Texas Instruments has introduced an analog-to-digital converter designed to detect power supply faults at nanosecond resolution, as data centre operators seek to improve reliability amid growing demand from artificial intelligence workloads.
The ADC34DSO75 was unveiled at the Open Compute Project (OCP) Global Summit this week. The 3-gigasample-per-second device combines high-speed sampling with on-chip digital signal processing to detect transient faults in low-voltage, high-current power rails.
TI said the chip enables continuous, low-power remote monitoring, potentially identifying brief power fluctuations that conventional monitoring systems miss. Such glitches can cause unplanned downtime, which the company estimates can cost data centre operators hundreds of thousands of dollars an hour.
The device is designed to capture fault events as they occur, reducing the need for engineers to investigate problems on site using oscilloscopes. TI said its approach offers remote diagnostic capabilities in a smaller and less expensive package than conventional monitoring methods.
Power stability has become an increasingly important challenge as data centres accommodate more demanding AI workloads. Failures in power delivery and signal integrity can disrupt operations, while the need for higher computing density puts additional pressure on infrastructure design.
TI also announced two components aimed at improving power delivery and signal integrity in AI data centres.
The LMI800PS is an 800V isolated hot-swap controller designed to protect power systems during load insertion and isolate faults. It integrates functions that would otherwise require five separate integrated circuits, including an isolated bias supply and PMBus telemetry. TI said the integration reduces board area by more than 60%.
The controller supports both ±400V and 800V power distribution architectures. Its reinforced isolation and fault protection are intended to prevent a failure in one load from disrupting a shared power bus.
The third component, the LMK6B, is a bulk acoustic wave-based oscillator designed to maintain timing accuracy in high-speed data transmission. TI said it delivers jitter of 9.6 femtoseconds at 625MHz, addressing the tighter timing tolerances associated with networking systems operating at transmission rates approaching 3.2 terabits per second.
“As AI workloads drive massive scale, the need for real-time visibility and architectural stability has never been greater,” said Karthik Vasanth, Vice-President and General Manager of TI’s Data, Converters and Clocks business unit.
