Open Access
Add to BibSonomy
Abstract
In this paper, we propose a low-loss assembly of optical elements [vertical cavity surface emitting laser (VCSEL)/photodiode (PD)] and a 90°-bent graded-index (GI) core polymer optical waveguide for compact high-speed optical transceivers. The air gap between the 90°-bent GI-core waveguide and VCSEL/PD chips are filled with a high-index resin, and we demonstrate a reduction not only of the Fresnel reflection loss but also of the coupling loss due to the lower divergence angle of VCSEL and fiber output beams. In this paper, the core size and NA of the 90°-bent GI-core in the polymer waveguide are redesigned to match the specific launch conditions realized with the gap filled with resins, resulting in theoretical estimates of insertion losses as low as 0.82 dB and 1.38 dB for the waveguides on the transmitter and receiver sides, respectively. Such low insertion losses are obtained not by step-index core but GI core waveguides. After redesigning the structure, 90°-bent GI-core polymer waveguides with a bending radius of 1 mm are experimentally fabricated using the Mosquito method, and we experimentally demonstrate that the insertion loss of the waveguide can be reduced by about 4 dB when the 100-µm air gap is filled with an optical adhesive (na = 1.51). In addition, an error-free 25.78-Gb/s data transmission is successfully demonstrated over a 100-m MMF link using the fabricated VCSEL based transceivers with 90°-bent GI core polymer optical waveguides.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
The amount of data traffic through the Internet keeps increasing as the Internet of Things (IoT), deep learning, and other cloud computing services grow. Hyper-scale data centers (DCs) and high-performance computers (HPCs) are required to collect and analyze huge amount of digital data from IoT devices and cameras to provide smart services. As more data is expected to be uploaded, the I/O throughput in DCs and HPCs will increase further. It is difficult to achieve high-speed and high-density data transmission with conventional electrical interconnects due to huge transmission loss and crosstalk [1,2] even in such short reach networks. Contrastingly, optical interconnects can achieve high-speed and high-density signal transmission. Thus, optical interconnects are commonly used in rack-to-rack datalinks in DCs and HPCs. Currently, high-speed and high-density optical transceivers for on-board or co-packaging on high-speed carriers have been proposed and developed [3–13].
When optical transceivers compliant to conventional small form factors are intended to be applied for on-board optical integration, a critical issue is how to couple optical devices and optical fibers with high density and low loss. In current optical transceivers, optical connectors with lenses integrated in them are used to connect optical fibers with a channel pitch of 250 µm. Lens-integrated optical connectors require many components. To couple an output beam from a high-speed vertical cavity surface emitting laser (VCSEL) with a large beam divergence angle, it has been necessary to use multiple lenses to focus the beam on a small spot. Here, VCSELs are integrated on a board surface to emit the beam vertically to the board, while the fibers are normally arranged parallel to the board. Thus, the VCSEL beam direction needs to be perpendicularly bent to couple to the fibers, because of which a 45° mirror is also integrated with lenses in the connectors. Therefore, it has been difficult to achieve high-density and high-efficiency optical coupling. Optical coupling techniques using 90°-bent fibers or polymer optical waveguides have been proposed for perpendicular light path conversion [14,15]. When optical fibers are applied, small-diameter bending would be possible, but the radiation loss could increase at the bending point, while the fatigue due to a long-term steep bending is a concern. Contrastingly, when polymer optical waveguides are applied, there is a possibility to form multiple cores by applying lithography, then vertical optical path conversion is just achieved by a perpendicular reflection on a 45° mirror formed at the waveguide edge. Hence, excess losses could be caused due to the incomplete reflection and mode-field mismatch compared to that in graded-index circular cores in multimode fibers (GI-MMF). Therefore, we have proposed a highly efficient and compact lens-less optical coupling between a VCSEL and MMF using a circular graded-index (GI) core polymer waveguide.
We reported that circular GI-core polymer waveguides can achieve low-loss optical coupling compared to conventional square SI-core optical waveguides [16,17]. Moreover, by optimizing the numeral aperture (NA) and core diameter of the waveguides that are fabricated using the Mosquito method, we experimentally demonstrated that a low insertion loss of about 2 dB can be achieved even with a bending radius as small as 1 mm [18]. Here, when the optical waveguides are integrated in VCSEL-based optical transceivers, an air gap is required between the VCSEL chip and the waveguide core to reserve space for the bonding wires that connect the VCSEL to the printed circuit board (PCB). In general, as the loop height of the bonding wire must be at least 50 µm, the minimum gap size would also be 50 µm. We showed in [18] that the insertion loss increases to as high as 7-8 dB with a gap of 100 µm. To address this issue, we proposed a simple method to fill the gap with a resin [18]. However, the refractive index of the resin filling the gap is not optimized. In general, such a resin or gel is inserted to reduce the Fresnel reflection, but we found that the resin filled in the gap allowed to decrease the divergence angle of emitting beam from VCSELs. Therefore, the core size and NA of the 90°-bent GI-core polymer waveguide that receive the VCSEL beam need to be redesigned based on such a specific launch condition.
In this paper, we propose an optimal structure of 90°-bent GI-circular-core polymer waveguide to couple with VCSEL and PD chips when filling their gaps by a resin. Then, we fabricate an optical transceiver with a 90°-bent GI core waveguide. The optimum structure with resin filling and the effect of resin insertion on the coupling loss are discussed. Then, evaluation boards are fabricated to use in optical signal transmission experiments where 90°-bent core polymer waveguides are integrated on both transmitter (Tx) and receiver (Rx) sides. Finally, we experimentally verify an error-free optical transmission of 25.78-Gb/s over a 100-m MMF (OM-3) link is possible.
2. Basic structure of proposed optical transceiver
Figure 1 shows a side view cross-section of the proposed optical assembly compared to a conventional one [18]. The conventional optical assembly, shown in Fig. 1(B)(b), employs an optical connector in which many optical components are required. For example, multiple lenses need to be integrated and accurately aligned in the optical connector to focus the widely diverged output beam from the VCSEL on the fiber core in the Tx side. The divergence angle of the beam tends to increase with an increased VCSEL modulation speed. Meanwhile, for the Rx side, focusing the output beam from an MMF on a small spot is required as well, because the active area of photodetectors needs to be small (∼20 µm) for high-speed signal detection. However, it could be difficult to realize a high-density optical assembly with the existing devices composed of many components.
Fig. 1. Side view cross-sections of optical engines with (A) 90°-bent GI core waveguide grouper with resin and (B) waveguide coupler with air gap [18].
We have already proposed an optical transceiver integrating a 90°-bent GI-core polymer waveguide, as shown in Fig. 1(B)(a) [18]. The optical transceiver consists of a VCSEL array, a PD array, and ICs (laser diode driver (LDD) and transimpedance amplifier (TIA)). In our previous publication [18], an air gap between the VCSEL chip and the waveguide cores existed, which increased the insertion loss of the waveguide. Therefore, we also proposed a simple method to fill this gap with a resin [18]. The resin coated on the VCSEL chip not only works as a sealing agent, but also reduces the emitting beam NA by adjusting the refractive index of the resin. Here, the resin material should exhibit low optical loss comparable to conventional matching oil or optical adhesive.
3. 90°-bent GI-core waveguide optimization with high index resin
We already found in [18] that the optimized structure (Tx side) for a 90°-bent core polymer waveguide with a bending radius of 1 mm is to have a 30-µm diameter GI circular core with an NA of 0.38 when a resin fills the air gap. However, even with resin filling, the spot size increases as the beam propagates across the gap. The final beam spot size when it reaches the waveguide input end is largely influenced by the emitting beam NA of the VCSEL and the refractive index of the resin. Hence, further optimization of the waveguide structure is required, depending on the refractive index of the resin filling the gap.
Figure 2(a) shows the simulation model for the insertion loss of 90°-bent waveguides using the ray trace analysis software, Zemax OpticStudio. Figure 2(b) shows the calculated results for the insertion loss with respect to the bending radius R over a range from 1 mm to 3 mm. Here, the gap size is fixed to 100 µm, while the core diameter d of the waveguide is varied as 30, 50, and 70 µm, and the refractive index of the filled resin is varied as na = 1.4, 1.6, and 1.8. For the simulation, the effective NA (NAeff) of the GI-core is defined by Eq. (1):
where ncore and nclad are the refractive indices of the core and cladding, respectively. In this simulation, the refractive index of the core is set to 1.585 assuming an organic-inorganic hybrid resin for the core [18], while the refractive index of the cladding is varied from 1.481 to 1.571 to investigate the waveguide NA’s effect on loss. It is indicated in Fig. 2(b) that the insertion loss tends to be lower as the refractive index of the resin increases. When the bending radius R = 1.5 mm and smaller, the core diameter d should be 50 µm to obtain the lowest loss, while when the bending radius is relaxed to 2.0 mm or larger, the loss of 70-µm core is rather lower than the loss of 50-µm core. This is because high-bandwidth VCSELs tend to emit a beam with a NA of 0.25 or higher (a wider divergence angle), and thus, a larger core size is required to accept the enlarged beam after propagating in the resin to result in a lower coupling loss between the VCSEL and waveguide. Here, a resin with a refractive index of 1.6 or higher is assumed to be used. Then, an NA of 0.25 and higher is required for the waveguide as well as the largest core size (70 µm) to achieve an insertion loss of 1 dB less than under a bending radius of 2.0 mm, as shown in Fig. 2(b). In addition, when the bending radius is 3.0 mm, the waveguide NA at which the lowest loss is observed is relaxed to approximately 0.21, because the bending loss does not significantly affect the total loss. Meanwhile, under a bending radius R of 1.5 mm, a waveguide NA of 0.3 allows an insertion loss of less than 1 dB, but a waveguide NA higher than 0.3 increases the insertion loss due to the coupling loss increase with MMF, which is attributed to the NA mismatch between the waveguide and connected MMF.Fig. 2. (a) Simulation model for 90°-bent GI-core polymer waveguide with resin at Tx side and (b) calculated insertion loss.
Figure 3(a) shows the breakdown of the calculated insertion loss when R = 1 mm, d = 50 µm and na = 1.8, compared to the similar breakdowns of the losses of waveguides with R = 1.5 mm, d = 50 µm and na = 1.8 in Fig. 3(b), and R = 2.0, 3.0 mm, d = 70 µm and na = 1.8 in Fig. 3(c) and (d), respectively. These three waveguides have the optimum structure showing the lowest insertion loss under three different bending radii. The results in Fig. 3 indicate that by increasing the waveguide NA, the bending and coupling losses at the VCSEL side decrease, while the coupling loss at the MMF side increases due to the NA mismatch. In all the figures in Fig. 3, the bending loss curves show a steep increase below a lower limit waveguide NA, which varies from 0.2 to 0.3 depending on the bending radius. In the case of a 1-mm bend radius, the bending loss is as high as 2 dB even at its lowest waveguide NA of 0.3, which decreases slightly when the waveguide NA increases to 0.4, exhibiting the lowest insertion loss of 1.2 dB (blue curve). Meanwhile, the lowest insertion losses at R = 2 and 3 mm are 0.5-0.8 dB. Remarkable total loss reduction is not necessarily observed even if the bending radius is relaxed and the waveguide structure is optimized for the relaxed bending radii. Therefore, since both a compact structure and an insertion loss lower than 1 dB are realized with a moderate waveguide NA (∼0.3), we can reach the conclusion that R = 1.5 mm should be the best condition for the Tx side.
Fig. 3. Breakdown of insertion loss (a) R = 1.0 mm, 50-GI, na = 1.8, (b) R = 1.5 mm, 50-GI, na = 1.8, (c) R = 2.0 mm, 70-GI, na = 1.8 and (d) R = 3.0 mm, 70-GI, na = 1.8.
On the other hand, the optimal waveguide structure on the Rx side is also calculated using a 90°-bent GI-core with bending radii of 1, 1.5, and 2.0 mm. In the model, an output beam from a high-speed VCSEL (NA = 0.25, aperture size = 6.7 µm) is directly coupled to a 0.05-m long GI-MMF, and its output beam is coupled to the 90°-bent GI core waveguide. Here, the variations of waveguide structure and the conditions of the resin inserted between the waveguide and PD remain the same as those for the waveguide on the Tx side. Figure 4(b) shows the calculated insertion loss. We find in Fig. 4(b) that the insertion loss is reduced as the refractive index of the resin increases in the same manner as the Tx side. This is due to the decrease in the divergence angle of the output beam from the waveguide core by the resin. In addition, the lowest insertion loss is observed when the core diameter is not 70 µm but 50 µm, which is independent of the bending radius. Here, a large waveguide core diameter allows a restricted mode launch condition at the coupling with the GI MMF, which could lead to low bending loss as well as low coupling loss. Contrastingly, the large core exhibits size mismatch with such a small active area of PD as 20 µm, resulting in high coupling loss with the PD. The increment of coupling loss with PD is more significant than the improvement in coupling loss with MMF and bending loss and thus, the 50-µm core waveguide exhibits the lowest loss. Meanwhile, although high waveguide NA contributes to reduce the bending loss, the coupling loss with PD increases because of the output beam expansion from the waveguide during the gap propagation. Hence, we find from Fig. 4(b) that R = 1.5 mm, a core diameter d = 50 µm, NAeff = 0.32, and na = 1.8 are the optimum structure of the 90°-bent GI-core polymer waveguide on the Rx side to achieve 2-dB insertion loss. When the bending radius is 2.0 mm, the lowest insertion loss is as low as 1.1 dB when the waveguide NA is 0.31 showing little difference with the case when R = 1.5 mm.
Fig. 4. (a) Simulation model for 90°-bent GI-core polymer waveguide with resin at Rx side and (b) calculated insertion loss.
Figures 5(a)-(c) show the calculated insertion losses of the 90°-bent GI core waveguides for Tx, Rx sides, and their total when R = 1.0, 1.5, and 2.0 mm, respectively with respect to the refractive index of the resin filling the gap. The dotted and dashed lines in Fig. 5 are the simulated insertion losses assuming a high-speed VCSEL (emitting a beam with a wide divergence angle) and a PD (with an active area of less than 20 µm) are connected with an MMF by the optical connector shown in Fig. 1(B)(b). As indicated by blue and red lines in Fig. 5(a), by optimizing the refractive index of the resin and the waveguide structure (core diameter and NA), the insertion loss can be lowered to 1.23 dB and 1.65 dB for the Tx and Rx sides, respectively even when the bending radius is as small as 1 mm. If the bending radius is slightly relaxed to 1.5 mm, the lowest insertion losses are 0.66 dB and 1.20 dB, respectively. Since the refractive index of commercially available optical adhesive is generally lower than 1.7, the realistic lowest insertion losses are estimated as 0.82 dB and 1.38 dB for Tx and Rx sides from Fig. 5(b) when na = 1.7, resulting in the total insertion loss of 2.20 dB. This is 3.8 dB lower than the total loss (6 dB indicated by green broken line) of the lens and mirror connectors: the loss improvements in the Tx and Rx sides are 1.68 dB and 2.12 dB, respectively. It is noted that the loss improvement is as small as 0.2 dB even if R is relaxed from 1.5 to 2 mm, while an 0.8-dB improvement is observed from R = 1.0 to 1.5 mm. Therefore, a bending radius of 1.5 mm should be selected, by which the 90°-bent GI core optical waveguide is expected to realize high-efficiency and high-density optical packaging.
Fig. 5. Calculated insertion loss with respect to refractive index of resin, na (a) R = 1.0 mm, (b) R = 1.5 mm and (c) R = 2.0 mm (Green broken lines: total loss, blue long and short dashed line: Tx side insertion loss, and orange long and short dashed line: Rx side insertion loss)
4. Loss-budget of 25.78-Gb/s optical link
As confirmed in the above section, filling the air gap with a high refractive-index resin, and then optimizing the waveguide structure corresponding to the specific refractive index of that resin reduces the insertion loss of the waveguide. Meanwhile, since the resin filling can change the boundary conditions on the emitting surface of VCSELs, the operating conditions of the VCSEL could also change, affecting the output beam characteristics. Therefore, to investigate the effect of resin insertion, the laser output characteristics (the I-L curve) are measured by dropping a matching oil on the emitting surface of the VCSEL.
Figure 6(a) shows the measurement setup for the I-L curve of a VCSEL. The sensor head of an optical power meter is placed 1-cm above the VCSEL aperture. Here, a droplet of matching oil covers the VCSEL surface, as shown in Fig. 6(a). The measured I-L curves are shown in Fig. 6(b), which indicates that the matching oil reduces the output power of VCSEL about 13% corresponding to a 0.6-dB power attenuation under 7-mA bias current, when the matching oil has a refractive index of 1.50 and higher. This could be attributed to the change in the reflection coefficient of the optical cavity of the VCSEL due to the replacement of air by a matching oil. When the refractive index of the resin exceeds 1.64, the I-L curves tend to be almost identical. Here, as shown in Fig. 5, the total insertion losses are 3.08 dB and 2.20 dB under the optimized waveguide structures for R = 1 mm and 1.5 mm, respectively. Therefore, the effect of VCSEL output power reduction due to resin insertion is small enough and thus, higher-speed and longer-distance optical signal transmission is expected by realizing 90°-bent GI-core structure using the Mosquito method.
The loss budget of a 25.78 Gb/s multimode optical link was calculated in our previous study [18]. The loss reduction by the resin injection is expected to improve the link margin compared to an optical link with a conventional lens and mirror structure. The results in Fig. 7(a) show the loss budget calculated using the 90°-bent optical waveguide with a bending radius of 1.5 mm and a resin with a refractive index of 1.7 in Fig. 5(b). For comparison, the results for the optical transceiver with conventional lens and mirror structure is also indicated. The optical power of the VCSEL decreases about 0.8 dB due to the resin insertion, as indicated at (1) in Fig. 7. Meanwhile, as indicated at (2) and (4) in Fig. 7, we can reduce the coupling losses about 0.82 dB (conventional structure: 2.50 dB) and even about 1.38 dB (conventional structure: 3.50 dB) for the Tx and Rx sides, respectively, by substituting the mirror and lenses for the 90°-bent core waveguide with the optimum structure. Figure 7(b) shows the link margin dependence on the bending radius. We can preserve a link margin as high as 3 dB, which is much larger than a 0.5-dB link margin for the conventional optical connector integrated transceiver [7]. Based on these results, a high-speed optical link with a sufficient link margin is expected by inserting a high refractive index resin between a VCSEL and an optimized waveguide fabricated using the Mosquito method.
5. Experimental results
5.1 Insertion loss of the 90°-bent GI-core waveguide
Since we already confirmed the loss reduction of 90°-bent GI-core polymer waveguides with a resin filling the air gap with a VCSEL on the Tx side in [18], in this paper we evaluate the insertion loss of the 90°-bent GI-core waveguide used in the Rx side. An active alignment is utilized for the optical axis adjustment between the waveguide core and PD to observe the highest photocurrent measured by the TIA when resins with different refractive indices are inserted between the waveguide and PD. The bend radius is set to 1 mm as a severe condition compared to the optimum while a gap of approximately 100 µm exists between the PD, as shown in Fig. 8(a). Meanwhile, the optical-field diameter (OFD) of the waveguide is measured to be 22 µm, slightly smaller than the designed core diameter (30 µm). Meanwhile, the waveguide NA is measured to be from 0.31 to 0.36. The OFD is defined as the diameter at which the near field pattern decreases to 1/e2 of its highest intensity under over-filled mode launch condition [18], referring to the mode-field diameter of single-mode fibers. Figure 8(b) shows the measured results. Here, the insertion loss is calculated from the photocurrent received by the TIA and the photosensitivity of the PD (0.45 A/W). When the air gap remains, a coupling loss as high as 7.5 - 8.5 dB is observed, independent of the waveguide NA. This is because the high NA waveguide that exhibits low bending loss emits an output beam with a large divergence angle. The beam spot size tends to expand while propagating the air gap, and exceeds the active area (20 µm) of the high-speed PD when it reaches the PD. On the other hand, when the gap is filled with a matching oil with a refractive index of 1.52 (not high enough), an insertion loss is reduced by as large as 4 dB. Calculated insertion losses with and without matching oil are also indicated in Fig. 8(b), which are underestimating the losses in both cases, but nevertheless indicate a remarkable loss (∼3 dB). The larger disagreement between measured and calculated results of the waveguide with matching oil can be attributed to the underestimated waveguide NA from its far-field pattern measurement, as such a strong waveguide NA dependence is confirmed in Fig. 4(b). We both experimentally and theoretically confirm that the index adjustment of the air gap not by a resin but even by simply a matching oil contributes to the loss reduction in the receiver side, as well. By filling the air gap between the optical devices and GI-core with a resin whose refractive index is optimized, the insertion loss is expected to be sufficiently reduced, which allows us to design a 25.78-Gbit/s 100-m MMF link.
Fig. 8. (a) Experimental setup for insertion loss measurement at Rx side and (b) measured insertion loss of fabricated 90°-bent GI-core waveguides.
5.2 25.78-Gb/s optical signal transmission
A 25.78-Gbit/s data transmission over a 100-m MMF link is experimentally demonstrated using the fabricated optical transceiver on which an optimized 90°-bent GI-core polymer waveguide is bonded by a resin (adhesive) with a refractive index na of 1.51 (NOA65, Norland Products, Inc.). Figure 9 shows photographs of the fabricated evaluation board for the transceiver, where Fig. 9(a) shows a top-view and Fig. 9(b) shows zoomed-in images of the optical assembly area. Arrays of GaAs-VCSELs in 850-nm band and PDs, SiGe-LDDs, and TIAs are integrated, and a clock and data recovery (CDR) circuit is also externally mounted for both Tx and Rx sides, as shown in Fig. 9(a).
Fig. 9. Top view of photograph of fabricated evaluation board (a) whole substrate and (b) optical assembly area.
The evaluation board is operated while placed on a Peltier element to keep the temperature of the optical devices constant. Here, for the I-L characteristics in Fig. 6(b), computational simulations show that the threshold current of the VCSEL varies by several hundred µA due to the resin insertion. Since even such a small variation of the threshold current could affect the VCSEL modulation particularly under a low bias current, the bias and modulation currents are adjusted to keep the operating current higher than 2 mA. The data rate, amplitude, and pattern length of the input electrical signal are set to 25.78 Gbit/s, 500 mVppd, and 231-1 PRBS, respectively. The bias and modulation currents applied to a VCSEL are set to 7.0 mA and 9.1 mAp-p, respectively. The 90°-bent GI core waveguide should ideally be fabricated directly on the VCSEL and PD chips as reported in [19], by which we can realize a gapless connection with the VCSEL / PD and waveguide core. However, in order to confirm the credibility of our designs for the waveguide structure including the gap with resin, in this paper, we independently fabricate 90°-bent GI core waveguides with the optimally designed structures [12]. An active alignment is utilized for the alignment between the optical devices (VCSEL and PD) and the GI core in the same way as that for the optical coupling loss measurement. After adjusting the position of the waveguide on the Tx side to maximize the optical power coupled from the VCSEL, the position of the waveguide on the Rx side is adjusted to maximize the photocurrent of the PD. They are connected to VCSEL and PD chips by filling their gaps with the adhesive, as shown by the photographs in Fig. 10. The bending radius of the GI core is set to 1 mm for both Tx and Rx sides as the most severe case, and the gaps between the waveguide core are filled with the adhesive, but the properties are compared before and after it is cured. Before being cured, it has a liquid state with a refractive index lower than 1.51.
Fig. 10. Photograph of optical transceiver on which a 90° bent GI core polymer waveguide is bonded by optical adhesive.
The optical waveforms after transmission through 3-m and 100-m long loop-back MMF links are shown in Fig. 11(a), compared with the received electrical waveforms with CDR at the Rx side for both links. The measured optical and electrical waveforms show clear eye openings even after 100-m MMF transmission. In Fig. 11(a), the waveforms from an optical transceiver without resin insertion are also shown for comparison. We find from Fig. 11(a) that the resin insertion dramatically improves the optical eye opening due to the loss reduction as mentioned above. The extinction ratios after 3-m and 100-m transmissions calculated from the eye waveforms are 4.45 dB and 4.17 dB, respectively, which indicate that high-quality signal transmission can be expected. Figure 11(b) shows the measured BER characteristics for 3-m and 100-m long MMF links. The results show that 25.78-Gbit/s error-free (BER < 10−12) transmissions are successfully achieved.
Fig. 11. (a) Measured waveforms and (b) measured BER characteristics of MMF-links composed of the proposed optical transceiver.
Here, curing the adhesive should stabilize the transceiver operation and maintain the quality of the optical assembly, while the volume contraction resulting from curing could be a concern because it could stress the connected optical devices (VCSEL, PD, and waveguide). Therefore, the uncured adhesives in both the Tx and Rx sides are UV-cured completely, and the MMF link performances are evaluated to investigate the effect of adhesive curing. Figure 12 shows the measured optical waveform and BER curves. There is no significant change in the eye waveforms before and after curing, where we find a 0.07-dB and a 0.3-dB improvement in the extinction ratio and minimum received optical power. The results indicate that the signal integrity is improved by resin curing. This is attributed to the increment of the refractive index of the adhesive after curing to further reduce the coupling losses between the waveguide and optical devices, resulting in noise reduction. Thus, we find the link power budget shown in Fig. 7 is an underestimate, and referring to the minimum received power obtained in Fig. 12, we can expect 1.7 dB (10.5 - 9.0 dB) of additional link margin and therefore 5.2 dB in total link margin even for 100-m MMF links. Such a sufficient link margin could be allotted to the coupling loss at both ends of the MMF with the waveguide caused by the Fresnel reflection and slight misalignment, which are not indicated in this budget estimation.
Fig. 12. Measured optical waveforms and BER curves to compare before and after filling the resin curing.
In this experiment, the polymer optical waveguides are fabricated by scanning a needle on a horizontal plane in the cladding monomer as designed in [12] to apply the waveguide to an optical transceiver and then the transmission characteristics of the 90°-bent circular GI-core polymer waveguides are evaluated. Meanwhile, the Mosquito method has a potential to form vertically 90°-bent cores by scanning the needle on a vertical plane. We succeeded in fabricating the vertically 90°-bent core waveguide with a bending radius smaller than 1 mm [19] to connect optical devices (VCSEL/PD) and MMF. If this needle-scan path is applied, multiple 90° bent cores can be formed in one waveguide. Furthermore, by scanning the needle in three dimensionally, even pitch converting structures from narrow-pitch (less than 100 µm) to 250 µm by 90°-bent cores can be formed. Such a waveguide allows ultra-high-density assembly required in optical transceivers for CPO (Co-Packaged Optics), which will be published elsewhere.
6. Conclusion
To realize high-bandwidth and high-density MMF links, we proposed an optical transceiver in which the air gap between a 90°-bent GI circular core waveguide and optical elements is filled with a high-refractive index resin. To achieve highly efficient optical coupling, we tried to optimize the design of the GI core waveguide, including the refractive index of the filling resin; Tx and Rx waveguide structures were designed and theoretically demonstrated to obtain insertion losses as low as 0.82 dB and 1.38 dB on the Tx and Rx side, respectively. The insertion loss on the Rx side was also experimentally measured using the fabricated 90°-bent GI-core waveguide with a bending radius of 1 mm, and we demonstrated that the matching oil (na = 1.51) injection improved the coupling loss by approximately 4 dB.
The application of 90°-bent GI-core polymer waveguides to optical transceivers was also discussed. A 90°-bent GI-core waveguide was applied to the fabricated optical transceivers on the evaluation board, and the eye opening was improved by filling the air gap with an adhesive. A 25.78 Gb/s data transmission over a 100-m MMF link was successfully demonstrated by the fabricated transceiver, for which error-free optical transmission was confirmed. Finally, curing the optical adhesive did not affect the BER characteristics maintaining the error-free transmission while increasing the physical stability of the connectors.
In the future, we will try to develop a gapless coupling method for even lower loss and to support > 50 Gb/s.
Funding
The 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. M. Pisati, F. D. Bernardinis, P. Pascale, et al., “A sub-250 mW 1-to-56Gb/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).
2. T. Norimatsu, K. Kogo, T. Komori, N. Kohmu, F. Yuki, and T. Kawamoto, “A 100-Gbps 4-lane transceiver for 47-dB loss copper cable in 28-nm CMOS,” IEEE Trans. Circuits Syst. I 67(10), 3433–3443 (2020). [CrossRef]
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 Photon. 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 Photon. 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. Sci. 8(1), 137–145 (2018). [CrossRef]
10. K. Nagashima, Toshinori Uemura, A. Izawa, Y. Ishikawa, and H. 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, pp.31–34 (2019).
14. H. Nasu, “Short-reach optical interconnects employing high-density parallel-optical, modules,” IEEE J. Select. Topics Quantum Electron. 16(5), 1337–1346 (2010). [CrossRef]
15. C.-T. Chen, P.-K. Shen, C.-C. Chang, H.-L. Hsiao, J.-Y. Li, K. Liang, T.-Y. Huang, R.-H. Chen, G.-F. Lu, and M.-L. Wu, “45°-mirror terminated polymer waveguides on silicon substrates,” IEEE Photon. Technol. Lett. 25(2), 151–154 (2013). [CrossRef]
16. 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]
17. 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]
18. N. Kohmu, M. Ishii, R. Hatai, and T. Ishigure, “90°-bent graded-index core polymer waveguide for a high-bandwidth-density VCSEL-based optical engine,” Opt. Express 30(3), 4351–4364 (2022). [CrossRef]
19. D. Suemori, M. Ishii, N. Kohmu, and T. Ishigure, “90-degree bent core polymer optical waveguide coupler for low loss lens-less light coupling between laser / photodetector and fiber,” in Proceeding of 2022 IEEE 72nd Electronic Components and Technology Conference, pp.436–440 (2022).
