Dropbox to the Rescue (Again)

A Convenient Method for adding Attachments to Wiki Pages

I just can't say enough about Dropbox.

I'm a heavy user of Wetpaint for my classroom and group wikis. One of my courses in software development requires that we pass a large number of code snippets around and while attachments to each wiki page works, it has some minor drawbacks:

• Wetpaint accounts (both and pay and free) limit the total number and size of attachments (understandably so...)
• Attachments appear at the very bottom of the wiki page, below the content and page discussion threads

Again, not a deal killer, just would be nice if we could mention a file in the body of the page, include a hyperlink in line with the text and continue with the prose.

Dropbox "Public Files" to the rescue. Simply:

• Copy the file you wish to share into the "Public" folder that was created in your Dropbox folder at install
• Right click (or control click) on the file to pop up the options menu
• Click on the Dropbox submenu
• Click on the "copy public link" menu item
• Highlight your link text in your blog or wiki and paste the public link
• Presto... all done

The public link to the "Top Secret.txt" demo file created in my Dropbox install appears below:

http://dl.getdropbox.com/u/141947/Top%20Secret.txt

After creating a link you can grab the Top Secret.txt file by clicking on the linked filename in this sentence.

The link is to the copy of the file that lives on the dropbox servers, so the link to your actual computer is still safe behind your log in and password.

I continue to be amazed...

You Can't Handle the Truth

Why American Universities offer a Ph.D. Degree in Analytical Chemistry

During the first day of my Analytical Chemistry and Instrumentation and Measurement courses my students and I cover the following facts:

• Economics is the study of Wealth
• Art is the study of Membership
• Politics is the study of Power
• Religion is the study of Values
• Science is the study of Truth

and we review the following definitions:

• Accuracy is how close measurements are to the Truth
• Precision is how close measurements are to Each Other

So the rest of the courses are conceptually simple:

Design and make measurements with the utmost care such that the results represent our current best understanding of the Truth.

This is why my professional sensibilities have been so bruised by the National Snow and Ice Data Center (NSIDC) and their February 18, 2009 comments concerning the drifting calibration of a sensor measuring the extent of Arctic Sea Ice.

The NSIDC publishes a daily image of the Arctic Sea Ice collected through the process below:

"NSIDC gets sea ice information by applying algorithms to data from a series of Special Sensor Microwave/Imager (SSM/I) sensors on Defense Meteorological Satellite Program (DMSP) satellites. These satellites are operated by the U.S. Department of Defense. Their primary mission is support of U.S. military operations; the data weren’t originally intended for general science use."

The modeled data is compared to independent measurements made by the NASA Earth Observing System Advanced Microwave Scanning Radiometer (EOS AMSR-E) sensor. Beginning in early January 2009, the SSM/I data measured a rapidly shrinking amount of sea ice (500,000 square kilometers (193,000 square miles) an area roughly equal to the size of California). After a closer look at the data, NSIDC discovered the SSM/I sensor had experienced calibration drift and was generating values that were low by close to 200,000 square miles.

I'm not surprised by calibration errors: measurement specialists are trained to find and correct them. What I am surprised by is the statement from the NSIDC concerning their use of a measurement instrument that incorporates a sensor with known baseline drift problems:

"Some people might ask why we don't simply switch to the EOS AMSR-E sensor. AMSR-E is a newer and more accurate passive microwave sensor. However, we do not use AMSR-E data in our analysis because it is not consistent with our historical data. Thus, while AMSR-E gives us greater accuracy and more confidence on current sea ice conditions, it actually provides less accuracy on the long-term changes over the past thirty years. There is a balance between being as accurate as possible at any given moment and being as consistent as possible through long time periods. Our main scientific focus is on the long-term changes in Arctic sea ice. With that in mind, we have chosen to continue using the SSM/I sensor, which provides the longest record of Arctic sea ice extent."

So the measurement folks at NSIDC have chosen "precision" over "accuracy", or as I tell my freshmen, choosing "the wrong answer consistently" over "the truth".

I wouldn't mind so much if I could simply pen a note to some obscure scientific journal reminding folks of the basics, but the problem is that the Political/Power folks use this data to prove the Arctic has lost snow and ice cover the size of California in as little as one month. I sure hope everybody knows how to swim.

What a Tangled Web

Functional Quantum Nodes for Entanglement Distribution
Contributed editorial appearing in
Scientific Computing 24:10, September 2007, pg. 20.

Our course in Laboratory Informatics continues on from a discussion of transducers, response models and calibration to a look at the systems and methods utilized for the collection of values from the many samples and disparate instruments throughout the laboratory and across the entire organization. The workhorse technology used to enable this ability is the laboratory network. The time spent generating a table of contents and cross-index for a paper notebook, producing photocopies and delivering them to an archive room to be filed in a dusty cabinet are thankfully joining the bygone days of pipetting by mouth. Electronic records, online data analysis, relational databases and other life-blood technologies routinely used by the business side of the organization are finding standardization and acceptance in the laboratory. Additionally, the networking and information technologies benefit from advances in the physical laboratory as well. This symbiotic relationship involves the integration of advances in energy, material and information into high-level systems that provide innovative solutions toward the common goals of bigger, faster, safer, cheaper, and the elusive – more accurate. An initial link between laboratories at the University of California Los Angeles (UCLA) and the Stanford Research Institute of Stanford University in 1969 evolved into the modern Internet through the addition of nodes, new technologies and standards developed and adopted by the growing number of members within the network. While the speed and size of the digital Internet continues to grow, laboratories have been working diligently on new systems based on research into quantum computing and quantum information science with the promise of increased performance and new capabilities.

A January 2005 SC article looked at emerging experimental research into quantum computing and recent major advances continue to demonstrate practical uses of the technology. Two primary differences between quantum computing and “classical” computing involve the quantum concepts of superposition and entanglement. In a classical system, a single value is represented by a large collection of individual components through the use of statistics. For example, let’s consider the average result of flipping 10 unbiased coins. We assign “heads” a value of 1 and “tails” a value of 0, flip each coin into the air and then record the outcome of the flip after it comes to rest. The statistical value of the arithmetic mean for the entire process can be obtained by summing all ten results and dividing the sum by the “degrees of freedom” (DOF) of the measurement. In this case, the DOF is 10 (generically, n, the number of flips) as each coin was “free” to produce a result of 1 or 0 during its flip. The probability of producing a 1 or 0 is “superpositioned” or spread across all ten coins by dividing the sum by the DOF. We can calculate the average difference or “variance” of the measurement by comparing the difference between each of the ten results and our calculated mean (we will need to square each difference to prevent negative values from destructively interfering with our calculation). As before, we calculate the average by summing the squared differences and dividing by the DOF. However, in this case the DOF is down to 9 (generically, n-1). Since we need to know the arithmetic mean to calculate each difference, we already have information about all ten flips. After we know the results of the first nine, we can use this data and the mean to determine the value of the tenth flip – it does not have any freedom to be 1 or 0; it must have the value dictated by the mean. The tenth flip is entirely correlated or “entangled” with the mean and it must assume a specific value based on the values of the other nine flips and the mean.

Classical measurements use these statistical tools to obtain information from large collections of items. But what happens when the number of items is reduced down to one or two? In the case of a single coin flip, the equal probability of being heads or tails “collapses” into a single value when measured. Before it is measured, the superposition of heads and tails can be visualized as the single coin is flipping though the air. Classical intuition says it can be thought of as rapidly oscillating between 1 and 0 or having both values simultaneously. A classical example of entanglement, however, would involve flipping two coins simultaneously. After the first coin has stopped and is measured, the second coin must always land with the opposite value to maintain the probability of 0.5. Our statistic calculation of variance also predicts this two-element system has one Degree of Freedom (n-1) and the second coin has no chance of having the same value as the first coin. Albert Einstein famously derided this classical analogy as “spooky action at a distance” and along with Boris Podolsky and Nathan Rosen posed what is known as the EPR Paradox in a 1935 paper that argued that something was missing from quantum theory.

Several theoretical approaches have been proposed to dispel the EPR Paradox and the 2001 approach described by collaborating researchers Luming Duan at the University of Science and Technology of China, Mikhail Lukin at Harvard University, and Juan Ignacio Cirac and Peter Zoller of the Universität Innsbruck (known as the DLCZ Protocol) has been demonstrated experimentally by Professor H. Jeff Kimble and his research group at the California Institute of Technology (Caltech) this past Spring. Like the pair of spinning coins described above, the DLCZ Protocol uses a pair of Cesium (Cs) atoms. Since it is experimentally difficult to work with single atoms, a pair of ensembles (each containing around 100,000 cooled Cs atoms) is used to increase the probability of interacting with a single Cs atom. The pair is first conditioned by ensuring all of the Cs atoms are in their ground state. Then a single photon is introduced into each ensemble with a chance it will collide with one of the Cs atoms through inelastic scattering (i.e., non-resonant Raman Scattering) producing a single Cs atom in a different ground state that can be designated as a digital “1”. This “writing” event can be confirmed by detecting the single Raman-scattered photon. There is also a chance the write photon will not interact with the Cs atoms, leaving the ensemble as a digital “0”. An entangled pair of ensembles is produced when one ensemble scatters the write photon and the other does not, yielding a “1” and “0” pair referred to as a “quantum node”. We can’t know which of the pairs has scattered the photon or that would constitute as a measurement and the entanglement would collapse. To keep this information unknown, the output of each ensemble is directed to a 50/50 beam splitter that is observed by two, single-photon detectors. If neither ensemble scatters a photon, nothing is detected and the pair is measured to be 0-0. If two scattered photons are detected, the pair is measured to be 1-1. If only one scattered photon is detected, then we have measured 1-0 or 0-1, depending on which ensemble did the scattering. Because of the 50/50 beam splitter, each single-photon detector is observing both ensembles simultaneously, so there is no way of knowing which ensemble scattered the photon and the pair remains superpositioned and entangled. Professor Kimbles’ group also prepared a second quantum node separated 3 meters from the first. Using photons having the same frequency as the Raman-scattered photons produced by the write pulse, the ensembles can be read by detecting the anti-stokes Raman-scattered photons produced by the Cs atom as it returns to it’s initial ground state. With the use of additional 50/50 beam splitters, the 1-0/0-1 superpositioned state of the first quantum node can be written to the second node without knowing which node is which. While there are several major technological hurdles to overcome before quantum networks become practical, this project demonstrates the ability to use classical data acquisition methods for the creation of functional quantum systems.

Image: First two nodes of the elementary quantum network at Caltech. The values of the initial quantum node (blue) are transferred to the remote quantum node (green). Photo by Nara Cavalcanti.

Related Web Resources

January 2005 Scientific Computing article “Change for the Better”
http://www.scientificcomputing.com/PRArchivebyIssue.aspx?RELTYPE=FE&YEAR=2005&MONTH=01&CommonCount=0

Quantiki – Wiki for Quantum Information Science
http://www.quantiki.org/wiki/index.php/Main_Page

Quantum Optics Group at California Institute of Technology
http://www.its.caltech.edu/~qoptics/home.html

Caltech Press Release on Entanglement Distribution
http://www.its.caltech.edu/~qoptics/Ensemble/index.html

The EPR Paradox in Quantum Theory
http://plato.stanford.edu/entries/qt-epr/

Quantum Optics and Quantum Information Group at University of Michigan
http://www-personal.umich.edu/~lmduan/

Quantum Optics Group at Harvard University
http://lukin.physics.harvard.edu/

Theory Group at Max Plank Institute for Quantum Optics
http://www.mpq.mpg.de/Theorygroup/CIRAC/wiki/index.php/Theory_Division.html

Quantum Optics Theory Group at the Universität Innsbruck
http://www.uibk.ac.at/th-physik/qo/index.html

Wave Which Way

Sensor Technology for Directional Underwater Sound
Contributed editorial appearing in
Scientific Computing 24:8, July 2007, pg. 14.

The American Philosophical Society (APS) was founded by Benjamin Franklin in 1743 and stands as the oldest learned society in the United States. Philadelphia native David Rittenhouse served as the society’s president from 1791 to 1796, after holding positions as Professor of Astronomy and Vice-Provost at what is now known as the University of Pennsylvania. In 1786, Rittenhouse published an article in the society’s journal, Transactions of the APS, titled, “Explanation of an optical deception.” An expert optical tool maker, Rittenhouse had strung fine hairs between the grooves of two co-linear screws and, by doing so, had invented a crude diffraction grating that spread out the colors of light when one looked through it. After Rittenhouse abandoned the device as an optical oddity, German physicist Joseph von Fraunhofer later reinvented the device when he strung thin wire between the grooves of two screws in 1813 and the diffraction grating has enjoyed continuous improvement ever since.

Dutch scientist Christiaan Huygens had proposed a model for light diffraction almost a century earlier, but it took the works of Fraunhofer, Thomas Young and Augustin-Jean Fresnel to reveal the importance of the wave model of light. Before that time, Sir Isaac Newton had proposed that light was a collection of different-colored particles (a forerunner of our modern model of the photon) that flew through the air in straight lines or rays and reflected from surfaces at predictable angles. With the advent of quantum mechanics in the mid 1800s, it was discovered that both models, the particle-ray model and the wave model, are equally valid ways to view light propagation. The choice of which way to model the propagation often depends on the interactions being observed. Reflection, absorption and scattering are described easily by the ray model, while the wave model is invoked to explain refraction, diffraction and dispersion. These optical models also are applicable to the motion of solids, liquids and gases. The classical Newtonian particle model is used most often to describe the kinematics of solids whereas the wave model is used to describe fluid motion.

Waves themselves can be viewed in different ways depending on their type. Light waves are viewed as primarily transverse since their electric and magnetic fields oscillate side-to-side while the light travels forward. Sound waves in air are modeled as primarily longitudinal since regions of high and low pressures oscillate along the direction in which the wave is traveling. A microphone is used to measure these high-frequency changes in air pressure and to record the sound waves. Underwater vibrations also are modeled as longitudinal pressure waves and can be detected using a similar hydrophone. However, surface waves are obviously transverse. This up and down movement of water is known as shear deformation and can be measured by observing the height and direction of the wave.

Interest in measuring underwater sound was sparked by the sinking of the Titanic in 1912 and the need to detect submarines during World War I. Unfortunately, those prevailing reasons are still with us today. Modern hydrophones can be very sensitive but, because of their reliance on longitudinal pressure waves, are omnidirectional. Hydrophones are arranged in both 2-D and 3-D arrays such that the arrival time of the pressure wave to each sensor can be triangulated to pinpoint both distance and direction to the source of the sound wave. According to the team of Research Engineer Francois Guillot, Professor Peter Rogers and Research Scientist David Trivett at the Georgia Institute of Technology, the U.S. Navy routinely tows hydrophone arrays that are thousands of feet long in order to obtain the desired directional resolution. Working under a grant from the Office of Naval Research, the Georgia Tech team developed a prototype underwater sound sensor able to measure both sound intensity and direction. The new device is sensitive to direction because it is based on the detection of the sound wave’s transverse shear deformation, much like detecting waves on the surface. The device includes a small paddle made of special composite material that has the same density as seawater. The paddle is attached to the main housing by a hinge that permits it to oscillate back and forth as it flutters with the passing wave.

The magnitude of the transverse shear deformation is extremely small and the team needed a sensitive way to measure the paddle vibration. Their method is based on the “optical deception” discovered by Rittenhouse. When light passes through an array of thin parallel obstructions, its waves diffract around them and the resulting constructive and destructive interference spatially separates the wavelengths. In 1978, Dr. Kenneth O. Hill, a scientist at the Communications Research Centre (CRC) in Ottawa, published a method for creating an array of parallel lines within the core of an optical fiber. Named after 1915 Nobel Laureate William Lawrence Bragg, Hill’s “fiber Bragg grating” transmits specific light frequencies while reflecting others back through the fiber. The frequency of the reflected light changes as the line spacing varies due to mechanical strain or temperature effects on the fiber’s linear expansion. The Georgia Tech team attached a single fiber containing two Bragg gratings to their sensor; one situated on the paddle and the other on the main housing. As the paddle oscillates, the distance between the gratings changes and its effect on the reflected light is monitored by a photodetector. While sensitive to its orientation with the wave, an array of these sensors is again needed to pinpoint direction and range of its source; however, the team suggests an array of these sensors will be more than five times smaller, easier to handle and less costly to operate. These are important improvements no matter which way you look at it.

Image: Using optical fibers, researchers at the Georgia Institute of Technology have created a sensor that detects the direction from which a sound is coming under water — an important improvement over current technology. Image courtesy Georgia Tech (Rob Felt).

Photographic Memory

Web Searching with Object Instance Recognition
Contributed editorial appearing in
Scientific Computing 24:7, June 2007, pg. 13.

Toward the end of last summer, LEGO introduced the latest incarnation of its most successful product line, the LEGO Mindstorms NXT Robotics Toolset. Complete with a 32-bit microcontroller, three rotational actuators, and four sensors for touch, sound, light and distance, these kits have become the foundation of our undergraduate Intelligent Systems course as well as my 11-year-old son’s favorite birthday present. Unlike traditional LEGO models, the Mindstorms are bundled with programming software used to animate the creations. While there is something to be said about the amount of imagination required to zoom around the house with a large-scale Rebel X-Wing fighter in hand, the prospect of designing your own X-Wing with actuating wings and blinking laser blaster complete with sounds sampled from the Star Wars DVD, permits an even greater opportunity to exercise additional areas of the designer’s brain.

And therein lies the most important lesson of the NXT Toolset — it does not include a biological brain. The instructions provide the steps necessary to assemble a humanoid robot and program it to perform some impressive movements, but breathing life into the creations requires good old imagination or the help of some special effects. The outer shell of the binocular ultrasonic range sensor is reminiscent of the 1986 film Short Circuit and it is our brain that infers that it is complete with stereo color vision, a photographic memory and object recognition. Alas, a quick glance at the schematic reveals it simply contains a collinear transmitter/receiver pair and a timing circuit that calculates the distance between itself and a planar object placed in its field of view. This is great stuff to a sensor enthusiast, but the robot has little chance of recognizing the face of its creator.

Researcher Larry Zitnick, currently with the Interactive Visual Media Group at Microsoft Research, is working on the development of systems capable of more than range finding. As a recent graduate of the Robotics Institute at Carnegie Mellon University, Dr. Zitnick is an expert in image analysis and has set to the task of developing algorithms having the ability to locate visual objects and recall their identity. The technology is being realized in a research prototype called “Lincoln” (available at lincoln.msresearch.us) that allows its users to search the Web for information about an object by snapping a color photograph of it with an Internet-enabled camera phone or PDA. Lincoln analyzes the digital photograph for words and printed images such as those appearing on a DVD jacket or in a magazine. The system then compares these features to those contained in a cataloged image database and, if found, the image’s keywords are used as the basis of a Web search. At the moment, the system is designed to work with books, magazines, posters, paintings and labeled products. In practice, the user sees an object of interest, snaps its picture, and Lincoln returns with Web information such as a product Web site, price or concert venues. Especially when using the small keypad of a PDA or the number pad of a Smartphone, the system saves wear and tear on the user’s thumbs.

There is no small amount of behind-the-scenes technical wizardry required for the handheld electronic device to display such a high functionality. The algorithmic brain of the system must extract image features and compare them to features in a database of images. As with all measurement systems, the immediate problem is one of calibration. The image of the object found in the “wild” is very unlikely to have been acquired under the conditions of lighting, resolution and perspective used to capture the image in the database. Similar to the triangulation used by the Global Positioning System to determine spatial location, Lincoln uses a triplet of three neighboring features within the image to compensate for location, magnification and perspective. The 2-D position of each vertex of the wild image triplet can be transformed mathematically to compensate for image translation, rotation and magnification and then compared with the canonical version of the image contained in the database. The problem becomes computationally large as the number of unique image triplets increases. Similar objects, such as people in the same family, require a large number of fine details in order to tell them apart. Even our brains are often confused by identical twins. For this reason, Lincoln does not work with people, plants or pets. Until our technology matures enough to include the production of biological neural network computers, our current maxim will hold true: A small amount of electronic ability requires an enormous amount of human imagination.

Now You See It

Modern Advances in Steganalysis
Contributed editorial appearing in
Scientific Computing 24:6, May 2007, pg. 18.

The United States during the 1970s is often portrayed by archive footage of long gas lines, leisure suits and discotheques with flashing floors. While the mainstream boomers were perfecting the use of cheesy pickup lines, others, including names like Bushnell, Gates, Jobs and Lucas were quietly developing the digital information revolution. Even though their efforts were wildly public, the importance of their advances was often eclipsed by the short-term political or cultural story of the day. Thankfully, my personal recollection is not rife with polyester and hairspray but of simpler things like elementary school, Star Wars and Pong. I especially recall my short-lived stint as a magician. Each appearance in the annual grade-school talent show would afford an opportunity to show off my most recent dazzling card and rope tricks. The thrill did not come from learning the mechanics of the trick, but from convincing the audience there was real magic in play. In addition to some fancy slight of hand, the real “trick” was in distracting audience attention to a grandiose, yet inconsequential movement while deftly pocketing a card or tucking a foam ball in the opposite hand.

The art of the feint has been in use for centuries but it is no less effective today. A common approach to making things disappear is to bury them under something obvious. In the case of hidden writing and messages, this technique is known as “steganography.” The Greek prefix “stego-“ causes stegosaurus to translate into “roof-lizard”, while the Greek prefix “stega-“ leads steganography to translate into “covered-writing”; however, both prefixes appear to be used interchangeably as shorthand terms for steganographic products. Unlike cryptography wherein messages are specifically encoded by a private sequence of symbols, steganographic writing and images appear as they should, but are difficult to see at first glance. When hiding digital images, it is common to replace the least-significant bits (LSBs) of a cover image with the bits of a secret image. In the image, titled “Lightening Jars”, the LSBs of each JPEG-encoded pixel have been replaced by those of embedded image of the Burlington, Vermont airport. The hidden image can be retrieved by processing the “stego-image” through a recovery algorithm.

Hiding messages in digital image, sound and video files is uncomplicated and can be achieved through the use of several free programs and web applications. The difficult problem is one of detecting hidden images in publically-transmitted media even though the method used to embed the images may be unknown. Termed “steganalysis”, local, federal and military agencies are particularly interested in knowing when digital files contain covert, illegal information. Professor Jessica Fridrich and her research team at SUNY Binghamton are especially active in the area of digital forensics and steganalysis. A simple scanning method would be to process a suspect image through a group of known recovery algorithms. While this counter measure may have been successful in the early days of digital steganography, the “bad guys” quickly combined their efforts with cryptography to hide an encoded message that cannot be easily recovered without a secret passkey. The problem then shifted to one of establishing the presence of any hidden payload within a digital file that can be handed over for decryption. One approach to payload discovery is to analyze the LSBs of a digital image to determine how well they “belong” to the rest of the image. A very “busy” cover image containing intricate details and sharp-cornered objects would be expected to have a specific level of high spatial-frequency components located in the LSBs of its digital file. A hidden payload that did not share the same frequency distribution would appear out of place. A close facsimile of the cover image’s expected high-frequency distribution can be acquired through analysis of the next-to-least significant bits, however, the bad guys may have modified them as well. Emerging intelligent classification algorithms trained on collections of similar images can be taught to analyze the complete frequency spectrum for anomalous components. While this cat and mouse game of give-and-take leads to vigorous research in counter-countermeasures, a most elegant solution involves the universal deployment of steganography for the technique of “digital watermarking.” In this process, encrypted information is intentionally embedded into digital files when they are created by the digital camera, camcorder, audio-recorder or imaging editing software. After the digital file has been distributed, steganalysis can be used to recover the watermark data to determine if the file has been altered. In addition to making the concealment of digital payloads extremely complicated, individual serial numbers can be included in the digital watermark to assist digital forensic investigators in cases involving terrorism, media piracy and child pornography.

Toys for Dots

Improved Nanodots for Data Storage
Contributed editorial appearing in
Scientific Computing 24:5, April 2007, pg. 12.

Every now and again, a sentiment emerges from a collection of varied sources that is inserted into a folder titled “Common Sense.” Micromanagement is one such term that burst into popular jargon along with phrases like paradigm shift and thinking outside the box. Rather than appearing in boardroom presentations, micromanagement is often heard around the water cooler, where it is used as a pejorative term by employees desiring more autonomy. In a society still evolving from a mechanistic industrial revolution, I view this term as the positive result of effective employee education and training. While upper management shepherds the big picture, the boots on the ground are best equipped to manage the particulars.

This approach works wonders in intelligent human systems, but often fails when applied to mindless machines. A misplaced semicolon or misspelled variable name can wreak havoc on the most sophisticated computer code. Attention to detail and vigorous micromanagement are prized expertise in physical and computer scientists. An example of these skills appearing earlier this year deals with the topic of magnetic data storage. Modern computer memory, known as random access memory (RAM), is available in two popular flavors. Dynamic RAM (DRAM) is based on electronic capacitors having charges that must be refreshed periodically, and static RAM (SRAM) uses electronic transistors that maintain their value once written. While both forms of RAM are very fast, they lose their values when power is removed. Slower, but longer-term, non-volatile memory is often provided by magnetic storage media commonly in the form of a hard disk drive (HDD). Pioneered by IBM in the mid-1950s, the first IBM-350 HDD contained 50 24-inch disks coated with ferromagnetic iron oxide, and had a storage capacity of 4.4 megabytes. Individual bits were recorded by magnetizing a section of the disk in a particular direction through exposure to a strong magnetic field. After the magnetic field was removed, each ferromagnetic bit would maintain its orientation, and could be measured by an additional sensor. Today’s HDDs boast capacities of several hundred gigabytes in form factors you can hold in your hand, but they are based on the same fundamental design of the IBM-350. These drives use thin films of cobalt (Co) alloy instead of iron oxide and are formatted into specific regions of tracks, sectors and bits at the factory. Improvements in storage capacity are achieved by shrinking the area required for each bit and increasing the sensitivity of the read sensors. However, at the sub-micrometer level, the magnetic fields of adjacent bits begin to interfere, and this presents a technological obstacle to further miniaturization.

One solution to this problem is to use electron beam lithography to fabricate individual stacks of Co alloy, known as nanodots. These nanodots are typically less than 100 nm in diameter, are tens of nanometers tall, and are patterned on silicon (Si) wafers in rectangular arrays with controlled spacing between the dots. Cobalt-palladium (Co/Pd) nanodots can be magnetized perpendicular to the plane of the Si base, much like an upright bar magnet. Sensing whether the magnetized dot is displaying its north or south magnetic pole to the read sensor provides a sharp signal, indicating a bit value of 1 or 0 that is isolated from neighboring dots. Unfortunately, initial studies of nanodots revealed a large variation in magnetic field strength needed to flip each dot’s magnetic orientation. Known as the switching field distribution (SFD), this limits how close the dots can be packed. If some dots require stronger field strength, they could fail to switch, or the increased strength could affect neighboring dots that switch more easily. Some scientists suggested the polycrystalline nature of traditional thin films was the source, while others pointed to variations in the lithography process.

Researchers at the National Institute of Standards and Technology (NIST) in Boulder, CO, along with colleagues at the University of Arizona, have fabricated various nanodot arrays using strictly controlled methods in an attempt to find the source of the large SFDs. Polycrystalline Co/Pd nanodots were fabricated using dc magnetron sputtering and single-crystal Co/Pd nanodots were grown using the more complicated process of molecular beam epitaxy (MBE). They found that the single-crystal nanodots exhibited similar SFDs as compared to the polycrystalline versions, suggesting grain boundaries and orientations were not a major contributor. However, their major discovery is that the SFD can be reduced significantly with the addition of Tantalum (Ta), a material whose main use is in the production of electronic capacitors, as a layer inserted between the nanodot and the Si wafer. Additional studies at this fundamental level can lead to the production of magnetic storage media that have densities greater than 1 terabit/in2, enabling a new generation of storage devices. As our nanotechnology continues to increase in sophistication, we will know our electronic micromanagement is effective when our intelligent electronic systems begin to complain about “nanomanagment.”

Image: False-color image of 50-nanometer cobalt-palladium nanodots made with a magnetic force microscope provides both topographic and magnetic profiles. The darker dots are magnetized in the up direction (representing 1 in binary code) and the lighter dots are pointing down (representing 0). Image courtesy of Justin Shaw.

A Scanner Darkly

Acquiring a Signature from Dark Matter
Contributed editorial appearing in
Scientific Computing 24:4, March 2007, pg. 13.

“If a tree falls in the middle of the woods and nobody is around to hear it, does it make a sound?” That’s one of those crowd pleasers that usually arrive after the third or fourth round of adult beverages. An even more interesting question, best debated sober, is, “If you hear a tree fall and there are no trees around, what made the sound?” Of course the correct answer is: “An invisible tree.” Illusionists make a good living enabling large objects to disappear, such as jet airplanes and statues of liberty. Rational audience members have no trouble stating, “It’s still there, I just can’t see it.” Astronomers have been using their own invisible tree for over 70 years, albeit under a different moniker: the existence of “dark matter.”

The ability to augment our sense of sight through development of the microscope and the telescope enabled giant leaps in science and our understanding of the nature of reality. These tools provide an opportunity to test our models and theories in both small and large environments. The extremely small realm of quantum physics often exhibits features and behaviors that run counter to our macroscopic Newtonian world. Observations of distant galaxies and clusters of galaxies permit scientists to verify our models of mass, velocity, acceleration and force on a very large scale. Instruments including the space-based Hubble Space Telescope (HST) and Chandra X-ray Observatory facilitate study of the composition and kinematics of objects too impossibly large to fit into a terrestrial chemistry or physics laboratory.

Analogous to a collection of marbles resting on the floor of a spinning merry-go-round, the giant stars contained within a swirling galaxy must have centripetal forces preventing them from careening off into space. Using calibrated models for the intensity of observed light, astronomers can predict the strength of gravitational forces needed to keep the rotating stars in formation. Surprisingly, the model suggests the galaxies should be more massive than they appear. Instead of overhauling our Newtonian models, astronomers very pragmatically suggested that a majority of the required mass is simply invisible to our light-gathering telescopes. In addition to no reflection or emission of electromagnetic radiation, it is theorized this dark matter possesses other exotic properties including collisionless interaction with normal, visible matter — akin to a ghost walking through walls. Without this property, we would be able to track dark matter from its effect on visible matter.

So, the task at hand is to design an experiment that tests for the existence of dark matter. While making visible matter disappear is the subject of vigorous research into stealth technologies, detecting the already invisible is a nifty little problem. Perhaps our ability to detect dark matter is cloaked by the interfering signal from visible matter. This is a classic problem in analytical chemistry and one that has resulted in the prominence of separations science. One way to find a needle in a haystack is to remove all of the hay. While this glib solution works well for laboratory chromatography and distillation systems, astronomers began the search for a “naturally separated” galaxy cluster to examine. They found one in the merging cluster 1E0657-56, nicknamed the “Bullet Cluster.” It is actually two clusters of galaxies, gasses and plasma found about four billion light years from Earth that collided around 100 million years ago. Because the stars contained in the galaxies are separated by large distances, they passed by each other undisturbed, in a manner similar to the coordinated dance line choreography of the Rockettes. However, the dense gasses and plasma collided much like a bullet fired through a cloud of smoke and resulted in the cluster’s descriptive epithet. The stars continued on their way while the gasses and plasma were held back by the collisions. Intensity measurements predict most of the visible mass should be found in the trailing gas cloud. According to its theorized collisionless behavior, the dark matter should be found traveling along with the swarm of exposed stars.

The next step in the solution is to measure the total amount of gravity displayed by the cluster through the use of gravitational lensing. Similar to the optical distortions induced by variations in air density on a hot summer day, images of galaxies behind the cluster are slightly distorted as their light passes through regions of strong gravitational fields. The amount of image distortion is quantified by comparing the images to those acquired from galaxies outside the effect of the gravitational lens in the same region of space. A contour map can be generated that represents the areas of the strongest gravitational fields and, by inference, the highest concentrations of mass. After comparing the gravitational lens data with intensity images captured by the HST and Chandra, it was found the regions of highest mass density for each passing cluster are located in front of the trailing luminous clouds of gas and plasma, thereby providing direct evidence for the existence of dark matter. This discovery was enabled through the use of very sophisticated telescopes in the hands of talented scientists. I find no shortage of poetry in the fact that, like their illusionist counterparts, making the invisible visible requires the use of smoke and mirrors.

Image: Composite image of the Bullet cluster. The white galaxies were imaged by the Hubble and Magellan telescopes. The x-rays emitted from the gas and plasma clouds were imaged by the Chandra X-ray observatory and appear in pink. Regions of detected dark matter obtained from gravitational lensing appear in blue. Image courtesy of X-ray: NASA/CXC/CfA/M.Markevitch et al.; Optical: NASA/STScI; Magellan/U.Arizona/D.Clowe et al.; Lensing Map: NASA/STScI; ESO WFI; Magellan/U.Arizona/D.Clowe et al.

Operation: Furious Speed Bump

Systematic Impediments to Laboratory Integration
Contributed editorial appearing in
Scientific Computing 24:3, February 2007, pg. 14.

Twenty years ago this month, I was a junior undergrad taking a course in laboratory instrumentation. As I recall, our project involved the development of a visible absorption spectrometer using a 12-volt auto lamp, some theater gels and a photodiode. We hooked it up to an analog-to-digital converter circuit that was ported into a 1-MHz Commodore64 personal computer running its own version of BASIC. I was pretty excited at the time to develop a program that asked the user to place the blank and sample cuvettes into our homemade wooden box, collect the intensity readings, calculate percent transmittance and absorption, display the results to the screen and print a report to the thermal printer. If this could be accomplished by some kid in college, I imagined professionals in industry would seize the computer revolution and have laboratories completely automated in short order. However, some folks in the 1950s envisioned flying cars by the turn of the century as well.

Fast forward the picture to today. I’m on the other end of the chalk, so to speak, teaching a course in laboratory instrumentation now called Laboratory Informatics. My juniors are working behind individual 2.8-GHz Pentium-4 computers with 1 GB of RAM running LabVIEW 8 under Windows XP; enough horsepower to operate an aircraft carrier. Having studied transducers and signal processing in a previous course, they have been set to the task of gathering laboratory measurements and stuffing them somewhere. We spend time learning relational database design, MS Access, statistics and confidence tests along with basic programming elements of state machines, multithreading and user interface design. These concepts are integrated into a working laboratory monitoring application to reinforce their importance and the roles they play in the supply chain of laboratory information.

So that we may defer a detailed discussion of instrument interfacing, we use a remote Windows Server 2003 machine to generate synthetic laboratory values and publish them to our student-side Ethernet network using National Instruments’ DataSocket protocol. This data services layer permits us to grab the values from anywhere on the network without requiring a local copy of instrumentation drivers. Our application reads the data, calculates an average with confidence values, and displays them on the screen and in a scrolling virtual strip chart. If the average value does not fall within the given tolerance level, an error is generated that sounds an audible alarm, sends an e-mail to a specified account, dumps the previous values leading up to the error to a file and documents the error, the time and the action taken as an entry into an MS Access database. It’s a nifty solution, but I’ve found the students much prefer working with “real” data being produced by actual processes in the laboratory. DataSocket allows us to write the analysis software and substitute real measurements for the virtual ones as they become available.

During one of last year’s senior projects, a student was using a CO₂ incubator for an experiment in our tissue culture laboratory. The student had spent several weeks perfecting his aseptic technique, eventually cultivating the needed cells. Near the conclusion of the project, the temperature controller on our 12-year-old incubator malfunctioned over the weekend and the student returned to find his project had mutated into a puddle of caramelized goo. Although too late to rescue that project, we purchased and installed a new incubator outfitted with digital control modules this past summer. Each module of the dual-chamber incubator has an RJ-45 jack that communicates using the RS-485 protocol. After all of the informatics system development we did during the semester, I thought it would be a simple matter to purchase an RS-485/Ethernet converter and plug the incubator into our monitoring system. But, like the doomed tissue culture project before it, I was in for a surprise. The installation and operation manual shipped with the incubator made no mention of the RS-485 port. There was no e-mail listed, but it did contain a phone number for technical support. After dialing the number, I was informed by the phone tree that the company was recently sold and my call was being rerouted. I sat in the queue for 10 minutes and had to hang up before my next class. A few days later this project recycled to the top of my to-do list, and I reached a technician. After describing what I was looking for, he said he had the documentation I needed and would fax it to me in short order. I stopped camping out at the fax machine three days later. Our laboratory coordinator gave me the cell phone number for our salesman and suggested that may be a quicker route to take. I dialed the number and, no kidding; the phone was in a bad cell and I was unable to leave a message. When I did get a hold of him a few days later, he said he would check into it. About a week later, I got a return message informing me the RS-485 communication software was an $800 option that should have shipped with the unit. After a fruitless search, our salesman needed our original PO number to request a replacement copy. By this time, the semester had ended and our IT department was simultaneously replacing the power supplies to our servers and upgrading our campus-wide purchasing software. They would be able to retrieve the information in a few days.

It is now eight weeks since we wanted to bring the incubators online. Granted, there is a lot of dead time between each of the impediments; but, in many multitasking research environments, that can be common. Our salesman has the needed purchase information and the order should be processed in a couple weeks. I guess my Academic Ivory Tower naiveté is to blame for assuming I would be able to download the communication commands from the manufacturer’s Web site. We turned the ordeal into an object lesson on the importance of information availability, project management and people skills. Perhaps my graduates will apply this experience in their own careers, and I will soon be able to retrieve instrument network commands through the wireless connection of my flying car.

Laboratory 3.0

Upgrading Laboratory Processes in a Flat World
Contributed editorial appearing in
Scientific Computing 24:2, January 2007, pg. 12.

I recently had the great fortune of viewing an online MIT OpenCourseWare presentation of a talk given by Pulitzer Prize-winning author, Thomas L. Friedman. He was addressing students on the topic of his book, The World is Flat: A Brief History of the Twenty-first Century, now available in an updated and expanded edition published in April 2006. Friedman is the foreign-affairs columnist for the New York Times and holds a Masters degree in Modern Middle East Studies from Oxford University. Before I get deluged by a wave of “so what?” from the technical audience of SC, it is difficult to overemphasize how well Friedman understands the evolution of modern computing, its effects on the global landscape and the imperative of information technology infusion into our research and development processes.

He asserts the era spanning 1492 until 1800, Globalization 1.0, was a time when Countries possessed the technology necessary to branch out, explore the world, collaborate and trade with other nations. After 1800, Globalization 2.0, Companies possessed the requisite technology and resources to interact internationally, forge partnerships with foreign suppliers and operate international offices. Beginning around 2000, the current era of Globalization 3.0, empowered by what the author refers to as “flattening forces”, enabled Citizens to have a global presence through which to leverage their individual talents and competitive advantages for the benefit of projects anywhere on the earth. It is Globalization 3.0 that made possible my purchase of the book while my family and I were on a day trip to Ocean City, New Jersey.

Strolling along the boardwalk and browsing through the many shops makes for a splendid autumn day. We came across a small bookshop and I knew I would most likely find a copy of The World is Flat within, even though we were standing among various pizza and ice cream shops on a tiny barrier island. Sure enough, I found several copies on the best sellers rack and carried one up to the counter. The proprietor used a handheld laser scanner attached to the computerized cash register to read the barcode on the dust jacket, recall the price, calculate the tax and announce the total cost. I presented my debit card, verified my identity by entering my Personal Identification Number on a touch screen and instantaneously transferred funds from my account in Pennsylvania to the bookstore. I left the store with my new purchase in hand along with a printed record of the entire transaction, including the item, a breakdown of the cost, the date and time, address of the store, and even the name of the register operator. This scenario is so commonplace that it often shrinks into the background and goes unnoticed. The flat world referred to by Friedman is one having little or no activation barriers to the availability of goods, services, labor and information.

The technology required to make my purchase effortless is the product of a large number of highly talented individuals that may include you, the reader of this column. So I am asking for your help. Upon returning to work here at the university after the day trip, I was dismayed to witness our laboratory processes from a flat-world perspective. I needed ingredients for a reaction, so I marched to the chemical storage room and looked up their shelf locations in an alphabetized binder. I gathered what I needed on a tray and noticed I took the last bottle of one chemical, vowing to order more. As I left the stock room, my hands were full so I failed to stop and fill out the faded form hanging from the clipboard on the back of the stockroom door. I noted I was the last person to make an entry describing what I took out of the room, the quantity and its current location… back in 2003. Arriving at the chemical hood, I gathered my laboratory notebook, opened a new entry with the date and time and proceeded to write down the name, manufacturer, CAS-number, lot number and purity of the ingredients. With that completed, I needed to decide which faculty member I would bother to witness my completed entry and verify my identity…ugh, I’m getting a headache simply writing about it. After recording all of this information manually, I had little time to perform the experiment before my next class.

Where are the barcodes and RFID tags on the chemical bottles? Are we the last small liberal arts university without a chemical inventory management and control system? Why can’t I just walk through a scanner in the stock room and have the computer log the chemicals into a database and have their names waiting for me in a menu ring displayed in my electronic laboratory notebook that knows I am me from the biometric scanner embedded in the stylus? I have been writing about some amazing data acquisition and analysis technologies in this column since 2000, the start of Friedman’s era of Globalization 3.0. My question to you, the reader, is this. Where can I find Laboratory 3.0? Is it in all of the major research universities? Does it exist widespread throughout industry? Are we almost there, or have we just started? As a foreign affairs columnist, Thomas Friedman is concerned American workers and students are unaware of Globalization 3.0. As a science and technology educator, my concern is that I’m unaware of Laboratory 3.0. Please pass along your experiences with information technology infusion into research and development laboratories at my email address below. And for those few folks who share my bewilderment, may I suggest that you use one of the electronic gift cards you received over the holidays to order the book online and have it delivered to your doorstep. Amazing.

Fact or Fiction?

Vetting Text with Natural Language Processing Algorithms
Contributed editorial appearing in
Scientific Computing 23:12, November 2006, pg. 10.

While teaching my instrumentation and measurement courses, I am often met with surprise at the statement, “Scientists are the members of our civilization charged with discovering, documenting and disseminating truth.” I then spend more than a few minutes justifying this assertion. To support this statement, think about the consequences facing other professions that don’t deal in facts. Exaggerations in advertising result in increased sales, poetic license is used to embellish solid stories into dramatic screenplays, and political mud-slinging is viewed simply as playing hardball. In contrast, the fabrication of scientific research findings has eradicated individual careers and entire organizations. If you need to know the truth of the matter, ask a scientist.

The U.S. Department of Homeland Security (DHS) is aware of this fact and has awarded a multimillion-dollar research grant to a cadre of computer science departments from the University of Pittsburgh, Cornell University and the University of Utah to develop automated algorithms capable of discerning fact from opinion in written text. The group is led by Professor Janyce Wiebe, Director of Pitt’s Intelligent Systems Program and benefits from the talents of Professor Claire Cardie of Cornell and Professor Ellen Riloff of Utah, all of whom are experts in Natural Language Processing (NLP). Even though reading and writing represent two-thirds of the foundational topics of education, computers currently only display aptitude for the final third of arithmetic. NLP is a field of algorithmic intelligence that strives to imbue computer systems with a conversational interface. While data mining is successful at finding relationships between price and sales figures, the data must be gathered and properly formatted by human operators capable of gleaning information from written reports. An NLP interface would increase the speed of this process and enable analysis of data appearing in the worldwide social database containing news reports, blogs and discussion groups we call the Internet.

A simple translation of text into digital representation of meaning is itself a difficult process; however, for data mining results to have any integrity, the original data must be true in the first place. Suffering from the axiom of “garbage in = garbage out,” NLP systems must possess the ability to discern between fact and opinion. If one were to read early sixteenth century texts concerning the nature of our universe, a simple poll would reveal a majority opinion that the earth is at the center. Heliocentric theories would be regarded as fringe heresy, even if they were accompanied by supporting facts. The modern concern of DHS is one of committing resources to an imagined threat or dismissing a real threat as being false.

The DHS grant calls for the development of accurate (read truthful) and robust techniques for extracting, summarizing, and tracking information about global events and beliefs from free text. As with all scientific instruments, this process is enabled by proper calibration. Before an analytical balance is used to make measurements, it must be shown the accepted standard concept of one gram. NLP instruments must be trained to recognize domain-specific patterns and relationships that identify the difference between asserted facts and subjective beliefs. This involves the use of traditional classification methods that have been trained to recognize statements as assertions when accompanied by words like “said” and “according to” and as subjective opinions when modified by transitive verbs such as “fears,” “suspects” or “suggests.” Subjective expressions are then further classified by their source so that they may be evaluated for their level of expert reliability. “…suggests it will snow tomorrow” is more reliable when appearing on a dispatch from the National Weather Service than on a Daily Horoscope page. The development of new scientific instruments is often accompanied by a refinement in our understanding of the universe. As a scientist, I am anxious to witness this tool’s ability to gather truth from such a vast source of information. As an editorialist, I am also anxious to see the results when it is applied to the front pages of our nation’s major newspapers.

Between the Lines

The Evolution of Revolutionary Broadcast Technology
Contributed editorial appearing in
Scientific Computing 23:11, October 2006, pg. 12.

Introductions to the topic of data acquisition eventually include a discussion of sampling. This is not due to outmoded homage to tradition, but simply because we do not have the technology to record an infinite number of items. Our present paradigm of the universe holds that time is a continuous variable and, as such, can be divided into ever smaller increments until the width of each interval approaches the limit of zero. With some experimental effort, spectroscopists can measure time slices on the order of femtoseconds (10^-15 seconds) and, as far as we can tell, not much happens in the physical world between subsequent measurements at this time scale. If we desired to record events occurring at this resolution, we would need at least one million gigabits of storage space every second — a pretty good practical definition of infinity at our current level of technology.

Acquiring and storing discrete data values over time often requires us to sample events at a much lower rate than they actually occur; a process known as frequency down-conversion. The well-known Nyquist-Shannon sampling theorem states that the sampling rate must be greater than twice the frequency of the event we wish to capture and analyze. Processes occurring faster than half the sampling rate produce nonsensical artifacts in the collected data as illustrated by the "wagon-wheel effect." When the spokes of a rotating wagon wheel are photographed by a motion picture camera operating at 24 frames-per-second (fps), the rotation appears as it should until the wheel achieves 12 rotations-per-second, after which the wheel looks as if it is rapidly spinning in the opposite direction. This optical illusion continues until the wheel appears to slow down and stop when it reaches the same rotation rate as the frame rate of the camera.

Even though 35-mm motion picture and still photography film are manufactured in continuous rolls, each camera advances the film while the shutter is closed and then holds the film stationary during a finite exposure. During playback, a cinematic projector also closes its shutter as the film is advanced, and then displays the next image for one twenty-fourth of a second. While this technology is still in wide use today, the Human Vision System (HVS) is capable of perceiving processes occurring faster than 24 fps. When displayed at its native acquisition rate, the HVS notices the shutter frequency as image "flicker." To minimize flicker, movie projectors operate their shutters at double (48 fps) or even triple (72 fps) the frame rate of the film facilitating frequency "up-conversion" to a rate faster than the HVS can perceive.

The electronic broadcast of black and white motion pictures in the U.S. was standardized by the National Television System Committee (NTSC) in 1941. Television cameras captured images as a stack of intensity scan lines at the same frequency as their 60-Hz AC power supplies. Limited by the vacuum tube technology of the day, the electronics could only record or display a little over 240 scan lines in one sixtieth of a second. The resulting images appeared flicker-free, but very small. The NTSC doubled the image height to 484 scan lines and mandated the use of image scan line "interlacing" developed by RCA in the 1920s. Given a vertical stack of scan lines, this technique alternately captures the 242 odd and 242 even lines of an image at 60 Hz, ultimately producing an "interlaced" image having an effective frame rate of 30 fps and a flicker-free projection rate of 60 Hz.

In addition to being fodder for a documentary on the history of film and broadcast television, frame rates and interlacing are currently hot topics in the area of digital signal processing (DSP). The cathode ray tube (CRT) monitors of the 1980s followed closely by modern liquid crystal displays (LCDs), plasma display panels (PDPs) and digital light processing (DLP) systems employ technology capable of updating the entire image in a single pass, known as progressive scanning. Current standards for high-definition television (HDTV) include the transmission of progressively-scanned 1280 (w) x 720 (h) images; a format known as 720p, and 1920 (w) x 1080 (h) images; known as 1080p.

The technical challenge for display manufacturers is how to convert the 440 (w) x 484 (h) interlaced images of legacy NTSC (480i) into high-quality content for display on expensive, flat-panel HDTVs. There are several popular methods used to accomplish this task, including ones for "deinterlacing" the images into progressive scan, interpolating image pixels to increase the spatial resolution, and predictive motion algorithms for frequency up-converting the 30 fps to as high as 60 fps. Each of these processes is afflicted by new "wagon-wheel" effects of their own, and DSP developers are racing to provide solutions in a bid to become part of a new standard.

Will Take a Little Time

New Chip-Scale Atomic Clocks
Contributed editorial appearing in
Scientific Computing 23:10, September 2006, pg. 16.

Historians continue to wrestle over an exact date for the start of the Industrial Revolution. Generally placed in the late 1700s, it is often correlated with the widespread use of the steam engine and coal as an energy source. As a technophile, I like to attribute it to the invention of tools having an intrinsic awareness of time. Hammers, chisels and saws have been in use for centuries, and complicated musical instruments arrived with the Renaissance, but each requires a human being for their operation. The intricate gears, valves and cogs of a steam engine and the pulleys and cables of a weaving machine grant them autonomy and, more importantly, the ability to work faster than their human counterparts. An increase in speed requires a precise knowledge of the system time so that all components work in concert. A faulty timing belt in a mechanical system often leads to catastrophic failure, as parts are in the wrong position at the wrong time. Modern electronics commonly rely on the vibration of a quartz crystal in an oscillation circuit as a “timing belt.” While not resulting in physical damage, an inaccurate or defective quartz oscillator can cause malfunction of the electronic device and the systems it controls.

The most precise oscillation circuit is based on the frequency of microwave radiation absorbed by a specific ground state of cesium (Cs). Since 1967, the International System of Units (SI) has based the value of one second on the Cs absorption of 9,192,631,770 Hz (ca. 9.2 GHz) radiation. The U.S. National Institute of Standards and Technology (NIST) in Boulder, CO, operates the NIST-F1 Cs atomic clock that serves as the primary time and frequency standard for the U.S., and contributes to the international group of atomic clocks that define coordinated universal time (UTC), the timing belt for the planet earth. While most electronic devices hum along happily using their quartz oscillators, the increasing globalization of network communication and global positioning systems (GPS) for military and civilian applications can benefit from the improved precision of an atomic clock, which is over five million times more precise than quartz.

The increased precision comes at the expense of size. NIST-F1 occupies 3.7 m³ and consumes 500 W of power. Smaller, less accurate atomic clocks are used by satellites, network providers and broadcasters, but the U.S. Defense Advanced Research Projects Agency (DARPA) would like to incorporate chip-scale atomic clocks (CSACs) into portable spread-spectrum communication, radar and GPS devices. Over the past few years, DARPA, NIST, universities and companies including Honeywell and Rockwell Collins, have been working on the CSAC project to develop a 1-cm³ atomic clock that consumes less than 100 mW.A major technical obstacle is the physics of 9.2 GHz radiation. It has a wavelength of 3.2 cm — much too large to fit into a 1-cm³ box filled with Cs gas. NIST scientists lead by Dr. John Kitching have solved this problem by using a recently discovered spectroscopic technique known as coherent population trapping (CPT). While Cs atoms can hop between low-energy levels in their ground state using microwave energy, they require near-infrared energy at 852 nm to reach an excited state — a wavelength small enough to fit inside a cubic-centimeter cavity. Although the mathematics are a bit involved, when 852-nm light is modulated at 4.6 GHz (exactly one-half of 9.2 GHz), upper and lower sidebands spaced at the correct 9.2 GHz are placed on the 852-nm carrier. When the modulation of the CPT circuit matches the exact spacing of the Cs ground state, the sidebands are absorbed and a greater amount of the 852-nm carrier intensity reaches the photodiode detector. The CPT technique achieves the same microwave absorption used by NIST-F1 using much shorter near-infrared light from a diode laser. Recent versions of the CSAC utilize more stable rubidium (Rb) atoms and have demonstrated a 1000x increase in precision over the use of quartz in a total package of 9.5 mm3 requiring 75 mW.

CSACs also may help to reduce the bandwidth congestion of consumer cell phones and WiFi devices. The frequencies allocated for communication using a specific technology are divided into separate channels that can be used simultaneously. The frequency width and number of channels is limited by the ability of each device to accurately transmit and receive within its assigned bandwidth. As timing precision increases, channels can be narrowed to make room for additional ones. When the DARPA CSAC program concludes this fall, the race will be on to commercialize the technology. I hope I’m not too late to pre-order my wristwatch version.

Image: Stylized Chip-Scale atomic clock. Frequency-modulated laser sends near-infrared light through the atomic gas cell and onto the photodiode detector. All components are created using micro-electro-mechanical systems (MEMS) fabrication technology. Image courtesy of DARPA CSAC Program.

There and Back Again

The Many Flavors of Image Stabilization
Contributed editorial appearing in
Scientific Computing 23:9, August 2006, pg. 14.

The first permanent photograph was created around 1826 by French scientist Joseph Nicéphore Niépce, and is currently housed in the Harry Ransom Center at the University of Texas at Austin. Coined heliography (Greek for "sun writing"), Niépce made special notice of the bleaching and stiffening effects sunlight has on organic polymers — much like the discoloration and cracking experienced by plush toys and plastic figurines placed on the rear deck of an automobile. He ultimately used a thin film of petroleum tar coated on a pewter plate as a recording medium. After exposing the film to several hours of sunlight shining through a pinhole camera, Niépce developed the film by washing away the unhardened tar with paint thinner to reveal a recorded image of the view from his home near Chalon-sur-Saône in Burgundy. Although the basic process is very similar to that of modern photography, the long exposure times required by heliography made it impractical for subjects other than still life.

Over the succeeding decades, exposure times have decreased dramatically with advances in film technology — even today's disposable flash cameras have an average exposure time of one one-hundredth of a second. However, the optimal length of exposure is determined by a combination of film sensitivity, camera design, lens configuration, lighting and subject distance. In a portrait studio, the camera is often mounted on a sturdy tripod and the subjects are asked to hold their pose in order to minimize the amount of motion blur introduced into the photograph. Digital photography replaces the photochemical film with a matrix of photosensitive electronic circuits known as a charge coupled device (CCD). After exposure, the digital image is recorded by reading the voltage values of each CCD picture element (pixel) and then the solid-state CCD is reset for the next image. In addition to obviating the film development step, digital camera design permits the elimination of the camera tripod support system as well. Digital camera manufacturers are offering image stabilization systems that fix the spatial position of the CCD even though the handheld camera body may be moving or shaking. For example, Konica Minolta incorporates an angle speed sensor into cameras equipped with its anti-shake (AS) system. While the shutter is opened, the sensor measures changes in camera position and moves the CCD relative to the camera by actuating the internal smooth impact drive mechanism (SIDM).

Video (motion) cameras can benefit from image stabilization technology as well. Although each frame of a video sequence may not suffer from motion blur, stationary observers (such as those seated in a movie theater) may become disoriented if the recording camera is shaking or unsteady. Developed by Garret Brown in 1973, the Steadycam system utilizes a collection of gimbals and springs to isolate the inertial frame of the camera from that of the camera support. The system works equally well with film and digital video cameras, but is impractical for use with consumer video products, including digital video cameras, video phones and PDAs. The miniaturization of personal electronic devices also makes Konica Minolta AS technology cumbersome as it masks the solid-state nature of the CCD.

Enabled by emerging digital signal processors (DSPs), digital image stabilization (DIS) provides motion compensation without physical movement of any camera component. The basic DIS process involves capturing an image using the camera's large CCD and only recording a subsection of the digital image. For example, a 3000 x 2000 pixel image is recorded from a 3016 x 2016 pixel CCD, resulting in a frame of eight unused pixels around the image. If the video camera is moved slightly before the next image is acquired (due to shaking), the original 3000 x 2000 scene is still captured, but is now located at a different position within the large CCD. The DSP compares the locations of objects in the new image with their locations in the previous one and determines if changes are the result of camera shake using its internal algorithms. The new image is stabilized by selecting a different 3000 x 2000 CCD pixel region centered on the position of the previous subsection. In this manner, the virtual 3000 x 2000 CCD is moved electronically.

Recently, Daniel A. Tazartes, Director of the Navigation and Applied Sensors Technology Center (NASTC) at Northrop Grumman, and his team, have taken image stabilization technology to the next level of complexity. While seated theater goers are stationary, helicopter pilots are decidedly not. A stabilized video camera on the ground or mounted to the helicopter can be very disorienting when displayed on a vibrating monitor and viewed by a vibrating pilot. NASTC has placed accelerometers on the camera, the display, and the pilot's helmet and these measurements and the video stream are fed into a processing unit. The processed video image is translated along the horizontal and vertical directions of the monitor to induce vibration and jitter. Viewed by a stationary observer, the display looks erratic, but appears stabilized to the pilot as the movement is synchronized to the relative motion of the vibrating display and the helmet-mounted accelerometer. New applications are arriving at a fast pace, so it may take some time for this technology to stabilize.

Image: The first successful permanent photograph created by Joseph Nicéphore Niépce ca. 1826. Currently on display in the Harry Ransom Center at The University of Texas at Austin.

New SWIS Army

Developing Collective Intelligence
Contributed editorial appearing in
Scientific Computing 23:8, July 2006, pg. 14.

This past spring, my family and I spent the long Memorial Day weekend at a campground in rural Lancaster County, PA. As the unofficial start of summer, many families head “down the shore” for some sun, sand and beach fries, while we prefer the less hectic pace of horse-drawn buggies, swimming and Hershey’s s’mores. While my son and I were rigging our lines for fishing in a small pond, he noticed a large number of tadpoles teaming near the water’s edge. As we approached them, they spotted our shadows and swam into deeper water, only to be terrorized by roaming sunfish that quickly shepherded them to shore. Not wanting to decimate the school of adolescent amphibians, we backed away and discovered an even larger pod of pollywogs in a nearby water-filled ditch. Localized in their relatively safe environment, they were easy to observe and study as they swarmed around in the puddle. The recent rain ensured their nursery had plenty of water, but I hoped the metamorphosis into air breathers initiated well before the arrival of the intense summer sun.

The schooling and swarming of insects and vertebrates has become an area of interesting research known as swarm intelligence (SI) and this column has looked at developments in genetic algorithms (GA), particle swarm optimization (PSO) and other biomimetic technologies of SI. Application of GA and PSO has primarily benefited simulations and control algorithms for the solutions of specific computationally hard problems, such as asset allocation and curve fitting. Recently, Professor Alcherio Martinoli and his Swarm-Intelligent Systems (SWIS) group at École Polytechnique Fédérale de Lausanne (EPFL) in Lausanne, Switzerland have applied SI to an actual swarm of miniature robots (bots). One of the SWIS projects is concerned with the development of a swarm-intelligent inspection system based on a collection of autonomous bots used for the examination of turbine blades within a jet engine. Usually requiring costly disassembly or the use of time-intensive borescopes, a swarm of mobile bots equipped with local sensors conceivably could enter the intact turbine engine and craw along the surfaces of the shaft and blades while collecting sensor readings.

The SWIS team unrolled the axis-symmetric geometry of the turbine into a flat “playing field” having vertical blade-shaped extrusions, in order to remove the technological obstacle of inverted locomotion. Instead of using a system-wide controller that plans the inspection routes and directs the deployment, each bot contains only local sensors for interaction with its environment, and communication with nearby members of the swarm. An onboard controller cycles through an internal finite state machine (FSM) having a default state causing each bot to roam and search for an obstacle. If the obstacle is determined to be a wall or another bot, the state is switched to one that avoids collision. When a blade is found, the FMS switches to circumnavigate the blade. After this task is completed, the bot pauses to transmit a beacon to nearby bots signaling the blade has been inspected, and then switches back to the search state.

More recently, Professor Martinoli and graduate students James Pugh and Nickolaus Correll have studied the efficiency of replacing the FMS with a PSO algorithm. Rather than using the beacon to inform local bots where not to search, each bot can serve as the physical embodiment of a virtual PSO particle using local neighbor optimization. When a blade is found (an optimal location for inspection), the bot communicates the location to its neighbors and the information is used to predict the search direction of additional blades. As bots meet different neighbors, they can communicate and update each other’s knowledge of the entire environment in a fashion similar to human search teams that report their progress to each other. In this way, the knowledge evolves inside the members of the swarm and the bots do not need to expend their limited energy in communication with a distant system controller.

Upon closer inspection, my son and I discovered the ditch was actually a spring head that flowed into the pond, and not stagnant water from a recent storm. In addition to spawning thousands of eggs in the hopes the supply will outlast the appetite of predators, it is evident the frog population benefits from dispersing their eggs throughout their tiny ecosystem in search of more optimal locations. I like to think my family displays similar collective intelligence by avoiding the bumper-to-bumper traffic on the way to the shore.

Image: A group of Alice robots (2 cm in length, originally designed at EPFL by Gilles Caprari) exploring a 2-D regular structure faithfully reproducing blade dimensions and inter-blade distances of an unfolded jet turbine. Picture by Alain Herzog.

Inductive Interrogation

Magnetic Field Response Measurement Sensors
Contributed editorial appearing in
Scientific Computing 23:7, June 2006, pg. 10.

One of the many great things about working behind the fence at Wright-Patterson Air Force Base is the diversity of truly amazing technology under development. Shuttling among the Air Force Research Laboratory facilities as an on-site research contractor is an experience that will not soon wear off. A palpable gee-whiz ether permeates the storied collection of hangers and test cells that house some of the most sophisticated state-of-the-art instrumentation in existence — the hybridization of academic smarts and military budgets.

My employer was charged with the development of optical diagnostics for the measurement of temperature, pressure and flow fields above, on and within aerospace test articles. Probing these values with light allowed us to gather our data without the cumbersome configuration of lead wires and tubing required by traditional electronic sensors that often modify the aerodynamics of the system under study. Owing to the human sensory apparatus, over time the models of sleek, futuristic flying machines melt into the background of perception causing the mundane to appear out of place.

Such was the case when a large-scale Boeing 747 center fuselage model was wheeled in by the U.S. National Transportation Safety Board (NTSB). An engineering marvel in its own right, the white, bulbous passenger plane was overtly obvious among the collection of sinister stealth. Upon closer inspection, the model contained a burst panel to study the debris trajectory from an explosion in the center wing tank, as proposed to have precipitated the 1996 TWA Flight 800 tragedy off of Long Island, NY, resulting in the loss of all 230 persons aboard. There are many conspiracy theories as to the actual cause of the calamity, most likely due to the layman's disbelief in the ability of a short-circuit in a simple fuel gauge sensor to cause such loss of life. A research group lead by Stanley E. Woodard, Ph.D., at the NASA Langley Research Center also reports the discovery of wire damage grounding the entire space shuttle fleet, and being the probable cause of the onboard fire resulting in the 1998 loss of the MD-11 Swissair Flight 111 out of New York and all 229 persons aboard.

In addition to improving the safety of wiring materials, a research team lead by Dr. Woodard is investigating the use of wireless sensors aboard aerospace vehicles to eliminate the danger due to arcing, fraying and chemical degradation of the wires providing power and data links to sensors. Traditional wireless laboratory measurement systems incorporate processors and transmitters that often need their own wired power supplies. Radio frequency transceiver (RFID) tags can transmit their data after receiving power from a radio wave, but the required silicon processors are often incompatible with their harsh measurement environment, and the radio frequency range available for aviation use limits the separation between the antenna and RFID tag to around 15 centimeters. In our optical diagnostics laboratory, we used infrared and visible electromagnetic radiation to interrogate our sensors remotely; however, we also had the luxury of mounting high-power lasers or banks of light emitting diode (LED) lamps external to our test articles.

To avoid the shortcomings of radio and optical electromagnetic radiation, NASA Langley has developed a wireless sensor system based on oscillating magnetic fields. The sensors are comprised by a spiral induction (L) antenna coil and a plate or interdigitated capacitor (C) that is in contact with the quantity to be measured. When the L-C sensor is stimulated by the magnetic field, a resonant current is established having an oscillation frequency proportional to the dielectric environment of the sensor. After the driving magnetic field is removed, the antenna coil of the L-C sensor emits a field at its characteristic frequency that is received by an interrogation antenna. The research team reports separation distances of up to 3.3 meters when powering the sensors at 1.5 watts.

Because different L-C sensors have individual resonant frequencies, several sensors can be used simultaneously in conjunction with a single control unit by sweeping the frequency over the range utilized by the sensors, typically 1 to 10 megahertz. Various sensors have been developed for the measurement of position, level, load, angular orientation, material phase transition, moisture, specific chemicals, rotation/displacement, bond separation, proximity, contact, pressure, strain and crack detection. While the patent application is still under review, NASA Langley is actively seeking partnerships and collaborations to commercialize the technology. Let's hope the induction time of its adoption is low so that the frequency of arcing events can approach zero.

Acquiring Minds

Distance Learning and Collaboration
Contributed editorial appearing in
Scientific Computing 23:6, May 2006, pg. 10.

I reached a milestone at the conclusion of this semester. I have now spent more time as a teacher in education than as a researcher in industry. The event was commemorated with little more than a “hey, I’m getting old” – about as much significance as it deserves. Upon reflection, I realize I have changed my target audience. Whereas I used to be primarily concerned with communicating with instruments, I’m now concentrating on collaboration with twenty-year-olds. Analogous to Management Information Systems (MIS) for business managers, our program offers a concentration in Information and Knowledge Management (IKM) for future laboratory and production managers. IKM is comprised of courses on the broad topic of laboratory informatics, including instrument and database interfacing, intelligent automation algorithms and data mining. Rather than adopt a course text and have me summarize my notes in chalk scribbles while the students sit in stunned rapture of my presentation abilities, the courses are run as if we were a team of professionals tasked with automating a laboratory in industry. In fact, we do have a research laboratory to automate – our research labs here at the school.

Much like the specialized research laboratories used by biology, chemistry and physics in our science building, our IKM concentration has created a dedicated laboratory for the study of laboratory informatics called the Virtual Control Room (VCR). The VCR looks like a typical control room with clusters of networked computers and a digital projection system serving as a room-wide heads up display (HUD). The instructor station is connected to the HUD and outfitted with a touch screen allowing chalk and dry-erase markers to be replaced by a mouse for clicking through the environment and a stylus for drawing play-by-play diagrams. Even though we have used technology to modernize the chalkboard, passively watching me click through PowerPoint slides is as disastrous as listening to me read my lecture notes. Coaching the students while they actively interface a remote instrument or develop an analysis algorithm is infinitely more desirable by all parties involved.

The days of “do your own work” and “don’t look at your neighbor’s paper” are slowly fading along with the industrial revolution. Our information economy requires effective communication and collaboration among team members. While “Please turn off cell phone” placards are sprouting up around campus classrooms, I encourage my students to leave their cell phones and PDAs activated while sitting at the VCR computers with email and instant messaging apps minimized. Effective management of personal information streams should be taught in school before it is acquired by experience on the job. Instead of worrying that students will pass answers to multiple-choice exams through phones, PDAs or calculators, our courses utilize “multiple-component” exams, requiring groups of students to collaborate on the solution to more sophisticated problems.

Audio and video beamed by telepresence applications extend the range of team communication effectively, but we need human communication protocols optimized for problem solving, similar to the hand signals used in baseball. The VCR is currently using the Wiki protocol developed by the Mediawiki Foundation. As the foundational technology of Wikipedia.org, our students are already very familiar with its operation and it has supplanted Google in our classrooms. Wikipedia is not a better search engine than Google; it is a better vehicle for information. It returns hyperlinked articles written in a common protocol rather than collections of links to pages only related by a few search terms. We have installed the open source Mediawiki on our VCR server and it has quickly become the collaborative notebook for our courses.

Until recently, assignments were handed out in hardcopy and individual responses would appear in the inbox of my email. Now a project is created as an article on the Wiki, subtasks are defined and listed as links to empty new pages, and student names are attached to the subtasks. Students populate the pages with their literature searches, calculations, experimental findings, procedures, code and additional questions. Every member of the team can view the status of any task and contribute if it travels into the realm of their expertise. The Wiki maintains a record of all modifications, including the content of the change, time, and author for auditing and tracking purposes. It also maintains a copy of all previous versions of each page so knowledge is not lost and the motivation behind changes can be reviewed. While the students are logging process values into a database to evaluate the performance of an experiment, I can review the Wiki logs to evaluate the performance of the team. I just have to keep our servers away from my Dean’s office so they don’t use them to evaluate me.