Bio: Computer history. Reverse-engineering old chips. Restored Apollo Guidance Computer, Alto. Ex-Google, Sun, Msft. So-called boffin. @kenshirriff@oldbytes.space
Location: Silicon Valley
how it started: how it’s going:
Extremely niche request: Does anyone know who were the designers of the Intel 82586 Ethernet/LAN coprocessor chip? Or have an errata sheet for this chip?
(A friend is writing a history of Ethernet.)
The Bendix Central Air Data Computer (CADC) was introduced in 1955. This complex mechanical analog computer uses tiny gears, cams, and differentials to determine airspeed, altitude, etc for fighter planes. I hooked a synchro up to the CADC and got some output from it. 🧵
Although the CADC performed computations mechanically, its outputs are electrical. It uses small synchros to convert shaft rotations to three-phase voltages. For accuracy, it uses two synchros for each measurement: coarse and fine.
I hooked up a synchro to the Mach air speed (fine) output. When I manually turned the gear inside the CADC, the synchro rotated to match. This shows that we have the right wiring, and is a step toward powering-up the CADC and getting it running.
Here's a closeup of the Mach synchros inside the CADC, surrounded by gears and differentials.
The CADC has two air pressure inputs and a temperature sensor input, which it converted to rotations. Cams took logarithms, differentials performed addition, then cams performed exponentiation. This implemented multiplication. Complex cams provided other functions.
My previous threads discuss the Bendix CADC and the operation of synchros in more detail.
Thanks to Joe for providing the CADC.
nitter.lacontrevoie.fr/kenshirriff/stat…
We recently got very nice Christmas gift boxes from @PCBWayOfficial. Most interesting was a ornament with blinking and flashing LEDs. I opened up an LED and took a look at the microscopic flasher IC inside. (Thread, CW flashing lights)
Here's the ornament before I got my hands on it. The LEDs go through multicolor cycles of blinking and fading. There's no microcontroller on this board, just the LEDs and the battery.
Here's a closeup of the tiny RGB LED in a surface-mount 0807 package. The package contains separate blue, red, and green LED dies connected by bond wires to a tiny silicon die.
This video shows the LED in action under the microscope. You can see that the package contains separate green, red, and blue LEDs that are illuminated individually.
I extracted the die by boiling the LED in a few drops of sulfuric acid. The silicon die is very small: here's the die under the microscope next to a few grains of sugar. Honestly, I'm surprised I didn't lose the die while decapping the LED.
The die looks like CMOS with a few hundred transistors. Just a simple circuit, not a microcontroller. I think it's based on shift registers, maybe pseudorandom LSFR. The closeup shows three shift register stages. The arrows indicate the connection from one stage to the next.
The die has three large zig-zag transistors to drive the three LEDS. They are connected to the black pads where bond wires were attached. The complex transistors at the left are probably a regulator to control the current to the LEDs.
Thanks to @PCBWayOfficial and @SGirl0311 for sponsoring us and sending us the Christmas gift boxes. Ornament design is by @akirasan. Ornament analysis performed with @curious_marc and @TubeTimeUS. I hope dissolving ornaments got you in the holiday sprit!
When you hook two synchros together, current through the windings produces magnetic fields and the torque rotates the synchros into alignment. When the two synchros are aligned, the current stops. This is how one synchro controls the other.
This clip shows the three output voltages from a synchro. The voltages vary as I turn the shaft. Depending on the position, the green, yellow, or blue output is the largest.
I'm using small aerospace synchros. Inconveniently, they run on 26 V 400 Hz. That works out to 73.5 volts peak-to-peak. Fortunately, @curious_marc has a vintage HP amplifier that could output that voltage. We drove it with a vintage HP function generator.
The US Navy produced a detailed guide to synchros in 1944 and a more modern (2012) guide. Most of the diagrams in this tweet are from those.
maritime.org/doc/pdf/synchro…
compatt.com/Tutorials/NEETS/…
I had a synchro unboxing thread yesterday with some more info.
A synchro is a cool device that transmits or receives rotation. I've hooked up two synchros: when I turn one shaft, the other shaft turns to the same position.
Synchros were popular before everything went digital, for example transmitting signals through a ship. Synchros send the shaft position over three wires with varying AC voltages.
A syncho is sort of a cross between a motor and transformers. It has a rotor winding and three fixed stator windings. The rotor is powered with AC and generates three AC signals on the stator windings, depending on the angle.
A synchro can electrically transmit a rotational position through wires. A synchro looks like a little electric motor, but sends the angular position of its shaft. I needed synchros for a project, so here's an unboxing thread, showing why military parts are so expensive.
Bendix developed the Autosyn line of precision synchros in the 1950s. I got one from 1987. It came in a sturdy box labeled "Fragile". Inside the box is a thick paper package with a "Method II Package" label. This refers to military spec MIL-P-116 for protection from water vapor.
Inside is a plastic bag with a humidity indicator and a pack of desiccant. Uh oh, the indicator is pink. Looks like some humidity got in during the past 35 years. But there's still more packaging!
Next is a a nice plastic box containing the synchro. Inside is a sheet stamped with all the tests that this synchro passed. Apparently each synchro must pass over a dozen tests. But that's not all!
The box also contains a strip chart showing the exact error of this synchro over the range of angles. Each synchro is individually tested so you can be sure you get a good one.
I knew that military components had more testing that commercial components, but I hadn't realized the extent of it. This explains why the servos cost hundreds of dollars each. (I got this one on eBay, much cheaper, but the pink indicator shows you get what you pay for.)
I hooked up the synchro to an analog air data computer from a 1960s fighter jet, but that's a topic for another thread.
The IBM 729 tape drive was an icon of computing in the early 1960s with its spinning tape reels. We finally fixed a malfunctioning drive at @ComputerHistory. Last week we tracked down a short. The drive still didn't work until we found a bad relay. Now it works. 🧵
Here's a video showing the repaired tape drive in action, controlled by the test panel that swings out from the IBM 1401 computer.
After fixing the short in the -48V line, the tape drive powered up, which was good. Unfortunately, it didn't respond to any buttons. Conveniently, these drives have a grid of neon bulbs on the back showing relay status. Relay 119 should have been closed but it wasn't.
I pored over the documentation trying to understand what operated relay 119. Stan suggested pulling the relay out and testing it. Stan and Brenda discovered that one of the relay's contacts was bad. They applied DeoxIT liberally, put the relay back, and the tape drive worked!
For more, see last week's thread on finding the short in the tape drive.
There are many subtle differences though; they must have redrawn the layout rather than simply copying the masks. E.g. the resistors on input pins.
One nice thing about Soviet integrated circuits is the rational naming scheme, rather than the random numbers of US chips. In KM1810BM86: K=consumer, M=metal-ceramic package, 18xx=monolithic microprocessor series, BM=microprocessor, 86=copy of 8086.
Here's a closeup of the part number 1810BM86 on the die.
I've labeled the die with the main functional blocks. Around the edges of the die, bond wires connect the chip to the 40 external pins.
The text "СДЕПАНО B CCCP" on the package translates as "Made in USSR". The chip illustrates the large technological gap between the US and Russia . This 8086 clone was made in 1987; at that point the Intel 80386 had been produced for two years.
