Electrical Engineering › Section Q · Thyristors, operational amplifiers, and power conversion › Topic Q.9
Practical op-amp limits, models, and data
The ideal op-amp rules get a design most of the way. The rest depends on how a real part falls short: small offsets, input currents, finite speed and limited common-mode range, all listed on its data sheet and all predictable with a few formulas.
Where real op-amps fall short
The ideal op-amp of Single-ended, differential, and operational amplifiers has infinite gain, no input current and zero output impedance, and with negative feedback it holds its inputs equal. Real parts depart from this in ways that matter more or less depending on the job. Their output is not exactly zero when the inputs are equal; their inputs draw tiny currents; their properties drift with temperature; their gain falls at high frequencies; their output can change only so fast; and they respond slightly to voltages both inputs share. Each imperfection has a data-sheet figure and a design remedy, and choosing an op-amp means matching those figures to the application.
Common-mode effects and input range
An op-amp, like any differential amplifier, has a small common-mode gain, measured by its CMRR as defined in Differential and instrumentation amplifiers. Values of roughly 70 dB to over 100 dB are typical, and inside a negative-feedback loop the op-amp's own common-mode error is small. A harder limit is the input common-mode range: the inputs work properly only within a stated span inside the supply rails. Outside it, some op-amps misbehave badly, with outputs that latch to one rail or even reverse phase. Never let an input go beyond either supply rail, which can happen at power-up, or when the signal comes from equipment powered separately.
Input offset voltage and drift
Short a real op-amp's inputs together and its output usually drifts to one rail. The cause is modelled as a tiny voltage source, the input offset voltage \(V_{os}\), in series with one input of an otherwise ideal op-amp; typical values range from microvolts in precision parts to a few millivolts in general-purpose ones. In a feedback amplifier it is multiplied by the noise gain \(1 + R_f/R_{in}\), the same factor for the inverting and noninverting circuits:
Many single op-amps provide offset-null pins for a trimming potentiometer. Offsets also drift with temperature, in microvolts per degree Celsius, so precision work uses low-drift parts or a stable temperature.
Worked example: output offset
An inverting amplifier has \(R_{in} = 1.0\,\text{k}\Omega\) and \(R_f = 100\,\text{k}\Omega\), a signal gain of −100, and its op-amp has \(V_{os} = 2.0\,\text{mV}\). 1. Noise gain: \(1 + 100/1.0 = 101\). 2. Output error: \(2.0\,\text{mV} \times 101 = 202\,\text{mV}\), of either sign depending on the part. 3. Compare with a 10 mV input signal, which gives 1.00 V of output: the offset is a 20% error. Check: the offset uses 101, not 100, because it acts at the op-amp's input like a signal applied to the noninverting input. Remedies are trimming, a lower-offset op-amp, or ac coupling if dc accuracy is not needed.
Input bias current
Real inputs draw small dc bias currents: nanoamps to about a microamp for bipolar inputs, picoamps for JFET and MOSFET inputs. Every input needs a dc path for its bias current to reach ground or a supply rail. A thermocouple wired across both inputs with nothing grounded gives no such path, and the circuit fails; grounding one lead fixes it. For the same reason, a split supply's centre must connect to signal ground. Bias currents also drop voltage across resistances. In the inverting and noninverting amplifiers, a compensating resistor equal to \(R_{in} \parallel R_f\) in series with the noninverting input cancels the effect of equal bias currents, leaving only their difference, the input offset current \(I_{os}\).
Worked example: bias-current error
An inverting amplifier has \(R_{in} = 10\,\text{k}\Omega\) and \(R_f = 100\,\text{k}\Omega\). Each input of its op-amp draws \(I_B = 80\,\text{nA}\), and the two differ by \(I_{os} = 20\,\text{nA}\). 1. With the + input grounded, the inverting input stays at 0 V, so its bias current must flow through \(R_f\): output error \(= (80\,\text{nA})(100\,\text{k}\Omega) = 8.0\,\text{mV}\). 2. Compensating resistor: \(R_{in} \parallel R_f = 9.09\,\text{k}\Omega\), nearest E24 value 9.1 kΩ. 3. With it, only the offset current matters: \((20\,\text{nA})(100\,\text{k}\Omega) = 2.0\,\text{mV}\). Check: compensation cuts the error by the ratio \(I_B/I_{os} = 4\).
Gain-bandwidth product
Most op-amps contain a small internal compensation capacitor that keeps them stable with feedback. Its side effect is that the open-loop gain falls at 20 dB per decade above a low corner frequency, often only a few hertz. Over that falling region, gain times frequency is nearly constant: the gain-bandwidth product, GBW. With negative feedback, the closed-loop bandwidth is about GBW divided by the noise gain: an op-amp with a 1 MHz GBW gives a gain of 10 up to about 100 kHz, or 100 up to about 10 kHz. Near this closed-loop corner the output lags: about 45° at the corner and approaching 90° a decade above. Uncompensated op-amps need an external capacitor and can oscillate without it.
Slew rate
Separately, the output can change only so fast: its slew rate, in volts per microsecond. A sine wave \(V_p\sin(2\pi ft)\) changes fastest at its zero crossings, at \(2\pi fV_p\), so the highest frequency at which the full sine can be produced without distortion is the full-power bandwidth:
An op-amp slewing at 0.5 V/µs can deliver a 10 V peak sine only up to \(0.5 \times 10^6/(2\pi \times 10) = 7.96\,\text{kHz}\). Beyond that the output turns triangular. GBW limits small signals; slew rate limits large ones, and both must be checked.
Worked example: choosing an op-amp
A microphone preamplifier needs a noninverting gain of 101 to 20 kHz, with outputs up to 1.0 V peak. Would an op-amp with a 1 MHz GBW and 0.5 V/µs slew rate do? 1. Bandwidth: \(1\,\text{MHz}/101 = 9.9\,\text{kHz}\), too low. 2. Required GBW: \(101 \times 20\,\text{kHz} = 2.02\,\text{MHz}\); with margin, choose 10 MHz or more. 3. Slew rate needed: \(2\pi \times 20{,}000 \times 1.0 = 0.126\,\text{V}/\mu\text{s}\), so 0.5 V/µs is enough. Check: here bandwidth, not slew rate, is the limit; splitting the gain over two stages of about 10 each would also work, since each then needs only about 0.2 MHz.
From vacuum tubes to data sheets
The first commercial general-purpose op-amp, from George A. Philbrick Researches in about 1952, used two twin-triode vacuum tubes and supplies of hundreds of volts. Transistor modules followed, and Fairchild's integrated 702, 709 and 741 arrived between about 1964 and 1968. Today there are hundreds of types: bipolar, JFET-input and CMOS; rail-to-rail; micropower; high-speed; high-current; and integrated instrumentation amplifiers. A data sheet's key figures are the supply range, GBW, slew rate, input bias current, offset voltage and drift, CMRR, input and output voltage ranges and output current. A model for circuit simulation captures them as an ideal op-amp plus offsets, input currents, finite gain with a falling response, and output resistance.
Common mistakes to avoid
First, multiplying the offset voltage by the signal gain of an inverting amplifier: use the noise gain \(1 + R_f/R_{in}\). Second, leaving an input with no dc path for its bias current, such as a floating thermocouple or an ungrounded supply. Third, using \(R_{in} + R_f\) for the compensating resistor: it is \(R_{in} \parallel R_f\). Fourth, dividing GBW by the wrong gain or forgetting it entirely. Fifth, checking GBW but not slew rate: a large, fast output can distort even when the small-signal bandwidth looks fine.
Key takeaways
Real op-amps have finite CMRR, a limited input common-mode range (inputs must stay inside the rails), input offset voltage multiplied by the noise gain 1 + R_f/R_in, and bias currents that need dc paths and are balanced by R_comp = R_in ∥ R_f, leaving the offset current. Parameters drift with temperature. Closed-loop bandwidth ≈ GBW ÷ noise gain, with about 45° lag at the corner. Slew rate sets the full-power bandwidth f_max = SR/(2πV_p). Op-amps evolved from 1950s tube modules to the 741 and today's specialised types; the data sheet decides.
Flashcards
Flashcard
How strongly the op-amp rejects a voltage both inputs share
Test your understanding
Question 1 of 5
A noninverting amplifier has R_f = 47 kilohms and R_g = 1.0 kilohm, and its op-amp has an input offset voltage of 1.5 mV. About how large is the output offset?
Question 2 of 5
An inverting amplifier has R_in = 22 kilohms and R_f = 220 kilohms. What compensating resistor should go in series with the noninverting input?
Question 3 of 5
An op-amp with a 3 MHz gain-bandwidth product is used as a noninverting amplifier with a gain of 50. About what is its closed-loop bandwidth?
Question 4 of 5
An op-amp has a slew rate of 0.5 V/µs. What is the highest frequency at which it can output an undistorted 5 V peak sine wave?
Question 5 of 5
A thermocouple is connected directly across an op-amp's two inputs, with neither lead connected to ground, and the circuit does not work. What is the fix?
