Session II of the 1975 Lake Arrowhead Workshop on "Advances in Storage for Minis and Micros" was devoted to low-cost tape devices. This article presents a summary of the discussions in that session. No attempt is made to present an all-inclusive coverage of low-cost tape devices, nor necessarily to present a balanced picture, but rather to present the actual material discussed at the workshop. As a result, several types of low-cost tape devices presently on the market are not covered. However, the unique gathering of individual practitioners in the field made this session a lively one, and the article that grew out of it represents a collection of insights that are not likely to be found elsewhere. Low-Cost Tape Devices L.C. Hobbs Hobbs Associates, Inc. Introduction As the price of minicomputers and microprocessors continues to decline, many new applications for computer technology are developing. However, in many cases the use of minicomputers or microprocessors in a particular application has been prohibited by the cost of the necessary peripheral equipment. Many applications that can benefit from the use of computer technology cannot justify the cost of the necessary input, output, and storage equipment, even if the minicomputer and microprocessor were free. Hence, there has been a continuing pressure to reduce the cost of peripheral devices. Perhaps the greatest cost improvements have been made in on-line and off-line secondary storage. The major contributors to enhanced storage capacity and reduced price have been low-cost reel-to-reel industry-compatible tape drives, magnetic tape cassette drives, magnetic cartridge drives, and floppy disks. Session III of this workshop was devoted to low-cost rotating memories, including floppy disk March 1976 drives; hence, they are not covered in detail here. However, it is important to note that in many applications floppy disk drives are a viable alternative and are directly competitive with tape devices in minicomputer and microprocessor systems and applications. The terminology in this area is not completely consistent and logical, but in most cases the term "reel-to-reel" is used to identify tape devices that handle a half-inch magnetic tape reel with start/stop characteristics and read/write electronics designed to provide industry compatibility. The term "cassette" is used almost exclusively to refer to the low cost Philips type cassette originally developed for audio applications and later modified and enhanced for digital recording. The term "cartridge" is generally used to refer to anything that does not fall into one of the two preceding classifications, particularly a device wherein the tape media is enclosed in some type of container or cartridge which also includes the two take-up reels. Until recently, cartridges were generally larger in both tape width 21 and tape length than cassettes, but last year a smaller version of the 3M cartridge was introduced. Reel-to-reel tape drives are used where larger capacity and industry compatibility is required, cassettes where low cost is the primary consideration, and cartridges where a compromise between the advantages and disadvantage of reel-to-reel devices and cassette devices is needed. Reel-to-reel devices are more expensive, whereas the lower-cost cassettes provide lower capacity and lower reliability (at least in the past). The list of possible applications for low-cost tape storage is almost as large as that for minicomputers and microprocessors themselves. The major categories of applications include minicomputer systems, terminals, instrument and test systems, data collection and data acquisition systems, data entry systems, educational systems, and small business systems. Of these applications, cassette devices are more likely to be used in terminals, data collection systems, data entry systems, and small business systems. Reel-to-reel devices are more likely to be used in minicomputer systems, instrumentation and test systems, and data acquisition systems. Cartridge Low-Cost Reel-to-Reel Tape Devices This discussion of reel-to-reel tape drives deals primarily with half-inch IBM format compatible tape drives applicable to minicomputers. It is limited to drives that are 19 inches x 24 inches or smaller, and excludes the "refrigerator" size units built to replace the IBM 2400 and 3400 series. This is certainly an arbitrary limitation, and even some of the faster 19-inch x 24-inch drives may not be applicable to minis. Cassettes, cartridges, and flexible disks are compared in a later table. (Flexible disks are properly the domain of tape engineers since they use contact recording.) The broad line of reel-to-reel products ranges from 7 inch reel tension arm 12.5 ips units to 10.5 inch reel vacuum column units up to 125 ips, some with self threading, all with PE or NRZI recording. Formatters are available for PE or NRZI with or without buffers, and incremental tape drives are available for special applications where a buffered, synchronous drive is not justified. Table 1 shows the typical characteristics of available tension arm units. The storage capacity listed is unformatted-that is, without allowance for inter- Table 1. Half-inch Compatible Tape Drive-Tension Arm REEL SIZE DECK SIZE 7 in. 7 in. Min Interface 8.75 x 19 8.5 in. 12.5 x 19 10.5 in. 19 x 14 MAX DATA RATE CAPACITY ACCESS TIME (UNFORMATTED) 40 20 15 MS 11.5 MBYTE 60 10 MS KBYTES/SEC 72 KB/Sec @ 45 IPS (120 KB/Sec @75 IPS) .7 7.50 23 MBYTE .8 8.50 8.33 MS (5 MS @ 75 IPS) 46 MBYTE 1 11.00 8.5 in. Incremental 12.5 x 19 Read 1000 Ch/Sec Write 500 Ch/Sec IRG Time 550 MS (800 BPI) 11.5 MBTYE 10.5 in. Incremental 19 x 24 Read 1000 Ch/Sec Write 500 Ch/Sec IRG Time 550 MS (800 BPI) 23 MBYTE devices offer an alternative in any of these applications depending upon the relative importance of cost versus capacity and reliability. In the following discussions, contributed by individual participants in the workshop session (see acknowledgements), a general overview is given of reel-to-reel and cassette devices, but the presentation on cartridge devices is restricted to a specific new type of cartridge device that offers enhancements and performance characteristics over previously available cartridges. However, it is important to note that two or three other cartridge devices using different types of cartridges are also in production and are in widespread use in the industry. 22 MEDIA COST FACTOR COST (.5) 8.50 11.00 record gaps. The efficiency is probably around 60% but subject to record length. Access time is really start time to the next record since the use of tape assumes random access is not a requirement. Data rate is the product of the maximum speed available and the maximum recording density-i.e., 1600 BPI. Defining a cost factor is always difficult, but in this case cost is normalized to 1 for the basic 10.5 inch reel tension arm tape drive. The purpose is to show the effect of reel size on cost. An 8.5 inch reel unit costs 80% of a 10.5 inch reel unit, and so on. Two special cases are shown in the table. One is a low-cost 7 inch reel "mini" at a cost factor of 0.5. This shows the cost difference for a unit with a read/ COMPUTER write head (as opposed to read while write) with NRZI and minimum frills. The 0.7 cost factor is a fully equipped mini with maximum speed, dual stack head, and PE data electronics as are the other units. The other special case is a 75-ips tension arm unit with 10.5 inch reels. This is a relatively low usage product designed for the application where high data rate is necessary, but with low standby power, low acoustic noise, etc. This is not the most costeffective way to get 75-ips operation, but it serves well for special applications where power and/or noise restrictions are especially severe. All of the units are available as 7-track or 9-track, with 9-track offered as NRZI or PE (1600-bpi Manchester code) or both. One other standard option is a read-only package, meaning, really, read anything7-track or 9-track; 200-, 556- or 800 bpi NRZI; or 9-track PE. This option is useful for computer output microfilm, off-line print stations, etc. All of the units may be obtained with formatter for full data recovery with NRZI, PE, or both, with various size buffers if necessary. "Industry standard" interface really exists so that formatters from one vendor work nicely with drives from another. are too short for reliable operation above the 75-ips range. The next higher category of drive offers a more complex tape path with longer storage columns, the possibility of autothread, and compatibility with easy load cartridges. These drives cost considerably more than the simpler drives, even if used at lower speeds, but are very cost effective at 125 ips compared to the larger "refrigerator" size drives that are really 200-ips drives slowed down. Group code recording (GCR) is discussed later but is shown in Table 2 to illustrate what happens to access time, data rate, and st6rage capacity. These vacuum column drives are also offered with NRZI, PE, both, read only, appropriate formatters, etc. One surprise is the relatively low usage of the PE format, even now. Only about 50% of the sales of compatible tape drives are PE, even though the format with its advantages of single track error correction and freedom from skew problems has been around for almost 10 years! One explanation is the complexity of data recovery as compared to NRZI, but the formatter to handle this is readily available and is well understood. One other explanation may be that skew (the inac,curate guiding of tape across the head) hasn't really Table 2. Half-inch Compatible Tape Drive-Vacuum Column SPEED RANGE MAX DATA RATE KBYTE/SEC ACCESS TIME CAPACITY (UNFORMATTED) COST FACTOR MEDIA COST 7.5 MS 46 MBYTE 1 11.00 120 5 MS 46 MBYTE 1.1 11.00 75-125 IPS 200 3 MS 46 MBYTES 1.5 11.00 125 IPS w/GCR 780 1.2 MS 180 MBYTES 1.7 11.00 25-50 IPS 80 50-75 IPS Most of the drives shown are available with 48 VDC primary power for telephone equipment application, as well as the usual 56/60 cycle AC. Table 2 shows the available ranges of competitive tape drives using vacuum column storage. The cost factor is again normalized to 1 for a 10.5 inch reel tension arm unit-that is, the cost factor of 1 here is the same as in Table 1. This shows that vacuum column tape units cost about the same as tension arm units in the 45-ips range, and less than tension arms at 75 ips, assuming no extra frills. The 75-ips vacuum column unit with the simplified tape path (i.e., no autothread) is one of the most cost-effective devices available. There is a break in the cost performance above the 75-ips area where costs increase more rapidly, as indicated in the chart. The lowest cost vacuum column tape path arrangement is not "autothread," and the storage columns March 1976 been such a problem after all. The manufacturers of mini drives have almost totally held to the head guide geometry of the 2401 series IBM drives. In so doing, one of the major interchange problems that faced the large mainframe manufacturers in the 60's, when "trough" guide machines from one supplier were asked to interchange with pc ' guide units from another supplier, is minimized. A detailed discussion of this interchange skew problem is more involved than can be discussed here, but Juan Rodrigues of Storage Technology Corporation has covered it nicely in the August 1975 Proceedings of the IEEE. Another surprise is the very low usage (5% to 10%) of hard faced heads though they have been around for several years, too. These heads offer three to five times the wear life of ordinary metal heads at initial cost increases of about 20% of a new head, not counting the installation expense. These are a good deal, and users are missing out! 23 Table 3. Small Storage Device Comparison DEVICE TYPE Half-Inch Compatible Tape w/7-Inch Reel (Min Configuration) RECORDING TRANSFER RATE CAPACITY KBITS/SEC (UNFORMATTED) DENSITY SOFT ERROR RATE RELATIVE COST MEDIA COST 100 MBITS 1000 BPI 360 (25 IPS) 10-9 1 7.50 22 MBITS 1600 BPI 160 (25 IPS) 10-8 .4 13.00 Philips Cassette 7 MBITS 800 BPI 19 (12 IPS) 10-7 .2 4.00 Flexible Disk 3 MBITS 3248 BPI 10-9 .25 8.00 3M Cartridge Table 3 compares the lowest half-inch compatible drive to other contact recording systems. Each type of device has its place, although the future of the Philips cassette seems questionable when dual density flexible disk is compared. This would mean about the same cost per bit stored for cassette and flexible disk, but with access and reliability somewhat better for flexible disk. Our best cost reduction efforts this past the line against inflation. The cost of heads, the biggest single item in tape units, is up substantially, partially due to the loss of production in Portugal. year have not even held But before considering where we are going, let's consider where are we not going. Much reduced prices for the half-inch tape drives appear unlikely. Most factors that determine price indicate the price should go up, even though the cost is already high compared to the cost of the processor in some applications. Our best cost reduction efforts this past year have not even held the line against inflation. The cost of heads, the biggest single item in tape units, is up substantially, partially due to the loss of production in Portugal. The recent changes to the UL requirements for plastics have also increased costs about 25% for many parts. This is a good standard and we fully support it, but the cost goes with the benefit. The cost of most materials is up; the new semiconductors are the only bright spot other than just trying harder. One popular idea that's been around awhile is that of a $1000 mini-i.e., a 7-inch reel half-inch IBM-compatible tape unit with the usual features to sell for $1000 in small OEM quantities. We would like very much to offer such a device, but we seem to be getting further away-not closer. Brushless DC motors don't seem to offer much since the brush types have very acceptable life. Major use of custom LSI devices in the product seems unlikely, though standard and special purpose devices look attractive. The production quantities 24 250 The control logic on a typical tape drive only costs about $9 now, so there isn't much room to move. The money is in heads, motors, and other mechanical parts. simply not high enough to justify custom LSI, since in most possible applications only one device of a given type per drive would be needed. The control logic on a typical tape drive only costs about $9 now, so there isn't much room to move. The money is in heads, motors, and other mechanical parts. Then what can we expect? One area where LSI looks good is in formatters. A single, small formatter for NRZI/PE from 12.5 to 125 ips mounted integrally to the tape unit is now possible. The potential savings in power supply and a package make this an attractive approach. Another activity that absorbs considerable effort is compliance with new standards for safety, RFI, acoustics, etc. More work will be done in these areas, just as in the automobile industry. Table 4 shows a partial list of standards and/or requirements that are becoming more important. As mentioned earlier, the result here is good, but there is a price associated with benefit. We expect to be borrowing from the "big boys" in a few areas and passing the improvements in performance along on lower cost drives. Group Code Recording (6250 bpi) is an obvious addition. The changes in storage and transfer rate were shown previously on the vacuum column chart (Table 2). IBM-compatible GCR on vacuum column units at speeds from 25 to 125 ips will be offered soon. The associated formatter will be considerably more complex than current units, but for those who need it, the benefits will be available with minimum system impact other than data rate-i.e., a 25-ips GCR unit could be used where 100 ips 1600 bpi was required with no system impact. Along with GCR, it is reasonable to expect changes to tape cleaners for more effective cleaning, capstan systems that use digital feedback devices rather than analog, more use of LED's, and more built-in test equipment. Emphasis will be on durability and serviceability-i.e., lower cost of ownership. are COMPUTER Table 4. Applicable Standards SAFETY STANDARDS U.L. 478 (10/1/75) . . . . . . . . . U.L. 114 . . . . . . . . . . . . C.S.A. C22.2 #154 . . . . . . . . . National Fire Protection Association . . . International Electrotechnical Commission . British Standards Institute . . . . . . International Commission on Rules for the Approval of Electrical Equipment . . . VDE (Association of German Elec Engrs) -0804 . . . . . . . . . . . . -0730 . . . . . . . . . . . . RFI . . . . . . . . VDE 0875/7.71 CISPR (International Committee on Radio Interference . . . . . . . . . FCC . . . . . . . . . . . . . . . . . . Electronic Data Processing Equipment Office Appliances and Business Equipment Data Processing Equipment, 1975 National Electric Code, 1975 IEC435 B.S. 3861 . . C.E.E.10, Part II . . . . Regulations for Telecommunication Devices Motor Operated Appliances . . Regulation for RFI Suppression, Conducted and Radiated . Publication 1, 2 Specification for CISPR - RFI Measuring Apparatus (150 KHz to 300 MHz) "Coming Up" . . . . . . ... . . . ACOUSTICS ISO 1996. Assessment of Noise with Respect to Community Response Lastly, there appears to be a need for a new ma storage device possibly based on one of the wi tape cartridges. It's too early to say what form su a device would take, but the availability of a ni media opens up several possibilities, one of whi will probably be made available for mini systems. A New Cartridge Tape Device Tape cartridges and drives have been availal for several years from 3M, IBM, Tridata, a others; but the search for an improved cartrid system has continued. The Emerson Tape Pac (see Figure 1) is a ni magnetic tape storage unit which provides the rel bility, interchangeability, capacity, and versatili to meet the needs of a wide variety of compul For an instrumentation tape recorder manufacturer, the bane of any application is tape threading-a highly sensitive precision instrument is exposed to the human operator where all sorts of horrible things can happen. The obvious need is an instrumentation cartridge. applications. A brief description of the background of the development will be of interest and will put the Tape Pac and our views on the future of storage in perspective. For an instrunentation tape recorder manufacturer, the bane of any application is tape threading-a highly sensitive precision instrument is exposed to the human operator where all sorts of horrible things March 1976 Figure 1. Emerson's Model 2000 Series Tape Pac System (photo courtesy of Emerson Electric) can happen. The obvious need is an instrumentation cartridge. In general, either self-threading or tworeel containers are highly impractical. High-density recording techniques are required to reduce the tape capacity needed to meet mission data requirements. At the same time, until a cartridge design could be found which provided the reliability, interchangeability, and precise tape handling (i.e., low wow and flutter, low time base error) instrumentation cartridges were not practical. A similar condition can be stated for tape storage for computers when the broad spectrum of computer tape applications are considered. A practical broadbase application of cartridges to computer storage requires the following: (a) high-density recording; (b) a design of sufficient precision and simplicity to yield reliability and interchangeability; 25 (c) a design which can be produced at low cost; and (d) a large enough capacity to meet most computer applications. A summary of the status of high density recording in both digital instrumentation and commercial computer applications will help establish the background for the development of the Emerson Tape Pac. In digital instrumentation applications, 16.6 kilobits/in. is commonplace, and 30 to 50 kilobits/in. systems are being evaluated. These higher densities are not ready for commercial applications because the tape needs improvement and the heads and electronics are too expensive. There has been much discussion on encoding techniques for optimum performance, and the balance between acceptable dropout performance and overhead for error correcting coding is still being explored. In commercial computer applications, 1600 BPI (9 track parallel) is routine. Although 6250 BPI (average-approximately 9000 BPI on tape) is available, it puts excessive demand on present tape quality and requires excessive overhead in error correcting. Hence, 3200 to 4000 BPI is the most practical operating density range. It is consistent with tape quality, saturation recording remains practical, and simple electronic recording and reproducing techniques are applicable. The Emerson Tape Pac was developed in this environment. This device uses a simple reel-to-reel configuration as shown in Figure 2. The C-2002 Tape Pac contains 600 feet of half-inch magnetic tape. Because of the precision die-cast housing construction, the use of ball bearings, and the simple two-guide design, tape transfer between reels is very uniform with low wow and flutter and small instantaneous speed variation. The accuracy and rigidity also provide reliability and interchangeability which are most important to any application. A brake is also provided internal to the Pac to prevent tape slacking in handling. The brake is designed to tighten the Pac when exposed to vibration. A spring-loaded hinged cover is also provided to protect the tape from damage during handling and exposure to the environment. The Model 2004 Tape Pac is driven peripherally by the twin-capstan Model 2005 Pac Drive, as illustrated in Figure 3. When in place, each capstan engages a reel. Each capstan in turn is driven by a motor. The relative speed control of each can be programmed to yield the proper tape speed and also precise control of tape tension. From end to end, acceleration or steady state, the tape tension can be maintained to 6 ± 1/2 oz. This results in long life and precision speed control. Fivethousand passes are readily obtained. At this juncture, the design has been tested at speeds from 15/1, ips through 240 ips. At 240 ips, or 1200 feet per niinute, the traverse time of the Tape Pac is 30 seconds. In some applications in which IBM compatibility is desired, start/stop times become an important consideration. The Tape Pac has no vacuum columns or compliance arms, but IBMcompatible recording can be produced up to speeds of 60 ips. 26 The characteristics provided by the Model 2004 Tape Pac and the Model 2005 Pac Drive are summarized below: * 600 feet-1 mil-half-inch tape (1000 feet-1/2 mil) * die cast housing * ball bearings * two tape guides * brake BALL BEARING MOUNTED REEL HUBS STORES 6DO FT OF 1 MIL X 1/2 INCH WIDE TAPE SPRING LOADED DUST DOORS PRECISION TAPE GUIDES NO OXIDE CONTACT WITH TAPE GUIDES TAPE PACK SIZE: DEPTH 5.0 INCHES 8.6 INCHES WIDTH HEIGHT 1.1 INCHES 2 LBS WE!GHT BRAKE MAINTAINS TAPE TENSION WHEN TAPE PACK IS OUT Of THE TRANSPORT Figure 2. Exploded view of Tape Pac (figure courtesy of Emerson Electric) Figure 3. Tape drive concept (figure courtesy of Emerson Electric) COMPUTER . * * * * * * * * * * * * * * * hinged covers low wow and flutter or low instantaneous speed variation reliable interchangeable good handling average tape speed-control-0.25% (a function of phase lock oscillator only) ISV-0.5% p-p at 30 ips tape tension 6 oz + 1/2-oz-end to end-accel/ decel tape passes-5000 minimum tape speeds-15/16 ips-240 ips start/stop times-35 milliseconds at 60 ips twin capstans individual reel peripheral drive phase locked tapeup reel control supply reel drive slaved to take up reel control acceldecel ramp control It can record IBM-compatible tape (7-track or 9track) at any density desired with NRZ-I or Phase Encoded quality consistent with existing designs. It can be used in a serial format up to 3400 bpi, 8 tracks, 180 megabits without redundancy or error correcting. Less than 1 in 108 permanent error rate has been observed after 5000 passes. A 10.8 KB/inch, 8-track serial design has been committed and 16.6 KB/inch densities definitely feasible. It is capable of time expansion or contraction over a wide speed range, and it eliminates tape handling problems inherent in conventional tape transport designs. Its inherent simplicity of design yields greater reliability, and the totally enclosed environment provides ruggedness and other requirements of hostile environment applications. Digital Cassette Tape Drives The Philips Cassette is a small, convenient, lowcost, reusable, non-volatile magnetic storage medium that can easily store lOOK bytes, or as many as 550K bytes in an ANSI 2-track format, or as many as 2 megabytes in a special 4-track configuration. One of the advantages of a cassette is that it comes in many lengths: standards are 150, 300, and 450 feet; noncertified cassettes contain as much as 600 feet. The Philips Cassette is the most widely used medium since paper tape and punched cards. There is a digital machine population of well over 500,000 units at the present time, and this number is growing at a rate of 150,000 units per year. This rate is predicted to continue for several years. On the basis of storage capacity, data rate, access time, cost (both entry costs and media costs), and media reliability, the comparison of the cassette with other storage devices is as follows: Storage Capacity (see Figure 4). At the high end, special cassette units can equal the storage capacity of a 3M cartridge unit using 4 tracks. ANSI-compatible units are about double the capacity of present day floppy disk units when two tracks are used, and lowMarch 1976 performance cassettes are roughly equivalent to 1000 feet of paper tape, or 1300 80-column punched cards. Data Rate. Data rates vary from very slow asynchronous, by character or bit rates, up to 6K bytes/ sec. However, typical transfer rates are generally in the range of 1K to 2K bytes/sec. Therefore, on the high end, this puts the cassette on a par with a 3M cartridge unit reading one track at a time. However, they are both an order of magnitude slower than a flexible disk unit which typically handles data at 31.25K bytes/sec. It is important to point out that data rates are not generally a major consideration when selecting a tape drive. Rates can easily be changed using buffering techniques, which are frequently required anyway; moreover there are other tape positioning times which can be much longer than the block transfer time. Access Time. There are two kinds of access time: the response time to a command when the unit is in position to read or write, and the block search time. Response Time. Typical numbers are 15 ms to 20 ms for intermediate tape speeds, but they can be as high as 50 ms to 100 ms at higher speeds (20 ips or higher). Search Time. The average for 300 feet of tape, assuming random block distribution, may be as low as 25 seconds, but on low performance units can be 100 seconds or more. Summary. The access times for cassettes are equivalent to the 3M cartridge units, but are over an order of magnitude slower than flexible disk units which average about 0.6 seconds for completely random data distribution (which usually isn't the case for cassettes or disks). Cost. Two kinds of cost can be considered: entry cost and media cost (see Figure 4). Entry costs are heavily dependent on application details, so a general comparison cannot be made. However, all pertinent system costs should be included when making such comparisons. For example, interface circuitry, data formatting circuitry, data buffers, software, and program storage requirements, if any, should all be included. Data stored on cassettes typically costs around 1.5 to 2.0 millicents/byte, but can be as low as 9.5 millicents/byte. This puts cassettes on a par with the 3M cartridge a\t the low end, and still half the cost/byte of a floppy disk at the high end. Media Reliability. In order to compare media reliability, one must be specific about application details such as hardware, duty cycle, and environment. There are occasional claims, usually from cassette competitors, that cassettes are only good for 300 passes or less. In some applications this is true, but typical numbers using certified cassettes, in a good transport, in an office environment, are 2000 to 5000 end-to-end passes, and when re-reading a short record, they are capable of 10K to 20K error-free passes. Present day digital cassette drives come in two basic versions: capstan drive units with as many as four motors, and hub drive units with up to two motors. Capstan Drive Units. This type of drive comes in two versions: dual and single capstan models. 27 CAPACITY (BYTES) 30M 3 10 M I REEL TO REEL (HALF-INCH, 2400 FEET) COST* (10-4 d/BYTE) 40 + FLOPPY DISK LOW PERFORMANCE CASSETTE 20 - (PAPER TAPE) 15 - - ANSI CASSETTE 1013M CARTRIDGE 5- HIGH PERFORMANCE CASSETTE 3M CARTRIDGE (4T) 2 M- HIGH-PERFORMANCE CASSETTE iM 1- 550 K ± ANSI CASSETTE (2T) 0.5 ± REEL TO REEL 250 K - FLOPPY (ONE SIDE) L LOW-PERFORMANCE CASSETTE 100 K -JI- (PAPER TAPE) 0.1 _L *Assuming complete utilization Figure 4. Tape cassette capacities and costs Dual capstan drives usually have two counterrotating capstans in order to "pull" the tape across the head in both directions. They use from one to four motors, but typically they have three motors (one capstan and two reel motors), two pressure roller mechanisms, and a head retraction mechanism. This makes this type of drive the most mechanically complex of the types that will be discussed. They generally use a high-inertia capstan system which rotates all the time, and therefore uses a "banging" pressure roller to start the tape. This is relatively hard on the tape and therefore reduces the media reliability. However, this type of unit has the lowest speed variation and time base error. Also, the control electronics is simple (especially on units with an AC motor), and the read/write circuitry can be of "fixed window" variety, which uses the fewest number of components. Single-capstan drives reduce the mechanical complexity somewhat by eliminating a capstan and pressure roller mechanism, and they generally use a lowinertia capstan which is capable of accelerating and decelerating the tape without the need for a "banging" pressure roller. Therefore, this type of drive is easier on the tape-and somewhat more reliable, but they either cannot run in the reverse direction, under capstan control, or they must defeat the pressure pad in the cassette. This is generally accomplished either by using cassettes Without pressure pa4ds, or by pushing the pad back so that it does 28 not drag on the tape. This type of unit has good speed accuracy; however, the T.B.E. (time base error) is generally worse due to the low inertia drive. The control logic and circuitry is usually more complex due to the servo; however, the read/write circuitry generally can still be of the "fixed window" variety. Hub Drive Units. These come in 2 versions also: constant tape speed units, and constant hub speed units. Constant tape velocity units generally have two motors which drive each reel. They don't use pressure rollers or capstan shafts, so the mechanical complexity is further reduced. They achieve constant, or nearly constant, tape velocity by using either a second track with a timing signal pre-recorded on it, or some type of motor EMF or tachometer differencing scheme for both hubs, or they use a programmed velocity profile which corrects the speed depending on the distance from the beginning of the tape. The units using timing tracks have good speed accuracy, but T.B.E. is worse than most capstan drive units. The other two types of units have relatively poor speed accuracy (from ± 5 to ± 1 0%), and are adversely affected by packing irregularities, tape thickness variation, and length of tape in the cassette. However, because there are no pressure rollers, this type of unit is easier on the tape than capstan drive units. The control logic and circuitry is probably the most complex because of the two servos and the motor coordination that is required. The read/write COMPUTER circuitry uses either a two-track or a single-track speed-tolerant decoding technique. Because of this, data density on tape is usually reduced and the circuit complexity is increased. Constant hub velocity units use either one or two motors and have the simplest mechanism of all the types of drives discussed. Since one of the reel speeds is constant (generally the takeup reel), the tape velocity varies with the radius of the tape pack, which is about 2.5 to 1 for a 300-foot cassette. However, the speed profile repeats from pass to pass so that fairly constant data rates can still be achieved. The control logic and circuitry are generally very simple, but the read/write circuitry is the most complex because it must be of the speed-tolerant variety in order to provide unit-to-unit interchange. Average data density on the tape is low because the maximum packing density is achieved only at the lowest tape speed. In addition to this, when two-track read/write schemes are employed, the cassette storage capacity is typically about 25% of a comparable capstan drive unit. Hub drive units reduce mechanical complexity, at the cost of electrical complexity and reduced storage capacity; however, the overall reliability is somewhat improved. On the other hand, capstan drive units still offer good reliability, improved edit and rewrite capability due to more precise tape handling, and good media life (though somewhat lower than hub drive units, it is still in the range of 2000 to 5000 end-to-end passes on a well-designed transport with a certified cassette). Cassettes will continue to be used in even greater numbers in the low-cost, low-performance applications but entry costs must be reduced and reliability must be improved. The use of high-performance cassette drives will tail off because there are fewer applications and because of competition from other types of memory devices. Because of a more liberalized attitude throughout the industry, new types of competitive devices for the cassette may appear on the scene. This is already apparent because of several new products which have been introduced, or are about to be introduced such as the "Reelette " which American Videonetics Corp. has introduced, and the Mini-3M cartridge which H-P has introduced in some of their equipment. Interdyne is very interested in this low-cost, high-volume OEM market, and has some new products in the works. Conclusion Throughout the history of the computer industry, electromechanical peripheral equipments have been hard pressed to keep pace with advances in internal logic and memory. However, continued development of simpler mechanisms and increases in bit density on magnetic storage media have tended to provide lower-cost, higher-capacity secondary storage media to complement lower-cost processors. In most cases, a reasonable system balance is possible now for minicomputers; but microprocessors (including several March 1976 thousand bytes of storage) priced in the few-hundreddollar range put new pressures on peripheral equipment development. In many cases the total cost of secondary storage devices is the major criterion rather than the cost per bit. Hence, very low cost storage devices such as magnetic tape cassettes and floppy disks take on added importance for smaller equipments and systems. Although speed and capacity can be compromised to achieve a lower price, in most applications reliability cannot be sacrificed. Hence, there is a continuing need for cost reductions in storage devices which must be achieved without sacrificing reliability. Since no single device is likely to provide the optimum characteristics of low cost, high speed, high capacity, industry compatibility, and random access, we will continue to see requirements for and enhancements in the cost/performance ratio of all types of secondary storage devices, including reel-to-reel magnetic tape drives, magnetic tape cassette drives, magnetic tape cartridge drives, and floppy disk drives. . Acknowledgements The information presented in this article is the direct contribution of the participants in Session II of the 1975 Lake Arrowhead Workshop, and full credit is due them for the material presented herein. The material on low cost reel-to-reel tape drives was contributed by Mr. Robert Tullos of Pertec, the material on the new cartridge tape device was contributed by Mr. John Coolidge of Emerson Electric, and the material on digital cassette tape drives was contributed by Mr. Richard Lewis of Interdyne. Linder Charlie Hobbs has worked in the design, analysis, and evaluation of electronic systems, information processing systems, computers, displays, terminals, memories, and peripheral equipments since 1948. He held technical and supervisory positions at RCA and UNIVAC and was manager of Data Processing Engineering at the Aeronutronic Division of Ford Motor Company. Since _ 1962, he has been president of Hobbs AssociNl a m ates, Inc., which provides computer timesharing and consulting services. Recent work includes evaluation and forecasting of technology and markets for computer equipment and systems for industrial and governmental organizations in the US, Canada, and Japan. He has published many papers on computer technology, has organized several panel discussions and workshop sessions, has edited a book, Parallel Processor Systems, has taught computer systems courses at UCLA since 1957, and holds thirteen patents. He has a strong interest in conferences, publications, and other means of disseminating technical information. Active in IEEE and AFIPS, Hobbs has served as chairman of the IRE Computer Standards Committee, chairman of the Philadelphia and Orange County Chapters of the Computer Society, chairman of the Orange County Section of IEEE, chairman of the 1967 FJCC, chairman of the Computer Group Administrative Committee, and as a member of the Board of Directors of AFIPS and the IEEE. 29