Dynamic systems: past history (derivative states) are relevant (): .

() = cos( 1())
Lumped: parameter is constant throughout system process = cosh( 1())
Distributed: system dimensions not equal to small over signal There are a total of n maxima and minima in the filters ripple.

Signal Energy: = 2 (). Can be infinite. The ripple, r, takes the place of .
1 /2
Signal Power: = lim /2|()|2 Gain equation: = 10 log10 [1 + 2 2 ( )]



Sinusoids: energy, finite power. Decays: finite energy&power Gain at : 2 2 ( ) = 1010 1
1
() = (), or () = 2 0 ()

2
1 1 10 10 1

() = (), or () = 0 Order determination: = cosh [
]
cosh1( )
1010 1
Odd and even signal components: () = () + () 1
1 1 (
101
)
() = (() ()), () = (() + ()) Passband frequency: = cosh [ ]
2 2
Periodic: x[n] = x(nT) then, [] = [ + 0 ] () = (( + 0 )T)
Signal Classification: continuous/discrete -> y-axis, Analog/digital -> x-axis Poles of a normalized Chebyshev lie on a semi-ellipse of the minor and major
1 1 1 1
Linearity: Additivity: 1 1 2 2 , 1 + 2 1 + 2 semiaxes: = sinh ( 1 ( )) & = cosh ( 1 ( ))

Homogeneity: () (21) (21)
Superposition: 11 + 2 2 11 + 2 2 Pole locations: = sin [ ] sin() + cos [ ] cosh()
2 2
Test, sub in: [] = 1 [] + 2 [], [] = 1 [] + 2 [] 1 1
1
= sinh ( ) , = 1, 2, ,
Time invariance: Delay input, output and compare, = [ + ]
1 = [] , 2 = [ + ] =
() = =
Causal: 0 , (), 0 ? () + 1 1 + + 1 + 0
Sampling: () = ()() = ()[() + ( ) + ( 2) + ] 0 ,
= ()( ) (Sample Train) = { 0 ,
1 2
() = ( ) (Impulses in Freqd) 1 + 2
Chebyshev Type 2 (Inverse Chebyshev) Filters:
Nyquists Theorem: minimum sampling frequency to recover a signal: > 2( )
2
1

2 . is max signal frequency (). Amplitude response: |()|2 = 1 | ()|2 =

2( ) 1
1+2

Quantisation Error: |[]| Elliptical Filters:
2
Nyquist rate: /2 Nyquist frequency: 2 Amplitude response: |()| =
1

Aliasing Frequency: = | |, where n is the closest integer 2 ()

1+2

multiple of to . Signal Manipulations:

Dynamic Range is the ratio of power of largest signal to smallest noise: = Down sampling: () = ()
10 log10 (

) (SNR) Up sampling: () = (/)


Matrix representation of a convolution via Toeplitz Matrices:
Convolution Integral: () () = () ( ) = () []
[] = = [][ ]
A Discrete-time LTI system is completely defined by the impulse response: [] =
Convolution Shift Property: () () = (); ( ) () = () = [] [ ]

( ) = ( ) [] = {[]} = [] {[ ]} = [] [ ]
Convolution with an Impulse: () () = () , = =
Graphically evaluate a convolution by flipping second function, shifting second BIBO Stability: requires = |[]| < , i.e. a finite impulse response.
function along time axis, evaluating convolution at that shift value as the area under FIR Systems:
the product of the curves. Known as Linear shift invariant, multiply accumulate delay system.
Fourier Series: () = 0 +

=1 cos(0 ) + sin(0 )
General form: = 1
=0 []
2 /2
2
2 /2 FIR systems have no poles and are feed forward exclusively.
= /2 () cos(0 ) , = /2 () sin(0 ) Filter has a length M, and an order N = M 1.

Z-Transform: (()) = =0 ()

= () Filter equation: () = 1 1
=0 ( ) = =0 () ( )
Time Shift: (( )) = () Impulse response: () = () = 1
=0 ( ) =
FVT: lim ()= lim (), lim () = lim(1 1 )() Phase response is a linear function of frequency.
0 k
1
Stability Bound: || < 1
2 3 1
Power Series: 1 + + ( ) + ( ) + =
1( )
Bilateral Transform: [] = = [] , e.g. [] = 0.9 [] +

1.2 [( + 1)] [] = =0 0.9

+ 1
= 1.2
Laplace and Z pole equivalents: =
ROC: =0| | < or |z| > |a|
1

Z Time Shift: [ 0 ] 0 () IIR Systems:

1 1
[ ][ ] [] [ 1][] [] + [1]

Known as linear time-invariant
1
Z Time Reversal: [] [ ] , = 1/ General form: () =
= []


IIR Systems require feedback in their implementation and have a non-zero impulse
Multiplication by : 0 [] ( ) , = |0 | response for infinite t.
0
() Window Types:
Multiplication by n: [] , =

Normalized filters: = 1.
Butterworth Filters:
1
Amplitude response: |()| =
2
1+( )

|()|2 = () ()
The number of poles of the filter is given by the filter order.
Poles lie of the negative real unit circle (for a normalized filter).

Pole locations given by: = 2 (2+1) , = 1, 2, ,

= cos ( (2 + 1)) + sin ( (2 + 1))
2 2

When specifying a filter, replace s with in the normalized filter equation.


Gain at : = 20 log10 |( )| = 10 log10 [1 + ( ) ]
2
Window Design Method:

1. Determine ripple values (same in pass and stop bands)

[(10/10 1)/ (10/10 1)] 2. Normalize frequencies to the sampling rate
Filter order: = log10
2log10 ( /) 3. Determine the peak error from ripple value

Cutoff frequency: = 1 4. Choose Window model from the peak error


[10 10 1]
2
5. Find the filter width (and hence order)
6. Find the cutoff frequency using the midpoint rule
Table: |3||2.0|2.0| -> 3 + 2 2 + 2 + 1 7. Use [] = [] []
Chebyshev Filters: () = 20 log10 (1 + 2 ) , () = 20 log10
1
Normalized Amplitude Response: |()| = ZOH adds a unit delay to the system
2 ()
1+2
1 1 ()
Model for ZOH: () = , () = , [] = (1 1 )( )
: , = 1 + 2 , = 100.1 1
 Second Order Systems:
2

2
2 +2 +
Underdamped (0 < < 1) P = 1 2
Overdamped ( > 1) P = 1 2
Critical ( = 1) P = (repeated)
1
= =
2
Transfer function of a lowpass filter:
2 1
() = =
( 1 )( 2 ) 2
( ) + 2 ( ) + 1

To obtain the filter information for a highpass filter, replace / with /. For
2
2
Bandpass:

Z Table (Laplace time Z):

(+1)() 1
Eulers Method: , Error -> 0 as T -> 0, = ( ) 1 () 1
1
()

1
1. Rewrite the function as a dynamic system: () + () = () + 2(), apply 1
1


inverse Laplace: () + () = () + 2() + ^() 2 (1)^2
1 (1 )
2. Replace continuous signals with digital equivalent and differentials with 1
(+1)() (+1)() (+)2 ( )^2 (+) (1)( )
approximating function (note - acausal): + () = + 2(). 1 ( ) 0 sin(0 )
(0 )
Rewriting: (+)(+) ( )( ) 2 +02 2 2 cos(0 )+1
(cos(0 ))
( + 1) () + () = ( + 1) () + 2() ( + 1) = ( + 1) + (0 )
2 +02 2 2 cos(0 )+1
2( 1)( ) ( 1)()
Z-transforming: () = () + 2( 1)() ( 1)() ()( + + 1) =
() ++1
()( + 2 2) =
() +22
1
Back Diff: [] = [] [ 1]

1
Forward Diff: [] = [ + 1] []

1 1 2 1
Central Diff: [] = [ + ] [ ]
2 2 +1

Lead/Lag Compensators:
+ 1 1
General form: () = , ()
+ 1 1
Lead: || < ||
Zero draws poles to the left; adds phase lead. Pole decreases noise.
Aim: Decrease rise-time and overshoot
Guide: Set a near desired wn, b= ~3-20x a
Lag: || > ||
Near pole/zero cancellation; adds phase lag.
Aim: Improve steady-state tracking
Guide: Set b near 0, a = 3-10x b
PID:
1
General form: () = + + = (1 + + )

1 0 + 1 1 + 2 2
() = + + [ ]=
1 1 + 1 1 + 2 2
Integral: increase dynamic order of signal tracking reduce steady-state error.
Derivative: Increases system damping.
2 1
Tustins Method: =
+1
Matched Pole-zero Method:
Replace poles/zeros in continuous domain with z equivalent.
+
Need to also scale discrete system by DC gain.
If () order > () order, multiply () by ( + 1) until equal order.
1.8
Rise time (10% - 90%):
0

Overshoot: = 12
4.6
4.6
Settling time (to 1%): = = , 0 =
0
Phase Margin: 100
State Space:
: state vector | : output vector | : input/control vector | : state matrix | : input
matrix | : output matrix | : feedforward matrix Control canonical form: = +
= +

= [ 1 ] = [ 1 ] = 1 1 + 2 2 + +

0 1 0
1 1 + 2 2 + + 0 0 1
() = = [ ]
+ 1 1 + 2 2 + +
1 1
0
0
= [ ] = [ 1 1 ] = 0

1

Filter Magnitude Response:

[ ] = 0 [ ] =
If you want to build a ship, don't drum up the men to gather wood, divide the work
and give orders. Instead, teach them to yearn for the vast and endless sea. -Antoine
Sallen Key: Low pass, High pass, Band Pass. de Saint-Exupery