OT-Novus 3-level RPN. What were they thinking?

[Matt Agajanian]

Hello all.

Perhaps you could shed some light on this mystery. A very intriguing curiosity is why the 4515 (Mathematician) incorporated a 3-level RPN without either R^ or Rv but the 4525 Scientist had a 4-level stack with a Rv manipulator which does facilitate R^ by way of a trio of Rv strokes. So, what was the point of a 3-register RPN stack without even stack manipulation keys?

[Walter B]

Albert Einstein: "Everything should be made as simple as possible, but not simpler."

[Matt Agajanian]

Good point. Although, with three registers, there's not much promise to evaluate complex expressions.

[Walter B]

Some years before even HP thought 3 levels were enough ;-)

[Andrés C. Rodríguez (Argentina)]

I think complex functions were way in the future at the time these products were made. Users combined their calculators with pencil and paper, as they did with classic slide rules. I wonder if there were complex slide rules of some kind...

[Lincoln R.]

What were they thinking? "Memory is expensive."

[Luiz C. Vieira (Brazil)]

Yet, the little beast is programmable... 102 program steps! If there is a forth stack register, chances are program steps would be reduced to hold its contents and surely extra ROM would be needed to accommodate the necessary handling operations for the extra register. Nothing comes for free, indeed.

Edited: 29 June 2012, 8:11 a.m.

[Kerem Kapkin (Silicon Valley, CA)]

Interesting that it does not have EEX funtion which limits its scientific use in the sense of what we expect from calculators today. It is almost like a programmable digital slide ruler with 8 digits.

[Steve Simpkin]

I acquired (for free) a Novus 4510 Mathematician in early 1977. This model was not programmable but I have fond memories of it as it was my first RPN calculator. It was soon replaced with an HP-25. As Kerem mentioned, it's biggest limitation was not the 3-level stack as much of the lack of exponents (EEX). This handicap severely limited its usefulness for scientific calculations.

http://www.thimet.de/calccollection/Calculators/Novus-4510/Contents.htm

[Michael de Estrada]

The real limitation IMO was that it only had one storage register, although it did at least have limited register arithmetic. One interesting feature that was not found on other single register calculators of its time was the M+x^2 capability, which made single variable statistics calculations possible. When the HP 45 came out with 9 storage registers instead of only one on the HP 35 plus full storage arithmetic, it completely obsoleted the HP 35. Also, display control was a big improvement over the HP 35.

[htom trites jr]

Three levels is sufficient for a straight numerical evaluation of a polynomial using Horner's method.

[Matt Agajanian]

Intriguing idea. That's a good point.

[db (martinez, ca.)]

I have that one and other 3 level stack RPNs. To me the biggest problem with Nat Semi / Novus's implementation of a three level stack is that the top register does not replicate. One can't do repetitive arithmetic by loading the stack and rolling it.

Lincoln R's answer above is spot on. It could be worse though. There were a lot of RPN 2 level stack machines. Even scientific and programmable ones.

[Michael de Estrada]

e.g. Sinclair Scientific

[db (martinez, ca.)]

Michael; no fair. you've been studying. but there's more!

from memory, because i'm locked out of my non hp rpn page by "an annoying web spider"........... the sinclair scientific programmable and at least one elektronika scientific model use a two level stack.

now to give the devil his due; no less an authority than dave hicks has said that a two level stack is NOT rpn, just consistently postfix. dave is "the dude" but i respectfully differ, if only to have more non hp RPNs to choose from.

edited to correct my stupidity

Edited: 30 June 2012, 1:52 p.m.

[Marcus von Cube, Germany]

The restrictions to these machines are similar to what TI had created with their SR-50 and SR-51 models: No parentheses and just multiply/divide/exponentiation before addition/subtraction. In order to compute a product of sums a memory register had to be used. It must have been a constraint imposed by the available memory. The 51 has linear regression but this occupies the whole machine so that almost no other function is available while a regression analysis is in progress.