PAPER A Simple Method of BER Calculation in DPSK/OFDM Systems over Fading Channels

Transcription

1 366 IEICE TRANS. FUNDAMENTALS, VOL.E88 A, NO. JANUARY 5 PAPER A Simple Metho of BER Calculation in DPSK/OFDM Systems over Faing Channels Fumihito SASAMORI a), Shiro HANDA, an Shinjiro OSHITA, Members SUMMARY In orthogonal frequency ivision multiplexing (OFDM) systems with ifferential phase shift keying (DPSK), it is possible to apply ifferential moulation either in the time or frequency omain epening on the conition of faing channels, such as the Doppler frequency shift an the elay sprea. This paper proposes a simple calculation metho, that is, an approximate close-form equation of the bit error rate (BER) in DPSK/OFDM systems mentione above over both time an frequency selective Rician faing channels. The valiity of the propose metho is emonstrate by the fact that the BER performances given by the erive equation coincie with those by Monte Carlo simulation. key wors: DPSK/OFDM system, Doppler frequency shift, elay sprea, bit error rate, close-form equation. Introuction Intensive research has been conucte on the aaptation of orthogonal frequency ivision multiplexing (OFDM) to high-rate wireless ata transmission systems because of its high spectrum efficiency an its resistance to multipath faing [], []. Besie the above avantages, however, intersymbol interference (ISI) an inter-carrier interference (ICI) give intolerable egraation to the bit error rate (BER) performance over faing channels [] [5]. The ISI, which is cause by the elay sprea, can be easily neglecte when multipath elay is within a guar interval (GI), while the ICI, which is cause by the Doppler frequency shift, cannot be cancele without ieal estimation at a receiver. In OFDM systems with ifferential phase shift keying (DPSK), it is possible to apply ifferential moulation either in the time or frequency omain epening on the conition of faing channels, such as the Doppler frequency shift an the elay sprea [6], [7]. Besies the ICI, the BER performance in those systems is influence by time or frequency correlation between ajacent symbols. Generally, the symbol perio in OFDM systems is much longer than that in single carrier systems in orer to reuce the effect of the elay sprea. Owing to this fact, inter-symbol fluctuation in the time omain cause by the Doppler frequency shift is le to be large, while inter-carrier fluctuation in the frequency omain cause by the elay sprea is le to be small. Consequently, there exists a merit to aopt the ifferential mo- Manuscript receive May 3, 4. Manuscript revise September 6, 4. Final manuscript receive October 5, 4. The authors are with the Faculty of Engineering, Shinshu University, Nagano-shi, Japan. a) fsasa@shinshu-u.ac.jp This work was supporte by MEXT, Grants-in-Ai for Young Scientists (B) ulation in the frequency omain epening on the channel conition. Theoretical analyses of the BER in OFDM systems with DPSK have been propose [7] [9]. In [8] an [9], the BER performance has been analyze consiering frequency selective Rayleigh an Rician faing channels, but only in the time omain ifferential moulation. In [7], taking account of the merit mentione above, both time an frequency omain ifferential moulations over Rayleigh faing channels have been investigate with close-form BER expressions. But the ICI nees to be actually calculate from the Rayleigh faing channel moele by a tappeelay line, which is calle as a elay profile moel. There exist many elay profile moels in frequency selective faing channels, then this forces the calculation of the BER to be complexity. Many theoretical analyses of factors in egraing BER have been conucte (e.g., []). Among them, a closeform equation of the BER coul be helpful in esign an evelopment of mobile raio systems, because it is important to have a quantitative unerstaning of the influence of each factor upon the BER. This paper proposes a simple calculation metho, that is, an approximate close-form equation of the BER in DPSK/OFDM systems over both time an frequency selective Rician faing channels, taking account of the ifferential moulation in both time an frequency omains. A simple means that the BER can be easily an promptly calculate by only substituting several parameter values of the channel conition into the erive close-form equation. The rest of the paper is organize as follows. Section contains a system moel. Section 3 presents a statistical erivation metho of the equation. In Section 4, the valiity of the propose metho is emonstrate by the fact that the BER performances given by the erive equation coincie with those by Monte Carlo simulation. Finally, the conclusion is in Section 5.. System Moel Fig. shows a block iagram of DPSK/OFDM systems using N-point IFFT/FFT in an equivalent low-pass system, where DPSK inclues the binary PSK (DBPSK) an the quarature PSK (DQPSK). In a moulation block, the followings are carrie out to a binary ata sequence: serialto-parallel conversion, DPSK moulation with ifferential coing in the time or frequency omain, OFDM moulation Copyright c 5 The Institute of Electronics, Information an Communication Engineers

2 SASAMORI et al.: A SIMPLE METHOD OF BER CALCULATION IN DPSK/OFDM SYSTEMS OVER FADING CHANNELS 367 Fig. Block iagram of DPSK/OFDM systems in equivalent low-pass system. Fig. 3 Vector of receive signal in space iagram. Fig. Differential moulation in time or frequency omain. carrier systems with DBPSK [], []. In this section, we will introuce a full etail of the extene statistical analysis incluing the frequency fluctuation cause by the elay sprea, an then a few comments will be mae on overlapping points. using IFFT, parallel-to-serial conversion an aition of the GI. It is assume that the ISI can be completely avoie with the GI which is longer than the maximum multipath elay. In a transmission block, an OFDM signal is subjecte to the faing an is ae the white Gaussian noise (AWGN). In a emoulation block, the esire binary ata sequence is emoulate from the receive OFDM signal through a reverse process of the moulation block. Fig. illustrates the metho of the ifferential moulationindpsk/ofdm systems, where f n (n =,,, N)is a subcarrier frequency, T s (= T + T g ) is an OFDM symbol perio, which is compose of a GI perio T g an a ata symbol perio T. With the ifferential moulation in the time omain, which is calle as DMT hereafter, the symbol #b is ifferentially encoe accoring to binary input ata consiering the symbol #a as a reference symbol. On the other han, with the ifferential moulation in the frequency omain, which is calle as DMF hereafter, #c is ifferentially encoe consiering #a as a reference symbol. 3. Statistical Analysis of BER Performance Generally in wireless communication systems, the BER is egrae by the time an frequency fluctuations over faing channels. In OFDM systems, the BER is furthermore egrae by the ICI because of the loss of orthogonality among subcarriers. We have investigate the effect of the time fluctuation cause by the Doppler frequency shift in the single- 3. Effect of Time an Frequency Correlations Fig. 3 shows a vector of the receive signal affecte by the AWGN an the Rician faing which has a line-of-sight (LOS) path an non-los iffuse path. Let z r an z be respectively reference an emoulating symbol vectors at a symbol timing, then an output signal v o is given by [] v o = Re ( ) z r z ( ) = r 4 r, () where r = z r + z r = z r z. () Assuming that a binary ata sequence is compose of all +, a bit error occurs when v o becomes minus an then the BER in DBPSK can be expresse as P b = Prob (v o < ) = Prob (r < r ). (3) In orer to calculate (3), mean values an variances of r an r must be investigate. Let A be an envelope of the irect wave through a LOS path. Assuming a frequency offset of local oscillator to be very small, mean values of r an r become E[r ] = E[ z r + z ] = A E[r ] = E[ z r z ] =, (4)

3 368 IEICE TRANS. FUNDAMENTALS, VOL.E88 A, NO. JANUARY 5 since z r an z have the same polarity. Let σ s an σ n be variances of the Rayleigh wave through the iffuse path an the AWGN, respectively, then variances of r an r become E [ (r E[r ]) ] = ( σ s + σ n) + ρσ s E [ (r E[r ]) ] = ( σ s + σ n) ρσ s, (5) where ρ is the covariance between z r an z,thatis,thefaing correlation. As shown in Fig., z r an z in DMT have a time interval of T s in the same subcarrier, then ρ means the time correlation. On the other han, z r an z in DMF have a frequency interval of /T in the same time, then ρ means the frequency correlation. The time correlation function ρ( t) is wiely known as ρ( t) = J (π f D t), (6) where f D is the maximum Doppler frequency, J ( ) isa Bessel function of the first kin of zeroth orer. On the other han, the frequency correlation function ρ( f ) can be erive by the Fourier transform of a multipath power elay profile S (τ) as follows [3]: ρ( f ) = S (τ)e jπ f τ τ. (7) A reference power elay profile with the rms elay sprea σ τ for most environments is specifie as [4] S (τ) = σ τ e τ στ, for τ, (8) which is calle as an exponential profile. From (7) an (8), ρ( f ) an its envelope characteristic ρ( f ) can be easily calculate as ρ( f ) = (9) + jπσ τ f ρ( f ) =. () + (πσ τ f ) Consequently, the covariance ρ in (5) can be obtaine by substituting t = T s an f = /T into (6) an (), respectively: J (π f D T s ), for DMT ρ = ( ) = + π στ + { } π(+α) στ, for DMF T Ts α = T g /T, () where α is calle as GI factor hereafter. Defining the variances of r an r as σ an σ,respectively, an rearranging (5), σ an σ are given by σ = σ s( + ρ) + σ n σ = σ s( ρ) + σ n. () From (4) an (), probability ensity functions of r an r can be expresse as p(r ) = r exp r + 4A σ σ p(r ) = r exp r σ σ I Ar σ, (3) which are calle as Rician istribution an Rayleigh istribution, respectively. Finally, the BER in (3) can be erive after integration an rearrangement as follows: P b = Prob(r < r ) { } = p(r ) p(r )r r r ( ρ) K+ = + ( exp K K+ K+ + ) K+ +, (4) where is the ratio of energy per bit to the spectral noise ensity (E b /N ) on the receive signal, an K (= A /(σ s)) is the Rician K-factor. The faing correlation ρ in () consequently causes the BER egraation of the emoulating symbol. The time an frequency fluctuations give rise to not only the BER egraation but also the interference to the other symbols of all subcarriers, which egraes the BER furthermore in OFDM systems. We will analyze the interference in the following subsections, which is base on the concept of complex Fourier series. Fig. 4 illustrates the complex Fourier coefficients c k which is given by c k = T T π jk f (x)e T x x, (5) where T is the Nyquist interval. Now f (x) is assume to be the interference which occurs at a certain subcarrier, the complex Fourier coefficients of f (x) are classifie into two cases:. c k (k ) is an alternating current (AC) component of f (x), an c k can be consiere as the inter-carrier interference (ICI) power,. c is a irect current (DC) component of f (x), an c can be consiere as the intra-symbol interference power only in the case of DQPSK. The etails will be escribe below. 3. Effect of Inter-Carrier Interference When a carrier cos(π f c t) passes through the faing channels, the receive carrier frequency is affecte by Doppler frequency f D cos θ, whereθ is a ranom phase with the uniform istribution of [, π). Then the receive carrier r(t) can be expresse as [5] r(t) = cos {π( f c + f D cos θ)t} cos(π f c t) (π f D t cos θ)sin(πf c t), for π f D t cos θ. (6)

4 SASAMORI et al.: A SIMPLE METHOD OF BER CALCULATION IN DPSK/OFDM SYSTEMS OVER FADING CHANNELS 369 Fig. 4 Complex Fourier coefficients. Fig. 5 Intra-symbol interference. The secon term of (6) means the ICI from the loss of orthogonality among subcarriers, which occurs for both DMT an DMF. From (5) an Fig. 4, the interference power to a subcarrier k (k ) can be obtaine as c k = T π jk T π f D t cos θe t t T = ( f DT cos θ), for k. (7) k It is foun from (7) that the interference power from a remote subcarrier k l (large number of k) is kept to be small. Therefore, by using the symmetricity of the complex Fourier coefficients an Euler zeta function, the total power of ICI I f D a as from subcarriers k ( k l k k l, k ) can be calculate k l a = c k I f D k= (π f DT cos θ), for k l, (8) 3 where it is assume that the number of subcarriers in OFDM systems is very large. Finally, the average power I a of ICI cause by the Doppler frequency shift can be erive by averaging (8) by uniformly istribute θ as follows: I a = π π (π f D T cos θ) θ 3 = (π f DT ) 6 = (π f DT s ), for DMT an DMF. (9) 6( + α) One thing which we woul like to mention here is that, against the elay sprea, the carrier frequency f c keeps invariable an the orthogonality among subcarriers consequently stays as it is. 3.3 Effect of Intra-Symbol Interference Fig. 5 illustrates a concept of the intra-symbol interference, where ψ is a phase rotation angle of the emoulating symbol z cause by the faing. This rotation gives rise to the egraation of the symbol energy an the interference to the quarature channel component. As mentione above, the ifferential etection in DMT is carrie out by the phase ifference between z r an z which have a time interval of T s in the same subcarrier. Then the ψ in terms of time can be expresse as ψ(t) = π f D t cos θ, for DMT, () which is cause by the Doppler frequency shift. On the other han in DMF, z r an z have a frequency interval of /T in the same time, an then the ψ in terms of frequency can be expresse as ψ( f ) = tan (πσ τ f ), for DMF, () which is cause by the elay sprea an is erive by taking the angle of (9) into account. Therefore, the receive carrier r(t) with the phase rotation ψ becomes r(t) = cos (π f c t + ψ) = cos ψ cos(π f c t) sin ψ sin(π f c t). () In (), cos ψ an sin ψ mean the egraation of the symbol energy an the interference to the quarature channel component, respectively. When assuming π f D t cos θ an πσ τ f, cos ψ an sin ψ in () can be approximate as cos (π f D t cos θ), for DMT cos ψ =, for DMF (3) +(πσ τ f ) sin (π f D t cos θ) π f D t cos θ, sin ψ = for DMT. (4) πσ τ f +(πσ πσ τ f ) τ f, for DMF Consequently, the egraation of the symbol energy can be neglecte, while the interference to the quarature channel component must be consiere as the intra-symbol interference. One thing to which we must pay attention here is that the intra-symbol interference occurs only in the case of DQPSK. The average interference power to the quarature channel component can be calculate by taking account of a DC component c of complex Fourier coefficient in (5). an its average power From (4), the interference power I f D cause by the Doppler frequency shift become Ī f D

5 37 IEICE TRANS. FUNDAMENTALS, VOL.E88 A, NO. JANUARY 5 ( I f D Ts ) = π f D t cos θt T s = (π f D T s cos θ) π Ī f D = (π f D T s cos θ) θ π = (π f DT s ), for DMT. (5) In the same way, the interference power I σ τ cause by the elay sprea becomes I σ τ = = = T ( π σ τ T T ) πσ τ ff { π( + α) σ } τ, for DMF. (6) T s Finally, the average interference power I quarature channel component can be expresse as to the, for DMT an DMF in DBPSK (π f I = D T s ), for DMT in DQPSK. (7) { } π( + α) σ τ T s, for DMF in DQPSK 3.4 Carrier-to-Noise Plus Interference Power Ratio Due to the inter-carrier an intra-symbol interferences mentione above, the BER expresse by (4) is more egrae. Here, the esire equation will be erive by introucing the Gaussian approximation [], [5]. Concretely speaking, it is necessary to erive the carrier-to-noise plus interference power ratio (CNIR) which involves the interference power estimate as the Gaussian noise. It is a point to notice that the two interference powers shoul be inclue in the CNIR of the receive Rayleigh wave since those interferences are cause by Rayleigh faing. /(K + ) in (4) inicates the E b /N of the receive Rayleigh wave [], which can be rewritten as K + = γ NI, for DBPSK γ NI, (8), for DQPSK where γ NI means the CNIR of the receive Rayleigh wave. By regaring the interference powers I a an I as the increase of isturbance powers besies the AWGN against the signal power, γ NI can be expresse as γ NI =, (9) γ N + I a + I where γ N is the CNR of the receive Rayleigh wave. Moreover, the CNR γ N can be calculate from the total E b /N EN of the receive signal (Rician wave) as follows: EN +α K+ γ N = = EN (K+)(+α), for DBPSK EN, (3) +α K+ = EN (K+)(+α), for DQPSK where /(+α) means the energy loss cause by the removal of the GI. From (9) an (3), (8) becomes K + = (K+)(+α), for DBPSK +I a +I EN. (3), for DQPSK (K+)(+α) EN +(I a +I ) Finally, substituting () an (3) into (4) leas the approximate close-form equation of the BER in DPSK/OFDM systems over both time an frequency selective Rician faing channels. The interference powers I a an I in (3) have been represente in (9) an (7), respectively. This metho can simply calculate the BER from the values of the maximum Doppler frequency normalize by the symbol frequency f D T s, the rms elay sprea normalize by the symbol uration σ τ /T s, the Rician factor K, the GI factor α an the E b /N of the receive signal EN. 4. Numerical Evaluation We will justify the propose metho by comparing the BER performances given by the erive equation with those by Monte Carlo simulation, which is precisely constructe base on the system moel in Section. Figs. 6 an 7 show the BER performances over Rayleigh an Rician faing channels, respectively. The elay profile moel in the simulation is assume to be an exponential profile in (8). It is foun from these figures that the approximate BER performances coincie with the simulation ones at any conition, that is, at any value of f D T s, σ τ /T s, K, α an EN. The equation can precisely express the following characteristics of the BER performance in OFDM systems over faing channels which are qualitatively an wiely known:. As shown in Fig. 6(a), the larger rms elay sprea σ τ /T s an the larger GI factor α cause the performance in DMF to be more egrae, because the frequency correlation in () becomes small, an moreover, the intra-symbol interference in (7) an the energy loss in (3) become large.. As shown in Fig. 6(a), the variation of σ τ /T s gives no effect on the performance in DMT, because each subcarrier in OFDM is affecte by only frequency-flat faing an the ISI which is cause by the elay sprea can be perfectly cancele by the GI. 3. As shown in Fig. 6(b), the performance in DMT is more sensitive to the Doppler frequency f D T s than that in DMF. That is because the ICI in (9) gives the effect on both DMT an DMF, while the time correlation in () an the intra-symbol interference in (7) give the effect on DMT only.

6 SASAMORI et al.: A SIMPLE METHOD OF BER CALCULATION IN DPSK/OFDM SYSTEMS OVER FADING CHANNELS 37 Fig. 6 BER performances over Rayleigh faing channels. Fig. 7 BER performances over Rician faing channels. 4. As shown in Fig. 7, the larger Rician factor K makes the performance to be further improve, because the transmission channel is ominate by a stable LOS path. 5. As shown in Fig. 7(a), on the assumption of the same parameter values in the equation, the performance is not affecte by the FFT point size N an the number of a tappe-elay line of the elay profile moel in the simulation. The large N only means the wie banwith, an then the frequency interval /T between ajacent subcarriers is efine by T s an α, not N. 6. As shown in Fig. 7(b), the BER performance of DQPSK is more egrae than that of DBPSK because of the intra-symbol interference in (7). Now we evaluate the characteristics mentione above by using typical Rayleigh faing channels, that is, IMT- Vehicular channel A/B [6] an ETSI/BRAN channel A [7]. Tables an show the elay profile moels an the simulation parameters, respectively, an Fig. 8 shows the BER performances. To simplify the evaluation, the number of subcarriers is equal to the FFT point size N, an then the OFDM symbol rate /T s can be calculate from = f b/ T s N, (3) Table Typical elay profile moels. (a) IMT- (6-ray moel) Vehicular A Vehicular B Delay (µs) Power (B) Delay (µs) Power (B) (b) ETSI/BRAN Channel A (8-ray moel) Delay (µs) Power (B) Delay (µs) Power (B) where f b / inicates the DQPSK symbol rate. It is also foun from Fig. 8 that the approximate results agree with the simulation ones, which is base on the fact that the BER in faing channel is etermine by the value of rms elay

7 37 IEICE TRANS. FUNDAMENTALS, VOL.E88 A, NO. JANUARY 5 Table Simulation parameters. Channel Moel IMT- ETSI/BRAN Veh. A Veh. B Channel A Moulation DQPSK FFT point size N 64 Number of Subcarriers 64 GI Factor α /4 Bit Rate f b (bit/s) 384k 8M OFDM Symbol Rate /T s (symbol/s) 3k 4k Doppler Frequency f D (Hz) 75 ( GHz, 4 km/h) (5 GHz, km/h) RMS Delay Sprea σ τ (µs) availability of DMF can be easily certifie by the approximate equation. Finally, we shoul note that the valiity of the approximate equation is confirme by the evaluation. 5. Conclusion We propose the approximate equation for easily an promptly calculating the BER in DPSK/OFDM systems over both time an frequency selective Rician faing channels. It is confirme by the performance evaluation that the propose equation can precisely express the characteristics of the BER performance over faing channels. It shoul be mentione that the analysis in this paper is conucte on the assumption of the absence of frequency offset in a local oscillator. We woul now like to go on to evelop this analysis by taking account of the offset. References Fig. 8 Fig. 9 BER performances over typical Rayleigh faing channels. Comparison of BER performances for DMT an DMF. sprea σ τ, not by the shape of elay profile [8]. This figure shows DMF has the avantage of low BER in the case of high Doppler frequency like IMT- channel moel. Fig. 9 shows an example of the comparison of BER performances for DMT an DMF, which are calculate by the approximate equation. In the case of small σ τ /T s,e.g. σ τ /T s = 5 4, it is foun that DMF shows the better performance on the conition of large f D T s,e.g. f D T s > 3. Generally speaking, the symbol perio T s in OFDM systems is much longer than that in single carrier systems, then the [] S.B. Weinstein an P.M. Ebert, Data transmission by frequencyivision multiplexing using the iscrete Fourier transform, IEEE Trans. Commun., vol.com-9, no.5, pp , Oct. 97. [] J.A.C. Bingham, Multicarrier moulation for ata transmission: An iea whose time has come, IEEE Commun. Mag., vol.8, no.5, pp.5 4, May 99. [3] P.H. Moose, A technique for orthogonal frequency ivision multiplexing frequency offset correction, IEEE Trans. Commun., vol.4, no., pp.98 94, Oct [4] T. Pollet, M.V. Blael, an M. Moeneclaey, BER sensitivity of OFDM systems to carrier frequency offset an Wiener phase noise, IEEE Trans. Commun., vol.43, no./3/4, pp.9 93, Feb./March/April 995. [5] Y. Li an L.J. Cimini, Jr., Bouns on the interchannel interference of OFDM in time-varying impairments, IEEE Trans. Commun., vol.49, no.3, pp.4 44, March. [6] H. Schubert, A. Richter, an K. Iversen, Differential moulation for OFDM in frequency vs. time omain, Proc. ACTS Mobile Commun. Summit 97, pp , Oct [7] M. Lott, Comparison of frequency an time omain ifferential moulation in an OFDM system for wireless ATM, Proc. IEEE VTC 99, vol., pp , May 999. [8] M. Okaa, S. Hara, an N. Morinaga, Bit error rate performances of orthogonal multicarrier moulation raio transmission systems, IEICE Trans. Commun., vol.e76-b, no., pp.3 9, Feb [9] J. Lu, T.T. Tjhung, F. Aachi, an C.L. Huang, BER performance of OFDM-MDPSK system in frequency-selective Rician faing with iversity reception, IEEE Trans. Veh. Technol., vol.49, no.4, pp.6 5, July. [] W.C. Jakes, Microwave Mobile Communications, IEEE Press, New York, 974. [] F. Sasamori an F. Takahata, Theoretical an approximate erivation of bit error rate in DS-CDMA systems uner Rician faing environment, IEICE Trans. Funamentals, vol.e8-a, no., pp , Dec [] F. Sasamori, F. Maehara, S. Hana, S. Oshita, an F. Takahata, Approximate equation of average bit error rate in DS-CDMA systems over faing channels, IEEE Trans. Commun., vol.5, no.3, pp.49 45, March 3. [3] P.A. Bello, Characterization of ranomly time-variant linear channels, IEEE Trans. Commun. Syst., vol.cs-, no.4, pp , Dec [4] M. Patzol an A. Szczepanski, Methos for moeling of specifie an measure multipath power elay profiles, Proc. IEEE

8 SASAMORI et al.: A SIMPLE METHOD OF BER CALCULATION IN DPSK/OFDM SYSTEMS OVER FADING CHANNELS 373 VTC-Spring, vol.3, pp , May. [5] T.S. Rappaport, Wireless Communications: Principles an Practice, Prentice Hall, 996. [6] ITU-R, Guielines for evaluation of raio transmission technologies for IMT-, Recommenation ITU-R M.5, 997. [7] BRAN WG3 PHY Subgroup, Criteria for comparison, ETSI/ BRAN Document no.37f, 998. [8] Y. Karasawa, T. Kuroa, an H. Iwai, The equivalent transmissionpath moel A tool for analyzing error floor characteristics ue to intersymbol interference in Nakagami-Rice faing environments, IEEE Trans. Veh. Technol., vol.46, no., pp.94, Feb Fumihito Sasamori receive the B.E., M.E. an Dr. Eng. egrees from Wasea University, Tokyo in 994, 996 an, respectively. Since, he has been a Research Associate in the Department of Electrical an Electronic Engineering, Shinshu University. His current research interests inclue sprea spectrum moulation an igital mobile communication systems. He receive the IEICE Young Engineer Awar in. He is a member of IEEE. Shiro Hana receive the B.E. an M.E. egrees from Shinshu University in 978 an 98 respectively, an the Dr. Eng. egree from Kobe University in 988. From 98 to 988, he was a Research Associate at Kobe University. From 988 to 994, he was with Nagano National College of Technology. Since 994, he has been an Associate Professor in the Department of Electrical an Electronic Engineering, Shinshu University. In 996, he was at the University of California, Davis, as a visiting researcher. His research interests inclue satellite an mobile communications systems, moulation an coing, an ata compression. He is a member of IEEE an SITA. Eucation. Shinjiro Oshita receive the B.E. egree from Osaka Electro Communication University in 967, the M.E. an Dr. Eng. egrees from Osaka University, Japan in 97 an 973, respectively. Since 973, he has been with Shinshu University, Nagano, Japan, where he is currently a Professor in the Department of Electrical an Electronic Engineering. His research interests inclue intelligent CAI an communications theory. He is a member of IEEE an Japanese Society for Information an System in