Smaller, Tougher and More Efficient: Why New Radiation Tolerant Computer Chips Matter

Space computing faces a difficult contradiction.

Satellites and spacecraft are being asked to process more data, make more decisions autonomously and support increasingly sophisticated workloads. At the same time, every additional component must operate within strict limits on physical space, electrical power, heat and weight.

Then there is radiation.

High energy particles from solar activity and galactic cosmic rays can interfere with semiconductor circuits, change stored values and cause systems to behave unpredictably. On Earth, the atmosphere and magnetic field provide substantial protection. In orbit and beyond, electronics must be designed with a much harsher environment in mind.

Researchers from Carnegie Mellon University and Sandia National Laboratories have developed a compact flip flop design intended to protect embedded FPGA configuration data against these radiation induced errors.

The work does not yet represent a complete new space processor. It targets one of the most common and important storage elements inside a chip. However, improving such a frequently repeated component could have an effect across the wider system.

Why Ordinary Computer Chips Struggle in Space

A modern computer chip contains vast numbers of transistors arranged into logic, memory and control circuits.

Among these circuits are flip flops: small storage elements that hold individual bits of information. They are fundamental building blocks in digital electronics, but they can also be vulnerable to radiation.

When an energetic particle interacts with a semiconductor, it can cause a single event upset, or SEU. This is often described as a soft error because it changes the electrical state rather than physically destroying the device.

That does not make it harmless.

If the affected bit controls a calculation, a system state or the configuration of programmable hardware, the consequences can range from corrupted data to incorrect operation. In an FPGA’s configuration memory, an upset can alter how part of the hardware itself is configured until the device is corrected or reprogrammed.

This is why space electronics cannot rely on ordinary reliability assumptions.

The Traditional Answer: Add More Redundancy

One established defense is triple modular redundancy, commonly shortened to TMR.

Instead of trusting one copy of a circuit, the system uses three. A majority voting circuit compares their outputs. If one copy produces an incorrect value, the other two can outvote it.

It is a powerful approach, but the protection is not free.

Triplicating circuitry requires additional silicon area. It can also increase power consumption, routing complexity and design overhead. These costs matter everywhere, but they become especially important inside small satellites and other systems operating under tight size, weight, power and cost constraints.

The challenge is therefore not simply to make a circuit radiation tolerant. It is to provide that tolerance without allowing the protection mechanism to consume too much of the system it is supposed to protect.

What the New Design Changes

The Carnegie Mellon and Sandia design is called a dual storage mode flip flop, or DSM flip flop.

Its key idea is reuse.

In the embedded FPGA configuration scenario examined by the researchers, configuration data is normally static after it has been loaded. Part of a conventional flip flop, the master latch, is therefore not being used in the same way during normal operation.

The DSM design reuses the master and slave latches as redundant storage elements and combines them with a C element for error correction. Instead of building three largely separate copies of the flip flop, it obtains additional resilience by making better use of circuitry that is already present.

This is an important distinction.

The researchers have not invented a completely new transistor or exotic semiconductor material. The innovation lies in how familiar components are arranged and used. The design was fabricated using a 22 nanometer FinFET process and compared with a standard D flip flop, a pulse DICE design and a TMR flip flop.

What the Radiation Testing Found

According to the DATE 2025 programme abstract, the DSM flip flop produced:

A 3,735 fold reduction in error cross section compared with the standard D flip flop.
A 455 fold reduction compared with the pulse DICE flip flop.
A physical area approximately 42% smaller than the TMR flip flop, while producing a similar error cross section.

The paper also received a DATE 2025 Best Paper Award in the conference’s test track.

Error cross section is a measure used to describe how susceptible a device is to particle induced errors under the tested conditions. A lower figure indicates that an upset is less likely, but it should not be read as a universal prediction of how often every future spacecraft will fail.

Radiation environment, shielding, circuit layout, manufacturing variation and the design of the wider system all still matter.

The public abstract also focuses on error tolerance and silicon area. Although a smaller design can support better efficiency, it does not publish a complete system level power or energy comparison. Claims about power savings should therefore remain measured until the full prototypes are evaluated.

Why Smaller Radiation Tolerant Computer Chips Matter

Reducing the area required for each protected flip flop can create several downstream advantages.

More useful logic in the same silicon area

If protection consumes less space, designers can use more of the chip for processing, memory, interfaces or other mission functions.

Lower manufacturing pressure

Smaller circuits can allow more devices to fit on a wafer and may improve the economics of producing specialised components, although final cost will depend on the complete design and manufacturing process.

Better use of limited spacecraft resources

Small satellites have strict limits on energy, cooling, volume and mass. Reducing unnecessary circuit duplication can help designers spend those resources where they create the greatest mission value.

A stronger foundation for reconfigurable systems

Embedded FPGAs allow hardware functions to be adapted for different workloads. Protecting their configuration storage more efficiently could make reconfigurable computing more practical in radiation sensitive environments.

NASA’s 2026 Small Spacecraft Technology State of the Art report describes small spacecraft as low SWaP C systems: platforms constrained by size, weight, power and cost. It also highlights the growing demand for more capable onboard processing and fault-tolerant features.

That is exactly where a more compact protection mechanism becomes interesting.

Could This Help Bring More AI Into Space?

Potentially, but this needs to be kept in perspective.

Earth observation satellites, autonomous spacecraft and communications platforms can benefit from processing data before sending it back to the ground. Local processing can reduce bandwidth use, shorten response times and allow a spacecraft to make decisions when a constant connection is unavailable.

Those capabilities may include AI inference, image filtering, anomaly detection, navigation assistance and workload acceleration.

Embedded FPGAs can be valuable in such systems because they provide hardware flexibility after manufacture. If their configuration storage can be protected with less area overhead, future designers may have more room for accelerators and mission specific logic.

However, the new DSM flip flop is not itself an AI accelerator, and the research does not suddenly make every terrestrial AI chip suitable for space. It is a building block that could make future radiation tolerant systems more capable and efficient.

That distinction matters.

The Next Test Is the Complete System

The research has already moved beyond simulation: the flip flop designs were fabricated and subjected to radiation testing.

The next challenge is integration.

Carnegie Mellon said the team was designing full system on a chip prototypes and planned to deploy the technology on a CubeSat in 2026. As of 28 July 2026, I could not find a public update confirming that this particular CubeSat deployment or its in orbit validation had taken place.

That means the technology should still be described as promising rather than flight proven.

The difference between a successful circuit level test and a reliable operational spacecraft is substantial. A complete system must account for power delivery, memory, software, interfaces, thermal behavior, packaging, multiple event effects and long duration exposure.

Resilience Is an Architectural Decision

There is also a useful lesson here beyond the space industry.

Traditional resilience often begins by adding more copies: more nodes, more paths, more storage and more failover capacity. Sometimes that is exactly the right answer. Sometimes it creates unnecessary cost and complexity.

The DSM flip flop demonstrates a different approach. The researchers examined how the circuit behaves in its real workload, identified an underused resource and repurposed it to create protection.

That is good systems engineering.

Resilience is strongest when it is designed into the architecture from the beginning, based on a clear understanding of the failure mode. It should not be treated only as an expensive layer added after the rest of the system has already been built.

Final Thoughts

This research is significant because it addresses one of the central tensions in space computing: how to increase capability without allowing reliability measures to consume too much area, energy or cost.

The most impressive part is not simply the reported improvement in radiation tolerance. It is that the researchers achieved it by using the existing flip flop structure more intelligently.

If the design continues to perform well inside complete systems and during an eventual flight test, it could help make future CubeSats, autonomous spacecraft and reconfigurable space computers more capable without sacrificing resilience.

Smaller does not automatically mean tougher.

In this case, smarter architecture may allow it to mean both.

Call to Action

Would you trust increasingly autonomous spacecraft to use more reconfigurable and AI enabled hardware, or should space systems continue to favor older, heavily proven designs?

Leave your thoughts in the comments below and follow Eagle Eye Technology for more practical analysis of emerging hardware, AI infrastructure, cybersecurity and resilient systems.

Sources