Search

Generative molecular design (GMD) is an increasingly popular strategy for drug discovery, using machine learning models to propose, evaluate and optimize chemical structures against a set of target design criteria. We present the ATOM-GMD platform, a scalable multiprocessing framework to optimize many parameters simultaneously over large populations of proposed molecules. ATOM-GMD uses a junction tree variational autoencoder mapping structures to latent vectors, along with a genetic algorithm operating on latent vector elements, to search a diverse molecular space for compounds that meet the design criteria. We used the ATOM-GMD framework in a lead optimization case study to develop potent and selective histamine H1 receptor antagonists. We synthesized 103 of the top scoring compounds and measured their properties experimentally. Six of the tested compounds bind H1 withKi’s between 10 and 100 nM and are at least 100-fold selective relative to muscarinic M2 receptors, validating the effectiveness of our GMD approach.

Drug-induced liver injury (DILI) is the most common cause of acute liver failure and a frequent reason for withdrawal of candidate drugs during preclinical and clinical testing. An important type of DILI is cholestatic liver injury, caused by buildup of bile salts within hepatocytes; it is frequently associated with inhibition of bile salt transporters, such as the bile salt export pump (BSEP). Reliable in silico models to predict BSEP inhibition directly from chemical structures would significantly reduce costs during drug discovery and could help avoid injury to patients. Unfortunately, models published to date have been insufficiently accurate to encourage wide adoption. We report our development of classification and regression models for BSEP inhibition with substantially improved performance over previously published models. Our model development leveraged the ATOM Modeling PipeLine (AMPL) developed by the ATOM Consortium, which enabled us to train and evaluate thousands of candidate models. In the course of model development, we assessed a variety of schemes for chemical featurization, dataset partitioning and class labeling, and identified those producing models that generalized best to novel chemical entities. Our best performing classification model was a neural network with ROC AUC = 0.88 on our internal test dataset and 0.89 on an independent external compound set. Our best regression model, the first ever reported for predicting BSEP IC50s, yielded a test set $R^2 = 0.56$ and mean absolute error 0.37, corresponding to a mean 2.3-fold error in predicted IC50s, comparable to experimental variation. These models will thus be useful as inputs to mechanistic predictions of DILI and as part of computational pipelines for drug discovery.

Hidden within our rapidly growing global streams of business, scientific, and communications data, is information that can reduce the global danger of nuclear weapons proliferation and use. Significant advances in machine learning, the explosion of available data and the ability of high performance computing that can be used to process massive amounts of data has the potential to provide a leap-ahead in the ability to detect proliferation and verify treaty commitments.

Machine learning is finding increasingly broad applications in the physical sciences. This most often involves building a model relationship between a dependent, measurable output, and an associated set of controllable, but complicated, independent inputs. We present a tutorial on current techniques in machine learning—a jumping-off point for interested researchers to advance their work. We focus on deep neural networks with an emphasis on demystifying deep learning. We begin with background ideas in machine learning and some example applications from current research in plasma physics. We discuss supervised learning techniques for modeling complicated functions, beginning with familiar regression schemes, and then advancing to more sophisticated deep learning methods. We also address unsupervised learning and techniques for reducing the dimensionality of input spaces. Along the way, we describe methods for practitioners to help ensure that their models generalize from their training data to as-yet-unseen test data. We describe classes of tasks—predicting scalars, handling images, and fitting time-series—and prepare the reader to choose an appropriate technique. We finally point out some limitations to modern machine learning and speculate on some ways that practitioners from the physical sciences may be particularly suited to help.

Adaptive optics (AO) is a technology that corrects in real time for the blurring effects of atmospheric turbulence, in principle allowing Earth‐bound telescopes to achieve their diffraction limit and to “see” as clearly as if they were in space. The power of AO using natural guide stars has been amply demonstrated in recent years on telescopes up to 3–4 m in diameter. The next breakthrough in astronomical resolution was expected to occur with the implementation of AO on the new generation of large, 8–10 m diameter telescopes. In this paper we report the initial results from the first of these AO systems, now coming on line on the 10 m diameter Keck II Telescope. The results include the highest angular resolution images ever obtained from a single telescope (0 \documentclass{aastex} \usepackage{amsbsy} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{bm} \usepackage{mathrsfs} \usepackage{pifont} \usepackage{stmaryrd} \usepackage{textcomp} \usepackage{portland,xspace} \usepackage{amsmath,amsx...

High-brightness and short-pulse-width ( approximately 200 ps) x-ray lasers offer biologists the possibility of high-resolution three-dimensional imaging of specimens in an aqueous environment without the blurring effects associated with natural motions. As a first step toward developing the capabilities of this type of x-ray microscopy we have used a tantalum x-ray laser (lambda = 4.483 nm) together with an x-ray zone plate lens to image a test pattern. The observed image shows a detector-limited resolution of approximately 75 nm and paves the way to three dimensional biological imaging with high spatial resolution (20-30 nm).

x-Ray computed axial tomography (CT) provides cross-sectional views of materials, components, and assemblies for industrial non-destructive evaluation. We have applied CT imaging to quantitatively measure the 3-D distribution ogf x-ray attenuation at reasonably high resolutions. In our industrial x-ray CT-studies, we have centered on two technical approaches: a first-generation translate/rotate CT system that consist of well-collimated (∼ 0.55 mm) photon source detector, and a third-generation scanner that uses a fluoroscopy detector.

The Image Content Engine (ICE) is being developed to provide cueing assistance to human image analysts faced with increasingly large and intractable amounts of image data. The ICE architecture includes user configurable feature extraction pipelines which produce intermediate feature vector and match surface files which can then be accessed by interactive relational queries. Application of the feature extraction algorithms to large collections of images may be extremely time consuming and is launched as a batch job on a Linux cluster. The query interface accesses only the intermediate files and returns candidate hits nearly instantaneously. Queries may be posed for individual objects or collections. The query interface prompts the user for feedback, and applies relevance feedback algorithms to revise the feature vector weighting and focus on relevant search results. Examples of feature extraction and both model-based and search-by-example queries are presented.

The problem of robust automatic road detection in remotely sensed images is complicated by the fact that the sensor, spatial resolution, acquisition conditions, road width, road orientation and road material composition can all vary. A novel technique for detecting road pixels in multisource remotely sensed images based on the phase (i.e., orientation or directional) information in edge pixels is described. A very dense map of edges extracted from the image is separated into channels, each containing edge pixels whose phases lie within a different range of orientations. The edge map associated with each channel is de-cluttered. A map of road pixels is formed by re-combining the de-cluttered channels into a composite edge image which is itself then separately de-cluttered. Road detection results are provided for DigitalGlobe and TeiraServerUSA images. Road representations suitable for various applications are then discussed

Laboratory and field demonstration results obtained as part of the DARPA-sponsored coherent communications, imaging and targeting (CCIT) program are reviewed. The CCIT concept uses a phase conjugation engine based on a quadrature receiver array, a hologram processor and a spatial light modulator (SLM) for high-speed, digital beam control. Progress on the enabling MEMS SLM, being developed by a consortium consisting of LLNL, academic institutions and small businesses, is presented.

Human analysts are often unable to meet time constraints on analysis and interpretation of large volumes of remotely sensed imagery. To address this problem, the Image Content Engine (ICE) system currently under development is organized into an off-line component for automated extraction of image features followed by user-interactive components for content detection and content-based query processing. The extracted features are vectors that represent attributes of three entities, namely image tiles, image regions and shapes, or suspected matches to models of objects. ICE allows users to interactively specify decision thresholds so that the content (consisting of entities whose features satisfy decision criteria) can be detected. ICE presents detected content to users as a prioritized series of thumbnail images. Users can either accept the detection results or specify a new set of decision thresholds. Once accepted, ICE stores the detected content in database tables and semantic graphs. Users can interactively query the tables and graphs for locations at which prescribed relationships between detected content exist. New queries can be submitted repeatedly until a satisfactory series of prioritized thumbnail image cues is produced. Examples are provided to demonstrate how ICE can be used to assist users in quickly finding prescribed collections of entities (both natural and man-made) in a set of large USGS aerial photos retrieved from TerraserverUSA

High-resolution surveillance imaging with apertures greater than a few inches over horizontal or slant paths at optical or infrared wavelengths will typically be limited by atmospheric aberrations. With static targets and static platforms, we have previously demonstrated near-diffraction limited imaging of various targets including personnel and vehicles over horizontal and slant paths ranging from less than a kilometer to many tens of kilometers using adaptations to bispectral speckle imaging techniques. Nominally, these image processing methods require the target to be static with respect to its background during the data acquisition since multiple frames are required. To obtain a sufficient number of frames and also to allow the atmosphere to decorrelate between frames, data acquisition times on the order of one second are needed. Modifications to the original imaging algorithm will be needed to deal with situations where there is relative target to background motion. In this paper, we present an extension of these imaging techniques to accommodate mobile platforms and moving targets.

The Solid-State, Heat-Capacity Laser (SSHCL), under development at Lawrence Livermore National Laboratory is a large aperture (100 cm{sup 2}), confocal, unstable resonator requiring near-diffraction-limited beam quality. There are two primary sources of the aberrations in the system: residual, static aberrations from the fabrication of the optical components and predictable, time-dependent, thermally-induced index gradients within the gain medium. A deformable mirror placed within the cavity is used to correct the aberrations that are sensed externally with a Shack-Hartmann wavefront sensor. Although it is more challenging than external correction, intracavity correction enables control of the mode growth within the resonator, resulting in the ability to correct a more aberrated system longer. The overall system design, measurement techniques and correction algorithms are discussed. Experimental results from initial correction of the static aberrations and dynamic correction of the time-dependent aberrations are presented.

The use of a Hartmann wave-front sensor to accurately measure the line-integrated electron density gradients formed in laser-produced and z-pinch plasma experiments is examined. This wave-front sensor may be used with a soft-x-ray laser as well as with incoherent line emission at multikilovolt x-ray energies. This diagnostic is significantly easier to use than interferometery and moire deflectometry, both of which have been demonstrated with soft-x-ray lasers. This scheme is experimentally demonstrated in the visible region by use of a Shack-Hartmann wave-front sensor and a liquid-crystal spatial light modulator to simulate a phase profile that could occur when an x-ray probe passes through a plasma. The merits of using a Hartmann sensor include a wide dynamic range, broadband or low-coherence-length light capability, high x-ray efficiency, two-dimensional gradient determination, multiplexing capability, and experimental simplicity. Hartmann sensors could also be utilized for wavelength testing of extreme-ultraviolet lithography components and x-ray phase imaging of biological specimens.

A high-speed Radon transform computer (RTC) has been designed and simulated that operates on the principles of the parallel projection pipeline engine (PPPE) processor suggested by J.L.C. Sanz, et al. (1988). The computer consists of an array of identical processing elements with fast communication links to each of its neighbors and uses forward- and back-projection formulas tests linear interpolation. Simulation and experimental timing tests have been performed on both a systolic array of 64 transputer processors and on arrays of commercial DSP (digital signal processor) chips. A custom VLSI implementation is currently under design which is expected to perform forward and inverse transforms of 512 by 512 images at or near video rates. >

The Imaging and Detection Program (IDP) within Laser Programs at Lawrence Livermore National Laboratory (LLNL) in cooperation with the Hughes Aircraft Company has developed a versatile, high performance, airborne experimental test-bed (AETB) capability. The test-bed has been developed for a wide range of research and development experimental applications including radar and radiometry plus, with additional aircraft modifications, optical systems. The airborne test-bed capability has been developed within a Douglas EA-3B Skywarrior jet aircraft provided and flown by the Hughes Aircraft Company. The current test-bed payload consists of an X-band radar system, a navigation system, a high-speed data acquisition, and a real-time processing capability. The medium power radar system is configured to operate in a high resolution, synthetic aperture radar (SAR) mode and is highly configurable in terms of waveforms, PRF, bandwidth, etc. Antennas are mounted on a 2-axis gimbal in the belly radome of the aircraft which provides pointing and stabilization. Aircraft position and antenna attitude are derived from a dedicated navigational system and provided to the real-time SAR image processor for instant image reconstruction and analysis. This paper presents a further description of the test-bed and payload subsystems plus preliminary results of SAR imagery.

We introduce a wave-optics based simulation code written to model a complete free space laser communications link, including a detailed model of an adaptive optics compensation system. We present the results obtained by this model, where the phase of a communications laser beam is corrected, after it propagates through a turbulent atmosphere. The phase of the received laser beam is measured using a Shack-Hartmann wavefront sensor, and the correction method utilizes a MEMS mirror. Strehl improvement and amount of power coupled to the receiving fiber results for both 1 km horizontal and 28 km slant paths will be presented.

Wave-front reconstruction with the use of the fast Fourier transform (FFT) and spatial filtering is shown to be computationally tractable and sufficiently accurate for use in large Shack–Hartmann-based adaptive optics systems (up to at least 10,000 actuators). This method is significantly faster than, and can have noise propagation comparable with that of, traditional vector–matrix-multiply reconstructors. The boundary problem that prevented the accurate reconstruction of phase in circular apertures by means of square-grid Fourier transforms (FTs) is identified and solved. The methods are adapted for use on the Fried geometry. Detailed performance analysis of mean squared error and noise propagation for FT methods is presented with the use of both theory and simulation.

Direct detection of photons emitted or reflected by an extrasolar planet is an extremely difficult but extremely exciting application of adaptive optics. Typical contrast levels for an extrasolar planet would be 109 - Jupiter is a billion times fainter than the sun. Current adaptive optics systems can only achieve contrast levels of 106, but so-called extreme adaptive optics systems with 104 -105 degrees of freedom could potentially detect extrasolar planets. We explore the scaling laws defining the performance of these systems, first set out by Angel (1994), and derive a different definition of an optimal system. Our sensitivity predictions are somewhat more pessimistic than the original paper, due largely to slow decorrelation timescales for some noise sources, though choosing to site an ExAO system at a location with exceptional r0 (e.g. Mauna Kea) can offset this. We also explore the effects of segment aberrations in a Keck-like telescope on ExAO; although the effects are significant, they can be mitigated through Lyot coronagraphy.© (2002) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Horizontal path laser communications are beginning to provide attractive alternatives for high-speed optical communications. In particular, companies are beginning to sell fiberless alternatives for intranet and sporting event video. These applications are primarily aimed at short distance applications (on the order of 1 km pathlength). There exists a potential need to extend this pathlength to distances much greater than a 1km. For cases of long distance optical propagation, atmospheric turbulence will ultimately limit the maximum achievable data rate. In this paper, we propose a method of improved signal quality through the use of adaptive optics. In particular, we show work in progress toward a high-speed, small footprint Adaptive Optics system for horizontal path laser communications. Such a system relies heavily on recent progress in Micro-Electro-Mechanical Systems (MEMS) deformable mirrors as well as improved communication and computational components. In this paper we detail two Adaptive Optics approaches for improved through-put, the first is the compensated receiver (the traditional Adaptive Optics approach), the second is the compensated transmitter/receiver. The second approach allows for correction of the optical wavefront before transmission from the transmitter and prior to detection at the receiver.

The development of high brightness and short pulse width (200 picoseconds) x-ray lasers now offers biologists the possibility of high-resolution imaging of specimens in an aqueous environment without the blurring effects associated with natural motions and chemical erosion. As a step toward developing the capabilities of this type of x-ray microscopy, a tantalum x-ray laser at 44.83 angstrom wavelength was used together with an x-ray zone plate lens to image both unlabeled and selectively gold-labeled dried rat sperm nuclei. The observed images show approximately 500 angstrom features, illustrate the importance of x-ray microscopy in determining chemical composition, and provide information about the uniformity of sperm chromatin organization and the extent of sperm chromatin hydration.

We have developed a high-resolution wavefront control system based on an optically addressed nematic liquid crystal spatial light modulator with several hundred thousand phase control points, a Shack-Hartmann wavefront sensor with two thousand subapertures, and an efficient reconstruction algorithm using Fourier transform techniques. We present quantitative results of experiments to characterize the performance of this system.© (2000) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Wavefront reconstruction techniques using the least-squares estimators are computationally quite expensive. We compare wavelet and Fourier transforms techniques in addressing the computation issues of wavefront reconstruction in adaptive optics. It is shown that because the Fourier approach is not simply a numerical approximation technique unlike the wavelet method, the Fourier approach might have advantages in terms of numerical accuracy. To optimize the wavelet method, a statistical study might be necessary to use the best basis functions or 'approximation tree.'

Sustained operation of high average power solid-state lasers currently requires an adaptive resonator to produce the optimal beam quality. We describe the architecture of a real-time adaptive control system for correcting intra-cavity aberrations in a heat capacity laser. Image data collected from a wavefront sensor are processed and used to control phase with a high-spatial-resolution deformable mirror. Our controller takes advantage of recent developments in low-cost, high-performance processor technology. A desktop-based computational engine and object- oriented software architecture replaces the high-cost rack-mount embedded computers of previous systems.

The wavefront controller for the Keck Observatory AO system consists of two separate real-time control loops: a tip-tilt control loop to remove tilt from the incoming wavefront, and a deformable mirror control loop to remove higher-order aberrations. In this paper, we describe these control loops and analyze their performance using diagnostic data acquired during the integration and testing of the AO system on the telescope. Disturbance rejection curves for the controllers are calculated from the experimental data and compared to theory. The residual wavefront errors due to control loop bandwidth are also calculated from the data, and possible improvements to the controller performance are discussed.

Results of experiments with the laser guide star adaptive optics system on the 3-meter Shane telescope at Lick Observatory have demonstrated a factor of 4 performance improvement over previous results. Stellar images recorded at a wavelength of 2 micrometers were corrected to over 40 percent of the theoretical diffraction-limited peak intensity. For the previous two years, this sodium-layer laser guide star system has corrected stellar images at this wavelength to approximately 10 percent of the theoretical peak intensity limit. After a campaign to improve the beam quality of the laser system, and to improve calibration accuracy and stability of the adaptive optics system using new techniques for phase retrieval and phase-shifting diffraction interferometry, the system performance has been substantially increased. The next step will be to use the Lick system for astronomical science observations, and to demonstrate this level of performance with the new system being installed on the 10-meter Keck II telescope.

The Keck AO system is nearing completion and is nearly ready for system integration in the optics laboratory at Keck Headquarters. because of the ambitious scientific objectives of this project, the AO system is quite complex in design. The optics bench alone has 33 motion control axes, and potentially 6 image planes. Calibration will play a major role in making the AO system function as planned. We have identified approximately 35 separate calibration procedures that will be needed to be performed during the integration phase in the laboratory. Many of these procedures are quite straightforward. Other calibrations are quite involved and are mini 'research projects' in themselves. It is this last category that will be discussed in this paper.

We will review the status of the natural/laser guide star adaptive optics facility that is being constructed for the Keck II telescope.

We have developed and tested a method for minimizing static aberrations in adaptive optics systems. In order to correct the static phase aberrations, we need to measure the aberrations through the entire system. We have employed various phase retrieval algorithms to detect these aberrations. We have performed simulations of our experimental setup demonstrating that phase retrieval can improve the static aberrations to below the 20 nm rms level, with the limiting factor being local turbulence in the A0 system. Experimentally thus far, we have improved the static aberrations down to the 50 nm level, with the limiting factor being the ability to adjust the deformable mirror. This should be improved with better control algorithms now being implemented.

The laser guide star adaptive optics system currently being developed for the Keck 2 telescope consists of several major subsystems: the optical bench, wavefront control, user interface and supervisory control, and the laser system. The paper describes the design and implementation of the wavefront control subsystem that controls a 349 actuator deformable mirror for high order correction and tip-tilt mirrors for stabilizing the image and laser positions.

Atmospheric turbulence severely limits the resolution of ground-based telescopes. Adaptive optics can correct for the aberrations caused by the atmosphere, but requires a bright wavefront reference source in close angular proximity to the object being imaged. Since natural reference stars of the necessary brightness are relatively rare, methods of generating artificial reference beacons have been under active investigation for more than a decade. In this paper, we report the first significant image improvement achieved using a sodium-layer laser guide star as a wavefront reference for a high-order adaptive optics system. An artificial beacon was created by resonant scattering from atomic sodium in the mesosphere, at an altitude of 95 km. Using this laser guide star, an adaptive optics system on the 3 m Shane Telescope at Lick Observatory produced a factor of 2.4 increase in peak intensity and a factor of 2 decrease in full width at half maximum of a stellar image, compared with image motion compensation alone. The Strehl ratio when using the laser guide star as the reference was 65% of that obtained with a natural guide star, and the image full widths at half maximum were identical, 0.3 arc sec, using either the laser or the natural guide star. This sodium-layer laser guide star technique holds great promise for the world's largest telescopes.

Our objective in this report is to develop methods to determine the output pupil wavefront using intensity measurements directly from the science detector. This wavefront can then be used to determine a reference wavefront which will precorrect for the non-common-path aberrations and produce the desired wavefront at the science detector. We describe two phase retrieval algorithms that can be used and a set of simulation studies of AO system calibration. We present the initial experimental results of applying this technique in calibration of the Lick Observatory laser guidestar AO system in a later paper.

A sodium-layer laser guide star adaptive optics system has been developed at Lawrence Livermore National Laboratory (LLNL) for use on the 3 meter Shane Telescope at Lick Observatory located on Mount Hamilton, near San Jose, California. The system is based on a 127-actuator continuous-surface deformable mirror, a Hartmann wavefront sensor equipped with a fast-framing low-noise CCD camera, and a pulsed solid-state-pumped dye laser tuned to the atomic sodium resonance line at 589 nm. Figure 1 shows a picture of the laser beam being projected from the dome of the Shane Telescope.

A sodium-layer laser guide star adaptive optics system has been developed at Lawrence Livermore National Laboratory (LLNL) for use on the 3-meter Shane telescope at Lick Observatory. The system is based on a 127-actuator continuous-surface deformable mirror, a Hartmann wavefront sensor equipped with a fast-framing low-noise CCD camera, and a pulsed solid-state-pumped dye laser tuned to the atomic sodium resonance line at 589 nm. The adaptive optics system has been tested on the Shane telescope using natural reference stars yielding up to a factor of 12 increase in image peak intensity and a factor of 6.5 reduction in image full width at half maximum (FWHM). The results are consistent with theoretical expectations. The laser guide star system has been installed and operated on the Shane telescope yielding a beam with 22 W average power at 589 nm. Based on experimental data, this laser should generate an 8th magnitude guide star at this site, and the integrated laser guide star adaptive optics system should produce images with Strehl ratios of 0.4 at 2.2 {mu}m in median seeing and 0.7 at 2.2 {mu}m in good seeing.

A prototype adaptive optics system has been developed at Lawrence Livermore National Laboratory (LLNL) for use on the 3-m Shane telescope at Lick Observatory. This system is currently based on a 127-actuator continuous-surface deformable mirror developed at LLNL, a high-quantum-efficiency low-noise fast CCD camera built for LLNL by Adaptive Optics Associates using a chip developed by Lincoln Laboratory, and a Mercury VME board containing four Intel i860 processors.

Ultra high resolution three dimensional images of a microscopic test object were made with soft x‐rays using a scanning transmission x‐ray microscope. The test object consisted of two different patterns of gold bars on silicon nitride windows which were separated by ∼ 5μm. Depth resolution comparable to the transverse resolution was achieved by recording nine 2‐D images of the object at angles between −50 to +55 degrees with respect to the beam axis. The projections were then combined tomographically to form a 3‐D image using an algebraic reconstruction technique (ART) algorithm. We observed a transverse resolution of ∼1000 A. Artifacts in the reconstruction limited the overall depth resolution to ∼6000 A, however some features were clearly reconstructed with a depth resolution of ∼1000 A.

We will describe an adaptive optics system developed for the 1 meter Nickel and 3 meter Shane telescopes at Lick Observatory. Observing wavelengths will be in the visible for the 1 meter telescope and in the near IR on the 3 meter. The adaptive optics system design is based on a 69 actuator continuous surface deformable mirror and a Hartmann wavefront sensor equipped with an intensified CCD framing camera. The system has been tested at the Cassegrain focus of the 1 meter telescope where the subaperture size is 12.5 cm. The wavefront control calculations are performed on a four processor single board computer controlled by a Unix-based system. We will describe the optical system and give details of the wavefront control system design. We will present predictions of the system performance and initial test results.

The architecture and major system components of the sodium-layer laser guide star system at LLNL will be described, and experimental results reported. The subsystems include the laser system, the beam delivery system including a pulse stretcher and beam pointing control, the beam director, and the telescope with its adaptive-optics package.

We present the design and implementation of a very compact adaptive optics system that senses the return light from a sodium guide-star and controls a deformable mirror and a pointing mirror to compensate atmospheric perturbations in the wavefront. The deformable mirror has 19 electrostrictive actuators and triangular subapertures. The wavefront sensor is a Hartmann sensor with lenslets on triangular centers. The high-bandwidth steering mirror assembly incorporates an analog controller that samples the tilt with an avalanche photodiode quad cell. An f/25 imaging leg focuses the light into a science camera that can either obtain long-exposure images or speckle data. In laboratory tests overall Strehl ratios were improved by a factor of 3 when a mylar sheet was used as an aberrator. The crossover frequency at unity gain is 30 Hz.