Personal calculator has key to solve any equation f(x)=0 (1979) [pdf]
people.eecs.berkeley.edu
people.eecs.berkeley.edu
I remember seminars at Berkeley where equipment manufactures show up to talk about their gear and Kahan would sit in the front row and grill them over their floating point implementations.
This is a great article, and reading it might give younger folks an idea of why HP users from the 1900s loved their calculators so much. They were designed for people who needed to get things done. And it wasn't just the algorithms inside. The physical machines were terrific, too. The pivoting keys left no doubt as to whether you had entered a digit (or whether, as in competitors' machines, the keystroke had entered the digit twice by mistake). And then there's RPN. Oh, lovely RPN -- trusty friend to those who needed correct answers and needed them in a rush.
Such elegance (of machines and writing) is rare nowadays.
Thanks for posting this, which brought back good memories.
The symbolic computation GUI is also a marvel of design that is unsurpassed to this day, especially in how you can select semantic blocks with a keyboard, and how good the rendering is. It felt like a human-centric interface rather than a computer-centric one. Although the general UI of the calculator was more of a confusing mess, so it had its flaws.
Well, I got mine for cheap from another student who "bought this useless garbage" so it's not a universal point of view.
Whenever I had to borrow or do something on a friends TI, I couldn’t stand it. The wobbly keys made typing all those parentheses even more of a nightmare!
These must be the team who were designing the HP-28S and user rpl - the beautiful love child of Forth and Lisp.
And what a calculator they put out!
It's hard to appreciate what you're holding in your hand when you have one of these machines. When I learned the 48s had both Kermit and a serial port, the clouds parted and the sun came out. Really extraordinary capability in a $99 calculator.
Always liked the form factor of the 28s, just needed a way to get code on and off (besides the printer port) of them.
But ya, i always like the 28 more than the 48.
Have you heard of RPL/2? A modern implementation of RPL meant for scientific computing? Its annoying to compile but its quite nice!
Also, assuming that a second guess isn't provided, how does the calculator choose one? I'd imagine there are numerical precision problems that make choosing one hard.
[edit]
Wait, is it using this algorithm? "False position": https://en.wikipedia.org/wiki/Regula_falsi . I thought it was using something primitive like the "secant method": https://en.wikipedia.org/wiki/Secant_method . But the article suggests that it tries to be more clever than that.
From memory it was a single 45 unit burn (about half the fuel) in the final moment.
Relatedly I learned a whole lot of algebra from reading a slide rule manual. I don't think I had a good intuitive understanding of logarithms until I saw them as pictures with arrows showing how they work on the slide rule. And then the manual was full of practical ways to reduce various common math problems to problems that are easy to solve with the slide rule, usually by figuring out how to express them as powers or logarithms.
As for the slide rule, I believe it was a Pickett booklet, likely one of the ones that have fortunately been archived here: https://www.sliderulemuseum.com/SR_Library_Pickett.htm
The Asimov slide rule book is also very good and I have a copy. I believe I inherited it, along with several slide rules, from my grandfather who had been an EE at the NIST and also fastidiously kept a lot of things he had bought for grad school. One of the slide rules has "US Government Property" silkscreened on with the scales which is fairly neat, I assume DoD or someone had had a large number made on contract like they used to do with office supplies. I also have a US Government Property set of drafting tools that I use when hand-drafting sewing patterns.