[Amen]’s Rockwell 920 calculator from the 70s was a very impressive piece of hardware for its time. It sported a 16-digit display, a printer, and it could run programs. It even had a magnetic card reader/writer that could be used to store programs and data externally. Seen through today’s eyes, it was less like a calculator and more like what we would call a single-board computer. They are also a window into another era, a time when many of the electrical design assumptions we take for granted hadn’t happened yet. When the time came to dig into what made the calculator tick, [Amen] had a lot of work to do just to get basic tools running.
For example, [amen]’s Blue Pill (an open-source, multipurpose test and measurement tool) is, on one hand, the perfect tool to snoop on the inner workings. However, those inner workings happen to use negative logic at -17 Volts, which means a logical zero is -17 V and a one is 0 V. Oh, and it uses an oddball clock rate, to boot. Since the Blue Pill doesn’t support -17 V negative logic (does anything?) a bit of custom work was needed to craft an interface. Once that was working, the Blue Pill was off to the races.
The unfamiliar elements didn’t end there. The pins on each IC, for example, are in a staggered layout quite unlike the DIP pattern most of us (and our tools, breadboards, and IC clips) are familiar with. As for the processor itself, [amen] has access to low-level documentation on Rockwell processors and instruction sets, but the timing diagrams are puzzling until one realizes the processor has two clock inputs at two different frequencies, resulting in what [amen] describes as four separate “clock phases”.
These design decisions were certainly made for good reasons at the time, and they even have a certain internal harmony to them, but it’s still a window into an era when the elements underpinning much of what we now have and work with had not yet happened.
Check out the video embedded below to see [amen] explain what it took to hook the Blue Pill up to a Rockwell 920. Also, if you’d like to see one of these vintage machines demonstrated in all its functioning glory, here’s a video of one being put through its paces.
14 thoughts on “Vintage Calculator Design Shows Just How Much We Take For Granted Today”
How many hundreds or thousands of time more powerful is the Blue Pill than the Rockwell 920? Like a super advanced alien doctor that breathes methane, fixing up a human.
The negative voltage would mean that is using ECL logic, right? That and the odd clock are quite interesting workarounds to get the performance the logic could do. All sorts of things like this are neat looks back in time at designs of then-some of which still hangs around now for special applications. All in all, pretty cool.
No, this is PMOS. ECL uses a precise negative voltage of -5.2V and small voltage swings to represent 1 and 0.
Correct, this isn’t ECL. ECL also uses ‘unsharp’ edges due to not wanting to saturate the driving transistors so they can switch quicker. Interesting type of logic.
I always thought PMOS was a bit strange, since it works on voltages which are negative in respect to ground. It can handle higher voltages than NMOS, though, and this advantage was important in low-power ICs which drove VFD displays directly.
Why didn’t they use NMOS and then a separate level shifter? PMOS has half the mobility, so you need twice the die area (and I presume that was precious).
Because PMOS predated NMOS?
Some weird cobbled together dingus from a four fingered solar system shows just how much we take for granted today? Not buying it.
HP calculators were good.
In hindsight even I don’t fully get that joke, but they made it work and it was still relatively early days. There’s nothing magical about 5 volts or any other standard. That said, the PC board shown is byootiful.
An interface to -17 logic is not a big deal. Early microprocessors used negative voltages. In the case Rockwell’s calculator, you just use -17V as the reference and then logic signals swing between 0 and +17 V. A couple of resistors would knock that wide voltage swing down to 3.3 or 5.0 V logic levels.
That’s exactly what I did. You have to be a bit careful, though, as your 0V is now at -17V if you attach to a computer for data dumping, as I did. Opto-isolated USB was useful there.
Yeah, I was playing along with the video and as soon as you said “I’ll just use -17 as my ground reference” I was all “Oh, be really careful, there’s a good reason why that’s not recommended!” And then you mentioned ordering an opto isolated USB cable and my concern ratched down a bit.
I seem to recall that when I got into the game most of the “interesting” MOS chips needed a negative supply rail, if not actually using it on the logic levels.
Had something to do with properly biasing the internal gates – I never looked into it but I seem to recall being told that the gates depended on current sinks that went off to the negative rail.
Oddly, it seemed to be a common cause of failure, since it was often bypassed poorly.
This is similar to the electromechanical calculator I used in 1963 on my first month working in a research lab as an undergrad. It reduced an estimated three months’ work with slike rule to a week. It could do logs and exponents, too.
Please be kind and respectful to help make the comments section excellent. (Comment Policy)