IN Brief:
- Independent customer testing measured better than 100dB channel isolation from 4GHz to 8GHz at 20mK.
- Channels on a 0.34mm pitch remained phase matched and delivered about 2dB insertion loss at 6GHz including connectors.
- RF behaviour remained stable after another cooldown, bending, and vacuum exposure.
QTREX Quantum has reported independent cryogenic testing of its additively manufactured microwave interconnect at 20 millikelvin, with adjacent-channel isolation measured at better than 100dB across the 4GHz to 8GHz band used by many superconducting-qubit control and readout systems. The unnamed quantum-computing company performed the measurements in its own cryogenic system using its own instruments, reference cables, and calibration standards, giving the result more weight than an internal room-temperature characterisation.
The tested structure used channels separated by a 0.34mm pitch across a 15cm run, with QTREX stating that no crosstalk could be detected above the test system’s noise floor. Phase matching was maintained at cryogenic temperature, while insertion loss was approximately 1dB at 1GHz and 2dB at 6GHz including the connectors used in the test arrangement.
Those parameters matter because superconducting quantum processors rely on microwave lines running from room-temperature control electronics through progressively colder stages towards the millikelvin environment around the device. Increasing qubit count can therefore increase the number of cables, connectors, thermal anchor points, and assembly operations inside a cryostat, making interconnect density and heat conduction part of the processor-scaling problem rather than a peripheral packaging issue.
Conventional coaxial cables provide well understood microwave behaviour, but every separate line occupies space and creates a conductive path towards the coldest stages, where available cooling power is extremely limited. QTREX is attempting to replace groups of individual cables with a monolithic additively manufactured structure containing multiple shielded transmission channels and shared walls between neighbouring signal paths.
The company says that geometry reduces the amount of metal needed for a given channel count and is designed to conduct roughly half the heat of an equivalent coaxial arrangement into the cooling stages, although the latest announcement is principally a microwave-performance result rather than a complete thermal benchmark. The practical value of the architecture will depend on maintaining both characteristics as channel count rises, since excellent isolation offers little advantage if the assembly introduces an unacceptable heat load into the cryostat.
Thermal cycling provides another demanding test because components repeatedly move between room temperature and millikelvin operation, creating dimensional changes that can affect conductors, dielectrics, joints, and connectors. QTREX says the RF behaviour remained unchanged after a second cooldown following a return to room temperature, while bending and vacuum exposure were followed by X-ray and optical inspection without cracks, delamination, or subsurface defects being identified.
The company compares its better-than-100dB isolation figure with a published 40dB crosstalk specification from a leading flexible cryogenic cable, although that comparison represents only one part of an interconnect decision. Engineers also have to consider insertion loss, phase behaviour, connector density, thermal conductivity, mechanical installation, serviceability, manufacturability, and the frequency range required by the particular qubit architecture.
The 4–8GHz test range nevertheless makes the result directly relevant to superconducting systems, where microwave control and readout commonly sit in that band and predictable phase and amplitude at the processor are essential. As more channels occupy the same cryogenic volume, coupling between neighbouring signal paths becomes increasingly difficult to tolerate because unwanted energy can disturb control pulses and complicate calibration across the system.
QTREX says it is finalising a binding agreement with the company that performed the tests and expects to identify the counterparty later, but the commercial discussion remains separate from the engineering milestone already reported. The more useful evidence at this stage is that the interconnect has operated at 20mK in another company’s cryogenic hardware and retained its measured behaviour across repeated cooldown and mechanical handling.
Further development will have to show whether those characteristics persist when the technology is scaled into larger assemblies and integrated with production connectors, thermal stages, processor packaging, and serviceable cryostat hardware. Additive manufacturing gives QTREX an unusual route to dense microwave geometry, but its value will ultimately be decided by whether isolation, loss, heat load, and mechanical reliability remain predictable when channel counts move beyond the test structure.
