Storage Technology Corporation’s (STK) recording head wafer costs were too high. Wafer production facility was underutilized, leading to high per-unit costs. STK had acquired a new wafer facility in 1998, but when an internal product was canceled, it was larger than needed.

Hewlett Packard (HP) was having some difficulty with their existing head supplier and was open to alternatives. Engineering was a key player in winning the business, with an action plan including the following: 1) definition of performance requirements and specifications; 2) understanding and planning for process changes needed to meet requirements; 3) ensuring that possible process changes required did not impact internal product already in manufacturing; 4) sizing capacity of existing tools; 5) sourcing components to ensure cost target attainment; 6) building prototype heads for evaluation by the customer; and 7) onsite involvement at HP’s manufacturing site in England to validate prototype drive build performance.

This was the organization’s first opportunity to work with external customers. Established program management system to ease the process. Engineers learned best practices of another company, and some of those disciplines were incorporated into STK’s business model.

First prototype build went extremely well, and STK was awarded 50% of the business. Subsequently, customer approached STK with every new product, continuing with an initial 50% share in each. With these new products filling the line, internal STK product costs were reduced by ~30% and about $12 million in additional revenue was brought in for 2005.

Additional OEM business with Hewlett Packard (HP) was causing a workload crisis. Three to four internal products were in various stages of engineering support, along with the several OEM programs. Layoffs had also reduced the manpower available. Each program had an engineer assigned to monitor technical and schedule issues. With the increasing number of programs, engineers were unable to concentrate on their technical specialties on top of the program duties.

Developed a team of four key players to comprise Program Engineering function, to handle all schedule, negotiation, and contract issues in order to free up the engineers’ time. Team handled initial discussions with potential customers on schedule, quantity and general technical requirements; met with Engineering to decide if quantity, performance and cost objectives could be met with reasonable certainty; and developed proposals for customers. They also worked with Legal to lay out contract issues, particularly the boundaries by which intellectual property would be delineated. Team published internal and external schedules, and collaborated with offshore suppliers to define production preparation activities and make sure tool capacity was available.

As development got underway, Program Engineering team handled weekly phone calls and less frequent, face-to-face status meetings. As production began offshore, they held weekly phone meetings between Program Management, Engineering and offshore engineers to map progress and handle questions. Team also published monthly KPIs for quality control and process monitoring. First programs were quite labor intensive, but with experience, they took on standard characteristics, and it became easier to manage additional programs to a sort of template.

STK assumed OEM programs with two other companies in addition to HP, with potential incremental revenue of $8-10 million in 2006.

Computing speeds were improving, but not as fast as heavy-duty scientific users wanted. In many cases, the only place where numerically intensive computing could be done was on dedicated mainframe systems (IBM, Cray, etc.), which was prohibitively expensive.

As member of three-person physicist team, worked with IBM Research and outside software developers to demonstrate one of the first practical applications of workstation clustering. Visited the IBM Research facility in Yorktown Heights to discuss projects with the scientists there. Worked with outside, parallel computing-software developers to port their products to IBM’s AIX system. Consulted user community on their needs. Assimilated the above to demonstrate parallel computing on a test platform of four systems in the Dallas facility.

IBM made a significant inroad as a viable alternative to the standard DEC/SUN installations that had a stranglehold in most labs and universities.

Recruited from IBM to lead design and development of first all-new internal recording head for Storage Technology Company (STK). Company was experiencing development problems because of antiquated thin-film facility and lack of internal thin-film head performance knowledge. This was STK’s first-ever, all-new tape storage system, comprised of a new format and new technology. Thus, a number of decisions had to be made that did not exist when reverse-engineering IBM designs.

Identified key issues, directed actions, and validated implementation of all components: Defined data location on tape and tolerances for format. Identified materials, including magnetic, wear and thermal characteristics. Established sensor technology—several options were available, but tool set was incapable of manufacturing some. Designed actuator appropriate for resonant frequencies and mass that enabled the required head movement. Developed servo technology required to track with sub-micron tolerances. Designed high-efficiency write transducer with newer, low-power electronics. Ensured proper head/tape surface finish and wear properties crucial to performance and yield. Integrated write and read characteristics of the head to precisely match the input requirements of the electronics. Established necessary processes to surmount initial yield deficiencies in order to achieve manufacturability.

Successfully launched 9840 tape drive, which became Storage Technology’s highest-selling system ever, with revenue exceeding $900 million in first 18 months.

A jump in technology was required for next-generation tape-storage product. Peak detection recording channels were declining, and the disk industry had already progressed on to the Partial Response Maximum Likelihood (PRML) channel to move forward.

As integration engineer, led a multifunctional team to work with a chip supplier to evaluate their channel with company components. Ordered and received initial evaluation chips, cards and software from the vendor. Completed evaluations with existing, down-level internal hardware, prompting promising results. Completed four-corners/stress tests, varying the environment to intentionally degrade performance and gauge robustness. Further analyzed system, using up-level components as available, before committing to production. Successfully completed tests using new plan-of-record hardware.

Chip was adopted and used to make the most-advanced and highest-capacity storage system in the industry at the time, 200GB. It is also being used in another company product currently in development.

Machining, Assembly and Test (MAT) in facility was too expensive. Even with full facility utilization, the proposed product cost targets would be exceeded. Much of the work was manual, and labor costs for US workers are high. Management understood that offshore manufacturing companies could beat these costs by 60% to 80%, but the potential downside was schedule and quality impact.

Launched manufacturing transfer effort with a nine-month target in order to maintain program schedules. Identified personnel with appropriate skills to travel to China for process transfer and training. Wrote, refined, and had documentation translated into Chinese. Crated and shipped production and test equipment—partial shipments at first, to keep production capability in the US while it was brought up to speed in China. Built and tested first prototypes in China using STK equipment and processes, but initial three months were unsuccessful in building product to specification.

With facility not yet capable of full production capacity, stationed two to three STK engineers in China at all times. Product was hand-carried back and forth between China and the US to accelerate the sorting out of process and test issues. Held daily morning and evening engineering meetings to resolve issues one at a time. Production ramp began about two months late, but within six months, made up the difference.

Facility continues to build most of company’s recording-head products today, with excellent quality and with savings of about $7.5 million annually. Supplier has more than caught up on production now, and an additional supplier was added in 2004.

After Tucson laboratory completed development of a new recording head for the IBM MagStar storage system, production and processes were to be transferred to the San Jose facility. Tucson development lab had process equipment dedicated to tape-head processes and materials. San Jose facility was primarily a disk fabricator, and shared most equipment with the larger disk operation. Additionally, most of the disk processes were different from tape processes.

Led manufacturing transfer effort, ensuring that all processes and test capability were duplicated in San Jose as closely as possible. Processed wafers in both environments and machining in one to compare wafer characteristics. Then processed wafers from one factory in both machining facilities to find machining process differences. Rewrote manufacturing specifications to account for differences in tools. Redefined test specifications: the tape to be used in production test was new, having been developed in parallel with the head. Traveled between sites numerous times, hand carrying test samples to expedite process.

Process was successfully transferred, passing the manufacturing readiness review. At transfer, the tape-head wafers had a better track yield than disk products, at ~97%.

Recording head program needed to boost production from 50 to 400 heads per week. Yields were less than 50% at final test, and manufacturing capacity was insufficient to support the necessary volume.

Concentrated efforts on maximizing yield, rather than attempting a brute-force approach. Led Engineering group to identify that a majority of the yield fallout was coming from perhaps 20-30% of the wafers. There was no way to test for the mechanical problem at wafer probe. Established a send-ahead procedure, whereby sample chips from each wafer were fast-tracked through production to get a yield figure. Those below a certain number would be scrapped.

Using this process, mechanical production bandwidth was utilized to its fullest, and yields quickly achieved above 80%. Achieved 400 head per week capacity, and yields at first ship were ~85%.

New technology recording heads for MagStar tape drives experienced data problems and sometimes damage in use. Identified that the new high-coercivity metal particle media used for the program had an extremely high surface resistivity compared to the older gamma-iron type. Static charge could build up between the tape and the head, and discharges could damage both head and tape.

Since this had never been seen before, as Lead Engineer, undertook considerable testing to understand the conditions that caused the problem, define and fix it. Directed sampling of output signals of data heads over time to understand the severity of the problem and conditions that might cause it. Tracked laboratory temperature and humidity. Evaluated media samples for resistivity, and realized that the specification allowed it to vary over five orders of magnitude. Tracked the phenomenon to a relatively small cross section of conditions, but enough to warrant a fix. Solved the problem through invention (patent) of an integrated resistor within the head structure to eliminate excess charge.

At an estimated possible field failure rate of 1% to 2%, savings were estimated at $500,000 to $1 million in the first two years on hardware alone, without even counting field service costs that could double or triple that figure. This concept was ahead of its time, and has become even more applicable and profitable with newer technologies.

IBM had zero acceptance in the scientific/engineering community in the use of IBM equipment for technical use. DEC, HP, SUN, SGI and even Apple were preferred for scientific applications, because IBM had no UNIX systems. When AIX was announced (IBM’s UNIX), the dilemma was how to get it out in front of the technical community.

As member of a team of three physicists from IBM’s National Engineering and Scientific Support Center (NESSC), traveled to national labs and universities to demonstrate the easy-to-use AIX systems. Converted users’ applications to run on AIX on the spot (from DEC or SUN, etc.). More complicated conversions were brought back to facility, but in all cases, the team ultimately converted the code and illustrated the IBM system performance capability.

As an overlay organization, the NESSC contributed a great deal to field successes with leading-edge computer users, though this is a small portion of IBM’s overall sales. Many of the lessons learned in this activity were passed on to the developers. IBM over the years has continued to improve on what it started in the workstation division, and continues to have a small but appreciable part of the UNIX market, as well as a considerable LINUX presence.

Testing within China-based supplier showed that signal amplitude of recording heads dropped over time. No performance impact had been seen at HP’s drive manufacturing facility, which was troubling because normally it was highly sensitive to production shifts, even if within specification.

Spearheaded an action team comprised of process and test engineering to determine the cause of the shift. Monitored key metrics, including film processes, mechanical processes, testers, and standards. Results showed no detectable change. Initiated detailed discussions with Chinese engineering staff to find if they knew anything that could possibly be related. They identified that a media batch changed at about the same time. Discovered that this batch had an anomalous degradation with use, not seen before or since.

Resolution of this problem protected $6.5 million revenue stream, and no yield loss was suffered (at a potential $130 per finished head) due to the blip in media wear characteristics. New and more stringent SPC procedures were established to include the tape as part of the tests, to ensure no component (head, media, tester, environment) is allowed to drift.