Contact Us

TPS3850H33QDRCRQ1

Brand: Texas Instruments
Texas Instruments TPS3850H33QDRCRQ1 is a power management component designed for AI Servers & Low-Altitude Economy (Drone and UAV) and related applications. In stock at Superb Automation — request a quote for competitive pricing and fast delivery.
quote now

Product Specifications

TPS3850H33QDRCRQ1 — Power Management & Signal Chain | Superb Automation

TPS3850H33QDRCRQ1

Texas Instruments — Power Management & Signal Chain

Product Overview

The TPS3850H33QDRCRQ1 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 VoltageV
Output VoltageV
Output CurrentA
Switching FrequencykHz
Efficiency%
Quiescent CurrentµA
Topology—
Package—
Operating Temp°C
Protection—

Applications

Industrial Automation Automotive Electronics Telecom Infrastructure Server & Data Center Power Medical Devices Consumer Electronics

Request Quote — In Stock, 100% New & Original, Full Warranty

TPS3850H33QDRCRQ1: a 3.3 V automotive supervisor with a 0.8 per cent window on both rails and a window watchdog that holds a processor in reset before a runaway loop reaches memory

TPS3850H33QDRCRQ1 is a voltage supervisor that watches one rail against an upper and a lower threshold at the same time, drives RESET low when either threshold is crossed, and runs a separate watchdog window on a WDI input so a processor that stops toggling or toggles too fast is caught before it writes bad data in a VSON body of 10 pins. The DRC body is ten pins on a 3.00 mm by 3.00 mm pad with wettable flanks, and the part exists because a single threshold supervisor cannot tell a rail that sagged from a rail that climbed on a load dump.

Order code and package

Catalogue stringTPS3850H33QDRCRQ1, where the 33 is the 3.3 V monitored level and the letter before it carries the window option
BodyVSON, DRC, ten pins, 3.00 mm by 3.00 mm, wettable flanks for optical inspection of the solder joint
Reel3000 pieces on a large tape and reel on the published row
Finish and moistureNIPDAU plating, moisture level 2, 260 C, one year floor life after the bag is opened
Temperatureminus 40 to 125 C ambient, which is AEC-Q100 grade 1 rather than an industrial range with an automotive label
Marking850PB
DocumentSBVS264B, January 2017, revised September 2021
Family splitthe addendum prints the same nine monitored voltages under two letters, G and H, and the pair carries the two window widths the feature list describes, plus or minus 4 per cent and plus or minus 7 per cent

Electrical parameters that decide the application

Supply rangeVDD 1.6 V to 6.5 V, so the supervisor runs from the rail it watches on a 3.3 V or a 5 V system and from a separate housekeeping rail when the monitored node is lower
Supply current10 uA typical and 19 uA maximum over the full temperature range, which is what makes a permanently connected supervisor acceptable on a battery node
Threshold accuracyplus or minus 0.8 per cent maximum on both the overvoltage and the undervoltage threshold from minus 40 to 125 C, not a 25 C figure
Hysteresis0.5 per cent typical with a 0.2 to 0.8 per cent window, so a rail that recovers slowly does not chatter the reset line
Adjustable member0.4 V falling threshold with a 0.3968 to 0.4032 V window on the H01 part, which lets a divider set any monitor point down to 0.4 V
Power on resetRESET is guaranteed low below 0.8 V of VDD, and the outputs are undefined below that, which is the state a cold crank passes through
Undervoltage lockout1.35 V, and below it RESET is held low regardless of the sense voltage
Reset delay, factory200 ms typical with the CRST pin open, 170 to 230 ms over temperature, and 10 ms typical with a 10 kOhm pull up to VDD, 8.5 to 11.5 ms
Reset delay, capacitor700 us at 100 pF up to 3.2 s at a 10 uF capacitor, following tRST equal to 3.22 times CCRST plus 0.381 ms, with a 375 nA charge current
Watchdog windowlower boundary 22.5 ms and upper boundary 55 ms with CWD open and both SET pins low, and the ratios 1/8, 1/2 and 3/4 are selected by the SET pins
Watchdog disableSET0 high with SET1 low turns the watchdog off, which the vendor states is for development and not for a shipped unit
Fault outputWDO is a separate open drain output that asserts only while RESET is high, so a fault log can tell a supply event from a software hang
Settle time500 us is required between changes on SET0 and SET1, which a controller that toggles the pins at boot has to respect

What the letters and the digits change

TPS3850H33QDRCRQ1the 3.3 V row on this page, Active and in Production, 3000 pieces large reel, marking 850PB
TPS3850G33QDRCRQ1the same 3.3 V level under the other letter, also Active, and the letter is the only thing that separates the two windows in the addendum
TPS3850H01QDRCRQ1the adjustable member, 0.4 V falling threshold, for a rail that no fixed option covers
TPS3850H09QDRCRQ10.9 V, for a core rail on a processor that runs below a volt
TPS3850G50QDRCRQ15.0 V, for the input side of a 5 V regulator rather than its output
TPS3850H25QDRCRQ12.5 V, the level that sits under a DDR termination rail or a 2.5 V interface

Points to check before this part goes on a BOM

The window is the whole point of this part. A single threshold supervisor on a 3.3 V rail trips when the rail falls, and says nothing when a regulator fails high and the rail climbs to 4.1 V. On an automotive board the second case is the common one, because a load dump or a failed feedback resistor pushes the output up, and a processor rated to 3.6 V absolute maximum is damaged before anyone notices. The TPS3850-Q1 compares the same sense pin against an upper and a lower threshold, both at 0.8 per cent, so a 3.3 V rail with a plus or minus 4 per cent window trips at 3.168 V and 3.432 V and leaves the rest of the tolerance budget to the regulator.

The addendum prints two letters for the same voltage and does not line them up with the two window widths in the table. That is a real trap on a purchase order. A BOM that says H33 and a BOM that says G33 are two different window options, both Active, both in the same body on the same reel size, and the only way to know which one a line needs is to read the tolerance stack of the rail rather than the part description. The safe move is to name the letter on the order and to ask the vendor to confirm the window on the acknowledgement, because the marking on the body, 850PB for the H33 row, is not something a goods inwards inspector will decode against a 7 per cent sibling.

The watchdog window has a floor as well as a ceiling, and the floor catches a different fault. A processor that hangs stops toggling and a conventional watchdog catches it. A processor whose clock has run away toggles too fast for the code to be doing anything useful, and a conventional watchdog is satisfied. The window here requires a toggle between 22.5 ms and 55 ms with the CWD pin open, so both a hang and a runaway clock produce a WDO assertion. The 0.5 per cent hysteresis on the threshold does not apply to the watchdog; what matters there is the delay accuracy, and the vendor quotes the factory programmed timing at 9.5 per cent in the feature list and 15 per cent in the section that describes the table, which is worth knowing before a window is set close to the software period.

CRST timing is set three ways and the pin decides which one on every reset. An open pin gives 200 ms, a 10 kOhm pull up to VDD gives 10 ms, and a capacitor gives a delay from 700 us to 3.2 s. The state machine measures the pin for 381 us each time RESET asserts, so a capacitor that has drifted or a resistor that has gone open changes the delay without any fault being reported. On a board where the reset delay is part of a power sequence, that measurement is the thing to verify on the first article, not the nominal capacitor value.

The two outputs do different jobs and the difference matters for diagnostics. RESET is the system reset and it can be wired to a processor reset pin or to the enable of a downstream regulator. WDO only asserts while RESET is high, which means a supply event clears a pending watchdog event rather than the two competing for the same line. A board that ties both to one net loses that distinction and will report a supply brownout as a software hang in the field, which is the sort of log that sends a team looking in the wrong place for a week.

Below 0.8 V of VDD the outputs are undefined, and below 1.35 V RESET is guaranteed low. That is a narrow but important band during a cold crank, when the rail passes through both regions on the way down and on the way back up. The processor has to be held in reset through the band, which the 1.35 V lockout does, and the sense pin must not be allowed to sit above its absolute maximum while VDD is low. A series resistor on SENSE, sized against the 50 nA idle current, keeps the pin inside its ratings when the monitored rail is still alive and the supervisor supply is not.

Working the watchdog window backwards from the software

Start with the longest legitimate gap between two WDI toggles in the firmware, not the shortest. A task that toggles once per main loop gives a gap that changes with every code path, and the gap during a flash write or a long filter pass is the one that sets the floor. If the longest honest gap is 12 ms, the 22.5 ms lower boundary is comfortable and the 55 ms upper boundary leaves 43 ms of margin before a hang is reported. If the longest honest gap is 30 ms, the same window leaves 25 ms and a single interrupt storm will produce a false reset.

Then check what the software does with the fault. WDO asserts for the reset timeout period, and RESET is not asserted by a watchdog event unless the board wires WDO back to the enable of something. A unit that only logs WDO and keeps running has a watchdog that reports faults and does not act on them, which is sometimes the right choice during a diagnostic build and never the right choice on a shipped product.

Choosing the reset delay

The reset delay has to outlast the slowest rail on the board, and the 200 ms default is generous for a processor and short for a rail that comes up through a soft start of 300 ms. The capacitor option covers the gap, and the value is easy to work backwards: 3.22 times the capacitance in microfarads plus 0.381 ms. A 1 uF capacitor gives about 3.6 s, which is far more than most boards need, so 0.1 uF for roughly 700 ms is the usual working point.

Capacitor tolerance is what the vendor warns about, and the arithmetic shows why. A 10 per cent capacitor on a 700 ms delay moves the delay by 70 ms, and a 20 per cent part moves it by 140 ms, which is more than the window between two rails on a tight sequence. A COG or X7R part with 5 per cent tolerance keeps the delay inside 40 ms. The 381 us measurement time is fixed and does not scale with the capacitor.

Where this order code fits

This is the supervisor on an automotive domain controller, a telematics unit, a camera module or an on board charger, where a 3.3 V rail feeds a processor that must be held in reset through a cold crank and must be stopped when its own firmware stops responding Teams comparing it with a neighbouring part start from our Power Management Cross Reference page, and teams replacing a part that has gone end of life start from the Date Code Management page.

Questions buyers ask

Does it monitor one rail or two? One rail against two thresholds, an upper and a lower, both at 0.8 per cent accuracy. Two rails need two parts, or the H01 adjustable member with a divider on a second rail.

What does the letter before the voltage mean? It is the window option. The addendum prints the same nine voltages under G and H and both are Active, so the letter has to be named on the order and confirmed on the acknowledgement.

Can the watchdog be turned off? Yes, SET0 high with SET1 low disables it. The vendor states that this is for development, so a shipped unit should not rely on it.

How long is the reset delay out of the box? 200 ms typical with the CRST pin open, 170 to 230 ms over temperature. A 10 kOhm pull up to VDD gives 10 ms instead.

What happens below 1.6 V of supply? Below 1.35 V RESET is held low and below 0.8 V the outputs are undefined. The part is not specified below 1.6 V, so the processor is held by the lockout rather than by the sense comparator.

SUPERB holds and sources this line for production and repair demand, quoted against the exact ordering string with date codes, packing and country of origin stated on the offer. When a part is moving, the BOM Shortage Recovery page sets out the checks that keep a substitute from failing on the bench.