TPS53659RSBR
Product Specifications
TPS53659RSBR
Texas Instruments — Power Management & Signal Chain
Product Overview
The TPS53659RSBR 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
TPS53659RSBR: the dual channel VR13 controller for server memory rails
TPS53659RSBR is a dual channel step-down controller written to the VR13 server specification, with the two channels arranged as four phases plus one or three phases plus two. It carries non volatile memory for its configuration and a PMBus interface for telemetry and for changing settings at run time, and it drives external NexFET power stages rather than holding switches of its own. The body is a 40-pin WQFN measuring 5 mm by 5 mm with a thermal pad, and the reel row on our page is the 3000 piece large reel.
The part sits in the memory power position rather than in the core position. The core controller in the same generation carries six phases and a different VID arrangement; this one carries five phases split across two rails, which is what a memory subsystem with a main rail and a termination or input output rail needs. The distinction matters because the two controllers are not interchangeable even though they share a package family and a PMBus interface.
What follows is from SLUSD38, November 2016, and from the packaging addendum: the conversion range, the frequency window, the two output ranges the internal converter can address, the phase shedding behaviour, the telemetry the PMBus carries and the carrier rows the addendum prints for this tail.
What the order code sits on
| Function | dual channel multiphase step-down controller, driverless |
| Specification | VR13 server serial VID compliant, backward compatible with VR12.0 and VR12.5 |
| Phase arrangement | four plus one, or three plus two |
| Conversion range | 4.5 V to 17 V input |
| Output range | 0.25 V to 1.52 V, or 0.5 V to 2.8125 V for the dual channel case |
| Resolution | 8 bit converter, selectable 5 mV or 10 mV step |
| Switching frequency | 300 kHz to 1 MHz with closed loop frequency control |
| Configuration | non volatile memory, parameters stored as new defaults |
| Phase management | dynamic phase shedding with a programmable current threshold |
| Transient | fast phase adding for undershoot reduction |
| Current limit | 16 selectable levels per phase, individual phase calibration and reporting |
| Telemetry | PMBus for voltage, current, power, temperature and fault conditions |
| Package | WQFN, 40 pins, 5 mm by 5 mm, thermal pad |
| Reel | 3000 pieces on a large reel, 250 on the small reel |
| Lead finish | nickel palladium gold with a gold flash on this row |
| Moisture level | Level-2-260C-1 YEAR |
| Operating temperature | minus 40 to plus 125 degrees |
Two output ranges from one converter
The eight bit converter can be configured for 0.25 V to 1.52 V with a 5 mV or 10 mV step, or for 0.5 V to 2.8125 V when the two channels are used as a pair. The second range exists because a memory rail and its companion rail are not always at the same level: a 1.2 V main rail and a 0.6 V termination rail are both inside the same window, and one converter covering both is cheaper than two controllers. The trade is resolution, because the same number of codes spread over a wider span gives a coarser step.
The step size matters at the low end. A 5 mV step on a 0.9 V rail is a little over half a percent, which is fine for a memory rail whose tolerance is a few percent; a 10 mV step on the same rail is a little over one percent, and the difference is visible when the rail is trimmed against a specification at the top and bottom of its range. A build that has to hit a tight window uses the 5 mV setting and accepts the narrower span.
Phase shedding and why server rails care
A server memory rail spends most of its life at a fraction of its peak current, and the efficiency at 10 percent load decides the idle power of the whole chassis. Phase shedding turns phases off as the load falls, so the remaining phases run at a higher fraction of their rating where the efficiency is better. The threshold is programmable and stored in non volatile memory, which means the behaviour survives a power cycle without a host writing a register.
Adding phases back is the harder direction, because the rail has already drooped by the time the controller reacts. Fast phase adding is the feature that shortens that window: the controller detects an undershoot and brings phases online ahead of the normal loop response. The result is a transient that is a few millivolts deeper than the steady state loop would give rather than tens of millivolts.
Per phase current calibration is what makes the shedding safe. Each phase reports its own current, and the controller balances the phases using those reports, so a phase that carries more than its share is visible before it becomes a thermal problem. Sixteen levels of current limit are available per phase, and the limit is set per phase rather than for the whole rail, which matters when one phase feeds a different section of the load.
The PMBus side of the part
The PMBus interface is what separates this generation from the analogue controllers before it. Voltage, current, power, temperature and fault conditions are all readable over the bus, so a baseboard management controller can log the rail rather than infer it. Programmable parameters are written over the same bus and stored in non volatile memory, which means a factory calibration can be done once and inherited by every board afterwards.
For a contract assembly line the practical use is the test fixture. A board that reports its own rail voltage and current over PMBus can be tested without a probe on the output, and a fixture that reads the telemetry can tell a marginal stage from a good one. The same interface lets a fixture set the rail to a test voltage that the product never uses, which is how a margin test is run without changing a resistor.
The bus also brings a constraint: the PMBus lines are shared with the host, and a board that leaves the bus unterminated or that puts a long stub on it will show communication errors at high frequency. The address is set by a pin strap, so two controllers on the same bus have to be strapped to different addresses, and a build that copies the strap from a single controller board will produce a collision.
Carriers, reels and reel geometry
| TPS53659RSBR | WQFN RSB, 40 pins, 3000 pieces, large reel, active |
| TPS53659RSBT | same body, 250 pieces, small reel |
| Reel geometry | 330 mm reel, 12.4 mm width, 5.3 mm pocket, 8 mm pitch |
| Lead finish | nickel palladium gold with gold flash on the large reel row |
| Moisture level | Level-2-260C-1 YEAR, one year floor life |
| Operating temperature | minus 40 to plus 125 degrees |
| Part marking | TPS53659 across two lines |
| Body | 5 mm by 5 mm with thermal pad |
The five by five millimetre body with a thermal pad is the reason this controller can sit next to the stages it drives. The pad is the return path for the gate drive currents, and a board that connects it to a plane through a handful of vias rather than a full array will show ringing on the drive signals. The data sheet gives the recommended via pattern, and it is worth following rather than shrinking, because the failure mode is intermittent and shows up as a stage that runs hot on some boards and not others.
Where this controller fits today
The applications are VR13 server memory rails and any board with an ASIC that needs two rails, which is the case the vendor prints alongside the server memory use. A board built around a VR13 processor with a memory subsystem that needs a main rail and a companion rail is the natural fit. Boards built to a later specification are looking at a later controller, and the demand for this part is repair, spare and bridge production rather than new platform work.
When a customer brings a board that needs this part replaced, the comparison is the specification generation first, then the phase arrangement, then the output range. A core controller from the same generation will not work in this position because the phase split and the VID arrangement differ. The fields are listed on our power management cross reference page, and where the original string is no longer produced the component obsolescence management page covers how a last time buy is sized before the shelf empties.
Questions buyers ask before a first order
Does it have integrated power stages? No. It is a driverless controller and needs external NexFET stages, inductors and current sense elements.
Can both channels be used at different voltages? Yes, the dual channel arrangement covers 0.5 V to 2.8125 V with the wider range selected, or 0.25 V to 1.52 V with the finer step.
Is the configuration retained without a host? Yes. Parameters written over PMBus can be stored in non volatile memory and become the defaults on the next power up.
What is the input range? 4.5 V to 17 V, which covers a 12 V server rail with its tolerance and a 5 V intermediate rail.