Reverse-engineering an early calculator chip with four-phase logic
righto.com
righto.com
To me the traces look more like they were laid out with tape, not hand-drawn. Back in the day we used a kind of thin black masking tape that was sold in various widths for this purpose. It had the property that you could form it into a bend radius that looks just like the traces on this board.
I agree that this doesn't look quite like the tape layouts I'm used to, but I only did tape on bare copper for one-offs, never tape on photofilm. Of course, the rules for those processes are somewhat different even if the basic mechanism is the same.
Also, those aren't "7044" and "7042" chips -- those are the date codes, weeks 44 and 42 of 1970, respectively. The IC marks are the top lines, "NRD2256" or "AC2261".
As this calculator is 20+ years older than me, can someone enlighten me as to why the dates would be like that? The only thing I can think of is that the manufacturing runs were done sequentially instead of all at once. Was that a common thing for ASICs back then?
I’m also curious why calculators were such a common thing then, yet they still cost over $1000 (in today’s money). I get it makes calculations quicker, but $1000 just seems excessive. Was it just the cost of manufacturing justified by the cost savings (in time) for businesses who bought them?
It still is common. It's more efficient to make a batch of 1000 widget-Xs and then a batch of 1000 widget-Ys, rather than 1 X and 1 Y and then 1 X, and so on. It saves changeover time for the machines in the process, in this case things like the wirebonders and so on. Not so much the wafer fab if the process was identical, though in those days it may have also been worth it there too.
> I’m also curious why calculators were such a common thing then
The magnitude of the speedup is really pretty huge. Slide rules helped but were only ~3 digit accurate; that's not always good enough. Sit down and do something like solve a simple quadratic equation with non-integral coefficients without any calculator assistance and you'll start to get an idea how bad it can be....
The tech was interesting though; their AL1 processor was based on four-phase logic like this calculator chip.
i see that the 4-phase logic is time-based. are there similar logic systems that are voltage-based, so instead of a high and low, you had very high, high, low, and very low? (i'm thinking a decimal computer would be neat but maybe too difficult given the relative benefits.)
Another example of 4-level logic is the Intel 8087, the math coprocessor chip used in the IBM PC. The microcode for math computations was too big to fit on the chip, so Intel used a 4-level ROM, with four transistor sizes for each storage location. So each "bit" gave you four different voltages. Comparators converted the four voltages into two bits. The result was twice the ROM density, allowing the microcode to fit. (In actuality, it was less than twice the density because of the additional circuitry.)
Finally, flash memory uses multiple voltage levels to store up to 4 bits per cell.
As far as decimal computers, there were lots of decimal computers in the 1950s and 1960s for business use. However, they used BCD (binary-coded decimal) or timed pulses, rather than 10 voltage levels.
Update: Found an oral history reference. The name was coined under the Eisenhower administration, along with it's etymological sibling Rocketdyne. Bad ass. https://airandspace.si.edu/research/projects/oral-histories/... Their inventions apparently include fundamentals to the field such as automated inertial navigation for flight control (1950), automated celestial navigation, and redundant flight control computers. They initially used memory based on the deferred transit of audio through crystal.