Resurrecting the Z3: How Relays and Memory Recreate a Lost Computing Masterpiece ✛ ✎
Context
December 1943. Berlin. A bomb strike destroys not just a building, but the world’s first fully automatic programmable computer, the Z3. Like a painting consumed by fire, the original vanished, leaving behind only circuit sketches and fragmentary memories. But just as restorers reconstruct lost artworks from drawings, engineers and historians set out to recreate this computing masterpiece—not to make an exact copy, but to preserve its spirit and logic. These numerical simulation of the past have become a bridge between eras.
Methods
The first copy was built in 1962 by Zuse himself at his company—it served as physical evidence in a patent dispute. It used modern small relays and changed the layout. In 2001, a hybrid was created in Berlin: control theory was handled by a microcontroller, and signal processing generated clock signals, but memory and Graphs remained relay-based. The Z3r project (2010) became a teaching tool—2,500 relays (about as many as stars visible to the naked eye) were fitted with LEDs, and 60 buttons were hidden under the console for manual optimization of bit manipulation. Finally, the Swiss reconstruction of 2024 came as close as possible to the original: scarce relays were replaced with diode arrays, and the topology of the wiring was done according to historical information theory circuit diagrams, preserving electromagnetic compatibility and even the tilt of stepping switches for visual effect.
Results
Every machine reproduces the key mechanism: 22-bit floating-point words, separate arithmetic for exponent and mantissa, sequential reading of commands from punched tape. The Z3r works like a transparent clock—you can see relays switching, registers shifting, the number being normalized. The Swiss version uses electromechanical stepping switches, making the rotation of the Digital Image Processing tangible. Tests show that all reconstructions execute the same algorithms as the Zuse Z3, albeit at different speeds.
Implications
These machines are not museum mummies. They have become a testbed: historians analyze the numerical simulation logic, engineers study the fault tolerance of relay circuits. Monthly public demonstrations in Solothurn, Switzerland prove that computing power isn't measured in gigahertz—sometimes the clicks of 80-year-old technology are enough to captivate an audience.
Future development
Interest in relay computers in education is growing: slow, visually rich operation is ideal for explaining the basics of computer architecture. Perhaps the next generation of replicas will feature a tactile interface—visitors will be able to manually program using patch plugs.
Impact
Affects the preservation of technical heritage and museum pedagogy. Breaks the stereotype that early computers were exclusively an Anglo-American invention.
Next steps
Creating an open digital model of the Z3 with a web simulator. Incorporating reconstructions into university courses on computing history.
Key open problems
Not directly linked to unsolved problems in physics. However, it raises an important question: how to recover lost digital systems when source code and documentation have vanished—a problem relevant to preserving the cultural heritage of the digital age.
The knowledge machine's view
This section is addressed to the author of the paper and their colleagues. Neighbouring works were found by meaning-based search across our archive, not by word overlap. We do not review: we show the field around the work.
What the work does
The paper describes four reconstructions of Konrad Zuse's Z3 computer, from the first one built by the inventor himself in the 1960s to the Swiss reconstruction of 2024. The author traces the evolution of approaches—from patent proof to research playground where historians and engineers study the logic and fault tolerance of relay circuits.
Where its strength is
The strength of the work lies in its comparative analysis of different strategies for recreating a historical computer, each of which brings a specific facet of the original into focus: logical structure, physical behavior, or didactic potential. This turns the reconstructions into a tool for probing the engineering mindset of the past.
Where it could go next
- It might be interesting to see how relay-circuit fault-tolerance approaches resonate with modern self-correction methods in quantum computing.based on: Quantum Computers Learned to Repair Themselves arXiv:2601.20818
- A related line of work involves simulating quantum effects on programmable devices—perhaps Z3's relay logic could offer an unexpected analogy for visualizing or emulating such processes.based on: Birth of Particles in an Expanding Universe on a Quantum Computer arXiv:2605.04099
- Another direction is comparing the discrete nature of Z3's computations with contemporary efforts to digitize and model complex physical systems, for instance in digital image processing for fusion reactors.based on: Digital Vision for a Fusion Reactor arXiv:2607.04390
Worth trying
- Test Z3's relay circuits as a tangible physical model for demonstrating error correction principles similar to those studied in quantum self-correcting systems.
- Investigate whether Z3's operational logic (sequential instruction execution, separate arithmetic units) can serve as a metaphor or simplified emulator for certain quantum algorithms.
Why this may matter
Such a comparison could deepen our understanding of how ideas carry over from early electromechanical computing to modern paradigms of fault tolerance and quantum information science.
Nearby works in our archive
- Wet Water and a Thinking Processorsimilarity 0.53
- Digital Vision for a Fusion Reactorsimilarity 0.52
- Quantum Computers Learned to Repair Themselvessimilarity 0.51
- A Fault Map of the Brain in Alzheimer's Diseasesimilarity 0.51
- Quantum Magic: Nonlocal and Everlastingsimilarity 0.51
- Quantum Drones Scan Ruins with Magnetic Visionsimilarity 0.50
- Birth of Particles in an Expanding Universe on a Quantum Computersimilarity 0.50
The archive is sparse around this work: the closest review is only 0.53, against a usual 0.64. It looks like few have drilled nearby.
The original Z3 weighed about a ton but performed only 3–4 additions per second — slower than a calculator in a 1980s wristwatch. Yet it used advanced binary floating-point arithmetic that wouldn't appear in mass-market processors for decades.
In William Gibson's novel 'The Difference Engine,' an alternate history is filled with mechanical computers. The Z3 is a real embodiment of such fantasies: relay logic, punched tape, and binary code seem to have stepped right out of steampunk pages, but 40 years earlier.
Key numbers
- 2500: relays in the Z3r reconstruction
- 1941: year of original's creation
- 3-4: additions per second performed by Z3
- 22: bits in a data word