MSK Tutorial
MSK Tutorial
n
w
To
Prepared By Mario Antonio Aldera
B.Sc. Electrical and
e
Electronic Engineering, Cape Town.
ap
C
of
ity
-i-
c_-:cc-:-:::c-:cu"'"-C -.-"=::oc.·:=-:.>•.,.··;,"~''"'··cc···•• i'·· .
.•'» · • · "'··")",
n
w
quotation from it or information derived from it is to be
To
published without full acknowledgement of the source.
The thesis is to be used for private study or non-
e
ap
commercial research purposes only.
C
of
DECLARATION
-ii-
ACKNOWLEDGEMENTS
f it " j .• i ' -
Dr. R.M. Btaun (Thesis ·supe~~isot);
I
Mr. M. G. Alder a for·' the'Y loan of ' the ' personal; computer on
which thi's· thes'is was. prepared;
·or. L.A. Alaera· for the photogr'!lphy~ which made the thesis
complete.
•.
,.
'
ABSTRACT
-iii-
modulation index tolerance sets the minimum attainable error
probability, independent of the signal-to-noise ratio.
The constructed MSK . modulator was found to of fer
satisfactory performance, and its hardware implementation
was simple. The coherent demodulator was more complex in its
implementation, and several modules in its structure were
designed, constructed, and their performance evalu~ated. The
non-coherent MSK demodulator was found ·to be simple to
construct, but it was theoretically shown that it is more
than 3 ·dB worse off in noise performance than the more
complex coherent demodulator.
-iv-
TABLE OF CONTENTS
Title Page i
Declaration . ii
Abstract iii
List of Illustrations ix
Nomenclature xiii
Chapter 1 INTRODUCTION 1
-v-
2.2.4(c) Reduction of MSK to SQPSK Format 34
2.2.4(d) Generation of MSK 35
· 2.2.4(e) Coherent MSK Demodulation 36
2.2.4(f) Spectral Efficiency 40
-vi-
5.2.2(a) Voltage Controlled Oscillator Module - 95
5.2.2(b) Frequency Doubler Module 97
· 5.2.2(c) 270° Phase Delay Module 98
5.2.2(d) Phase Detector Module 99
5.2.2(e) Loop Filter Module 101
5.2.3 Phase-Locked Loop Performance Evaluation 105
5.3 Symbol Clock Recovery Module 110
. 5 .4 Adder/Subtractor Module 112
5.5 Coherent Down-Conversion Module 114
5.6 Detection Filter Module 116
-vii-
chapter s CONCLUSIONS 159
-viii-
LIST OF ILLUSTRATIONS
Title
-ix-
3~4 Degradation Due to Filtering on MSK and SQPSK 57
-x-
5.17 Carrier Phase Noise Measurement 108
5.18 MARK Frequency PLL Output Spectrum 109
5. 19. Symbol Clock Recovery Module 111
5.20 Adder/Subtracter Module 113
5.21 Regenerated Quadrature Phasers · 114
5.22 Coherent Down Conversion Module 115
5.23 Design Curves for Lowpass Detection Filters 116
5.24 Detection Filter Module 117
5.25 Constructed Coherent. MSK Receiver 118
-xi-
7~16 PLL Open Loop Transfer Function 151
.7.17 Loop Filter Circuit 152
7. 18 Recovered Clock Spectrum 154
7.19 Static Phase Adjustment Circuit 155
7. 20 Data Latch Module 156 .
7~21 Constructed Non-Coherent MSK Demodulator 157
-xii-
NOMENCLATURE
-xiii-
ISI Intersymbol Interference. ISI is the
interference of adjacent pulses with
the amplitude of the present pulse.
PD Phase Detector.
-xiv-
oo
Q(y) = Jy e-x2/2 dx
si ( f) =
:~n (rrf) ·
sine (f) = ------
rtf ·
-xv-
CHAPTER 1
INTRODUCTION
-1-
coherent MSK modem.
-2-
l
CHAPTER 2
-3-
H{f)
TRANSMIT End RECEIVE End
9--------1
- 1·0
I t t 1 1' ~
Ts Ts Ts Ts t
B f
<p(f)=kf
1 ~ S!t-nTsl
..
=
J~B 1.ej:2n:ft. df
-4-
= 2B s i ( 2 rtB t )
h(t)
28
t
-3/28 -1/8 -1/28 1/28 1/8 3/28
0
h( t) = l/T.,, si (n:t/T.,,).
-5-
of the received pulse h(t), the responses due to all the
other input impulses pass through zero, thus avoiding
intersymbol interference (!SI) which can result in detection
errors at the receive end.
n = f 5/B = 1/Ts . 2T ..
= 2 B/s/Hz.
0(f) = k.f
=> 0'(f) = k
-6-
A skew-symmetric transmittance function, Yi(f), is added to
the brick-wall amplitude characteristic of the low-pass
filter, with Yi(f) defined by the relation;
-7-
----------------------------------- ---
0(jw) ~kw
-8-
+5
-5
-10
-15
iii
:=. -20
::c
-25 1 - . - - - - + - - · er= 0.5
I<
""
.!! 1 - - + - - - - - - + - - er= 0.3
C>
N
II
>
-30
m<t tt ttitttt?fl----t--t----1-- er= o'
-35
IN = Nyquist frequency
-40 = i of symbol rate
-9-
pulses. The most important feature of the eye diagram is the
eye opening, and it is desirable to have a fully-open eye
(loo·% eye opening) as ·this means that no IS! is present.
Fig.2.4 illustrates an eye diagram for a system with a
filter oc = 0.3 [2.1].
-10-
Rec~lling the well known result of the ideal matched filter
[2.2], we know that the optimum receiver filter for an input
pulse shape p ( t), corrupted by AWGN of two-sided spectral
density Sn(w) = N0 /2 [W/Hz] is given by;
where,
... ( 2. 2)
= 1/2rr J: 00
IP(w) j 2 dw ... ( 2. 3)
-11-
made by the decision threshold. If y(T 0 ) is positive, the
data bit is assumed to be a "!", and if it is negative, it
is assumed to be a ~0". When this is completed, the
integrator is reset to zero, and the process repeats.
Because of its operation, such a receiver is often referred
to as an integrate, ~ample and ~ump (IS&D) receiver.
AWGN
n(t) p(t)
Message
To y(t) y(TO ~ DECISION Data Out
----+-l-----..
±p( t)
±pit)+ n(t) 1
0
dt
THRESHOLD
... (2.4)
-12-
.
where Epq = IT ° p(t).q(t) dt
0
-13-
2.2 BANDPASS (CARRIER) SYSTEMS
AWGN
n(t)
+A rl
-A_J L@ D>O ~ '1' sent
®
Random D<O :;. 'o' sent
Polar NRZ
Input Sample at
t=Tb
cos(w,t + 90 ) cos(WJ + 90 )
Carrier Local Oscillator
-14-
changes by ±180°, depending on the input data value. The
transmitted signal is then corrupted by AWGN in the
channel before it is ·received at the receive end. It is
then mu! ti plied by the local oscillator of the receiver
(which is precisely phase-locked to the carrier at the·
trans~itter), before being processed by the IS&D.circuitry
-15-
Nyquist filtering to the signal. It was shown previously
that ~ = 0 filtering allows for the transmission of pulses
at· baseband at a rate f 0 through a bandwidth f 0 /2. Were
such a low-pass filter to be placed at point (A) · in
Fig.2.6, then the output spectrum of the modulator would
extend . from f 0 f 0 /2 to f 0
+ f 0 /2. We are therefore
transmitting f 0 bits per second through a bandwidth of f 0
Hz, yielding a maximum spectral ~fficiency for BPSK of
1 B/s/Hz. Ii is also possible to perform the filtering at
point (B) in Fig.2.6. The filter would then be a bandpass
filter of bandwidth f 0 Hz, which is commonly referred to
as the Double-Sided Nyquist bandwidth. In practice, the
spectral efficiency achieved with BPSK is typically 0. 8
B/ s /Hz [ 2 . 3] .
-16-
Q
'o' .,.
-17-
Sample
I-Channel I-Channel at t=Ts
!.A
AWGN
n ( t)
+
DEMUX · MUX
Sample
at t=Tb+ T5
O-Channel 0-Channel
-18-
•.
Th
+A -
,. Input
Data -A ___;___
1 2 3 4 5 6 7 a 9 10 - 11
+A
I-Channel
-A
_J Ts ·1 I5 I9
1 1 3 3 5 7 7 9
+A
Q-Chanriel
-A
2
---
2
I4 4
I. 6 6. a 8
I 10 10
-19-
are combined by means of a parallel-to-serial converter
which yields the output data stream at a bit rate of fb
Bits/s.
-20-
baseband symbols with a low-pass filter of cut-off
frequency f 9 /2 (Hz], or with a band-pass filter at RF with
a bandwidth of f 0
(Hz]. This enables us to transmit at a
rate of f 0 B/s through a bandwidth of f 9 = f /2 (Hz]. We
0
•.• (2.9)
01 00
• •
11 • • 10
-21-
are not "Staggered"). Considering the constellation
diagram in Fig.2.12, it is apparent that as the two
channels can both change at the same time, it is possible
for the' modulator to change from the state "00" to "11",
or from "01" to "10". This results in the carrier passing
throu~h the origin of the diagram. i.e. the amplitude of
the carrier is reduced to zero at certain times. This
results in the QPSK waveform having 100% AM modulation
present.
-22-
·regeneration [ 2. 4]. The error probabilities and Spectra
for QPSK and SQPSK are identical under linear conditions.
Hod. Index
Control
Polar NRZ
----
Random
vco
Data
h = ... ( 2. 10)
-23-
where;
h = FM modulation index
~f = peak-to peak frequency deviation of VCO
fb= bit rate of modulating waveform
Fig. 2 .14 shows the frequency map of the VCO. Note that
this is not the spectrum of the generated CP~FSK.
MARK SPACE
frequency frequency
Af /2 M/2
0 f
-24-
... ( 2 .. 11)
where;
... ( 2. 12)
-25-
I There are two discrete spectral lines in its
spectrum, which indicate that some power is
wasted ·transmitting two carriers which
convey no data;
and
T0
= p(t).q(t) dt
J0 .
. .. ( 2. 13)
-26-
frequency is much larger than the bit rate
hence the right-hand term in 2.13 may be
ignored. Equation 2.13 then reduces to;
... ( 2. 14)
1·0
-0·217
-27-
demodulator structure for CP-FSK. This was deliberate as
it is not generally possible to construct coherent
receivers for arbitrary h values, only for h being one
half, integers, and integers plus one-half [ 2. 6]. It is
far simpler to demodulate CP.-FSK by means of discriminator
detection [2.7], but then the phase information contained
in the carrier is lost (with the subsequent degradation in
error performance). The results presented in [2.7] suggest
that the optimum value for h on wideband channels lies
between 0.7 and 0.8.
rolloff of the BPSK and SQPSK spectra. The reason for the
more rapid rollof f is inherent in the continuous phase
nature of the CP-FSK waveform. The BPSK and SQPSK schemes
allowed for abrupt changes in carrier phase, whereas the
CP-FSK scheme gradually changes the carrier phase over a
bit time while preserving phase continuity at the bit
transition boundaries. This aspect of CP-FSK will be more
fully discussed in section 2.2.4.
-28-
YcrrsK(w)
A 2 Tb/2
-29-
·immediately tells us that the frequency map for MSK
consists of two frequencies spaced by;
6f
=> where N is an odd integer • • • ( 2 • 17 )
2
fm = f0 - l::i.f /2
= 70.144 - (0.5)(0.5)(2.048) [MHz]
= 69.632 MHz.
f9 = f 0 + l::i.f /2
= 70.656 MHz.
-30-
where e~, the recursive phase constant, is calculated
from;
... ( 2. 19)
0(t=O) = ek
and
-31-
100°
9{)0
Additional
Phase 0°1'---+---oi.---+----,f'---"'t---+--t•
6Tb
-90°
-1ao·
fm = (136)f 0
and
f 9 = (138)f 0 •
-32-
It would thus appear that MSK is at a 3 dB disadvantage in
error performance compared to antipodal signaling (i.e. we
would need to double the average energy per bit-to-double
sided noise density ratio µ in order to obtain the same Pe
as BPSK. or SQPSK).
Pe(MSK) = Q(v[4E/2N 0 ])
= Q(v'2µ) . . . ( 2 . 21 )
-33-
with modulation indices unequal to multiples of O. 5 are
difficult, if not impossible to construct. MSK is
therefore a compromise. between excellent noise performance
and receiver complexity.
where;
-34-
power spectra.
Input
Data PHASE- CARRIER
DEMUX LOCKED Osc.
f=1/4Tb fc
fb Clock
-90° -90°
f5 =fb/2
DIFF.
ENCODE YQ(tJ
Q-{hanne\
-35-
In practice, this may prove to be a complex method of MSK
generation, and a simpler implementation is to use the.CP-
FSK visualization of . MSK to generate it. This will be
further discussed in Chapter 4.
+
PLL
.-72
2fm
fb Clock
PLL
72.
2fs 1-Channel
DIFF.
IS~D
DECODE
( 1•
Q-Channel
-36-
later that the MSK spectrum contains no di~crete spectral
components (making it an efficient signaling scheme), and
so a non-linear operation (squaring) is necessary to
regenerate these frequencies. In fact, by squaring the MSK
signal, the two spectral lines 2f~ and 2fe appear out of
the continuous spectrum. APPENDIX A shows mathematically
how this arises. The squared spectrum is actually Sunde's
FSK (with h = 1) as . the frequency· doubling· effectively
doubles the MSK modulation index to 1. 0. A plot of the
squared MSK spectrum is to be found in Fig.5.3.
we get;
-37-
Using the simple trigonometric relation cos ( cx:+J3) +cos ( cx:-J3)
= 2.cos (cx:).cos(/3) the expression reduces to;
and
SMsK(t).So(t) in the Q~Channel.
... ( 2. 23)
-38-
Similarly, the Q-Channel product becomes;
... ( 2. 24)
Symbol
Clock
I-Channel 0
a-Channel O
-39-
quadrature ~nd that each channel yields one-half the
transmitted bits. The falling edge of the symbol rate
clock ( f"' = 1/2T0 ) as shown can be used to sample !-
Channel at maximum eye opening, while the Q-Channel could
be sampled on the rising edge of the clock.
The bit rate clock is obtained from the 2fm and 2fa
frequencies by the relation;
... ( 2. 25)
... ( 2. 26)
-40-
0
~l-'-<-+-+-1-·Hl-+-+l-4~~4+:-!+>
1.0 2.0 ),O 4.o S.ti, 6.0 1.0 S.o 9~1> 10.a
(l' .. j' c }T ... ftQfU\.\l..Iim tAE:aVDICT 011'Stt no.: CAktUER (KZ./Jll1'/S£C)
-41-
pulse shapes of the schemes. MSK uses sinusoidal symbols,
while SQPSK uses rectangular symbols. Intuitively, one can
se~ that the sinusoid has more continuous derivatives than
the rectangle. This implies that it is smoother than the
rectangular waveshape, hence more power will be
concehtrated in the lower frequency region of the spectrum
(hence the. rapid spectral rollof f) . An interesting
correspondence [ 2 .14] provides a simple proof that any
·baseband pulse shape satisfying the same conditions as MSK
does (i.e. symmetric etc.) will always have a wider main
lobe than that of a rectangular pulse of the same
duration. There clearly exists a trade-off between pulse
smoothness (which· sets the main lobe width) and rate of
spectral rollof f. In closely packed radio channels, MSK
may well be at a disadvantage relative to SQPSK which has
a narrower main lobe width [2.12].
B .
P 0 b = 1 - {
I-B
GMsK(f) df . .. ( 2. 26)
-42-
0
~-1--·-~-
__l_µ~:·
;+t~fUtlit~~nttH
~
:-JJ~-H-~4+-kl~~J~i+tt
0
c z.o ll,:i 6.o
:!; -
e.c• 1c.o 12.0
7'•'0•Sl:l!:~' ~01\.".A:.!ZtO M.:.~'l:JTI!
1i..o
(~/!IT/StC)
1e.c
.
1s.o 20.0
-43-
cos (rrf/f-,,,)
1 - ( 2 f I fa ) 2
H( f) = ~[l - sin (rr/2oc)(2f/fs l)]
cos (rrf/f.,.)
$ (l+oc)f.,./2
0
... ( 2. 27)
•.•-.-.,...----.,...----..,.-..,.-----,...---~----.
... ( 2. 28)
-44-
The filtering need not be Nyquist filtering, and Chebyshev
filters, for exampl~, may be used [2.17]. On~ must expect
a certain amount of performance degradation, but this may
be minor when compared to the simplicity of the non-
Nyquist filtering.
.. ... - 01
.... ....
/ "' ' '\
I
I \
11 I \ 00
I
I
\ I
\ I
-45-
performed. Note that as a consequence of the Gray coding, a
symbol error only results in a single bit error (as opposed
to 2 bit errors were the coding not Gray). The probability
of a bit error is thus one-half the probability of a symbol
error. This fact should be borne in mind when viewing MSK as
an M-Ary PSK scheme (where one always calculates in terms of
symbols, not bits).
-46-
REFERENCES
-47-
Vol. COM-20, June 1972, pp.466-470.
-48-
and probability of error performance of MSK and
offset QPSK systems", IEEE Trans. on Comm., Vol.
COM-27, Dec. 1979, pp.1794-1801.
-49-
CHAPTER 3
3.1 INTRODUCTION
-50-
0. 5. A deviation from this value will result in
performance degradation and an analytical
investigation · into the effects of incorrect
modulation index will be presented.
TRANSMITTER
Fl F2
LIMITING
AMPLIFIER
RECEIVER
NOTE:
[!)- PRE LIMITER ~ILTER, F 1 , IS 4 POLE CHEBYSHEV,
0.1 dB RIPPLE, OOUBLE SIOED BAllDllllDTH B.
-51-
In this model [3.1], the pre-limiter filter F1 (which
performs the main bandlimi ting at IF) is ,a 4-pole O. ldB
ripple Chebyshev filter ·Of 3dB bandwidth B Hz. The limiting
amplifier produces a constant output envelope with no phase
shift from input to output. This is an accurate model of
ILA' s used in terrestrial microwave links. The output RF
filter (F 2 ) is modeled as an ideal brick-wall filter of
linear phase and having 50% excess bandwidth (the sharp cut-
off can only ·be approximated in practice). The input filter
of the receiver, F3 , is such · that it presents a Raised
Cosine spectrum of 50% excess bandwidth at the decision
threshold input (i.e. oc = 0.5). This model was implemented
on a computer and the results were obtained by simulation.
-52-
•
;;
~ • •. ,
..
;rir 11'•1 l .. t f:J t <r IOuh l,..• nd*d
{ -lO
di 11o.t114w 1t t11.
p
j
"'"' l"I t f'l fo 1u : ' ·:-o>c
r.,J .. Cl:.1·~r """'· •
'Jl'f'J('.
•
f ••• J.r.
'·'
!re~ .
er .,l
(1)
;; i
'.!,
,.
>
::
> •
- 1'
; ...
••
~ •1(1
.
c
.,.
-4(• ' -~I)
~ i -lt
'
i ... 1.0
Cb)
;r b}
,,.
: , ... i
'-
. ,~
.., !.CJ
(<)
.., : r., f..
,: r h)
., .,
!
.. · 10"
.. 0
•~ r.
·'' ·••
.• _,. -~ •lt
.,.
g
:t
'
k.
.....
..
-:·Ii
••• , -•t
>
1
C·.;. J. t• LI
: .... ,. l u. ., : .:i I.'
:~r t; .
<'1.l E;)
(di (•)
-53 -
( c) After filtering with BTb = 0.75, the first sidelobe is
·attenuated to 51dB below the main lobe, but is
partially restored to 24dB below after the hard-
limiting.
(d) In all the cases above, the main lobe of MSK is·nearly
fully restored after hard-limiting.
-54-
+11
~.
-1 ~ - ,,,.
I
Filtered then llmilerl
Unf i l tercd
Fi 1 tered
One .can see that the effect of filtering the MSK waveform is
to smoothe out the hard transitions, such as those in the
time interval (t 1 ,t2 ) in the I-Channel. The reason for the
smoothing is that the filtering removes the high frequency
components of the waveform, thereby eliminating the sharp
transitions. This, in effect, reduces the spectral occupancy
of the waveforms which is the reason for the filtering
function.
-55-
s~ectral qualities. However, it was shown above that pre-
filtering modified the symbol shapes (it smoothed them), so
the vector sum of of the I-and Q-Channel symbols at the
limiter input is no longer constant. One must therefore
expect the hard-limiter to affect the MSK spectrum.
-56-
the signaling energy results in degradation of the modem
error performance.
2
c..""
"'
z
Vi
c
c OPSK
-~
"C
"'~.
"'
Cl
0.5 0.6 0.7 0.8 0.9 10 1.1
MSK
-57-
practice, as the filtering used guarantees that an oc = 0.5
Raised Cosine spectrum is present at the decision threshold
input. Some authors realized their channel filtering by
means of 7 Pole, O.ldB ripple Chebyshev filters [3.2,3.3].
The results presented in [3.2) are shown below in Fig.3.5.
I I
I I · - - - · MSK
I I
I I - - OFFSET OPSK
_, I I
<
0 I I
0 I I
c.
;:: I I
:z
< I \
,_o2 \ \
w \ \
....~ \
<
_,
"'0::
:z
0
;::
<
0
<
0::
"'1
"'
0
w
u
:z
<
~
0::
~
0::
' ,,
' ' Pe • 10"'
-:;-------------l
' ' P• • io-7
..."' Pe• 10-4 '~ ....
""~
oL._________i__ _ _ _ _ _ _ _ _--~
_~:::========l
O.~ 1.0 2.0
BT - CHANNEL BANDWIDTH NORMALIZED TO BINARY
DATA RATE (HZ/BIT/SEC!
-58-
3.4 THE EFFECTS OF INCORRECT MODULATION INDEX ON THE MSK
ERROR PERFORMANCE ·
-59-
REFERENCES
-60-
CHAPTER 4
-61-
ru
2· 048 MHz
PRBS
Generator
DIFF.
ENCODER - ~
LTC vco
MSKOutput
Yeo Vc1
-62-
The generalized configuration of a PRBS generator is shown
in Fig.4.2.
. I
7
---.. 1 2 3 4 5 6 7 _,, •
I n. ,, . m
I
I
I
PRBS 0 utput
{l
''
-63-
K = 2 16 - 1
= 65 535
+SV +SV
9 9
Clock
8 ---a
Din 74LS164 Din 74LS164
1 1
+ SV 2 O.o 0. H f21..----~0.E~o.G_o._H .....1
1-6--.lr-13-'
L---------. I 10 12 13 pRB s 0 ut
:=51\1-----+SV-~ .,_______
\e_____~---------<:ff.1.-----------'
74 L 586
~~~ -~~~-i~::~~-] 2 00
p PRBS (f) - K [ f E S(f-n/KT 0 ) + (l/K 2 )S(f)
2 rr To n=-oo
n<>O
... (4.1)
where (form= 16),
-64-
K = 65 535
and
T0 = (2.048 E6)- 1 = 488.28 ns.
-65-
.4o PRBS SPECTRUM MKR A 4. 10 MHz
REF .0 dBm ATTEN 10 dB -.22 dB
PEAK
LOG ..
10
dB/
MA!=IK ;:R A
.
4. 10 MHz llA!vi~ ~(\
/'
. I
~·
-.22 dB
. •• I
I~
~
~
~i I \ ·( "'\
( \
,.,,.,,
VA
FC
SB
\ I ~ ~
'
I
f
CORR
-66-
where yE(n) is the present (encoded) bit, yE(n-1) is the
previous encoded bit and YL(n) i$ the present data bit to be
encoded~ The symbol 8) represents modulo-2 addition
(equivalently the EXOR function). Fig.4.5 shows the TTL
realization of the differential encoder.
YE (n)
..-~~~~~~JI~'--~~~~~
>---_._I--ID o-
74LS86 C-lk-~74LS
74
Note that the coder converts the NRZ-L data bits into NRZ-M
format, and this does not affect the spectral properties of
the generated MSK waveform.
-67-
Good high speed switching performance (at 2MHz)
Fast settling time (minimal oscillation)
·Low output impedance (to achieve a high slew rate)
Output voltage variable between lV and 14V
High input impedance voltage Control inputs
1· TTL compatible input.
ISV
OV__J
-68-
+15V
100 22u
--- ...
2N2222
100p Tl P31 C
2N 2907
1K5
BFY90 Output
74HC14
68
MJE 2955
-69-
capacitor and aids in the faster switching of the second
transistor [4.3]. Note, in general, the low resistance
values used in the the circuitry: this is to reduce the RC
time constants in the circuit and hence increase the
switching speed.
When the input to the LTC is low (OV), the first transistor
is OFF, and so is the second~ The output is then provided by
the 68 Q collector resistor of the second transistor, which
is supplied by a voltage level Vco (note that the resistor
is low in value in order to decrease the output impedance of
the module} . When· the input to the LTC is high ( SV), the
first trans is tor is ON, and so is the second. The second
transistor is saturated, hence the output of the circuit is
z Vc 1 - 0.2V. Note that due· to the asymmetry of the
the output stage, one expects the rise-time to be shorter
than the fall-time (the output is active pull-up, and
passive pull-down).
-70-
st~bility, while the +15V Vee supply is also well decoupled.
50 Q output impedance.
-71-
762
nu
1n +SV
270
88601 I rp
I47p
-72-
not damped, would introduce ringing on the control voltage
waveform (thus generating spurious FM modulation). A series
damping resistance of 270 Q is included, and this provides
the correct amount of damping (without degrading the slew
rate) .
-73-
to be inadequ~te as the sharp transitions on the modulating
waveform cause current surges in the circuit and affect the
MOSFET' s operating point (causing unwanted AM in the VCO
output). The MOSFET quiescent current was optimized by
modulating the VCO with a 2 MHz square wave and observing
its output on a spectrum analyzer. Were no AM present, the
resultant Bessel Power Spectrum would be symmetrical about
the carrier frequency. With AM present, however, destructive
interference ·occurs on the odd lower sideband spectral
components as the AM spectrum is even, and the Bessel
spectrum has the property;
-74-
Fig.4.9 Modulator VCO Transfer Function
-75-
constant (zero) input voltage is shown in Fig.4.10.
MARK =:R D.
1 121 . l2J kHz I
I
-41. 24 dB i I
I
I
I
I I I
I
I I N IV II
- I
I ~h~' l \ I
I
I
I i i !
VA
FC
SB
I I
I
I
~
\1~ I
I
v~,J\ !
I
l\N.JVV~' I I vv~J\J,
l~I·
CORR I I
~""u~
~ I ~ J/J'NJI!
'\...
,J. ,...._ ~
I 1M '·"al.I. ~ I
>IV V r '
·~~~
I I
i I I I I I
CENTER 62.16 MHz SPAN 512Jl2J.0 kHz
RES BW 1 kHz VBl.-.J 112!12! Hz SWP 15 sec
-76-
Fig.~.11 Completed MSK Modulator
-77-
MKR A 3 . 13 MHz
ATTEN 20 dB -1.69 dB
I
I.
. II
I .
MARKER A
3.13J MHz
wf\11
'
I
i
l
~
I
I
-78-
lobe level, in good agreement with Feher's results
shown in Fig.3.2.
-79-
REFERENCES
-80-
CHAPTER 5
PLL
2f rn
-72 + l:
( )1. "!-'
~ _ff
PLL
2f s
-7 2 z:
IF AMP +2
&. @
BPF I-Channel
"'X... DIFF. ©
~
DEC.
MUX
DIFF.
DEL
-81-
(1) The Frequency Doubler Module;
-82-
The nominal drive level for this module is 0 dBm. The
circuit diagram is shown in Fig.5.2.
+5V
560
TO Oownconversion Module
..
r
220p
SBL 1 To MARK
I PLL
MSK Input • L +5V
560
1
•
R 220p
68 To SPACE
I PLL
MAR1
-83-
the output of the module.
LOG
II (>
10
. dB/ I
I
I I •
I I I ~ri,J i I I 1
I
I ,/;J;/y.,~o
. fvlvr ,, v'V\ Vi
tJ' I!
I
I I
I
1·
I
\
' I
'\ ; ~. ... .r~\IW,
J
VA SB
FC
1
CORR
The two spectral lines are clearly seen, and they are at the
frequencies;
-84-
:Note that the· lines are spaced 2. 05 MHz apart (as read off
from the markers on the plot), and this agrees well with the
predicted value of 2.048 MHz (equation 2.25). One can also
see that the · ef feet of the frequency doubling is to widen
the main lobe to 6 MHz, which shows that MSK has a 50%
narrower main lobe than Sunde's FSK (h = 1).
-85-
while allowing as little noise as possible through the
bandpass of the PLL's. This ensures that· the recovered
carriers· have minimal ·phase noise present, allowing for
efficient detection. As the theory of PLL's is extensive,
only the topics relevant to their use in the coherent MSK
demodulation structure will be discussed. The actual
hardware implementation will be described, and finally,
their performance will be evaluated.
270°
----- X2
fout=fin
¥'out= ~in
-86-
The X2 block squares (hence frequency doubles) the VCO
output frequency, hence explaining the terminology "Square
Law Tracking Loop" often found in the literature when
describing the PLL shown in Fig.5.4. The 270° phase
shifter has an amplitude gain of 1, and we ignore it in
the derivation of the loop transfer function (its presence
. .
... (5.1)
••. (5.2)
The loop filter sets both the Type and Order of the loop.
By Type, it is meant the number of perfect integrators in
the loop. All PLL's are at least of Type I, as the VCO is
a perfect integrator (the 1 Is term in its gain constant
implles that it is an integrator). Were F(s) to be
implemented by a passive RC f~lter, the loop would still
be of Type I, as the passive network (always of low-pass
configuration) does not provide a good approximation of a
perfect integrator. However, with F ( s) realized with an
active filter, the Type of the loop would at least of Type
II (depending on the number of integrators), as the active
filter provides a good approximation of a perfect
integrator. The order of the loop is determined by the
-87-
d~gree of the denominator polynomial of the closed loop
transfer function. F(s) is often approximated be means of
an·active filter to be;
ST 2 + 1
F(s) = ------- ... (5.3)
ST1
2K 0 Ka(ST 2 + l)/T 1
H(s) = ----------------------------- ••• (5.4)
s2 + 2K 0 Ka(T 2 /T 1 )s + 2K 0 Ka/T 1
-88-
... ( 5. 7)
and
... ( 5. 8)
Log IG I
' I
Passi~,! Active
filter t' filter
I',
I '
I
I
I
I
I
I
I
I
I
I
I w • w" I
I . I
I I
I I
I I
o.--~-4~~.;.-~~~~~~~~-L~w
: I
II Passive
. I
-90
-135
-180
Phase (deg)
One can see that the initial -6 dB/octave slope due to the
-89-
VCO (recall that the VCO gain constant contains a 1/s
term) is increased to -12 dB/octave at a frequency of
l/AT 1 (rad/s), where A is the DC gain of the loop filter
(this is typically a very large number, so the breakpoint
is close to DC). After this pole, the phase response of
the lbop approaches -180°. At a radian frequency of l/T 2 ,
F(O)
T2
Log w
-90-
are defined;
-91-
VCO 1no FM)
I
I
I
I FM sidebands SPaced t.w
I
t ,t t t
VCO (with FM)
" "
Lcomponent
at input signal
frequency
(rad/s] . . . ( 5 . 11 )
-92-
To make AwPI as large as possible, it is desirable to make
the DC gain of the loop F(O) large. The time taken for the
loop to pull-in for an initial frequency offset of ~f, and
loop noise bandwidth BL (with '5 = · 0. 707) is termed the
Pull-in time, and is given by;
-93-
5.2.l(d) Hold-in Range (~wH). Once the PLL has acquired
lock, there is a range of frequencies, the Hold-in Range,
over which the PLL will track the input signal without
losing lock. The limit set on the Hold-in range is
determined by which loop components (PD or loop filter)
saturates first. It should be borne in mind that as the
input frequency varies, the phase detector output has to
vary in order to adjust the VCO frequency to agree with
the input frequency. The phase detector has a maximum
output limit of a few hundred millivolts (for diode ring
PD's) and this sets the Hold-in range. Based on this fact,
[5.3] gives the Hold-in range as;
-94-
5.2.2 Design and Construction of the PLL Modules
sv
390
220p
I To Frequency
sv
1150211 16p
• 390
Doubler
0
RFC
rnw
I56p MARZ
220p
I
86801 f I12p I47p
t 70MHz
Output
-95-
..............................................................................
~
VCO tank circuit Q by the low output impedance of the loop.
filter. In order to restrict the range of frequencies
attainable by the vco~ a 12 pF capacitor is connected in
shunt with the varactor (this prevents the PLL from
acquiring the wrong frequency).
--:-.-71.-
.-
- ·---·--· --····
- --· -- -
---·-·-·----
. -----·--· 71
- . 1 . . - -
.. _·- L_ ___ _-2 ___ 3 _:_> -~ 4.. -~~-__:L 5-.:.c_i__:._: ::-Q :::: .·
--~·----'-----'-----'-----'-.:...;__...:.;· --~----·~- ;
._.-_--_..;__· . : ~. : ' - - - - .: : : - . . •··
-96-
K0 = 8.42 E6 rad/s/V.
vin=O ..sv
The value of K0 for both the MARK (2fm) and SPACE (2fs)
PLL's are similar.
SY
sv 390
5(1J l 0
SY To 270
SRA1 MAR2
L • 390
•
R
MAR1 140 MHz
MAR 2 Output
-97-
phase shifter. The output levels of this module are z2dBm.
-98-
· of the free space wavelength, we get;
le = 0.66*0.75*2.1542
= 1. 066 m,
-99-
cpR-q>L
dVout
=> = 0.272.sin 0 [V/rad]
dVout
Ka = -----
d0 10=900
= 0.272 [V/rad],
-100-
It should be noted that as the maximum output level of the
phase detector is les·s than 300 mV, it is necessary to
amplify this level to a more usable range for the VCO
(i.e. the range OV to 15V). It is for this reason that an
active .filter is used to realize the loop filter, as it
can have gain (the passive filter cannot).
15V
To VCO
SY
-A( sCR 2 + 1)
F(s) = ----------------------.
sCR 2 + 1 + (l+A)(sCR 1 )
-101-
'.
The DC gain A is (without R3 present) normally a very
large number (typically 10s). The presence of R3 reduces
this gain to F(O), which is still large (typically 100).
The reason for including R3 will be explained later. With
A large, F(s) may be reduced to;
F(O) = 100.
-102-
bandwidth is dependent on wn, we desire to make wn small.
Experience indicated that wn ::; 2n:.1000 [rad/s] provided
satisfactory results .. In the ensuing calculations, we
assume wn = 2n:.1000 [rad/s].
-r 1 = 116.0 ms
and
T2 = 225.1 µs.
R1 = 11. 6 KQ
and
R2 = 22.5 Q.
,
-103-
-~+-.+1·' llF-·--·L-+f i· i~. ill ·
+-+------+ ~---G"-~f T·r••;--1-~ I~ f ri 1l11 ··•·· ....•
-20 itc--:~.,-,.,.,.~~~~'::t-t-t-H-·-~--1
c
;;.-
;
. :§'
J -so
·40
.+.;...~~-l.;.;.:.!'-"-4"'-"-'-'-'f;.;.;.!,
--:·-T-·--r1 t··l1·1i----.. ,- - ·- ·-i ·1
I --:-
1·
i. f t • :~ . I ..
:
;
Irr·
l- . .:
'l
!
..!?
2 -60 +-;--~..,......;~-,-.,-,+,.-....,-+-
- -- -··--
'
• : 1 i .. --
l I • 1
I l ii· ,:
iOHz ~ • ' • iOkHz
-104-
one must interpret this value as meaning that the PLL can
track widely varying input frequencies. The Lock-in range
and Pull-out frequencies are seen to be small, and this is
a result of the small value of wn.. However, the noise
bandwidth is also small because of the chosen value of wn.
-105-
®
Amplitude Amplitude
(dB I !dB)
f f
-106-
I
(\ (\ (\ (\ Ill Ill n·
·'I
I I I I 1
I .\ .
u v v v \lj ~ Ill
-107-
-----p
s
10dB/
. f
.
RES BW 1kHz VBW 300 Hz SPAN 100kHz
RES BW [Hz] ]
Psse =
lHz 1 [Hz]
-108-
l
I
I
!
i
I I
vA ssl I./
FC i
r-~~~~-+~~-;-~~m-~~~~~¥-r.-~--i~~-+~~--+-~____,,
CORR
-109-
· plot, we get;
~(lOkHz) - -37.54 - 30 + 2 .. 5
= -65.04 [dBc/Hz]
It has already been shown (equation 2.25) that the bit rate
clock is given by the difference frequency of the 2fm and
2fa carriers. The first step then, is to mix the two 140 MHz
outputs of the MARK and SPACE frequency PLL's. The
difference frequency is then extracted by means of a low-
pass filter. The circuit diagram of the module is shown in
Fig.5.19.
-110-
From SBL 1 SY
SPACE
Pll L
2·6uH
Q---iA
R 74LS
From 0 74. Q
MARK
PLL
100K
SY SY
100K 10tH<
® 15 14 7
Q 13 I-Channel
Q
s Q-Channel
-111-
sampled at the falling edge of the symbol clock, while the
Q-Channel is sampled at the rising edge of the clock. As the
sampling is optimal at the maximum eye opening, the I- and
Q-Channel sampling waveforms (their falling edges) must be
capable of being variable relative to the rising edge of the
symbol ·clock (i.e. the sampling instant for each channel
must be variable to adjust it for optimum operation).
-112-
subtracters are required. It was dee ided to use Trif ilar-
wound broadband transformers [5.5] as the
addition/ subtraction circuit elements. Fig. 5. 2 0 shows the
circuit diagram.
ror Q-Channel
1or
• I-Channel
The MARK frequency PLL feeds the 0°-0° power splitter, and
each output of this splitter feeds an input to the the 0°-0°
power combiners. The SPACE frequency PLL feeds the 0°-180°
power splitter, and each output of this splitter feeds the
other two inputs of the 0°-0° power combiners. The upper
power combiner therefore effectively subtracts the PLL
outputs, while the lower one adds them. The
addition/subtraction process is linear provided that the
transformers are not operated in their non-linear
(saturated) region. For the powers used in this application,
the transformers are operated well out of their non-linear
region.
-113-
The isolation between the two input ports was measured (at
an input power level of 5 dBm) by connecting an RF source to
one input port, and measuring the power level at the other
input port. The isolation was found to be the same for both
ports, · and its measured value was 36.5 dB. This is
considered adequate, and the good performance is due to the
ii:therent symmetry of the transformer arrangement.
-114-
•
downconversion to baseband of the received MSK waveform. As
was discussed in section 2.2.4(e), the I-Channel is obtained
by forming the product; .
I-Channel= sI(t).sMsK(t),
SBL 1
L
MSK
Input
.
L
SBL1
-115-
5.6 DETECTION FILTER MODULE
- - SIMULATED
BER• 10.t
DATA RATE• llO MBPS
LINEAR CHANNEL
~HANNEl Fil TERS: 1.f'OlE CHEBYSHEV & PHASE EQUALIZED
(MEASURED RESPONSES)
DETECTION Fil TERS: 2.f'OlE BUTIERWORTH
.I •
CHANNEL Fil TER • 58 MH~
r-......
0 BACK TO BACK
X 113 MHz ~
l> llOMHJ
6IJ MH> PHASE' o 66 MHJ PHASE EQUALIZED
EOUALliED
~ • 61JMHJ
.-- -
80 MHJ '-. .............. ~
- -
~
-·
BACK TOBACK
113 MHJ
~'
.............. -..A,. A A . ---
-
-
0
0.30
I
O.Jli "
0.'11 O.~
i:,,__
0.50
lldT
~
0.56
-- 0.80
0
0.116
a
0.70
-116-
The horizontal axis shows the 3 dB bandwidth ( Bct), symbol
duration (T) products .. The vertical scale indicates the
performance degradation to be expected relative to ideal
matched filter detection, for various levels of channel
filtering.. The lowest curve (for back-to-back modem
operation) is the . relevant one for our design. For this
c&rve, least degradation occurs at;
20·9uH
-117-
j
The constructed Coherent MSK Receiver is shown in Fig.5.25.
-118-
REFERENCES
-119-
CHAPTER 6
where fb is the bit rate clock and f 9 and fm are the SPACE
and MARK frequencies respectively. For the situation where
the modulation index h is not 0.5, we have;
-120-
f0 ' = 2(0.5 ± Sh)f 0
- fb ± 26hfb,
2-048 MHz
PRBS MSK Output
ru &. LTC
OIFF ENCODE
INTERF.tH
CIRCUITRY
fb FREQUENCY f b
'--~~~~~~--
COM PAR AT 0R
-121-
(proportional to the frequency difference between the two
inputs) is produced. This is used to drive the interface
circuitry which derives," from the error signal, the voltage
control inputs to the level translation circuitry necessary
to maintain the modulation index at its correct .value.
Instability of the loop is avoided by making the loop time
constant large, thus . introducing a form of dominant pole
compensation.
... ( 6. 1)
where;
-122-
Vi = Amplitu~e of input waveform
v cl = Forward voltage drop of a diode (=0.6V)
fi,f2 = The input frequencies to be compared
Ci = Input capacitor value
R = Output load resistance of the charge pump.
The charge pump operates off a single (+lSV) s~pply, and the
necessary level translation to bipolar (±lSV) format is done
by means of an inverting summer op-amp. Fig. 6. 2 shows the
circuit for the module.
15K 30K
.__~10K
10K
1K8
SV1
-123-
For a lOKQ potentiometer at the output of the charge pump,
R z 2.SKQ (Thevenin equivalent resistance), and hence
C1 = lnF. For f 1 = f 2 , the output potentiometer is set to
give V u t
0
= 7 .SV (i.e. Vcc/2). The smoothing capacitor is
. chosen to be large to ensure that the · response of the
circuit is relatively slow to enhance the overall loop
stability.
-124-
Fig.6.3 Measured Transfer Function of
Frequency Comparator Module
-125-
to its circuit diagram shown in Fig.6.4.
10K
1SK
1SK
SK
-1sv·
Iset SPACE I
1S1<
1SK
1SV
'Error Signal
Input SK
-1SV
lset MARK I
-126-
function as the preset resistances that were originally used
to set Vco and Vci in the level translation circuitry
described in section 4; 1. 3) . Note that the sense of the
error signal is inverted because the op-amp is operating in
the inverting summer configuration.
The op-amp thus multiplies the input by -0.5 and adds this
to a constant level of 7.SV. The 7.SV source is generated by
amplifying the -SV source (set by the zener) by a gain
of -1.5. Note that the sense of the error signal is again
inverted and this results in the output of the lower path
varying in-phase with the error signal. For zero voltage
input to the module (i.e. for correct modulation index), the
output of the lower path is dependent only on the voltage
set by the preset resistance, which sets the nominal MARK
frequency.
-127-
.•
which sets the nominal SPACE frequency.
-128-
From the VCO curve (Fig. 4. 9), the gain of the VCO at the
MARK frequency is 0.427 MHz/volt, while at the SPACE
frequency it is 0.178 MHz/volt. By a first order
approximation then, we can set the gain of the SPACE
frequency control path to be 0.427/0.178 = 2.40 times larger
than that of the MARK frequency path, thus offsetting the
gain differential due to the non-linearity of the VCO.
-129-
CHAPTER 7
Pe = Q(v'°2il)
-130-
detection may be imp~emented, resulting in considerably
simpler receiver circuitry. By disregarding the phase
information available ·to us, a degradation in noise
performance must be expected. Because MSK is an orthogonal
signaling scheme (see section 2.2.4(b)), the probability of
error is given by the expression [7.1];
Pe = ~e-µ/2,
MSK IF
Input DISCRIM - DECISION DATA Data Output
I NATOR 1--- THRESHOLD LATCH
-131-
--·--- ·-·---------------
-132-
MSK l nput
SBL-1
L
7·77uH
50
-133-
An 8:1 transformer is u~ed, and the resultant ·1oad presented
to the tuned circuit is;
RL = (8) 2 .50
::::: 3. 2 KQ.
Ci = C2 = 1.55 nF
Li = 7.77 µH
-134-
Fig.7.3 Frequency Discriminator Transfer Function
-135-
I,
~B
COPI
-136-
74LS14
10K
-1SV
I 22uf
-137-
it TTL compatible. The 74LS14 Schmitt trigger is used to
buffer the module output for the circuitry that follows. The
signal at this point in the ·circuitry is used for two
functions;
-138-
47 R e-~overed
1
VA SS
~
\I
/ i\ I \;I
\ ~
r r\ v' \
. .JI""'
FC I
! I
I
CORR I 1
I I I I
I I I
I
II II I
! i
CENTER -175.0 kHz SPAN 10.00 MHz
RES BW 100 kHz VBW 3 kHz SWP H1'l0 msec
-139-
lie on the clock frequency and multiples thereof). In order
to regenerate a deterministic component, a non-linear signal
processing operation has to be applied to the waveform.
- -------------
@ FWR
d
Random dt I· I
TTL
Waveform
VCO
....____________ Recovered
Clock
-140-
@o
.© 01----'
0 \....---'
0 \....---'
+
0 1---------. - - - - -
-141-
This clearly · shows how the random RZ waveform may be
decomposed into the sum of a deterministic waveform (the
clock waveform), and . a random waveform. It is the
deterministic compo.nent which regenerates the clock line.
The waveform at point C in Fig.7.7 is then fed to the Phase-
Locked Lo9p (PLL) which acts as a narrow bandpass filter to
extract the clock line at a high carrier to noise ratio.
-142-
74HCOO
74LS14
-143-
;p Regenerated Clock Lines MKR L:::. 2.05 MHz
REF .0 dBm ATTEN 10 dB -27.87 dB
PEAK
LOG ~
10
dB/
I
II
i'
l "trv,,I'\, ~/ ~v,
N? ~-lit~ i,~w ~~
/11'\M
VA SB
FC 1/ fv-.1,, \f'rv.W ~tJ
I
CORR v I
l
-144-
spectrum implies that a clock frequency is present in the
waveform, and may be extracted by a PLL.
-145-
10K .
3-SSuH
-146-
1K8
82K
0·1u
r-
-~ BC 108 ·
330p
56
39K
1K 10u
Note how the 2.048 MHz clock line has been amplified and the
clock harmonics suppressed. It is onto this spectral line
that the PLL locks.
-147-
lP Spectrum After SPF MKR A 2 . 05 MHz
REF -8.0 dBm ATTEN 10 dB -26.23 dB
~ }
PEAK
LOG
10
dB/
MARK ~R A ('>
2·.05 MHz
-26. ~3 dB
I
I I
I
'
VA
FC
SB
l \
\,w / y~ I
CORR
~ I I I
I I~ M..Jw..
IL~ A
. . . .I
II
I A. . -
CENTER 4.825 MHz SPAN 10.00 MHz
RES SW 3 kHz VBW 3 kHz SWP 3.12l sec
-148-
. '
-149-
7.4.3(c) Configuration of the VCO. To avoid false locking
of the PLL, the frequency excursion of the VCO must be
limited to only the .frequencies of interest. The lower
frequency limit is set at 1.6 MHz, and the upper limit at
2.4 MHz. Using the design curves for the PLL, we obtain;
R2 = 10 KQ
C1 = 150 pF.
9 4
c; Vcont vco out
--=- inh
~~J
11 12
nn
:> ••
15K~ ~· 10K 150p
R1
n'n
Rz
rr '7
c,
-150-
where Vmax and Vmin are the maximum (SV) and minimum (OV)
control voltages applied to the VCO.
The loop can be made stable by ensuring that the PLL open
loop transfer function approaches the 0 dB gain point at a
rate of -6 dB/octave or less. The open loop transfer
function is shown in Fig.7.16.
Open Loop
-12dB/oct
Gain
!dB)
-151-
The initial ~6 dB/octave rollof f is due to the presence of
the VCO in the loop (it is an integrator, as was discussed
in section 5.2.l(a)). A pole (inserted by the loop filter)
then increases the slope to -12 dB/octave. This is reduced
back to -6 dB/octave by the loop filter zero at a
frequency f 1
[Hz]. The gain slope continues to descend at
this rate and the unity gain . frequency is reached at a
frequency f2 • Provided that f2 is chos'en to be 3 to 5
·times f 11
sufficient phase margin will be provided, and
the loop will be stable ( 7. 4]. The loop filter that is
used is the Lag-Lead filter, as is shown in Fig.7.17.
-152-
I
R3 = 470 KQ
and
R4 = 390 Q.
-153-
tlo Re.covered
Clock Spectrum MKR L::.. 10.0 kHz
REF -8.0 dBm ATTEN 10 dB -54.67·dB
PEAK
.LOG
10
dB/.
VA SB
FC
CORR
-154-
the recovered clock with respect to the data eye-opening.
The circuit to perform this static phase adjustment is
shown in Fig.7.19.
-------SV
74LS
9 123 - 12
Q
Clr
11
SV
-155-
sv .
1 4
0 ata
0ata - 2 0 a s
In put _ 3 74LS nutput
Cl ock Input 74
-156-
Fig.7.21 Constructed Non-Coherent MSK Demodulator
-157-
REFERENCES
-158-
CHAPTER 8
CONCLUSIONS
-159-
I
. I
APPENDIX A
For brevity, we let YI(t) = yI, and y 0 (t) = Yo· Also, rr/2Tb
= 2rr~f = ~w, and 2rrfc = wet.
-Al-
Using identities (i), (ii), and (iii), we may transform the
expression to;
-A2-
(c) The term ~.cos 2(wc - t:.w)t is a similar spectral
component at a. frequency of twice the MARK
frequency.
-A3-
Frequency Doubled ·
MSK Spectrum
--
BPSK Spectrum
/
/
''
I
I
/
' \
\
\
----- ......
- I
I \
\
\
, ,,. ,,,....
----- ., --- '
-
....... '
I \
-A4-
APPENDIX B
-Bl-.
a
' b 01 ,, /
'' ,, ,,
''
''
11 '' 00
c ' a
d 10 ''
~e = ±rrh
= ±rr/2.
A0 = ±rr(0.5 + oh)
-B2-
= ±rc/2 ± rcoh
= ±rc/2 ± <I>.,.'
0.
'' t=T
'' ,,
' ', ,, /
' /
'' ' ,, ""
', ,,"'
t=2T
,
,, "
/ "
,, / "
-B3-
error.
-B4-
Set A. Set B Set C Set D
PA= (4).(1/16)
= 1/4.
-BS-
A similar analysis foF five• and six-bit sequences was
performed, and the results are shown in Fig.B.5 and B.6.
-B6-
I
to;
-B7-
. - 6 MAR 1989