How far can tape guide rollers go? Is a 3-piece design the future?

Tape continues to be the cheapest, most convenient way to store large amounts of data. Cartridges of 200 and 300GB in half-inch-wide format are now common. Soon, 500GB and 1TB capacities in one cartridge will be available.

Increasing cartridge capacity is directly connected to increasing the track density. However, track density is primarily dependent on tape path stability. Other factors, such as a wide-band servo with the ability to follow a jittery track, only partially compensate for poor guiding. The servo wouldn't need to be so robust if the track stability were more controlled. Therefore, steady tape guiding is fundamental, and the issues influencing it must be thoroughly understood. Rollers have come to play a primary role in guiding the tape, but how far can rollers go?

Long Paths in the Past

In the IBM 3590 era of long tape paths and large tape drive decks (Figure 1), unevenly stacked tape on the reels had a long time to adjust to smooth level motion over the head. Tape was mainly guided from the cartridge reel to the machine reel over a series of air-bearing guides, called D-bearings. This was completely satisfactory for 36-track tape.

[FIGURE 1 OMITTED]

These guides supported tape on a film of air supplied by tiny air holes in the smooth surface and connected to a small air pump. The tape's bottom edge was registered against a guide edge, usually ceramic (Figure 2). On the top edge, ceramic pads on flexible fingers pressed gently down to register the tape against the bottom guide edge.

This worked fine for wide tracks in the region of 250 microns. However, when tracks become 25 microns--as today's LTO, Quantum, and StorageTek drives are--problems arise. Lateral tape vibrations caused by the tape rubbing against the guiding surfaces become significant. These movements are sometimes more rapid than the tracking servo can follow. Movements above 500-800 hertz are common, and many tracking systems cannot follow them. This leads to significant position error signal, or PES. For example, if a tape moves 10 microns and the head can only follow 9 microns of movement, 1 micron of PES is generated.

The trend toward smaller and cheaper drives prompted the abandonment of hydrostatic bearings--that is, bearings with air supplied to them to float the tape in a static mode. Hydrodynamic bearings were substituted, which develop an air film when the tape gets up to speed and flies over the surface. This was thought to be a cost-effective approach because no air pump was needed. Moreover, the air pump was a space liability in the small and compact 5.5-inch form factor of modern half-inch cartridge tape drives. However, guide vibrations and edge wear problems persisted. The embracing of rollers was the result.

[FIGURE 2 OMITTED]

Tape guide rollers had been used in many tape transport systems, especially video systems. Compact rollers were common in half-inch tape video decks of professional broadcast level. Some manufacturers had as many as 13 rollers in the complex helical scan tape path. Because the tape moved relatively slowly, 3-4 inches/second, severe edge wear did not develop. Rollers were thought safe.

Short Paths Now

However, tapes speeded up. Two meters per second, then four, and then six became common in order to have a high data rate and to read a 600m length cartridge tape in a reasonable amount of time. Even more challenging, tape path lengths became very short. Instead of the ample tape path of about 1m long, as in the 3590 family of tape decks, paths were shortened to about 15mm (Figure 3). This forced the tape coming off the reels to meet the first guide roller at 5mm distance rather than 20-25mm. The result was poor guiding and edge wear.

Severe edge pressure came from the flanges at rollers 1 and 4. It is the job of these rollers to reduce lateral excursions that come from uneven wrapping on the reels, called stagger wrap or scatter wrap. If these could be reduced, the problem of guiding tape would be made much easier. However, rollers 1 and 4 cannot do the whole job. Most manufacturers agree that the tape has to be guided by stages. To this end, various vendors have designed rollers with the outer ones different from the inner ones. They are either wider or have a different surface finish.

[FIGURE 3 OMITTED]

Guiding and Flying

The outer rollers push the tape toward the inner rollers or D bearings next to the head. These outer elements do the brunt of steering the tape. The inner elements do the final guiding and also damp out vibrations normal to the tape plane to present a smooth motion over the head.

The tape flies over the outer rollers, touching the surface only occasionally. One can check this out by examining the outer roller rotation with a strobe light to see that it is irregular and not synchronous with the tape velocity. This slipping allows the tape lateral freedom to adjust to the roller flange's guiding pressure. On the inner elements, it is necessary for the tape to adhere to the surface in order to damp out vibrations.

[FIGURE 4 OMITTED]

Consequently, the outer roller usually has a smooth surface to support an air film and the inner roller has a grooved surface to bleed off the air film. There are arguments about how aggressive the groove should be and whether this influences the read/write quality. The answer must lie with testing a particular geometry. I know that it is possible to have grooved geometry on rollers 2 and 3 next to the head that does not affect the read/write characteristics.

[FIGURE 5 OMITTED]

With a D bearing next to the head, it is not possible to damp out normal tape vibrations. Moreover, the fingers and guiding pads can excite frequencies at just the place one doesn't want them--next to the head. This is why the tendency is to get away from D bearings next to the head and use rollers. D bearings were fine in the large deck days, but designers have learned that short distances make them risky in the small deck era.

Roller Characteristics

Since the choice narrows to rollers, what kind of rollers should be used? May they not also present risks that were not evident in the large decks and longer tape paths? The answer is: Yes, they do present risks to damaging tape in their own way. Figure 4 shows the tape pressing hard against the guiding flange. Since this is a high stress point on the tape, it is important that the roller be gentle to the tape at this location. However, this is usually not the case. Designers have chosen one of three styles, none of which is easy on the tape. There is a 1-piece roller with a small corner radius, a 1-piece roller with a corner undercut, and a 3-piece roller with a spring loaded flange.

[FIGURE 6 OMITTED]

The 1-piece rollers are chosen because they are cheaper since they are machined out of a single piece of metal (Figure 5). However they must have a tight radius where the flange meets the barrel. A practical radius would be .50mm (.020"). However, this is 70 times the thickness of 7-micron thick tape used today. This is so large compared to the tape thickness that it bends the tape outward and, after many cycles, scallops and stretches the edge beyond yield.

The second type (Figure 5) has an undercut at the flange/barrel junction to eliminate the radius problem. However, the undercut is itself so large compared to the tape that the tape folds into the undercut, especially under tension. After a number of cycles, it has a pronounced crease on the edge from the undercut.

The third type (Figure 6) suffers from the same problem. In an effort to make the flanges exactly the width of the tape in order to guide better, the upper flange is spring-loaded to bear on the upper edge while the lower edge is registered on the lower flange. The roller barrel is slightly shorter than the tape width and the effect is to create the same undercut as before.

The answer, I believe, is to make a 3-piece roller (Figure 7) with exceptionally smooth flanges and a barrel, perhaps .05-.10mm (.002-.004 inch) wider than the tape. This presents a smooth guiding surface and a firm corner for the tape to bear against without bending it outward, as with a radius, or inward, as with an undercut.

Flange Finishes

Flanges on a 1-piece roller can be very abrasive to the tape. This simply follows from the fact that they often have machine marks as a result of being made too fast on an automatic lathe machine with too dull a cutting tool. These marks can be like a file to the tape edge. Even the finest feed will produce them, and tape rubbing across them wears. Moreover, there is evidence that tape drawn across this surface vibrates much the same as a violin bow vibrates the string when drawn across it. The answer is to make the flange separateiy and finish it to a fine surface condition. Ceramic flanges should be considered for this as they can be finished to a polish for a reasonable price. I believe this approach is cost-effective in the long run, allowing smooth running, non-wearing tape which supports high density tracks.

How Far?

How far, then, can rollers go? I believe that with the above suggestions put into practice, a 4-roller tape path should take the tape drive capacities into the 1-terabyte territory. Care taken with these approaches, together with precision-but-current machining practices, can reduce high-speed lateral motions enough to support an 8-10 micron track width. Other factors, of course, are needed for 1TB. These include closely spaced read/write elements to support a 10-micron track width, very fine head gaps together with hot media to support high lineal density, and matching channel improvements. The track-following servo needs to be capable of 1 KHz bandwidth. I do not believe that tape speeds need to increase significantly over 6m/sec for capacity reasons. However, quick access time may prompt speed increases to the 12m/sec region.

[FIGURE 7 OMITTED]

What about beyond that? The 1TB half-inch (12.5mm) tape drive is only several years away. What about doubling capacity several times again, or an order of magnitude? For those considerations, a 1-micron track is needed.

[FIGURE 8 OMITTED]

To do that, tape path technology may have to return to an old idea: Air-supported D bearings. However, they would have to be bearings without pressure fingers. Fingers induce vibrations that cannot be tolerated. I have seen a porous, air-fed ceramic bearing developed over the past several years that shows promise. The tape is registered against the lower support guide by tilting the bearing less then a degree outward. The result is lateral motion of less than .5 microns amplitude at 4 m/s lineal speed (Figure 8). This compares favorably with tape lateral excursions at the head of 10 to 20 microns typical in present LTO drives. Several companies and universities are using this technology now to do advanced testing. It also has the potential to form the basis of the next-generation servo writers.

Track Following Improvements

As tape speeds get faster and tracks get finer, track following becomes more difficult. PES for a given system increases in proportion to the increase of speed. Simple testing can demonstrate this. It follows simply from the fact that doubling the speed, for example, doubles the rapid lateral movements of the tape. A lateral frequency that was 500 Hz is now 1000 Hz. Therefore the bandwidth of the servo system has to keep up with the increase in speed. However, it is not easy to increase bandwidth. The simple brute force approach, pouring more current into the actuator coil, won't make up for other actuator weaknesses.

Refinements in servos will be needed. At present, two methods are used to move the head: moving along ceramic rails on roller bearings, and moving it on flexures. The roller/rail combination suffers from having to overcome static friction to get it moving. This leads to hysteresis and makes movements of .5-micron almost impossible. But that is what will be needed to follow tracks of less than 10-micron pitch. The other approach, flexures, has resonant problems that are inherent in any beam/spring system. They offer freedom from static friction and therefore hysteresis, but their resonant frequencies have to be skillfully chosen to out of the range of track-following interest--either far above or far below.

Is it possible, then, to design a track-following scheme that has neither flexures nor roller bearings? One method might be a piezo structure within the head matrix to finely adjust individual read/write elements. Another might be to utilize frictionless air bearing guides. This virtually eliminates hysteresis and resonance problems, making minute movements possible with very little coil current. Such a system is under development.

Conclusion

Increasing track density and shorter tape paths has strained the ability of guiding elements to deliver smooth motion over the head. Hydrodynamic or hydrostatic air bearings have the serious drawback of needing spring fingers. Rollers made in one piece have problems including undercuts at the flange that draw tape in or radiuses bend it out. This lowers the effectiveness of the flange to guide. Moreover, flanges themselves often are rough and cause wear. Nevertheless, if these problems are corrected, rollers have the possibility steadying tape sufficiently to support 1TB on a half-inch tape.

Beyond that, returning to a high-precision D-bearing without fingers shows the promise of supporting track densities down to 1-micron width. A similar technology for head actuators allows track-following minute movements at high speeds. That should support cartridge capacities in the 10TB range and beyond. Just for comparison, 1.2 TB was the total capacity of the first practical cartridge silo-sized library from StorageTek when it was introduced in 1985. It contained 6000 cartridges of 3480 style containing 200MB each.

By Gary Collins, P.E

Gary Collins, president of Collins Consulting (Boulder CO), is a professional mechanical engineer and has consulted for the storage industry for over 20 years.

[e-mail]