The inventor had worked for years in silence, hemmed in by the indistinct clamour of machines that ran too loud, too hot, too crude. Inside the walls of the Institute, he had come to dread the daily testing of gears—massive, squealing systems whose vibrations and faults were hidden beneath layers of steel and indifference. Physical experiments were expensive, slow, and inconclusive, and yet, they were demanded without exception. The cost was not just financial; it was existential.
From this mechanical chaos, the idea emerged. A method—a simulation—that would subtract the distortions, isolate the true core of mechanical interaction, and reduce the complex interplays of contact to residual compliance. No longer would engineers be enslaved to grinding prototypes and shuddering transmission benches. Instead, he proposed to simulate the contact of gears, subtract the dynamically responding eigenmodes, and model the residual bulk compliance through calculated matrices. These matrices, generated before contact interactions, would feed into simulations of quasi-static flexibility. The output would not be guesswork but clarity: meshing stiffness, transmission error, angular speed—delivered as values, charts, even videos. A vision of NVH—Noise, Vibration, Harshness—decoded and tame.
It was, he believed, a liberating invention. He filed it with the European Patent Office, his hope sealed in a sheath of claims and procedural formality.
But from the very first letter, a creeping dread returned.
The European Patent Office, in its labyrinth of corridors and deadened hallways, did not receive the invention as the inventor had hoped. The reply came from the Examining Division, unsigned but final in its language.
They found no invention.
The method, they stated, was nothing more than a simulation running on a standard computer. Claim 1, which described the steps of residual compliance calculation, simulation execution, and output generation, did not transcend the realm of abstract data processing. The use of the simulation output for assessing NVH, they claimed, was not described as computer-implemented but left to human discretion. The claim thus extended beyond what had been originally disclosed. It violated Article 123(2) EPC.
The inventor protested. The simulation was technical, he insisted. The matrices, the modeling, the output—they were not mere abstractions but tools meant to guide real-world design, to inform tangible outcomes in gearboxes and transmissions.
The Office remained unmoved. The claim, they said, was nothing more than a program running on a computer. A standard one. The application was refused.
The inventor appealed.
He submitted a new request, repositioning a former auxiliary request as the main one. It now excluded any human interpretation of NVH, and focused entirely on the computer-executed parts: modeling, simulation, and data output. He filed new auxiliary requests, tailored to rebut each procedural objection, each semantic doubt. He revised his language. He made his calculations more explicit. He sharpened his intent.
The Board of Appeal responded with a summons. Oral proceedings were to be held. Before them, however, came another letter. A communication that, in its dryness, rang with finality.
They repeated the earlier concerns. Claim 1 lacked clarity. It still extended beyond the content originally filed. And most damningly, it lacked inventive step. The method amounted, in their eyes, to nothing more than a series of calculations performed by a standard computer. The simulation results were not used for anything specific and technical; they were only information, meant to be interpreted by humans, not machines.
Nonetheless, the inventor prepared. He filed two further auxiliary requests just days before the hearing.
On the day of the oral proceedings, he entered the room like a man walking into a tribunal of ghosts.
The Board of Appeal sat high and remote. Three of them: the Chair, and two silent figures flanking him. They did not wear robes, nor did they speak at first. When they did, their voices were measured and slow.
The inventor spoke of engineering. Of saving time and weight and energy. He pointed to the matrices—how the eigenmodes were filtered, how this improved simulation speed and design accuracy. He cited paragraph 22 of his application. He showed how his method, though digital, altered the real world by changing the way gearboxes were designed, tested, and manufactured.
The Chair listened, then responded:
“The subject-matter of Claim 1, however drafted, differs from a conventional computer only in the program that it executes. Unless that program achieves a technical effect beyond normal computation, it cannot confer inventiveness.”
The inventor stood, trembling slightly. “But the simulation avoids physical testing. It speeds up real design. The output affects engineering decisions.”
“Affecting human decisions is not, in itself, a technical effect,” the Chair said. “As confirmed in G 1/19, points 97 and 98.”
The inventor sat down.
They deliberated, briefly.
The main request and first auxiliary request were rejected under Article 123(2) EPC.
The second auxiliary request, which removed the NVH-assessment step, survived that hurdle. But not the next.
Inventive step, said the Board, was still lacking. The method was a simulation, its result was information. That information might be useful, but the utility lay in human hands, not in the simulation itself. The technical effect was not inherent in the claimed method. Any technical improvement emerged only through the intervention of an engineer, who might use the output wisely—or not.
He raised one final argument. Energy consumption, heat dissipation—weren’t these technical?
“Any software running on hardware consumes energy and produces heat,” the Chair replied. “That is not a sufficient technical effect.”
He mentioned speed again. His method, he claimed, was faster than any prior art simulation.
“A mere speed comparison,” said the Chair, “is not a suitable criterion. See T 1227/05, Reasons 3.2.5.”
Finally, he challenged the COMVIK approach itself. He questioned its legality. He claimed it was backward-looking, incompatible with Article 56 EPC. That it reduced invention to semantics.
The Chair looked at him for a moment before replying.
“The Enlarged Board reaffirmed the COMVIK approach in G 1/19. We see no compelling reason to deviate.”
The decision was read aloud.
None of the admitted requests was allowable.
The appeal was dismissed.
And so the inventor walked out of the European Patent Office for the last time. Not as a bearer of innovation, but as a petitioner whose ideas dissolved under scrutiny—not for their lack of vision, but for their failure to satisfy the cold syntax of Article 56 EPC.
He did not stop inventing. But he no longer submitted applications. His simulations became internal tools, proprietary and unpublished. The company still used them, quietly optimizing gears without prototypes, refining transmissions without tests. And though his matrices saved time, money, and material, the method remained legally invisible.
Somewhere, buried in the EPO archives, was the phrase: “gaining knowledge about the designs… is in itself not a technical effect.”
He knew it by heart.
Practical Guidance for Future Cases
The story of the inventor is not singular, nor unique. The fate of this patent application offers critical lessons:
- Simulations must demonstrate a technical effect: A calculation, however complex, must lead to or control a further technical process or interaction with physical reality—not merely inform a user. Decision G 1/19 has set the boundaries.
- COMVIK remains the doctrine: No matter how philosophically contested, the COMVIK approach governs how inventive step is assessed in mixed (technical/non-technical) inventions at the EPO.
- Do not rely only on speed or internal computation: Faster does not mean inventive. Procedural optimizations must tie back to a technical use or advantage over prior methods—reflected in the claimed scope.
In the patent world Kafka might have imagined, the bureaucracy is real—but so is the escape. The key is not to argue with the architecture of the system, but to understand and out-draft it.
Based on T 1676/23 (Contact modeling/SIEMENS).