2.1.

Angle Multiplexing

Patented AM can significantly extend the path length between multiple free-space apertures, resulting in a longer distance a wave travels, and a longer period of storage for storage in motion applications. At a given data rate, this increases the amount of data that can be stored by the loop. A schematic of one implementation of AM is shown in Fig. 3. In this system, $M$ apertures in a plane (top side of Fig. 3) are facing $N$ apertures of another plane (bottom side of Fig. 3). Starting at the bottom left of the $N$ aperture plane, the laser beam travels to the aperture on the bottom left of the $M$ aperture plane and is reflected to another aperture in the $N$ aperture plane. It is then steered to another aperture of the $M$ aperture plane that sends it back to the initial aperture located at the bottom left of the $N$ aperture plane. At this location, light is sent to the next aperture in the upper $M$ aperture plane. By repeating this process, light travels to all the paths between all $M$ apertures of the top plane and two of the $N$ apertures of the lower plane. When the last aperture of the upper $M$ aperture plane is reached (located on the top-right), light is sent to a third aperture in the bottom $N$ aperture plane. The process is then repeated until the laser beam has traveled all possible paths between the $N$ apertures of the bottom plane and the $M$ apertures of the top plane. One can readily show that there are $M\times N$ paths in total. Hence, the path length was increased from $\min (M,N)\times L$ to $M\times N\times L$, where $L$ is the distance between the two planes. The above is an example path. We have explored and optimized the hop order for AM.

Fig. 3

Principle of AM between two planes of $M$ apertures on the top side and $N$ aperture on the bottom side. The red line represents the laser light travelling through the AM systems. For clarity, all the paths between 2 of the $N$ aperture side and all of the $M$ aperture side are represented.

Our initial implementation of the AM concept used a combination of mirrors and volume holographic gratings (VHGs). A light ray is transmitted by a VHG unless it comes in at certain angle and wavelength, in which case it is reflected. Multiple angles can be embedded in a single VHG to steer the light in different directions according to the incoming angle of incidence. VHGs are diffractive elements consisting of a periodic phase or absorption perturbation, throughout the entire volume of the element. When a beam of incident light satisfies the Bragg phase-matching condition, it is diffracted by the periodic perturbation. Volume holograms were first treated by H. Kogelnik in 1969 by the so-called coupled-wave theory.⁷ For volume phase holograms, it is possible to diffract 100% of the incoming reference light into the signal wave, i.e., full diffraction of light can be achieved. Moreover, by adequately designing the thickness of the VHG, low angular selectivity in the milliradian range can be obtained. However, VHGs possess a high dispersion: for a broadband signal with wavelengths spanning more than 100 nm, it is necessary to employ dispersion-compensating components such as thin gratings. We have considered methods of making VHGs broadband enough for our application. Other approaches to design high angular selectivity apertures using low dispersive approaches for broadband operation are also currently pursued at LyteLoop. It will be easier to implement a broadband AM system if we can make it all reflective. Later we show a limited all reflective data storage system using AM.

2.2.

Broadband Wavelength and Spatial Mode Division Multiplexing

In LyteLoop fiber, using common fiber, we are limited to the near infrared (NIR) with wavelengths from 975 to 1800 nm, we have shown that it is possible to reach a storage capacity from 2 to 10 petabytes using a fiber system of 10,000 km 50 cores, 23 modes per core.⁸ In LyteLoop satellite, Lyteloop cell, and LyteLoop tube systems, storage capacity can be increased by extending the wavelengths of the lasers from NIR down to the visible and even UV light. Using wavelengths from 266 to 3400 nm, we have a bandwidth of 1004 THz, which is many times more than in current fiber long haul systems. Using tens of layers of dielectric coating, broadband apertures with high reflectivity can be obtained, as shown in Fig. 4.

Fig. 4

Mirror reflectivity as a function of wavelength.

In addition to wider bandwidth, for free space, one can then use space-division-multiplexing using, e.g., OAM or HG modes, as discussed above. For fibers, we can also use spatial modes and multiple cores.

2.3.

Near Ideal All-Optical Regeneration

In a storage in motion system based on laser communication, losses need to be compensated. Using state-of-the-art amplifiers poses two challenges: the first challenge is bandwidth, and the second is noise introduced by optical amplifiers exponentially increases the optical power consumption as the number of these amplifiers increases. In recent years, it was shown that using a chain of theoretical “ideal regenerators,” it might be possible to significantly reduce the power consumption in long haul fiber systems.⁹ An ideal regenerator is a mathematically ideal device that has the following two properties:

Optical coherent signals have both of their quadratures modulated. For this reason, a near-ideal regenerator can either first separate each quadrature and separately regenerate their amplitude levels before recombining them, as shown in Fig. 5(a), or sequentially regenerate optical signal as in Fig. 5(b). Amplitude regenerators can be achieved using saturable amplifiers while phase-sensitive amplification can be used to regenerate phase. In LyteLoop fiber, LyteLoop cube, and LyteLoop tube systems, the controlled environment and the fact that all regenerators are in the same place can be harnessed to increase the efficiency and near ideality of the regenerators. Using our patent-pending concepts of sharing components and functionalities that are located in one place, we are developing regenerators with the potential of being scaled to larger numbers of spatial and wavelength channels [space division multiplexing (SDM)-wavelength-division multiplexing (WDM)], with the ability to share the pump and hence its coherence among signals and near-ideal regenerators.

Fig. 5

Principle of a near-ideal regeneration for a coherent optical signal where (a) quadratures are separately regenerated and recombined and (b) amplitude is first regenerated then phase is regenerated.

To illustrate the gain in optical power consumption, we have calculated the theoretical lower limit of optical power consumption for a LyteLoop fiber system at the quantum Shannon limit. The fiber under consideration has 50 cores and 23 spatial modes per core and a total length of 10,000 km. We assumed its absorption is the same for all modes and all cores. It is shown in Fig. 6(a) with a minimum loss of $0.18\mathrm{dB}/\mathrm{km}$ at 1550 nm. The results are shown in Fig. 6(b) for ideal regenerators positioned at different spans. For each number of bits/symbol, the bandwidth of the optical signal was adjusted to reach a total capacity of 0.98 petabytes. The results show that it is possible to reduce the theoretical minimum optical power of keeping data in motion to below $10\mathrm{mwatts}/\mathrm{Tbyte}$ with the right spacing and coding design.

Fig. 6

(a) Calculated absorption spectrum of a silica based multicore and multimode fiber with 0.18 dB/km at 1550 nm where all modes are supposed to have the same theoretical absorption. The black curve is the total absorption, the red curve the Rayleigh contribution and the blue curve the NIR absorption. (b) Power consumption (left vertical axis) and capacity (right vertical axis) as a function of #bits/Symbol/Hz for different spacing between ideal regenerators. The capacity is kept constant at 0.98 Petabytes while the bandwidth is reduced.

We have calculated that similar optical power reduction can be obtained in LyteLoop cell, LyteLoop tube, and LyteLoop satellite. It is important to note that these savings occur if the optical signal is not detected in the electric domain. In other words, as long as read/write operations in the storage in motion system do not occur, it is possible to have a low idle power consumption.

3.1.

Fiber-Based Proof of Concept

Using state-of-the-art long-haul telecommunications systems, we have set up a storage loop with 1.5-Gbytes capacity. The loop is composed of a point-to-point 1000-km long-haul WDM coherent transmission system as shown in Fig. 7(a). Each terminal consists of six transponders that convert $2\times 100\mathrm{gpbs}$ ethernet stream onto one 200 gbps optical transport unit, stream carried by a dual-polarization 16-QAM modulated C-band WDM 37.5-GHz channel. All six WDM channels are multiplexed into the transmission line that is composed in each direction of 10 spans of 100-km SMF-28 with amplifiers and ROADMs in between the spans for signal amplification and WDM channel equalization. Excluding 25% forward error correction overhead, an aggregated of 1.2 Tbps is propagated over $2\times 1000\mathrm{km}$ in both direction, which corresponds to a storage capacity of 1.5 Gbytes. To write a file, a proprietary algorithm converts the file to be stored in the loop to 100 Gbps ethernet packets that are fed through an ethernet switch to one of the terminal ethernet port. By connecting all the other ethernet ports in a daisy-chain configuration using ethernet cables, the 100 Gbps stream is propagated 12 times back and forth over the 1000-km transmission system. At the last port, a loopback is used to propagate the stream back to the input ports. In this manner, the whole capacity of the storage loop is filled with only a 100-Gbps line feed, without the need to generate a 1.2 Tbps data stream. When the stream has looped back to the first ethernet port, it is sent to the second port of the ethernet switch using a custom-made multi-fiber push-on cable (ribbon cable supporting multiple fibers) to separate the ethernet ingress and egress streams from/to the transponder to/from the switch. By default, the switch will redirect the stream to the port that is connected to the transponder input port.

Fig. 7

(a) Storage in a 2000-km fiber loop using a long-haul bidirectional 1000-km WDM system. Eth: ethernet (b) graphic interface for storage system access and monitoring.

When the user decides to access a file, the switch reroutes a copy of the packets corresponding to the desired file to the server. It is also possible to erase any file by disabling the switch rerouting to the storage loop. A user interface and monitor dashboard were created to seamlessly perform any disk operation on the storage loop and monitor the storage space already used and the storage space still available [Fig. 7(b)]. It is noteworthy that in this system, parity-check is performed at every loop. Files were stored for over months with no errors, corresponding to a distance traveled of nearly 300 billion km.

3.2.

Free-Space Prototype

A free-space prototype is currently being built. At completion, it will consist of a transmitter system having 70 wavelength channels spanning from 550 to 1550 nm. Each wavelength will be composed of 10 spatial channels, either OAM modes or HG modes. In total, 700 channels at 200 Gbps will be generated for an aggregate throughput of 140 terabits per second.

The optical path will be based on the patented AM system with 40 apertures facing 60 apertures 10-m apart. The total optical path will therefore be 48 km since both directions of propagation will be used.

The free-space prototype will be built in four phases. In phase 1, one-dimensional (1D) AM has been demonstrated using two different approaches. The later phases concentrate on generating more wavelengths and storing them in AM setups with increased optical path length. This paper focuses on our experimental results in phase 1.

The group has demonstrated the feasibility of the transmitter system design, by connecting 10 C-band channel transceivers propagating through 10 HG modes and back. The SNR was calculated per addition of mode and per channel. Figure 8 shows a schematic of the experimental set up.

Fig. 8

Schematic showing experiment conducted using 10 DWDM channels and 10 OAM modes.

Figure 9 shows that the SNR reduces with the addition of HG modes. This can be attributed to crosstalk between the spatial modes. The SDM devices are sensitive to alignment and vibrations. With better alignment and vibration isolation techniques, we expect these SNR values to improve.

Fig. 9

Graph showing how the SNR changes with the addition of HG modes.

Figure 10 shows the variation in SNR with channel frequency. In this experiment, all 10 spatial modes were sourced from 12 unique C-band channels and were split into 10 copies using a $1\times 10$ optical splitter. As can be seen, the SNR is above 18 and consistent through the 10 channels. This shows that the SDM device is capable of producing HG modes consistently over 10 C-band channels

Fig. 10

Graph showing the variation of SNR with channel frequency.

Figure 11(a) shows the HG modes generated by the prototype system at 1550 nm, recorder using an IR CCD camera. Figure 11(b) shows a ray-tracing simulation of part of the AM system.

Fig. 11

(a) IR CCD camera images of the HG modes generated at 1550 nm by the free-space LyteLoop prototype. (b) Ray-tracing simulation of the AM system with eight apertures on each side (64 optical paths).

One of the critical aspects of phase 1 was demonstrating AM in 1D, meaning all of the apertures are in a single plane. In phase 2, we have apertures in two planes, so arrays on each side will become two dimensional (2D). In phase 1, we have two lines of apertures on each side. The achieved goals of phase 1 are shown in Table 2. Initially, we demonstrated AM using a combination of VHGs and mirrors. One issue with this approach is the VHGs are not broadband, but we will have to use wavelength channels at multiple wavelengths. Then, because of the interest in broadband operation, we also demonstrated 1D AM using mirrors.

Table 2

Overview of results achieved in the three-AM demonstrations.

There were three-AM demonstrations completed in the lab for phase 1. The first (phase 1a) used VHGs and had 1D aperture arrays on each side, the second (phase 1b) used VHGs and had two-dimensional aperture arrays on each side, and the third used only mirrors and had 1D aperture arrays on each side. All three demonstrations to be discussed were done with a 633-nm Gaussian beam with the intention of adding more wavelengths and spatial modes in future phases. The objective of these lab demonstrations was to reduce risk for an eventual LyteLoop product. The geometry of the AM subsystem is such that this reduces more risk for tube and cell than it does for space, because the experiment we conducted was more similar to the tube or cell, geometry.

3.2.1.

Phase 1a and phase 1b—angle multiplexing with VHGs

VHGs were chosen for phase 1a and 1b because of their physical properties of angular selectivity and diffraction efficiency. The optical path length of a laser in a 10-m long modified 4f cavity was successfully extended to 800 m with the phase 1a approach and 160 m in the phase 1b approach. The cavity utilized two large, curved mirrors (mirrors 1 and 3) on side A and another two large, curved mirrors (mirrors 2 and 4) on side B. Many of the apertures were simply virtual spots on these large mirrors that the light repeatedly bounced off. VHGs and a flyback mirror were mounted in front of the large mirrors to create the necessary propriety geometry for AM to function properly. For phase 1b, two 1D VHGs were mounted diagonally, and orthogonal to each other, at both mirror 1 apertures to effectively steer the light in 2D. All VHGs for both phase 1a and 1b were aligned for maximum throughput and the correct angular separation using piezoelectric motors. Figure 12 shows AM demonstrated experimentally, using mirrors and VHGs, in phase 1a.

Fig. 12

(a) Experimental setup for phase 1a with VHGs. (b) Experimental setup for phase 1b with VHGs.

A critical aspect of any AM setup is signal loss, since after roughly 20 dB of loss signal is unable to be read by our current transceivers. Maintaining high mirror reflectivity is thus crucial. Transmissive VHGs were shown to have $\sim 5\%$ loss per pass, which severely lowered transmission after many bounces through the system. This loss was measured for phase 1a and is shown in Table 3.

Table 3

Transmission power loss during propagation due to mirror and grating losses for phase 1a.

Location	Power (mw)	Power (dBm)	Net transmission (%)
After the ingress beam expander	15.34	11.86	100
At mirror 3, aperture 1 after first grating alignment	13.47	11.29	87.8
At mirror 3, aperture 2 after second grating alignment	6.60	8.20	43
At mirror 3, aperture 3 after third grating alignment	3.19	5.04	20.8
At mirror 3, aperture 4 after fourth grating alignment	1.59	2.01	10.4
At mirror, after egress mirror and steering mirrors	0.80	−0.97	5.2

Figure 13 shows a demonstration of AM using gratings.

Fig. 13

Phase 1a in operation.

3.2.2.

Phase 1—All mirror angle multiplexing

A new geometrical approach to AM that only uses mirrors, no gratings, was designed and demonstrated in the lab after phase 1a and phase 1b were complete. Unlike the VHG demonstrations, it was inherently broadband and bidirectional, and it had eight apertures on one side and six on the other for a total of 98 hops. It also had less loss. The same four large, curved mirrors from phase 1a and phase 1b were used in a similar axial and radial orientation with several smaller mirrors added to redirect the light at certain locations.

The input power was 16.20 mw and the power at the egress was 3.08 mw for a transmission of 19%. All the loss came from the scattering and absorption of the mirrors. The system had a total of 98 hops. This can be calculated by $Hops=2*(M_{\mathit{Tot}}{N}_{\mathit{Tot}})+2$ where $M_{\mathit{Tot}}$ and $N_{\mathit{Tot}}$ are the total number of apertures on each side. The additional two hops come from the two ingress hops that were not counted before the input into the AMAM system. Mirrors 1 and 3 had four apertures each and mirrors 2 and 4 had three apertures each, so there are double that many apertures on each side, and thus there are $2*(8*6)+2=98Hops$. Figures 14 and 15 show AM operating using mirrors. It is noteworthy that the number of hops can readily be increased by modifying the inclination angles of the ingress/egress mirrors as well as that of the mirrors.

Fig. 14

AMAM as demonstrated in the lab as seen from the mirror 1, mirror 3 side. Mirror 2 is on the right and mirror 4 is on the left.

Fig. 15

AMAM as demonstrated in the lab as seen from the mirror 2, mirror 4 side. Mirror 1 is on the left and mirror 3 is on the right.