine sweat ducts, which transport sweat to the surface of the skin, can be seen. These appear as bright spiral features and are labeled “SD.” Figure 6(d) shows the nailfold region of the finger. The nail (N), cuticle (C), and nail fold (NF) regions can clearly be distinguished.

Fig. 6

B-frames of various samples. These images have dimensions of 4 mm across by $\sim 1.1\mathrm{mm}$ in depth. (a) Highly scattering translucent tape. (b) Surface of a metal screw. (c) Human in-vivo finger-tip. Note that interesting structural features such as sweat ducts (SD) and the dermal (D) and epidermal (ED) layers can easily be identified. (d) Human in-vivo nail fold (NF) region of the finger. The NF, cuticle (C), and nail (N) can easily be distinguished.