# 8.

3-1
8.3 Generalize to CPM [P4.3.3]

8.3.1 CPM Signals

• Generalize from CPFSK to continuous phase modulation (CPM) by
allowing:
non-rectangular frequency pulses
frequency pulses with durations , 1 LT L >

o Example: 2REC – and we can go to L-REC:

o Example: 2RC (raised cosine) – and we can go to L-RC:

8.3-1

8.3-2
o Example: Gaussian MSK (GMSK), with parameter : BT
The frequency pulse is a Gaussian filtered rectangular pulse:

The frequency pulse has area 1/2, as usual:

• Why bother with CPM?
o Smoother transitions, so lower bandwidth;

o Slower rise of phase pulse (takes symbols to get to ½) means
smaller frequency shift, so lower bandwidth;
L

o Longer rise of phase pulse to ½ means a chance of increasing
between alternative signals
min
d
8.3-2

8.3-3
• The generic CPM transmitted signal is then
( , )
( , ) 2 ( )
( , )
k o
k
j t
t h I q t kT
s t P e
φ
φ = π − +
=

I
I
I
φ

where is usually taken to be zero with coherent detection.
o
φ

• Typical CPM signal trajectories:

o First, the total phase ( , ) t φ I

o Next, the complex signal : ( , ) s t I

8.3-3

8.3-4
8.3.2 CPM States and Trellis

• The state description of general CPM is more complex than for the CPFSK
special case. Consider the effect of the past on ( ) t φ in interval n,
, shown below for 3RC, which has ( 1) nT t n T ≤ < + 3 L = .

All symbols are
taken as +1 here.

1 1
1 1
1
1
current pulse
phase state correlative state , ,
the state ( , , , )
( , ) 2 ( ) 2 ( )
n n n L
n n n n L
n
k k n
k n L k n L
I I
I I
t h I h I q t kT hI q t nT
− − +
− − +

≤ − = − +
θ
σ = θ
φ = π + π − + π −
∑ ∑
I

o The phase state is not the same as the signal phase . It is the
sum of all saturated phase pulses, up to and including .
n
θ (nT φ )
n L
I

o The correlative state comprises all past symbols for which the phase
pulse is still changing across interval n.

o And, of course, the remaining influence is the current symbol
n
I .

8.3-4

8.3-5
• So the state affecting interval n, ( 1) nT t n T ≤ < + is

1 2 1
phase
correlative state
state
, , , ,
n n n n n L
I I I
− − − +
⎛ ⎞
⎜ ⎟
σ = θ

⎜ ⎟
⎝ ⎠

, summarizing relevant history,

and the state update is

( )
1 1
mod2
n n n L
hI
+ − +
θ = θ + π π

( )
1 1 1
, , , ,
n n n n n L
I I I
+ + − −
σ = θ …
2 +

The allowable values of the phase state
n
θ are the same as in full response
( ), like CPFSK. 1 L =

• Example: binary transmission with 2REC

The phase pulse is

o The phase in interval n, ( 1) nT t n T ≤ < + is determined by
(a) the new bit ;
n
I
(b) the legacy from the past:
1 n
I

, which is still causing the phase to
change; and , which is static, from saturated pulses.
2
k
k n
h I
≤ −
π
∑ n
= θ
So the state is
( )
1
,
n n n
I

σ = θ .
8.3-5

8.3-6

o The phase state is determined by
n
θ
2 n
I

and previous bits. The
possible values of are determined by h alone, not L or M.
n
θ
For 2 3 h = :

o Transitions: Since , the
1 1 − n n n
h I
+
θ = θ + π
1 n
I

part of the state
(
1
,
n n n
)
I

σ = θ tells you whether
1 n+
θ will increase from or not.
n
θ
For example, (0,+1) must go to (1, x)
(0,-1) must go to (2, x)

o Sketch a trellis using state
( )
1
,
n n n
I

σ = θ

8.3-6

8.3-7
• Caution: Unlike CPFSK, the phase (nT) φ at the start of interval n is not
sufficient itself as a state variable.

o If it were, then the phase trajectory ( ) t φ in ( 1) nT t n T ≤ < + would be
determined solely by and the data (nT φ )
n
I .

o But in is affected by previous symbol values in
the phase state and the correlative state, and
( ) t φ ( 1) nT t n T ≤ < +
(nT) φ cannot
unambiguously represent both components of the state.

o Example: 2REC, 2 3 h = , again. We have
1 1
( ) 2 ( )
3
n n n n
nT hq T I I
− −
π
φ = θ + π = θ +
The phase can take on 3 values: (nT φ )
Solid dots: phase state
n
θ
Open dots: the 3 possible values of
starting phase ( ) nT φ

These phases cannot unambiguously represent
n
θ and
1 n
I

. For
example, phase 0 can arise from

0
3
π
+ (
1
1
n
I

= + ) or from
2
3 3
π π

( ).
1
1
n
I

= −
8.3-7