INA128U
Product Specifications
INA128U
Texas Instruments — Power Management & Signal Chain
Product Overview
The INA128U 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
INA128U: the plain grade eight pin SOIC in a 75 piece tube
The INA128U is the plain grade of the eight pin SOIC body in a 75 piece tube, the carrier that sits between a bench prototype and a production reel. The body measures 4.9 mm by 6 mm and runs on the reflow line with the rest of a modern board, and the tube holds enough parts for a few hundred channels with spares. It is the code most often quoted for a first build of a new instrument.
The part itself is a three op amp instrumentation amplifier with a current feedback input stage: 50 µV maximum offset, 0.5 µV per degree drift, 5 nA of bias current, 8 nV per root hertz of voltage noise, 120 dB of common mode rejection, 700 µA of quiescent current and a gain range from 1 to 10,000 set by one resistor with G = 1 + 50 kΩ divided by RG. The bandwidth is 1.3 MHz at unity gain and 200 kHz at a gain of 100, which is the property that separates this family from a general purpose op amp front end.
The INA128U rows as the addendum publishes them
| Part class | Precision instrumentation amplifier, three op amp, current feedback |
| Vendor status | Active, production |
| Data sheet | SBOS051G, revised January 2026 |
| Grade | Plain, the family offset and gain error rows |
| Gain equation | G = 1 + 50 kΩ divided by RG |
| Gain range | 1 to 10,000 V/V |
| Input offset voltage | 50 µV maximum |
| Offset drift | 0.5 µV per degree maximum |
| Input bias current | 5 nA maximum on the standard rows |
| Voltage noise | 8 nV per root hertz at 1 kHz, 0.2 µV peak to peak from 0.1 to 10 Hz |
| Common mode rejection | 120 dB minimum |
| Bandwidth | 1.3 MHz at a gain of 1, 200 kHz at a gain of 100 |
| Input protection | Plus or minus 40 V without damage |
| Supply range | Plus or minus 2.25 V to plus or minus 18 V |
| Quiescent current | 700 µA typical |
| Body | SOIC, D outline, 8 pins, 4.9 by 6 mm |
| Carrier | 75 pieces per tube |
| Moisture level | Level 2, 260 degrees, one year floor life |
50 µV of offset and half a microvolt per degree
The offset row for this family is 50 µV maximum with 0.5 µV per degree of drift, and the pair is what separates it from a general purpose op amp front end. Over a 40 degree excursion the offset moves 20 µV, which on a 10 mV full scale bridge is 0.2 per cent, so a channel built on INA128U can hold its accuracy across the cabinet temperature without a heater or a chopper. The same excursion on a 10 µV per degree part moves 400 µV, which is 4 per cent of that full scale and larger than any plausible bridge nonlinearity.
The offset is quoted at a gain of 1 and divided by the gain at higher settings in the vendor table, which matters when a channel runs at a gain of 100: the input referred offset falls to a fraction of a microvolt while the gain error does not fall at all. That asymmetry is why the accuracy budget at high gain is set by the resistor rather than by the amplifier, and why a 0.1 per cent resistor is worth more than the next grade of amplifier.
Current feedback input stage and bandwidth at a gain of 100
The INA128 and INA129 use a current feedback input stage, and the reason is visible in one number: 200 kHz of bandwidth at a gain of 100. A voltage feedback instrumentation amplifier loses bandwidth in proportion to gain, so a part with 1.3 MHz at unity gain is down to a few kilohertz at a gain of 100. The current feedback topology holds the curve much flatter, which is what lets INA128U serve a multiplexed data acquisition front end at a gain that would otherwise be too slow to scan.
Current feedback has a cost that the data sheet states plainly: the feedback impedance has to be resistive and short. The gain resistor should sit within a few millimetres of the RG pins and should be a metal film part, because inductance in that path shows up as peaking in the step response rather than as a gain error. At a gain of 1000 the resistor is around 50 Ω, where a centimetre of trace is a measurable fraction of the impedance, so the placement advice is not decorative.
The D body at 4.9 by 6 mm against the P body at 9.8 by 9.4 mm
The family ships in two bodies, an eight pin SOIC measuring 4.9 mm by 6 mm and an eight pin plastic DIP measuring 9.81 mm by 9.43 mm, and the difference is a process decision rather than an electrical one. The D body runs on the reflow line with the rest of the board and is the default for a new build; the P body needs plated through holes, a separate insertion step and either wave or selective solder, and on a board with a handful of DIP parts that second process usually costs more than the part.
The reason to keep the P body is mechanical. A socketed DIP can be replaced in the field or swapped during calibration without a rework station, and the longer leads absorb a mismatch between the board and a connector. A probe amplifier that is trimmed on the bench is a genuine case for the socket. On INA128U the die and the pinout are the same in both bodies, so a board can move from one to the other without an electrical change, and the footprint is the only thing that has to be revisited.
8 nV per root hertz of voltage noise against 0.17 pA of current noise
The voltage noise row is 8 nV per root hertz at 1 kHz and the current noise is 0.17 pA per root hertz on the standard rows, which places the crossover source resistance at about 47 kΩ. Below that the voltage noise dominates and the amplifier is as quiet as the source; above it the current noise dominates and a FET input part would be quieter. On a 350 Ω bridge the voltage noise is the whole story and 8 nV per root hertz over a 10 Hz bandwidth is 25 nV, which against a 10 mV full scale is 2.5 parts per million.
The 0.1 to 10 Hz noise is quoted as 0.2 µV peak to peak, which is the figure that matters for a slow channel: a 10 mV full scale signal with 0.2 µV of low frequency noise has a resolution floor of 20 parts per million, better than most 16 bit converters can use. A channel that needs more resolution than that is limited by the sensor, and the amplifier stops being the item to change. On INA128U the noise is flat enough that a higher gain does not improve the input referred figure below the crossover.
120 dB of common mode rejection and the source imbalance that removes it
The common mode rejection row is 120 dB minimum on the family, measured with a 1 kΩ source imbalance, and it is the number that decides whether a bridge channel can reject a 50 Hz disturbance. Rejection degrades with source imbalance at roughly 1 dB per 100 Ω above the test condition, so a channel with a 10 kΩ mismatch between the two input legs has thrown away most of the 120 dB. In practice that mismatch comes from the sensor rather than from the board: a 350 Ω bridge with one leg 5 Ω high, or a thermocouple with a 2 kΩ cable resistance on one side only.
The fix is a pair of matched resistors at the input, or a buffer for each leg, or a sensor with a trimmed output impedance. The REF pin is part of the same problem: the vendor notes that 8 Ω in series with REF is enough to degrade rejection on a typical part, so the reference has to be driven from a low impedance source such as an amplifier or a shunt reference with a capacitor, not from a divider alone. On INA128U the rejection also falls with frequency, and the vendor curve of rejection against frequency is the one to read for a switching supply disturbance rather than the dc row.
Plus or minus 40 V of input protection and the bias current return path
The inputs survive 40 V beyond the rails without damage, which covers the ordinary fault of a transducer cable being plugged in while the rack is powered, and it covers a miswired bridge. It does not cover a sustained fault at the supply pin, and it does not protect the source: a 40 V fault drives current through the protection network and into whatever is driving the amplifier, so a series resistor at the connector is still worth the board space. The protection also stops being transparent above roughly a megohm of source resistance.
The bias current return path is the second half of the input requirement. Each input needs a dc path to a defined potential, and a floating source such as a transformer or a capacitive sensor will let the inputs drift until the amplifier saturates. The vendor shows the ordinary arrangements, and the one to avoid on INA128U is a path through a high value resistor on one leg only, because that converts the bias current into a differential offset. Matched paths on both legs, or a path to the reference rather than to ground, keep the error common mode.
Buying INA128U in the tube
Seventy five pieces is the right quantity for a first build and the wrong quantity for a production run, and the crossover is around a few hundred boards. Above that the reel rows of the same die are cheaper per piece and faster to feed, and the only change on the line is the carrier setting. Because the die and the body are identical, a programme can move from tube to reel without a qualification run.
The plain grade is the right bin when the channel has a trim or the full scale is large. On a 10 mV full scale the 50 µV offset is half a per cent of range, which is small enough to leave alone; on a 1 mV full scale it is five per cent and needs the A grade or a trim. Our amplifier cross reference page lists the grades and the near equivalents we check, and the traceability report page covers what is recorded against each tube.
Questions that arrive with this part number
Whether the U and the P codes are interchangeable is the first question and they are electrically, with the same die, the same pinout and the same data sheet, and they are not mechanically, because the SOIC footprint and the DIP footprint are different land patterns. The second question is whether the 75 piece tube can be split across builds, and it can, with the level 2 floor life giving a year of storage once the tube is opened inside a dry cabinet.
The third is about the reference. On a bipolar supply the REF pin goes to the analogue ground the converter uses, and on a single supply it goes to a buffered mid rail. A divider alone has a source impedance far above the 8 Ω the vendor names as the point where rejection starts to degrade, so the reference is one of the few places where a cheap resistor pair is a real mistake.