# Texas Instruments | Conditional Stability in Feedback Systems | Application notes | Texas Instruments Conditional Stability in Feedback Systems Application notes

```Application Report
SLVA947 – February 2018
Conditional Stability in Feedback Systems
Daniel Hartung
ABSTRACT
Conditional stability can make a device appear unstable at first glance. This application report attempts to
give insight about why a conditional stability system could be stable, as well as provide a mathematical
approach to proving the stability. These methods are applied to the part TPS7H1101A-SP, because its
response is consistent with being conditionally stable.
1
2
3
4
5
6
7
Contents
Introduction ................................................................................................................... 2
Closed Loop Gain............................................................................................................ 2
TPS7H1101A-SP Stability .................................................................................................. 3
Nyquist Stability .............................................................................................................. 7
Nyquist Analysis of TPS7H1101A-SP..................................................................................... 8
Conclusion .................................................................................................................. 11
References .................................................................................................................. 11
List of Figures
1
Simple Block Diagram of Control Loop ................................................................................... 2
2
Frequency Response of TPS7H1101A-SP ............................................................................... 3
3
Frequency Responses of Models.......................................................................................... 4
4
Transient Responses of Models ........................................................................................... 5
5
Transient Response of TPS7H1101A-SP
6
7
8
9
10
............................................................................... 6
Nyquist Plot Example ....................................................................................................... 7
Right Half Plane Pole ....................................................................................................... 8
Frequency Response of TPS7H1101A-SP ............................................................................... 9
Nyquist Plot of TPS7H1101A-SP ........................................................................................ 10
Nyquist Plot of Mirrored Response ...................................................................................... 10
All trademarks are the property of their respective owners.
SLVA947 – February 2018
Submit Documentation Feedback
Conditional Stability in Feedback Systems
1
Introduction
1
www.ti.com
Introduction
Traditional stability measurements involve measuring the loop gain of a part. This measurement is
accomplished by breaking the loop at the feedback resistor divider and injecting a signal. The gain and
phase of the signal are measured from one side of the injected signal to the other. This measurement is
done over a wide range of frequencies. A Bode plot, the frequency response of the part, is created from
the measurement. Traditional stability says that the phase margin of the part is the phase where the gain
of the part goes to zero, and the gain margin is the negative of the gain when the phase crosses over to
zero. If the gain margin is large and positive and the phase margin is large and positive, then the part is
stable. While this way of doing stability works well, it only takes into account simplistic systems. More
specifically, the process does not take into account systems where the phase crossovers occur before the
gain crossover, but increase to a positive value before that gain crossover occurs. The TPS7H1101A-SP
device has a frequency response similar to what has been described and is used as an example.
2
Closed Loop Gain
The first step in showing how a system with a frequency response like the TPS7H1101A-SP is stable is
obtaining an understanding of why it is stable. To obtain this understanding, users must look at the
following control loop structure (see Figure 1), where X is the input and Y is the output.
Figure 1. Simple Block Diagram of Control Loop
G(s) and H(s) are transfer functions that describe the feedback loop structure. When H(s) is positive the
loop is in negative feedback, and when H(s) is negative the loop is in positive feedback. When the loop
structure is in positive feedback, the function that describes the complete control loop is given in
Equation 1.
(1)
Equation 1 is unstable when the loop gain, G(s)H(s), is equal to 1, so this behavior is avoided. In negative
feedback, Equation 2 describes the closed loop gain.
(2)
Similar to positive feedback, the equation for negative feedback is unstable when the loop gain is equal to
–1. An important point to take from these two equations is that positive and negative feedback are very
similar to one another. For large values of the loop gain, the equations for closed-loop gain are almost just
out-of-phase with each other. For small values of the loop gain, the equations are almost exactly the
same. It is only for values very close to 1 that the equations deviate drastically.
If the loop gain is 100, in negative feedback the closed loop gain (Y/X) would be G(s)/101, while in positive
feedback the closed loop gain would be –G(s)/99. In either case, the gain is finite and thus the closed
control loop is not necessarily unstable. This is similar to how when designing a closed-loop control
system, the desired loop gain when in positive feedback (or the frequency of the gain margin and beyond)
is small. This occurrence is due to the previously mentioned idea that positive and negative feedback are
2
Conditional Stability in Feedback Systems
SLVA947 – February 2018
Submit Documentation Feedback
TPS7H1101A-SP Stability
www.ti.com
almost the same for small loop gains. If the loop gain is 0.01, then the closed loop gain is G(s)/1.01 in
negative feedback and G(s)/0.99 in positive feedback. What really matters is the phase around the
crossover frequency of the device, which is where loop gains near 1 occur. A device is stable as long as it
is in negative feedback at the crossover frequency. Experiments were conducted on control loops with
positive feedback before the crossover frequency, where the phase of the crossover frequency was
controlled to be positive and negative [1]. These experiments showed the control loop as stable when the
phase margin was positive and unstable when the phase margin was negative.
This finding is not to say that having a dip of positive feedback is desired before the crossover frequency.
If the crossover frequency were to move to an area where there was positive feedback, it would cause the
device to become unstable. Intuitively, negative gain margin of a device does not necessarily mean that a
device is unstable. Gain margin is meant to show that for changes in the gain of a device over process,
there is a buffer between the gain dropping to 0 and where the phase crosses over into positive feedback.
It is more useful to look at the absolute value of that gain margin, because the gain of a device is stable
during positive feedback for both very small gains and very large gains.
3
TPS7H1101A-SP Stability
The transient response of the TPS7H1101A-SP device was throughly tested, and the device is known to
be stable from these transients. Because the transients are stable, the frequency response of the device is
also be stable.
The frequency response of the TPS7H1101A-SP device was found using the TI EVM, with VIN = 3.3 V,
VOUT = 2.5 V, and IOUT = 2.3 A, measured in Figure 2.
Figure 2. Frequency Response of TPS7H1101A-SP
The main difference between this frequency response and commonly seen responses is the fact that the
phase increases for frequencies lower than 1 kHz, but higher than DC. As explained in Section 2, this
finding does not mean the device is unstable. The fact that the gain is so high means the device is stable,
unless there is a very large, negative change in the gain of the device. The negative phase is a result of
the fact that the device is sometimes in positive feedback for very low frequencies.
SLVA947 – February 2018
Submit Documentation Feedback
Conditional Stability in Feedback Systems
3
TPS7H1101A-SP Stability
www.ti.com
The device remains in negative feedback for its DC gain and frequencies very near it. This fact is
important because the phase of the DC gain determines the DC operating point of the device. While Bode
measurements will not go down to DC, remember that the device acts the same as any other LDO when it
comes to its DC operating point, and the phase trains off the Bode plot coming down to 180 degrees. It is
because of these factors that it is safe to say the device is in negative feedback for frequencies at and
near DC.
Connecting the previous ideas to gain margin and phase margin, the phase margin of the device is the
same as any other device. The phase at the crossover frequency is the phase margin of the device. The
gain margin of the device is a little different. The device, instead of having one gain margin, has two.
These gain margins are the absolute value of the gain at each point of phase crossover. The highfrequency phase crossover happens at higher frequencies than the graph shows, and the low frequency
gain margin happens around 700 Hz.
When it comes to gain margin, the more gain there is at the point of phase crossover the better. It does
not matter if it is in the positive or negative direction. This occurrence is due to what was previously
discussed, where for very large and small gains positive and negative feedback is similar and stable. What
does matter is that the phase being negative is no where near the crossover frequency.
The TPS7H1101A-SP device, with the Bode plot shown, is stable for all frequencies. The fact that at low
frequencies the phase is different than traditional LDOs does not effect the stability of the part.
Two models were created similar to the TPS7H1101A-SP device, where one model included the loop that
added the positive feedback into the part and one model took that loop out. Figure 3 shows the frequency
response of the two parts. The model that includes the positive feedback loop shows the gain in pink and
the phase in red. The model without the positive feedback loop shows the gain in yellow and the phase in
green.
Figure 3. Frequency Responses of Models
4
Conditional Stability in Feedback Systems
SLVA947 – February 2018
Submit Documentation Feedback
TPS7H1101A-SP Stability
www.ti.com
Notice that the models become the same for frequencies after about 1 kHz. This shows the positivefeedback loop only changes the response for frequencies before the crossover frequency, and thus the
two models have the same crossover frequency and phase margin. Crossover frequency and phase
margin are the two major contributors that determine the transient response of a control loop, and this was
supported when the models had their transient responses measured, shown in Figure 4.
2.55
Simulated Transient
2.54
2.53
2.52
Conditionally Stable
2.51
Always Stable
2.5
2.49
2.48
0
0.0005
0.001
0.0015
Figure 4. Transient Responses of Models
SLVA947 – February 2018
Submit Documentation Feedback
Conditional Stability in Feedback Systems
5
TPS7H1101A-SP Stability
www.ti.com
The models, despite one having positive feedback, have the same transient response. This supports the
idea that positive feedback is only an issue for frequencies near the crossover frequency. Parts like the
TPS7H1101A-SP that have positive feedback for some frequencies not near their crossover frequency
can be just as stable as parts that do not include positive feedback. The models are supported when
looking at a load step from 2.75 A to 0 A for the TPS7H1101A-SP, as shown in Figure 5.
Figure 5. Transient Response of TPS7H1101A-SP
6
Conditional Stability in Feedback Systems
SLVA947 – February 2018
Submit Documentation Feedback
Nyquist Stability
www.ti.com
4
Nyquist Stability
Nyquist stability is a stability method that was used as the basis for the traditional Bode plot stability that is
used today. The stability method focuses on the denominator of the negative-feedback closed-control
loop. Nyquist diagrams is a transform of the transfer function of the control system that plots it on real and
imaginary axises. Figure 6 shows an example of what a Nyquist diagram may look like.
Figure 6. Nyquist Plot Example
The diagrams are also always symmetrical because half of the frequencies the diagram plots are positive
and half of them are negative. In the graph, the blue line is the positive frequencies and the red line is the
negative frequencies. The arrows in the diagram point toward higher frequencies for positive frequencies
and lower frequencies for negative frequencies. These facts are important for later when the Nyquist
diagram of the TPS7H1101A-SP device is found on bench and it looks like only half of the diagram is
there. Additionally, no arrows are present in the diagram. Part of the reason only half of the graph is there
is that the positive frequencies are measured, and the negative frequencies are assumed to be
symmetrical around the x-axis.
To determine if a system is stable using Nyquist stability, use Equation 3 [2].
where
•
•
•
Z = The number of right half plane (RHP) poles of the closed-loop system
P = The number of RHP poles in the loop gain system
N = The number of clockwise rotations around –1 (a counterclockwise rotation around –1 is counted as a
negative rotation)
(3)
SLVA947 – February 2018
Submit Documentation Feedback
Conditional Stability in Feedback Systems
7
Nyquist Analysis of TPS7H1101A-SP
www.ti.com
The closed-loop system is stable when there are no RHP poles in the system. This fact is because an
RHP pole in a transfer function becomes an exponential with a positive exponent in the time domain, as
given in Figure 7.
Positive Exponential
RHP Pole
1
at
e
s-a
Inverse Laplace Transform
Figure 7. Right Half Plane Pole
RHP poles can be identified in a Bode plot as a pole where instead of the phase decreasing, the phase
increases. RHP poles correspond to an unstable equation in time, therefore if there are any in the closed
loop function, the control loop is unstable. Thus for Equation 3, the closed-loop transfer function of a
system is stable when Z ≤ 0.
5
Nyquist Analysis of TPS7H1101A-SP
The different closed-loop response of the TPS7H1101A-SP device changes the equation that the Nyquist
plot must consider. While most applications of Nyquist plots have the encirclements around the point
(–1,0), the TPS7H1101A-SP device has the encirclements around the point (1,0).
This difference is because the TPS7H1101A-SP, and other parts, create a different denominator than the
denominator that traditional Nyquist stability considers. While Nyquist stability has a denominator that
looks like 1 + G(s)H(s), the TPS7H1101A-SP device has a denominator that looks more like –1 +
G(s)H(s). This discrepancy is due to the instability point and how Bode plots relate to Nyquist plots.
To obtain an understanding of why the point changes from (–1,0) to (1,0), the time at which a control loop
is considered unstable must be examined. The transform that creates Nyquist diagrams plots the
magnitude and phase of the data in the Bode plot on a circular plot, with the magnitude as the radius and
the phase as the angle from the left-hand side x-axis. This plotting is very similar to polar graphs. In
traditional bode plots for power control loops, the point at which a device is considered unstable is at 0 dB
and 0 degrees of phase margin. Because 0 dB corresponds to 1 in amplitude this makes the point at
which the device is unstable at (1,0). While traditional bode plots for power control loops start at 180
degrees and decrease to 0 degrees, this is not true of traditional Nyquist stability. Traditional Nyquist
stability starts at 0 degrees and decreases to –180 degrees [2]. For traditional Nyquist stability, the angle
would therefore be –180 and the magnitude is 0 dB, which is the point (–1,0). This idea has a basis in the
mathematics behind Nyquist stability.
Nyquist stability can be built from Cauchy's argument principle [2], also known as the principle of variation
of the argument. Cauchy's argument principle is used to prove Nyquist stabilityReference [2]. The source
shows how Nyquist stability can be built from Cauchy's principle by setting the function f(s) = 1 + L(s),
where L(s) is the loop-transfer function. Where f(s) would have encirclements around the origin, the 1 in
the equation simply shifts the origin over to (–1,0). With the TPS7H1101A-SP device, the denominator 1 –
L(s), can be found from the bode plot frequency response, starting at 180 degrees. This act changes the
equation that plots the Nyquist stability from f(s) = 1 + L(s) to f(s) = 1 - L(s). This act rotates the point
(–1,0) to (1,0) as the negative rotates the function by 180 degrees. The direction of the arrow is conserved
from this rotation.
8
Conditional Stability in Feedback Systems
SLVA947 – February 2018
Submit Documentation Feedback
Nyquist Analysis of TPS7H1101A-SP
www.ti.com
The first step to determining the stability of the TPS7H1101A-SP device is looking at the Bode plot of the
frequency response and looking for any RHP poles. Figure 8 shows the frequency response of the
TPS7H1101A-SP device for convenience.
Figure 8. Frequency Response of TPS7H1101A-SP
Looking at the Bode plot there is one instance of the gain decreasing and the phase increasing. There is
what looks like a RHP pole in the region of 10 to 100 Hz, but while the phase generally increases by only
90 degrees, the phase ends up increasing by around 250 degrees. Given the phase change, it looks like
there may be three RHP poles in the response, however given that the gain only decreases by around 20
dB/decade this seems unlikely. An explanation that fits this behavior is that there are two RHP poles and
one zero. Adding these poles and zeros together around the location gives a slope of –20 dB/decade and
still has that very large phase change. This occurrence means that for the Nyquist stability equation P = 2.
The next step in determining the stability of the device is finding the number of encirclements around 1
(instead of –1, as previously explained in this section). Figure 9 shows a Nyquist plot that was taken using
the same machine and setup that the Bode plot used.
SLVA947 – February 2018
Submit Documentation Feedback
Conditional Stability in Feedback Systems
9
Nyquist Analysis of TPS7H1101A-SP
www.ti.com
Figure 9. Nyquist Plot of TPS7H1101A-SP
Remember that as explained in Section 4, the results must be mirrored about the x-axis to get the full
picture, because negative frequencies are not measured. What is not shown on the plot is that the
direction of the plot is counter clockwise. This occurrence is because the low frequencies measured start
on the far left of the graph and then follow the blue line as the frequencies increase. The 1 on this plot is
the right most point of the 0 dB circle. If the results are mirrored about the x-axis, as shown in Section 4,
there are –2 encirclements (or 2 counterclockwise encirclements) around point 1.
Figure 10. Nyquist Plot of Mirrored Response
10
Conditional Stability in Feedback Systems
SLVA947 – February 2018
Submit Documentation Feedback
Conclusion
www.ti.com
One encirclement is caused by the half circle at the bottom of the plot, and the other encirclement is
caused by the half circle at the top of the plot, which means that N = -2.
Using Equation 3 for Nyquist stability, Z = P + N becomes Z = 2 + (–2) and thus Z = 0. Because Z ≤ 0, the
closed-loop system for the device is stable.
The phase margin of the device can also be concluded from the Nyquist graph. The angle from the
positive x-axis, to the point at which the line intercepts the 0 dB circle, is considered the phase margin.
Looking at the Nyquist graph, the point at which the line intercepts the circle is at a little above 60 degrees
from the positive x-axis. This correlates well to the 60.23 degrees of phase margin that were found in the
Bode plot for the control loop.
6
Conclusion
Control loops that have phase crossovers before their crossover frequency are not always unstable. The
equations for closed loop gain show how the phase that really matters is the phase of the crossover
frequency. This fact is because for very large and very small gains, positive and negative feedback are
very similar. As long as the phase at the crossover frequency indicates the device is in negative feedback,
the device is stable. A minor tweak to Nyquist stability can be used to verify the stability of the device.
Both Nyquist stability and Bode plot stability can be used to find the phase margin of the device.
7
References
1. Roberge, James. Lecture 17: Conditional Stability. Video. MIT OpenCourseWare, Massachusetts
Institute of Technology. 2017.
2. Åström, Karl Johan and Murray, Richard M. Frequency Domain Analysis. In Feedback Systems: an
Introduction for Scientists and Engineers, 267–292. Princeton University Press, 2011.
SLVA947 – February 2018
Submit Documentation Feedback
Conditional Stability in Feedback Systems
11
Texas Instruments Incorporated (‘TI”) technical, application or other design advice, services or information, including, but not limited to,
reference designs and materials relating to evaluation modules, (collectively, “TI Resources”) are intended to assist designers who are
developing applications that incorporate TI products; by downloading, accessing or using any particular TI Resource in any way, you
(individually or, if you are acting on behalf of a company, your company) agree to use it solely for this purpose and subject to the terms of
this Notice.
TI’s provision of TI Resources does not expand or otherwise alter TI’s applicable published warranties or warranty disclaimers for TI
products, and no additional obligations or liabilities arise from TI providing such TI Resources. TI reserves the right to make corrections,
enhancements, improvements and other changes to its TI Resources.
You understand and agree that you remain responsible for using your independent analysis, evaluation and judgment in designing your
applications and that you have full and exclusive responsibility to assure the safety of your applications and compliance of your applications
(and of all TI products used in or for your applications) with all applicable regulations, laws and other applicable requirements. You
represent that, with respect to your applications, you have all the necessary expertise to create and implement safeguards that (1)
anticipate dangerous consequences of failures, (2) monitor failures and their consequences, and (3) lessen the likelihood of failures that
might cause harm and take appropriate actions. You agree that prior to using or distributing any applications that include TI products, you
will thoroughly test such applications and the functionality of such TI products as used in such applications. TI has not conducted any
testing other than that specifically described in the published documentation for a particular TI Resource.
You are authorized to use, copy and modify any individual TI Resource only in connection with the development of applications that include
the TI product(s) identified in such TI Resource. NO OTHER LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE TO
ANY OTHER TI INTELLECTUAL PROPERTY RIGHT, AND NO LICENSE TO ANY TECHNOLOGY OR INTELLECTUAL PROPERTY
RIGHT OF TI OR ANY THIRD PARTY IS GRANTED HEREIN, including but not limited to any patent right, copyright, mask work right, or
other intellectual property right relating to any combination, machine, or process in which TI products or services are used. Information
regarding or referencing third-party products or services does not constitute a license to use such products or services, or a warranty or
endorsement thereof. Use of TI Resources may require a license from a third party under the patents or other intellectual property of the
third party, or a license from TI under the patents or other intellectual property of TI.
TI RESOURCES ARE PROVIDED “AS IS” AND WITH ALL FAULTS. TI DISCLAIMS ALL OTHER WARRANTIES OR
REPRESENTATIONS, EXPRESS OR IMPLIED, REGARDING TI RESOURCES OR USE THEREOF, INCLUDING BUT NOT LIMITED TO
ACCURACY OR COMPLETENESS, TITLE, ANY EPIDEMIC FAILURE WARRANTY AND ANY IMPLIED WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT OF ANY THIRD PARTY INTELLECTUAL
PROPERTY RIGHTS.
TI SHALL NOT BE LIABLE FOR AND SHALL NOT DEFEND OR INDEMNIFY YOU AGAINST ANY CLAIM, INCLUDING BUT NOT
LIMITED TO ANY INFRINGEMENT CLAIM THAT RELATES TO OR IS BASED ON ANY COMBINATION OF PRODUCTS EVEN IF
DESCRIBED IN TI RESOURCES OR OTHERWISE. IN NO EVENT SHALL TI BE LIABLE FOR ANY ACTUAL, DIRECT, SPECIAL,
COLLATERAL, INDIRECT, PUNITIVE, INCIDENTAL, CONSEQUENTIAL OR EXEMPLARY DAMAGES IN CONNECTION WITH OR
ARISING OUT OF TI RESOURCES OR USE THEREOF, AND REGARDLESS OF WHETHER TI HAS BEEN ADVISED OF THE
POSSIBILITY OF SUCH DAMAGES.
You agree to fully indemnify TI and its representatives against any damages, costs, losses, and/or liabilities arising out of your noncompliance with the terms and provisions of this Notice.
This Notice applies to TI Resources. Additional terms apply to the use and purchase of certain types of materials, TI products and services.
These include; without limitation, TI’s standard terms for semiconductor products http://www.ti.com/sc/docs/stdterms.htm), evaluation
modules, and samples (http://www.ti.com/sc/docs/sampterms.htm).
Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265