Neil Dennehy

[email protected]

Fundamentals of Linear
Stability

by
Ken Lebsock
Learning from the Past, Looking to the Future
Learning from the Past, Looking to the Future
 Fundamentals  of    
3 Linear  
 Thank  you  Neil,  S tability  
I’m   pleased  to  be  here.  

Webcast  Presenta,on  by:  Kenneth  Lebsock  Ph.D.  

 
NASA  Engineering  &  Safety  Center  (NESC)  
GN&C  Technical  Discipline  Team  (TDT)  Deputy    

NASA Engineering & Safety Center (NESC) GN&C TDT 3  

 Introduc,on  
1)  Prequel  
2)  Stability  of  Aerospace  Systems  
3)  Quan,fying  Stability  

NASA Engineering & Safety Center (NESC) GN&C TDT 4  

 Prequel  
•  Roughly  speaking,  a  linear  system  is  said  to  be  stable  when  it  returns  to  its  steady-­‐
state  condi,on  aSer  a  disturbance  and  unstable  if  it  diverges  aSer  a  disturbance.  
Stability  is  a  System  Property.  

•  This  presenta,on  deals  with  only  the  first  of  the  two  main  categories  of  commonly  
used  approaches  to  analyze  the  stability  of  a  linear  system:  
1)  Frequency  Analysis  Methods  based  on  the  Nyquist  Criterion  and  displayed  
graphically  with  Nyquist  Plots,  Nichols  Charts,  or  Bode  Gain  and  Phase  Plots  
2)  Pole-­‐Zero-­‐Configura,on  Methods  based  on  the  Routh-­‐Hurwitz  Criterion  and  
displayed  graphically  in  Root  Locus  Plots  
 
•  The  importance  of  Frequency  Analysis  Methods  lies  in  the  fact  that  they  can  be  
used  to  determine  the  rela,ve  degree  of  system  stability  by  producing  the  so-­‐
called  phase  and  gain  stability  margins.    
–  These  stability  margins  are  needed  for  frequency  domain  controller  design  
techniques.  

 
  NASA Engineering & Safety Center (NESC) GN&C TDT 5  
 Stability  of  Aerospace  Systems  

•  Stability  is  of  concern  for  any  

control  system  but  it  is  especially  
important  in  the  a\tude  control  of  
aerospace  vehicles  which  may  be  
inherently  unstable  in  the  absence  
of  ac,ve  control  (e.g.  NASA  X-­‐29).  
•  Gain  margin  and  phase  margin  
represent  the  tolerance  of  a  control  
loop  to  perturba,ons  in  loop  gain  
and  phase  delay.    
•  Stability  requirements  specified  as  gain  and  phase  margins  are  convenient  because  
they  can  be  measured  in  a  closed  loop  control  system  by  ar,ficially  introducing  
varia,ons  in  the  loop  gain  or  the  phase  delay  un,l  instability  is  observed.  
•  These  stability  margins  can  be  found  analy,cally  by  examina,on  of  the  Open  loop  
ferFunc,onfer  func,on  Bode  gain  and  phase  plots  as  func,ons  of  frequency,  the  
Nichols  chart  gain  and  phase  cross  plot,  or  the  Nyquist  plot  of  Real  vs.  Imaginary  
parts.  

NASA Engineering & Safety Center (NESC) GN&C TDT 6  

 Quan,fying  Stability  

•  In  this  presenta,on  of  control  loop  design  in  the  frequency  domain,  different  
stability  metrics  will  be  discussed:    
-  Gain  Margin  and  Phase  Margin    
-  Stability  Margin    

•  Three  examples  of  determining  Gain  and  Phase  Margins  from  the  Open  Loop  
Transfer  Func,on,  L(jω) ,  will  be  demonstrated:  
-  Bode  plots  of  magnitude    and  phase  of  L(jω) vs.  frequency  
-  Nyquist  plot  of  the  real  and  imaginary  parts  of  L(jω)
-  Nichols  cross-­‐plot  of  magnitude  vs.  phase  of  L(jω)
 
•  The  more  compact  and  precise  Stability  Margin  is  found  by  two  methods:  
-  Graphically  from  the  Nyquist  plot  of  the  real  and  imaginary  parts  of  L(jω)
-  Analy,cally  from  the  maximum  value  of  the  Sensi,vity  Func,on  S(jω)  
 
 

NASA Engineering & Safety Center (NESC) GN&C TDT 7  

 Transfer  Func,ons  

1)  Open-­‐Loop  Transfer  Func,on  

2)  Complementary  Sensi,vity  Transfer  Func,on  
3)  Sensi,vity  Func,on  

NASA Engineering & Safety Center (NESC) GN&C TDT 8  

 Open-­‐Loop  Transfer  Func,on  

d n

r+ e u + +v x+ + y
_ C(s) G(s)

Basic  LTI  SISO  Feedback  Control  Loop    

•  The  Open  Loop  Transfer  Func,on    L(s) = G(s) C(s) where  G(s) is  the  Transfer  Func,on    
of  the  Plant  Dynamics  and  C(s) is  the  Transfer  Func,on  of  the  Controller.    

•  The  symbol  s is  the  Laplace  variable.  In  the  frequency  domain  analysis  we  use  the    
subs,tu,on:  s => jω  where  j  is  √-­‐1,  and  ω  is  the  frequency  in  rad/sec.  

•  The  Open  Loop  Transfer  Func,on  describes  how  the  system  output  x  would  respond    
to  an  input  r  if  the  feedback  loop  were  not  closed.    
 
NASA Engineering & Safety Center (NESC) GN&C TDT 9  
 Complementary  Sensi,vity  Transfer  Func,on  

|  T(jω) | (dB)   •   The  Complementary  Sensi,vity  Func,on  is  

the  Closed  Loop  Transfer  Func,on:  
 
0 T (s) = L(s) = x(s)
  1+ L(s) r(s)
-3  
•   T(s) describes  how  the  system  output  x    
  responds  to  an  input  r  when  the  feedback  
loop  is  closed.    
 
•   The  Closed  Loop  becomes  unstable  as:    
ωbw  
Log ω  (rad/sec)          L(s) =>  -­‐1  
       
•   The  ideal  Closed  Loop  Transfer  Func,on  has     unity  gain  and  zero  phase  shiS  at  low    
frequencies  so  that  the  reference  input  r  is  tracked  perfectly.  Resonant  peaking  ,  just    
under  the  closed  loop  3dB  bandwidth,  is  undesirable  but  oSen  inevitable.    
 
•   We  desire  that  the  Closed  Loop  Transfer  Func,on  rolls  off  at  frequencies  that  are  
higher  than  the  system  bandwidth  in  order  to  filter  out  delitrious  sensor  noise.    

NASA Engineering & Safety Center (NESC) GN&C TDT 10  

 Sensi,vity  Func,on  
|  S(jω) | (dB)   •   The  Sensi,vity  Func,on:    
                           S(s)
             =       1 = T(s)
Ms   1+ L(s) L(s)
0   is  the  ra,o  of  the  output  of  the  closed  
  loop  system to  the  output  of  the  open  
loop    system.  It  describes  the  effect  of  
feedback  on  the  output.    
 
•   The  Sensi,vity  and  Complementary  
Sensi,vity  Func,ons  are  constrained  at  
ωms  
all  frequencies  by:  S(jω) + T(jω) =1  
Log ω  (rad/sec)    
 
 
•   Plant  disturbances  with  frequencies  such  that  |S(jω)|  <  1  (i.e.  0  dB)  are  akenuated.  

•   Disturbances  at  frequencies  where  |S(jω)|  >  1  are  amplified  by  feedback.      

•   The  maximum  Sensi,vity,    Ms=

                   S(j
       ω
       )            ,  occurring  at  the  frequency  ωms  ,  is  thus  a    
∞
measure  of  the  largest  amplifica,on  of  the  low  frequency  Plant  disturbances.    

NASA Engineering & Safety Center (NESC) GN&C TDT 11  

 Bode  Plots  and  Stability  

1)  Gain  and  Phase  Margins  

2)  Calcula,on  of  Stability  Margins  from  L(jω)  
3)  Stability  Margins  from  Bode  Plots  
4)  Stability  Margins  from  Bode  Plots  (cont.)  

NASA Engineering & Safety Center (NESC) GN&C TDT 12  

 Gain  and  Phase  Margins  

•  Gain  margin  and  phase  margin  are  two  independent  measures  of  rela,ve  stability.  
They  measure  how  "close"  a  system  is  to  crossing  the  boundary  between  stability  
and  instability.  

•  Gain  margin  is  the  amount  of  change  in  the  value  of  the  gain  of  the  Open  Loop  
Transfer  Func,on  L(jω) ,  from  its  present  value,  to  that  value  that  will  make  the  
Bode  magnitude    pass  through  the  0  dB  at  the  same  frequency  where  the  phase  is  
-­‐180  degrees.    

•  Phase  margin  is  the  amount  of  pure  phase  shiS  (no  change  in  magnitude)  that  will  
make  the  phase  shiS  of  L(jω) equal  to  -­‐180  degrees  at  the  same  frequency  where  
the  magnitude  is  0  dB  (1  in  absolute  value).    

•  The  Gain  and  Phase  margins  may  be  found  from:  

-  Bode  plots  of  magnitude    and  phase  of  L(jω) vs.  frequency  
-  Nyquist  plot  of  the  real  and  imaginary  parts  of  L(jω)
-  Nichols  cross  plot  of  magnitude  vs.  phase  of  L(jω)
 
 
 
NASA Engineering & Safety Center (NESC) GN&C TDT 13  
 Calcula,on  of  Stability  Margins  from  L(jω)  

•  Simple  algorithms  are  used  to  calculate  the  gain  and  phase  margins  
directly  from  the  Open  Loop  Transfer  Func,on.  

•  Iterate  to  find  the  frequency,  ωgm  ,  at  which  Arg[L(jω)] =  -­‐180°  
-  The  gain  margin  is  -­‐  Abs[L(jωgm )]
 
•  Iterate  to  find  the  frequency,  ωpm  ,  at  which  Abs[L(jω)] =  1    (i.e.  0  dB)
-  The  phase  margin  is    Arg[L(jωpm )] +  180°  

•  The  Bode  magnitude  plot  should  be  examined  carefully  to  determine  if  
there  are  mul,ple  crossings  of  the  0  dB  line.  

•  Also  check  if  resonant  peaks  occur,  especially  in  the  frequency  range:  
   ωpm< ω < ωgm .  
 

 
NASA Engineering & Safety Center (NESC) GN&C TDT 14  
 Stability  Margins  from  Bode  Plots  
Magnitude HdBL vs Freq HradêsecL
•   Illustra,ve  Example   10.
wpm wgm
•  Open  Loop  Transfer  Func,on:   0.

7.6(s 2 +0.1s +0.55) -10.

gm
  L(s )=
s (s +1)(s +20)(s 2 +0.06s +0.5) -20.

-30.
•  The  analy,c  calcula,on  of  the  Gain  and   -40.
Phase  Margins  indicates  that  the  System  
is  extremely  stable.   0.1 1 10

Phase HdegL vs Freq HradêsecL

-  At  the  -­‐180°  Phase  Cross-­‐Over  Frequency  
ωgm =  4.37  rad/sec, gm =    34.5  dB.   -100.

-120.
-  At  the  0  dB  Gain  Cross-­‐Over  Frequency,   -140.
ωpm  =  0.41  rad/sec,    φm =    68.6°.   fm
  -160.
•  The  control  loop  easily  meets  a  stability   -180.
requirement  of  gm = 6 dB Φm = 30°.   wpm wgm
-200.
0.1 1 10

NASA Engineering & Safety Center (NESC) GN&C TDT 15  

 Stability  Margins  from  Bode  Plots  (cont.)  
Magnitude HdBL vs Linear Freq HradêsecL
0
•  The  tenuous  stability  of  the  system  is  
obvious  in  a  blowup  of  the  Bode  plots  in   -3
the  frequency  range  0.5  –  0.9  rad/sec.  
  -6
•  The  magnitude  has  a  local  maximum  at  
Minimum
Radius from
-­‐0.33  dB  at  ω  =  0.69  rad/sec.   -9
Nyquist Point

•  The  phase  angle  has  a  local  minimum  of     0.5 0.6 0.7 0.8 0.9

-­‐178.5  deg  at  ω  =  0.73  rad/sec.  

Phase HdegL vs Linear Freq HradêsecL
•  Neither  of  these  two  extreme  values  
uniquely  defines  a  precise  measure  for   -120

the  stability  of  the  system  because  they  

do  not  occur  at  the  same  frequency.   -140
Minimum
Radius from
Nyquist Point
•  The  worst  case  stability  occurs    at  a   -160
frequency  somewhere  between  these  
two  cases  where  L(jω) approaches the -180
closest to -1.     0.5 0.6 0.7 0.8 0.9

NASA Engineering & Safety Center (NESC) GN&C TDT 16  

 Nyquist  Plots  and  Stability  

1)  Stability  Margins  on  the  Nyquist  Plot  

2)  Blowup  of  Nyquist  Plot  for  the  illustra,ve  Example  
3)  Gain  &  Phase  Margins  from  Nyquist  Plot  for  Example  
4)  Gain  and  Phase  Margins  Bounds  from  the  Sensi,vity  Func,on    
5)  Stability  Margin  from  Maximum  Sensi,vity  for  Example  
6)  Calcula,on  of  Stability  Margin  from  S(jω)  

NASA Engineering & Safety Center (NESC) GN&C TDT 17  

 Stability  Margins  on  the  Nyquist  Plot  

•  Recall  that Ms,  the  maximum   Nyquist  Plot  (for  a  generic  L(jω)  
magnitude  of  S(jω), is  also  the    
Im L(jω)  

minimum  of  |1+L(jω)|.    
•  sm  is  the  shortest  distance  from  the   Δm

Nyquist  curve,  L(jω),  to  (-1, 0 j),  the  
cri,cal  point,  i.e. sm  = Min|1+L(jω)|.    
 
Re L(jω)  

-1

•  An  alterna,ve  way  to  express   sm  

margins  is  by  a  single  number,  sm  ,  
φm  
the  Stability  Margin.  
•  Therefore  the  maximum  Sensi,vity  is  
also  a  measure  of  minimum  stability   L(jω)    

since  Ms=1/sm.    
•  The  Sensi,vity  maximum  is  a  more   •  gm  =  20Log10[1/Δm]  =  Gain  Margin  
compact  indicator  of  stability  than  a   •  φm =  Phase  Margin  
pair  of  gain  and  phase  margins.     •  sm    =  1/Ms =  Stability  Margin  

NASA Engineering & Safety Center (NESC) GN&C TDT 18  

 Blowup  of  Nyquist  Plot  for  Example  L(jω)  
7.6(s 2 +0.1s +0.55)
L(s )=
s (s +1)(s +20)(s 2 +0.06s +0.5)
Nyquist Plot
•  Ms,  the  maximum  value  of  the  magnitude  
of  the  Sensi,vity  occurs  at  some  frequency  
ωm , i.e.    Ms =   | S(jωm)|  =    1 /  |1+L(jωm)|. 0.2

”
•  The  denominator,  |1+L(jωm)|, reaches its’ +1
0.0
minimum value at the same frequency, ωm.  

L I jwm M

Imag @LH jwLD ô

sm
-0.2
•  The  magnitude  of  the    posi,on  vector  from  
the  Nyquist  point,  (-­‐1,  0j), to the L(jω)
locus reaches a minimum value, at ω = ωm:   -0.4
sm  =  |1+L(jωm)|  
w
-0.6
•  Therefore  the  closest  approach  of  the  
Open  Loop  Transfer  Func,on,  L(jω), to the
Nyquist point,  (-­‐1,  0j), is: -1.0 -0.8 -0.6 -0.4 -0.2 0.0
sm =  1 /    Ms Real @LH jwLD ô

NASA Engineering & Safety Center (NESC) GN&C TDT 19  

 Gain  &  Phase  Margins  from  Nyquist  Plot    
for  Example  
Phase  Margin  is  found  from  the  
Intersec,on  of  L(jω) with  the   0.000
-0.002
Unit  Circle  on  the  Nyquist  Plot -0.004 w Dm
-0.006
•  φm =  68.6  deg -0.008
-0.04 -0.03 -0.02 -0.01 0.00
0.0
Expanded  Region  near  Nyquist  Plot  Origin  
-0.2

-0.4
fm
The  Gain  Margin  is  found  at  the  
Intersec,on  of  L(jω) with  the  Nega,ve  
-0.6

-0.8
Real  Axis  using  an  Expanded  Scale  
-1.0 around  the  Origin  of  the  Nyquist  Plot  
w≠ •  gm  =  20Log10[1/Δm]  =  34.5  dB  
-1.2

-1.4
-1.0 -0.8 -0.6 -0.4 -0.2 0.0

NASA Engineering & Safety Center (NESC) GN&C TDT 20  

 Gain  and  Phase  Margins  Bounds  
from  the  Sensi,vity  Func,on  

•  The  rela,onships  between  the  maximum  Sensi,vity,  Ms,  and  the  lower  bounds  of  the  
gain  and  phase  margins  are  given  by  the  following  inequali,es:  
 
gm ≥ 20Log10 # Ms & (dB) φm ≥ 2ArcSin # 1 & ≥ 1 (rad)
! $ ! $
and
  ! " Ms -1 % #" 2Ms &% Ms

•  Typical  specifica,ons  for  maximum  Sensi,vity  magnitude  are  in  the  range  of  1.33  to  2  
which  corresponds  to  gain  and  phase  margins  of  (12  dB  and  45°)  to  (6  dB  and  30°).  
 
•  These  inequali,es  are  useful  even  for  poorly  behaved  Open  Loop  Transfer  Func,ons.  

•  Be  very  careful  if  the  lower  bounds  of  the  gain  and  phase  margins  found  from  these  
inequali,es  are  rela,vely  small  compared  to  the  gain  and  phase  margins  found  from  
the  Bode  plots.  

NASA Engineering & Safety Center (NESC) GN&C TDT 21  

 Stability  Margin  from  Maximum  Sensi,vity  
for  Example  

•  Example   1.0
Nyquist Plot

  7.6(s 2 +0.1s +0.55)

L(s )=
s (s +1)(s +20)(s 2 +0.06s +0.5)
0.5
•  The  magnitude  of  the  Sensi,vity  is  at  
its    maximum  value  of  Ms  =  4.0  when  

Imag @LH jwLD ô

sm
the  frequency  is  ωm=  0.714  rad/sec   0.0

  Sensitivity Function Magnitude HLog-Log PlotL

5.0
-0.5
Ms = 4

2.0
    -1.0
» SH jwL »

1.0 w

Real @LH jwLD ô

-1.0 -0.5 0.0 0.5 1.0
0.5

0.2 sm = 1/Ms = 0.25

Frequency w HradêsecL
0.1 0.2 0.5 1.0 2.0 5.0 10.0
gm ≥ 2.50 dB and Φm ≥ 14.4°

NASA Engineering & Safety Center (NESC) GN&C TDT 22  

 Calcula,on  of  Stability  Margin  from  S(jω)  

•  A  simple  algorithm  is  used  to  calculate  the  Stability  Margin,  sm ,  directly  
from  the  Sensi,vity  Transfer  Func,on.  

–  Iterate  on  frequency  to  find  the  maximum  of  the  absolute  value  of  the  
Sensi,vity  func,on, Ms = Max[  |S(jω)| ]  .  

–  The  Stability  Margin  is  simply  sm = 1/Ms.  

•  Check  the  Nyquist  plot  to  confirm  that  the  value  of  sm  found  by  itera,on  is  
the  closest  approach  of  L(jω)  to  the  (-1, 0 j) point.  
 

NASA Engineering & Safety Center (NESC) GN&C TDT 23  

 Nichols  Plots  and  Stability  

1)  Gain  &  Phase  Margins  on  Nichols  Plot  

2)  Nichols  Plot  Stability  Margin  Contours  
3)  Stability  Boundary  on  Nichols  Plot  
4)  Blowup  of  Nichols  Plot  for  Example  1  

NASA Engineering & Safety Center (NESC) GN&C TDT 24  

 Gain  &  Phase  Margins  on  Nichols  Plot  

•  Example   Magnitude HdBL vs Phase HdegL

  10
  7.6(s 2 +0.1s +0.55)
L(s )= fm = 68.6°
s (s +1)(s +20)(s 2 +0.06s +0.5)
  0
•  The  Gain  and  Phase  Margins  are  
readily  apparent  on  the  Nichols  
plot,  perhaps  more  so  than  on  the   -10
Bode  plots:     gm = 34.5dB
gm = 34.5 dB Φm = 68.6°
-20
 
•  However  the  loop  approaching  the    
 (0  dB,  -­‐180°)  point  indicates  that   -30
there  is  far  less  loop  stability  than  is   LH jwL w
indicated  by  the  gain  and  phase  
margins  on  the  Nichols  plot.   -40

•  The  Nichols  plot  needs  an  overlay  

of  equal  Stability  Margin,  sm ,   -50
contour  lines  to  precisely  define   -200 -180 -160 -140 -120 -100
stability.  

NASA Engineering & Safety Center (NESC) GN&C TDT 25  

 Nichols  Plot  Stability  Margin  Contours  

Nichols Plot Stability Margin Contours<

•   Realis,cally  the   0
Stability  Margin   sm =
1
sm =
3

specifica,ons  should   -2
4 8

be  in  the  range  of:   sm =

1
sm =
5
2 8
   ½  ≤  sm  ≤  ¾.  

Gain » LH jwL » HdBL

-4
3
sm =
4
• Then  the  stability   -6

contour  will  lie  

between  the  blue  and   -8

red  contours.  
-10

• The  locus  of  the  open  

-12
loop  transfer  func,on,   -180 -170 -160 -150 -140 -130
L(jω), should not pass Phase –° LH jwL HdegL
above and to the left of Nichols  Plot  Contours  equivalent  to    constant  radius,  sm,    
the specified stability about  the  (-1, 0 j) point  of  the  Nyquist    Stability  Criterion  
margin contour on the
Nichols Plot.  
NASA Engineering & Safety Center (NESC) GN&C TDT 26  
 Stability  Requirements  on  Nichols  Plot  
•  Stability  requirements  are  commonly  
specified  as  a  pair  of  numbers  for  example:    
gm = 6 dB Φm = 30°.   Nichols Plot
0
fm = 30°
•  The  actual  margins  are  easily  read  on  a  
Nichols  plot  as  the  points  where  the  L(jω)  
locus  crosses  the  gain  and  phase  axes.   -2
 

Gain » LH jwL » HdBL

Straight Line
•  OSen  a  straight  line  is  naively  drawn  
between  these  two  points  to  define  the  
-4
stability  requirements    boundary  under  the  
assump,on  that  gain  and  phase  errors  
sm =1ê2
combine  linearly.  
 
-6
•  This  straight  line  assump,on  is  not  correct.   gm = 6 dB

•  The  stability  requirements  boundary  on  the  

Nichols  plot  should  be  a  Stability  Margin  
-8
contour  similar  to  the  ones  shown  on  the   -180 -170 -160 -150 -140
preceding  chart.   Phase –° LH jwL HdegL

NASA Engineering & Safety Center (NESC) GN&C TDT 27  

 Blowup  of  Nichols  Plot  for  Example  1  

•  Example  1   Nichols Plot

  0
-165.6°
  7.6 (s 2 +0.1s +0.55)
L (s ) = sm =
1

  s (s +1) (s +20) (s 2 +0.06s +0.5) 4

-2 -2.5 dB
•  The  closest  approach  to  the   Nyquist
unstable  (-1, 0 j) point  of  the  

Gain » LH jwL » HdBL ô

Worst Case
Nyquist  Criterion  occurs  at:     -4

ωm =  0.714  rad/sec  
yielding  the  simultaneous   Straight Line
changes  of  Gain  and  Phase:   -6
Approximation
of 6dB & 30°
Δg = 1.8 dB Requirement
  Δφ = 10.6 deg
-8
•  This  appears  on  the  Nichols  plot   w
as  the  point  of  tangency  between   LH jwL
the  Open  Loop  Transfer  Func,on  
and  the  Nyquist  Stability  Margin   -10
-180 -175 -170 -165 -160
sm = ¼ contour  curve.   Phase –° LH jwL HdegL

NASA Engineering & Safety Center (NESC) GN&C TDT 28  

 Summary  and  Conclusion  

1)  Stability  Summary  

2)  Loop  Design  Tradeoffs  
3)  Stability  Margin  Rela,onships  
4)  References  

NASA Engineering & Safety Center (NESC) GN&C TDT 29  

 Stability  Summary  

•  Stability  requirements  are  oSen  specified  in  terms  of  gain  margin  and  phase  
margin.  These  are  independent  margins  describing  how  much  either  the  gain  or  
phase  alone  can  be  varied  before  the  system  becomes  unstable.  These  two  
margins  are  represented  as  separate  independent  points  on  the  Nichols  plot.  

•  However  the  true  stability  objec,ve  is  to  specify  how  "close"  a  system  is  to  
becoming  unstable  for  any  combina,on  of  gain  and  phase.  This  is  compactly  and  
precisely  measured  by  the  Stability  Margin  which  describes  the  closest  approach  
of  the  Open  Loop  Transfer  Func,on  to  the  (-1, 0 j) point  on  the  complex  plane  in  a  
Nyquist  plot.  

•  Stability  margin  contours  can  be  drawn  on  the  Nichols  plot  to  show  the  Nyquist  
boundary  of  different  combina,ons  of  gain  and  phase  devia,ons  that  can  be  
tolerated  with  a  certain  level  of  robustness  to  instability.    
 
•  The  gain  and  phase  margins,  as  well  as  the  Nyquist  Stability  Margin  contour  
connec,ng  them,  are  defined  in  the  frequency  domain.  The  margins  and  the  
contour  can  be  verified  in  ,me  domain  simula,ons  by  injec,ng  different  
combina,ons  of  gain  and  phase  perturba,ons  from  nominal.  
 
 
  NASA Engineering & Safety Center (NESC) GN&C TDT 30  
 Loop  Design  Tradeoffs  

•  Loop  design  can  be  viewed  as  simultaneously  tuning  the  Sensi,vity  Transfer  
Func,on  to  achieve  disturbance  rejec,on  and  loop  stability  goals  and  tuning  the  
Complementary  Sensi,vity  Transfer  Func,on  to  achieve  performance  and  noise  
rejec,on    goals.    
 
•  S(jω)  Sensi,vity  Transfer  Func,on  (Disturbance  Rejec,on  Transfer  Func,on)  
-  Rolling  off  S(jω)  suppresses  low  frequency  load  disturbances.  
-  However  decreasing  S(jω)  at  low  frequencies  increases  its  maximum  value,  Ms,  
which  reduces  stability.  This  is  the  result  of  the  Bode  Integral  being  Constant.  

•  T(jω)  Complementary  Sensi,vity  Transfer  Func,on  (Closed  Loop  Transfer  

Func,on)  
-  Tracking  Performance  may  be  improved  by  increasing  the  bandwidth  of  T(jω).    
-  However  limi,ng  the  bandwidth  by  rolling  off  T(jω)  is  necessary  to  suppress  
high  frequency  sensor  noise.  
 
 
 
NASA Engineering & Safety Center (NESC) GN&C TDT 31  
 Stability  Margin  Rela,onships  

•  One  of  the  fundamental  goals  of  the  tradeoffs  involved  in  loop  tuning  is  to  shape  the  
Sensi,vity  Transfer  Func,on  so  that  its  maximum  value,  Ms,  is  compa,ble  with  the  
control  loop  stability  requirements.    

•  Recall that the maximum value of S(jω)  is the inverse of the Stability Margin:

Ms = 1/sm sm= 1/Ms  

 
•  Typical  specifica,ons  for  the  maximum  Sensi,vity  Magnitude  and  the  Stability  Margin  
are  in  the  range  of:  

4/3  ≤  Ms  ≤  2    3/4  ≥   sm  ≥  1/2    

 
•  Respec,vely  these  corresponds  to  gain  and  phase  margins  of  approximately:    
 
(12  dB  and  45°)  to  (6  dB  and  30°).  

NASA Engineering & Safety Center (NESC) GN&C TDT 32  

 References  

 
 
 
 
1)  Feedback  Systems:    
An  Introduc4on  for  Scien4sts  and  Engineers  
By  Karl  Johan    Åström  &  Richard  M.  Murray  
Available  free  from  Google  Books  
 
2)  Respect  the  Unstable    
IEEE  Control  Systems  Magazine,    
Vol.  23,  Num.  4,  pp.  12-­‐25,  Aug.  2003.  
By  Gunter  Stein  
Available  free  at  NEN  GN&C  Reading  Room  

NASA Engineering & Safety Center (NESC) GN&C TDT 33  

Learning from the Past, Looking to the Future
Tim Crain – November 2014

Learning from the Past, Looking to the Future Page: 35