19

I was cleaning out some junk when I encountered an interesting chip. It looks like a remnant from the Apollo space program with its all-metal body with the marking ADC581B 8533 6686. Next to the 8533, there's a company logo resembling a sine wave, which I initially mistook for the Texas Instruments logo. It's fairly large, around 1.5 to 2 inches long, with a noticeable heft for its size. There are 16 gold plated legs on each side.

I looked all over the web for a datasheet or any information on it, but couldn't find anything. The thought of it being some sort of radionuclide powered device and leading to another Goiânia briefly entered my mind. So far, however, no blue flashes or sparkles.

| Photo of large metal-cased chip

psmears
  • 678
  • 4
  • 6
user148298
  • 2,215
  • 5
  • 36
  • 51

2 Answers2

40

It's an analog to digital converter. Several years after this ad, they changed their logo to the one on your 1985 (week 33) chip. Magazine ad from "Electronics" May 1979.enter image description here

GT Electronics
  • 4,074
  • 7
  • 15
32

As @GT says, it's a 12-bit ADC, made by Hybrid Systems. Logo from USPTO.

enter image description here

According to this website:

​Company: Hybrid Systems Corporation Based: Bedford, Mass.

Founded: 1967 (merged with Harris Semiconductor spin-off Data Linear in 1986 & became Sipex Corp.)

Founders: Donald B. Bruck

Specialty: Manufacturer of hybrid integrated circuits, discrete component modules, thin-film networks and modular sub-assemblies for the industrial, military / aerospace and research market

This is a hybrid circuit, meaning that inside the hermetic can is not just a monlithic IC chip but a sort of ceramic (alumina) printed circuit board, typically containing multiple semiconductors (often bare wire-bonded dies), perhaps some passives similar to modern SMT parts, and resistors made by directly depositing thick film (printed) and/or thin film (deposited) resistive elements on the substrate (and firing them in an oven at relatively high temperatures- which, along with dimensional stability- is why the ceramic, and optionally trimming the values using abrasive or laser methods). Image from this web page:

enter image description here

The Electronics magazine advertisement says they are using thin-film networks for the critical resistors (probably an R-2R network with about 24 resistors) which means they would sputter or vacuum deposit rather than print those elements, and then they would trim them to get the required tolerances. Today that all can be done on a single IC chip, including the trimming and thin-film networks.

Spehro Pefhany
  • 376,485
  • 21
  • 320
  • 842
  • Is 12-bit normal? I notice that Harris Semiconductor also produced a 12-bit computer called 6120. – Lorraine Jan 24 '22 at 15:28
  • 8, 10, 12 and 16 are the most common. Now days, the most common ADC is one function in a microcontroller. Discrete chips like this are more common if higher resolution ADC is needed (18 to 32 bit) or better linearity, or faster acquisitions or ...? – GT Electronics Jan 24 '22 at 16:00
  • 2
    Yes, 12-bits is a common ADC bit width. As are 8, 10, 14, 16, 24 etc. The ADC bit width is only loosely related to the computer or MCU bit width. 12 bits means the resolution is 1 part in 4096 which is a reasonable resolution when compared with something like a synchro/resolver or the accuracy and stability of most analog sensors for that matter. There are physical reasons why any ADC bits beyond 20 are usually pretty much noise (that's one part in a million resolution) though 24 bit and even 32 bit converters are offered. For whatever reason it's rare to see odd numbers of bits. – Spehro Pefhany Jan 24 '22 at 16:02
  • @SpehroPefhany: For ADCs that accept differential inputs, it's common to have an even number of digits *including* polarity, or to have an even number of digits *plus* polarity, though it may be that the ones which have an even number of digits including polarity operate as a successive-approximation ADC over the entire range, while those that have an even number of digits plus polarity bisect the range based on which input is higher, and then produce an even number of digits after that. – supercat Jan 24 '22 at 16:44
  • 2
    @OmarL 12-bit ADC is **still** extremely common, particularly in industrial automation. I run dozens of industrial machines and almost exclusively they all use 12-bit ADCs and DACs, including half a dozen systems designed and built just a few years ago. Many industrial PLCs are still built with 16-bit native word sizes and a 12-bit ADC fits in there with three bits left over for channel selection, which is handy on modules that multiplex 4 or 8 channels into a single readout word. – J... Jan 25 '22 at 00:14
  • 1
    Keep in mind that there are industrial and precision 12-bit converters that are good to a fraction of an LSB and ones like the free '12-bit' converter on the Expressif ESP SOC devices that is more like an uncalibrated 8-bit converter (more than 10x worse). – Spehro Pefhany Jan 25 '22 at 02:53
  • @SpehroPefhany "10x worse?" 10x which number? If the initial number were, let's say 40. 10x would be 400. So I assume "10x worse" would be, -360? If my math isn't correct, please advise. Also, advise how you would do your calculation if things were only "0.5x worse". – GT Electronics Jan 25 '22 at 06:34
  • 3
    @GT >10x the LSBs error – Spehro Pefhany Jan 25 '22 at 08:44
  • 1
    @SpehroPefhany Yep. I pay around $700 per channel for analog I/O on my machines and they're very accurate. The ESP ADCs are about as bad as I've seen - the linearity curves look like scribbles from a kindergarten class. In a way, it's amazing that they work as well as they do given that you get fifteen of them in a package that you can buy with pocket money. – J... Jan 25 '22 at 18:14