Rob Barrett I am a member of the User Systems Ergonomics Research group at the IBM Almaden Research Center. Although a physicist by training (Washington University in St. Louis and Stanford University), I have always worked with computers and enjoy designing and building systems, whether out of mechanics, electronics, or software.

Current Projects

WBI (Web Browsing Intermediaries) is an approach to personalizing the information on the web so that it is tailored to individual users' needs. More generally, WBI is a computational model for the web based on the idea of intermediaries. You can think of intermediaries as computational elements which add value to information transactions, by doing things such as:

• providing authentication,
• remembering and ordering information previously viewed,
• enabling collaboration between users that are concurrently viewing the same information
• transforming information to better suit the device I'm using to view it, and
• provide navigational enhancements (e.g. new links).

You can read more about WBI in a recent paper or you can download the WBI Personal History application to try out for yourself.

deFacto

The brainchild of Gentry Underwood, deFacto seeks to improve the "pushing" of information through a dynamic user model. In the "push" view, computers actively present information to a user (for example, through a scrolling ticker) without the user directly asking for it. Companies such as PointCast have made a business of this. But most of these systems assume that people's information needs can be defined by a series of checkboxes that say "I am interested in sports news", or things like that. In deFacto, we assume that users' interests change rapidly and that the computer can discern these interests by watching people as they work. deFacto uses a collection of sleuths to observe the user's activities, gather relevant information, and then present that information. You can read more about deFacto in a recent paper.

Previous Projects

AIM & FIRE

AIM & FIRE (aka RAT) developed new ways of searching for textual information.

With AIM, users describe information in terms of metadata, things such as what language it is written in, what genre of writing is desired, how timely the information is, what subject area it is in. These descriptors go beyond simple word matching searches.

With RAT, as users view the results of their searches, they can mark certain results as "relevant" or "trash". RAT then provides assistance by recommending words to add or remove from the original search to produce better results.

Negative Inertia for TrackPoint

Our research group, led by Ted Selker has led the development of the TrackPoint pointing device. My particular contribution is the Negative Inertia algorithm, which made its debut in TrackPoint III. The idea behind Negative Inertia is that pointing needs to be highly responsive without being uncontrollable. If the gain of the pointing device is too high, it is responsive but tends to be difficult to use. By introducing a dynamic transfer function, we were able to increase the responsiveness without losing control -- producing a more usable device.

Timing-Based Servo for Tape Storage

This was one of my favorite projects because of the many interesting aspects of the problem, because it resulted in a nice product, and because I enjoyed working with Tom Albrecht and Jim Eaton. The problem is how to keep a read/write head accurately aligned to a magnetic tape as flies by at several meters/second, stretching and vibrating all along the way. To add more complication, the tape is dirty so that signals go in and out, and the sensor being used to measure the servo signal on the tape can also be unstable.

We developed an approach based on a very narrow servo read head (which limits the error due to servo read head instabilities) which reads timing marks on the tape. These timing marks continuously vary across the tape, providing a smooth signal which indicates the position of the head. This signal is then fed back to a head positioning system to maintain accurate alignment. This was the easy part. The tough part was developing a way to write continuously varying timing marks on the tape. Tom and I designed a dual gap write head which produces the necessary pattern.

This servo technique has been incorporated into the LTO (Linear Tape-Open) standard.

More technical information is available in the proceedings of INTERMAG'96 or from our patent.

Charge Storage

My thesis work covered a number of applications of scanned probe microscopes, such such as the Scanning Tunneling Microscope, the Atomic Force Microscope, and the Scanning Capacitance Microscope. A couple of the resulting technologies have been productized by Park Scientific Instruments.

One of the most interesting applications of these scanned probes is to both manipulate and examine surfaces at (or near) the atomic scale. I developed the Charge Storage approach with my thesis advisor Cal Quate. This technique uses the scanning capacitance microscope to store information by trapping charge in the insulating films of a silicon nitride - silicon dioxide - silicon structure. The trapped charge can then be detected through its depleting effect on the underlying silicon. We have demonstrated the ability to store large numbers of bits (500,000) at a high density (26 Gb/sqin). We have also demonstrated the ability to read and write this charge at data rates of 1 Mb/sec. There are lots of open issues still, but as the superparamagnetic limit quickly approaches for magnetic data storage, this may be an option for continuing the dramatic improvements in data storage that we have seen over the past decades.

For more information, see

• R.C. Barrett and C.F. Quate, "Large-Scale Charge Storage by Scanning Capacitance Microscopy," Ultramicroscopy 42-44, 262 (1992).
• B.D. Terris, R.C. Barrett, and H.J. Mamin, "Scanning Capacitance Detection and Charge Trapping in NOS," Proceedings of the SPIE, Los Angeles, CA (1993).