Open Access
Add to BibSonomy
Abstract
In this paper, we present a low-loss optical assembly utilizing a 90°-bent graded-index (GI) core polymer optical waveguide on vertical cavity surface emitting laser (VCSEL) based optical transceivers. The proposed assembly can replace conventional components such as mirrors and lenses for realizing subminiature optical engines applicable to on-board integration. To minimize the total insertion loss of the waveguide when connected to a high-speed VCSEL and a GI-core multimode fiber (MMF) at each end, the characteristics of the beam emitted from VCSELs are measured and taken into consideration for the waveguide design. In order to confirm the effect of insertion loss reduction by the waveguide numerical aperture control, 90°-bent GI-core polymer waveguides are fabricated applying the Mosquito method. The fabricated waveguide exhibits a total insertion loss as low as about 2 dB at 850-nm wavelength, which includes the coupling losses at both ends, bending, and propagation losses. We also investigate a way to reduce the insertion loss when a gap exists between the waveguide and VCSEL chip. We theoretically and experimentally confirm that filling the gap with a high index resin can reduce the coupling loss by 5 dB.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Over the last decades, with the remarkable surge of network traffic in hyper-scale data centers (DCs) and in high-performance computers (HPCs), large capacity data transmission has been required for the DC and HPC networks. In the case of conventional electrical interconnects, achieving both high-speed and high-density data transmission can be difficult due to large transmission loss and crosstalk [1,2]. Contrastingly, optical interconnects have an advantage in their ability to transmit high-speed signals with high wiring density. Therefore, optical interconnects have been widely deployed in inter-rack links in DC and HPCs [3–8]. In order to satisfy the higher-bandwidth demand, even on-board optical interconnects as well as inter-rack optical interconnects may be required. Therefore, high-speed and compact optical engines have been proposed and developed for on-board optical interconnects [9 –13].
If conventional pluggable optical engines compliant to small form factors are used for on-board interconnects, an important issue is how to achieve high-density assembly at the coupling points between active optical devices (light sources and photo detectors) and fibers. In current small form factor compliant optical engines, lens integrated optical connectors are commonly used to couple to optical fibers with a channel pitch as large as 250 μm [7–9]. In addition, many components are required for the lens-based assembly. In particular, light beams emitted from high-speed light sources need to be focused to have a small spot size using multiple lenses in the optical connector. The requirements that the components preserve lens focal lengths, lens diameters, and collimated beam diameters would be an obstacle for high-density optical assembly.
Meanwhile, a couple of techniques to utilize step-index (SI) core polymer optical waveguides have been proposed and developed [14–17]. For off-board links, multimode fibers (MMFs) have been frequently deployed because of the wide tolerance in their alignment in connections, namely low total link cost. So far, optical coupling techniques between active optical devices and polymer waveguides with a 45° mirror at their ends have been proposed. Meanwhile, we have reported that circular graded-index (GI) core polymer optical waveguides exhibit a higher connectivity with MMFs compared to conventional square or rectangular SI-core polymer waveguides [18]. By applying the Mosquito method we developed, circular GI-core waveguides can be fabricated easily [19]. As we reported in [19], low-loss optical coupling could also be realized by applying a 45° mirror to the GI-core polymer waveguides. However, lenses are still required even for the GI-core waveguides with 45° mirrors particularly if the waveguides have thick cladding layers, or when the output beam from the waveguides is coupled to high-speed photodiodes with a small active area. Hence, we proposed to replace the coupling components such as the lenses and mirror with an alternative, a 90°-bent GI-core polymer waveguide. In addition, we also preliminarily demonstrated how to fabricate compact 90°-bent (with a small bending radii) GI-core polymer waveguides [20,21].
In this paper, we discuss the optimum structure for the 90°-bent GI-core polymer waveguides to be an assembly on VCSEL-based optical engines for MMF links. To achieve efficient optical coupling, we design the optimum numeral aperture (NA) of the GI-core waveguides and fabricate them. Then, we experimentally confirm that a low insertion loss of about 2 dB is attained by adjusting the NA, even under a bending radius as small as 1 mm.
2. Optical engine with a 90°-bent GI core polymer optical waveguide
2.1 Basic structure of the optical engine
Figure 1(a) shows a side view cross-section of the optical engine designed in this paper compared to conventional one shown in Fig. 1(b) [20]. We already proposed an optical engine with a 90°-bent GI-core polymer waveguide [20], as shown in Fig. 1(a). The optical engines are composed of a VCSEL array, a photodiode (PD) array, and ICs (laser diode driver (LDD) and trans-impedance amplifier (TIA)). Thermal vias are formed in the substrate close to the ICs for efficient heat dissipation.
Fig. 1. Side view cross-sections of optical engines with (a) proposed and (b) conventional structures.
In the conventional optical assembly shown in Fig. 1(b), an optical connector in which lenses and a reflection mirror are integrated is used. However, there are many components required for this assembly. In particular, the light coupled to high-speed detectors needs to be focused onto a small spot using multiple lenses in the optical connector. Hence, it is difficult to achieve a high-density optical assembly with existing devices.
Meanwhile, in the proposed optical assembly shown in Fig. 1(a), a 90°-bent GI core polymer waveguide is inserted between the optical device and MMF. We already confirmed that such a waveguide with 90°-bent GI-cores could be fabricated using the Mosquito method where a bending radius as small as 1 mm was realized [21]. Here, waveguide parameters of the core diameter, bending radius, and NA influence the waveguide insertion loss (the bending and coupling losses). In the Mosquito method, these waveguide parameters can be controlled by adjusting several fabrication conditions such as: (i) the combination of the core and cladding materials, (ii) the core monomer dispensing pressure, (iii) the interim time for monomer diffusion, (iv) the core monomer viscosity, (v) the needle inner diameter, (vi) the needle length, (vii) the needle-scan velocity, and (viii) the needle-scan path.
2.2 Link loss-budget of the 26-Gbaud optical link
The total loss budget of a 26-Gbaud MMF optical link is designed, using the proposed optical engine with 90°-bent GI-core polymer waveguides referring to [20]. Figure 2 shows a schematic link structure and the designed loss budget diagram. The transmitter (Tx) optical power is set to 0 dBm with the optical modulation amplitude (OMA) assuming the emitted power from a VCSEL at 80 °C. As the operational temperature limit of the VCSELs (in Fig. 2 (1)) [9] is specified to be 80 °C, a power of 0 dBm is regarded as the worst case. The optical coupling loss at the Tx side between the VCSEL and an MMF is set to -3.5 dB (in Fig. 2 (2)) [7–9]. Here, such a high coupling loss could be caused by the diffraction of the VCSEL beam in the “lens-less” coupling in this structure. The propagation loss of a 100-m long MMF cable is estimated to be 1.5 dB including the bit-error-rate (BER) power penalty due to modal and chromatic dispersions as shown in Fig. 2 (3). The coupling loss at the receiver (Rx) side between the MMF and a high-speed PD with a 25-μm${\varnothing}$ active area is set to -2.5 dB under a lens-less coupling as shown in Fig. 2 (4) [7–9]. Moreover, the crosstalk penalty is estimated to be -1.0 dB in Fig. 2 (5) [19]. The optical sensitivity of PD with TIA is set to about -9.0 dBm (denoted as Criterion (broken line) in Fig. 2) [22]. Hence, the link margin is estimated to be 0.5 dB.
2.3 Loss of 90°-bent GI-core waveguide
To realize an optical transmission greater than 26-Gbaud, a power margin of 0.5 dB is not large enough as shown in Fig. 2, hence it is important to keep the insertion loss of the 90°-bent GI-core waveguide as low as possible. The insertion loss of GI-core polymer waveguides in the Tx side can be classified into (1) the coupling loss (between the light source and the waveguide), (2) the bending loss, (3) the propagation loss, and (4) the coupling loss (between the waveguide and MMF). The propagation loss of GI-core polymer waveguides is determined mainly by the core materials, and a typical value is approximately 0.04 dB/cm at 850-nm wavelength when we fabricate them using organic-inorganic hybrid resins supplied by Nissan Chemical. Hence, in the case of a 5-cm long waveguide, the propagation loss is calculated to be about 0.2 dB, which is smaller than other losses, as shown in Fig. 2(3).
On the other hand, both coupling losses(1 and 4) and bending loss(3) depend on the NA of the polymer waveguides: coupling loss(1) and bending loss(3) should decrease while coupling loss(4) could increase with increasing the waveguide NA. Therefore, the NA of the GI-core waveguide needs to be optimized for minimizing the sum of these three factors.
3. Design and analysis of 90°-bent GI-core waveguide
3.1 High-speed VCSEL characteristics
Since the waveguide needs to be coupled to an MMF at its output end, the waveguide core diameter should not differ significantly from that of the MMF. So, the waveguide core diameter is designed to be close to the MMF core diameter, from which the VCSEL beam diameter and divergence angle are key parameters of coupling loss (1) at the waveguide input end. In order to optimize the waveguide NA, high-speed VCSELs are characterized for which specific evaluation boards are designed and fabricated.
Figure 3(a) shows a top-view photo of the evaluation board on which high-speed VCSELs are integrated. The evaluation board is composed of a high-frequency dielectric-substrate (MEGTRON-6, Panasonic Corporation), 50-Ω matched coplanar microstrip lines (CMLs), and a 4-channel VCSEL bare chip. The pads of the bare chip and the CMLs are connected by 20-μm${\varnothing}$ gold wire bonding as shown in Fig. 3(a). Figure 3(b) shows the measurement setup for near field pattern (NFP) and far field pattern (FFP) of the VCSELs. The VCSEL chips evaluated in this paper are summarized in Table 1. The spot size and divergence angle, thus the NA, are calculated from the measured NFP and FFP. Figures 4(a) and (b) show the calculated NA and spot size with respect to the bias current under DC power supply, while Figs. 5(a) and (b) under a 25.78 Gbit/s modulation.
Fig. 3. (a) Photograph of fabricated VCSEL evaluation board and (b) measurement setup of high-speed VCSEL
Fig. 4. Measured bias current dependence (only DC power supply) of (a) NA and (b) beam width of VCSELs.
Fig. 5. Measured bias current dependence (modulated at 25.78 Gbit/s signal) of (a) NA and (b) beam width of VCSELs.
Table 1. High-Speed VCSELs for evaluation
We find from Figs. 4(a) and 5(a) that the NA of the VCSELs increases with increasing bias current, even in the case of single-mode VCSEL (6). In multimode VCSELs, since only low-order modes emit under a low bias near the threshold (close to single-mode operation), a low NA is observed, particularly in VCSELs (1), (2), (4), and (5). Meanwhile, due to diffraction, VCSEL(3) shows higher NA than other multimode VCSELs, independent of the bias current, because its spot size is much smaller than those of the other multimode VCSELs, as shown in Figs. 4(b) and 5(b). In VCSEL(3), the aperture (spot size) could be deliberately reduced to less than 5 μm in order to increase the bandwidth to 21 GHz, which would sacrifice beam directionality in favor of a high NA. When the VCSELs are used in actual high-speed optical engines, a bias current higher than 7 mA should be applied under modulation. From the results shown in Figs. 4 and 5, we find that the NA and spot size measured under DC bias applied are almost identical to those under modulation. Thus, we believe that the modulation speed exhibits no effect on the insertion loss of the polymer waveguide. Here, the results in Fig. 5 are the time-averaged NA and spot size. Hence, in the following calculations, the NA and spot size measured under DC power supply are used for the analysis, and we determine the representative beam NA emitted from all the VCSELs evaluated in this paper as 0.25 or higher, as shown in Figs. 4(a) and 5(a).
On the other hand, little bias current dependence is observed in the spot size in all the VCSELs, as shown in Figs. 4(b) and 5(b). The spot size of VCSEL(3) is about 4 μm and that of VCSEL(4) is about 6 μm. Although these values may look to be almost the same, a 1.5 time difference is observed in the spot size after 10-μm gap transmission at the input end of the core, due to the beam divergence. On the other hand, since the NA of VCSEL(3) is higher than that of VCSEL(4), with increasing the gap between the VCSEL and GI core is, the beam size of VCSEL (3) increases to exceed the beam size of VCSEL(4). Therefore, in general, low-NA VCSELs are expected to exhibit efficient optical coupling since beams from low-NA light sources are less likely to diverge resulting in a directional beam.
In optical engines on which VCSEL chips are wire bonded, it is difficult to apply butt-coupling with waveguide cores, because the wire is a physical obstacle as shown in Fig. 3(a). Therefore, beams from VCSELs need to propagate through a small air gap between them to increase the beam width due to diffraction, which is affected by the NA shown in Figs. 4(a) and 5(a), as well as the original spot size in Figs. 4(b) and 5(b). The beam width win(d) after propagating a distance d from the VCSEL chip surface is expressed by Eq. (1).
Fig. 6. Calculated VCSEL’s beam width dependence on the propagating distance in air gap. The bias current is (a) 5 mA, (b) 7 mA, and (c) 9 mA
We find from Fig. 6(b) that the gap distance should be smaller than 35 μm and 60 μm to obtain beam widths of 30 μm and 50 μm after the gap, respectively. These gap distance limitations are required for high efficiency coupling with 30-μm core waveguides (90°-bent core waveguide) and 50-μm core fibers, respectively. Hence, VCSELs with low NAs tolerate the gap distance, as indicated by the curves of VCSELs (4) and (5). As shown in Table I, these two VCSELs exhibit both high bandwidth and low NA.
3.2 NA and bending radius control in GI-core polymer waveguide
To achieve efficient optical coupling, the insertion loss of 90°-bent GI-core polymer waveguides are simulated assuming an input beam from a VCSEL with the measured characteristics shown in Figs. 4 to 6. Figure 7(a) shows the simulation model for the 90°-bent GI-core waveguide using ray trace analysis software, Zemax OpticStudio. The light from a VCSEL (NA = 0.25, spot size = 7 μm, assuming 7-mA bias current) is coupled to a GI-core waveguide to show the minimum coupling loss. A 5-mm long GI-core waveguide is bent perpendicularly with a bending radius of 1 mm at a 1-mm distance from the input end. First, the gap distance is set to 1 μm, assuming almost no gap, then specific gap distances are investigated. After propagating through the GI-core, the output light from the waveguide core is coupled to a 50-μm GI core MMF (NA = 0.22), and enters a 25-μm${\varnothing}$ PD. Waveguide insertion loss is simulated by varying the waveguide NA. Here, the effective NA (NAeff) of the GI-core waveguide for the simulation is defined by Eq. (2).
where ncore and nclad are the refractive indices of the core and cladding of the polymer waveguide, respectively. In this simulation, the refractive index of the core is set to 1.585 (assuming a core polymer), while the refractive index of the cladding is varied to calculate the waveguide NA dependence.Fig. 7. (a) Simulation model for 90°-bent GI-core polymer waveguide, (b) Calculated NA dependence and (c) distance dependence of insertion loss [21]
Figure 7(b) shows the calculated waveguide NA dependence of the insertion loss. The calculated results for several different core sizes from 10 to 70 μm are also shown in Fig. 7(b), which are indicated as 10-GI, 30-GI, 50-GI, and 70-GI. We find from Fig. 7(b) that 30- and 50-μm GI cores show lower insertion losses than the other two. Moreover, by optimizing the NA and core size, the lowest insertion loss of about 0.7 dB can be achieved under a bending radius as small as 1 mm. The higher NA of the waveguide could increase the coupling loss with an MMF (NA = 0.22) due to NA mismatch, while the total insertion loss decreases because the bending loss reduction contributes more to the total than the coupling loss.
At the input end, the size mismatch between the GI core and input VCSEL beam tends to increase the coupling loss. In addition, the diverging beam from the VCSEL due to diffraction, excites a wider range of mode orders in the waveguide core, exhibiting an over-filled lunch (OFL) condition. Since the higher-order modes are sensitive to core bending, bending loss should increase compared to that under a restricted mode launch condition with a small beam spot. Thus, it is important to control the NA of the VCSEL as well as the waveguide core diameter to reduce the total insertion loss. Figure 7(c) shows the calculated gap distance dependence of the waveguide insertion loss. The NAeff and core size of the waveguide are set to 0.35 and 30 μm, respectively. When the VCSEL NA is higher than 0.35, the gap between the VCSEL and the waveguide core must be less than 65 μm to keep the insertion loss lower than 3.5 dB. Hence, by optimizing the GI-core design and controlling the NA of high-speed VCSEL, a more efficient optical coupling specified in the loss budget (Fig. 2) can be achieved under a bending radius of 1 mm.
When applying the waveguide to optical engines, it is also important to investigate the effect of alignment tolerance between the waveguide and VCSEL on the insertion loss. The alignment tolerance is currently under evaluation, and the results will be presented elsewhere.
4. Waveguide fabrication
The 90°-bent GI-circular-core polymer waveguides designed above are experimentally fabricated using the Mosquito method. The Mosquito method has a unique feature that arbitrary core alignment patterns can be formed, since the needle tip is placed in the cladding monomer and can scan in three dimensions [19,23]. First, the fabrication conditions for perpendicularly bending the core with a 1-mm bending radius or smaller are investigated.
Figure 8 schematically shows the needle-scan path to dispense the core monomer for a 90°-bent GI circular core polymer waveguide with a 1-mm radius. We already employed the scan paths shown in Fig. 8(a) for a bending radius larger than 5 mm [13]. To decrease the bending radius to 1 mm or less, the needle-scan velocity is an issue because some cladding monomer flows are caused by the needle scan, which can disrupt core alignment. The needle-scan path is redesigned as shown in Figs. 8(b) and (c). The needle-scan path in Figs. 8(c) takes the effect of the monomer flow into consideration. For instance, a bending angle θ in the scan path is set less than 90°, as shown in Fig. 8(c). Photos (1) and (2) in Fig. 9 show the top-view of fabricated polymer waveguides under different needle-scan paths. The fabricated waveguides formed by a conventional needle-scan path are also shown in (3) and (4) in Fig. 9. The result in (4) indicates that it is difficult to fabricate waveguides with a bending radius smaller than 1 mm using a conventional scan path. However, we confirm that a bending radius smaller than 1 mm can be successfully formed by optimizing the needle-scan path as shown in (1) and (2) in Fig. 9. Here, the fabricated waveguide size is as large as 25-mm high, 35-mm wide, and 1-mm thick. Since the 90°-bent GI core waveguide in this paper is designed for the basic evaluation of bending loss, the waveguide is fabricated by scanning the needle on a single plane parallel to the substrate surface, and the core length before and after bending is not minimized. Therefore, the length of the waveguide from the input end to the bending point is 20 mm, which can be shortened by dicing, polishing, or other end terminating processes.
Fig. 8. Needle-scan path designs for GI-core waveguide with small-bending radius (R = 1mm)
(a) conventional path, (b) θ = 90° and (c) θ <90°.
Fig. 9. Top-view photographs of fabricated polymer waveguides dispensed on different needle-scan paths (1),(2): proposed path (shown in Fig. 8(b),(c)), (3),(4): conventional path..
Next, the fabrication conditions for a high-NA GI-core waveguides are investigated. We already confirmed in the simulation results shown in Fig. 7(b) that a waveguide NA higher than 0.3 is required to keep the insertion loss sufficiently low. Hence, the higher-NA waveguides are fabricated using a cladding material with low refractive index. Meanwhile, we keep using the same core material, organic-inorganic hybrid resin that we have used in the Mosquito method. In order to vary the refractive index of the cladding, the waveguides are fabricated under conditions (1)-(5) in Table 2. The effective core diameter and NA of the fabricated waveguides are measured. Here, the cladding materials used are “OrmoClad” and “OrmoCore” from Micro Resist Technology GmbH, the refractive indices at 850-nm wavelength of which are 1.528 and 1.550, respectively. These materials are used in their pure form or as a mixture with a specified ratio for the cladding.
Table 2. Fabrication conditions of GI-core polymer waveguide.
Here, the NA of the waveguides fabricated using the Mosquito method is affected by the interim time as well as the refractive indices of the core and cladding materials. The interim time is defined as the time right after the core monomer is dispensed to when the UV exposure for curing starts. We already found that the waveguide NA decreases as the interim time increases due to the monomer diffusion. So, the optical-field diameter (OFD) and waveguide NA variation with respect to the interim time are measured using the experimental setup shown in Fig. 10(a). The OFD is defined as the diameter at which the NFP decreases to 1/e2 of its highest intensity, referring to the mode-field diameter (MFD) of single-mode fibers (SMFs).
Fig. 10. (a) Experimental setup for NFP and FFP measurements, (b) measured OFD, and (c) NA of straight GI-core polymer waveguides fabricated in Table 2 with respect to interim time
The light from an LED at 850nm (incoherent light) is coupled to the fabricated polymer waveguide via a high NA SI-core MMF (NA = 0.44, OFD = 55 μm), and the NFP and FFP are measured at the output end of the polymer waveguide.
Figure 10(b) shows the measured OFD, where we find little OFD variation with respect to interim time, probably because the boundary of the core and cladding monomers (OrmoClad and OrmoCore) is less likely to shift even if the monomer diffusion progresses. The larger molecules in the core monomer do not diffuse significantly, and therefore the boundary is stationary. Meanwhile, Fig. 10(c) shows that NA tends to decrease significantly with increased interim time, particularly when the OrmoClad monomer is used in the cladding as indicated by (1), (3), (4), and (5). The smaller size molecules in the OrmoClad monomer can diffuse into the core monomer more rapidly, which means OrmoClad and the core monomer are highly miscible. However, by using the OrmoClad monomer its pure form, a waveguide with sufficiently high NA (> 0.3) is successfully fabricated due to the large index difference, where the interim time should not be too long. Hence, efficient optical coupling between high-speed VCSELs and GI-core waveguides is expected.
5. Experimental results
5.1 Insertion loss of 90°-bent GI-core waveguide
The insertion losses of 90°-bent GI core polymer waveguides with different NAs are experimentally measured using the setup shown in Fig. 11(a). A high NA SMF (NA = 0.28, MFD = 3 μm) is used as a lunch probe fiber for the waveguides by butt-coupling them without using matching oil. Employing this launching condition, the input beam width and NA remain stable during the measurements for multiple samples compared to the direct coupling of the tested waveguides and a VCSEL chip. Meanwhile, a 50-μm GI core MMF (NA = 0.22, OFD = 47 μm) is butt-coupled to the output end of the polymer waveguide, which reproduces the actual MMF links using the waveguide coupler we propose in this paper.
Fig. 11. (a) Experimental setup for insertion loss measurement and (b) measured and calculated NA dependence of insertion loss [21].
The measured insertion losses of the fabricated waveguides are shown in Fig. 11(b), where the marks of 30-GI (Measured) correspond to the experimental results of 30-μm GI-core waveguides. We find the results, particularly at 5-mm bending radius (shown by the red squaremarks), are in good agreement with the calculated results indicated by a solid curve. We confirm from Fig. 11(b) that some fabricated waveguides exhibit an insertion loss close to 2 dB even for 1-mm bending radius when the waveguide NA is higher than 0.3. However, the insertion losses of the waveguides with an NA higher than 0.3 are not always low, and some waveguides show an insertion loss higher than 3.0 dB. Such a high insertion loss with such a large variation could stem from the core index profile being close to an SI. The interim time is kept very short to discourage monomer diffusion when fabricating the high-NA waveguides. Thus, some waveguides with an NA higher than 0.3 would not have GI cores but rather a semi-GI or even SI profile. Therefore, the disagreement between the measured and calculated losses when NA > 0.3 and R = 1 mm (shown by blue marks and solid curve) could also result from insufficient index gradation formation. For low insertion loss, both high NA and GI cores are very important.
5.2 Optical coupling between light source and GI-core waveguide
In the measurement setup shown in Fig. 11(a), the fabricated 90°-bent core polymer waveguide is coupled to a VCSEL source via a high-NA SMF (NA = 0.28) since the output beam from the high-NA SMF could be regarded as a beam from high-speed VCSEL chips as mentioned above. However, as mentioned in Section 3.2, when the waveguides are integrated on the actual VCSEL based optical transceivers, a gap needs to be inserted between the VCSEL and the waveguide core to allow for a bonding wire, as shown in Fig. 12(a). In general, at least 50-μm height is required to the loop of the bonding wire. Thus, the minimum gap size should be 50 μm as well. High coupling loss could be caused by the gap, as calculated in Fig. 7(c). Hence, the insertion losses of 90°-bent GI core waveguides are measured under a direct coupling to a VCSEL chip with a gap for the boding wire, as shown in Fig. 12(a).
Fig. 12. (a) Experimental setup for insertion loss measurement and (b) measured insertion loss of fabricated 90°-bent GI-core waveguides.
The output light from a VCSEL (NA = 0.25, spot size = 6.7 μm, corresponding to VCSEL (5) in Figs. 4 and 5) is coupled to a 90°-bent GI-core (core size = 30 μm). The other end of the GI-core is connected to a 1-m long OM3-MMF to guide the output light to an optical power meter. In this measurement, the gap distance between the VCSEL and the waveguide is set to 100 μm. As shown in the measured results in Fig. 12(b), huge insertion loss as high as 7-8 dB is observed, independent of the waveguide NA, which is caused by input beam divergence during the 100-μm gap propagation. Such a huge loss exceeds the link loss budget allocation estimated in Fig. 2 (indicated by broken line, “Criterion” in Fig. 12(b)).
To address the high coupling loss issue, we propose a simple solution: to fill the gap with a resin. A resin coated on VCSEL chips could generally work as a sealing material, and by adjusting the refractive index of the resin, it could reduce the Fresnel reflection as well. We focus on its ability to reduce the beam NA from the VCSEL in this paper. The measured insertion loss is shown in Fig. 13, where the gap is filled with a matching oil with a refractive index of 1.64. These results indicate that such a loss of as much as 5.0 dB can be saved just by filling the gap with oil. An insertion loss of 3 dB in Fig. 13 can provide a power margin in the loss budget. The insertion loss of the conventional coupling element (mirror and lens) with VCSEL(1) in Table 1 is about 2 dB at the transmitter side and about 0.5 dB at the receiver side [8]. Hence, the measured insertion loss of 90°-bent core waveguide with a high refractive index resin is slightly higher than the conventional one. However, the NA of the fabricated waveguide is optimally designed to couple to a VCSEL with NA = 0.25 without an air gap. Since the effective NA of the VCSEL changes by filling the gap with a resin, as shown in Fig. 13, the loss can further decrease by optimizing the waveguide NA for this effective NA of VCSEL with the resin. Furthermore, the Mosquito method can form waveguide cores directly on a VCSEL chip as shown in Ref. [24]. The direct core formation just on VCSELs enables a gapless coupling. Hence, by optimizing the waveguide NA and even core size, an insertion loss of less than 1 dB can be expected, as shown in Fig. 7. Then, the total insertion loss can be smaller than that of the conventional optical engine. It is noted that the loss reduction in Fig. 13 indicates that the waveguides with lower NA show lower insertion loss when filling the gap with a high index material, although their insertion loss is higher if the gap remains unfilled (shown by red marks). With a matching oil, the input beam NA and beam width are reduced due to a smaller refraction angle into the oil compared to the air, by which higher order modes in the waveguides are launched with less power, resulting in a restricted mode launch condition. Therefore, both bending loss and the coupling loss with GI-MMF with a 0.22 NA decrease. We confirm that filling the gap with a “cured high-index resin” (not uncured oil) should be effective to reduce the insertion loss as well. In this preliminary experiment, we do not optimize the refractive index of the oil or resin. As explained, we need to redesign the core size and NA of the 90°-bent GI core in the waveguide based on the specific launch condition which results from filling the gap with resin. The redesigned results will be published elsewhere. Furthermore, we already reported a proposal in [24] to dispense the core directly above a VCSEL chip in the Mosquito method. By forming the core directly using the Mosquito method, not only can the gap be filled with a resin but also the VCSEL beam can be tightly confined in a specific core to further decrease the coupling loss.
Fig. 13. Measured insertion loss of waveguide before and after filling the air gap with high index oil.
6. Conclusion
For a high bandwidth and density MMF links, VCSEL based optical engines integrating 90°-bent GI circular core polymer waveguide are proposed. To achieve efficient optical coupling, we tried to control the NA of the GI-core waveguide. Since it is important to reveal the characteristics of high-speed VCSELs for the waveguide NA control, the output NA and spot size of the VCSEL beams were measured, and the insertion loss of 90°-bent GI circular core polymer waveguides is calculated from the measured VCSEL characteristics. According to the calculated results, the insertion loss could be reduced by a high-NA (> 0.3) waveguide. By adjusting the fabrication conditions, such as the needle scan path and the refractive indices of the core and cladding materials, high-NA (> 0.3) 90°-bent GI-core waveguides were successfully fabricated. Moreover, we demonstrated low insertion loss (approximately 2.0 dB) in the fabricated high NA waveguides.
In addition, the application of the 90°-bent GI-core waveguide to optical engines is also discussed. A gap of more than 50 μm is required for the coupling between a VCSEL and a waveguide due to a bonding wire. When there is a gap of 100 μm, the insertion loss exceeded 7 dB. Then, in order to decrease the beam divergence and spot size of the VCSEL output beams, we filled the gap with a high refractive index oil. Then, the insertion loss of the waveguide was successfully reduced by 5 dB using an oil with a refractive index of 1.64.
Since the Mosquito method can form the cores on multiple layers in the vertical direction by simply varying the needle tip height when dispensing the core, that is, three dimensional wiring patterns, it can be applied to multi-channel designs by repeatedly stacking 90°-bent GI-cores in the vertical direction. In addition, we are investigating another method for the multi-channel design: in the Mosquito method a GI-core is directly formed on a VCSEL aperture [24] to start dispensing the core monomer on it. By utilizing this technique, not only multi-channel design can be supported, but also gapless optical coupling between the VCSEL and the GI-core is possible.
We are trying to directly assemble the proposed 90°-bent waveguide on a high-speed VCSEL chip to realize subminiature optical engines to transmit a >25-Gbit/s optical signals in MMF links.
Funding
Foundation for Technology Promotion of Electronic Circuit Board, Research Grant.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. T. Norimatsu, T. Kawamoto, K. Kogo, N. Kohmu, F. Yuki, N. Nakajima, T. Muto, J. Nasu, T. Komori, H. Koba, T. Usugi, T. Hokari, T. Kawamata, Y. Ito, S. Umai, M. Tsuge, T. Yamashita, M. Hasegawa, and K. Higeta, “3.3 A 25Gb/s multistandard serial link transceiver for 50dB-loss copper cable in 28 nm CMOS,” in Proceedings of IEEE International Solid-State Circuits Conference, pp.60–61 (2016).
2. M. Pisati, F. D. Bernardinis, P. Pascale, C. Nani, M. Sosio, E. Pozzati, N. Ghittori, F. Magni, M. Garamoazzi, G. Bollati, A. Milani, A. Minuti, F. Giunco, P. Uggetti, I. Fabiano, N. Codega, A. Bosi, N. Carta, D. Pellicone, G. Spelgatti, M. Cutrupi, A. Rossini, R. Massolini, G. Cesura, and I. Bietti, “A sub-250 mW 1-to-56 Gb/s continuous-range PAM-4 42.5 dB IL ADC/DAC-based transceiver in 7 nm FinFET,” in Proceedings of IEEE International Solid-State Circuits Conference, pp.116–117, (2019).
3. F. E. Doany, B. G. Lee, D. M. Kuchta, A. V. Rylyakov, C. Baks, C. Jahnes, F. Libsch, and C. L. Schow, “Terabit/sec VCSEL-based parallel optical module based on holey CMOS transceiver IC,” J. Lightwave Technol. 31(4), 672–680 (2013). [CrossRef]
4. M. Chacinski, N. Chitica, S. Molin, N. Lalic, and O. Sahlén, “25.78 Gbps data transmission with 850 nm multimode VCSEL packaged in QSFP form factor module,” in Proceedings of Optical Fiber Communications Conference 2013, Paper OW1B.1 (2013).
5. Y. Tsunoda, T. Shibasaki, H. Oku, J. Matsui, T. Shiraishi, S. Ide, T. Mori, Y. Koyanagi, and H. Tamura, “24 to 34-Gb/s ×4 multi-rate VCSEL-based optical transceiver with referenceless CDR,” in Proceeding of Optical Fiber Communications Conference 2016, paper Th4D.4, (2016).
6. K. Nagashima, T. Kise, Y. Ishikawa, and H. Nasu, “A record 1-km MMF NRZ 25.78-Gb/s error-free link using a 1060-nm DIC VCSEL,” IEEE Photonics Technol. Lett. 28(4), 418–420 (2016). [CrossRef]
7. N. Kohmu, Y. Matsuoka, T. Takai, N. Chujo, and H. Arimoto, “An ultra-compact 25.78-Gbit/s × 4-ch active optical cable with a high heat-dissipation structure,” in OptoElectronics and Communication Conference, 1-3N-3 (2017).
8. Y. Matsuoka, N. Kohmu, Y. Lee, H. Arimoto, T. Takai, N. Chujo, M. Sato, S. Komatsuzaki, A. Ogura, K. Yamazaki, and Y. Sunaga, “A compact 310-Gb/s optical transceiver for high-density optical interconnects,” IEEE Photonics Technol. Lett. 29(3), 334–337 (2017). [CrossRef]
9. N. Kohmu, T. Takai, N. Chujo, and H. Arimoto, “A 25.78-Gbit/s × 4-ch active optical cable with ultra-compact form factor for high-density optical interconnects,” Appl Sciences 8(1), 137–145 (2018). [CrossRef]
10. K. Nagashima, A. Uemura, Y. Izawa, H. Ishikawa, and Nasu, “A very high-dense on-board optical module realizing >1.3 Tb/s/Inch2,” in Proceeding of IEEE Electronic Components and Technology Conference, 532–537 (2017).
11. T. Aoki, S. Sekiguchi, T. Simoyama, S. Tanaka, M. Nishizawa, N. Hatori, Y. Sobu, A. Sugama, T. Akiyama, H. Muranaka, T. Mori, Y. Chen, S.-H. Jeong, Y. Tanaka, and K. Morito, “Low crosstalk simultaneous 16-channel × 25 Gb/s operation of high density silicon photonics optical transceiver,” in Proceeding of European Conference on Optical Communication, Paper Tu.1.C.3 (2017).
12. H. Nasu, K. Nagashima, T. Uemura, A. Izawa, and Y. Ishikawa, “>1.3-Tb/s VCSEL-based on-board parallel-optical transceiver module for high-density optical interconnects,” J. Lightwave Technol. 36(2), 159–167 (2018). [CrossRef]
13. N. Kohmu, M. Ishii, and T. Ishigure, “A 53-Gbit/s/ch active optical cable utilizing GI polymer waveguide for high-density on-board optical interconnects,” in Proceeding of 2019 IEEE CPMT Symposium Japan, 31–34 (2019). [CrossRef]
14. M. Hikita, S. Tomaru, K. Enbutsu, N. Ooba, R. Yoshida, M. Usai, T. Yoshida, and S. Imamura, “Polymeric optical waveguide films for short-distance optical interconnects,” IEEE J. Sel. Top. Quantum Electron. 5(5), 1237–1242 (1999). [CrossRef]
15. R. Dangel, C. Berger, R. Beyeler, L. Dellmann, M. Gmür, R. Hamelin, F. Horst, T. Lamprecht, T. Morf, S. Oggioni, M. Spreafico, and B. J. Offrein, “Polymer-waveguide-based board-level optical interconnect technology for datacom application,” IEEE Trans. Adv. Packag. 31(4), 759–767 (2008). [CrossRef]
16. N. Bamiedakis, J. Beals, R. V. Penty, I. H. White, J. V. Degroot, and T. V. Clapp, “Cost-effective multimode polymer waveguides for high-speed on-board optical interconnects,” IEEE J. Quantum Electron. 45(4), 415–424 (2009). [CrossRef]
17. H. Nasu, N. Nishimura, Y. Nekado, and T. Uemura, “Polymer waveguide-coupled solderable optical modules for high-density optical interconnects,” IEEE 66th Electronic Components and Technology Conference, 1087–1092 (2016).
18. R. Kinoshita, K. Moriya, K. Choki, and T. Ishigure, “Polymer optical waveguides with GI and W-shaped cores for high-bandwidth-density on-board interconnects,” J. Lightwave Technol. 31(24), 4004–4015 (2013). [CrossRef]
19. Y. Morimoto and T. Ishigure, “Design for polymer optical waveguides realizing efficient light coupling via 45-degree mirrors,” Opt. Express 27(8), 10839–10853 (2019). [CrossRef]
20. N. Kohmu, M. Ishii, and T. Ishigure, “High density electrical and optical assembly for subminiature VCSEL-based optical engine,” IEEE Trans. Compon. Packag. Manufact. Technol. 12(1), 27–36 (2022). [CrossRef]
21. N. Kohmu, M. Ishii, and T. Ishigure, “High-efficiency optical coupling between VCSEL and 90-degree-bent graded-index core polymer waveguide with numeral aperture optimization,” in Proceeding of 2021 IEEE CPMT Symposium Japan, 150–153 (2021).
22. T. Takemoto, Y. Matsuoka, H. Yamashita, Y. Lee, H. Arimoto, M. Kokubo, and T. Ido, “A 50-Gb/s high-sensitivity (−9.2 dBm) low-power (7.9 pJ/bit) optical receiver based on 0.18-μm SiGe BiCMOS technology,” IEEE J. Solid-State Circuits 53(5), 1518–1538 (2018). [CrossRef]
23. K. Soma and T. Ishigure, “Fabrication of a graded-Index circular-core polymer parallel optical waveguide using a microdispenser for a high-density optical printed circuit board,” IEEE J. Sel. Top. Quantum Electron. 19(2), 3600310 (2013). [CrossRef]
24. J. Nakamura and T. Ishigure, “Application of the Mosquito method for implementing waveguides on surface emitting light sources,” in Proceeding of 2018 IEEE CPMT Symposium Japan pp.123–124 (2018).
