IEEE TRANSACTIONS ON BROADCASTING, VOL. 36, NO. 3, SEPTEMBER 1990
SP-QPSK: A NEW MODULATION TECHNIQUE FOR SATELLITE AND LAND-MOBILE DIGITAL BROADCASTING
Hisakazu Katoh* Kamilo Feher
Digital Communications Research Laboratory Department of Electrical Engineering and Computer Science
University of California, Davis Davis, CA 95616
Fellow, IEEE
ABSTRACT
A new modulation technique, "SP-QPSK" (sinusoidal shaped d4-QPSK) suitable for land-mobile and satellite digital broadcasting systems applications is introduced. In digital data and/or sound broadcasting systems, which may have a relatively low bit rate transmission requirement, the residual phase noise introduced by the Doppler shift of moving vehicles presents a high bit error floor in coherently demodulated systems. To avoid this problem, non-coherent detection methods may have to be employed. Even though these systems require a higher C/N than their theoretical coherent counterparts in stationary AWGN environments, the overall performance of non-coherent systems is frequently superior in a mobile radio environment. In addition, to satisfy the high power and spectral efficiency requirements of emerging digital broadcast systems, nonlinear, saturated amplifiers may become essential subsystems.
We introduce a new modulation technique, SP-QPSK, which combines the advantages of the IJF-OQPSK narrowband satellite systems and of the d4-QPSK systems which have been adopted as the second generation land- mobile cellular standards, i.e., the US digital cellular standard.
Two-symbol-interval (TSI) QPSK modulation techniques [ l ] , such as IJF-OQPSK, SQAM, etc., have been developed and are in use in several satellite communication systems which require efficient nonlinear power amplification and reduced out-of-band radiation. However, these schemes need coherent detection due to the fact that they have "offset"
I and Q baseband channels by half a symbol duration. This
"offset" mode is not suitable for conventional differential non- coherent reception. On the other hand, for differential or discriminator detection, d4-QPSK [2; 3 and 61 is an improved QPSK-based modulation technique proposed for the US digital cellular standard. However, it has a major drawback in nonlinearly amplified systems, i.e., a significant out-of-band spectral radiation. We invented a new modulation technique, sinusoidal shaped d4-QPSK, or for short, "SP-QPSK," which has the combined advantages of the TSI-QPSK and also of the d4-QPSK systems.
2 . TSI-xI4-QPSK Systems
Even though the d4-QPSK modulator is basically a QPSK (four state) modulator, it has eight signaling phase 3ates as illustrated in Fig. 2.1. The principles of this modulation technique are described in [2; 3 and 61. The The performance of our new generation of SP-QPSK
systems is investigated by computer simulations a r d experimentally. Digital signal processing implementati.xi techniques have been used in the experimental prototype design.
We demonstrate that nonlinearly amplified SP-QPSK has a 1OdB lower out-of-band radiated power than conven- tional QPSK and it is suitable for differential and discriminator detection. Improved performance and simplified (non- coherent hardware) receivers could lead to novel digital broadcasting applications of this powerful modulation technique.
1 . INTRODUCTION
Digital data and sound broadcasting have been proposed for future satellite and/or terrestrial broadcasting system applications [ 8 ] . Some of the most important modulation-demodulation (modem) requirements include:
(1). nonlinear power amplifier to attain increased power efficiency of the transmitter; (2) discriminator or differential noncoherent detection for Doppler shift-caused phase noise- controlled mobile radio systems.
Fig. 2.1 Phase status of d4-QPSK
The connection between two states indicate the possible phase transition.
* Visitor at UC Davis 1989-1990
0018-9316/90/0900-0195$01.00 0 1990 IEEE
phase shift as a function of the information symbols is illustrated in Table 2.1. At every sampling instant each orthogonal axis may have a value of (I, 0, or -1) or (0.707 or
-
0.707). In our new modulation scheme we propose to connect both these values with an appropriate smooth curve which realizes TSI-based x/4-QPSK schemes. Fig. 2.2 shows the block diagram of an SP-QPSK modulator. In our design Nyquist filters used in conventional d4-QPSK are replaced by TSI waveform shaping circuits [l].
ReCtangUlar Triangle sinusoidal
Table 2.1 Phase shift as a function of information symbol
y(t)=D(i-l) : i-lStci
y(t) =( D ( i) - D ( i-1) )t/T+D(i-1) : i-16tCi y(t)=((D(i-l)-D(i))cos(a*t/T)+D(i-l)+D(i)I/2
: i-l6t<i
- -
"/?i n p u t C-- S/P m a p p i n g
-
Fig. 2.2 Block diagram of SP-QPSK modulator
-
o u t p u t - w a v e s h a p e
Fig. 2.3 and Table 2.2 show illustrative wave shapes which could be suitable for this design. A TSI wave shape may be regarded as the product between a binary on-off signal and the wave shaping circuit. The theoretical spectral density of these wave shapes is illustrated in Fig. 2.4 and Table 2.3 [4; 51.
1
0.9
Rectangular
Triangle
Sinusoidal
0.8
0.7
0.6
E
E 0.5
0.4
Parzen
Table 2.2 Expression of TSI wave shape [Ref. 11
spectrum 2nd side lobe -3dB width
TlsinlZnfT) -13dB lax
T.sinlrfT)' -27dB lda
-32dB 194
4 -53dB L82
2rfT T
2 (*fT)
2WfT(l-(fT)')
8,*fT/4
y(t) =(D( i-1) -D( i) ) (1-6( t/T) * ) (l+t/T) +D( i) y(t) =2 ( D ( i-1) - D ( i) ) (l-(t/T) 3+D(i)
:i-lSt<(2i-1)/2 : (2i-11/2<t<i y(t)=((D(i-l)-D(i))*
(0.42+0.5cos(~*t/T)+O.O8cos(2n*t/T))+D(i-l) : i-lSt<i
I I
note : T symbol duration i current sample instance
D(i) current sample value (-1, -0.707, 0, 0.707, 1)
S p e c t r u m of Wave Shape
0
-10 -20
-
-30 - 4 0-50
a
-60-10
-80
-90
- 100
0 2 4 6 8 10
ParZen
- Sinusoid -
Frequency [ I / T H r ]
Rcctongulor - Trionglc
Fig. 2.4 Theoretical characteristics of TSI wave shape (in Frequency domain)
-.,
c40 . n .
l/T[Hz]
(a) linear channel (b) hard limiter channel
Fig. 2.5 Simulated Nyquist filtered (a=0.2) power spectrum
lTT[Hz]
(a) linear channel (b) hard limiter channel
Fig. 2.6 Simulated rectangular shaped power spectrum
ITT[Hz]
(a) linear channel ( b ) hard limiter channel
Fig. 2.7 Simulated triangle shaped power spectrum
1 m z 1
(a) linear channel (b) hard limiter channel
Fig. 2.8 Simulated sinusoidal shaped power spectrum
I '
* .
*.
, I I < * e t . t , ,, I ' ( I "(b) hard limiter channel Fig. 2.9 Simulated Partzen shaped power spectrum
Sinusoidal Wave Shape
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1
0 -0.1 -0.2 -0.3 - 0 . 4 -0.5 -0 6 -0.7 -0.8 -0.9 - 1
Fig. 2.10 Baseband wave form on one orthogonal axis of SP-QPSK
. .
LPF
L i m i t e r
I
input
output p/ s
Threshold Comparator
Fig. 3.1 Block diagram of FM discriminator
Here, U(t), V(t) correspond to (1). As U(t) and V(t) have sinusoidal waveform, output signals can be easily computed.
Substituting (4) for (3) shown in Fig. 2.6, has the same spectrum as a conventional
binary PRBS signal. Triangular-shaped waves, Fig. 2.7, have a lower out-of-band spectrum. However, as there are discontinuities, it may not be suitable for specific band-limited channel applications. Sinusoidally shaped waveforms have a low out-of-band spectrum both in linear channel and in nonlinear channels. It has approximately a 10 dB spectral improvement at Af = 1.6/Tb from the carrier frequency. In addition it has a smooth waveform which has continuous derivatives. The Partzen shape, illustrated in Fig. 2.9 based on Ref. [4], also has a smooth waveform and has a low out-of- band spectrum. However, the second peak at f = 2.5/Tb is larger than that of the sinusoidally shaped signal. Hence sinusoidal shaping is the best one among the investigated TSI d4-QPSK waveforms. The sinusoidal shaping rule is rewritten here:
U ( t ) = { ( l ( i - 1 ) - I ( i ) ) c o s ( x t / T ) + I ( i - l ) + l ( i ) } / 2 ( l a ) V ( t ) = { ( Q ( i - l ) - Q ( i ) ) c o s ( x t / T ) + Q ( i - l ) + Q ( i ) } / 2 ( l b ) where U(t) and V(t) are the orthogonal signals of SP-QPSK, I(i) and Q(i) are the values at the sampling instants i, and T is symbol duration.
Fig. 2.10 shows a computer generated baseband waveform of an SP-QPSK modulator.
3 . DEMODULATION METHODS
For power efficient applications, saturated amplification may be required. This mode of operation is frequently used in numerous communication applications, especially in satellite broadcasting. Because of nonlinear amplification of non-constant envelope bandlimited QPSK type of signals, the received demodulated signal is distorted as compared to the constellation shown in Fig. 2.1. The information is contained in the phase shift between two symbol durations. Hence the combination of a hard limiter (extremely fast AGC) and an FM discriminator is suitable for stable reception. In Fig. 3.1 a block diagram of the demodulator, based on FM discrimination, is illustrated. Instead of Integrate and Dump filtering, which is often used for analysis of FM discriminators, Butterworth lowpass filters are used because band-limited channels are assumed in this case.
The output of the discriminator becomes four-level; each level expresses 3x14, ~ 1 4 , 4 4 , -3x14 phase shift between two-symbol duration. Therefore gray encoded signals can be decoded. Data is regenerated from differentially detected phase information. Transmitted data is recovered with parallellserial conversion and gray code conversion recovers the original information from these four levels. The analysis of the output of the discriminator is as follows:
The input signal of the discriminator is
S(t) = A COS (2xfct
+
$(t)} (2) here f, is a carrier frequency. When z is a small amount of delay the output of the differential detector after lowpass filter is(3) In this case as S(t), formula (2), is the sum of two orthogonal signals, phase information satisfies the next.
(4) O(t) =sin {$(TI - Q(t-7))
tan ($(t)) = V(t) / u(t)
U2(t)+V2(t) dW(t-T)+v(t-T)
The computed waveform of (5) is Fig. 3.2. As shown in this figure, four phase states are detected. Therefore a three- level baseband threshold detector can distinguish these four symbols [3; 61.
d i f f e r e n t i a l d e t o u t p u t
4 E
Fig. 3.2 Simulated output of FM discriminator
LPF
H 5-
i n p u t OI P/ S o u t p u t
LPF
5-
Threshold C o m p a r a t o r
Fig. 3.3 Block diagram of differential detection
Fig. 3.3 is the block diagram of the non-coherent, differential detection. In a relatively low bit rate system a long delay line is necessary for differential detection. The differential operation may be performed at a low IF frequency. In this case a DSP implementation is suitable.
The output signal at this point is as follows:
Differential operation is processed at each multiplier.
The output signals after these lowpass filters are I(t) = U(t) U(t-T)
+
V(t) V(t-T)Q(t) = V(t) U(t-T) - U(t) V(t-T)
where U and V correspond to the baseband signal expressed in Eq. (1). The computed waveform of these outputs is shown
in Fig. 3.4. As it needs only a two-level threshold detector.
accuracy of the detection is more improved than the F!;
discriminator detection using the three-level threshold decision mentioned above. Except for the difficulty 01 long delay line, the differential detection may also be suitable for the demodulator.
Fig. 4.1 Baseband waveform of SP-QPSK (I channel output)
(a) Eye diagram on linear channel
(b) Eye diagram on hard limiter channel Fig. 3.4 Simulated output waveform of differentiai
detection
4 . EXPERIMENTAL RESULTS
Performance of our SP-QPSK is measured with experimental hardware designed at UC Davis. DSP techniques have been used in this 250 kb/s modem design.
Fig. 4.1 and Fig. 4.2 illustrate the measured baseband waveforms of one orthogonal axis and a constellation diagram generated mutually from this modem.
Spectrum spreading is also measured. Fig. 4.3 shows the comparison of spread spectrum among IJF-OQPSK, SQAM and SP-QPSK. SP-QPSK has 10 dB advantage compared with QPSK and has the same spectrum as the IJF- OQPSK nonlinearly amplified system [8].
For demodulation an FM discriminator was designed.
Fig. 4.4 is the band-limited spectrum at the IF input of the demodulator. Fig. 4.5 is the experimental waveform at output of the FM discriminator.
Fig. 4.2 Constellation of SP-QPSK
(H : 200 kHz/div, V : 10 dB/div) Fig. 4.4 Received spectrum
(c) IJF-OQPSK (upper) SP-QPSK (lower)
(d) SQAM (a=0.8) (upper) SP-QPSK (lower)
( e ) IJF-OQPSK (upper) SP-QQPSK (lower) (after hard limiter)
bit rate : 250 kbts horizontal : 100 kHz/div vertical : 10 dB/div Fig. 4.3 Measured spectrum
(f) SQAM (a=0.8) (upper) SP-QPSK (lower) (after hard limiter)
Fig. 4.5 Measured waveform of demodulator (FM discriminator output)
5 . CONCLUSION
A new modulator "SP-QPSK is described. It has a good performance in nonlinearly amplified channels. It can be realized with simple hardware. Therefore it is useful for power and spectrally efficient systems, especially digital broadcasting. As a demodulator an FM discriminator is suitable for eliminating the Doppler-caused phase noise effect of mobile receivers.
6 . REFERENCES
K. Feher, Advanced Digital Communication.
Englewood Cliffs, NJ: Prentice-Hall, 1987.
P.A. Baker, "Phase Modulation Data Sets for Serial Transmission at 2000 and 2400 Bits per Second,"
AlEE Transactions on Communications arid Electronics, July 1962, pp. 166-1 71.
Y. Akaiwa and Y. Nagata, "Highly Efficient Digital Mobile Communication with a Linear Modulation Method," JFFF Journa I on Selected Areas
in
Corn-, Vol. SAC-5, No. 5 , June.
F.J. Harris, "On the Use of Windows for Harmonic Analysis with the Discrete Fourier Transform,"
Proc.
-
IEEE, Vol. 66, No. 1 , January 1978, pp. 51-84.E.O. Brighan, The Fast Fourier Transform and Its Applications. Englewood Cliffs, NJ: Prentice-Hall, 1988.
C-L. Liu and K. Feher, "Noncoherent Detection of d4-QPSK Systems in a CCI-AWGN Combined
N.G. Kingsbury, "Transmit and Receive Filters for QPSK Signals to Optimize the Performance on Linear and Hard-Limiter Channels," IEE Proc., Vol. 133, Pt. F, No. 4, July 1986.
D. Pommier, P.A. Ratliff, "New Prospects for High Quality Digital Satellite Sound Broadcasting to Mobile, Portable and Fixed Radio Receivers," Published by EBU (European Broadcasting Union), ITU-WARC-ORB
W.,
Geneva, September 1988.. .
Interference Environment," JFFF VTC -89, pp. 83-94.
