Designing an FPGA Calculator from Scratch

Designing an FPGA Calculator from Scratch(baltazarstudios.com)

74 points by zdw 3 days ago | 10 comments

defrost 2 days ago |

Ten minutes in it looks like a great project walkthrough from design to physical device build.

Good start for anyone interested in the guts of going from logic gates to math() primitives ( add, mult, tan, sin, etc ).

Two snippets from the lede, one from a chapter heading:

  This is a scientific BCD calculator that uses binary-coded decimals, the same internal number format HP used in its scientific calculators going back to the 1970s. It represents every decimal digit as a 4-bit nibble, which means perfect decimal accuracy, no floating-point conversion errors, and an architecture that is genuinely shaped by the problem it solves. 

  Across ten chapters, you will follow full arc: the architectural decisions and tradeoffs, the numerical algorithms (addition, multiplication, CORDIC for trig, logarithms), the custom CPU design and its 12-bit instruction set, a hand-written two-pass assembler in Python, the microcode that runs on that CPU, a scripting layer for high-level key functions, and finally the physical board with its battery, display, and keyboard.

Chapter 6 (of 10):

  No general-purpose CPU has nibble-addressable memory and addressing modes designed to walk a 16-digit BCD mantissa — so this post designs one.

I like it.

foota 2 days ago |

I would love to have some real application that needs an FPGA :) Someday perhaps.

avmich 1 day ago | |

You would perhaps need to change the viewpoint for that. Theoretically, there is nothing which can't be achieved - functionally - without FPGA. However, that doesn't mean some problems can' be solved more conveniently using FPGA, and the solutions turn out better in some regards.

atultw 1 day ago | | |

Could you share some of those applications which are better solved with an FPGA? As a student I have some ideas but am interested to hear more.

mrheosuper 1 day ago | |

i was thinking of using FPGA to control led matrix, the algorithm is not hard, it's just there are too many pin to control, they need high clock rate if you want high color depth, a using MCU bit bang is not really a choice.

RicoElectrico 1 day ago | | |

And Colorlight i5/9 boards are made for that very application; it just so happens they're a reasonable minimal devboards for ECP5 FPGAs. I don't think right now you can get anything more capable for the price (Yeah, there are super cheap decommissioned miners with Zynq, but there's almost no I/O fanned out.)

i_don_t_know 1 day ago | |

Hard real-time control systems is one application.

This is a scientific BCD calculator that uses binary-coded decimals, the same internal number format HP used in its scientific calculators going back to the 1970s. It represents every decimal digit as a 4-bit nibble, which means perfect decimal accuracy, no floating-point conversion errors, and an architecture that is genuinely shaped by the problem it solves. Across ten chapters, you will follow full arc: the architectural decisions and tradeoffs, the numerical algorithms (addition, multiplication, CORDIC for trig, logarithms), the custom CPU design and its 12-bit instruction set, a hand-written two-pass assembler in Python, the microcode that runs on that CPU, a scripting layer for high-level key functions, and finally the physical board with its battery, display, and keyboard.