Multi-channel Coaxial — Core Carrier For High-Speed Signals in AI

The core demand of AI high-speed signals is "high transmission rate", and the upper limit of transmission rate depends on the bandwidth and loss characteristics of the carrier. The operating frequency band of multi-channel coaxial can cover DC to 145GHz, with mainstream products stably supporting 67GHz~110GHz. It perfectly adapts to the transmission requirements of 224G/448G SerDes technology, providing sufficient bandwidth support for AI large model training, high-definition machine vision and other scenarios. 

AI systems are mostly deployed in complex environments such as industrial workshops, data centers and outdoor scenes. External electromagnetic interference and internal channel crosstalk can easily cause signal distortion and high bit error rate, affecting the accuracy of AI decision-making. Multi-channel coaxial builds an all-round anti-interference barrier through double-layer shielding structure and channel isolation design.

On the one hand, the shielding structure of double-layer aluminum foil + braided mesh can effectively block external electromagnetic interference with a shielding effectiveness of over 90dB, avoiding external noise from intruding into the transmission link.

The core working mode of AI systems is "parallel processing". Whether it is multi-sensor data collection, multi-GPU collaborative training or multi-module signal interaction, synchronous transmission of multiple signals is required, with timing deviation controlled at the microsecond or even nanosecond level. Otherwise, data misalignment and processing delay will occur, affecting system performance.

Multi-channel coaxial achieves high consistency of each channel through strict production process control:

• Length control: the length tolerance of each channel is within ±1mm to ensure consistent signal transmission delay.
• Impedance matching: the impedance deviation of each channel is controlled within ±1Ω to avoid signal reflection and timing offset caused by sudden impedance changes.
• Phase control: through precise calibration, the phase deviation between channels is controlled within ±1ps.

Current AI equipment is developing towards miniaturization, high density and integration. Both edge AI terminals and AI computing modules in data centers put forward strict requirements on the volume, weight and compatibility of transmission carriers. Multi-channel coaxial adopts an integrated design that integrates multiple coaxial cable bundles into one. Compared with a combination of single coaxial cable bundles, its volume is reduced by more than 30% and weight by 40%, which can be easily embedded in compact AI equipment and solve the wiring problem of space-constrained devices. Ultra-thin multi-channel coaxial cables (diameter <1mm) can even be embedded in AI glasses frames to achieve stable transmission of real-time AR interactive signals.

Meanwhile, multi-channel coaxial supports a variety of standardized interfaces such as GPO, GPPO and G3PO, which can seamlessly connect with various devices such as AI chips, GPUs, sensors and test instruments without additional adaptation and transformation, reducing the integration cost of AI systems.