APPLICATION

MSK and its Application to Wireless Data Transmission

1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
1. Introduction
Bandwidth Efficient methods for transmission of data have been and continue to be of great importance to the data
dependent community. A community that is experiencing exponential growth. Minimum Shift Keying (MSK) is a continuous
phase modulation (CPM) technique that offers advantages in performance and ease of implementation. This paper will
provide some insight as to how a simplified MSK modem can be implemented and modeled. The model presented is
suitable for a system simulator such as Ptolemy from UC Berkeley.
1.1 Background
The academic field of "Data Transmission" is loaded with modulation schemes. Most involve translation of data bits or
patterns into a unique combination of phase, frequency or amplitude. Some of the more notable techniques are listed in
Table 1.
MODULATION TECHNIQUE COMMON
ACRONYM
Frequency Shift Keying FSK
Multi-level Frequency Shift Keying MFSK
Continuous Phase Frequency Shift Keying CPFSK
Minimum Shift Keying MSK
Gaussian Minimum Shift Keying GMSK
Tamed Frequency Modulation TFM
Phase Shift Keying PSK
Quadrature Phase Shift Keying QPSK
Differential Quadrature Phase Shift Keying DQPSK
/4 Differential Quadrature Phase Shift Keying /4 DQPSK
Quadrature Amplitude Modulation QAM
Table 1: Modulation formats [6].
Each of the modulation formats listed in Table 1 is suited to a specific application. In general, schemes where two or
more bits are represented by a symbol (e.g. QAM, QPSK) require better signal to noise ratios (SNR) than two-level
(binary) schemes for similar bit error rate (BER) performance. Additionally, in a wireless system, schemes that have more
than two levels (m-ary) generally require greater power amplifier linearity.
Most implementations of the modulation formats listed in Table 1 are synchronous. When data rates exceed 1200
bits/second or when the transmission medium is subject to non-ideal affects (e.g. fading or SNR < 25 dB) synchronous
data transmission is preferred over asynchronous. Synchronous data transmission is characterized by the presence of a
clock which is synchronous to the data. The clock is embedded in, and therefore recoverable from, the modulated signal.
MSK is a synchronous modulation format.
Another important consideration in data transmission is bandwidth. Digitally modulated data, composed of sharp "one
to zero" and "zero to one" transitions, results in a spectrum rich in harmonic content that is not well suited to RF
transmission. Hence, digital modulation formats that minimize bandwidth (BW) consumption are in vogue. As implied
earlier, digital modulation involves the mapping of changes in data states to changes in amplitude, frequency, phase, or
some combination of the three. After smoothing the transitions (discontinuities) in phase, frequency or amplitude, we can
see, through Fourier analysis, BW consumption is reduced. An entire family of modulation formats, categorized as
MSK and its Application to Wireless Data Transmission 2 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
continuous phase modulation (CPM) minimize BW consumption by eliminating phase discontinuities. CPM state changes
are represented by non-abrupt changes in phase and frequency while the amplitude of the carrier envelope remains
constant (i.e. phase modulation or frequency modulation).
1.2 Theory of Continuous Phase Modulation
The simplest form of constant envelope modulation is binary FSK. This type of modulation is composed of two sinusoids
s t t
1 1
( ) cos (1)
and
s t t
2 2
( ) cos . (2)
Where s
1
(t) is selected for a logic zero and s
2
(t) is selected for a logic one. An implementation of the this type of FSK
system would appear as shown in Figure 1.
cos
1
t
cos
2
t
BINARY DATA IN
FSK
OUTPUT
Figure 1:Discontinuous phase FSK modulator.
This type of FSK is referred to as discontinuous phase or non-coherent FSK because there is no coherence
between the phase of
1
,
2
and the modulating signal (BINARY DATA IN). CCITT standard V.21 (300 baud) is an
example of non-coherent FSK (f
1
= 1080 Hz and f
2
= 1750 Hz) it has also been applied to other low data rate systems
where bandwidth efficiency is not an issue.
An alternative method, to that depicted in Figure 1, for generating FSK is to use a voltage controlled oscillator (VCO)
or frequency modulator. The key aspect of the VCO is, for instantaneous frequency deviation proportional to the
modulating signal m(t), the phase must be proportional to the integral of the modulating signal. In mathematical terms
e t A j t
C
( ) Re{ exp[ ( )]} (3)
where, (t) =
C
t +
C
+ (t)
e(t) = modulated carrier (FM is a type of angle modulation)
A
C
= amplitude of unmodulated carrier

C
= angular frequency of unmodulated carrier

C
= initial carrier phase angle
(t) = instantaneous phase angle due to the modulating signal m(t).
MSK and its Application to Wireless Data Transmission 3 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
Rewriting equation (3) applying Eulers equation for exponential relations,
e t A t t
C C C
( ) cos[ ( )] + + (4)
where,
( ) ( ) t K m d
t

. (5)
Here, m(t) is the modulating signal and K is a constant of proportionality. Substituting (5) into (4) yields the equation
for continuous phase FSK (CPFSK).
e t A t K m d
C C C
t
( ) cos[ ( ) ] + +

. (6)
In the case where m(t) is bipolar (logic 1 = +1, logic 0 = -1) the integration in equation (5) can be carried out as follows.
First we must have an expression which represents the bipolar data signal as shown below.
m t S r t nT
n
n
( ) ( )

(7)
where, S
n
= 1, according to bipolar data polarity
r(t) = rectangular pulse with amplitude 1/(2T) and duration T.
Substituting (7) into (5) yields
( ) ( ) t K S r nT d
n
n
t

(8)
( ) ( ) t K S r nT d
n
n
t

. (9)
For the interval 0 t < T
( ) t K S
T
d
n
n
t

1
2
0
(10)
( ) t K S
T
t
n
n

1
2
. (11)
Using the result from equation (11), equation (6) can be written as
e t A t KS
T
t
C C n n
n
( ) cos[ ] + +

1
2
. (12)
MSK and its Application to Wireless Data Transmission 4 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
Where, nT t < (n+1)T and
n
is adjusted according to the phase of the previous symbol
n-1
. As a result, if the
modulating signal (NRZ binary data) has no impulses, the modulated carrier e(t) will not have phase discontinuities. A
general representation for CPFSK modulator is shown in Figure 2.

cos(t)
NRZ
BINARY
DATA IN
FSK
OUTPUT
VCO
e(t)
m(t)
Figure 2:CPFSK generation.
Looking at the term K in equation (12)referred to as a constant of proportionality earlierfor constant envelope
modulation systems (e.g. FM) K is proportional to the peak frequency deviation of the VCO. For CPFSK, the modulation
m(t) is NRZ data and
K f T
d
2 . (13)
Where, f
d
= (f
2
- f
1
(
T = symbol or bit period
f
1
= logic zero (S
n
=-1)
f
2
= logic one (S
n
=+1).
The term f
d
T is commonly known as the modulation index h. Substituting 2h for K in (12) yields
e t A f t
hS
T
t
C C
n
n
n
( ) cos[ ] + +

2

(14)
for, nT t < (n+1)T.
For MSK, h = 0.5 so (14) can be written as
e t A f t
S
T
t
C C
n
n
n
( ) cos[ ] +

_
,
+

2
4
(15)
for, nT t < (n+1)T.
Consequently, if S
n
is +1, the frequency of the carrier is shifted higher to
f f
T
C 2
1
4
+ . (16)
When S
n
is -1, the frequency of the carrier is shifted lower to
f f
T
C 1
1
4
. (17)
MSK and its Application to Wireless Data Transmission 5 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
Subtracting (17) from (16) yields,
f f
T
2 1
1
2
(18)
Equation (18) states, for h = 0.5, the frequency separation between f
1
and f
2
is equal to half the bit rate. Figure 3illustrates
a set of four waveforms for a typical MSK modulator.
T
t t
t t
S
n
=-1
S
n
=+1

n
=0
n
=
(a) (b)
(c) (d)
e(t) e(t)
e(t) e(t)
Figure 3: Typical MSK modulator outputs for (a) logic 0 and
n
=0, (b) logic 0 and
n
=, (c) logic 1 and
n
=0, and (b)
logic 1 and
n
=.
In Figure 3 it should be noted that e(t) for a logic 1 (S
n
= +1) has only one more zero crossing than a logic 0 (S
n
= -1). In
other words, zero crossing detector must discriminate down to the occurrence of a single zero crossing in order to
accurately demodulate the data signal. According to [7], a non-coherent zero crossing detection scheme will perform
nearly as well as a coherent detector if the FSK signals are orthogonal. For two signals to be orthogonal over an interval
T, their inner product must be zero,
x t y t dt
T
( ) ( )
*

0
0
(19)
where, y(t) x(t) and y
*
(t) is the complex conjugate of y(t). If x(t) and y(t) are real then (19) becomes
x t y t dt
T
( ) ( )

0
0
. (20)
MSK and its Application to Wireless Data Transmission 6 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
For any FSK system where
1
=2f
1
and
2
=2f
2
, equation (20) can be written as
cos( ) cos( )
1 2
0
0 t t dt
T

. (21)
Performing the integration in (21) yields
sin( )
( )
sin( )
( )




1 2
1 2
1 2
1 2
2 2
0

+
+
+

T T
. (22)
For MSK equation (18) can be rewritten in terms of radian frequency


1 2

T
. (23)
And the radian carrier frequency is

C

+
1 2
2
. (24)
Substituting (23) and (24) into (22) yields
sin( ) 2 0
C
T . (25)
Which has the solution

C
k
T

2
. (26)
Where k is an integer.
As a result, if equation (26) is satisfied, f
1
and f
2
will be orthogonal. In fact, equation (23) represents the minimum
separation between f
1
and f
2
. Hence, the name Minimum Shift Keying. This modulation scheme was first proposed by
Doelz and Heald [3] which resulted in a US Patent, assigned to the Collins Radio Company. Figure 4 is a simplified block
diagram for a MSK modulation system.
NRZ
DATA
SOURCE
VCO
MODULATION
INDEX = 0.5
Figure 4: MSK modulator.
A modulation scheme is of little practical value if it can not be demodulated in an efficient manner. The definition of
efficiency is somewhat application dependentin general cost, size, and power consumption must be minimized while
performance must be maximized. The performance of a MSK demodulator can be quantified by measurement of the
MSK and its Application to Wireless Data Transmission 7 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
signal-to-noise ratio (SNR) versus BER. SNR is considered by some as an inaccurate method for comparison of
performance because of its dependence on noise BW. In other words, the SNR can be significantly altered by varying the
system BW. Unless all systems have identical BW it is difficult to make comparisons. A more accepted measure of
performance is E
b
/N
0
versus BER. E
b
/N
0
is related to SNR as shown below,
E
N
S
RN
S
N
B
R
b n
0 0

_
,

_
,
. (27)
Where, S = signal power
R = data rate in bits per second
N
0
= noise power spectral density (watts/Hz)
E
b
= energy per bit
B
n
N
0
= N = noise power
B
n
= noise BW of IF filter.
Figure 5 is a plot of typical performance characteristics for ideal coherent and non-coherent MSK detectors.
Non-Coherent
Characteristic
Ideal Coherent
Characteristic
Figure 5: BER versus SNR where noise BW equals the bit rate.
2. Implementation
In this section we describe the implementation of a 1200/2400 bit/second MSK modem. That is, a MSK modulator and
demodulator including carrier detection and data synchronization.
2.1 MSK modulator
Depending on selected bit rate and digital data input, the modulator generates one of the 1200 Hz, 1800 Hz, or 2400
Hz frequencies. The frequencies for different data rates and bit polarity (logic one or zero) are given in Table 2.
MSK and its Application to Wireless Data Transmission 8 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
Baud rate
(BPS)
TX data
in
MSK tone
frequency
1200 1 1200 Hz
1200 0 1800 Hz
2400 1 2400 Hz
2400 0 1200 HZ
Table 2: Tone frequencies for 1200/2400 bps MSK modem.
The modulator generates tones that are phase continuous and change frequency at the zero crossings as shown in Figure
6.
2400 BPS case
1200 BPS case
TX data
Figure 6: MSK modulator output waveforms.
The modulator block diagram is shown in Figure 7. Primarily, the modulator consists of a digitally controlled oscillator and
a low pass filter. The modulation index is 0.5 and the carrier frequency is 1500Hz for 1200 baud and 1800Hz for 2400
baud case. This way of implementing the modulator is sometimes referred to as serial modulation.
DCO
Digital I/P
Data rate select
MSK tone O/P
Figure 7: MSK modulator block diagram.
2.2 MSK demodulator
Demodulation is the process of converting the MSK sinusoidal signals to digital ones and zeros. Figure 8 below shows the
block diagram of a non-coherent detector. The front end of this detector consists of a band pass filter whose frequency
response plot looks similar to Figure 15 for 1200 bps and Figure 16 for 2400 bps. The filter output is sliced to produce a
digitized form of the MSK signal, with positive half of the sine wave transformed into high going pulse and the negative
half into low going pulse. The monostable is a zero crossing indicator, it produces a pulse every time the input signal
changes polarity. The frequency discriminator then differentiates between the two MSK tones based on the zero crossing
MSK and its Application to Wireless Data Transmission 9 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
signal frequency. This detection scheme is non-coherent because the receiver is neither phase nor time synchronized to
the transmitter.
MSK signal I/P
Digital PLL
&
S/H
Comparator
Sync. data O/P
Recoverd Clock
Carrier indicator
Un-sync. data O/P
BPF
Limiter
Carrier detector
Descriminator
LPF
Mono
Zero Crossing
Filter
Figure 8: Non-coherent MSK demodulator.
The carrier detector output indicates if MSK signal is being received. The carrier detector works off of the zero
crossing indicating signal. The zero crossing filter discriminates between good and bad occurrences of the zero crossings
and counts the bad zero crossing occurrences over a pre-defined period. The comparator compares the bad zero crossing
count with a pre-set limit to make a decision on the carrier detector output.
The phase locked loop generates a reference clock running at the selected baud rate. The unclocked data output is
sampled on the rising edge of this clock to provide synchronized data output. The period of this synchronous clock is
adjusted every once per cycle if required to line up the rising edge of the clock with the middle of the unclocked data
output.
2.3 Transmit and Receive Synchronization
The modulator provides a transmit reference clock to the user, this clock needs to be used to feed the digital input data to
the device.
MSK and its Application to Wireless Data Transmission 10 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
TX SYNC
TX DATA
RX SYNC
Clocked Data
Dont care
Hold
data
Change
data
Hold
data
Hold
data
Change
data
Hold
data
Change
data
Hold
data
Change
data
Figure 9: Transmit and receive timing diagrams.
The de-modulator provides the recovered clock and data in reference to this clock. The timing diagrams of the TX and
RX sync clocks are shown above. These two clocks can be used to enable synchronous communication between the
modem and data terminal equipment.
Figure 10below illustrates an example of how the modem can communicate synchronously with any instrument that has
synchronous RS-232C interface.
MX-COM
MODEM
Sync. RS-232 C Connector
RX Sync
TX Sync
TX Data
RX Data
1
L
e
v
e
l

C
o
n
v
e
r
s
i
o
n
C
i
r
c
u
i
t
Figure 10: Synchronous RS-232C interface.
2.4 Carrier Detector
The carrier detector is an important feature of the de-modulator. The carrier detect indicator may be used to wake-up or
shut down some circuitry in the system to reduce the power consumption.
The carrier detector response time is user controllable via an external input. By altering the state of the external input
the user is able to trade sensitivity for response time. In other words, the faster the response the more likely the carrier
detect will false on noise and if falsing must be minimized then response time must be compromised.
MSK and its Application to Wireless Data Transmission 11 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
3. Modeling
With increasing demand to move products quickly to market and to keep costs down, verifying a design through simulation
before building a breadboard saves time and money. The rapid growth of PC based simulation tools brings modeling
capability to a larger number of design engineers. The verification process can be broken down into three steps:
1. Creating a model of the design - Software modeling is a process of emulating hardware blocks of a design. Models can be
designed on a hierarchical building block approach. The lowest level models may emulate simple hardware functions such as
filters, amplifiers, or digital gates. Several lower level models can be joined together to create more complex models such as
modulators or demodulators. These higher level models can be joined together to create even higher level models such as a
complete communication system.
2. Simulating the function of the model - Simulation is the process of stimulating a model with various inputs and observing
the models response. The response of the model can be recorded in plots and data files for analysis.
3. Analyzing the plots and data collected from the simulation - Analysis is the process of verifying the plots and data against
design goals. This process helps the designer ensure correctness of or see how changes may improve the design.
The models, simulation, plots and pseudo code included in this section were derived using a modeling tool called
Ptolemy. Ptolemy is a UNIX based modeling tool that has been ported to run on a PC based UNIX system called Linux.
Linux and Ptolemy are freeware software packages and are available to be downloaded from several sites on the internet.
One such site is sunsite.unc.edu. The Ptolemy home page is http://ptolemy.eecs.berkeley.edu. Other archives sites can
be found using a web browser search engine. There are several other PC based modeling tools available from various
software companies.
The scope of this document is to model an MSK communications scheme. Each topics will be broken up by function.
The topics will include modeling MSK generation and detection (including a description of bit synchronization and carrier
detection) and modeling MSK at the system level. The conclusion will review the topics and results presented in this
document.
Discrete Time Models
All the models presented below are in the discrete time domain with successive samples uniformly spaced in time. The
complete transmitter->channel->receiver model is a multirate system, meaning that different sampling rates are used at
various stages in the model. For example the input data bits to be transmitted are sampled at the baud rate either 1200 or
2400 samples per second. The transmitter models output sample rate is N times the baud rate. The majority of the
receiver also runs at this N times rate. We typically used N=60 for the simulation results presented below. The emphasis
of this section is on modeling the basic functions included in a typical serial MSK modem offered by MX-COM. RF
interfaces and channel models are discussed to aid the reader in developing a complete system model however they are
not fully developed. In order to simulate frequency offsets between the transmitter and receiver an
interpolating/decimating block that changes the sample rate by M/N can be inserted between the transmitter and receiver.
For a 0.1% frequency offset let M=1000 and N=999. For in-depth coverage of multirate processing see [2].
In the following sections, pseudo-code is provided to show the algorithm or function performed by each of the blocks
in the model. The authors hope the pseudo-code will aid the readers in modeling and performing simulations of their
system designs. The pseudo-code is presented in a simple manner ignoring certain programming considerations such as
memory allocation, minimal buffering, type declaration, and scheduling of execution sequence. The models below focus
on the core algorithm not the implementation details. Various system simulation environments such as Ptolemy have their
own unique ways of dealing with these considerations which are beyond the scope of this paper.
The following conventions are used in the pseudo-code.
// starts a comment which ends at the end of the line
= assignment or comparison depending on context
+ , - , , / all have their normal meaning. Division is also shown as
x
y
and multiplication as x y
Y[x] x is the index of array Y, x and Y are used here as an example - they may have any name. Inputs and
outputs of each model are treated as an arrays. When referring to an input or output, increasing values of x
progress forward in time. Thus Input[x-1] would refer to the previous input to the model at sample x. If a block is
multirate, then it consumes more or fewer input samples than it creates output samples. For example the MSK
modulator model creates N output samples for every input bit. That is upon processing Input[i], it will create
Output [ ] N i , Output [ ] N i +1 , Output [ ] N i + 2 , up to Output ( ) [ ]
N i +1 -1
MSK and its Application to Wireless Data Transmission 12 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
F(x) function F( ) evaluated with parameter x. For example:
sin(x) is the sinusoidal function of trigonometry
cos(x) is sin(x + /2)
sinc(x) is sin(x)/x and has the value 1 at x equal 0
sign(x) returns 1 for 0 and positive x and -1 for negative x
for i=0 to N-1 { command sequence } iterates command sequence N times with i increasing by 1 each
iteration.
while (expression) { command sequence } repeats command sequence as long as expression remains true.
if (expression) { seq. 1 } else {seq. 2} executes seq. 1 if expression is true otherwise executes seq. 2.
3.1 MSK Modulation
MSK modulation is the process of converting digital data to sinusoids that represent digital data values. In MSK
modulation digital data must be converted to a bipolar representation of the data. The bipolar data is then modulated with
a modulation index of 0.5. The resulting frequencies are equal to the carrier frequency plus and minus one forth of the
baud rate. The initial phase constant of the VCO determines whether data transitions occur at the signal peak (i.e. cos
MSK) or the zero crossing (i.e. sin MSK).
Digital Data
to Bipolar
Conversion
VCO with
Modulation
Index of .5
MSK
Output
Digital
Data
Input
Figure 11: Typical MSK Modulator.
3.1.1 1200 Baud MSK Signals
Figure 12 shows 1200 baud MSK data modulated with a 1500 hertz carrier. MSK sine waves are continuous phase and
can only start at zero for a phase constant of zero and pi radians. The first four traces show the sin MSK symbol shapes
for data 1 and 0, while the fifth one shows the eye diagram. See Figure 19 for a sequence of bits at 1200 baud.
3.1.2 2400 Baud MSK Signals
Figure 13 shows 2400 baud MSK data modulated with a 1800 hertz carrier. All four symbol shapes are shown followed by
the corresponding eye diagram. See Figure 20 for a sequence of bits at 2400 baud.
MSK and its Application to Wireless Data Transmission 13 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
Pseudo-Code Listing 1 MSK Modulation Model
__________________________________________________________________________________
I/O, Variables, and Constants
input[ ] // transmit data bit array
output[ ] // output array
N // interpolation rate -> output sample rate = N x baud rate
2p[i] // phase constant in radians 0, for sin MSK /2, 3/2 for cos MSK
p[i] // phase constant at sample i has value of 0 or 0.5
Map[0] = 1 // input data 0 mapped to bipolar data 1
Map[1] = -1 // input data 1 mapped to bipolar data -1
c // carrier normalized to the baud rate
// 1200 Baud Value = 1.25 = 1500(Carrier)/1200(Baud Rate)
// 2400 Baud Value = 0.75 = 1800(Carrier)/2400(Baud Rate)
_____________________________________________________________________________________
Pseudo-Code for MSK modulator
dk = Map[input[i]]; // Map input data value (dk) for this symbol.
for n=0 to N-1 { // Generate N output samples per input bit
tx eye diagram
data 0 at 0 rad
data 0 at PI rad
data 1 at 0 rad
data 1 - PI rad
0 1/6 1/3 1/2 2/3 5/6 1
Time (Bits)
Figure 12: 1200 Baud sin MSK.
tx eye diagram
data 0 at 0 rad
data 0 at PI rad
data 1 at 0 rad
data 1 - PI rad
0 1/6 1/3 1/2 2/3 5/6 1
Time (Bits)
Figure 13: 2400 Baud sin MSK
MSK and its Application to Wireless Data Transmission 14 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
[ ] [ ]
output N i n sin 2 c
dk
4.0
n
N
p i + +

_
,

1
]
1

_
,
;
}
p[i+1] = fractional remainder of
[ ]
p i c
dk
4.0
+ +

_
,

; // calculate phase constant for next input bit

_____________________________________________________________________________________
3.2 MSK Demodulation
MSK Demodulation is the process of converting Sinusoids to digital data. MSK demodulation looks for specific
frequencies that represent digital data. Shown in Figure 14is a block diagram for a typical MSK demodulator. MSK
demodulation can be broken down in to three sections, signal demodulation, bit synchronization and MSK carrier
detection.
Demodulation
Bit Synchronization
Carrier Detection
Clock
Output
Clocked
Data
Output
Unclocked
Data
Output
Carrier
Detect
Output
Comparator
w/Hysteresis
Sliding
Window
Averager
Invalid
Zero
Crossing
Indicator
Input
Receive
Filter
Zero
Crossing
Detector
Mono-
Stable
Discrim-
inator
Post
Detect
Filter
Compar
-ator
Data
Retiming
Digital
PLL
Receive
Input
Figure 14: MSK Demodulation Block Diagram.
3.2.1 Demodulator
The MSK detector demodulates the digital information from the received analog signal. There are six blocks that make up
the detector. Each block processes the output of the previous block. The output of the sixth block is recovered digital data.
The blocks are listed below and will be discussed in order of signal progression.
(a) Input Receive Filter
(b) Zero Crossing Detector
(c) Monostable
(d) Discriminator
(e) Post Discriminator Low Pass Filter
(f) Data Slicer/Comparator
3.2.1.1 RX Filter
The receivers input filter is a band pass filter centered around the main lobe of the MSK spectrum. It is designed to have
a passband width equal to the bit rate (i.e. 2100-900 = 1200Hz and 3000 - 600 = 2400Hz). The frequency domain
MSK and its Application to Wireless Data Transmission 15 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
impulse responses for both baud rates are shown below in Figure 15 and Figure 16. These graphs show the magnitude
responses of linear phase finite impulse response (FIR) filters. The filter weights are based on a windowed difference of
sinc( ) functions such as:
sinc
f t
f
sinc
f t
f
window_ function(t)
H
S
L
S
2 2

_
,

_
,

1
]
1
1
(28)
where:
f
H
is the upper cutoff frequency of the bandpass filter
f
L
is the lower cutoff frequency of the bandpass filter
f
S
is the sampling frequency of the bandpass filter
Most system simulation packages include tools for generating filter coefficients based on spectral templates, thus
coefficients are not tabulated. If an infinite impulse response (IIR) filter is used then it should be group delay equalized
over the passband. See section 3.4 for a simple model of an FIR filter.
For thorough coverage of digital filters refer to a digital signal processing text book such as [2],[5] or [8].
-72
-60
-48
-36
-24
-12
0
12
0 300 600 900 1200 1500 1800 2100 2400 2700 3000 3300 3600
G
a
in

(
d
B
)
Frequency (Hz)
Figure 15: Rx 1200 Baud Filter.
-72
-60
-48
-36
-24
-12
0
12
0 600 1200 1800 2400 3000 3600 4200 4800 5400 6000 6600 7200
G
a
in

(
d
B
)
Frequency (Hz)
Figure 16: Rx 2400 Baud Filter.
3.2.1.2 Zero Crossing Detector
The zero crossing detector monitors the incoming sinusoid signal from the RX input filter at a rate of N samples per symbol. When the
signal passes through zero the output is pulsed high for one sample period. The zero crossing detector then monitors for the next zero
crossing.
MSK and its Application to Wireless Data Transmission 16 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
-1
input 0
1
0
1
output
t1 t2 t3
Time
Figure 17:Plot of Zero Crossing Function.
Pseudo-Code Listing 2 Zero Crossing Detector
____________________________________________
I/O
input[i] = current input sample
output[i] = current output sample
____________________________________________
Pseudo Code
if sign(input[i]) != sign(input[i-1])
output[i] = 1;
else
output[i] = 0;
________________________________________
3.2.1.3 Monostable
The monostable monitors the pulsed output signal from the zero crossing detector. When a pulse is detected the output
of the monostable is set high and the timer is restarted. The value of the timer is set to 1/3 a symbol time for 1200 baud
and 1/2 symbol for 2400 baud. When the time-out period of the monostable expires the output is set low. If a new pulse
is detected on the input before the timer has expired, the timer is restarted and the output remains high. Figure 18 shows
the function of the monostable.
0
output
1
0
1
input
t1 t1+Tms t2 t2+Tms t3 t3+Tms
Time
Figure 18: Plot of Monostable Function.
MSK and its Application to Wireless Data Transmission 17 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
Pseudo-Code Listing 3 Monostable Model
___________________________________________
I/O, Variables, and Constants
input[i] // Current input sample
output[i] // Current output sample
T
ms
= N/3 // 1/3 Bit time @ 1200 Baud
T
ms
= N/2 // 1/2 Bit time @ 2400 Baud
timer = 0 // initialize timer first time
___________________________________________
Pseudo Code for Monostable
if input[i] = 1 then timer = T
ms
;
if timer > 0 then output[i] = 1;
else output[i] = 0;
if timer > 0 then decrement timer;
___________________________________________
3.2.1.4 Discriminator
The discriminator uses the output signal of the monostable to reconstruct the digital information. In Figure 19 and Figure
20, the operation of the discriminator is displayed. The signals in these figures are under perfect conditions without
passing through the receive filter. Under more realistic conditions the output the discriminator looks much less perfect and
hence needs to pass through the Post Discriminator Filter to look more like the transmitted bits. See the discriminator
output trace in Figure 31 to understand the need for the low pass filter.
3.2.1.4.1 1200 Baud Case
In Figure 19, the TX data in signal is included as a reference to the digital data being decoded. The TX MSK out signal is
the modulated signal that is applied to the zero crossing detector. As seen in this signal, a data bit one has 2 zero
crossings while a data bit zero has 3 zero crossings per bit time. The monostable output signal is the input to the
discriminator. (Refer to the monostable for an explanation of its output signal). The input signal to the discriminator is
time shifted to create two additional signals. The first is shifted by one sixth of a bit time and the second is shifted by one
third. These three signal are logically NANDed together to produce the discriminator's output signal. For a logic one the
zero crossings are spaced such that the output of the monostable is low for one sixth of a bit time before the middle of the
bit and again before the end of the bit. The logical NANDing of the three signals creates a pulse width equal to one half of
the bit period for each of the two zero crossings in a logic one. This creates a digital one with the appropriate bit timing.
For the zero bit, the zero crossings keep the monostable output high so the NAND output stays low.
3.2.1.4.2 2400 Baud Case
In Figure 20, the TX data in signal is included as a reference to the digital data being decoded. The TX MSK out signal is
the transmitted signal for which the discriminator is decoding. As seen in this signal, a data bit with the value of one has 1
zero crossings and a data bit with the value of zero has 2 zero crossings per bit time. The monostable output signal is the
input to the discriminator. The input signal to the discriminator is time shifted by one half of a bit period and is logically
NANDed with the input. The logical NANDing of the two signals creates the proper bit period. For a zero, the zero
crossings keep the monostable output high so the output of the NAND stays low.
MSK and its Application to Wireless Data Transmission 18 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
tx data in
tx msk out
monostable out
delay 1/6 symbol
delay 1/3 symbol
discriminator out
1 2 3 4 5 6 7 8 9 10
Time (Symbols)
Figure 19:1200 Baud Discriminator Signals.
tx data in
tx msk out
monostable out
delay 1/2 symbol
discriminator out
1 2 3 4 5 6 7 8 9 10
Time (Symbols)
Figure 20: 2400 Baud Discriminator Signals.
Pseudo-Code Listing 4 Discriminator Model
_____________________________________________________________________________________
I/O, Variables and Constants
input[i] // current input sample
output[i] // current output sample
Taps[] = {0, N/6, N/3} //1200 Baud
= {0, N/2} // 2400 Baud
state // variable used to compute logical And of delayed inputs
_____________________________________________________________________________________
Pseudo Code for Discriminator // runs at N x baud rate
state = 1 // initialize to true for each input sample.
for x = 0 to size of Taps-1
state = state & input[i - Taps[x]] // compute logical And of each delayed input
output[0] = complemented value of state // logical Invert
_____________________________________________________________________________________
MSK and its Application to Wireless Data Transmission 19 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
3.2.1.5 Discriminator Low Pass Filter
The discriminator low pass filter is designed to pass frequencies up to half the bit rate. This corresponds to the highest
frequency needed to pass an NRZ data sequence 101010... As seen in spectral plots in Figure 21 and Figure 22, the
discriminator low pass filter rolls off sharply beyond its cutoff to remove high frequency noise out of the discriminator.
These plots depict a finite impulse response filter with weights of a windowed sinc( ) function. See section 3.4 for a finite
impulse response filter model.
-72
-69
-66
-63
-60
-57
-54
-51
-48
-45
-42
-39
-36
-33
-30
-27
-24
-21
-18
-15
-12
-9
-6
-3
0
3
0 600 1200 1800 2400 3000 3600
G
a
in

(
d
B
)
Frequency (Hz)
Figure 21: 1200 Baud LPF.
-72
-69
-66
-63
-60
-57
-54
-51
-48
-45
-42
-39
-36
-33
-30
-27
-24
-21
-18
-15
-12
-9
-6
-3
0
3
0 1200 2400 3600 4800 6000 7200
G
a
in

(
d
B
)
Frequency (Hz)
Figure 22: 2400 Baud LPF.
3.2.1.6 Bit Slicer
The bit slicer is simply a comparator with a threshold half way between a one and a zero. That is it outputs 1 for inputs
greater than 0.5 and output 0 for inputs less than 0.5. Section 3.4 shows pseudo-code for a comparator. That model
provides for hysterisis; however for use as a bit slicer both the upper and lower threshold can be set to 0.5 to essentially
eliminate the hysterisis.
3.2.2 MSK Carrier Detection
The carrier detector is made up of three blocks, the bad zero crossing detector, sliding window filter and Comparator with hysterisis.
The carrier detect section monitors the information out of the zero crossing detector to determine whether the zero crossing are timed
properly for a 1200 or 2400 baud MSK signal. If the zero crossings are properly spaced a logic level is output to indicate that a valid
MSK signal is being received.
Bad Zero
Crossing
Detector
Comparator
w/Hyteresis
Sliding
Window
Averager
Carrier
Detect
Output
Carrier
Detect
Input
Figure 23: Carrier Detect Block Diagram.
MSK and its Application to Wireless Data Transmission 20 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
3.2.2.1 Bad Zero Crossing Detector
The Bad Zero Crossing Detector monitors the zero crossings of the filtered input signal. The time between zero crossings
is compared against a referenced. If the Bad Zero Crossing Detector determines that a zero crossing is not properly time
spaced the output pulses high, otherwise the output remains low. Figure 24 and Figure 26 show the timing windows used
for 1200 and 2400 baud. Figure 25 and Figure 27 show that with a properly timed input signal (i.e. zero crossings within
the window) the output is low. With an improperly timed input (i.e. zero crossing outside the window or end of window
reached) the output pulses high. The windows are based on the ideal zero crossing spacings +/- 5% of the bit time. Thus
each baud rate would have two valid windows, however the receive input filter introduces a slight intersymbol interference
at symbol boundaries and thus the windows need to widen towards each other to account for this. Since the windows of
the 1200 Baud case become very close, a single window is used.
invalid zc
good zc
0 1/6 1/3 1/2 2/3
Time (Symbols)
Figure 24: 1200 Baud Valid Zero Crossing Window
.
input
window
output
0
Time
Figure 25: 1200 Baud Bad Zero Crossing Detector.
MSK and its Application to Wireless Data Transmission 21 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
invalid zc
good zc
0 1/6 1/3 1/2 2/3 5/6 1 7/6
Time (Symbols)
Figure 26: 2400 Baud Valid Zero Crossing Window.
input
window
output
0
Time
Figure 27: 2400 Baud Bad Zero Crossing Detector.
Pseudo-Code Listing 5 Bad Zero Crossing Detector
_____________________________________________________________________________________
I/O, Variables and Constants
input[i] // current input sample
output[i] // current output sample
START // Array of valid-window START times
{Nx0.2833} // 1200 baud - one window
{Nx0.45, Nx0.8} // 2400 baud - two windows
STOP // Array of valid-window STOP times
{Nx0.55} // 1200 baud - one window
{Nx0.6, Nx1.05} // 2400 baud - two windows
counter // used to keep track of time since last zero crossing occurred
OK // flag used to indicate properly timed zero crossing
MSK and its Application to Wireless Data Transmission 22 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
_____________________________________________________________________________________
Pseudo-Code for Bad Zero Crossing Detector
OK = TRUE; // flag defaults TRUE for no zero crossing input
if (input[i] = 1) { // if zero crossing occurs this sample
OK = FALSE; // then flag defaults FALSE and window is checked
for j=0 to size of STOP and START arrays // loop over all valid-windows
if ((counter >= START[i]) & (counter <= STOP[i])) // check if counter is in window
then OK = TRUE; // set flag TRUE if in window
reset counter; // counter is reset at every zero crossing
}
if (counter > max. value in STOP array)
then { OK = FALSE; reset counter; } // window expired -> spacing is too large for MSK
else increment counter;
output[i] = logic inverse of OK; // output pulses high for bad zero crossings
_____________________________________________________________________________________
3.2.2.2 Carrier Detect Sliding Window Filter
The sliding window filter maintains a count of the number of bad zero crossings within the last L samples. It can be
modeled as a finite impulse response filter of length L with all coefficients set to 1. See section 3.4 for pseudo-code of a
finite impulse response filter. To mimic response time of MX-COM's serial MSK modems the length of the window should
be eight symbols so L=N x 8 where N is the number of samples per symbol.
3.2.2.3 Carrier Detect Comparator
The carrier detect comparator monitors the output of the sliding window filter to determine if the zero crossing spacings
are consistent enough to look like MSK. It has hysterisis to prevent chattering in noisy conditions. The comparator's
upper threshold is set to five while its lower threshold is set to one. Thus, it trips to the detect state when there are no bad
zero crossings in the last L samples and trips to the non-detect state when there are more than five. Section 3.4 shows
pseudo-code for a comparator with hysterisis. That models output should be inverted to have an active high MSK carrier
detect.
3.2.3 Bit Synchronization
The Bit Synchronization section re-times the data from the demodulation section. There are two blocks in this section, the Phase Lock
Loop block (PLL) and the re-timing block. The PLL creates the clock that is used to re-time the data.
Figure 28: Bit Synchronization Block Diagram.
MSK and its Application to Wireless Data Transmission 23 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
input data
clock output
1 2 3 4 5 6 7 8 9
Time (Bits)
Digital PLL with N=12
Figure 29: Bit Synchronization Plot.
3.2.3.1 Digital Phase Lock Loop
The PLL aligns the output clock rising edge to the middle of the input NRZ data bit. When the PLL is not aligned with the
incoming data a correction factor is applied to the PLL clock period until it becomes synchronized with the incoming data.
As seen in Figure 29 the output clock starts out of phase with the data and it becomes synchronized by the end of the
plot. In Figure 29 the correction factor used is one 12
th
of a bit period to more clearly show the locking process
Pseudo-Code Listing 6 Digital PLL
_____________________________________________________________________________________
I/O, Variables and Constants
input[i] // current input sample; input is NRZ data
output[i] // current output sample; output is clock which locks to NRZ data changes
X // the correction factor (Note: for MX-COM MSK modems X = 5% bit time)
N // number of samples per symbol
early // flag set TRUE if data transition is early
late // flag set TRUE is data transition is late
counter // a general counter
reset // flag set TRUE when algorithm determines it is time to reset counter
_____________________________________________________________________________________
Pseudo Code for First Order Digital PLL
reset = 0;
if counter<N/2-(0.5X) and input[0] not equal to input[1] // set early flag
then early = 1;
if counter>N/2+(0.5X ) and input[0] not equal to input[1] // set late flag
then late = 1;
if counter is equal to N - 1 - X and early and NOT late // apply correction factor
then reset = 1; // by resetting early
if counter is N-1 and NOT early and NOT late // apply no correction
then reset = 1; // by resetting on time
if counter is N-1 and early and late // reset on time since flags conflict
then reset = 1;
if counter is equal to N - 1+ X // apply correction factor
then reset = 1; // by resetting late
if counter > N/2 // set output clock phase based on counter state
then output = 0;
else output = 1;
MSK and its Application to Wireless Data Transmission 24 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
if reset is equal to 1 then { // reset flag is true then reset counter and early and late flags
counter = 0;
late = 0;
early = 0;
}
else increment counter;
_____________________________________________________________________________________
3.2.3.2 Data Re-timing
The date re-timing block aligns the data to the clock recovered by the PLL. The most common solution is a D flip-flop. The
data is applied to the D input of the flip-flop and the PLL output clocks the data through the flip-flop. A model of a D flip-
flop is shown in section 3.4.
3.3 System Simulations
Figure 30 shows an MSK system consisting of a transmitter, noisy channel, and a receiver. By simulating a complete
system with the models defined above, various signals are plotted in Figure 31. The transmit data bits have a 16 bit
preamble for bit synchronization. The receive filter output shows how noise degrades the band limited MSK signal. This
degradation is more clearly seen in the receive filter eye diagrams, Figure 32 and Figure 33. The discriminator output
clearly shows the need for the post discriminator low pass filter. The unclocked output exhibits jitter (i.e. variable length bit
duration) and shows the need for sampling it in the middle of the bit to reliably recover it. The clocked data output shows
relatively little jitter in the re-timed data. Additionally the carrier detect output and internal signals are shown to help the
reader understand their function. While the transmitter is not enabled many bad zero crossings are indicated. These
keep the sliding window filter output from going below the carrier detect threshold. After the start of transmission few bad
zero crossings are seen and the carrier detect output trips high when the sliding window filter output goes below the
threshold.
MSK
TX
+
MSK
RX
INPUT
BITS
OUTPUT
BITS
Transmitter Receiver Channel
NOISE
Figure 30: MSK Communications System with Noisy Channel.
MSK and its Application to Wireless Data Transmission 25 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
Tx Bits
Rx Filter Out
Discriminator Out
Unclocked Data Out
Rx Sync. Clock
Clocked Data Out
Carrier Det. Bad ZC
Carrier Det. Filt. Out
Carrier Det. Out
Tx Start End of Preamble
Time
Figure 31: System Simulation Signals for 1200 baud at 12dB SNR
0 1/2 1
Time (Bits)
Figure 32: RX Eye Diagram for 1200 Baud at 12dB SNR.
MSK and its Application to Wireless Data Transmission 26 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
0 1/2 1
Time (Bits)
Figure 33: RX Eye Diagram for 2400 Baud at 12dB SNR.
3.3.1 MSK Simulations with various SNR
Figure 34 shows the simulated bit error rate performance of the system depicted in Figure 30. The channel is assumed to
be additive white gaussian noise (AWGN). In the simulations the noise source produces white noise from DC to half the
sampling rate. Thus the total noise power is much higher than the noise in the bit rate bandwidth. Equation (29)
computes an effective SNR from a measured SNR assuming white noise over the measurement bandwidth.
SNR SNR 10log
BW
BW
Eff Meas
Meas
BitRate
+ (29)
which can be rearranged as:
SNR SNR 10log
BW
BW
Meas Eff
Meas
BitRate
(30)
To adjust the effective SNR to 12dB assuming:
AWGN source is a random variable with zero mean and variance =
2
= 1 i.e. dB power = 0dB.
MSK peak signal level = 1 i.e. signal power = 20log(RMS voltage level) = 20log
1
2

_
,

= -3dB
and a sampling rate 60 times the baud rate i.e. BW
Meas
= 30 times the baud rate
Correction Factor = 10log
BW
BW
10log
30
1
Meas
BitRate
= 14.77dB
thus SNR
Meas
= 12dB - 14.77dB = -2.77dB
The SNR without adjusting noise power level is:
SNR 10log
Signal Power
Noise Power
Meas
= -3dB
so it is necessary to decrease the noise power by 0.23 dB so:
noise variance = 10
0 23
10
.
= 0.9484
to arrive at other effective SNR we can simply adjust the noise variance by a scale factor:
MSK and its Application to Wireless Data Transmission 27 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
scale factor = 10
12
10
SNR
The table below shows the required noise variance for SNR near 12dB with the system sampling rate equal to 60 x baud
rate.
Effective SNR (dB) Noise Variance Scale factor from reference
13 0.753 0.794 = 1/1.259
12 0.948 1 i.e. 12dB is reference
11 1.194 1.259
10 1.503 1.585 = (1.259)
2
Table 3: Noise variance for various effective SNR.
1e-5
1e-4
1e-3
1e-2
1e-1
5 6 7 8 9 10 11 12 13
B
E
R
SNR (dB)
Figure 34: Simulated Bit Error Rate Performance.
3.3.2 FM System
Because of the relatively narrow bandwidth required by MSK, it can often be transmitted over existing analog radio links
with little or no modification. Figure 35 is a block diagram of FM transmission of MSK. Pseudo-code models for the key
blocks of this system are shown below.
FM
VCO
Transmitter Receiver Wireless Link
MSK
TX
MSK
RX
FM
Discriminator
Figure 35: FM Block Diagram.
3.3.2.1 FM modulation model
Pseudo-Code Listing 7 VCO based FM modulator model
_______________________________________________________________________________
MSK and its Application to Wireless Data Transmission 28 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
I/O, Variables and Constants
input[ ] // input sequence
output[ ] // output sequence
IntIn = 0 // variable used to store integral of input[] sequence up to current time.
step // difference between consecutive input samples normalized to interpolation rate M.
carrier // FM carrier frequency normalized to interpolation rate M
deviation // FM deviation frequency normalized to carrier frequency.

c

2 carrier
M
// carrier frequency constant

d

2 carrier deviation
M
// deviation frequency constant
_______________________________________________________________________________
PSEUDO-CODE // run for each sample, output sample rate is
// M times input sample rate
// use linear interpolation between input samples
step =
input[i] input[i -1]
M

;
for j=0 to M-1 {
IntIn = IntIn + input[i-1] + step j ; // simple integrator
output[M i + j] = ( ) sin
c d
+ + (i M j) IntIn ;
}
_______________________________________________________________________________
3.3.2.2 FM Discriminator Model
A simple heterodyne discriminator model based on Figure 35 can be built with multipliers, delays and filters. Pseudo-code
is shown for the mixer and the discriminator.
X
local
osc
IF
BPF
T
90
X
INPUT
LPF
OUTPUT
MIXER DISCRIMINATOR
Figure 36: Heterodyne Discriminator Block Diagram.
Pseudo-Code Listing 8 Mixer
MSK and its Application to Wireless Data Transmission 29 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
_____________________________________________________________________________________
I/O, VARIABLES and CONSTANTS
RFinput[i] // RF input
IFoutput[i] // IF output
// local oscillator frequency normalized to sampling rate of block

c

2 f
f
c
s
// note f
c
is offset from the RF carrier by the IF frequency.
_____________________________________________________________________________________
PSEUDO-CODE for MIXER
IFoutput[i] = RFinput[i] sin(
c
i);
_____________________________________________________________________________________
Pseudo-Code Listing 9 FM Discriminator
_____________________________________________________________________________________
I/O, VARIABLES and CONSTANTS
IFinput[i] // input to discriminator
DSoutput[i] // discriminator output
T
90
// Delay equal to one quarter of the IF center frequency
_____________________________________________________________________________________
PSEUDO-CODE for DISCRIMINATOR
DSoutput[i] = IFinput[i] IFinput[i - T
90
];
_____________________________________________________________________________________
3.4 General Purpose Models
This section list various general purpose models which the authors hope are self explanatory.
Pseudo-Code Listing 10 Finite Impulse Response Filter Model
_____________________________________________________________________________________
I/O, VARIABLES and CONSTANTS
input[i] // current input sample
output[i] // current output sample
L // FIR filter order
j // index to previous samples
sum // used to accumulate tap x weight products
weights[0..L-1] // array of filter coefficients
_____________________________________________________________________________________
PSEUDO-CODE // run for each sample, input and output sample rates are equal
sum = 0
MSK and its Application to Wireless Data Transmission 30 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
j = 0
while j<L {
sum = sum + (input[i-j] x weights[j]);
increment j;
}
output[i] = sum;
_____________________________________________________________________________________
Pseudo-Code Listing 11 Comparator with Hysterisis
_____________________________________________________________________________________
I/O, VARIABLES and CONSTANTS
input[i] // current input sample
output[i] // current output sample
Hth // upper threshold
Lth // lower threshold
_____________________________________________________________________________________
PSEUDO-CODE // run for each sample, input and output sample rates are equal
if (input[i] > Hth) // output trips high
then output[i] = 1
else if (input[i] < Lth) // output trips low
then output[i] = 0
else output[i] = output[i-1] // output stays same as last output
_____________________________________________________________________________________
Pseudo-Code Listing 12 D type flip-flop
_____________________________________________________________________________________
I/O, Variables, and Constants
D[i] // current input sample
C[i] // clock input
Q[i] // current output sample
state // used to store sampled
_____________________________________________________________________________________
Pseudo-Code
if (C[i] = 1) & (C[i-1] = 0) // rising edge of clock input
then state = D[i]; // sample input
Q[i] = state; // output state
_____________________________________________________________________________________
MSK and its Application to Wireless Data Transmission 31 APPLICATION
1997 MXCOM Inc. www.mxcom.com Tele: 800 638-5577 910 744-5050 Fax: 910 744-5054 Doc. # 20830084.001
4800 Bethania Station Road, Winston-Salem, NC 27105-1201 USA All trademarks and service marks are held by their respective companies.
3.5 References
[1] J.B. Anderson, T. Aulin and Carl-Erik Sundberg, Digital Phase Modulation, Plenum Press, New York, NY, 1986.
[2] R. E. Crochiere and L. R. Rabiner, Multirate Digital Signal Processing, Prentice Hall, Inc., Englewood Cliffs, NJ, 1983.
[3] M.L. Doelz and E.H. Heald, Minimum Shift Data Communications, US Patent No. 2,977,417, March 28, 1961.
Assigned to the Collins Radio Company.
[4] W. D. Greeg, Analog and Digital Communication, John Wiley and Sons, Inc. New York, NY, 1977.
[5] R. A. Haddad and T. W. Parsons, Digital Signal Processing - Theory, Applications, and Hardware, W. H. Freeman and
Company, 41 Madison Avenue, New York, NY, 1991.
[6] Kostedt, F. and Kemerling, J., Practical GMSK Data Transmission, Wireless Design and Development, vol. 3, No. 1,
pp. 21-25, January 1995.
[7] E. A. Lee and D.G. Messerschmitt, Digital Communications, Kluwer Academic Publishers, Boston, MA, 1988.
[8] A. V. Oppenheim and R. W. Schafer, Digital Signal Processing, Prentice Hall, Inc., Englewood Cliffs, NJ, 1975.
[9] J.G. Proakis, Digital Communications, 3rd Ed., McGraw-Hill, Inc., New York, NY 1995.
[10] T. S. Rappaport, Wireless Communications - Principles and Practice, ISBN 0-13-375536-3, Prentice Hall PTR, Upper
Saddle River, NJ, 1996
[11] Sklar, B., Digital Communications: Fundametals and Applications, Prentice-Hall, Englewood Cliffs, NJ, 1988.