Theory

Consider the PID controller configuration with the unstable Maglev plant shown in Figure:

Maglev plant

𝑅 𝑠 𝐸 𝑠 1 𝑢 𝑠 𝐷𝐶 𝜔𝑛 2 𝑌 𝑠
+ 𝑘𝑐 1+ + 𝑃 𝑠 = 2
𝑇𝑖 𝑠 𝑠 + 2𝜉𝜔𝑛 𝑠 − 𝜔𝑛 2
- -

𝑇𝑑 𝑠

PID Controller

The PID controller in above figure has the derivative mode acting on the output instead of
the error . As described earlier, this is commonly done in industry, to prevent the
“derivative kick” from happening for step changes in setpoint signal R .
We can derive the closed loop transfer function CLTF between and R as

Y s  DC  n2 Kc  s  1T 
CLTF   3  i
(1)
R  s  s   2n  Td  DC  n  s  n  K c  DC   1 s   DC  K c n2 Ti
2 2 2

If we compare the denominator of CLTF in (1) in terms of the standard second order system
along with a pole pCLTF , we have

Kcn2  DC 
s   2n  Td  DC   s    K c  DC   1 s 
3 2
n
2 2
n
Ti (2)

  s  pCLTF  s 2  2CLTFnCLTF s  n2CLTF 
where the term + is chosen (or designed) to cancel the numerator term( + 1⁄ ) in
equation 2 (so that we get a standard second order form that is easy to use for design).

The RHS of (2) can be expressed as

2 2 2
+ (2 + ) +( +2 + ) + (3)

From (2) and (3)

s3   2n  Tdn2 ( DC )  s 2  n2  Kc  DC   1 s  Kcn2 ( DC ) Ti

(4)
   
 s3  2CLTFnCLTF  pCLTF s 2  n2CLTF  2CLTFnCLTF  pCLTF s  n2CLTF pCLTF

Comparing the coefficients of corresponding powers of s on both sides of equation 4 gives

o Constant: n2CLTF pCLTF  Kcn2 ( DC ) Ti (5)

2 2
o s term: −1 = +2 (6)

2 2
o term: 2 + =2 + (7)

From equations 5, 6, 7, we obtain

n2CLTF
Kc  (8)
n2 ( DC )

2CLTF nCLTF
Ti  (9)
n2

2CLTFnCLTF  2n  1
Ti
Td  (10)
n ( DC )
2
From the given PO and using the well-known relation

=√ (11)

we can obtain the corresponding value of .

From the given and using the approximate relation
4
ts  (12)
CLTFn CLTF

we get

= (13)

where is found from (11).

Having obtained from (11) and from (13), we can find the required K c , using

equations 8, 9, 10.

The steady state error (or offset) to step inputs is zero, due to the presence of integral action.

Remark: It is important to note that the relation (12) is quite approximate. See any standard book on
control systems, or the references link on the main page of this site.