Progress in second-generation holographic data storage

Transcription

1 Progress in second-generation holographic data storage Mark R. Ayres *, Ken Anderson, Fred Askham, Brad Sissom Akonia Holographics, LLC, 221 Miller Dr., Longmont, CO, USA, 851 ABSTRACT Holographic data storage (HDS) remains an attractive technology for big data. We report on recent results achieved with a demonstrator platform incorporating several new second-generation techniques for increasing HDS recording density and speed. This demonstrator has been designed to achieve densities that support the multi-terabyte storage capacities required for a competitive product. It leverages technology from an existing state-of-the-art pre-production prototype, while incorporating a new optical head designed to demonstrate several new technical advances. The demonstrator employs the new technique of dynamic aperture multiplexing in a monocular architecture. In a previous report, a monocular system employing angle-polytopic multiplexing achieved a recording density over 7 Gbit/in 2, exceeding that of contemporaneously shipping hard drives [1]. Dynamic aperture multiplexing represents an evolutionary improvement with the potential to increase this figure by over 2%, while still using proven anglepolytopic multiplexing in a monocular architecture. Additionally, the demonstrator is capable of two revolutionary advances in HDS technology. The first, quadrature homodyne detection, enables the use of phase shift keying (PSK) for signal encoding, which dramatically improves recording intensity homogeneity and increases SNR. The second, phase quadrature holographic multiplexing, further doubles density by recording pairs of holograms in quadrature (QPSK encoding). We report on the design and construction of the demonstrator, and on the results of current recording experiments. Keywords: holographic data storage, big data, optical data storage, holography 1. INTRODUCTION The Akonia Holographics AP1 demonstrator platform was designed and constructed in 213, with data density experiments commencing in early 214. Development is ongoing, with many key technical features yet to be engaged. Nevertheless, as of this writing, AP1 has already demonstrated the recovery of data written at a raw areal density of 1.35 Tbit/in 2, exceeding that of shipping hard drives (currently at approximately 1 Tbit/in 2 [2]). 2. SECOND GENERATION HDS Despite a number of highly publicized efforts, no commercial HDS product has ever been sold to a general market. However, the Bell Labs spin-off, InPhase Technologies, Inc., produced an advanced pre-production HDS drive capable of storing 3 GB on a removable 5 ¼ disk [3]. We shall take the features of the InPhase Tapestry prototype as typical of first generation HDS. These features include an off-axis architecture recording 32 angle-polytopic [4] multiplexed holograms in a single location, and direct-detection resampling data channel [5] operating on 1.4 megapixel holographic data pages. The InPhase drive achieved a peak areal density of 512 Gbit/in 2 in a photopolymer medium with a 1.5 mm thick recording layer, with read and write data rates of 2 MB/s. Other first generation HDS systems include the monocular [6], collinear [7] and bit-wise architectures, though none were developed to a state nearing commercial readiness. By second generation HDS, we denote a set of systems and materials innovations that will increase storage density and speed by more than an order of magnitude. These include: * mark@akoniaholographics.com; phone ; akoniaholographics.com

2 1) DRED TM Media Formulation The DRED formulation represents a significant advance over the two-chemistry recording medium developed by researchers from InPhase. The two-chemistry medium is so-named because it consists of two separate, intermixed polymer systems. The first system, the matrix, forms the solid recording layer, and the second remains unpolymerized until exposed to light [8]. Akonia has demonstrated a factor of six increase in dynamic range (M/#) using DRED technology, and is currently optimizing it for commercialization. 2) Dynamic Aperture Multiplexing Dynamic aperture multiplexing fully capitalizes on the shared beam path of a monocular objective lens. Dynamic aperture multiplexing increases the angular scan range by dynamically adjusting the data page size, increasing storage density by over 2%. 3) Quadrature Homodyne Detection Homodyne detection is the method of blending a coherent reference field with a signal and detecting the interference pattern between the two. This has the effect of amplifying the signal, eliminating nonlinear effects of coherent noise, and allowing the detection of phase as well as amplitude. Akonia has developed a novel algorithm that allows homodyne detection to be performed by combining two images taken with the reference phase changed by 9 o. We estimate this will boost the signal to noise ratio enough to more than double storage density, as well as providing a host of other benefits [9]. 4) Phase Quadrature Holographic Multiplexing The ability to detect the phase of a hologram presents another opportunity to increase storage density. A second hologram with a 9 o phase difference from the first can be recorded at each reference angle. This technique, phase quadrature holographic multiplexing, provides yet another doubling of storage density [1]. 2.1 AP1 Demonstrator Platform The AP1 demonstrator was constructed from an InPhase prototype, and hence inherited a high level of functionality from that platform. Components and subsystems incorporated with little or no modification include the SLM and detector, tunable 45 nm laser, electronics and firmware, disk loader, and various actuators and servo systems. Many optical components have also been retained, but the objective lens and reference scanner assemblies have been replaced with a newly designed.85 NA objective lens and scanner capable of delivering diffraction-limited planar reference beams over the entire ±6 o angle scan range. Figure 1. The Akonia Holographics AP1 demonstrator platform. The completed AP1 demonstrator is shown in Figure 1. The objective lens is housed within the assembly at center, and the disk is visible in the bottom of the photograph. Figure 2 is a schematic of the main optical paths. The aperture

3 sharing element (ASE) combines the signal and reference beams, allowing angular regions to be allocated in real time between the two beams for dynamic aperture multiplexing. Figure 2. AP1 schematic. Since AP1 uses the legacy 1.4 megapixel SLM, laser, and electronics from the 2 MB/s InPhase prototype, it cannot directly demonstrate data transfer rates that would be attractive for a commercial product. Akonia has instead developed a model based on a modern 4K-class SLM that supports competitive 2 MB/s recording and recovery speeds, even without the benefit of the coherent homodyne data channel, which would further boost transfer rates. Fortunately, however, the legacy components can support recording densities equaling those achievable with newer components. More media M/# is required since the data payload of each hologram is smaller, but once that is achieved switching to a larger page size will actually result in a system with more margin. Hence, the overarching goal of the AP1 demonstrator is to achieve full density with no asterisks, thereby demonstrating with high confidence the feasibility of the technology for commercial products. 2.2 DRED Recording Medium Although the number of holograms per book is currently the same as for Tapestry, the higher-density AP1 architecture requires considerably more dynamic range. Density experiments are currently being conducted using a DRED formulation with an M/# of 16 per 2 µm thickness, resulting in a total M/# of 12 in the 1.5 mm thick recording layer. This exceeds the M/# of the InPhase Tapestry formulation by a factor of four. The media is bonded between two 13 mm plastic disk substrates using the InPhase Zerowave TM bonding process [11]. The resulting disk is fitted with a hub and cartridge, and has sufficient recording area for over a hundred automated recording experiments. Akonia owns InPhase manufacturing equipment capable of large-scale media production. 3. EXPERIMENTAL PROGRESS Density demonstration experiments are performed by recording a grid, typically of 6 9 books of angularly-multiplexed holograms spaced at a book pitch of 34 µm. Gridding is necessary because the exposure footprint of each book is considerably larger than the book spacing, so several neighbors are required on all sides of a given book to replicate the hologram overlap conditions that will obtain in a full-capacity medium. In the case of a 6 9 grid, the inner 2 3 books are so overlapped, and thus at density. For a large production medium, only those books at the outer edges of the medium or of a recording partition will not be fully overlapped, and the peak data density will dominate. This crucial step is sometimes omitted from reported density demonstrations.

4 3.1 Monocular-Mode Results AP1 features a monocular architecture with both signal and reference beams delivered through the same high-na objective lens. As such, it is capable of exceeding the performance of previous monocular systems [1][12] even without invoking the second-generation innovations. In the first phase of AP1 experiments, the system was operated in conventional monocular mode, i.e., using a fixed-size data page and a disjoint reference angle range. For the following experiment, 32 holograms were recorded in each book of the 6 9 grid. The data region of each page contains 66,29 pixels, so the raw areal bit density is 66,29 bits ( 34 µm) 32 pages 2. (1) = 1.35 Tbit/in 2 The data pages were recovered using the 4:3 oversampling data channel of the InPhase Tapestry [5]. The channel uses known reserved block pixel patterns embedded within the page to determine a quality metric, µ 1 µ SNR 2 log 1, (2) σ1 + σ where µ 1 and µ are the means, and σ 1 and σ are the standard deviations of the detected ones and zeros, respectively. Figure 3 shows a typical detector image of a 4 db hologram recovered at density, along with the sampled reserved block SNR map of the page Figure 3. Left: Detected data page recovered at density; Right: Regional SNR map of data page. Figure 4 shows the aggregate SNR for each hologram in the first density book within a grid written at 1.35 Tbit/in 2. The average SNR of the holograms in the book is 3.2 db, and the lone hologram below 2 db (at -37 o ) was degraded by a known dirt speck on an optical surface. Also shown is the diffracted power of holograms, demonstrating the uniform peak and trough (between hologram) signal levels. The Tapestry data channel also incorporates several layers of error correcting codes of the types that will be used in the channel for AP1. A low density parity check (LDPC) code with rate.5 and code word length 32,768 is used within each data page. The LDPC code will correct all bit errors within a recovered page with an extremely high level of probability if the SNR of the page is 2. db or higher. Additionally, an outer Reed-Solomon code can correct up to 1% of the pages within a book that are uncorrected by LDPC. Thus, the density book of Figure 4 would easily be recovered free of bit errors using the full Tapestry channel, and AP1 has demonstrated holographic data recording and recovery at a record-breaking areal density of 1.35 Tbit/in 2. Further code development in response to empirical channel characteristics will be required to meet an exacting production bit error rate specification. Of particular importance in these results is the demonstration of recording in the extremely high dynamic range DRED medium without incurring debilitating write-induced optical scatter noise. Some fraction of the dynamic range of the medium inevitably goes to writing noise rather than signal, so one early concern was that this effect could dominate at very high dynamic range levels. An indication of the scatter level is given by the detected power levels at reference -5 Though the hardware-based Tapestry error-correcting channel remains fully functional, the bring-up of an SNR/BER-only channel for AP1 was deemed the first priority.

5 angles far from any recorded holograms, e.g., the diffracted power levels of Figure 4 at reference angles below -6 o and above -3 o. Ongoing testing at Akonia has indicated that, while these levels are strongly influenced by media formulation and preparation, there is no fundamental correlation between M/# and scatter level. This result gives us confidence that both higher SNRs and higher densities will be achievable as the system and materials are refined. Raw bit density, for example, will be doubled directly by phase quadrature multiplexing, and SNR will be greatly increased by PSK modulation and homodyne detection. 5 x Hologram SNR [db] Diffracted Power [W] Reference Angle [deg] Figure 4. SNR and diffracted power of monocular holograms recovered at an areal density of 1.35 Tbit/in Dynamic Aperture Multiplexing Overview Dynamic aperture multiplexing improves storage density by greatly increasing the scan range available for angle multiplexing. This is accomplished by dynamically altering the portions of the available angular aperture used for the signal and reference beams. Figure 5 illustrates the general principle of dynamic aperture multiplexing, showing the angular location of reference beam (green dot) as it transits from the left edge of the figure (-6 o reference beam angle) to a position on the right half (+1 o reference beam angle). The black rectangle represents the footprint of the SLM image in the angular map, and the gray sub-region of the SLM indicates the active data page for the corresponding reference beam angle. As the reference beam scans from left to right, the data page is trimmed to maintain a constant angular separation between the reference beam and the closest edge of the data page, establishing the minimum Bragg selectivity for the gratings within the hologram. a) b) c) Figure 5. Angular aperture maps for reference beam angles a) 6 o, b) 2 o, c) +1 o. Dynamic aperture multiplexing more efficiently utilizes the available grating space not only by recording more holograms, but by recording larger data pages at the start of the angular scan range. However, the decreasing page size leads to diminishing returns in density and higher M/# usage as higher scan angles are used. Additionally, data transfer rates decline as more small pages are included. Figure 6 illustrates this trade-off. Although AP1 is capable of reaching

6 2.84 Tbit/in 2 by recording 1131 holograms, this would require more than twice the dynamic range needed to reach 2. Tbit/in 2 with 455 holograms. For this reason, Akonia anticipates operating in the region below the knee of the curve in a commercial product, although higher-density demonstrations may be performed in the laboratory Areal Density [Gbit/in 2 ] Transfer Rate [MB/s] 1 5 Read Write Holograms Multiplexed 3.3 Dynamic Aperture Multiplexing Progress Figure 6. Density vs. number of holograms for dynamic aperture multiplexing. Dynamic aperture multiplexing is currently being actively developed on AP1. Holograms of varying page size have been recorded and recovered, and final modifications are being made to various subsystems in preparation for density experiments. Since the number of holograms required per book to achieve 2. Tbit/in 2, 455, is only marginally larger than the 32 already demonstrated, we believe that this milestone will be passed in the near future. Dynamic aperture multiplexing is an evolutionary improvement, merely adding more available angles to an angle-multiplexing system. It requires only more M/# and sufficient noise performance, both of which are close at hand. 3.4 Quadrature Homodyne Detection Progress Quadrature homodyne detection is also being actively developed at Akonia. Quadrature homodyne detection is expected to double storage capacity by dramatically improving SNR, as well as improving data speeds. Homodyne detection also enables phase quadrature multiplexing, which will further double storage density. Testing of phase quadrature multiplexing is currently being planned. Because AP1 is still configured for direct detection in pursuit of the dynamic aperture milestones, a separate test system has been constructed implementing the homodyne data channel and using PSK signal modulation. The results from this system are extremely positive, with SNR improvements exceeding even the simulated predictions. Please refer to Homodyne detection of holographic memory systems, to be presented later in this session, for details on the results of these experiments. 4. CONCLUSION We have presented experimental results from a second-generation holographic data storage demonstrator platform. Though the second-generation features have not yet been fully deployed, the demonstrator has achieved a world-record raw areal bit density of 1.35 Tbit/in 2 operating in a conventional first-generation monocular mode. While density results have not yet been confirmed for dynamic aperture multiplexing or for quadrature homodyne detection, progress has been ongoing and both features remain on track to meet or exceed their original performance expectations.

7 Successful demonstration of the second-generation innovations will spark a renaissance in holographic data storage. The big data storage tsunami is generating enormous opportunities for cold storage, archival storage, and near-line storage of the sorts that will be the target of initial HDS commercial offerings. These second-generation innovations not only leapfrog the capabilities of competing technologies, they lay the foundation for Moore s-law growth for years to come. REFERENCES [1] K. Shimada, T. Ishii, T. Ide, S. Hughes, A. Hoskins, K. Curtis, High density recording using Monocular architecture for 5GB consumer system, Proc. SPIE755, Optical Data Storage 29, 755Q (29). [2] T. Coughlin, New Areal Density Point for Cloud Storage HDDs, Forbes, 7 April 214. [3] K. Curtis, L. Dhar, W. L. Wilson, A. Hill, M. R. Ayres, Holographic Data Storage: From Theory to Practical Systems, John Wiley & Sons, Ltd. (21). [4] K. Anderson, K. Curtis, Polytopic multiplexing, Opt. Lett. 29, (24). [5] M. Ayres, A. Hoskins, K. Curtis, Image oversampling for page-oriented optical data storage, Appl. Opt. 45, (26). [6] K. Anderson, et al., U.S. Patent 7,742,29, Monocular holographic data storage system architecture, June 22, 21. [7] H. Horimai, X. Tan, and J. Li, Collinear holography, Appl. Opt. 44, (25). [8] L. Dhar, A. Hale, H. E. Katz, M. L. Schilling, M. G. Schnoes, F. C. Schilling, Recording media that exhibit high dynamic range for digital holographic data storage, Opt. Lett. 24, (1999). [9] M. R. Ayres, U.S. Patent 7,623,279, Method for holographic data retrieval by quadrature homodyne detection, Nov. 24, 29. [1] M. R. Ayres, Coherent techniques for terabyte holographic data storage, Optical Data Storage Topical Meeting (ODS 21), May 21. (Invited paper). [11] S. Campbell, et al., US Patent 5,932,45, Method for fabricating a multilayer optical article, August 3, [12] T. Ishii, et al., Terabyte Holographic Recording with Monocular Architecture, 212 IEEE International Conference on Consumer Electronics (ICCE).