Storing data holographically is an attractive proposition. Using the whole volume of a storage medium, instead of just the surface, allows us to encode much more information in a small space. In addition, the fact that data can be recorded and read out in parallel makes information access and transfer extremely fast. The fundamental problems in achieving holographic storage have been the lack of a viable storage medium and the need for a recording system that can take full advantage of holography's possibilities. Current developments are addressing both issues.
Conventional optical storage systems achieve increases in density by decreasing the size of the marks on the surface of the storage disk. The primary advantage of holographic storage comes from using the volume of the medium and not just the surface to store information.
To write the data in a holographic storage system, a beamsplitter divides an input laser beam into two parts: a signal beam that carries the data and a reference beam. A spatial light modulator (SLM) encodes the data onto the signal beam by translating the bit stream of 0s and 1s into an optical checkerboard pattern of light and dark pixels, arranging the data in an arrayor pageof approximately 1 Mb. When the reference beam and the signal beam intersect in the storage medium, they record the hologram. By varying the angle or wavelength of the reference beam or by shifting the position of the medium, one can record multiple holograms in the same volume, thereby vastly increasing the data density.
To read the data, a reference beam with characteristics identical to that used to store the data diffracts off the index modulation (hologram), reconstructing the data page with the stored information. A detector captures an image of the reconstructed data page and reads the data in parallel. This parallel recording and readout of data provides holography with its potential for very fast transfer rates. materials improvements
The major challenge to implementing holographic storage has been the development of a suitable storage medium.1 Our group addressed this issue by developing a new class of materials that satisfy the stringent criteria for commercial viability. The material offers high dynamic range, high photosensitivity, dimensional stability, optical clarity, manufacturability, nondestructive readout, thickness, and environmental and thermal stability.
Typical photopolymers use a single chemical process for bonding molecules together both to form the medium and perform the recording. We developed a polymer system with materials that use two distinct chemistries that are independent, yet compatible. One chemistry forms the medium and controls the mechanical, manufacturing, and data/archive lifetime parameters. The second chemistry comes into play during the recording process. These two do not interact or interfere with each other, thus enabling high dynamic range with extremely good dimensional stability during recording.
In addition to developing the materials, we also developed manufacturing processes to enable cost-effective fabrication of high-quality optical media. In order to record large pages of data containing 1 million bits with a good signal-to-noise ratio (SNR), the medium must have very good optical quality as measured by transmitted wavefront. The wavefront quality as manufactured by our process is about λ/8 per cm2 at 633 nm. demo system
As the first step in commercializing a holographic storage drive, our company has focused on developing test beds for each of the primary functions: channel, servo, mechanical, and optical sub-systems. The test beds allow us to develop the various functions of the drive in parallel before we integrate the functions into a complete prototype drive. All test bed components and mechanics are adjustable, including the wavefront quality of the beams. We have also developed all the test bed electronics. We can currently support 10-MB/s transfer rates with designs based on field programmable gate arrays. The other critical drive components include a singlemode, single-frequency gallium nitride diode laser generating 30 to 50 mW at 405 nm, a ferroelectric-liquid-crystal SLM, and a 1280 x 1024 Complementary metal-oxide semiconductor active pixel sensor operating at 1000 frames; all are commercially available or under development with industry partners.
We have built a small, portable demonstration system that records and plays encoded digital video in the MPEG 4 format to demonstrate the feasibility of holographic storage. The demo performs angle multiplexing to record multiple, superimposed holograms in the volume of a medium between 0.8- and 1-mm thick. A precision galvo mirror controls the reference beam angle and a set of scan optics images the beam onto the recording medium. The system stores holograms with a spacing of 0.065° and has control hardware and a hardware input/output channel. It operates in a normal room environment with no active vibration isolation.
Figure 1. The optomechanical subassembly includes a ferroelectric liquid crystal driven by control electronics and a CMOS camera. INPHASE TECHNOLOGIES
The optomechanical subassembly in the demonstration system contains the SLM, the Fourier transform optics, and the mount for the medium (see figure 1). This design allows for robust operation and easy, independent alignment of the module. The optical path from modulator to camera is less than 10 cm.
The non-uniform light intensity of the data page image limits a system's SNR. Part of the reason for this non-uniformity is the variation in intensity of the beam illuminating the modulator. This system records data page images with extremely high fidelity, which means the recorded images are essentially identical to the input images. Two issues affect fidelity: the imaging quality of the optical system and the ability of the media to record that image; a single recorded image should be identical in SNR to the image transferred through the system. Single holograms easily exhibit SNRs in the range of 9 to 10 dB, with less than 0.5% pixel saturation. The optical quality of the medium makes this fidelity possible by minimizing distortion. In most cases, hologram reconstructions are indistinguishable from input images, which means the media will in essence record all input images with no loss of information.
The physical interface board controls and directs the recording and readout functions. The servo board controls the galvo mirror for angular addressing of the holograms, as well as the x, y, and θ motion of the readout camera for system alignment and feedback. We're designing the application-specific integrated-circuit electronics to run at 20 MB/s, roughly 50% of which is overhead.
We've developed a first-generation servo system using marks mastered into the medium. We have demonstrated functions including seek, tracking, and sector decoding. We have also been able to perform fully automated disk interchange between two development platforms. This is the first demonstration of holographically recorded information transfer between two independent systems. taking data
Figure 2. Intensity of the recording varies with angle in a set of about 90 holograms representing better than 5 MB of MPEG 4 video.
Using an unobstructed scan range of approximately 11°, the demonstration system records a stack of 90 to 100 holograms, more than 5 MB of data in a single location (see figure 2). The media is recordable, write-once material. As recording proceeds, the absorption of the material decreases and the amount of exposure needed slowly increases. We used a recording schedule to equalize the diffraction efficiency of each recorded page. It is important to note that the sensitivity of the material falls off nearly linearly with exposure over the range we used, with the first exposure at 50 ms and the final at 75 to 100 ms. This contrasts with other photopolymer media and inorganic photorefractive materials, in which the response falls off exponentially. In addition, the medium used for demonstration purposes is at least 10 times less sensitive than the commercial medium, because that lets us operate the device while open to ambient light, allowing us to observe recording and readout.
Figure 3. In the first (top) and last (bottom) recovered holograms of a data set written in a single location, both data pages have signal-to-noise ratios of about 7 dB.
With proper scheduling, the first and last exposures record holograms with comparable SNRs. In the demonstration, the first and last recovered page from a data set each had an SNR of about 7 dB (see figure 3); the average SNR of the data set was 6.5 dB, typical for this platform.
In our system, all holograms with an SNR above approximately 2 dB can be recovered error free. To get our results, we derived the hologram histograms using the entire data page with no pre-processing. The overlap of the data distributions was exaggerated due to the intensity variations across the image, as mentioned above. In practice, we apply equalization algorithms to the recovered data page before decoding to eliminate this variance.
Our demonstration system serves as the mile marker on the path to commercialization of holographic storage. Holographic technologies will provide the highest-density removable media with transfer rates much faster than those of other optical-data-storage devices. Ultimately, the technology will be integrated into a multitude of products, ranging from corporate storage systems to consumer devices similar to CD readers. The long-term vision is that holographic technology will one day take on a prominent role in the storage landscape. oe
1. L. Dhar, C. Boyd, et al., Optical Data Storage 8, OSA Technical Digest Series, p. 113 (1998); Photopolymers for Digital Holographic Data Storage; L. Dhar, M. Schnoes, et al., "Holographic Data Storage," part 5, Optical Sciences Book Series, Springer Verlag (2000).
William Wilson, Kevin Curtis, Lisa Dhar
William Wilson is chief scientist, Kevin Curtis is CTO, and Lisa Dhar is vice president of media development at InPhase Technologies, Longmont, CO.