FixedFlow Get in touch

In development · Runs in your browser

FixedFlow

Work out Q-format math in a browser tab before it goes wrong on the target.

SQ0.1516 bits · signed

Stored integer
3276 0x0CCC
Represents
0.0999755859375
Error
−2.441 × 10⁻⁵ −0.0244%
Range
−1 to 0.999969482421875
Resolution
2⁻¹⁵ = 0.000030517578125

Truncated toward zero, so the stored value is a little closer to zero than the one you typed. The error is always less than one step.

Pick a format and type a number. This converter runs in your browser and follows FixedFlow’s engine: it truncates toward zero, and it clamps anything out of range. Formats here stop at 32 bits.
Title
FixedFlow
Dwg no.
Q4-001
Rev
2026.1
Status
In development
Drawn by
DWMM Holdings

Notes

What it does.

FixedFlow is a web calculator for fixed-point (Q-format) math. It shows conversions, range, resolution and the rounding error you’ll really get. I wrote it because doing this by hand while writing DSP firmware is slow and easy to get wrong.

  1. 01

    Q-format conversion showing range, resolution and rounding error side by side

  2. 02

    A multi-variable analyzer with charts, compared against a floating-point reference

  3. 03

    Runs entirely in your browser, with no backend and no sign-up

Three tools

The converter is the first of three.

FixedFlow writes unsigned formats as Qm.n and signed ones as SQm.n, with the sign bit counted separately, so SQ7.8 is a 16-bit word. Resolution is always 2⁻ⁿ.

  1. Bit Visualizer

    Two operands and an operation (+, −, ×, ÷) in one format. It shows the sign, integer and fraction bits of each, the result in decimal, hex or binary, the error against the floating-point answer, and badges for overflow and precision loss.

  2. Format Advisor

    Model a chain of stages, each with its own format and the ideal value at that point. For each stage it checks the range, shows the quantization error and the theoretical SNR (6.02N + 1.76 dB), and suggests a format when the value doesn’t fit.

  3. Analyzer

    Write an integer expression in C style, give each variable a type from int8_t to uint32_t and a range, and it sweeps every combination against a floating-point reference. You get charts, error statistics and a CSV export. Long sweeps run in batches you can cancel.

Where it stands

Built, tested, not launched yet.

It’s a React and TypeScript app with no backend, so the math runs in your browser. There are 38 unit tests in Vitest for the engine, the Advisor and the Analyzer, and 15 Playwright tests for the pages.

  • Not launched. It hasn’t launched yet, so this site doesn’t link to it.
  • C support is partial. The Analyzer reads expressions with + - * / and parentheses. Compiling real C to WebAssembly is planned.
  • No error histogram yet. The Analyzer charts error over the sweep; the histogram is on the list.

Specification

What it’s built with, and when.

Pulled from the repository each time this site is built.

State
In development
Discipline
Hardware systems
Shape
Installed software
Started
Apr 2026
Last change
Apr 2026
Commits
13
  • TypeScript
  • React
  • Vite
  • Tailwind CSS
  • TypeScript 99%

Next

Get FixedFlow

Ask about availability, a build for your platform, or a license.

This site isn't finished. Anything on it may still change, and none of it is binding until the site is live.