← Lab

QUANTA

In progress · Started , updated

An open-hardware physical entropy generator with two independent sources that can be queried from the outside.

Fig. 1The QUANTA A-1 board

Drag to turn it.Light 60% · D9 21, D10 18 · K1 and K4 closed

A drawing made from screenshots of the design of the A-1 board, which has not been manufactured yet, with the parts in their relative positions. The emitter light is infrared and is painted visible here. The lids, the motion and the traces belong to the drawing and are not measurements. In the drawing the jitter of the screen trace grows with the square root of the light, as shot noise would, and the relays switch in an invented order so the matrix can be seen.

The last chapter of Cinco mil años tirando huesos ends with an instrument that does not exist yet. It draws randomness from two physical phenomena, watches itself while it does so, and lets anyone ask it from the outside how things are going. That instrument is QUANTA, and this experiment is the attempt to take it off the page and onto a circuit board.

Two noises and a watchman

A loaded die looks a lot like a fair one until somebody throws it a thousand times, which is why QUANTA is not content with handing out numbers. It draws them from two independent physical sources. The first is a differential avalanche pair, two identical branches pushed to the edge of breakdown, where the current fires off in cascades nobody can predict, and subtracting one from the other keeps that noise. The second is balanced optical detection, an LED that shares its light between two photodiodes whose signals are subtracted as well.

On top sits a gauge, a small program that audits both sources in real time, and both can be queried from the outside. Everything is open, from the hardware and firmware to the software and documentation.

The book against reality

The starting point was a seven-part technical manufacturing book, plus a master document, with the whole board designed on paper. In July 2026 I checked it against SPICE simulations, datasheets and recalculations, and it did not come out well. The enclosure it promised was the size of a flat lighter, and according to its maker’s drawing the screen alone was too wide to fit. The promised throughput could not be met under the book’s own rules, and some bandwidths fell below specification. The book’s performance figures were discarded, and since then the repository ranks what is measured first, the audit second and the book last.

Until there is a board, the closest thing to a measurement is the simulation bench. It simulates the electronics of both sources and draws, for each test point, the signal the oscilloscope should show on the day there is something to probe.

Plot of the simulated voltage on one avalanche branch over a hundred microseconds, a noise that wanders inside a shaded tolerance band.

The signal the simulation expects at TP-AVA, the test point of one avalanche branch, with its tolerance band. It is simulated, not measured. It comes from the draft in the QUANTA SPICE bench ↗ (labels in Spanish).

A board that was never paid for

The first board, the A-0, got as far as a schematic, a routed layout and a manufacturing order. It was cancelled on 19 July 2026 because of a specification error, since the OLED screen and the button had no hardware behind them, and the order was never paid. Its file is kept in the repository as an archive and as a quarry for the next one.

Top 3D view of the A-0 board, green, with dozens of labelled components, six relays in a column in the middle and three red parts at the bottom left.

3D render of the A-0, the cancelled board, as it was left in its file. It comes from its files in the repository ↗.

Its successor, the A-1, is both a validation instrument and the product. Its schematic is closed and free of electrical errors, and it is being routed right now. It has not been manufactured yet, so there is no physical entropy measurement, and QUANTA is not randomness certified by a Bell test either.

Until the board exists and has been measured, QUANTA has no throughput or size figures. The audit already showed what the ones written in advance are worth.

Log

  1. Routing the A-1

    Routing · In progress,

    With the schematic closed, the board is being routed. It has not been manufactured yet.

  2. Schematic closed

    A-1 · Done, · replaces the A-0

    The A-1 is both a validation instrument and the product. Its schematic is closed and passes the electrical check with no errors.

  3. A-0 cancelled

    A-0 · Cancelled, · cancelled

    The A-0 was cancelled because of a specification error. The OLED screen and the button had no hardware behind them, and the manufacturing order was never paid.

  4. Auditing the book against reality

    Audit · Done,

    I checked the technical book against SPICE simulations, datasheets and recalculations, and serious errors came up. There was a geometrically impossible enclosure, a promised throughput that could not be met under the book’s own rules, and bandwidths below specification. The book’s performance figures were discarded.

  5. Technical manufacturing book

    Book · Done, date pending

    The starting point was a seven-part manufacturing book, plus a master document, with the complete board design.

Links