Ken Shirriff

Computer history. Reverse-engineering old chips. Restored Apollo Guidance Computer, Xerox Alto. Ex-Google, Sun, Msft. So-called boffin.

Blog: https://righto.com
Bluesky: https://bsky.app/profile/righto.com
Github: https://github.com/shirriff

Public Key

npub1mgq5f2207ye4p2q55ge5qytuc4qsj6zw7cu79k735yq04c5as9wqqysjel

NIP-05 Address

[email protected]

Profile Code

nprofile1qqsd5q2y498lzv6s4q22yv6qz97v2sgfdp80vw0zm0g6zq86u2wczhqpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0ds2lnuuk

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1mgq5f2207ye4p2q55ge5qytuc4qsj6zw7cu79k735yq04c5as9wqqysjel

Show more details

Last Notes

2026-06-28 16:30:48 UTC

- reply

Credits: IOP boards provided by Richard, 4 Pi board from @npub1xvg…62t2.
Photo of the IOP and CPU courtesy of RR Auction. IOP documentation from Mike Stewart. Exploded diagram from https://ibiblio.org/apollo/Shuttle/IBM-74-A31-016%20-%20Space%20Shuttle%20Advanced%20System,%204%20Pi%20-%20Prototype%20Input,%20Output%20Processor%20(IOP).pdf#page=145

2026-06-28 16:29:31 UTC

- reply

This diagram shows how the circuit boards (called pages) were plugged into the I/O Processor's aluminum-alloy case. For storage, the system used magnetic core memory, larger pages at the back. This system didn't use a microprocessor; it was built from 11 logic pages crammed with simple chips.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/828/677/897/496/599/original/bdb367657aa33065.png

2026-06-28 16:29:01 UTC

- reply

IBM had a series of aerospace computers called System/4 Pi. These computers all used standard-sized boards (top). Except that the I/O Processor's boards (bottom) were one inch wider. I guess they needed the extra space for more circuitry.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/828/673/973/376/821/original/985dec12c33bec0c.png

2026-06-28 16:25:31 UTC

The Space Shuttle had five general-purpose computers that controlled, monitored, and navigated the Shuttle. Each computer consisted of two boxes: the CPU (right) and the I/O Processor (IOP, left). The IOP connected the computer to 24 high-speed networks. Let's look at two boards from the IOP... 1/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/828/658/722/675/690/original/bdfb7e5580604b74.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/828/660/601/395/337/original/ad557b5e6d305fc9.png

2026-06-13 16:21:58 UTC

Intel introduced the 8087 floating-point chip in 1980 to make math faster. The chip does tangents, square roots, and so forth, but fundamentally everything is computed through additions and subtractions, performed by an adder at the heart of the chip. Let's look at the adder's circuitry... 1/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/743/713/202/353/709/original/a352daf872448c4b.jpg

2026-06-07 19:02:50 UTC

- reply

Yes, the thyratron is very similar to an SCR (also known as a thyristor: thyratron transistor).

2026-06-07 17:29:42 UTC

- reply

For more on the IBM 604 and this thyratron tube module, see my latest article: https://www.righto.com/2026/06/ibm-604-thyraton-tube-module.html

2026-06-07 17:29:23 UTC

- reply

Pluggable tube modules were an important innovation. They made it easy to repair tube-based systems, they packaged components densely, and they paved the way for IBM's tube computers in the 1950s. IBM featured the modules in this 1952 ad from Time magazine: "Fingers you can count on". 4/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/710/002/138/525/318/original/ac23113bcb5f6077.png

2026-06-07 17:28:12 UTC

- reply

The tube is not exactly a vacuum tube since it has a bit of xenon gas inside. It is called a thyratron: when it turns on, the xenon ionizes, allowing a large current to flow through the tube. In the IBM 604, thyratron tubes provided high-current pulses to the electromagnets to punch holes in cards.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/709/995/768/003/934/original/c8d064af13fc1d4e.png

2026-06-07 17:26:34 UTC

- reply

The tube module acts as a high-current switch: I hooked it up to control a light bulb. The first button turns the module on with a small signal. The second button cuts the power to turn the module off. 2/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/709/988/997/608/709/original/0e140e699149097a.mp4

2026-06-07 17:24:35 UTC

The IBM 604 Electronic Calculating Punch was introduced in 1948. Not quite a computer, it was a programmable calculator that became very popular due to its low price of $550/month, and over 5600 were produced. I have a vacuum tube module from this system, so I powered it up... 1/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/709/983/493/765/921/original/e4c470945bdfb3ad.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/709/984/028/780/884/original/381d7e497e49bec4.png

2026-05-30 17:48:19 UTC

- reply

I did this work with the Opcode Collective, especially Smartest Blob and Gloriouscow.
The full ROM image: https://8087.martypc.net/#dzi=tiles%2Ftiles.dzi
A description of the 8087 microcode: https://github.com/a-mcego/granite/blob/main/tools/8087mc/bin/8087.md

2026-05-30 17:47:35 UTC

- reply

We are reverse-engineering the meaning of the 8087's microcode. For more details, see my latest article:
https://www.righto.com/2026/05/microcode-inside-intel-8087-floating.html

2026-05-23 16:54:05 UTC

- reply

For more on the Spacelab computer, see my latest article: https://www.righto.com/2026/05/reverse-engineering-spacelab-computer.html

2026-05-23 16:49:16 UTC

The Space Shuttle could hold a flying laboratory called Spacelab in its cargo bay. Three French-made computers ran Spacelab. I opened up a Spacelab computer and found that instead of a microprocessor, it is built from a multitude of simple chips. Let's take a closer look at the computer...
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/624/912/700/890/522/original/214214663330bbfb.png

2026-03-29 22:22:12 UTC

- reply

Fair enough :-)

2026-03-29 17:30:27 UTC

- reply

I forgot to explain the 4 Pi name. IBM's System/360 mainframes got their name because they supported a full 360º circle of applications, from business to scientific. The 4 Pi aerospace computers, however, were designed for the 3-D world, a "sphere" of applications. While a circle has 360º, a sphere has 4π steradians of solid angle.

2026-03-29 17:11:56 UTC

- reply

There were many more models of System/4 Pi computers, including refrigerator-sized computers used on submarines; see my article for details: https://www.righto.com/2026/03/ibm-4-pi-computer-history.html

2026-03-29 17:10:28 UTC

- reply

In 1991, the Space Shuttle upgraded its computer to the AP-101S, replacing two 60-pound boxes with one box. Since there were five computers on the Shuttle, this change freed up 300 pounds for additional cargo. The CPU was implemented in three boards full of chips, one below. These boards were much larger than the earlier 4 Pi boards.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/313/560/937/224/462/original/52c1c1399755da8c.png

2026-03-29 17:05:00 UTC

- reply

The Space Shuttle used a more advanced System/4 Pi computer called the AP-101B. A separate computer, the I/O Processor, provided 24 data buses to all parts of the Shuttle. These boxes still used individual TTL chips and magnetic core memory.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/313/536/095/110/947/original/e1a3f1289b90810a.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/313/541/468/850/195/original/8e27dd8ace3af0aa.png

2026-03-29 16:59:44 UTC

- reply

IBM's 4 Pi computers were built from simple TTL integrated circuits and used magnetic core memory for storage. "Flat-pack" integrated circuits on multi-layer circuit boards provided dense logic. The board below was round to fit inside a tactical missile, but most of the computers used standardized rectangular boards called "pages".
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/313/507/522/967/045/original/1e4a3ca7fe6f2bf6.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/313/520/789/902/233/original/78fc68c03f8df9f3.png

2026-03-29 16:53:11 UTC

In 1967, IBM introduced the System/4 Pi line of aerospace computers, packing mainframe performance into a compact box. 4 Pi computers powered everything from military aircraft to the Space Shuttle to sonar systems on submarines. Thread...
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/313/469/577/392/917/original/dc58a800e6094d8b.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/313/477/112/469/007/original/bd51682e80a94354.png

2026-02-14 22:09:42 UTC

- reply

The 8087 doesn't help with 16-bit integer math.

2026-02-14 22:08:57 UTC

- reply

Early calculator chips are pretty weird. They operate in decimal and are serial, so they operate on one bit at a time with a one-bit ALU. So all the data is constantly in motion.

2026-02-14 17:45:31 UTC

- reply

Instruction decoding in the 8087 has many more complications: hardcoded instructions handled by the "Bus Interface Unit", instructions that load a constant from the constant ROM, and handling the instruction queue. For more details, see my latest blog post:
https://www.righto.com/2026/02/8087-instruction-decoding.html

2026-02-14 17:43:59 UTC

- reply

The microcode inside the 8087 chip uses conditional jumps for many purposes. Some conditions examine the instruction, so the microcode can act differently for different instructions. Here's the PLA that stores target addresses for microcode jumps. Target 0 is special: it does a three-way branch depending if the instruction is multiplication, division, or addition/subtraction.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/070/210/295/765/263/original/97320097bb05411c.png

2026-02-14 17:38:43 UTC

- reply

Like many processors, the 8087 uses a "Programmable Logic Array" (PLA) to decode the bit patterns in instructions. A PLA consists of grids of transistors, with some of the transistors missing. The pattern of transistors controls the PLA's function. The grid is a dense and flexible way of implementing this logic.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/070/191/247/489/545/original/f545e53425b42f7a.png

2026-02-14 17:32:53 UTC

In 1980, Intel released the 8087 floating-point chip, making math much faster. I'm reverse-engineering this chip, 46 years later. Most of its instructions are implemented in microcode, but some are implemented in hardware. Let's look at the circuitry that decodes instructions and decides what to do...
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/070/157/546/401/938/original/d0355c62ff44b38c.jpg

2026-02-05 05:10:34 UTC

- reply

Thanks, I meant the 1402 card reader/punch. I've edited it now.

2026-02-04 22:29:15 UTC

- reply

The card reader/punch (IBM 1403) can punch 250 cards per minute, so the heart cards zipped out at high speed. However, this machine can't print text on cards, so I needed to go back to the keypunch...
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/014/711/450/155/721/original/d90d0f36f1425b31.jpg

2026-02-04 22:26:04 UTC

- reply

To duplicate the heart cards, I toggled a short program into the 1401 computer to read a card and print out copies. The 1401 has a special feature called "column binary" that allows it to read and punch cards even if the hole pattern isn't valid.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/014/701/175/234/140/original/1a6d83ce385e38b6.jpg

2026-02-04 22:23:00 UTC

- reply

To make the heart, I had to figure out the right sequence of characters to form the hole pattern. Inconveniently, some holes don't form valid characters, so I had to "multi-punch" in some columns. This worked for one card, but the keypunch can't duplicate multi-punched cards.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/014/674/913/974/809/original/8b3dcdb557214c3c.jpg

2026-02-04 22:16:41 UTC

The Computer History Museum has a presentation tonight on the history of computer dating: "Algorithms of Love". I made a special punch card for the event. It wasn't easy to make this card so I'll explain how I did it...
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/116/014/664/909/132/010/original/2081b2230f6ba348.jpg

2026-01-23 17:51:13 UTC

- reply

If you want more on the 8086 processor's ALU, see my latest blog post:
https://www.righto.com/2026/01/notes-on-intel-8086-processors.html

2026-01-23 17:49:35 UTC

- reply

The control signals in the 8086 need to be amplified to drive the circuitry for 16 bits. The driver circuit is very unusual: it is a "bootstrap" circuit that uses stray capacitance to boost the output voltage higher than normal, similar to a charge pump. The transistor with +5 on the gate acts as a diode, preventing the extra voltage from escaping. On the die, you can see that the drive transistors are much larger than the regular transistors.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/945/661/335/648/174/original/86972c1629f867b7.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/945/661/833/356/559/original/f4ad96cd678a030a.png

2026-01-23 17:44:15 UTC

- reply

Each bit of the 8086's ALU is implemented with an unusual circuit. It uses two lookup tables (trapezoids), loaded from six control signals. This reprograms the ALU's logic to implement the desired operation. This is similar to how an FPGA uses lookup tables to implement different logic functions.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/945/643/865/364/083/original/bfe68d6627e856d7.png

2026-01-23 17:41:29 UTC

- reply

What's the XI mux? To make the 8086's microcode more compact, it often doesn't specify the ALU operation, using a special code "XI" instead. The hardware looks at the machine instruction, determines if it is an ADD, SUB, XOR, or so forth, and substitutes the correct ALU operation. Thus, one microcode routine can implement 8 different machine instructions.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/945/631/671/040/629/original/fa7522f7b00f42db.png

2026-01-23 17:35:57 UTC

- reply

The 8086's microcode specifies what ALU operation to perform along with which temporary register to use. The control circuitry holds this information in latches. A PLA (Programmable Logic Array) decodes the instruction into dozens of control signals. The Lookup Table (LUT) PLA generates six signals that configure the ALU to perform the desired operation.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/945/606/310/637/415/original/2c3ef21544a485a0.png

2026-01-23 17:31:57 UTC

The arithmetic/logic unit (ALU) in the Intel 8086 processor (1978) is more complicated than you might expect, performing 28 different operations from addition and logical AND to shifts and BCD adjustment. A special control circuit reconfigures the ALU for each operation. Let's look closer...
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/945/586/323/699/700/original/102fa0e7b5e6779e.jpg

2025-12-30 19:51:11 UTC

If you're interested in obscure details of the microcode in the Intel 8087 floating-point chip, I have a new blog post...
https://www.righto.com/2025/12/8087-microcode-conditions.html

2025-12-09 18:33:38 UTC

In 1980, Intel announced the 8087 Math Coprocessor, a chip that made floating-point 100 times faster. I opened up the chip, took photos of the silicon structures, and analyzed its circuitry. It's a very complex chip for its time. Let's take a look inside...
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/691/041/901/490/910/original/1cdcfcb63068720f.png

2025-09-15 17:22:01 UTC

I've been using HTML forever, but I just learned that <div /> isn't the same as <div></div>.
In the first case, the div isn't closed and everything that follows is inside it.

2025-09-10 16:28:19 UTC

Can anyone identify these mystery terminals? The first is from "Los placeres ocultos" (1977). The second is from "Policia" (1987). (Someone wants help identifying computer hardware in Spanish movies.)
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/180/922/622/920/348/original/3775c09f6a038a85.jpg
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/180/923/259/523/681/original/2499d8b2885e7420.jpg

2025-09-09 20:33:19 UTC

Can anyone recognize this front panel from a mainframe computer from a 1978 Spanish movie? The black color resembles IBM System/370 panels, but the knobs don't look right. Other information suggests maybe an IBM-compatible FACOM/Fujitsu computer.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/176/229/183/333/474/original/11621ba59a67038f.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/176/230/628/398/559/original/7f8cfa6e50c0e0d8.jpg

2025-09-06 15:30:13 UTC

- reply

Marilou Schultz's rug is on display at SITE Santa Fe's current exhibition: "Once Within a Time"; let me know if you see it!
Photo credit: rug photo from First American Art Magazine.

2025-09-06 15:29:48 UTC

- reply

To learn more about the 555 timer rug, see my latest post: https://www.righto.com/2025/09/marilou-schultz-navajo-555-weaving.html

2025-09-06 15:29:16 UTC

- reply

Marilou Schultz first made a chip rug in 1994, when Intel commissioned a rug based on the Pentium as a gift to AISES (American Indian Science & Engineering Society). The Pentium weaving used natural dyes, while the 555 weaving uses aniline dyes and some metallic threads for more intense colors.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/158/057/875/855/920/original/e58557ce2c42d82a.png

2025-09-06 15:28:14 UTC

- reply

Marilou Schultz based the rug on a photo by Antoine Bercovici (Siliconinsider). He used a special dark field microscope that produces a black background, highlighting the metal wiring on top of the silicon. The rug (left) mostly matches the photo (right), but there are some artistic changes.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/158/055/265/875/669/original/d17a9422b875d9bf.png

2025-09-06 15:27:04 UTC

Celebrated Navajo (Diné) artist Marilou Schultz recently completed a striking weaving. Although this rug may appear abstract, it is a representation of the wiring inside an integrated circuit. It shows the 555 timer, said at one point to be the world's most popular IC. Let's take a closer look...
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/158/050/948/973/697/original/1c67c7967bb775ca.png

2025-08-19 14:08:50 UTC

- reply

That QR code is the coolest thing. Nice how the black parts show the underlying structure, including the alternating squares of the timing pattern, while the colored parts have the data

2025-08-18 18:49:18 UTC

The latest issue of the journal Science mentions a magnetic compound Cr2Gr2Te6. The element Gr confused me, but it turned out to be a typo for Ge, germanium. Strangely, I found multiple papers with the same typo in the same context, so I wrote a short blog post about it.
https://www.righto.com/2025/08/Cr2Ge2Te6-not-Cr2Gr2Te6.html
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/051/260/012/006/032/original/b4d763331f5a9876.png

2025-08-17 14:46:00 UTC

How did Intel's 386 processor connect to the outside world? Tiny wires connect to the bond pads around the edges. Special input/output circuits—visible in zoomed-in blocks—send and receive signals while protecting the chip from dangers that could destroy it. Let's take a look... 1/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/044/643/703/148/847/original/ade5aadb4db85882.png

2025-08-09 21:29:33 UTC

- reply

Here's one more scan of the 386 for your viewing enjoyment...
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/115/000/927/586/381/110/original/1a3b0b8050871852.png

2025-08-09 16:49:07 UTC

- reply

For more photos and information on the 386 package, see my latest blog post. Thanks to Jon Bruner and Lumafield for the CT scans.
https://www.righto.com/2025/08/intel-386-package-ct-scan.html

2025-08-09 16:48:24 UTC

- reply

Here are the remaining four layers inside the package: one signal layer and three power layers. There are four power layers altogether so the chip's I/O circuitry and logic circuitry can be powered separately. That way, current surges from I/O don't affect the CPU.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/999/814/940/400/506/original/e87e706cd64b30c0.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/999/815/489/913/022/original/3a0535796477af65.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/999/815/802/643/766/original/4f93d34930af14b8.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/999/816/463/499/658/original/dec0e7a371ccf122.png

2025-08-09 16:45:03 UTC

- reply

One of the six power layers inside the Intel 386 processor's packaging. It's almost a solid conductive plane; the grid of holes joins the ceramic above and below. Most of the pins are connected to lower layers, so only a few pins (red dots) reach this layer.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/999/810/561/736/923/original/def2ccedc0cd3ceb.png

2025-08-09 16:43:59 UTC

- reply

The CT scan of a signal layer inside the 386's package. The wide abstract-looking regions connect the pins (red dots) to the chip in the middle. Unexpectedly, I spotted thin wires running to the edges of the chip. These are used during manufacturing to electroplate the pins with gold.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/999/804/296/926/732/original/b2bbce69021da18e.png

2025-08-09 16:42:20 UTC

- reply

The silicon die of the 386 processor is mounted in the middle of the ceramic package. Tiny bond wires connect the die to two tiers of contacts that surround it. Intricate but hidden wiring inside the ceramic connects these contacts to the 132 gold-plated pins. 2/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/999/800/784/471/090/original/c93a49f63aaa3e6e.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/999/801/823/991/849/original/44334b84e6e6e32c.png

2025-08-09 16:41:24 UTC

Intel's 386 processor was popular in late 1980s microcomputers. From the outside, the 386 chip is a boring ceramic square. But I obtained a 3-dimensional CT X-ray scan from Lumafield, revealing six layers of complex wiring hidden inside the ceramic package. Let's take a closer look... 1/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/999/798/287/025/630/original/b6b7263a4a45c5a4.png

2025-08-02 15:48:17 UTC

- reply

The chip uses these components to turn an FM radio signal into audio by combining an oscillator, mixers, filters, amplifiers, and other circuits. If you want to know more about this FM radio chip, I go into great detail in my blog post: https://www.righto.com/2025/08/reverse-engineering-analog-TDA7000.html

2025-08-02 15:48:10 UTC

- reply

The most unusual components in the chip are the "varicap diodes", which help adjust the radio's frequency. These special diodes act as capacitors with a value that can be controlled by changing the voltage.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/959/952/713/675/600/original/7618026ed7e72556.png

2025-08-02 15:47:30 UTC

- reply

Another important part of the chip is the resistors, which impede the flow of electricity. They are typically long and zig-zagging. The ends of the resistors are connected to the metal wiring. The arrows indicate the holes in the insulating oxide layer where these contacts take place.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/959/950/899/190/720/original/87eafe22cbeca0a6.png

2025-08-02 15:46:58 UTC

- reply

Transistors are the main components of the chip, amplifying signals and performing other functions. Modern chips use CMOS transistors, but this chip uses NPN and PNP transistors. These diagrams show how the transistors look on the die, along with cross sections showing the silicon layers.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/959/945/345/806/985/original/58435326da175ddd.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/959/947/586/432/711/original/139386b4f7303f94.png

2025-08-02 15:45:41 UTC

- reply

This close-up of the die shows how the metal wiring (white lines) connects the underlying silicon structures (gray and green). Unlike modern chips, this chip only has one layer of metal, so the wiring needs to twist and turn when other wires get in the way.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/959/942/774/410/322/original/449d60d870154dcf.png

2025-08-02 15:45:07 UTC

- reply

Here's a die photo of the FM radio chip, the TDA7000. You can see the greenish silicon, with white metal wiring on top. The 18 white squares around the edges are the bond pads that connect the die to the chip's pins, with gold wires attached. I've labeled the main functional blocks of the chip. 2/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/959/939/647/229/975/original/cd11258761bb6f6f.png

2025-08-02 15:44:12 UTC

In 1983, Philips produced the first FM radio receiver on a chip, leading to products such as the FM radio wristwatch. Let's look at the tiny silicon die inside this chip and see how it works. 1/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/959/936/994/946/905/original/91629cbe2c94ae88.png

2025-07-21 16:38:46 UTC

- reply

This box has thirteen orange digits at the top, which look like Nixie tubes. But they are a different technology called edge-lit lightguide display. Each digit has ten plastic sheets and ten lightbulbs. Each sheet has dots etched in the shape of a number. Lighting a sheet lights up that number.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/892/203/701/092/439/original/cd6176f12e31314c.png

2025-07-21 16:37:44 UTC

We obtained a mysterious box of 1960s electronics. I reverse-engineered it and with much effort, we got it running. It turns out to be a test unit for testing NASA's Up-Data Link, a system from the Apollo moon landing to control the spacecraft from the ground. Let's take a look inside... 1/n
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/892/199/248/415/500/original/4055c8cc57e3f577.png

2025-07-12 02:30:18 UTC

- reply

Texas Instruments introduced LED-based opto-isolators in 1964. They used an infrared GaAs LED shining onto a silicon photo-transistor. This device had the curious name of "molecular multiplex switch/chopper"; making small devices through "molecular electronics" was the hot thing back then. The device cost $275 (almost $3000 in current dollars; nowadays, an opto-isolator costs a few pennies. 2/3
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/837/899/854/387/798/original/b315a286594d7302.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/837/900/540/397/085/original/7261613364c052f1.png

2025-07-12 02:15:40 UTC

NASA used this "Up-Data Link Test Set" to test one of the electronic systems onboard the Apollo spacecraft. I've been reverse-engineering this box and I found that it uses LED-based opto-isolators. I didn't realize that they had LEDs and optoelectronics in the mid-1960s, so I investigated... It turns out that opto-isolators existed back then but were extraordinarily expensive. 1/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/837/825/895/886/494/original/eeac334015ce1042.png

2025-07-11 00:57:12 UTC

- reply

And if you'd rather see inside the Globus visually, check out my blog post that examines its gears, cams, differentials, and other mechanisms in detail. It's essentially a mechanical analog computer. Thanks to Marcel for providing the Globus. 4/4
https://www.righto.com/2023/01/inside-globus-ink-mechanical-navigation.html

2025-07-11 00:56:52 UTC

- reply

Lumafield's scanner is very cool if you want three-dimensional scans of objects for failure analysis, reverse engineering, or inspection. You can play with the Globus scan for yourself at this link, zooming, rotating, and slicing the Globus. 3/4
https://voyager.lumafield.com/project/d848dd54-d594-479f-a723-a463547ea7aa

2025-07-11 00:56:27 UTC

- reply

This photo shows the Globus with its rotating globe and a crosshair showing the spacecraft's location. Dials next to the globe indicate latitude and longitude, while the dial at the bottom indicates when the spacecraft is in light or darkness. The knobs control orbital settings. 2/4
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/831/873/991/670/722/original/b01ccf7e06149f21.png

2025-07-11 00:55:22 UTC

The Globus is a navigational instrument that uses a rotating globe to show the position of the Soyuz spacecraft above the Earth. Inside the Globus, a complicated system of gears and motors positions the globe. Jon Bruner from @npub1jsp…ggr4 created a three-dimensional X-ray scan for us. 1/4
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/831/862/779/820/382/original/e45eb7c1d2002304.mp4

2025-06-25 17:34:03 UTC

In the 1960s, this absurdly large camera was used to make masks for integrated circuits. The layers of the integrated circuit were drawn at large scale and then optically shrunk to make the glass masks that were used in manufacturing.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/745/201/551/682/693/original/c17a4bb37ddc8429.png

2025-06-20 03:02:53 UTC

- reply

It turned out that the computer's START button had a strand of wire that shorted one of its contacts. As a result, it would keep re-starting as long as you held the button, blowing through the HALT or single-step until it crashed. The contacts bounced a bit, causing the random gaps we saw.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/713/459/741/661/700/original/5b7e77fd908d7ab7.png

2025-06-20 03:01:09 UTC

- reply

You push the green START button on the console to start the computer after a HALT, and also to single-step it. Hmmm. For convenience, there's also a START button on the card reader and one on the printer. We found that everything worked fine with those buttons. Hmmm.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/713/454/783/296/234/original/98f25f1d9fdbd9f5.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/713/456/124/385/314/original/dd9b9c16c891bef7.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/713/457/226/551/709/original/dbbe7d96b9cc404d.png

2025-06-20 02:59:46 UTC

- reply

I hooked up an oscilloscope and discovered that the computer wasn't running a single instruction cycle. Instead, it ran a few cycles (the pulses below), stopped briefly, ran a few more, irregularly stopping and starting, "building up steam" until it ran thousands of cycles. But why this pattern?
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/713/452/563/450/784/original/2120e7238b75cac3.png

2025-06-20 02:58:19 UTC

- reply

The IBM 1401 computer lets you run code a single cycle at a time, so you can see what happens at each step. (The arrow on the control panel.) But when I tried to single-step through the bad HALT instruction, the computer immediately died in a distant memory location. Now we had two problems. 2/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/713/446/074/748/367/original/0f3ebe6d7b121630.png

2025-06-20 02:57:39 UTC

Last week, the vintage IBM 1401 computer at the Computer History Museum started behaving strangely: it wouldn't halt. More specifically, if you had two HALT instructions in a row, it would halt for the first, but when you continued, it crashed mysteriously. Here's how we fixed it.... 1/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/713/444/182/512/594/original/f544d480f5bac21d.png

2025-05-17 19:02:38 UTC

How thin is spider silk compared to the wiring on a chip? Here's a close-up of a spiderweb strand diagonally across an Intel 386 processor. The spiderweb is 4 µm thick, just a bit thinner than this 1985 chip's 5 µm metal wiring. Wiring in modern chips is 100 times thinner.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/524/719/690/200/261/original/723071f94cdf898d.png

2025-05-16 22:00:30 UTC

- reply

Transistors in a matchbook? We considered replacing the bad transistor with one of these vintage Texas Instruments transistors that come in a cool package. But
@curiousmarc.bsky.social
found a different transistor that worked better. Next week I'll verify that it fixes the tape drive. 5/5
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/519/754/651/367/773/original/75aa3e6e6f8172a8.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/519/756/272/495/535/original/101fadb8d0b83618.png

2025-05-16 21:58:40 UTC

- reply

We tracked down the fault to this circuit board and found the transistor in the lower left had failed.
@npub1vgw…46mg replaced the bad transistor with one from his extensive collection. This board is a "hex inverter". For comparison, a 1970s-era hex inverter IC is at the right.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/519/750/015/431/953/original/e8bd76244fa9924a.png

2025-05-16 21:57:58 UTC

- reply

Unfortunately, the computer sat in an unheated garage in Germany for many years, so many of the parts suffer from corrosion. You can see green corrosion around the base of some of these transistors. Every so often, one of the transistors will fail, causing a computer malfunction.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/519/745/608/236/342/original/592dbf75d6ffdc76.png

2025-05-16 21:56:59 UTC

- reply

The IBM 1401 was the most popular computer of the early 1960s. It has rows and rows of circuit boards—each the size of a playing card—to implement its logic. This computer predates silicon, so it is implemented with germanium transistors.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/519/740/755/450/343/original/6decec7290ed6674.png

2025-05-16 21:55:49 UTC

Tape drive repair at the Computer History Museum. I helped figure out why the tape drives wouldn't backspace. The problem turned out to be a bad germanium transistor on an inverter card. 1/5
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/519/735/109/459/784/original/a9d30dd925d284c4.mp4

2025-05-10 16:46:32 UTC

- reply

The Kindle 11 uses an MT8113 processor. This is a 1 GHz dual-core Cortex A-7, which is 32-bit like the 386. The 386 ran at up to 40 MHz. That would put your Kindle at roughly 50 times as fast. (Apples and oranges blah blah blah.)

2025-05-10 16:31:50 UTC

- reply

If this isn't enough about the 386 processor's prefetch unit, see my latest blog post. It has more photos, schematic diagrams, and lots of explanation:
https://www.righto.com/2025/05/386-prefetch-circuitry-reverse-engineered.html

2025-05-10 16:28:35 UTC

- reply

This close-up of the 386 processor's die shows the different parts of the prefetch unit. There's a lot going on just to hold 16 bytes of instructions.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/484/477/253/419/159/original/ea7fe8c5ea38a887.png

2025-05-10 16:27:51 UTC

- reply

Another part of the prefetcher converts an 8-bit value to a 16- or 32-bit value. This is tricky if the value is negative, since the upper bits must be filled with 1's. Here's a close-up of the sign extension circuitry that fills upper bits with the correct value.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/484/475/201/835/225/original/93dc6d99f9c69408.png

2025-05-10 16:27:15 UTC

- reply

How hard can it be to add 1 to a number? The problem is that incrementing a 30-bit number can require carry the 1, carry the 1, repeated 30 times: slow. In the 386's prefetcher, Manchester carry chain and carry skip circuits make this fast, implemented with complicated dynamic logic.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/484/471/126/460/803/original/9cfbde4438b41413.png
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/484/472/573/220/621/original/e88e0b31f016a261.png

2025-05-10 16:25:28 UTC

- reply

The 386 processor is a 32-bit processor, so it works in 4-byte chunks. This network shifts bytes around so the chunk is aligned correctly. Bits go down the vertical wires, across horizontal wires, and then out the bottom. Energizing different wires provides the desired shift and lines up the bytes.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/484/465/048/138/654/original/9c84259172b1c141.png

2025-05-10 16:24:11 UTC

Anyone use Intel's 386 processor from 1985, the first 32-bit x86? To improve performance, it preloaded instructions into a 16-byte prefetch queue (highlighted in red). Why does such a small queue take up so much of the chip? Let's look under the microscope at its complicated circuitry... 1/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/484/459/842/162/685/original/1e8828f331603f65.png

2025-05-01 17:42:53 UTC

Intel's 386 processor (1985) moved the x86 architecture to 32 bits, but it needed to be backward compatible with earlier 16 and 8-bit processors. As a result, it needed complicated circuitry for its internal registers: six different circuits for 30 registers. Let's look at the silicon circuits. 1/N
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/433/808/533/161/186/original/96d10f995fda6bdc.png

2025-04-13 16:42:29 UTC

- reply

Yes, the Retro Chip Tester generated the right checksum for the ROM, but the chip failed in use. I suspect that the badly-programmed bit didn't settle on the output fast enough, causing a timing issue. We also saw one ROM address where the logic analyzer showed the right byte on the bus but the 6502 clearly read the wrong byte. This also suggests a timing issue (or maybe a voltage level issue).

2025-04-13 16:23:12 UTC

- reply

For more details on this Commodore PET repair, see my blog post:
https://www.righto.com/2025/04/commodore-pet-repair.html
Thanks to @npub1vgw…46mg
and @npub1xvg…62t2 for their help with this project. Thanks to Mike Naberezny for providing the PET.

2025-04-13 16:22:17 UTC

- reply

To make a long story short(er), we had programmed the replacement ROM with not enough voltage, so it had a couple of bad bits, enough to make the computer crash.
Finally, we got the computer running and I ran a simple graphics program, similar to ones I made when I was a kid.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/331/569/005/503/091/original/be7a8cfafed1b033.mp4

2025-04-13 16:19:29 UTC

- reply

Now the failures were very strange: the computer booted up and I could type in a simple "hello world" program. But when I ran it, the program printed random floating point numbers and then crashed. There was also a mysterious 2 by 2 checkerboard character near the middle of the screen.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/331/558/943/447/469/original/a352f0ce5992d915.png

2025-04-13 16:18:44 UTC

- reply

Even with the replacement ROMs, the PET still wouldn't boot. Marc attached his vintage logic analyzer to the 6502 so we could trace what the processor did, byte by byte. I found that the memory test was failing on the first byte. We found three bad memory chips, but the system still failed.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/331/557/032/648/715/original/3f409449221934cf.png

2025-04-13 16:18:06 UTC

- reply

We used a "Retro Chip Tester" to test the ROM chips and found two bad ROMs were producing bad signals. Unfortunately, the PET's ROMs are weird, unavailable chips, so I got two adapter boards and used normal ROMs. The adapters are in the third row of chips from the front.
https://assets.oldbytes.space/assets.oldbytes.space/media_attachments/files/114/331/554/375/953/103/original/5e1685a8cb0326c5.png