When setting the gain of a difference amplifier, the integrated instrumentation amplifier (In-Amp) wins for precision DC and low-frequency sensor interfaces (like strain gauges, load cells, and thermocouples) because it completely decouples gain setting from Common-Mode Rejection Ratio (CMRR). The discrete single-op-amp difference amplifier wins for high-frequency signal routing, simple level shifting, and cost-sensitive consumer electronics where source impedances are low and perfectly matched. You cannot swap a discrete design into a high-impedance biomedical sensor front-end without destroying your signal-to-noise ratio, just as you shouldn't waste a $7 AD8221 In-Amp on a simple 12V-to-3.3V audio level shifter.
The Single Physical Difference: Input Buffering Topology
The single physical difference that drives every other performance metric between these two circuits is the input stage topology.
A standard discrete difference amplifier uses a single operational amplifier with a four-resistor network. The input signals are fed directly into the op-amp through resistive voltage dividers. This means the circuit's input impedance is relatively low (dictated by the series resistors) and, critically, the input stage is unbalanced. The non-inverting input sees a high-pass filtered path to ground, while the inverting input sees a virtual ground.
An integrated instrumentation amplifier (like the classic 3-op-amp topology) places two high-impedance buffering op-amps at the inputs before the difference stage. The gain is set by a single external resistor ($R_G$) between the inverting nodes of these input buffers. The final difference stage is internally laser-trimmed to unity gain. This physical separation of the gain stage and the difference stage is what allows In-Amps to maintain massive CMRR even at high gains.
Real-World Component Specification Sheet
Here is how the topology difference translates to hard datasheet numbers across common industry parts. Note how the discrete LM358 requires external precision resistors to even approach the baseline CMRR of the integrated parts.
| Parameter | Discrete (LM358 + 0.1% Resistors) | Integrated (TI INA128P) | Integrated (ADI AD8221BRZ) |
|---|---|---|---|
| Topology | 1 Op-Amp, 4 External Resistors | 3 Internal Op-Amps | 3 Internal Op-Amps |
| Gain Equation | $G = R_f / R_{in}$ | $G = 1 + (49.4k\Omega / R_G)$ | $G = 1 + (49.4k\Omega / R_G)$ |
| Input Impedance | $\approx 10k\Omega$ to $100k\Omega$ | $10^{10} \Omega$ (Differential) | $10^{12} \Omega$ (Differential) |
| CMRR (at G=10) | $\approx 60$ dB (Resistor limited) | $106$ dB | $120$ dB |
| Input Bias Current | $\approx 20$ nA (Bipolar) | $\approx 2$ nA | $\approx 0.5$ nA |
| Approx. Cost (2026) | $0.25 (Total BOM) | $4.50 per IC | $7.20 per IC |
The Math: How Gain is Set and the CMRR Trap
To understand why the gain of a difference amplifier behaves so differently in these two topologies, we have to look at the math and the CMRR tolerance trap.
Discrete Difference Amplifier Gain
For a discrete circuit, the output voltage is defined by the ratio of the feedback resistor to the input resistor:
$V_{out} = \frac{R_2}{R_1} (V_2 - V_1)$
If you want a gain of 10, you might choose $R_1 = 10k\Omega$ and $R_2 = 100k\Omega$. However, for the circuit to reject common-mode noise (like 60Hz mains hum), the ratio of $R_2/R_1$ must perfectly match the ratio of the non-inverting divider $R_4/R_3$.
The Trap: If you use standard 1% tolerance resistors, the worst-case mismatch drops your CMRR to a dismal 48 dB. Even if you upgrade to expensive 0.1% precision resistors, your CMRR is physically capped at roughly 68 dB at a gain of 10. Furthermore, if you need to change the gain later, you must swap two resistors while maintaining perfect matching, or you will destroy the circuit's noise rejection. (For a deeper look at resistor mismatch impacts, see this guide on instrumentation amplifier basics).
Instrumentation Amplifier Gain
An In-Amp shifts the gain equation to the input buffers, leaving the difference stage at a fixed, laser-trimmed unity gain:
$V_{out} = \left(1 + \frac{49.4k\Omega}{R_G}\right) (V_2 - V_1)$
To get a gain of 10 on an INA128, you simply place an $R_G$ resistor of $5.49k\Omega$ between pins 1 and 8. Because the internal difference stage resistors are laser-trimmed on the silicon die to 0.01% matching, the CMRR remains at a massive 106 dB regardless of what external $R_G$ you use. You can change the gain with a single, standard-tolerance resistor without affecting noise rejection.
Component Comparison Matrix
When deciding which topology to use for your PCB or breadboard, evaluate your design against these four concrete criteria.
| Criteria | Discrete Op-Amp Difference Amp | Integrated Instrumentation Amp |
|---|---|---|
| Source Impedance Handling | Poor. Requires low-impedance sources ($<1k\Omega$) to prevent loading errors and gain attenuation. | Excellent. High-impedance inputs ($>1G\Omega$) do not load down fragile sensors like piezoelectrics or pH probes. |
| Gain Adjustability | Difficult. Requires swapping matched resistor pairs; altering gain ruins CMRR if not perfectly matched. | Trivial. Single $R_G$ resistor sets gain; CMRR is entirely independent of the external gain resistor. |
| Bandwidth / Slew Rate | High. Limited only by the chosen op-amp (e.g., an OPA2188 can handle MHz signals easily). | Moderate to Low. Internal compensation limits most In-Amps to a few hundred kHz (e.g., INA128 BW = 1.3MHz at G=1). |
| PCB Footprint & BOM | Large. Requires 1x SOIC-8 op-amp + 4x 0603 resistors + decoupling caps. | Small. Single SOIC-8 or MSOP-8 IC + 1x $R_G$ resistor + decoupling caps. |
Where They Are NOT Interchangeable & Cost Realities
While both circuits subtract $V_1$ from $V_2$ and multiply by a gain factor, they are absolutely not interchangeable in edge-case applications.
Do not use a discrete difference amp for Wheatstone bridges. A typical strain gauge load cell outputs a differential signal in the millivolt range, superimposed on a 2.5V or 5V common-mode voltage. The source impedance of the bridge is often $350\Omega$ to $1k\Omega$. If you use a discrete difference amp with $10k\Omega$ input resistors, you will load the bridge, introduce severe gain errors, and the 48 dB CMRR from 1% resistors will allow the 5V common-mode voltage to leak through and saturate your ADC. You must use an In-Amp here.
Do not use an In-Amp for high-frequency RF or audio level shifting. If you are shifting a $\pm 10V$ audio signal down to a $0-3.3V$ single-ended ADC range at 100kHz, an AD8221 ($7.20) is a massive waste of budget and will likely struggle with the required slew rate. A discrete circuit using a $0.15 TLV9002 and four $0.02 resistors will handle the bandwidth easily and cost pennies.
The Final Decision Framework
- Choose Discrete When: You are designing a cost-sensitive consumer product, dealing with audio or high-frequency signals ($>500$ kHz), performing simple voltage level translation, and your source impedance is strictly below $1k\Omega$.
- Choose In-Amp When: You are interfacing with DC or low-frequency precision sensors (thermocouples, RTDs, strain gauges, ECG/EKG biopotentials), your source impedance is high or unbalanced, and you need a CMRR greater than 80 dB to reject industrial mains noise.
For a comprehensive breakdown of how Analog Devices engineers approach the CMRR limitations in modern sensor designs, refer to their application notes on de-mystifying instrumentation amplifiers. Ultimately, the gain of a difference amplifier is not just a multiplier—it is a trade-off between precision, bandwidth, and BOM cost. Choose the topology that protects your signal's weakest link.
