An instrumentation amp is a specialized integrated circuit that amplifies the tiny voltage difference between two input terminals while aggressively rejecting any voltage common to both inputs. If you are trying to read a 5mV signal from a sensor sitting on top of a noisy 12V power rail, this is the silicon that saves your design from returning garbage data to your microcontroller.
What Instrumentation Amps Change in a Real Circuit
When you need to measure a small differential signal, your first instinct might be to wire up a standard operational amplifier (like an LM358 or TL072) in a difference configuration. That approach works in textbook simulations, but it falls apart on the bench. A standard op-amp difference circuit has relatively low input impedance, which loads down high-impedance sensors, and its Common Mode Rejection Ratio (CMRR) relies entirely on the physical matching of four external resistors. If those resistors are off by even 0.1%, your CMRR degrades drastically, and power supply noise bleeds right into your signal.
Instrumentation amps change this dynamic entirely by using a three-op-amp internal topology. The two input pins are fed into high-impedance buffer stages (often >109 ohms), meaning they draw virtually zero current from your sensor. The third op-amp handles the differential subtraction using internal, laser-trimmed resistors. This yields a massive CMRR—often exceeding 100dB—right out of the silicon, without requiring you to hunt for 0.01% tolerance resistors.
The Math: A Worked Numeric Example
Let’s look at a classic application: reading a 50kg load cell (strain gauge) with an INA128 instrumentation amp to feed a 5V Arduino ADC.
- Load Cell Excitation: 5.0V DC
- Load Cell Sensitivity: 2 mV/V
- Full-Scale Differential Output: 5.0V × 2 mV/V = 10 mV at 50kg.
- Target ADC Voltage: 5.0V (for the Arduino's 10-bit ADC).
To map 10mV to 5V, we need a voltage gain (G) of 500. The INA128 sets gain using a single external resistor ($R_G$) connected across pins 1 and 8, governed by the datasheet formula:
G = 1 + (50,000Ω / $R_G$)
Solving for $R_G$:
500 = 1 + (50,000 / $R_G$)
499 = 50,000 / $R_G$
$R_G$ = 50,000 / 499 = 100.2Ω
In practice, you would use a standard 100Ω 1% metal film resistor. This yields an actual gain of 501. At full scale (50kg), the output will be 10mV × 501 = 5.01V, which is perfectly within the 5V ADC range (with a tiny software offset correction).
Where You Meet Instrumentation Amps in Practice
You will find these chips anywhere a physical phenomenon is converted into a tiny electrical signal that must survive a noisy environment. Think of common-mode noise like a pair of water pipes raised 10 feet in the air; the water pressure differential (your signal) inside is what matters, not the 10-foot elevation (the common-mode voltage) that affects both pipes equally.
| Part Number | Manufacturer | Typical CMRR | Key Characteristic | Approx. Price (1pc) |
|---|---|---|---|---|
| INA128 | Texas Instruments | 120 dB | Low power, general purpose, classic DIP-8 | $6.50 |
| AD620 | Analog Devices | 100 dB | Low cost, widely available, 1-resistor gain | $4.20 |
| INA333 | Texas Instruments | 110 dB | Zero-drift, micro-power, ideal for medical/thermocouples | $8.10 |
| AD8421 | Analog Devices | 114 dB | High speed (10 MHz bandwidth), low noise | $7.80 |
Source: Datasheets from Analog Devices and Texas Instruments. Pricing reflects 2026 distributor averages for SOIC/DIP packages.
Bench War Story: Setup, Numbers, and What Went Wrong
Theory is clean; the workbench is not. Here is a real-world scenario that highlights the most common way hobbyists and junior engineers brick their sensor readings.
- The Setup: We needed to measure the current draw of a 24V DC motor. We placed a 10mΩ shunt resistor on the high side (between the 24V supply and the motor). At the max motor current of 10A, the shunt drops 100mV. We wired an INA128 across the shunt, set the gain to 50 (using a 1kΩ $R_G$), expecting a 5V output at max current. We powered the INA128 from the Arduino’s 5V rail (V+ = 5V, V- = GND).
- The Numbers: The differential signal was 100mV. The common-mode voltage (the voltage present at both inputs relative to ground) was roughly 24V.
- The Outcome: The output pin of the INA128 was pegged erratically near 5V, regardless of motor speed. The ADC read a solid 1023.
- What Went Wrong: We violated the Input Common-Mode Voltage Range. The INA128 was powered by 5V, meaning its internal transistors could only handle input voltages within a few volts of its supply rails (roughly 1V to 4V). We were feeding 24V into the input pins. The internal op-amps instantly saturated and phase-inverted.
What People Commonly Confuse Them With
It is easy to misuse terminology and order the wrong silicon. Here is what instrumentation amps are not:
- Difference Amplifiers: A difference amp (like the TI AMP04 or a discrete 1-op-amp circuit) subtracts two signals but lacks the high-impedance input buffers. If you connect a difference amp to a high-impedance Wheatstone bridge, the external resistors will load the bridge, causing massive gain errors and non-linearity.
- Operational Amplifiers: An op-amp (like the ubiquitous NE5532) is an open-loop building block with massive raw gain. An instrumentation amp is a closed-loop system. You cannot wire an instrumentation amp as an integrator, a comparator, or an active filter; it is strictly designed to multiply the difference between its two inputs by a fixed gain.
- Isolation Amplifiers: While an instrumentation amp rejects common-mode voltage noise, it still shares a common ground reference with the rest of the circuit. If you need galvanic isolation to protect a microcontroller from a 400V AC mains fault, you need an isolation amplifier (like the ISO124), which uses internal capacitive or magnetic barriers to pass the signal without an electrical connection.
Frequently Asked Questions
Do I need a negative power supply (dual rails) to use an instrumentation amp?
Not necessarily, but you must understand the output swing limits. If you power an INA128 with a single 5V supply (V+ = 5V, V- = 0V), the output cannot swing all the way to 0V; it typically bottoms out around 0.8V above the negative rail. If your sensor output can be negative, or if you need to measure down to exactly 0V, you either need a negative supply rail (e.g., -5V) or you must bias your sensor's common-mode voltage up to mid-supply (2.5V) so the signal stays within the amp's linear range.
Why is my output stuck at 1.5V or 2.5V when both inputs are grounded?
You likely left the REF (Reference) pin floating. The REF pin dictates the "zero" baseline of the output. If you want 0V out when the differential input is 0V, you must tie the REF pin directly to your circuit ground. Leaving it floating allows stray electrostatic fields to bias the output stage, resulting in a phantom DC offset. For advanced applications, you can drive the REF pin with a DAC to dynamically shift the output baseline.
How do I handle RF interference (EMI) on the input wires?
Instrumentation amps have excellent low-frequency CMRR, but their CMRR degrades at high frequencies (above 10kHz). If your sensor wires act as antennas picking up local Wi-Fi or switching regulator noise, the internal transistors can rectify that RF energy into a DC offset error. Always place a low-pass RC filter directly at the input pins of the amp. A 100Ω series resistor and a 100nF X7R capacitor across the inputs will kill RF interference without affecting your low-frequency sensor data.
For deeper topology analysis and internal schematic breakdowns, the All About Circuits semiconductor textbook provides excellent open-source reference material on the three-op-amp architecture. Mastering this component bridges the gap between reading a sensor on a clean desk and deploying it in a noisy, real-world electrical cabinet.
