Modified time-based servo enables increased track density on tape.

In 1995, IBM introduced a new tape drive incorporating a novel track following servo system. The 3570's time-based servo system, developed by Tom Albrecht of the IBM Almaden research center, is one of the key enabling technologies now used in all generations of the successful Ultrium line of LTO tape drives.

Track following servo systems have kept the industry on track to achieve 1 terabyte of storage on a commercial single reel tape cartridge within the next few years. Time-based servo is an elegant design, which has endured 10 years of success with little change. It appears, however, that modifications may be required to move beyond 1 TB. Longitudinal errors in placement of the servo pattern generate false position information during drive servo operation. The magnitude of these errors can become unacceptable at the high track densities necessary to support multi-terabyte data cartridges.

The LTO time-based servo pattern is used to align recording heads during read/write operations (Figure 1). As tape handling has improved and track densities have increased, a curious phenomenon has appeared. Closed loop tracking performance, normally a function of lateral tape motion and head/actuator performance, is also influenced by vertical position within the servo band.

[FIGURE 1 OMITTED]

System modeling shows why this happens. Longitudinal spacing error of the servo frame, when demodulated, generates non-uniform error with respect to vertical position within the pattern. Researchers at the Imation Oakdale laboratory have developed a modified time-based pattern, which avoids frame spacing error. The new pattern provides a velocity reference for servo decoding without relying on frame-to-frame distance. Experiments show the enhanced tracking capability of this pattern at elevated track densities.

Time-Based Servo Operation

Time-Based servo bands, which are written to tape during the manufacturing process, are used as a reference to position all data tracks for the life of the cartridge. The servo bands are comprised of a series of repeating patterns (servo frames) down the length of tape. In a simplified version of this pattern, each frame is made up of two large magnetic stripes written at an angle with respect to one another.

The servo pattern is decoded in a drive by measuring the distance (a) between frame stripes at the servo read gap position while tape is transported longitudinally across the recording head. A geometric transformation yields the vertical position y of the read gap with respect to each servo frame on tape. This distance is actually computed by measuring the time between stripe pulses at a known tape speed, hence the term "Time-Based Servo". At roughly 200 um tall with a repeating frame distance of 100 um, a single servo band can provide information necessary to position many data tracks.

In an actual drive, the recording head is mounted on a movable actuator and its position is constantly being monitored and corrected to compensate for lateral tape motion. The actuator is adjusted using a position error signal (PES), which is the difference between the measured position computed from the servo pattern and the desired position of the head. When the head is properly registered, meaning the PES is sufficiently small, data tracks are written to tape. The ability to stay "on track" is what helps move the industry to higher and higher track densities. This simple, yet effective, servo system can resolve less than a micrometer of vertical motion over the entire 200 um range.

The Servo Writing Process

The time-based servo pattern is factory written as part of the cartridge build process. The servo write head is a mirror image of a single servo frame with two non-parallel write gaps on the recording surface. The pair of stripes is written simultaneously by pulsing the write head with current while transporting the tape on a precision tape deck. By controlling tape speed and the timing of write pulses, the pattern is repeatedly "stamped" down the tape with precise control over frame-to-frame spacing. Unfortunately, some modulation in frame spacing is inevitable due to fluctuations in tape speed and tension. What are the consequences of frame spacing error?

This question is not as simple as it might seem. For a fixed amount of frame spacing error, the resulting error in computation of vertical position y is dependent on location within the pattern. For instance, near the bottom of the servo pattern, a I percent frame spacing error results in roughly a 3 um error in vertical position. At the top of the pattern, the same one percent error produces about half the vertical position error.

Frame Spacing Error

Suppose the distance between servo frames varies from frame to frame down the length of tape. If the rate of change is beyond the track following capability of the head/actuator, the result will be a residual PES. If the magnitude of this PES error is large enough, the drive will believe it is "off track" and a servo fault will occur.

During normal drive operation, it is impossible to distinguish between PES from lateral tape motion and PES from frame spacing error. From a practical standpoint, PES from frame spacing error is simply unwanted noise in the servo system, and leads to uncertainty when computing the true position of the recording head. Using a clever diagnostic mode, however, it is possible to derive the contribution of frame spacing error. Because PES from frame spacing error is sensitive to vertical pattern position, it's contribution to overall PES can be estimated by running multiple passes of the same section of tape with the servo read gap at different vertical locations on the servo pattern.

It is common to evaluate servo performance by measuring the standard deviation of PES while track following. A good a rule of thumb for a well-designed system is to limit the maximum allowable PES standard deviation to about 2 percent of the track width. A 2 terabyte LTO form factor product would require a track width of roughly 5 um (5000 track per inch). A 5 um track width would require a PES standard deviation of 100 nm or better, which is about the point at which frame spacing noise dominates the PES signal for the current time-based pattern.

Modified Servo Pattern

Efforts to reduce the influence of frame spacing error at the Imation Oakdale laboratory have led to the development of a modified version of the time-based servo pattern. The pattern, known as the "N" pattern, is comprised of three simultaneously written servo stripes per frame. Because two parallel stripes within a frame are used for reference velocity, the requirement to tightly control frame-to-frame distance vanishes.

Another trick used to quantify PES noise is to intentionally write a servo pattern with all servo stripes oriented parallel to one another. This effectively eliminates lateral tape motion from the PES computation and the resulting PES signal is a very good estimate of total measurement noise, including frame spacing induced PES error. The standard deviation of PES noise calculated for each meter of tape is consistently below 30 nm, about 1/3 the estimated PES noise caused by frame spacing error in current LTO products.

The Future of Tape

The future of tape storage is rapidly heading toward multi-terabyte tape cartridges. To achieve this, data track widths below 10 um and closed loop tracking of under 100 nm will be required. In addition to the tracking requirement, new tape systems still need to be backward compatible with current time-based servo written tapes to support a smooth migration path. The "N" pattern for time-based servo provides compatibility with today's systems along with high resolution for tomorrow's multi-terabyte systems.

Douglas W. Johnson is a servo engineer at Imation Corp., Oakdale, MN

www.imation.com