Bode - Plot - and - Op - Stability 1-15
Bode - Plot - and - Op - Stability 1-15
Wherever possible the overall technique used for this series will be "definition by example" with
generic formulae included for use in other applications. To make stability analysis easy we will use
more than one tool from our toolbox with data sheet information, tricks, rules-of-thumb, SPICE
Simulation, and real-world testing all accelerating our design of stable operational amplifier (op amp)
circuits. These tools are specifically targeted at voltage feedback op amps with unity-gain bandwidths
<20 MHz, although many of the techniques are applicable to any voltage feedback op amp. 20 MHz is
chosen because as we increase to higher bandwidth circuits there are other major factors in closing the
loop: such as parasitic capacitances on PCBs, parasitic inductances in capacitors, parasitic inductances
and capacitances in resistors, etc. Most of the rules-of-thumb and techniques were developed not just
from theory but from the actual building of real-world circuits with op amps <20 MHz.
This first part reviews some fundamentals essential to ease of stability analysis and defines some
nomenclature which will be used consistently throughout the entire series.
Data Sheet Info
Tricks
Rules-Of-Thumb
Tina SPICE Simulation
Testing
Goal: To learn how to EASILY analyze and design Op Amp
circuits for guaranteed Loop Stability using Data Sheet Info,
Tricks, Rules-Of-Thumb, Tina SPICE Simulation, and Testing.
The frequency response for the magnitude plot is the change in voltage gain as frequency changes,
specified on a Bode plot as voltage gain in dB vs frequency in Hz. Bode and plotted semi-log with
frequency (Hz) on the x-axis, log scale, and voltage gain (dB) on the y-axis, linear scale. Preferred y-
axis scaling is a convenient 20 dB per major division. The other half of the Bode plot is the phase plot
(phase shift vs frequency) and is plotted as degrees phase shift vs frequency. Bode phase plots are
semi-log with frequency (Hz) on the x-axis, log scale, and phase shift (degrees) on the y-axis, linear
scale. Preferred y-axis scaling is a convenient 45° degrees per major division.
Magnitude Plot
Phase Plot
Magnitude Bode Plots require voltage gain to be converted to dB, defined as 20Log1010A, where A is
the voltage gain in volts/volts (V/V).
dB A(dB) = 20Log10A where A = Voltage Gain in V/V
A (V/V) A (dB)
0.001 -60
0.01 -40
0.1 -20
1 0
10 20
100 40
1,000 60
10,000 80
100,000 100
1,000,000 120
10,000,000 140
The slope of voltage gain with frequency is defined in +20 dB/decade or -20 dB/decade increments on
a magnitude Bode plot. They can also be described as +6 dB/octave or -6 dB/octave (see Fig. 1.4)
which can be proved by:
A single pole response has a -20 dB/decade, -6 dB/octave rolloff in the Bode magnitude plot. At its
location (fP) the gain is reduced by 3dB from the dc value. In the phase plot the pole has a -45° phase
shift at fP. The phase extends on either side of fP to 0° and -90° at a -45°/decade slope. A single pole
may be represented by a simple RC low pass network as shown in Fig. 1.5. Note how the phase of a
pole affects frequencies up to one decade above and one decade below the pole frequency.
Pole Location = fP
Magnitude = -20dB/Decade Slope
Slope begins at fP and continues down
as frequency increases
Actual Function = -3dB down @ fP
Phase = -45°/Decade Slope through fP
Decade Above fP Phase = -90°
Decade Below fP Phase = 0°
Zero
A single zero response has a +20 dB/decade, +6 dB/octave "roll-up" in the Bode magnitude plot. At
the zero location (fZ) the gain is increased by 3 dB from the dc value. In the phase plot the zero has a
+45° phase shift at fZ. The phase extends on either side of fZ to 0° and +90° at a +45°/decade slope. A
single zero may be represented by a simple RC high pass network (Fig. 1.6). Note how the phase of a
zero affects frequencies up to one decade above and one decade below the zero frequency.
Zero Location = fZ
Magnitude = +20dB/Decade Slope
Slope begins at fZ and continues up as
frequency increases
Actual Function = +3dB up @ fZ
Phase = +45°/Decade Slope through fZ
Decade Above fZ Phase = +90°
Decade Below fZ Phase = 0°
On a Bode magnitude plot it is easy to find the frequency location of a given pole or zero. Since the x-
axis is a log scale of frequency this technique allows a ratio of distances to accurately and quickly
determine the frequency of the pole or zero of interest. Fig. 1.7 illustrates this "Log Scale Trick."
Log Scale Trick (fP = ?):
2) L/D = Log10(fP)
3) fP = Log10-1(L/D) = 10(L/D)
A (dB)
5) L = Log10(fp’) x D
L = Log10 (3.16) x 2cm = 1cm
where fp’ = fp normalized to the
1-10 decade range –
fP = 31.6 fP’ = 3.16
Fig. 1.7: Log Scale Trick
Intuitive Component Models
Most op amp applications use combinations of four key components-- op amp, resistor, capacitor, and
inductor -- and to facilitate stability analysis it is convenient to have "intuitive models" for them.
Our intuitive op amp model for ac stability analysis is defined in Fig. 1.8. The differential voltage
between the IN+ and IN- terminals will be amplified by x1 and converted to a single-ended ac voltage
source, VDIFF, which is then amplified by K(f) (representing the data sheet Aol Curve: open-loop gain
vs frequency). The resultant voltage, VO, is then followed by the open-loop, ac small-signal, output
resistance, RO, with the output voltage appearing as VOUT.
Our intuitive resistor model for ac stability analysis is defined in Fig. 1.9. The resistor has a constant
resistance value regardless of the operating frequency.
T 1.00k
R(f) Magnitude
750.00
Resistance (ohms)
500.00
AM1
VR
+
A
250.00
+
VG1 R1 1k
R(f) = VR / AM1
0.00
1 10 100 1k 10k 100k 1M 10M 100M
Frequency (Hz)
XC(f) Magnitude
VG1 C1 1p
150.00G
XC(f) = VR / AM1
XC(ohms)
100.00G
50.00G
Hi-f XC
0.00
T
6.50M
AM1
+
A VL
+
XL(f ) = VL / AM1
DC XL
XL(ohms)
3.25M
1.63M
Hi-f XL
0.00
The lower part of Fig. 1.14 illustrates the traditional control-loop model which represents a gain circuit
with feedback. The top part of Fig. 1.14 depicts the sections of a typical op amp circuit with feedback
which correspond to the control-loop model. This model can be called the op amp loop-gain model.
Note that the Aol is the op amp data sheet parameter Aol, and is the open-loop gain. β is the amount of
output voltage, VOUT, that gets fed back, created in this example by a resistor network. Deriving
VOUT/VIN we see that the closed-loop gain function is directly defined by Aol and β.
network
RF
network =VFB/VOUT
VOUT
VFB
RI
-
VOUT
+ RF
+ VFB
VIN
- RI
+
Acl: Closed Loop Gain
VIN
Aol VOUT
From this model we can derive the criteria for stability in a closed-loop op amp circuit:
VOUT/VIN = Aol / (1+ Aolβ)
If: Aolβ = -1
Then: VOUT/VIN = Aol / 0 Æ ∞
Stability Criteria:
At fcl, where Aolβ = 1 (0dB), Phase Shift < +180°
Desired Phase Margin (distance from +180° Phase Shift) > 45°
Fig. 1.15: Derivation Of Stability Criteria
Loop Stability Tests
Since loop stability is defined by the magnitude and phase plot of loop gain (Aolβ) we analyze its
magnitude and phase by breaking into the closed-loop op amp circuit, injecting a small-signal ac
source into the loop, and then measuring amplitude and phase to plot the complete loop-gain picture.
Fig. 1.16 shows the equivalent control-loop block diagrams for the op amp loop-gain model and the
technique we will use for the loop-gain test.
When analyzing a circuit built in SPICE for simulation, the traditional loop-gain test breaks the closed-
loop op amp circuit using an inductor and capacitor. A very large value of inductance ensures the loop
is closed at dc (a requirement for SPICE simulation is to be able to calculate a dc operating point
before performing an ac Analysis) but open at the ac frequencies of interest. A very large value of
capacitance ensures that our ac small signal source is not connected at dc but is directly connected at
the frequencies of interest.
Fig. 1.17 illustrates the SPICE setup schematic for the traditional loop-gain test.
network
RF
Op Amp Loop Gain Model
RI
VFB Op Amp is “Closed Loop”
-
VOUT
+
+
VIN
-
Before simulating a circuit in SPICE we want to know the approximate outcome. Remember GIGO
(garbage-in-garbage-out)! β and 1/β, along with the data sheet Aol curve, provide a powerful method
for first-order approximation of loop-gain analysis. In future sections tricks and rules-of-thumb will be
presented for computing β and 1/β. Fig. 1.18 defines the β network for op amp circuits.
20
Fig. 1.19: Loop Gain Information From Aol Plot And 1/β Plot
Plotting the 1/β on the Aol curve there is an easy first-order check for stability called rate-of-closure,
defined as the "rate-of-closure" of the 1/β curve with the Aol curve at fcl, where the loop gain goes to
0 dB. A 40 dB/decade rate-of-closure implies an UNSTABLE circuit, because it implies two poles in
the Aolβ plot before fcl which can mean a 180° phase shift; a 20 dB/decade rate-of-closure implies a
STABLE circuit. Four examples are shown in Fig. 1.20 with their respective rate-of-closure computed
below.
fcl1: Aol - 1/β1 = -20 dB/decade - +20 dB/decade = -40 dB/decade rate-of-closure: Unstable
fcl2: Aol - 1/β2 = -20 dB/decade - 0 dB/decade = -20 dB/decade rate-of-closure: Stable
fcl3: Aol - 1/β3 = -40 dB/decade - 0 dB/decade = -40 dB/decade rate-of-closure: Unstable
fcl4: Aol - 1/β4 = -40 dB/decade - -20 dB/decade = -20dB/decade rate-of-closure: Stable
β) = 1
At fcl: Loop Gain (Aolβ
Rate-of-Closure @ fcl =
(Aol slope – 1/β slope)
*20dB/decade Rate-of-Closure @ fcl =
STABLE
**40dB/decade Rate-of-Closure@ fcl =
UNSTABLE
fp1
135
Frequency
90
(Hz)
10 100 1k 10k 100k 1M 10M
45 fz1
fcl
fp2
-45
Fig. 1.22: Loop Gain Plot From Aol Curve & 1/β Plot
The VOUT/VIN closed-loop response is not always the same as 1/β. In the example in Fig. 1.23 we see
that the ac small-signal feedback is modified by the Rn-Cn network in parallel with RI.
At fcl Aolβ = 1 (0dB).
No Loop Gain left to correct for errors.
Aol VOUT/VIN follows the Aol curve.
100
80
fcl
60
40
VOUT/VIN Note:
20 1/β is the AC Small Signal
SSBW
(Small Signal BandWidth) Closed Loop Gain for the
0
Op Amp.
1 10 100 1k 10k 100k 1M 10M VOUT/VIN is often NOT the
Frequency (Hz) same as 1/β.
References
After earning a BSEE from the University of Arizona, Tim Green has worked as an analog and mixed-
signal board/system level design engineer for over 23 years, including brushless motor control, aircraft
jet engine control, missile systems, power op amps, data acquisition systems, and CCD cameras. Tim's
recent experience includes analog & mixed-signal semiconductor strategic marketing. He is currently a
Strategic Development Engineer at Burr-Brown, a division of Texas Instruments, in Tucson, AZ and
focuses on instrumentation amplifiers and digitally-programmable analog conditioning ICs.