TPS25990ARQPR
Product Specifications
TPS25990ARQPR
Texas Instruments — Power Management & Signal Chain
Product Overview
The TPS25990ARQPR from Texas Instruments is a power management / signal chain IC delivering high efficiency, precision regulation, and robust protection. Texas Instruments (TI) is a global leader in analog ICs, power management solutions, and embedded processors.
Superb Automation partners with leading brands including Qorvo, TI, ADI, XILINX, NVIDIA, Intel, ST, MACOM, Renesas, and Coilcraft to provide complete BOM and supply-chain solutions.
Key Features
- High-efficiency power conversion with low quiescent current
- Wide input voltage range with precision output regulation
- Integrated over-current, over-temperature, and short-circuit protection
- Small footprint packaging for space-constrained designs
- AEC-Q100 qualified options available for automotive applications
- Flexible sequencing and soft-start control
Key Specifications
| Input Voltage | V |
| Output Voltage | V |
| Output Current | A |
| Switching Frequency | kHz |
| Efficiency | % |
| Quiescent Current | µA |
| Topology | — |
| Package | — |
| Operating Temp | °C |
| Protection | — |
Applications
Request Quote — In Stock, 100% New & Original, Full Warranty
TPS25990ARQPR 60 A stackable eFuse with PMBus telemetry
The TPS25990ARQPR is a 2.9 V to 16 V eFuse built for currents that no small package part can carry alone. The integrated FET has a typical 0.79 milliohm resistance, the part is rated for 50 A root mean square and 60 A peak, and several of them can be connected in parallel with active state synchronisation so that the current is shared during start-up and in steady state. The body is a 26 pin QFN of 4.5 mm by 5 mm.
The feature that changes how the part is used is the PMBus interface. A host can read input voltage, output voltage, input current, input power, temperature and fault status, reconfigure the protection thresholds, and command a power cycle without touching a pin. Configuration can be stored in on chip non-volatile memory, and a blackbox recorder keeps the last events with relative time stamps. For a board that is expected to report why it shut down, that is a different class of component from a fuse with a flag pin.
Key parameters taken from the TPS25990 document
| Input range | 2.9 V to 16 V operating, 20 V absolute maximum, minus 1 V withstand at the output |
| On resistance | 0.79 milliohm typical for the integrated FET |
| Current rating | 50 A root mean square and 60 A peak for one device |
| Scaling | Parallel connection of several devices with active state synchronisation and current sharing |
| Overcurrent threshold | Programmable from 7 A to 50 A with five per cent maximum accuracy |
| Threshold setting | Set by the IMON resistor and an internal reference of 1 V, 0.5 V, 0.3 V or 0.1 V, so 1.1 k with 1 V is the 50 A point and 1.1 k with 0.3 V the 7 A point |
| Response | Circuit breaker in steady state with a programmable blanking timer, active current limit during start-up |
| Fast trip | 280 ns typical on a short |
| Telemetry | PMBus interface for input and output voltage, input current, power, temperature and fault status |
| Configuration | On chip non-volatile memory, plus blackbox recording of several events with time stamps |
| Energy | Assured FET safe operating area of 8 watt seconds |
| Supply | 10 V to 16 V on VDD for in-system non-volatile memory programming, 4.5 V to 16 V without it |
| Package | VQFN-HR RQP, 26 pins, 4.5 mm by 5 mm |
| Carrier | 3000 pieces on a large reel, moisture level 2 at 260 degrees with a one year floor life |
| Marking and grade | T25990, minus 40 to plus 125 degrees junction |
The table is read from the vendor document. The supply row is the one that catches a first build: the part needs 10 V on VDD to program its non-volatile memory in system, so a 5 V rail that powers the part in normal operation is not enough to store a configuration.
Where the numbers matter in a real build
Seventy nine hundredths of a milliohm at 50 A is about 2 W, and that number is the reason the part exists. A discrete solution with two 1.5 milliohm FETs, a gate drive, a sense resistor and a controller occupies far more area and adds the sense resistor loss on top. The 26 pin body spreads the current across many pins, and the document notes the 29 mil separation between the power pins and ground, which is a creepage figure for a 16 V rail that has to pass a safety review.
Paralleling is not simply tying the inputs and outputs together. The devices synchronise their operating state so that all of them enter the inrush phase at the same time and share the current, which avoids the case where the first device to turn on carries the whole load and shuts down while the others are still off. The primary device acts as the controller and the others follow, which means the configuration in non-volatile memory has to be the same in every device on the rail, and a mixed batch of configuration images is a failure mode that shows up only at start-up.
The overcurrent threshold is set by a resistor on the IMON pin and an internal reference, with four reference levels available. With a 1.1 k resistor and the 1 V reference the threshold is about 50 A, and with the 0.3 V reference it is about 7 A on the same resistor. That gives a system builder a 7 to 50 A span without changing the resistor, but it also means a configuration that reads the wrong reference is a factor of three away from the intended limit. The five per cent maximum accuracy is the number to use for the protection margin.
The blanking timer is programmable through the interface rather than by a capacitor, so a rail can allow a peak above the threshold for a defined window and then break the circuit. Combined with the 280 ns fast trip, that covers both a load that surges and a load that shorts. The assured safe operating area of 8 watt seconds is the figure that bounds how long the FET can be in the linear region during a fault, and it is the number a thermal argument should quote rather than the steady state dissipation.
The telemetry is what makes this part useful beyond protection. Reading input current on a rail lets a board report its own load without a separate shunt, and reading temperature next to the FET gives an early warning before a thermal shutdown. The blackbox recorder keeps the sequence of events, which matters when a unit comes back from the field with no fault code on the host.
- The configuration image in non-volatile memory, which has to be identical across every device on a parallel rail.
- The VDD rail during programming, since 10 V is needed to store a configuration in system.
- The reference level and the IMON resistor together, because the threshold depends on both.
- The blanking window against the real peak load, not against the typical load.
- The safe operating area figure for any fault that leaves the FET in the linear region.
- The creepage and clearance of the power pins on the board, since the rail is a 16 V high current path.
Which projects this part fits
The vendor lists servers and high performance computing, network interface cards, graphics and accelerator cards, datacenter switches and routers, input hot swap, and fan trays. The combination of 60 A, telemetry and a small body points at accelerator cards and switch line cards, where a rail has to be protected, measured and reported without a discrete shunt and a controller.
Questions that come up before a purchase order is released
How many devices can be paralleled?
The document describes a primary controller with additional devices following it, with active state synchronisation for start-up and steady state. The practical number is set by the current sharing accuracy and the board area rather than by a hard limit in the electrical table.
Does the part need a microcontroller?
It works as a standalone eFuse with threshold settings from its pins and memory, but the telemetry and the reconfiguration need a PMBus host. A board without one gets protection and a flag rather than a full register map.
Can the thresholds be changed after assembly?
Yes, over PMBus, and the setting can be stored in non-volatile memory. Programming that memory in system needs 10 V on the VDD rail.
What happens on a short circuit?
The fast trip path responds in about 280 ns, and the assured safe operating area bounds the energy the FET can absorb during a slower fault.
Is the current measurement accurate enough to replace a shunt?
The document gives a five per cent maximum accuracy on the overcurrent threshold, and the analog monitor pin is specified separately. For a rail where the reading drives a protection decision that is enough; for a metering function the figure has to be checked against the system budget.
How SUPERB handles this line
Parts like this are bought against a rail budget and a firmware image, so the sourcing work is in matching the configuration and the reel rather than in the part number alone. Our power management cross reference carries the family mapping, our last time buy and long term planning page covers the buffer we hold on a part that goes into a platform rather than a product, and our supplier quality audit page covers how a reel is traced back to its lot before it reaches a line.