This page is just to help me remember that the quality and performance of computers is changing all the time. Many technologies I once regarded as state of the art, or even impossible, are now routine, or even obsolete. Science fiction author Dr Gregory Benford remarked, in a variation of Clarke's Law, that "any technology distinguishable from magic is insufficiently advanced". He may be right, but I'd like to at least remember the days when I thought I understood bits of it.
The abacus was possibly in use in Egypt by 450 B.C., and in a different form in China by 600 B.C. Went from stones in grooves on loose sand, through grooves in wood, to a frame and metal rods with beads. The Chinese style has two beads at the top, five at the bottom. The Japanese form has one bead at the top, four below, and forces a carry operation when a column exceeds 9.
The young Blaise Pascal invented an eight digit, geared adding machine in 1642.
Analog computers developed also in the seventeenth century, John Napier's bones provided logarithm calculations, while William Oughtred invented the slide rule. By the 19th Century, we find items like Lord Kelvin's tide prediction machine. In the 20th Century, Dr Vannevar Bush made an analog computer in 1925, and an advanced version was used during WWII to calculate artillery firing tables. Another analog computer was used for fire control in the B-29 bomber.
The eccentric genius Charles Babbage succeeded in 1822 in making a Difference Engine, for the solution of polynominal equations (x squared, + x, + a = 0). He worked until his death in 1841 on his more general computer, the Analytical Engine.
Dr Herman Hollerith's punch cards, developed for the 1890 US Census, led to the IBM punch card.
Alan Mathison Turing, eccentric and tactless British mathematician and logician born in 1912, suicided with cyanide at age 41 after being forced to accept treatment with hormones to "cure" his homosexuality. He is known for his war work decipering the German Enigma code machine, and for the idea of the Turing test. In the 1930's Turning wrote a paper "Computable Numbers", which outlined the design of a machine that could compute anything that could be computed. www.abelard.org/turpap2/tp2-ie.asp
Dr John Vincent Atanasoff and graduate student Clifford Berry of what is now Iowa State University in Ames, Iowa. Designed and built a machine for solving systems of simultaneous equations in 1939-1940. Had a seven electronic valve add and subtract unit. Punched card I/O, using high voltage to punch the cards for output. Memory drum was a large array of capacitors physically rotated to synchronise the computation process.
In 1944, Howard Aiken and IBM completed the Mark 1 Automatic Sequence Controlled Calculator. It contained 3000 relays, and could multiply two 23 digit numbers in about four seconds.
John W Mauchly and J Presper Eckert built ENIAC. Mauchly visited Atanasoff in Iowa and saw this work before doing ENIAC. However Mauchly also did notes on the concept in the 1930s. US Government contract W-670-ORD-4926 signed 5 June 1943, for University of Pennsylvania to develop Electronic Numerical Integrator and Computer (ENIAC). This 1800 square foot machine, with 18,000 valves and 11,000 switches, could do 5000 additions a second, and was completed in 1946.
The next big step was EDSAC (Electronic Delay Storage Automatic Calculator), the first system to incorporate a stored program. The idea of the stored program has been crucial to computer development.
In 1983, IBM had US$40 billion turnover and 369,000 employees, and the next nearest data processing companies were Digital Equipment (US$4,826 million), Burroughs, Control Data, NCR, Sperry, Hewlett Packard, Wang, Honeywell, and Xerox. Apple were number 11 with US$1084 million turnover and 4582 employees. Intel were #66, with US$1121 million turnover. Compaq (who bought Digital in 1998) were #91, with US$111 million turnover, and 614 employees. Personal Computer revenue in 1982 for IBM was US$500 million, for Apple US$664 million, for Tandy US$466 million, for Commodore US$367 million, and all had growth rates for the year exceeding 60%.
Stanford engineering graduates Bill Hewlett and Dave Packard started their partnership in a one car garage in Palo Alto in 1938. They made an audio oscillator to test audio gear, and sound equipment for Disney's Fantasia.
Intel's 4004, the first 4 bit microprocessor, in 1971. The 8 bit 8008 came a year later. The 8086 in 1978. The first IBM PC, using an 8088, in 1981. The 80286 in 1982. The first 32 bit Intel chip, the 80386 in 1985. The 486 in 1991, the Pentium in 1993.
Masatoshi Shima designed the Intel 8080 and later went to Zilog with Fedrico Faggin to design Z8000. He worked at calculator maker Busicom from 1969 to 1971, and collaborated with Intel's Ted Huff in the design of the 4004 chip set. His technical education as an undergraduate was in organic chemistry.
A magnetic recording patent was granted in 1898.
IBM Type 305, an experimental five megabyte magnetic disk memory. Multiple disks used, the unit was about the size of two washing machines. October 1955 development, released 13th September 1956 as the 305 RAMAC. Perhaps the first magnetic disk memory. It had 50 disks, each 24 inches in diameter. I think this was the first disk drive.
1962, the first removeable disk drive. Companies planned on making lots of money selling replaceable disk packs.
1973, the first Winchester drive, from IBM, ancestor to all current (1990's) hard drives. It has two 30 million character spindles, hence name 30-30, and thus the name Winchester, after the rifle. Other ideas about the origin of the name come from IBM having a facility in Winchester, England, another on Wincester Road in San Jose, and maybe even a scientist named Winchester. It used an air bearing for the read write heads.
Control Data and others developed the Storage Module Device (SMD) interface, which was computer independent. It let system designers make a simple interface to drives, and buy the drives from any vendor. Later drive interfaces like ATA, and SCSI, and IDE are all made to a standard, not proprietary, and are thus cheaper.
Floppy Disk Drives
The first floppy disk drive, from IBM, in 1970. It used 8 inch floppies. The 8" flexible plastic disk is smoothly coated with a magnetic oxide. There is a square jacket around this, which includes a jacket wiper to remove contaminants from the disk surface. A smoke or dust particle, or human fingerprint, is far thicker than the oxide surface and can destroy data. These disks were rotated at 360 RPM by a large mains powered synchronous motor and a pulley belt (countries like Australia with 50 Hz mains used a different diameter pulley).
The read write heads were moved by a stepper motor driving a worm gear. When track 0 was reached, a mechanical switch put a pulse on the Track 0 control line. A hole in the diskette was detected by a photocell, and this was used to indicate when the start of the first sector had been encountered. For a period, some drives expected disks to have 32 hard sectors and a sensor hole at the start of every sector.
The heads were moved from Track 0 to the desired track by the stepping motor. The head had to be raised from the track, and then lowered again (loaded) at the correct track. The unload operation took about 3 millisecond. Stepping took about 50 millisecond per track, and loading took about 15 millisecond before the head settled.
A popular early format for diskettes was the IBM 3470, with 77 tracks, each with 26 sectors of 128 bytes. The diskettes were single sided, and single density. Data was transferred at 250 kilobytes per second. The earliest format was frequency modulated (FM), with clock pulses 4 micro-seconds apart written to disk to form data cells. A magnetic transition within a cell was a one, no transistion was read as a zero. This made the electronics very simple, as you could recover data with a simple monostable, or a phase locked loop.
Double density recording using modified frequency modulation (MFM) mostly eliminated the clock pulses and used a self clocking format. A clock pulse was only inserted if a pair of zeros were occurred together.
The control lines on older disk drives followed this patten below. All odd numbered pins were ground lines. 8" drives used 50 pin connectors, 5.25 and 3.5 drives used 34 pin connectors.
5.25" 8" Function 4 18 Head load (-in-) 6 32 Drive select 4 (-in-) 8 20 Index sector pulse (-out-) 10 26 Drive select 1 (-in-) 12 28 Drive select 2 (-in-) 14 30 Drive select 3 (-in-) 16 Motor on (-in-) 18 34 Step direction (in or out) (-in-) 20 36 Step pulse (-in-) 22 38 Write data (-in-) 24 40 Write gate (enable write) (-in-) 26 42 Track 0 pulse (-out-) 28 44 Write protect (-out-) 30 46 Read data (-out-) 32 14 Side select (-in-) 34 22 Ready (-out-) (If not available index pulse may be used)
In 2000 in the USA, the number of internet users increased 52% to 68.7 million. DSL growth was 86% to 2.3 million. Cable grew 19% to 4.1 million. Combined high speed access was about 10%. Free advertising supported internet providers had 14.8 million active subscribers. AOL had 39% of the market. Internet television had 1.2 million subscribers. Report by AFP from Telecommunications Reports International. Majority are still on dial up services.
The Qwerty keyboard, produced by Christopher Sholes in 1873, was deliberately designed to slow down typists, because early typewriters tended to jam when used at any great speed. Not until the 1930's was the more efficient Dvorak Simplified Keyboard produced by Dr Dvorak of the University of Washington, Seattle. DSK improves typing speed by an average of 35%, cuts trainng time by 50%, and reduces fatigue. 70% of all common words can be typed by the middle row of keys, three times better than the Qwerty leyboard.
Modems (modulator and demodulator) change digital signals from equipment into analog sounds for transmission through the telephone system (known in various countries as POTS - plain old telephone system, or as PSTN - public switched telephone network), which was originally intended only for voice.
I seem to recall seeing book sized boxes offering all of 50 bps (bits per second) back in the 1960s or 70's.
Early modem standards like Bell 103 (300 bps) and 202 (1200 bps) or CCITT V.21 (300 bps) tended to be asynchronous (they start and stop each character they send with special start and stop bits, which means an 8 bit byte takes 10 bits, so 300 bps means 30 characters a second). Later modems like V.22 bis at 2400 bps were synchronous, sending special characters to correct their timing. I can't actually recall whether any early modem tried AM - amplitude modulation, or loudness. The early 300 bps modems used FSK - frequency shift keying, like an FM radio - and changed frequency to indicate a one or zero. This was soon replaced with PSK - phase shift keying or phase modulation, changing the position of the wave form with relation to the original carrier cycle. This wasn't sufficient at higher speeds, and a combination of phase and amplitude known as quadrature amplitude modulation was used.
The need for error correction introduced extra bits in patterns, or trellis code modulation (TCM). Modems rapidly stopped being a dumb bit of electronics and incorporated their own computer, and automatic repeat on error systems like MNP were included.
At speeds higher than 9600 bps, adaption to phone line conditions was needed. Multiple narrow carriers (512 in the original Telebit Trailblazer) each carrying part of the signal pushed speeds up, ending at 33 kbps.
Several companies, such as Micro Instrumentation and Telemetry Systems (MITS) and Heathkit sold kit calculators from 1971.
The 1 MHz 6502 based Apple was designed by Steve Wozniak, and was advertised in April 1977. It had an 8k BASIC, 4k of RAM and cost US$1298. It was also advertised as a bare board for only US$798.
This was a 68000 based system, cheaper than an Apple Macintosh, using Digital Resaerch's GEM operating system and GUI. I got one in October 1987 from Seahorse computers in Blaxland, complete with black and white monitor and a mouse. A few weeks later I added desoldered the space left for extra memory chips, and added sockets for more RAM (to expand it from 512 kB to 1 MB). Had to add bypass caps, locate a bad joint, and add an inline resistor in one line to reduce voltage overshoot. In November I used an external Adaptec 8451 controller and 20 MB Miniscribe 3425 hard drive. I also fudged a double sided floppy drive into the Atari, to replace the single sided drive. In April 1988 the monitor failed, and repairs took over a month.
B and S Minimap
Designed and built in Wentworthville, a suburb of Sydney, Australia around the late 1970's as a school system. It used a 2650 CPU, had an 8K BASIC in ROM and 16 K of RAM. It had video out, RS232 and cassette tape connectors. Peripherals included a punch card reader through which you pushed cards manually, a 32 character wide thermal printer, an 8 bit input box and an 8 relay output box. A user could install a faster cassette tape interface (3000 baud instead of the variable 2 to 1200), and also a light pen. A disk drive working off the RS232 was planned, and could be shared amongst 16 MiniMaps. A prom burner was also planned.
Dick Smith System 80
This was a TRS80 clone from (I think ECCA) Hong Kong that was announced around 1 Sept 1979. Dick Smith stores didn't know about it two weeks later, nor did Computerland.
In 1979 I investigated an Olivetti electronic proportional spacing daisy wheel typewriter as a printer, but it cost A$2345. An IBM Selectric, in the form of a 73 Input Output terminal about three years old would cost about A$1400 to restore, with a new cost about twice a Selectric, around A$1138.
IBM Personal Computer Buses
Video Electronics Standards Association (VESA) held a meeting at the Fall 1991 Comdex, and in December 1991 formed the VL Bus committee. Started with 15 members and rapidly grew to 120 companies. They wanted a low cost modular local bus that was scalable and extensible but compatible with ISA, EISA and Micro Channel. It was to be available without patent protection, and have enough bandwidth for video graphics. VL Bus 1.0 was approved on 28 August 1992. It was intended for the Intel 486, but allowed for 386 as well. Interface speed across the 32 bit VL Bus connector was a maximum of 40 MHz, but 66 MHz was possible on the motherboard. Zero wait state writes, and one wait state reads, with two wait states for speeds above 40 MHz. Early board designs would not always work from PC to PC, but mostly they did. Later upgrades allowed 64 bits.
Intel announced the Peripheral Component Interconnect (PCI) local bus specification, with the June 1992 revision providing processor independent 20 to 33 MHz synchronous interface to the CPU and cache memory via a PCI bridge. The multiplexed address and data meant lower pin counts and lower manufacturing costs, but much tighter control was needed of the electrical characteristics, with many companies using Intel's PCI chip set. Early PCI bus systems, especially for the 486, essentially just didn't work reliably, however PCI took over as 586 CPUs became more common. It was actually originally designed so that the graphics and disk controller were on the motherboard, and it was not until the Fall 1992 Comdex that it was announced there would be a standardised connector by December 1992.
Ohio Scientific Instruments
Began operations in Hiram, Ohio around 1975. In Byte #6 there was an advertisement for a 6502 based computer trainer for US$99. This could be traded in on an OSI Superboard kit, or for US$10, for a blank circuit board and a set of instructions for these computers designed by Mike and Charity Cheiky, founders and owners of OSI.
(Mike developed OSI to give Charity something to do. He designed all the early products by himself. Charity was president and ran the business end of things. Mike was the chief engineer. *AGT)
Mike and Charity demonstrated their 400 board at the Trenton NJ Computer Show of 1976, showing how it could be built with either a 6502 or a 6800 processor. This board had space for 8 memory chips, 2102, to provide 1024 bytes. OSI also had video boards and memory boards, and you could buy the bare boards for US$29. The 400 series boards were designed on the basis that cheap is better, and the 420 memory board took that to the nth degree. No sockets for the memory chips. The board was single sided. The bus was done with Molex pins.
One really nice feature of the OSI boards was they were large and uncluttered, easy to build, and didn't produce as much interference as the more compact designs others used. The top of the line model at this point cost US$675. What is more, you could buy 16K memory boards, an unheard of amount of memory in those days.
The OSI Challenger computer was one of the first personal computers to be supplied with a floppy disk drive, one of the first that could handle multiple users, and the very first to use a Winchester hard drive. This was at a time when computers like the Apple II and TRS-80 were still using cassette tapes.
(The Challenger was one of the first personal computers to be sold already assembled, rather than as a kit such as the MITS Altair or the SWTP 680. The OSI 74 MB Winchester was in another league entirely from the 5MB Seagate hard drive that was introduced later. The initial production run of the Challenger was ten units, which all went to dealers. I received #5. *AGT)
They had a wide range of 6502 based computers by 1977, ranging from the bare 500 board with 8K BASIC and 4k RAM at US$298, through the CII with a case for US$598, an 8 inch floppy model with 16k of RAM for around US$2000, the CIII with 32k of RAM, two 8 inch floppy drives and serial I/O for US$3481, to a US$6000 addition with a 74 MB hard drive.
OSI provided serial terminal systems, 24x24 character displays, up to 64x32 character displays, sound output as standard. They had voice input boards with a Votrax module, X10 AC control interface boards, joystick boards, ADC and DAC boards, prototyping cards, home security interfaces, a telephone interface, and an eprom burner. Probably the largest range of boards ever to come from one small manufacturer.
Unfortunately, the OSI software support was pathetic. Their attitude seemed to be that they produced computers, not manuals. Towards the end they did have three excellent hardware manuals, produced by SAMS as part of their "Photofacts" series. OSI were bought out by MA-COM, who got rid of Mike Chieky, dropped the hobby systems, and eventually went bankrupt.
(Software support and documentation support are two different things. One of Mike's major marketing errors was not allowing third parties to produce software for OSI computers. This was at a time when many people were producing software for CP/M-based programs that ran on Intel 8080 or Z80 processors. OSI could not compete with the large selection of software available under CP/M. OSI was bought by MA-COM as you say. MA-COM did not get rid of Mike Cheiky. He left voluntarily because he did not agree with the way MA-COM was running the company. He and Charity formed a new company named Dreisbach Electromotive. MA-COM sold the OSI division to a Swedish company. It survived for a while with a new name, then disappeared. *AGT)
(An additional little-known fact: OSI made a technology transfer transaction that sent its processor design to Yamaha. Yamaha used it as the basis for their computer musical instrument synthesis technology that later showed up in their electronic keyboards. *AGT)
*AGT Corrections by Allen G. Taylor, Former OSI Western Regional Manager, Former OSI Far Eastern Sales Manager. Author, SQL for Dummies, and Using Hard Disks with PCs. Many thanks.
Charles D Tandy acquired a failing electronics store in New England in 1963. It was called Radio Shack. Tandy knew nothing about hobby electronics, but was great at handling specialist hobbyists via small stores with motivated managers, via his experience with leather hobby stores. He opened an average of 470 stores and dealers a year for 15 years. By the time he died in 1978, Tandy had 7100 stores. Tandy cut stock from the 10,000 to 30,000 then typical to about 2400 items. Stock control weeded out all the slow selling items. He dropped mail order, and all the associated clerical systems.
Tandy announced the 4 MHz Z80 based TRS-80 Model I on 3rd August 1977. Designed by Steven Leininger, who also directed the design of the Model II, the Model III and the Tandy Color computer, the Model I included a 4k level 1 BASIC in ROM, and 4k of RAM, and cost US$399.95. Supplies were scarce. Byte reviewed it in April 1978, and the reviewer reported ordering it on August 4, and taking delivery on October 11 when another customer cancelled.
Steven Leininger joined Radio Shack in the fall of 1975. He was an electronics enthusiast since putting together an electronics kit in his South Bend, Indiana home when ten years old. He was initially into audio, electric guitars and amplifiers, but in his last semester in junior year at Purdue University, signed up for more digital courses. With batchelor's and master's degrees in elctronic engineering, he moved to Santa Clara in 1974. His girlfriend Susanne also moved, when the cost of phone calls got too great. Steven was working on National Semiconductor's SC/MP microprocessor, and moonlighting at a computer store. There he met some Texans from Radio Shack, including Jack Sellers (who became General Manager of Tandy Business Products), and the Leiningers moved to Texas where Steve became project engineer for Radio Shack.
The initial kit computer idea was scrapped (too many customers had problems building kits). They went with a standard typewriter keyboard as well as less intimidating, TV video output was cheaper than a printer, and cassette tapes would work for storage. In Februray 1976, Leninger and Sellers decided to use a Z80 and dynamic RAM, and incorporate the changes into the design by April. The hardware design went well, but the software person disappeared, so Leininger had to write that as well.
The last fixes to the BASIC were made in a two day session at the factory, ending at 2 a.m. A few days later, with a simple game running on the machine, Charles Tandy inspected it, and decided to make 3000 for initial sale. If they didn't work, they would be used "in store" for inventory control. Charles Tandy wanted them on sale in August 1976, and they made the deadline by hours.
Computer Museums and Collections
- Obsolete Computer Museum
- Lots of personal computers from the late 1970s and early 80s. www.obsoletecomputermuseum.org/