High-Current E-Fuses Bring Compact Circuit Protection to Data Centers
As graphics processing units (GPUs) and other AI chips are deployed into data centers, it’s becoming critical to regulate the huge amounts of power racing into them. Sudden inrushes of current can overload the hardware or cause it to overheat, potentially resulting in system failures.
To address that, Alpha and Omega Semiconductor (AOS) introduced new e-fuses that support up to 60 A of circuit protection in a single package. The AOZ17517QI series targets the high-reliability requirements of data centers and telecom infrastructure.
In today’s data centers, all of the critical power rails are monitored and protected by a hot-swap device to prevent costly disruptions. These devices keep the main power bus from being interrupted during short circuits or other overcurrent faults.
To enhance this protection, e-fuses combine power switches, digital controllers, and current sensing in a single chip that continuously monitors the input current. If it exceeds the set limit, the device will limit the current up to its maximum rating. If the overcurrent condition lasts too long, the e-fuse disconnects the load. In that way, it plays the same role as a traditional fuse while adding resettable protection and — in many cases — integrated diagnostics.
New AOS e-fuse: Power MOSFET + High-Performance Controller
Leveraging the company's advanced co-packaging technology, the new e-fuse integrates a robust power MOSFET and a high-performance controller IC with protection features. The trench MOSFET at the heart of it comes with low on-resistance (RDS(on)) and a wide safe operating area (SOA) to prevent current racing into the system from straining the device. The power MOSFET acts as the gatekeeper for the system, isolating the load from the input bus when the transistor is turned off and relaying current to the load when turned on.
The e-fuses are designed to be deployed on the 12-V DC bus. While the 48-V DC architecture is taking over many of the world’s largest AI data centers, 12 V remains widely used for power distribution inside the rack.
“High-reliability systems such as servers and other types of telecom applications have a critical requirement for input bus protection,” explained James Wang, Power IC Marketing Manager at AOS. “Traditional protection methods have used a hot-swap controller and an external discrete FET to meet these needs. However, this approach is being increasingly dropped in favor of a more compact solution offered by an e-fuse.”
AOS said several of its 60-A e-fuses can be placed in parallel to handle higher currents and spread out the current, which is important when hot swapping hardware into the rack. In hot swapping, a new server or other electronic module is swapped into the rack while the other components keep running as usual. When the hardware hits the backplane, the sudden increase in current can overwhelm the power circuits or even the processors in the server, potentially causing components to degrade or even the entire system to fail.
By spreading out the current over several different devices, the heat can also be reduced. This is important because the heat increases the RDS(on) of the MOSFET, which impedes power delivery to the load.
AOS said it integrates accurate analog current and voltage monitoring signals to detect short circuits or other faults. The e-fuses are equipped with a wide range of protection functions to react to these fault conditions, enabling a streamlined system power-up or the ability to hot plug into the backplane. The e-fuses feature programmable overcurrent protection, short-circuit protection, undervoltage lockout, overvoltage clamp, thermal-shutdown protection, startup SOA management, and programmable soft-start.
Operating range for the e-fuses is 4.5 to 20 V, with an absolute maximum rating of 27 V. They come in a compact 5- × 5-mm QFN-32L package, with prices starting at $1.80 in 1,000-unit quantities.
