1. MSK PSD Quadrature MSK baseband waveforms are
Bandpass MSK signal (IQ representation)
IQ waveforms are orthogonal (independent)
PSD of complex envelope g(t) is
and since x(t) & y(t) have same basic
shape
m(t) = ±1
ECE 4710: Lecture #30

2. MSK PSD MSK PSD is Pulse shape is truncated cosine (or sine) over 2Tb
FT of truncated cosine is
ECE 4710: Lecture #30

3. MSK Two types of MSK PSDs for both MSK types (I & II) are the same
Type I
Pulse shape on x(t) and y(t) alternates between positive and negative half cosinusoid
Differential encoding  Fast Frequency Shift Keying (FFSK)
Yields one to one relationship between ±1 m(t) data and fH / fL
Type II
Pulse shape on x(t) and y(t) is always a positive half cosinusoid
No one to one relationship between m(t) data (±1) and fH / fL
fH / fL determined by m(t) data and encoding
PSDs for both MSK types (I & II) are the same
ECE 4710: Lecture #30

4. Note that MSK is Binary M = 2
MSK PSD
Complex envelope PSD
Observations
FNBW
MSK  0.75 R
QPSK  0.5 R
50% smaller!!
1st sidelobe
QPSK   13.4 dB
MSK  23 dB!!
MSK
QPSK or OQPSK
Note that MSK is Binary M = 2
ECE 4710: Lecture #30

5. MSK vs. QPSK BW MSK signal has 50% larger FNBW relative to QPSK
Need wider channel BW for RF signal
Spectral efficiency is not as good
MSK sidelobes are much smaller than unfiltered QPSK
Truncated cosine pulse shape for MSK
Rectangular pulse shape for unfiltered QPSK
Adjacent Channel Interference (ACI) from MSK is very good compared to unfiltered QPSK
ECE 4710: Lecture #30

6. GMSK
MSK baseband waveforms can be filtered to further reduce sidelobe levels
Raised cosine (RC) filter for QPSK
Eliminate all sidelobes
Satisfies Nyquist criterion  No proper sampling point within symbol period
RC filter cannot be used for filtering MSK envelope
MSK is constant envelope  enables non-linear Class C PA
RC filtered MSK would have sidelobes regenerated by non-linear Class C PAs
ECE 4710: Lecture #30

7. GMSK
Gaussian MSK = GMSK  use Gaussian shaped filter to further improve spectral efficiency of MSK
Filter rectangular m(t) waveforms prior to generating IQ baseband waveforms x(t) & y(t)
Before data are frequency modulated on carrier
Cannot filter x(t) & y(t) as that would make MSK not have a constant envelope
Gaussian filter transfer function
ECE 4710: Lecture #30

8. GMSK Gaussian Filter ECE 4710: Lecture #30
Significantly reduces spectral sidelobes
Affect on FNBW is very minor for reasonable filter BW
Does NOT satisfy Nyquist criterion
Will cause unwanted ISI if BG is too narrow
How do we quantify reasonableness or narrowness?
Bandwidth-Bit Duration Product  BG Tb
MSK FNBW is 0.75 R = 0.75 / Tb
BG Tb provides normalized measure of filter BW wrt signal BW
Tradeoff  narrow filter BW vs. increasing ISI
ECE 4710: Lecture #30

9. GMSK ECE 4710: Lecture #30 Gaussian Filter 0.3 GMSK MSK QPSK or OQPSK
Reasonable BG Tb
0.3 to 0.5
Greater than 0.5?
sidelobe levels not
reduced enough
Less than 0.3?
ISI becomes too large
GMSK normally specified by
amount of filtering
Example: 0.3 GMSK is
GMSK with BG Tb = 0.3
MSK
QPSK or OQPSK
0.3 GMSK
ECE 4710: Lecture #30

10. GMSK
Even though m(t) data is shaped by Gaussian filter response GMSK is still a constant envelope modulation method just like MSK
Gaussian filtered data used to frequency modulate carrier
Carrier envelope remains constant
Non-linear Class C amps used
Excellent DC to RF efficiencies (80-90%)
Enables long operation of mobile devices relying upon battery power supply
Constant envelope also means GMSK is less susceptible to signal fading and interference which occurs during transmission thru channel
ECE 4710: Lecture #30

11. GSM Global System for Mobile (GSM) ECE 4710: Lecture #30
First digital standard developed for cellular telephone
Developed in Europe in late 1980’s and widely deployed in early 1990’s
Long before digital PCS cell phones in U.S. in late 1990’s
Significant use in non-European markets (Asia, South America, etc.)
Most widely used 2G cell phone standard in the world
Leads market share of all other technologies by factor of 3-4
T-Mobile and AT&T Mobile in U.S. used GSM after 2002
Modulation method for GSM is 0.3 GMSK
ECE 4710: Lecture #30

12. MSK Generation Type 1 MSK with differential encoding
Very simple use of FM Tx with differentially encoded m(t)
Doesn’t require separate IQ waveforms, x(t) & y(t), generated from m(t)
No need for two IQ oscillators
Simple = Fast  Fast Frequency Shift Keying (FFSK)
ECE 4710: Lecture #30

13. MSK Generation Type I MSK with no differential encoding
IQ waveforms must be generated !!
Parallel method of generation
ECE 4710: Lecture #30

14. MSK Generation Type II MSK ECE 4710: Lecture #30
MSK is specific case of BPSK
Serial generation filters BPSK RF (bandpass) signal with off-center BPF
ECE 4710: Lecture #30

15. MSK Spectral Efficiency
MSK is special case of BFSK with minimum DF
NNBW spectral efficiency is better than BPSK but worse than QPSK
30-dB BW spectral efficiency is 4  better than QPSK
Low sidelobe levels  low ACI
Also better than 64 QAM!
Note that BPSK with RCF & r = 0.5 has h =  same as MSK
ECE 4710: Lecture #30

16. MSK & GMSK MSK & GMSK for mobile radio applications
Low sidelobe level and low ACI
Large number users spaced very close together in frequency domain with minimal interference between channels
Support large number of users
Non-coherent Rx for demodulation
No carrier synchronization
Simple & inexpensive Rx
Constant envelope
Non-linear Class C PA with high DC to RF efficiency
Long battery life for mobile units
ECE 4710: Lecture #30