Dispersion studies demonstrate that waveguide layout can be used to enhance the bandwidth performance of multimode polymer waveguides for use in board-level optical interconnects, providing >40 GHzxm without the need for any launch conditioning.
Short-reach optical interconnects have attracted significant interest for use in datacommunication links within data centre and supercomputer environments [1]. Due to the ever increasing demand for higher interconnection bandwidth, conventional copper-based electrical interconnects impose a performance bottleneck owing to their inherent disadvantages. Optical technologies provide a promising solution to this issue owing to their advantages over their electrical counterparts: higher bandwidth, lower crosstalk and improved power efficiency [2]. Multimode polymer waveguides are considered to be an excellent candidate for use in board-level optical interconnects enabling the formation of electro-optic (EO) printed circuit boards (PCBs). Siloxane polymer materials in particular, enable the direct integration of such waveguides onto standard PCBs owing to their good thermal and mechanical properties [3]; while the large core dimensions employed (typically 30 to 70 μm) offer relaxed alignment tolerances, thus enabling low-cost system assembly with pick-and-place tools [4].
However, the highly-multimoded nature of these waveguides has raised important questions concerning their bandwidth performance and their ability to support very high on-board data rates (>40 Gb/s). We have recently presented bandwidth studies on a 1 m long polymer multimode spiral waveguide using time domain measurements under various launch conditions [5], and demonstrated the potential of transmitting very high data rates (>100 Gb/s) over a single waveguide channel using refractive index engineering and launch conditioning [6]. In this paper, we present a novel approach to achieve bandwidth enhancement in such waveguides by using waveguide layout and exploiting the mode-filtering properties of passive waveguide components such as waveguide bends and crossings. Mode-filtering schemes have been used in multimode fibre systems [7] and mode-selective ring resonators [8] and couplers [9] for few-moded systems to increase data capacity. A similar approach is proposed here in the context of highlymultimoded board-level optical interconnects using simple passive waveguide components such as bends and crossings, providing a useful tool in the design of optical backplanes. Dispersion studies carried out on the waveguide bends and crossings show that the suppression of higherorder modes in these waveguides components results in a bandwidth enhancement up to ~1.5 times over similar straight waveguides and indicate that intelligent waveguide layout, together with the appropriate choice of refractive index profile, can ensure bandwidth-length product (BLP) >40 GHz×m to support very high on-board data rates while maintaining low loss performance and without the need for any launch conditioning at the waveguide input. The results highlight the potential of this technology for use in high-performance board-level optical interconnects. Dispersion measurements are carried on these waveguide components using a short pulse laser source and an autocorrelator. The experimental setups used are shown in Fig. 3
A femtosecond erbium-doped fibre laser operating at ~1574 nm (TOPTICA FFS) is utilised as the light source and a frequency-doubling crystal (MSHG1550-0.5-1) is employed to generate second harmonic pulses at wavelength of ~787 nm. The output light is coupled into a cleaved short 50/125 μm MMF patchcord via a 10× microscope objective (NA = 0.25) while the other end of the 50/125 um MMF patchcord is cleaved and mounted on a translation stage to enable butt-coupling into the waveguides. A near field image and the far field intensity of the 50/125 μm MMF input are shown in Fig. 3(c) and 3(d) respectively. The launch employed in the measurements is neither highly restricted or overfilled corresponding therefore, to a relatively realistic launch condition that could be encountered in a real-world system. The output light from the waveguide is collected with a 16× microscope objective (NA = 0.32) and is coupled to an autocorrelator to detect the transmitted optical pulses. The width of the received signal pulses are estimated from the autocorrelation traces using curve fitting with common pulse shapes (i.e. sech 2 or Lorentzian). The frequency response of the waveguide components is extracted by taking the Fourier transform of the back-to-back and waveguide links and subtracting one from the other. As a result, the -3 dB bandwidth of the waveguide components can be determined. A similar setup is used for the measurement of the loss performance of the components. The loss measurements are conducted at 850 nm using a multimode VCSEL source. A pair of microscope objectives is employed to couple the light into the short 50/125 um MMF patchcord. The cleaved end of the input fibre is butt-coupled to the waveguide input facet while at the waveguide output, a 16× microscope objective is used again to collect the output light and focus it onto an optical power meter head (HP
This content is AI-processed based on open access ArXiv data.
