Fullscreen HTML5Med Res (??MB)HTML4

Data Storage Systems Part 1: by Dr David Maddison My articles on Computer Memory last year concentrated on ‘ephemeral’ storage such as RAM. I also mentioned more permanent storage systems like punch cards, magnetic drums and core memory. These two articles take a more complete look at permanent and semipermanent data storage of today, the future and the past. T hose two earlier articles on Computer Memory were in the January & February 2023 issues (siliconchip. au/Series/393). There is some overlap between this series and that one because computers didn’t always distinguish between temporary and permanent storage, especially in the early days. Partly, that was because RAM was so expensive per kilobyte, and it was necessary to use slower but cheaper storage to ‘swap out’ the contents of RAM. That allowed the computer to work with more data without needing a lot of expensive memory chips. We refer to the more permanent storage systems as ‘secondary storage’; this is long-term data storage, which retains its state when the system power is off. It is typically used to store an operating system, programs and (of course) for data storage. In contrast, ‘primary storage’ is volatile memory the computer uses during operation as programs run. An example of secondary storage is a hard disk or solid-state drive (SSD). By definition, it is a permanent part of the computer. A hierarchy of computer storage is shown in Fig.2. Offline storage is much like secondary storage but is removable, transportable media such as a USB flash drive or an optical disc like a DVD. Once connected to the computer, it behaves much like secondary storage. It is typically used for transporting data 14 Silicon Chip between computers without a physical connection, or for backing up data, including off-site backups. With some offline storage, the recording medium is kept in longterm physical storage, for backups of important information such as bank records. Being completely offline means it cannot be accessed or damaged without authorisation. Such storage might be for historical and archival records, such as old government census data. Tertiary storage is where the data is accessible to a computer, but not permanently connected to it. An example is a large tape library requiring a robotic arm to retrieve and insert a tape into an appropriate reading mechanism. This is also called nearline storage; it is almost online, but retrieving the data storage medium takes time. Cloud storage might be considered a form of secondary storage that a third party manages. It is located remotely from the user and may span multiple servers. Its main advantage is that it is more convenient, as it can be accessed from various locations. Disadvantages include an unknown risk of unathorised access (it depends on many factors such as the company managing it), an unknown risk of data loss (it has happened...) and the fact that the cloud storage company could go out of business and cut you off from your data. Therefore, cloud data still needs to be backed up like any other data. There are and have been many different secondary storage technologies; this article focuses on the more Fig.1: a blue IBM-style 80-column card encoded with almost the full Extended Binary Coded Decimal Interchange Code (EBCDIC) character set, shown at the top. Source: https://w.wiki/8R5y Australia's electronics magazine siliconchip.com.au significant ones, as well as some of the more unusual and interesting systems. We won’t go into as much detail on systems that were already covered in the aforementioned Computer Memory article. Technologies covered The entries below are arranged chronologically, based on the earliest use of the technology. There will be some overlap between the later versions of one technology and the earlier versions of its replacement. Entries marked with an asterisk (“*”) were covered in some detail in the Computer Memory article. In this first article, we have details on: • Punched cards* • Paper tape* • Drum memory* • Core memory* • Rope memory* • Magnetic tape* • Magnetic cards • Floppy disks • Bubble memory* • Optical discs • Magneto-optical discs The follow-up next month will concentrate on: • Hard disks • Flash memory* • Solid-state drives (SSDs) Plus the following possible future technologies: • 5D optical storage • Holographic storage • DNA storage Fig.2: ways that memory and storage can attach to a computer. Fig.3: a Canon Canola 167P calculator/computer (1971) with punched card program storage. Similar machines were used in NSW high schools in the early 1970s. Image courtesy of John Wolff, www.johnwolff.id.au Punch(ed) cards Punched cards are pieces of cardboard with holes in them representing the data – see Fig.1. The most recent and common form of punched card was the IBM 80-column card at 7⅜ × 3¼ inches (187 × 83mm). They were introduced in 1928 for tabulating machines. Not all modern punch cards were in IBM format, though; the Canon Canola 167P (Fig.3) would be familiar to many readers who were NSW high school students in the early 1970s. Fig.4: durable Mylar replaced paper in punched tape for industrial use, such as machine control. This tape was among the last to be produced in 1979. Source: https://w. wiki/8R62 (CC BY-SA 3.0). Punched paper tape Punched paper tape is similar to punched cards, except it is continuous; see Fig.4. This format has been obsolete since the early 1980s. Drum memory Drum memory was invented by siliconchip.com.au Australia's electronics magazine February 2024  15 Austrian Gustav Tauschek in 1932. Data was recorded on a drum coated in magnetic material. It was invented much earlier than the modern computer because it was used to record and tabulate data from punched card machines. The original device could store 62.5kB of data. Drum memory was used as RAM on some early computers but also as secondary storage in the 1950s and 1960s. It was the first type of secondary storage for computers – see Fig.5. The ERA 1101, renamed UNIVAC 1101, was built by Engineering Research Associates in 1950 and was one of the first stored program computers (ie, it was not programmed by rerouting wires). Programs were stored Fig.5: an early drum drive, circa 1951, at the Computer History Museum, Mountain View, California. The scratches on the drum surface are damage due to misaligned heads. Source: https://w.wiki/8R63 (CC BY 2.0). on a drum system of about 48kB. The drum was 22cm in diameter, spun at 3500 RPM and had 200 read-write heads. One of the last drum memory devices created was the IBM 2301, introduced in 1968 for the System/360 mainframe. It cost US$80,000 and had a storage capacity of about 4MB. It had an access time of 8.6ms, a transfer rate of 1.2MB/s and was used for memory paging (supplementing main memory to create a virtual memory extension). The drum was about 60cm high and 30cm in diameter, and the entire cabinet was about 2m tall and had a 1 × 2m footprint. Drum memory was not manufactured after the 1970s, although as late as 1980, PDP 11/45 computers that used drum memory and ran Unix were still in use. US Minuteman ICBM missile “Launch Control Centers” used drum memory until the mid-1990s. Perhaps the ultimate development of magnetic drum storage was the Univac FASTRAND, a giant 2.4m-long machine weighing about 2276kg. FASTRAND II stored the equivalent of 99MB (8-bit bytes). The FASTRAND III (Fig.6) had a higher data density, holding about 50% more data. Both the II and the III models had two counter-rotating drums, as the model I with a single drum had serious gyroscopic precession problems; only a few were made. Drum memory was the forerunner of hard disk drives and was eventually replaced by them. You can watch a video titled “1963 Sperry Rand UNIVAC FASTRAND Magnetic Drum, Computer History Archives, Unisys Educational” at https://youtu.be/luPM6XaKZuU The video mentions that such drives were used in OTC’s automatic message relay system in Paddington, Sydney, which was decommissioned in 1988. For further information on that, see siliconchip.au/link/abrn Another video about a 1960s-era minicomputer with drum storage titled “Meet my new Litton Minicomputer (it has Drum Memory)!” is at https://youtu.be/2yRcyQUIA5g Magnetic core memory Fig.6: a FASTRAND III drum drive from 1969 at https://gwdg.de/ – Source: https://www.radiomuseum.org/museum/d/rechnermuseum-der-gwdggoettingen/.html 16 Silicon Chip Australia's electronics magazine This memory was commonly used from around 1955 to 1975 as the main memory in computers, but it was also a form of non-volatile memory as it would retain its data when the power siliconchip.com.au Fig.7: an IBM core memory from the 1950s or 1960s. Source: https:// collections. museums victoria. com.au/ items/394677 (CC BY). was off. It comprised a grid of toroidal cores, which could be individually magnetised to store bits of information (see Fig.7). The YouTube video “Building the Core64 Interactive Magnetic Core Memory Kit” at https://youtu. be/7K6Qu-mNDms might interest our readers. Also see www.core64.io/ Besides covering core memory in the Computer Memory article last year, we also had a dedicated article on it in the March 2014 issue (siliconchip. au/Article/6937). Core rope memory Rope memory is a fascinating type of ROM (read-only memory) using magnetic cores with multiple sense, set/ reset and inhibit wires going through (or bypassing) them. This type of memory was used in the Apollo Guidance Computer. It had a much higher density than erasable magnetic core memory, which could only store one bit per core. With rope memory, up to 192 bits could be stored per core. The precise way it worked is very complicated. The best way to understand it is to watch these videos: • “Apollo Core Rope Memory (Apollo Guidance Computer Part 30)”: https://youtu.be/hckwxq8rnr0 • “Core Rope Memory Built and Explained - F-J’s Physics - Video 169”: https://youtu.be/WBHdNpAC7X4 • “DRUM MACHINE USING NASA TECHNOLOGY - Rope Core Memory Sequencer”: https://youtu.be/ zytjONYkU94 (also see Fig.8). Magnetic tape Magnetic tape was a common method of data storage on earlier computers, and it is still used today for backups and archival storage. Earlier tapes used ‘open reels’, but modern tapes are contained with cartridges. Today, magnetic tape is generally cheaper per gigabyte than other storage media but also slower, so it is used where speed is not so important. Magnetic tape was first used on the UNIVAC I computer on half-inch (12.7mm) metal tapes. There were eight tracks of data. Six tracks contained 128 characters per inch; one was for parity (error checking), and one was a clock signal. Those tapes were heavy and cumbersome. IBM computers from the 1950s used half-inch (12.7mm) wide plastic tape siliconchip.com.au Fig.8: the top of this device has an eightcore core rope memory, made with large cores as it is a demonstration unit. Source: https:// youtu.be/ zytjONYkU94 Storage capacity units The following are standard SI units for storage capacity. These measurements are often applied to the capacity of storage and networking capacity. ● 1 kilobyte = 1000 bytes (103) ● 1 megabyte = 1,000,000 bytes (106). ● 1 gigabyte = 1,000,000,000 bytes (109) ● 1 terabyte = 1,000,000,000,000 bytes (1012) ● 1 petabyte = 1,000,000,000,000,000 bytes (1015) ● 1 exabyte = 1,000,000,000,000,000,000 bytes (1018) A byte usually contains 8 bits. Similar terms can be used to refer to storage by number of bits (kilobit, megabit, gigabit etc). When referring to RAM, the same terms are sometimes used to refer to numbers based on the powers of two. For example, a kilobyte can sometimes refer to 1024 bytes (210), a megabyte to 1,048,576 bytes (220) etc. To reduce confusion, per the IEC, they are now called kibibyte (KiB, 210 bytes), mebibyte (MiB, 220 bytes), gibibyte (GiB, 230 bytes), tebibyte (240 bytes) etc. The names may seem strange, but the motivation is that “bi” are the first two letters of the word “binary”. Unfortunately, you sometimes see the use of mixed bases, eg, one “megabyte” may refer to 1000 × 1024 or 1,024,000 bytes. Thankfully, that is relatively uncommon. Australia's electronics magazine February 2024  17 ◀ Fig.9: the IBM 729 tape drive was popular in the 1960s. This bank of 729s is at the Computer History Museum in Mountain View, California. Source: Ken Shirriff, https://ibm-1401.info/729-Info.html ◀ Fig.10: the last nine-track, half-inch tape drive produced, the Qualstar 3400. It could be attached to a PC. Source: www.bitsavers.org/pdf/ qualstar/Qualstar _3400_Brochure.pdf coated with ferric oxide, much like audio tape. Lengths of 1200ft (365m) and 2400ft (730m) became standard. A tape reel size of 10.5 inches (267mm) was used, although smaller reels and shorter lengths were available. Earlier IBM tapes, introduced in 1952, used seven tracks (six data bits and one parity across the tape), while later ones, introduced in 1964, had nine tracks (eight data bits and one parity). Seven-track tapes had a recording density of 100, 200, 248, 556 or 900 characters per inch, while nine-track tapes stored 800, 1600 then 6350 characters per inch. Thus, the shortest tapes at the lowest recording density had a capacity of about 1.44MB. The longest tapes at the highest recording density had a capacity of around 182.88MB (but due to block size considerations, more like 170MB). During the late 1950s to the 1960s, the IBM 729 Magnetic Tape Unit (seven tracks) was a common tape unit in various versions – see Fig.9. The last half-inch nine-track tape drives were the Qualstar 3400 series from the USA in 2003; see Fig.10. Such drives interfaced with PCs and were presumably used to transfer data from old tapes. The nine-track format dominated offline tape storage until the early 1990s. Another type of tape was DECtape (Fig.12), introduced in the 1960s and used with many Digital Equipment Corporation computers such as the PDP-8 and PDP-11. These tapes were ¾-inch wide (19mm) and 260ft (79m) long. Each tape could store 184,000 12-bit PDP-8 words. DECtape had six data tracks, two mark tracks, two clock tracks and 18 Silicon Chip a data density of about 350 bits per inch. The tape system was considered highly reliable and durable. DEC­ tape is derived from LINC tape (1961), which was a public domain technology as the US taxpayer had funded its development. DECtape II was introduced in 1978, with very narrow (3.8mm) tape in a cartridge, giving a 256kB capacity. At the time, DECtape was considered a major advance for storing a computer’s operating system over the alternative of paper tapes, which could not support time sharing. The drum and disk drives of the time were expensive, unreliable and of limited capacity. Many early home computer systems used audio cassettes (Compact Cassettes) to store data. Some Compact Cassette tapes had a special formulation for digital data, and the tape length was usually shorter than audio tapes. Computers that used (or could use) cassettes included various Commodore computers (VIC-20, C64, C128 etc), ZX Spectrum, Sony MSX, Amstrad CPC 464, BBC Micro and various Ohio Scientific computers, among others. “Pocket computers” like the Sharp PC-1211 (TRS-80 Pocket Computer PC-1) and PC-1500 (TRS-80 Pocket Computer PC-2) also used cassette tapes. The Commodore computers used the Datasette (Fig.13), which was considered reliable but slow. It used a digital recording scheme on standard tape and transferred data at around 50 bytes per second. Various vendors developed ‘fast loader’ software to load data from cassettes much faster than the default methods used by computer manufacturers. Formats for cassette data storage Australia's electronics magazine included Frequency Shift Keying (FSK), first developed by RCA for their prototype home computer of the early 1970s. It was called FRED or Flexible Recreational Educational Device and had a built-in cassette drive. The Hobbyist Interchange Tape System (HITS) was introduced in 1975 by Jerry Ogdin for general hobbyist use. It used Pulse Width Modulation (PWM). The original article on HITS can be downloaded from siliconchip. au/link/abrt The Kansas City Standard (KCS) was introduced in 1975 by S-100 bus computer manufacturers and used FSK. KCS and its variations were used for numerous computers, including the Acorn Electron, BBC Micro, Dick Smith Super-80, Exidy Sorcerer, Microbee, MITS Altair 8800, Ohio Scientific, Sega SC-3000, Sony MSX and various Casio calculators. Particularly interesting variations of KCS included the encoding of software on a flexible vinyl 33⅓RPM record distributed in the May 1977 issue of Interface Age. KCS was also used to Fig.12: DECtape and DECtape II (lower right). Source: https://w. wiki/8R6W (CC BY-SA 3.0). siliconchip.com.au distribute software over the air in 1979 or 1980 via the Dutch broadcaster Nederlandse Omroep Stinging. The Apple I and ][, Atari computers and the TI-99/4 had their own versions of cassette interfaces. A ZX81 computer could load from tape at 300 baud (bits per second), while the ZX Spectrum could load at 1500 baud without speed loader software. The 1982 Dick Smith Wizzard computer used cassette tape, as demonstrated in the video titled “The Dick Smith Wizzard - Part 2 - Cassette Storage Module” at https://youtu.be/ bXKFag4x6EU The D/CAS (Data/CASsette) or streamer cassette was a professional form of Compact Cassette for digital recording. It used media optimised for data, and there was a notch in the case to identify this special format. Storage capacities started at 200kB; 600MB was possible by 1990 (see siliconchip.au/link/abru). It wasn’t only personal computers that used Compact Cassette for storage. The Burroughs B1700 mainframe of the 1970s could be booted from Compact Cassette tape! The DC100 (Data Cartridge 100) by HP and 3M was released in mid-1976. It was originally used in the HP9820 calculator and a range of other HP calculators, terminals and computers, such as the HP85. It had a formatted storage capacity of 560kB on 140ft (43m) of tape. The format was available for other companies, but the take-up rate was poor. It was a scaled-down version of 3M’s DEC300 cartridge, which had 300ft (91m) of tape and 2.5MB capacity. A variation of the DC100 cartridge, the DC150, was used for DECtape II, mentioned above. The ZX Microdrive (Fig.11) was introduced by Sinclair Research for use with the ZX Spectrum home computer in 1983. It was an endless loop tape drive containing 5m of 1.9mm-wide magnetic tape. It could store around 85kB, taking into account bad sectors. Video tape was also used for backups. The Danmere Backup was introduced in 1996 and could store between 750MB to 4GB on a video cassette, depending on the settings and model. There was also the Magurex Video Backup System for the Commodore Amiga and the Russian ArVid (2GB of data on an E180 tape). These systems were in use from about 1992 to about 1998 but had limited popularity. See the video titled “LGR Oddware - Danmere Backer VHS Hard Drive Backup System” at https:// youtu.be/TUS0Zv2APjU I recall Dick Smith Electronics selling one of these systems, possibly the Danmere. QIC tape (Quarter Inch Cartridge) Fig.13: the Commodore Datasette. It could store about 100kB per 30-minute side on standard audio cassette tape, but with special speed loading software, that could be extended to 1MB per 30-minute side. Source: https://w.wiki/8R6X Fig.14: the internals of a Sony LTO-3 cartridge. Note the RFID chip in the lowerleft corner. Source: https://w.wiki/8R6Z (CC BY-SA 4.0). Fig.11: a ZX Microdrive (opened) in comparison to Compact Cassette tape. Both hold about the same amount of data (about 100kB nominal), but the cassette takes 20 minutes to load fully, and the Microdrive 10 seconds. Source: https://w.wiki/8R6Y siliconchip.com.au Australia's electronics magazine was introduced by 3M in 1972. The tape is ¼-inch (6.35mm) wide, and the cartridges are very robust, with a heavy aluminium baseplate. The original tape cartridge was the DC300, which held 200kB on 300ft (91m) of tape and formed the basis of the DC100 tape and the DECtape II formats. Other formats were QIC-11 (20MB), QIC-24 (45MB or 60MB), QIC-120 (125MB), QIC-150 (150MB), QIC525 (525MB) and QIC-1350 (1.35GB), among others. Travan was another derivative of the QIC format intended for PC backup use, with 8mm-wide tape. Tape types included QIC-80 (80MB-500MB), TR-1 (400MB), TR-1EX (500MB), QIC3010 (340MB), TR-2 (800MB), QIC3020 (670MB), TR-3 (1.6GB), TR-3EX (2.2GB), QIC-3080 (1.2-1.6GB), TR-4 (4GB), QIC-3095 (4GB) and TR-5 (10GB). Linear Tape-Open (LTO) or Ultrium (Fig.14) is a successful and popular attempt to make a universal, open standard for tape for backups, archives and data transfer. It is under the control of Hewlett Packard Enterprise, IBM and Quantum via the LTO Consortium (www.lto.org). The original version, LTO-1, was released in 2000 and had a native capacity of 100GB. The current (2021) version is LTO-9, with a native capacity of 18TB per cartridge (advertised by its compressed capacity of 45TB). Future versions of LTO are planned with native capacities as follows: LTO-10 (36TB), LTO-11 (72TB), LTO12 (144TB), LTO-13 (288TB) and LTO14 (576TB). LTO tape is 12.65mm wide (‘½in’). The length was 609m for LTO-1, increasing to 1035m for LTO-9. Each tape has a passive RFID non-contact February 2024  19 memory chip inside that stores various identification information about the tape and user data. There is also a bar code specification for LTO tapes, for use in a tape library or for general identification. LTO is designed with a certain amount of compatibility with older versions. Generations 1 to 7 can read tapes from two generations prior and can write to tapes of the previous generation. LTO-8 can also read and write LTO-7 tapes, while LTO-9 can also read and write LTO-8 tapes. Otherwise, older tapes need to be migrated to newer versions. As with all other media formats, given that the earlier LTO tapes can be up to 24 years old, it is essential to migrate old data to newer versions as older media may degrade. Manufacturers specify that LTO tapes will retain their data for between 15 and 30 years. Tape libraries are a convenient way to store large collections of tapes. They may stored on shelves for manual retrieval or, more likely today, in automated systems with robotic media retrieval – see Figs.15 & 16. Card Random Access Memory (CRAM) CRAM was a product of NCR Corporation and became available for their NCR Century series computers in 1962 (see Fig.17). It comprised cartridges with either 256 or later, 512 plastic cards with a magnetic recording surface, each 3in x 14in (76 × 356mm). Each card had a unique notch pattern at one end by which it was suspended by rods. By rotating the suspending rods, an individual card could be selected. It was released from the cartridge and then read, after which it was returned. The capacity was either 5.5MB or 11MB per cartridge. CRAM was quite successful, according to the document at siliconchip.au/link/abro: “NCR was the first company to incorporate bulk storage as an integral element of online inquiries. Bulk storage provided accessibility to a larger capacity than could be cost-justified on secondary storage devices such as disk drives. The cost/bit was reduced by using removable media, transport mechanisms, and read/write stations.” So, it was cheap enough to enable the storage of online data for purposes such as bank balance enquiries. Such a machine is in the Museums Victoria Collections (siliconchip.au/link/ abrp). The original CRAM product brochure can be seen at siliconchip. au/link/abrq Other magnetic cards The HP-65, introduced in 1974, was the first calculator to use a magnetic card for storage. The card would store 250 bytes per side – see Fig.18. Another calculator that used magnetic cards was the Texas Instruments TI-59, which was introduced in 1977. Shown in Fig.19, it was also the first calculator series to use removable ROM modules with pre-written applications containing up to 500 steps. The card would hold 240 bytes per side for a total of 480 bytes, and the calculator itself had a memory of 960 bytes. There was a ROM module for a US Marine Corp version of the related TI-58C for Harrier ‘jump jet’ takeoff and landing calculations; siliconchip. au/link/abrr Fig.15: an LTO tape library with a robotic arm to store and retrieve tapes automatically. Source: Fujifilm (www.techradar. com/news/heres-the-cheapest-way-to-store-a-huge-1000tb-ofdata-online). 20 Silicon Chip Fig.17: an NCR CRAM unit. Source: NCR product brochure (https://archive.computerhistory. org/resources/text/NCR/NCR. CRAM.1960.102646240.pdf, p27). In 1969, IBM introduced the Magnetic Card Selectric Typewriter, an early word processor that could record, store and play back keystrokes. It used magnetic cards for storage (see Fig.20). They were like a combination of a punched card and a floppy disk. Each card could store about 5000 characters, compared to a punched card with just 80. There is a video of it titled “1969 IBM Mag Card Selectric Typewriter MC/ST Electronic Word Processing Magnetic Storage automation” at https://youtu.be/bW_jJjUarp0 Floppy disks A floppy disk is a flexible disc with a magnetic coating within a protective sleeve (usually square). The name ‘floppy’ was used because those sleeves were originally flexible, although rigid housings were used Fig.16: the IBM TS4500 Tape Library at KEK, Japan’s “High Energy Accelerator Research Organization”. Its capacity is 100 petabytes (100PB). Source: https://w. wiki/8R6a (CC BY-SA 4.0). Australia's electronics magazine siliconchip.com.au Fig.18: an HP-65 calculator with a magnetic card that passes through the machine as the program is loaded or stored. Source: https://w. wiki/8R68 (CC BY 2.0). Fig.19: a TI-59 calculator with magnetic card storage. Source: https://w. wiki/8R69 (CC BY-SA 4.0). starting with the 3.5in version. They were a common storage medium from the 1970s to the 1990s. Development of the floppy disk was started by IBM in 1967, and the first 8in (20cm) floppy was introduced in 1971 as the IBM 23FD, called the Minnow, with ~80kB (81,664 bytes) of storage, equivalent to over 1000 punch cards. The drive was read-only and was used to load microcode onto System 370 mainframe computers. The first 8in floppy drive with read/ write capability was the Memorex 650, which had a capacity of 175kB and was introduced in 1972. In 1973, IBM introduced the 8in Diskette 1 as part of its 3740 data entry system (Fig.22), which popularised the floppy disk. It had a capacity of 242,944 bytes formatted. There is an interesting related IBM document, “IBM 3740 Data Entry System System Summary and Installation Manual Physical Planning”, available from siliconchip.au/link/abrs The 8in floppy disk was developed to a peak capacity of around 1.2MB in 1977. A 5.25in (13⅓cm) disk and drive was introduced in 1976, the Shugart SA-400 Minifloppy, with a nominal capacity of 110kB (formatted capacity 87.5kB). This product became extremely popular. By 1978, Tandon introduced a 360kB double-sided, double-density format and, in 1979, the TM-100 drive (Fig.21). It appears that it wasn’t immediately used by any of the popular PC manufacturers. The original Apple ][ of 1978 used SA-400 drive mechanisms and had a capacity of 113kB. Atari released a similar 90kB drive in Fig.20: the IBM Selectric MC-82 with a magnetic card reader. Source: https://w.wiki/8R6A (CC BY-SA 3.0). 1979, while Commodore had a 170kB drive, also in 1979. The original IBM PC from 1981 had an optional floppy disk drive with 160kB per side. Support for 180kB per side (360kB total) was not offered until 1983. The TRS-80 Model III (1980) used Tandon TM-100 drives with a total capacity of 360kB. 5.25in floppies reached a maximum capacity of 1.2MB by 1982. In 1982, the Microfloppy Industry Committee (MIC) released the 3.5in (8.9cm) disk specification. A single-­ sided disk was released in 1983 with a formatted capacity of 360kB, or 400kB on the Apple Macintosh, followed by a double-sided disk of 720kB or 800kB on the Mac, and 880kB on the Amiga. In 1986, a 3.5in floppy was released with a formatted capacity of 1.44MB or 1.76MB on the Amiga. A 2.88MB “Extra High Density” (ED) 3.5in floppy disk was introduced in 1987. The Video Floppy (VF) disk was a 2in (50mm) floppy disk for recording analog video, usually as a series Fig.21 (left): a Tandon TM100-2A 5.25in floppy disk drive, as used on the original IBM PC, with an initial capacity of 320kB (increased to 360kB with DOS 2.0). Source: https://w. wiki/8R6D Fig.22 (right): the IBM 3740 Data Entry System popularised the floppy disk. On top of it are four 8in floppy disks, a Diskette 1 box and an oddly shaped CRT monitor. Source: https://w.wiki/8R6C (CC BY-SA 2.0). siliconchip.com.au Australia's electronics magazine February 2024  21 Floppy disk hacks Some early 5.25in floppy disks were sold as single-sided, and the “writable” side was indicated by a notch on one side. However, the media was actually writable on both sides. Some people used a paper hole puncher or special punch to make a notch on the other side so they could turn the disk upside-down and write data on both sides. This trick worked only with single-sided drives, such as for the Apple ][ or Commodore 64. Similarly, the capacity of single-density 720kB 3.5in floppy disks could be increased to 1.44MB by using a special punch to tell the drive it was a double-density disk. 3.5in, 5.25in & 8in floppy disks. Source: Eric Chan – www.flickr.com/photos/186773210<at>N06/52405767023 of separate independent still images. It was introduced in 1981 by Sony for the original Mavica “still video” cameras, which stored images in analog rather than digital format. It was also later used by Canon, Minolta and Panasonic. The disk had multiple medical and industrial imaging applications throughout the 1980s and 1990s. A data variant called the LT-1 was also produced that could store 793kB of data. Iomega introduced the Bernoulli Box floppy disk in 1982. The original disks were rather large at 21 × 27.5cm. Capacities of 5MB, 10MB or 20MB were initially available. It was discontinued in 1987. Bernoulli Box II was released later in a smaller 5.25in form factor with capacities of 20MB, 35MB, 44MB, 65MB, 90MB (late 1980s), 105MB, 150MB, and in 1993, 230MB. At the time of its introduction, standard floppy disks had a capacity of 1.2MB and hard drives around 30MB. Disk-ruining head crashes were still a problem with floppy and hard disks at the time. However, the Bernoulli principle enabled the head to be drawn toward the fast-spinning disk without touching it, so theoretically, it was impossible for the head to hit the media. Several ‘bump tests’ by reviewers confirmed this. Floptical disks were high-capacity floppy-like disks introduced in 1991 that used optical tracking with magnetic read/write. They were intended to replace conventional floppy disks. Their formatted capacity was 20.3MB in the same 3.5in form factor as a standard floppy disk. They contained an optical track for accurate read/write head tracking, but the data was still written and read magnetically. The drive could also read standard 720kB and 1.44MB standard 3.5in floppy disks. The Iomega Zip drive was introduced in 1994 (Fig.23), initially with a capacity of 100MB, then 250MB and 750MB. It became the most popular of the high-capacity floppy products but was eventually displaced by cheaper CD-R and CD-RW drives and media, then later, USB flash drives. ZIP disks were a different form factor to 3.5in floppies and incompatible with them. By 2003, the sales of ZIP disks and drives had declined dramatically. The Iomega Jaz was sold by Iomega from 1995 to 2002, initially with a 1GB capacity, increased to 2GB in 1998. However, like the PocketZip, they never became very popular. The Imation LS-120 SuperDisk had a capacity of 120MB, doubled with the subsequent LS-240. They were sold from 1997 to 2003 and were conceptually similar to the Flopticals mentioned above. They were intended as a replacement for the 3.5in 1.44MB floppy disk and had the same form factor. The SuperDisk drives could also read and write regular 3.5in floppy disks and could format such a disk to 32MB, although any alteration to the data required the whole disk to be rewritten. The SuperDisk had limited success, partly because Iomega’s ZIP disk had been on the market for several years at the time of SuperDisk’s release. Also, Fig.24: the Japanese Fujitsu FM-8 computer from 1981 had optional bubble memory storage, originally 32kB but later 128kB. It was the first PC with such an option. Source: https://w. wiki/8R6F (CC BYSA 4.0). Fig.23: an Iomega ZIP drive and 100MB disk. This is the external model; internal versions were also made. Source: https://w.wiki/8R6E (CC BY 2.0). 22 Silicon Chip Australia's electronics magazine siliconchip.com.au USB flash drives were becoming available and popular, and the cost of CD burners and media was falling. Caleb Technology released the UHD144 in 1998. It could read and write conventional 3.5in floppies and its own 144MB disks. Compared to other high-capacity disks, the disks were inexpensive, but the product did not survive competition from the Iomega ZIP, the Imation LS-120 and the CD-ROM. The company went bankrupt in 2002. The Iomega PocketZip or Clik! was introduced in 1999 as a small 40MB disk but never became popular and, like other floppy disk technologies, was replaced by flash memory devices. The Sony HiFD was released in 1998, and like some others, could read and write conventional 3.5in floppies. It had a capacity of 200MB. Unfortunately, the product suffered many problems, such as head crashes. It was re-released in 1999, but its reputation meant it was doomed to failure. Bubble memory We mentioned this type of memory in Part 2 of our article about Computer Memory. Briefly, individual bits of data are kept in the form of magnetic domains or ‘bubbles’ in a thin film of a substance such as gadolinium gallium garnet. The bubbles remain even when power is removed. It was introduced commercially in 1977 (see Fig.24) but became obsolete in the 1990s. It was once seen as a rugged alternative to hard drives, with a similar storage density to early drives, but that was quickly surpassed. Optical discs The idea of the modern optical disc came from David Paul Gregg in 1958. He was awarded US Patent 3,350,503 on it in 1967. The patent mentions the ability to record digital data. This invention and several related ones led to the development of the LaserDisc for analog data, the CD (Compact Disc), MiniDisc, DVD, Blu-ray and many derivatives. Optical discs store data in the form of pits and lands in the substrate. They are read by a laser, as shown in Fig.25. For writable media, the pits are also made by a laser. For mass production, the data is written all at once with a stamping machine rather than a laser. LaserDiscs were launched in 1978, storing video and audio data as analog signals (later versions included digital audio). Despite being analog, fundamentally, the information was still stored on the disc as a series of pits and lands like later fully digital CDs and DVDs. LaserDiscs were not generally used as a data storage medium, although in 1984, Sony produced a little-known digital LaserDisc format that could store 3.28GB of data per disc. The extent to which it was commercially used is not clear. There is a reference to it in the video titled “The Computer Chronicles - Japanese PCs (1984)” at https://youtu.be/rbh1XP4kCT4?t=954s LaserDiscs were officially discontinued in 2009, but had failed to be popular long before that, unlike the physically smaller DVD format, which was wildly successful. The Compact Disc (CD) was invented by Sony and Philips and released in 1982 as the Digital Audio Compact Disc for sound recordings. The CD-ROM (ROM = read-only memory) was announced in 1984 for data storage, but a suitable file format specification was not released until 1986. That was the “High Sierra” format, developed by Microsoft, Philips, Sony, Apple and DEC. Standard CD-ROMs have a capacity of 650-700MB, depending on how close to the edge the data is written. If some of the ‘rules’ are ignored (eg, lower data integrity), capacities of up to 900MB per disc are possible. One of the first products on CD-ROM was the Grolier Academic Encyclopedia. These discs were widely used for distributing software and in game consoles in the 1990s and early 2000s. They were also used for data backups of hard disks and for making copies of audio CDs. Regular CDs were 12cm in diameter, although mini 8cm CDs came along later, with a significantly reduced capacity. Eventually, people realised they didn’t have to be round, and all sorts of oddly shaped mini CDs were made for promotional purposes. However, ‘slot loading’ type compact disc drives only supported the full-size 12cm CDs, limiting the usefulness of the smaller versions. Besides audio discs and CD-ROMs, CDs were produced in many other versions. The CD-R became available in 1990 and could be written once and read many times (WORM), according to a specification released in 1988. The CD-RW was introduced in 1997 and could be written to, read and erased many times. Fig.25: a comparison of how data is stored on CDs, DVDs, HD DVDs and Blu-ray discs. Legend: track pitch (p), pit width (w), minimum length (l), laser spot size (⌀) and laser wavelength (λ). siliconchip.com.au Australia's electronics magazine February 2024  23 Fig.26: an IBM 3363, an early WORM drive with a formatted capacity of 200MB. Source: www. ardent-tool.com/docs/pdf/brochures/ ibm-3363-opticaldrive&cartridge.pdf CD-MO used magneto-optical technology, similar to the MiniDisc, but was never released commercially. Another CD format was Kodak’s (initially proprietary) Photo CD, introduced in 1991 and designed to contain 100 high-quality photos for display on the CRT TVs of the day. However, the format failed to gain widespread market acceptance and was discontinued around 2004. Picture CD was another Kodak product that followed Photo CD. DVDs (Digital Versatile Discs) were released in Japan in 1996 and other countries from 1997-1999. They can store any digital data, but video was initially the primary use. A standard non-rewritable DVD-ROM with one side and one layer can store 4.7GB of data (DVD-5); a single-sided, duallayer disc 8.5GB, and with two sides and dual layers, 17GB (DVD-18). As with CDs, commercial prerecorded discs are stamped rather than “burned”. Prerecorded movie discs are typically in either DVD-5 (single-side, single-­ layer) or DVD-9 (single-side, dual-layer) format. Single-side, dual-layer discs use Reverse Spiral Dual Layer (RSDL), a technique where the data is first written from the inside of the disc outwards. The laser wavelength is then changed to penetrate the first layer, and read the second layer. The second layer of data is written from the outside of the disc inwards. This allows a seamless change of layers for movies or other continuous data streams. As for writable DVDs, there are two write-once versions (DVD+R, DVD-R) and two rewritable versions (DVD+RW, DVD-RW). The less common DVD-RAM was designed to act like a removable hard disk. The difference between the “+” and 24 Silicon Chip “-” formats is that DVD-R was developed by Pioneer in 1997 and approved by the DVD Forum (www.dvdforum. org), while DVD+R was developed by Sony and Philips in 2002. There are technical differences in the method of recording and reading data. Both have compatibility problems with some drives, although the “+” versions are slightly better. Another type of DVD is HD DVD (High-Density DVD), with around triple the capacity of a regular DVD (15GB instead of 4.7GB per side and layer), up to 60GB for dual side, dual layer. This format was on the market from 2006 to 2008 but was supplanted by Blu-ray. Regular DVDs are the same size as standard CDs at 12cm in diameter, but there were also 8cm diameter mini DVDs with reduced capacity. Blu-ray was introduced in 2006 and is the same diameter as CDs and DVDs at 12cm. It has a capacity of 25GB (single layer), 50/66GB (dual layer), 100GB (triple layer) or 128GB (quad layer) for the BDXL write-once variant (specification released 2010). Blu-ray is mainly used for video and games. Standard Blu-ray discs only support a video resolution of up to 2K (1080p), so Ultra HD Blu-ray was introduced in 2016 to support 4K (3840 × 2160 pixels). BDXL and HD Blu-ray discs are incompatible with standard Blu-ray players and with each other for reading and writing. Optical Disc Archive (https://pro. sony/en_AU/products/optical-disc) is a proprietary Sony product introduced in 2012 and marketed as an alternative to Linear Tape Open (described earlier) with greater durability and a longer life – see Fig.27. It uses a cartridge containing 11 optical discs with three layers on each side for a capacity of 5.5TB in the largest cartridge. Australia's electronics magazine Fig.27: a 5.5TB Optical Disc Archive cartridge. The discs themselves are similar to, but not the same as, Blu-ray discs; they are Archival Discs (AD), which were jointly developed by Sony and Panasonic and designed to last at least 50 years. There were other optical disc formats that did not become popular, such as GD-ROM (Gigabyte Disc Read-Only Memory), a special format developed by Yamaha and used in Sega game consoles from around 1999 to 2006. Its purpose was to make copying the discs more difficult, but it also offered increased capacity compared to standard CDs of about 1GB. UDO (Ultra Density Optical) discs are a WORM technology intended for archival use with an expected life of 50 years, introduced by Sony and Plasmon in 2003. UDO 2 discs were released in 2007 with a capacity of 60GB. The discs are still available, although the format is not widely supported. M-DISC is a technology for DVD, Blu-ray and Blu-ray BDXL designed for extreme longevity, claimed to be up to 1000 years. They are readable by standard DVD players from 2005 Fig.28: a Sony MDW80 MiniDisc. Source: https://w.wiki/8Uen siliconchip.com.au and by standard Blu-ray and Blu-ray BDXL players. They are writable by most drives made since 2011. Other optical and magneto-optical systems An early example of an optical WORM drive for PCs that preceded the widespread adoption of CDs was the IBM 3363 (Fig.26). It was introduced in 1987 and intended for use with the IBM Personal System/2. It used a polycarbonate optical disc in a 5.25in cartridge and had a formatted capacity of 200MB. The MiniDisc (MD) was introduced by Sony in 1992 (see Fig.28) and discontinued in 2013. It was an erasable 65mm magneto-­optical disk in a caddy, meant for audio recording and intended to replace cassette tape. MiniDiscs could record 60, 74 or 80 minutes of audio using unique digital compression developed by Sony. To write data, a laser would heat a spot on the disk, altering its magnetic characteristics and allowing it to be magnetised, after which a magnetic head would write to it. To read the data, a laser sensed the altered polarisation of light due to the magnetic field of the spot. MD Data was a magneto-optical medium introduced in 1994. It used the same technology as the audio MiniDisc, although the caddy was slightly different to prevent insertion in a MiniDisc player. The disks stored 140MB, more than the 100MB of Iomega’s Zip drive, which was released at about the same time. However, MD Data was regarded as slow and discs were expensive. They were primarily used in Sony’s digital cameras, some other Sony products and a Sharp camera. The last product to use it was introduced in 1997. In 1999, MD Data2 (also called MDView) was released. This could hold 650MB of data but was only used in one Sony camera and some audio products. MiniDisc’s successor was Hi-MD, released in 2004, intended for data storage. It could store 1GB but was discontinued in 2011. Next month The second and final article in this series next month will continue where this one left off, covering the more modern storage technologies mentioned in the introduction. SC siliconchip.com.au The first terabit storage system – on photographic film! The IBM 1360 was the first computer storage system to store one terabit of data (125GB). It evolved from a mid-1950s CIA requirement to store vast numbers of printed documents. A system called “Walnut” was produced and delivered to the CIA in 1961 that could store 99 million photos of documents. 200 small boxes each contained 50 pieces of photographic film, each holding 99 photos in a 3×33 array for a total of 990,000 photos. Each set of 200 boxes was kept in a “document store”, and there could be up to 100 of those. Individual pieces of film were retrieved by an automated process. This system was developed into “Cypress”, using a superior film type, and IBM tried to commercialise it as the 1350 Photo Image Retrieval System. The same basic system was developed into the 1360 Photo-Digital Storage System (see Fig.29). It stored digital data on 35 × 70mm photographic film in a black and clear pattern, as shown in Fig.30. Each piece (or “chip”) had 32 data frames in a 4 × 8, holding a total of 6.6Mbits. 32 chips were held in a box called a cell. Data was written to unexposed film using an electron gun; it was then automatically developed. If data had to be updated, a chip was removed and replaced by a new one. There was extensive data redundancy, so there were 4.7Mbits of usable space per 6.6Mbit chip. There were 75 “trays” holding 30 cells each for a total of 2250 cells per “cell file unit” for half a terabit of data. Systems with more than one cell file unit achieved one terabit of storage or greater. The system at Lawrence Livermore National Laboratory kept one terabit. Only five 1360 machines were delivered in 1967 and 1968; the last system was shut down in 1980. No 1350 machines were delivered. The original IBM manual is available from siliconchip.au/link/abrv and there are videos titled “The First Terabit Server -The 1967 IBM 1360” (https://youtu.be/ twso8Nj7fLI) and “IBM 1360 Photostore Cell” (https://youtu.be/4-Jvd7lOjWA). Fig.29: the IBM 1360 PhotoDigital Storage System, circa 1965. It was the first secondary storage system to store one terabit of data. Source: https://w.wiki/8R6G Fig.30: a piece of photographic film from an IBM 1360 showing the data storage pattern, with a sewing needle for scale. Source: IBM press kit (https://w.wiki/8R6V). Australia's electronics magazine February 2024  25