Ken Shirriff

@kenshirriff

Restored Apollo Guidance Computer, Xerox Alto. Reverse-engineering old chips. Analog computing. Added ₿ to Unicode. Charger teardowns. Arduino IRremote library.

Silicon Valley
Vrijeme pridruživanja: svibanj 2010.

Tweetovi

Blokirali ste korisnika/cu @kenshirriff

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @kenshirriff

  1. 1. velj

    Video showing the analog computer's digital counter, wired to the 100 Hz clock, counting down from 255. The digital section has some intermittent problems, maybe due to oxidized connectors. The oscillator quit working shortly after this video.

    Prikaži ovu nit
    Poništi
  2. 1. velj

    The analog computer we're restoring has a digital section with eight flip flops and 8-bit counter. We replaced bad capacitors in the power supply.Also replaced a burnt-out bulb with the last spare we had, but another burned out a few minutes later :-(

    The analog computer is the size of a mini-fridge. The cover is removed, showing the circuitry inside.
    The power supply board for the digital section. We replaced the three blue filter capacitors that went bad. The board has a large transistor on a heat sink.
    The digital section is programmed by a patch panel. 16 lights to the left show the status of the flip flops and counter.
    Prikaži ovu nit
    Poništi
  3. proslijedio/la je Tweet
    22. sij
    Prikaži ovu nit
    Poništi
  4. 21. sij

    I explain how the Soyuz space clock works in 's latest video. Includes closeups of the circuitry in this Soviet digital clock from 1984.

    Poništi
  5. 19. sij

    Another die shot of Am2901 bit-slice processor. In this one, I dissolved the metal layers so you can see the silicon transistors. Compare with original photo. Register file in center, 4-bit ALU in lower right. Empty regions used for buses. Bipolar transistors look like rectangles

    Die photo of Am2901 chip with metal layer removed. Die is bluish, with small transistors.
    Detail of die photo showing many transistors. They appear as a complex pattern of rectangles.
    Die photo of Am2901 chip. The bond wires are connected to pads around the chip.
    Detail of die photo with metal removed, showing transistors. They are overlapping rectangles.
    Poništi
  6. 18. sij

    Die photo of an AMD Am2901 bit-slice processor, used in minicomputers such as the PDP 11/44. The 4-bit chips were combined to process larger words. The chip used bipolar transistors, not CMOS. Two layers of metal on the die. Chip was introduced in 1975, this die dated 1983.

    Die photo of the Am2900 showing the complex circuitry of the chip. Bond wires are connected around the outside.
    The Am2900 integrated circuit with the top removed. The die is visible inside the integrated circuit.
    Closeup of the die, showing two layers of metal wiring.
    Closeup of the die showing text 4301X, 1983 AMD.
    Poništi
  7. 17. sij

    The Soyuz spacecraft used a digital clock on the control panel. Surprisingly complex, the clock had over 100 ICs and a switching power supply crammed inside. I open the clock and explain what all these chips do. Details:

    The Soyuz spacecraft clock is a gray box with two rows of LED digits on the front, as well as switches and buttons.
    The clock opened up. The circuit boards are fanned out, showing the circuitry inside the clock.
    The switching power supply board has transistors, inductors, ICs, and other components.
    One of the digital logic boards. It has rows of surface-mounted flat-pack integrated circuits.
    Poništi
  8. 15. sij

    Nice tiger chip art on Dallas Semiconductor wafer from 1987. Interesting NPN/PNP transistor test circuit as well as feature widths from 50 μm down to 1 μm. "I'd rather be modeling." The 9256 and 2256 part numbers don't match products; maybe this was a development wafer?

    Chip art on a semiconductor wafer. An orange tiger jumping out of a green building.
    Test circuitry on an integrated circuit wafer. Square structures for NPN transistors and circular structures for planar PNP transistors.
    Text on a silicon chip wafer: "I'd rather be modeling" (C)(M) Dallas Semiconductor 1987
    Test circuitry has lines varying from very wide 50 μm line to very narrow spiraled 1μm line.
    Poništi
  9. 10. sij

    Unusual Dallas Semiconductor wafer is entirely test circuitry, with 20 test blocks repeated over the wafer. Everything from transistors of different sizes to flash memory cells. Not shown: resistors of many types, anti-fuses, capacitors, isolation, and other chip structures.

    Microscope photo of silicon die showing complex test circuitry.
    A silicon wafer.
    Test circuitry on the die with transistors of various sizes. Labeled "30V PCH". Multiple pads for source, gate, and drain.
    Test circuitry labeled "Flash Memory". One test cell with "1.5 um rules", one with "1.2 um rules".
    Poništi
  10. 9. sij

    I'm a couple of weeks late with this Christmas tweet, but here's a punch card wreath. These were a thing in the 1960s when computers used punch cards.

    A wreath made from punch cards. The cards are bent to form points and they are spray-painted gold. Seasonal greenery is in the middle of the wreath.
    A closeup of the wreath, showing punch cards with holes.
    A punch card with the text "Made by -- Karen W Bowers and Suzanne Worthington"
    Poništi
  11. 9. sij

    Continuing repairs of the card reader/punch . The punch unit has many belts, gears, and cams, and they all need to be exactly timed to the angle of the driveshaft so cards are punched without jamming. A dial (photo 2) shows the angle in the punch cycle.

    Volunteers at the Computer History Museum repairing the IBM 1402 card reader/punch.
    Inside the card punch, many gears and belts.
    The card reader/punch opened up with the front panel removed, showing the gears and mechanisms inside.
    A cam unit with switches activated by rotating cams.
    Poništi
  12. 7. sij

    The Dallas Semiconductor 2435 is a three-pin chip to identify a battery and sense its temperature over 1-wire interface. The die is surprisingly complex and has some interesting chip art: a padlock labeled "Security Products", a DIP with thermometer, and a battery(?) with dog tag

    Die photo of the DS2435 chip.
    Chip art on the die: A padlock with "Security Products" and a DIP with a thermometer.
    Die: Dallas Semicon. (M) (C) 1994
    Chip art on the die: a battery(?) with dog tags and the text DS2435.
    Poništi
  13. 25. pro 2019.

    Merry punch-card Christmas! I punched this card on the IBM 1401 computer and the 1402 card reader/punch. Punch cards normally don't have a hole pattern like this, so I used the "column binary" feature. IBM charged $101 per month extra for column binary support.

    A punch card with the hole pattern spelling out MERRY XMAS, along with a Christmas tree.
    The IBM 1401 computer, with some of the internal circuitry exposed.
    A punch card with MERRY XMAS.
    The IBM 1402 card reader/punch. This large unit was used to punch the card.
    Poništi
  14. 22. pro 2019.

    Analog computer restoration update. Hooked up an integrator using the patch panel and got a smooth ramp. Two integrators solve second order differential equation; solution is sine and cosine waves. Two speeds: drawing takes seconds on slow speed, sub-millisecond on fast speed.

    The analog computer is a box the size of a mini-fridge. It has colorful patch panels on the front, as well as potentiometer knobs, a voltmeter, and buttons.
    Oscilloscope trace showing integration. It starts off at -10 volts and then has a smooth ramp up to 10 volts, where it flattens out.
    Closeup of the patch panel which has a grid of holes with colorful labels. A tangle of yellow wires provides connection.
    Oscilloscope display showing sine and cosine curves.
    Poništi
  15. 13. pro 2019.

    Restoration of 1969 analog computer update: powered it up and used an op amp to multiply by 10. Input: .7V on potentiometer, output 7V. It is programmed by plugging patch cables into the patch panel. Next step integration, but will need more patch cables.

    The analog computer is a mini-fridge sized box. It has a patch panel on the front for programming, as well as potentiometers and a meter.
    A closeup of the patch panel. Yellow wires are plugged into the panel to control it.
    The meter of the analog computer shows 0.7V, while a multimeter shows 7.00 volts.
    One of the modules from the analog computer. It has a bunch of circuit boards, as well as gold-plated fingers on the front to connect with the patch cords.
    Poništi
  16. 25. stu 2019.

    Restoring an analog computer from 1969; we fixed the precision power supply. Large transformers and 8 complex cards in the power supply, but no manual, so I had to reverse-engineer it. One board has high-precision 0.01% (not 1%) resistors.

    The analog computer is a box about the size of a mini-fridge. It has a colorful plug board on the front, along with dials and a meter.
    The power supply contains two large power transformers, along with large cylindrical electrolytic capacitors, all mounted in a metal frame.
    The power supply takes up the lower part of the computer. The transformers are in the box on the left, while the card cage on the right holds 8 circuit boards to regulate the voltage.
    This printed circuit board implements an op amp. The board is covered in components: resistors, capacitors, transistors, and a metal-can integrated circuit.
    Poništi
  17. 16. stu 2019.

    Mysterious "Micropositioner" artifact showed up . Opened it up and it's not a positioner, but a highly-sensitive polarized relay from the 1950s, activated by 40μA. Inside, someone left a folded-up note documenting the pinout. It has tube-style octal pins.

    Poništi
  18. 15. stu 2019.

    The IBM 360/20 mainframe stored microcode on a stack of Mylar sheets with transformers. Holes punched through the wiring in the sheet created paths through or around the transformers for 1's or 0's. My writeup on TROS (Transformer Read-Only Storage):

    A TROS module is rectangular, about 15" long. It has a stack of 128 Mylar sheets connected to green circuit boards.
    A closeup of a TROS Mylar sheet, showing the holes punched through the wiring. The diagram on the right shows how the hole pattern controls the current path through or around the transformers.
    The IBM System/360 Model 20 is a computer a bit bigger than a washing machine. The front has been removed showing the circuitry inside, and 8 TROS modules on the left. Photo from Wikipedia.
    A closeup of the TROS module, showing the connections between the Mylar sheets and the circuit boards.
    Poništi
  19. 12. stu 2019.

    Inside a sonic delay line, 50 feet of coiled nickel wire provides 5ms delay for sound pulses. This stored 11 kilobits of pixel data in IBM's 1965 video terminal. Transducers (center, lower left) generated and detected pulses. Thanks to Alan Park for photo.

    Inside a sonic delay line, 50 feet of coiled nickel wire holds data as sound pulses. A circuit board is on the left.
    Outside view of an IBM sonic delay line. It is a thin rectangular module with 8 pins sticking out.
    A woman typing on an IBM 2260 terminal.
    Poništi
  20. 7. stu 2019.

    A 1965 IBM terminal used sonic delay lines for memory. It stored pixels as sound pulses that traveled along a long nickel wire, but vibrations from heavy footsteps could corrupt the display. The full story of how it led to 80x24 and 80x25 displays:

    Man typing on an IBM 2260 terminal.
    The sonic delay line held about 50 feet of nickel wire. Bits were sound pulses along this wire. The delay line module is a flat module with wires sticking out.
    The IBM 2260 terminal has a CRT display and a keyboard.
    Woman typing on an IBM 2260 terminal.
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·