How a 1974 Sinclair Scientific calculator works

Found 1974 Sinclair Scientific calculator workings link.
http://hackaday.com/2013/08/30/ken-shirriff-completely-reverse-engineers-the-1974-sinclair-scientific-calculator/

Article here. Hackaday says - Read it now. http://files.righto.com/calculator/sinclair_scientific_simulator.html
http://en.wikipedia.org/wiki/Sinclair_Scientific

Great stuff with history, schematics and code. Reading the stuff in the article about how multiplication worked in the TI TMS 080x
http://datamath.org/Chips/TMS0803.htm ,
realized it was like a mechanical calculator,
http://en.wikipedia.org/wiki/Mechanical_calculator ,
and it struck that it might be an interesting project to actually make a mechanical scientific calculator. Never thought about possibility before.

Starting point would be some common manual mechanical calculator like -
http://en.wikipedia.org/wiki/Odhner_Arithmometer
Interestingly the most commonly produced of these are so common - though unseen in the modern era of electronic calculators - as to have practically no antique value - and can be obtained for little money (unless you want to start with a rare Curta). http://en.wikipedia.org/wiki/Curta

Arithmometer pretty much works like the ALU math http://en.wikipedia.org/wiki/Arithmetic_logic_unit in the TMS0803/Sinclair... with a little help from the user during shifting. A Monroe has capability for automatically doing the shifts for multiply and divide... making it a four function mechanical calculator
http://www.hpmuseum.org/ffhand.htm
http://en.wikipedia.org/wiki/Monroe_Calculator_Company

The relation of the TI chip to four function mechanicals may not really be an accident. Note the connection to Busicom http://en.wikipedia.org/wiki/Busicom
- who was making arithmometers when they partnered with Intel to make the 4004 microprocessor.

So then one needs registers (there are only two in most mechanicals - not counting a "check" counter in some) and an instruction cache of some sort (to get the micro code for shifts and such). And some glue/flow mechanisms to stick the whole thing together. Of course one might consider these solved problems if one considers a pad and pencil to be RAM and a calculator manual (or other paper documented process) to be ROM, but part of the point of a calculator is to automate or mechanize as much as possible, to take human error out of the process.