As AI workloads push data center architectures to the absolute breaking point, operators are battling power instability and signal degradation that lead to wildly expensive downtime. Texas Instruments used the Open Compute Project (OCP) Global Summit to tackle this head-on, debuting a slate of analog solutions built specifically to scale AI infrastructure, most notably, the industry’s first analog-to-digital converter capable of 24/7, nanosecond-resolution remote monitoring.
Here is how TI is addressing the critical bottlenecks in power and signal integrity:
- Catching Nanosecond Transients: The new ADC34DSO75 is a 3 GSPS (gigasamples per second) ADC that completely changes how operators monitor low-voltage, high-current power rails. Traditional setups routinely miss the microscopic power-rail glitches that cause servers to crash. By pairing ultra-high-speed sampling with advanced on-chip digital signal processing, this chip catches transient faults in real time. It is the first fully remote, event-based telemetry solution on the market, effectively eliminating the need to wheel an oscilloscope out to the rack for debugging. For hyperscalers, this means securing unmatched diagnostic capabilities while minimizing downtime—which can easily cost hundreds of thousands of dollars an hour.
- Taming 800V Power Delivery: With data centers aggressively scaling rack voltages, TI also rolled out the LMI800PS, the industry’s first 800V isolated hot-swap controller. Instead of a complex, five-IC discrete setup, this single high-density package handles in-rush current and power path protection while integrating an isolated bias supply and PMBus telemetry. It cuts board space by over 60% and simplifies high-voltage designs. Crucially, it secures the power bus with 5kVrms reinforced isolation and more than 8mm creepage, ensuring that a single load failure won’t cascade and take down the shared bus. It supports both ±400V and 800V distributions, making it highly adaptable for modern rack architectures.
- Locking Down Signal Integrity: As networking transmission rates race toward 3.2Tbps, the margin for timing errors effectively vanishes. Even microscopic jitter can spike bit error rates and destabilize the entire system. TI’s answer is the LMK6B, a bulk acoustic wave (BAW) oscillator that locks in an industry-leading jitter performance of just 9.6fs at 625MHz, giving next-gen AI networks the pristine signal integrity required to keep massive data streams moving without disruption.
Leadership Comment
“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 for TI’s Data, Converters and Clocks business unit. “By delivering industry-leading solutions for high-speed telemetry, advanced power protection and precision timing, TI is enabling hyperscalers and operators to design the next generation of AI data centers with optimized performance and minimal downtime.”
More details
Also at OCP, TI Power Electronics Engineer David Reusch will present “800V Data Center DC/DC Design Considerations, Topology Comparisons and Trade-Offs,” on Oct. 15 at 2:45 p.m. in room 230A.
See ti.com/ocp for more information.
For more information on the ADC34DSO75, read the technical article “The power glitches AI data centers can’t see.”
