The Floppy Disk
Mylar Memories
Alan Field Shugart was born on the 27th of September in 1930 in Los Angeles. In his early adult life, he wasn’t all too certain on the path his life should take so he changed his major in college four times in four years eventually graduating from the University of Redlands with a degree in engineering physics in 1951. By this point, he was twenty years old (soon to be twenty one), he was married, he had a child, and he was broke. This pressure led him to sign on with IBM as a customer engineer. He wasn’t exactly excited to go to work for IBM or even to go to work as a customer engineer (this was viewed as the lowest engineering position at IBM), but his start date would be the day after graduation and Shugart needed money. This early position saw him frequently fixing punch card accounting machines.
Shugart transferred to the IBM Research and Development laboratory in San Jose at 99 Notre Dame Street in 1955 to work on magnetic storage. About fifty people worked at this location at the time, and around fifteen people were working on magnetic storage. Shugart showed up to work one day shortly after starting in his new position to find Don Johnson pouring iron oxide paint onto a spinning twenty four inch disk of aluminum from a Dixie cup. This was with zero protections, no clean room, and generally very little advanced equipment to speak of. This first experiment developed into a project to build a random access storage system. The hope at the time was to put fifty disks on a vertical shaft. This would provide something around five million characters of random access storage (about 3.75MB), and the entire drive system would weigh around one ton. This original design provided an average access time of six hundred milliseconds and a maximum access time of one and a half seconds with twelve hundred RPM. A character for this device was stored as six bits with a parity bit, and the thing had neither automatic error correction nor address marks. It was effectively a raw disk with records and tracks, and all of the logic done with vacuum tubes with each needed circuit having been required to be bespoke. As the development of the technology progressed, the total number of surfaces was expanded to one hundred (the fifty disks were dual sided), but the total number of read/write heads was still just two. This was largely due to the air supply that provided the air bearing via forced air that ensured consistent gapping between the disk surface and the head. The air was routed through tiny holes in the head carrier with a single air supply for both heads. Given air compressor technology of the time, it was impractical to have more than two heads. This then meant that when accessing data on a separate disk the two heads would swing horizontally off the disk, move vertically to which ever disk was needed, and then move horizontally again to the track needed. The drive was released on the 14th of September in 1956 as the RAMAC 350 Disk File or IBM 350 disk drive system for the IBM 305 RAMAC (Random Access Method of Accounting and Control). The disk drive rented for about $750 per month (around $8402 in 2024), and the air compressor rented for an additional $150 (about $1680 in 2024). The patent for this was awarded to Bill Goddard and assigned to IBM.
A major advancement came to this field in 1958 with the increased understanding of gas-lubricated bearings. The arms and heads were designed to be extremely aerodynamic and essentially fly just above the disk surface at a distance of about two hundred millionths of an inch (12.7 microns). Without the requirement of an air supply, the drive could have a read/write head for every disk surface, and the total cost of ownership of a drive could be significantly reduced. The majority of the experimental work on this second phase of disk technology within IBM was done by Russ Brunner, Jack Harker, Ken Haughton, and Al Osterlund. The break through disk drive to first utilize this technology was the IBM 1301. This was roughly the same size as the 305 but it could store ten times the data with access times under two hundred milliseconds. Density for the 1301 was five hundred thirty five bits per inch with fifty tracks per inch, meaning twenty six thousand five hundred bits per square inch. The 1301 also brought modularity to the design with two disk modules of twenty disks each that were individually accessed.
Developing the 1301, Shugart recalls:
But the first disk drive with a head on each surface didn’t come easy. We stumbled several times along the way, including making the recording disks out of thin sheet steel instead of aluminum, oxidizing the surface to obtain magnetic properties, and then sandwiching the two thin platters around aluminum honeycomb to get the structural rigidity and flatness. And they worked fairly well, but there were thousands of flaws. We could never get the surface good enough for practical use. We built some probe heads and tried to do vertical recording on there steel disks; and we discovered a glue with temperature characters perfect for matching the magnetic element to the element carrying shoe. The probe heads didn’t work and the perfect glue turned out to be water soluble at one hundred fifty degrees or so.
We had lots of interesting projects in those early days of magnetic recording. Two of them come to mind, both of those projects run by Ray Herrera. One of the projects was called “project rusty nail,” because one day I asked Ray what was the cheapest way of making a magnetic head and he said wrapping a piece of wire around a rusty nail, and I said lets try it and we did. I don’t recall the outcome of that project. We had another project that we called project beer can and the concept was to get a very, very inexpensive small magnetic drum. We noticed in the beer industry they were just starting to extrude aluminum beer cans and the surface looked really, really good, so we ordered several cases of beer cans and coated some and tried to record on them. I don’t recall the conclusion of the project either. Biggest problem was getting IBM purchasing to order the beer cans.
After IBM published material concerning gas bearings in three articles titled A Gas Film Lubrication Study in the July 1959 issue of the IBM Journal of Research and Development, they were actually beaten to market by Bryant Computer Products who shipped a flying-head slider technology drive to the U.S. Army Signal Corps in November of 1960 for the MOBIDIC project (MOBIle DIgital Computer). The MOBIDIC was a transistorized computer packed into the trailer of tractor-trailer truck. The Bryant drive was also the first commercial use I’ve been able to find for zoned recording. The Bryant design used twenty six disks of thirty nine inches each with a rotation speed of twelve hundred RPM with the disks arranged on a horizontal spindle. Meanwhile, the IBM 1301 was first made available on the 2nd of June in 1961 at a price of $2100 per month (or around $21400 in 2024). For the rare company or institution that chose to buy a 1301, the list price was $115500 (nearly $1.2 million in 2024). It used twenty five disks per module allowing up to fifty disks in a single 1301 unit. Each disk was twenty four inches in diameter, and they operated at eighteen hundred RPM. The Bryant model had a slightly higher maximum access time, and slightly lower bandwidth, but offered total capacity in a system of upto 205MB. It was, however, the IBM that was the market success, and the IBM that set the market standard.
At this point, the two major advancements in disk technology were the gas lubricated bearing and modular storage. These were combined with a disk size reduction to just fourteen inches in the IBM 1311, and the modules were made removable with IBM 1316 disk packs. Each disk pack was four inches tall, weighed ten pounds, and offered 2MB of storage.
Storage density increases became normal and expected with each subsequent model of drive. The 2311, for example, offered 7.25MB of storage per disk pack. The ability to easily remove these disk packs meant that offline storage was limited only by the number of disk packs on-hand. This was a big change in the industry as this modularity meant that data could be easily moved between machines, easily warehoused, and easily archived for the first time. The 2311 and 1311 disk packs were forward compatible, and this created a defacto standard in the storage industry. Shortly after the disk packs hit the market, plug-compatible storage companies began to arise like Memorex, Century, Dysan, and Marshall.
The fourth step in the evolution of magnetic storage was the reduction in size and weight of the read/write head. The low mass head was also extremely low cost and it didn’t have to be removed or unloaded from the drive. Instead, the head could land at spin down or take off at spinup. Further, this method of action allowed for a reduction in the complexity of the arm actuator carrying the head. Importantly, this was all a single, sealed (mostly) unit that required very little maintenance. The original research and development work for this was done by Armin Miller. Miller developed modified frequency modulation (MFM) for magnetic data recording which had the practical effect of doubling the recording density of regular FM encoding. The MFM work was done while he was working at Ampex Corporation. Of course, Miller wasn’t satisfied with changing the world once, so for an encore he developed the first low mass, three-point-suspension read/write head for disc drives at his company DataDisc (founded in 1962). His design was allowing DataDisc to achieve thirty two hundred bits per inch at a time when three hundred bits per inch was normal. Another practical upshot of Miller’s head design was that the heads could be stopped and started arbitrarily, even when in the middle of a platter, without a head crash. IBM purchased a license for Miller’s technology, and the resulting IBM product was the IBM 3340 Direct Access Storage Facility. The 3340 was announced in May of 1973 and was shipping by November. The initial intended design for this product was to have two drives of 30MB each in a single box. The 30/30 design brought about the nickname of Winchester and this nickname was later applied to all drives that operated similarly.
The IBM System/360 was designed and implemented with microcode in use by both the CPU and the peripheral controllers. The storage for this microcode was implemented via transformer read-only storage (TROS) holding one thousand nine hundred twenty bytes. The wiring for the transformers was printed on sheets of Mylar (sort-of like a flexible circuit board). These were referred to as tapes. Each tape had two loops of wiring referred to as word lines. The word lines went either through or around each transformer. Each tape had sixty transformers and two hundred fifty six drive lines, yielding two hundred fifty six words of sixty bits. The Mylar tapes were then stacked on transformers with a bar placed across the top of each transformer to close the transformer core with a wire wrapped around the that top bar to act as a sense line.
By around 1967, Shugart was the Direct Access Storage product manager, and TROS storage wasn’t going to be used for microcode on the System/370. In its place was semiconductor memory. This introduced simplicity in design and manufacturing, but it also brought about a new problem; semiconductor memory was volatile and microcode would need to be loaded into the machine some how at power on. In 1967, Dave Noble was assigned to work on the Initial Control Program Load (ICPL) Program. He worked alone for a few months to determine what the ICPL would need to do, and how it would need to do it. At first, Noble thought of using tape, and while magnetic tape would work for this, some kind of random access medium was desired to ease file access and data organization. This new random access storage technology needed to be economical and provide the ability to load either microcode or diagnostics as desired while also being removable. This would have another benefit where microcode would be easily updatable with new media.
Mylar was in use at IBM, iron oxide was in use at IBM, low mass drive heads were in use at IBM, platters on a spindle were in use at IBM, and the marriage of the four brought us the floppy disk in IBM’s project Minnow. There was a problem though. The first eight inch floppy disks were just bare Mylar sheets coated in iron oxide. This being the case, they’d easily get dirty or damaged or both. Herb Thompson and Ralph Flores thought up a clever, slim, durable jacket that doubled as a cleaning mechanism for the surface of each disk. According to Shugart, the floppy disk would never have been successful without the jacket, and while history remembers Shugart as the man who created the floppy, Shugart himself gives the credit to Dave Noble. The patent for the floppy disk was issued on the 6th of June in 1972 with the inventors listed as Ralph Flores and Herbert Thompson. The drive patent was issued on the 18th of July in 1972 with the inventors listed as Warren Dalziel, Jay Nilson, and Donald Wartner.
The first drives available to customers were shipped in 1970 as the IBM 23FD for loading S/370 microcode. These were single-sided, read-only, somewhat slow, and low capacity holding just 81.6K.
Computers are always hungry for more storage, and this created both competition in the storage industry and not a small amount of drama. IBM’s storage competitors Byrant and Telex were having some problems by the late 1960s, and IBM would experience some of their own. In December of 1967, twelve members of the San Jose R&D office left IBM to start Information Storage Systems which focused on making IBM compatible peripheral equipment. This was quite the shock for IBM which traditionally had enjoyed good employee loyalty, and those twelve individuals became known as the dirty dozen within IBM. Their company, ISS, began selling disk packs, disk drives, tape drives, and printers exclusively through Telex, and all were fully compatible with IBM computers.
In the Summer of 1969, after having gotten the Minnow project underway, Shugart left IBM for Memorex where he became the Vice President of Product Development for the disk drive division, and around two hundred other IBM personnel followed shortly thereafter. In 1972, Shugart’s team at Memorex shipped the Memorex 650 Flexible Disc File. This was the first read/write floppy disk drive. The disks for this drive had a capacity of 187.5K.
In 1973, Shugart left Memorex to start his own company, Shugart Associates. The initial plan for this company was to develop small business computer systems including floppy disk drives and printers. Shugart burned through his initial funding quite quickly, and after two years with no products completed, his financiers wish to focus on floppy disk drives. Shugart disagreed. Whether he was fired or quit is unclear, but Shugart Associates went on to create the five and a quarter inch floppy disk drive while Alan Shugart went on to cofound Seagate with Finnis Conner in 1979. The five and a quarter inch disk and drive followed after An Wang of Wang word processors suggested to Jim Adkisson and Don Massaro at a meeting in Wang’s conference room in late 1975 that the eight inch format was too expensive and too bulky for his products. He really needed a $100 drive (or about $566 in 2024) rather than a $200 drive (around $1132 in 2024). Steve Jobs needed the same thing for Apple, but Wang was the largest customer of Shugart Associates at the time. Adkisson then began work on the product which was released in September of 1976. The first model was the SA-400 Minifloppy. The price of media and their capacities were listed in the December 1976 issue of BYTE with SA-400 Minifloppy costing $4.50 (about $24 in 2024) and having a size of 110K (unformatted).
The article discussing then state-of-the-art storage technologies focused heavily on Altair and IMSAI. This is, of course, sensible as these were the leading microcomputers of the time and among the first microcomputers ever sold commercially. The earliest microcomputers would have made use of paper tape and audio cassettes for data storage and loading but floppies proliferated quickly. IMSAI even offered a dual eight inch floppy drive for the IMSAI 8080.
Of course, the Shugart eight inch drive, also listed in BYTE, had a very serious impact on computing history. It was an SA-800 that Gary Kildall used to develop CP/M. IMSAI then licensed CP/M in 1975 and shipped it on eight inch floppy disks and a little later on five and a quarter inch floppy disks. CP/M’s use of floppies as well as the shipping of VisiCalc on floppies made the floppy disk a defacto standard. It became so popular that even the more affordable 8 bit machines like the Atari 400 and 800 as well as the Commodore VIC-20 and 64 had floppy drives available for them.
Shugart Associates was acquired by Xerox in December of 1977 for a stock trade worth $41 million (about $201 million in 2024) as floppy disk sales hit $96 million (about $482.7 million in 2024) and drive sales hit $289.6 million (about $1.45 billion of which Shugart represented $226 million in 2024 dollars).
The five and a quarter inch disk was the dominant format by 1979. Both the Apple II in 1977 and the IBM PC in 1981 would ship with five and a quarter inch drives available but without an eight inch option (though the Apple Disk II was actually released in June of 1978). Steve Jobs eventually got exactly what he’d wanted as the Apple Disk II system used Shugart drives. The Apple II and VisiCalc were such a big deal that Apple quickly became Sugart Associate’s largest customer with over $25 million per year (about $117 million in 2024) in floppy disk drives sold at $100 each (around $467 in 2024). The impact of the Shugart five and a quarter inch drive can be felt today due to its dimensions at three and a quarter inches tall, five and three quarter inches wide, and eight inches deep. Those are the dimensions of nearly all PC compatible drive bays. That the drive bays were this size set the dimensions of the first hard disk drives, and later ODD drives.
In 1981, Sony introduced the three and quarter inch floppy disk drive. The first drives were 218.8K unformatted or 161.2K formatted. The double-sided version launched in 1982. The first computer to use this format was the Sony SMC 70 in 1982 which was designed for professional video applications. It shipped with a drive compatible with single-sided, double-density disks. An important note here is that while the SMC 70 was the first computer to ship with a three and a half inch drive, it wasn’t the first product with one. That distinction goes to the Sony Series 35 word processor which was released one year earlier. The original Sony disks are not compatible with later disk drives. The Sony disks had a different number of tacks, different iron oxide coating thickness, and the shutter wasn’t sping loaded. Later disks, however, could be used in the SMC 70; it was just that the early disks could not be used in later drives.
In 1982, the Microfloppy Industry Committee was formed to establish a standard for the new floppy disk format. This was announced in June of 1982 with membership open to all interested companies. Shugart was among the first members but a total of twenty three different companies ended up joining. Again, however, it was Shugart Associates creating a bit of a standard as the SA-300 was plug-compatible with any computer that had used a five and a quarter inch drive. The pinout would remain the same for the lifetime of the floppy drive but the connector would change. The first single-sided drives shipped in 1983 followed by double-sided drives in 1984.
In 1984, the Apple Macintosh shipped with only a single-sided three and half inch disk drive of 400K. 1985 saw Atari adopt single-sided 360K disks for the ST, and Commodore adopt double-density 880K disks for the Amiga. In 1986, IBM shipped the 5140 with dual 720K double-sided double-density disk drives. That year also saw Apple move to double-sided 800K disks.
When IBM introduced the IBM PC AT in 1984, it brought with it 1.2MB five and a quarter inch disk drives to replace the 360K drives and disks of the IBM PC and IBM PC XT. Later models of the AT would include three and half inch disk drives with the thirty four pin IDC connector that became a default in PC compatibles shipping with three and half inch disk drives. The 1.44MB double-sided high-density format was first seen on the IBM PS/2 in 1987, and the Macintosh IIx in 1988. In 1990, double-sided extra-high-density disks and drives appeared first with the NeXTcube, NeXTstation, and IBM PS/2 model 57 with a storage capacity of 2.88MB. While this format did see mild success with inclusion in some high profile machines, the 1.44MB disks and drives were already plentiful and cheap.
Three and a half inch floppies remained in use through the 1990s despite the rise of CD-ROM, Iomega Zip, and Minidisc. The decline and fall of the three and a half inch disk was only truly assured with the introduction of USB flash drives which offered a far higher speed of operation, a smaller size, silent operation, and higher capacities. Ironically, the floppy’s zenith was in 1998 with two billion floppies being sold each year, which was the same year Apple introduced the iMac with no floppy disk drive. That the iMac was a wildly successful product was as good an indicator as one could get that the floppy drive and disk were dying. In 2007, PC World in Europe announced that it would be ending sales of floppy disks and drives once their existing stocks had sold. Sony ceased making floppy disks and drives in 2011. If you desire floppy disks or drives today, they are still available from floppydisk.com while supplies last.
Without Shugart you'd need a different name for this blog!