Computer Engineering 15

○ All amplifiers have limited frequency response.

■ The gain of DC-coupled amplifiers roll-off at high frequencies due to
unwanted parasitic capacitors (on every circuit node!)
■ The gain of AC-coupled amplifiers roll off at high frequencies, but also at
low frequencies due to coupling and bypass capacitors.

Operational Amplifiers
Outline
● Why use Operational Amplifiers? (“opamps”)
○ The wonderful effects of using negative feedback.
● Characteristics of ideal opamps
● Basic amplifiers using opamps
○ Inverting and non-inverting amplifiers, unity-gain buffers
○ Summing amplifiers
● Difference & Instrumentation amplifiers
● Integrators, Differentiators, and Active Filters
● Non-ideal effects in real integrated circuit opamps
○ Finite DC gain, offset voltage, input bias and offset currents
○ Output voltage and current limits
○ Finite bandwidth, slew rate, and full-power bandwidth
● Summary of key concepts

Why use Operational Amplifiers

● Opamps are a key building block in feedback systems.
● Negative feedback makes everything better.
○ More accurate amplifier gain (closed-loop)
○ Higher or lower input & output resistance (whichever is needed)
○ Higher amplifier -3dB bandwidth (closed-loop)
● But, we have to pay for this.
 Computer Engineering 16

○ Negative feedback trades high open-loop amplifier gain for everything else we
need.
● So, operational amplifiers are designed to have very high open-loop voltage gains of
100,00 (100 dB) or more, with the intention of using them in closed-loop feedback
circuits which trade away the extra gain for other desirable performance characteristics.

Key Concepts for Negative Feedback

●
● Negative feedback systems sample the output signal of an amplifier and feedback a
fraction of the output to the input, where it is subtracted from the source signal.
○ A = Xo / Xi = gain of the forward amplifier (the “open-loop” gain)
○ Β = Xf / Xo = the gain of the feedback path (the “feedback factor”) = the fraction of
the output signal fed back to the input
○ Af = Xo / Xs = the gain of the feedback amplifier (the “closed-loop” gain)
● Negative feedback makes the closed-loop gain feedback insensitive to changes in the
open-loop gain.
● Even if A varies with the process, voltage and temperature (PVT variations), the gain
with feedback stays ~ constant.
○ - β is usually set by ration of resistors, which stays ~ constant.

Opamp Symbols

●
 Computer Engineering 17

● Opamps require + and - power supplies to operate, but are often shown without these
explicitly drawn to keep schematics simple and easy to understand.
○ Everyone knows power supplies are required, so why draw them?
○ The purpose of a good schematic is to effectively communicate information, so
reduce “clutter” to make them easier to read.

Characteristics of Ideal Opamps

●
● Opamps always have both a + and - input to make it easy to apply negative feedback
around them.
● The output voltage, Vo = A(v2-v1) where v2 = v+, v1 = v.
● Outputs can be single-ended (measured with respect to ground) or differential with Vop,
Von (Vodm = Vop - Von)
● Ideal opamps have:
○ Infinite gain (A = ∞) → there is 0 V between the + and - inputs
○ Infinite bandwidth (f-3dB = ∞)
○ Infinite input resistance (Ri = ∞) → i = 0 into the + and - inputs
○ Zero output resistance (no signal is lost, even for small loads)
○ Zero gain for common-mode inputs (only differences are amplified)

Differential & Common-mode signals

● Opamps are designed to amplify the difference between the + and - inputs, but NOT
amplify any voltage common to both inputs (like noise)
● This is called common-mode rejection
● It is often useful to redraw the input signals as the combination of a differential signal
and a common-mode signal.
 Computer Engineering 18

Inverting Amplifier

●
 Computer Engineering 19

● Using the Ideal Opamp assumptions that the:

○ Voltage between the + and - input terminals = 0 (since A = ∞)
○ Input currents = 0 (since Ri = ∞)
■ Nodal analysis shows that voltage gain is:

○ The - input is a “virtual ground”

Non-Inverting Amplifier

●
● Using the Ideal Opamp assumptions that the:
○ Voltage between the + and - input terminals = 0 (since A = ∞)
○ Input currents = 0 (since Ri = ∞)
■ Nodal analysis shows that voltage gain is:

○ Here the voltages at both the + and - inputs = vi since A = ∞ causes vid = 0 V.
 Computer Engineering 20

Non-inverting Amp as a Feedback Amp

●
● Feedback is provided by the R1-R2 voltage divider circuit:

●
● Which is the same gain found using ideal opamp analysis!
○ This is a more exact equation and can be used to find Af when A ≠ ∞

Unity-gain Buffers

● For a non-inverting amp:

● So if R2=0 and R1 = ∞ (R1 is an open circuit) then:

● Unity-gain buffers are a special case of the non-inverting amp, with a voltage gain = 1
 Computer Engineering 21

Inverting Summing Amplifiers

●
● Additional inputs can be added to an inverting amp to create a Weighted Summing
Amplifier
○ Vo is the sum of all the inputs, each with its own separate gain.
○ All input currents combine at the summing node and flow into Rf

Summing Amps with + and - inputs

●
● What if we need gains that are not just negative? Answer: Use two stages, take 2
summing amps and run its output through the input on a second one.
● Multiple amplifier stages can also be cascaded to provide for both positive and negative
gains.
○ Note that the output signal from the 1st stage sees an additional gain of -Rc / Rb
as it passes through the 2nd stage, resulting from an overall positive gain.
○ This does require multiple opamps → higher cost, power, etc.

Non-Inverting Summing Amplifiers

● Another way to do the previous topic, however, is not as flexible but often used.
 Computer Engineering 22

●
● Non-Inverting summing amps can also be built by adding multiple resistors to the +
opamp input, but since this node isn’t a virtual ground the gains for v1, v2, etc., are not
independent of each other.
○ Using superposition:

Difference Amplifiers

●
○ Difference amplifiers are used to amplify the difference between 2 inputs, and not
amplify the common-mode

○
 Computer Engineering 23

Common-mode Rejection Ratio

● Common-mode Rejection Ratio or CMRR is a measure of how well we achieve our
goal of making the gain for differential inputs >> the gain for common-mode inputs

● Ideally Acm = 0, so CMRR = ∞

● For practical circuits CMRR can be very high (60-120dB), but not infinite
● CMRR is typically limited by how well devices match
○ In this Difference Amplifier, the ration R4/R3 must match R2/R1
○ CMRR for opamps typically depends on the matching of devices inside of the
operational amplifier (e.g., transistors, resistors)

Differential Input Resistance

●
 Computer Engineering 24

● The Differential Input Resistance is found by summing voltages around the input loop,
with 0V between + and -:

○
○ Note the tradeoff between gain and Ridm:
■ R1 needs to be small for high Adm
■ R1 needs to be large for high Ridm
■ Hard to get both at the same time.

Instrumentation Amplifiers

●
○ Instrumentation Amplifiers provide much higher CMRR and input resistance but
require additional opamps.
■ In this version, the 1st stage gain is the same for Vidm and Vicm.
○ If we make V2 and V1 the same the common-mode voltages will be the same.
 Computer Engineering 25

●
○ The previous drawing just combined the two R1 and removed ground.
○ Much lower common-mode gain, Acm, can be achieved in the 1st stage (and
therefore much higher CMRR) just by removing the wire between node X and
ground!
■ In this new version, the 1st stage gain for Vidm is the same as before.
■ But the 1st stage gain for common-mode Vicm is reduced to only 1.
■ Note that zero current flows in both R1 and R2 due to Vicm

Instrumentation Amplifiers

Amplifiers with General Impedances

●
 Computer Engineering 26

● All the opamp circuits we’ve been studying work just as well if we replace the resistors
with impedances.
○ Z1, Z2 don’t have to be just resistors, these elements can be
frequency-dependent elements like capacitors or inductors too.
○ The ideal Opamp assumptions still hold as long as the loop gain Aβ >> 1 (starts
to fail at high frequencies, as A(ω) drops off)
● Can use this to build integrators, differentiators, and filters.

Integrators

●
○ By replacing the resistor in feedback around the opamp with a capacitor we get
an Inverting Integrator.

○
 Computer Engineering 27

Integrator with finite DC gain

● By adding a resistor in feedback in parallel with the capacitor we can limit the low
frequency gain:

● Integrators with ∞ DC gain are usually avoided since small DC voltages cause Vo to
ramp to VDD or VSS
○ Since C = an open circuit at DC no negative feedback exists.
 Computer Engineering 28

Differentiators

● By replacing the resistor at the input with a capacitor we get an Inverting Differentiator

●
● Note that differentiators are very sensitive to high frequency noise.
 Computer Engineering 29

ActiveF Lowpass Filter

● An integrator with Rf in parallel with C can also be used as an Active Lowpass Filter

●
● At ω > ω-3dB the gain magnitude rolls off at -20dB/decade

Non-ideal effects in IC opamps

● Actually, integrated circuit opamps have a number of non-ideal effects that affect
performance and must be considered in real-world applications.
● DC non-idealities include:
○ Finite DC gain (the gain is not really finite)
○ Offset voltage
○ Input bias and offset currents
○ Output voltage and current limits
● AC non-idealities include:
○ Finite bandwidth (the bandwidth is not really infinite)
○ Slew rate and full-power bandwidth

Effect of finite DC open-loop gain

 Computer Engineering 30

● We assume that ideal opamps have an infinite DC gain. Real IC opamps have high, but
not infinite gain. This non-ideal effect causes small errors in the actual gains we get in
our amplifiers. Recall for a feedback amp:

● Example: β = 0.1 is used to set Af to ~ 10 what is the value of Af as A gets small?

Effect of finite DC open-loop gain

● Inverting and non-inverting amplifier
 Computer Engineering 31

Opamp Input Offset Voltage

● The opamp’s input offset voltage is the voltage that must be applied between the + and
- inputs to set Vo = 0.
○ Appears as a shift in the Vo vs Vidm DC transfer curve.
○ Can be modeled as an extra VOS source at the opamp input.
○ VOS is caused by mismatch between transistors inside the opamp.
○ ^ has image reversed, should look like this: with the Vos
 Computer Engineering 32

●
○ Input offset voltages are amplified by the gain, just like any input
■ This can cause errors at the amplifier output.
○ Many opamps have extra terminals which allow the user to null out (cancel) the
input offset voltage.
○ Can also buy low VOS opamps when needed

Opamp input bias offset currents

● Opamps built with BJTs have input bias currents and input offset currents which can
cause errors in an opamp’s output voltage.
 Computer Engineering 33

● Not a problem for opamps that use MOSFET inputs, since MOSFETs typically have zero
gate current.
○ Some newer “nanoscale” MOST processes do have small gate leakage currents
which must be considered.

Opamp input bias and offset currents

●
● Fortunately, it’s usually easy to minimize errors due to input bias currents by adding a
resistor in the + opamp terminal
○ Choose R3 = R1 || R2 to set VO = 0 due to IB
● Still left with a small error due to IOS:
○ VO = IOSR2
● This typically reduces the output voltage error by approximately 10x.

Opamp output voltage & current limits

 Computer Engineering 34

● The opamp output voltage is limited by the supply.

● In the example shown the amplifier gain was set to 10.
● If V1 peaks at 1.5 V this means VO should peak at 15V. BUT, if +/- 13V power supplies
are used clipping occurs.
● Clipping can also occur if the opamp’s output current io needs to be larger than the max
to reach the desired VO.
○ VO = iLRL → must be able to supply enough current for small RL

Finite Bandwidth effects

● Gain-Bandwidth Product is constant anywhere on the slope between fb and fT!

○
○ Note -20dB = 10x drop in |A|, and a decads increase in frequency = 10x
● Opamps typically have high gains at low frequency, which roll off to lower values as
frequency increases.
○ Internal compensation causes gain to drop at -20dB/decade.
○ Lower gain can cause frequency dependant errors in VO
 Computer Engineering 35

Slew rate errors

●
● Slew Rate is the maximum slope at the output voltage for fast inputs
● For slower inputs VO is exponential
○ VO(t) = V(1-e(-t/Ʈ))
● If the input changes too fast the opamp’s output voltage may not be able to keep up with
the input.
● How fast VO can change is limited by the Slew Rate

Full-power Bandwidth