An Inter-Cell Interference Power Level Feedback Technique for One-Cell Reuse OFDM/TDMA using Subcarrier Adaptive Modulation Scheme

Transcription

1 An Inter-Cell Interference Power Level Feedback Technique for One-Cell Reuse OFDM/TDMA using Subcarrier Adaptive Modulation Scheme Kazunari YOKOMAKURA, Seiichi SAMPEI Graduate School of Engineering, Osaka University 2-, Yamadaoka, Suita, Osaka -, JAPAN Norihiko MORINAGA Department of Information Technology, Hiroshima International University Hirokosingai --, Kure, Hiroshima -2, JAPAN Abstract This paper proposes an inter-cell interference (ICI) power level feedback technique to cope with dynamically varying ICI thereby achieving stable operation of a multi-level transmit power control (MTPC) in one-cell reuse time division multiple access (TDMA) using orthogonal frequency division multiplexing (OFDM) based adaptive modulation scheme (AMS). In this system, ICI level of each subcarrier is estimated by calculating a vector error of received symbol. As for the notification of ICI, we employ an analog quadrature amplitude modulation (QAM) and code division multiplexing (CDM) to achieve high frame efficiency and high accuracy of notified information, in which we divide all the subcarrier into some blocks, and notify only the average value for each block, where only the values higher than a threshold are fed back. Computer Simulation confirms that the proposed scheme can estimate and notify ICI power level with high accuracy. Keywords OFDM based adaptive modulation scheme (OFDM AMS), multi-level transmit power control (MTPC), inter-cell interference (ICI), one-cell reuse TDMA system, vector error, code division multiplexing (CDM) I. INTRODUCTION As the wide spread of broadband access via the Internet, we can use applications via the Internet with wireless systems such as wireless local area network (WLAN) and mobile phone system, and thereby broadband services that support more than Mbit/s is strong requirement for beyond third generation (BG) wireless communication systems []. As a candidate for BG wireless access technology, the systems such as a variable spreading factor-orthogonal frequency and code division multiplexing (VSF- OFCDM) [2] and a multiple-input multiple-output with orthogonal frequency division multiplexing (MIMO-OFDM) have been studied. We have proposed a one-cell reuse time division multiple access (TDMA) system that employs an orthogonal frequency division multiplexing (OFDM) based adaptive modulation scheme (AMS) (OFDM AMS)[], [] with a multi-level transmit power control (MTPC) [] (OFDM AMS/MTPC) in which modulation parameter and transmit power for each subcarrier are controlled according to the received signal to interference plus noise power ratio (SINR) for each subcarrier under inter-cell interference (ICI) existing cellular conditions []. The most important issue to realize this system is how to estimate ICI level of each subcarrier and feed back the information to the transmitter side. Therefore, this paper proposes a method of estimating interference plus noise power level for each subcarrier by measuring a vector error of each received symbol, and notifying it by using an analog quadrature amplitude modultion (QAM) and a code division multiplexing (CDM) schemes without so much overhead in each TDMA slot. II. PROPOSED SCHEMES Fig. shows the concept of OFDM AMS/MTPC based TDMA system. In this system, modulation parameter and its transmit power for each subcarrier are controlled according to the SINR of each subcarrier, that is, when the SINR is high, the modulation Fig.. Desired Signal Undesired Signal Received SINR BS MS High Rate Low Rate f Independent Channel BS2 Received SINR MS2 the concept of OFDM AMS/MTPC based TDMA system parameter for high user rate is assigned to the subcarrier, but when the SINR is low, a modulation parameters for low user rate is assigned. In single cell system in which no ICI exists, this case has already developed in [], in which delay profile information (DPI) and noise level are embedded in the preamble of TDMA slot to feed back the signal to noise power ratio (SNR). On the other hand, in cellular system in which ICI exists, we have to feed back interference plus noise power of each subcarrier which is varying dynamically in frequency domain by not only frequency selective fading channel but also the MTPC. A. estimation of signal power for each subcarrier This system has to estimate the received power for each subcarrier in a soft decision decoding or interference plus noise level estimating due to the different transmit power for each subcarrier by the MTPC which employs the subcarrier transmit power control (STPC) []. Fig. 2 shows the concept of a method of estimating received signal power for each subcarrier. This concept is that the differential between the estimated received SINR and the target SINR for each subcarrier is the implement of signal power for each subcarrier. Concretely, the received power is estimated as follows, step. In first frame after link establishment, we employ carrier hole as modulation parameters of all the subcarrier. The received power for each subcarrier is equal to interference plus noise power level. An interference plus noise power for k th subcarrier is P I+N (k) = 2 R (k) 2 () where R (k) is the received symbol of k th subcarrier. step.2 f

2 Rx SINR Estimated SINR Target SINR Quadrature Channel Reference Symbol PI+N (n,k) d 2 d Received Symbol k Subcarrier Increment of signal power S(k) Estimated signal power S(k) = S (k) S ( k) S : Average signal power of subcarrier Fig.. Reference Symbol Reference Symbol Candidates Received Symbol In-phase Channel A method of estimating interference plus noise power level Fig. 2. the concept of estimating signal power of each subcarrier We estimate an average received power for each subcarrier. The average received power of k th subcarrier is S = C Rx N s (2) where S is the average power, C R x is received power estimated by the pilot signal, and N s is the number of subcarriers in an OFDM symbol. step. We estimate the received power of each subcarrier on the assumption that we don t employ the STPC. The received power for k th subcarrier is S (k) = A H(k) 2 S () where S (k) is the received power for each subcarrier, H(k) is a normalized channel response for k th subchannel. A is a constant which satisfies following equation : N s C Rx = A S H(k) 2. () k= step. We estimate the SINR of each subcarrier without the STPC. The SINR for k th subcarrier is S INR(k) = S (k) P I+N (k) where P I+N (k) is the interference plus noise power level estimated in a previous frame. step. However, The power allocated each subcarrier is controlled by the MTPC to satisfy the target SINR according to the modulation parameter. The difference S (k) between the target SINR and the received SINR is S (k) = S INRML(k) th S INR(k) where S (k) is the difference, and ML(k) is modulation parameter which represents the number of bits transmittable at a symbol we can know from modulation level information (MLI) subframe. step. We calculate the received power for each subcarrier. Because S (k) is the signal power implement for each subcarrier, the estimated the received power for each subcarrier is () () Ŝ (k) = S (k) S (k). () where Ŝ (k) is the received power for each subcarrier. In after frames, we repeat after step.2, and use it in a soft decision decoding or estimation of interference plus noise power level. B. estimation of interference plus noise power level Fig. shows a method of estimating interference plus noise power level of each subcarrier. Proposed scheme estimates interference plus noise power of each subcarrier by a vector error which is the distance between the received symbol and the reference symbol. Even though the reference symbol is able to be regenerated from the hard decision value of the received symbol, this system lowers the required SINR by a channel coding, so the reference symbol is regenerated from bits decoded by a soft decision decoder to regenerate reliable it in proposed scheme. The vector error for k th subcarrier in n th OFDM symbol is e(n, k) = R(n, k) S Re f (n, k) 2 = R(n, k) r n,k e jθ n,k S Tx (n, k) 2 () where R(n, k) is the received symbol, S Re f (n, k) is the reference symbol, S Tx (n, k) is the transmitted symbol regenerated from decoded bits, and r n,k e jθ n,k is the channel state (r n,k : envelope, θ n,k : phase). Additionally, e(n, k) is averaged within a frame to mitigate the affect of the reference symbol error. The average vector error e(k) is e(k) = N f rame N f rame n= e(n, k) () where N f rame is the number of OFDM symbols in a frame. In this case, the received SINR for each subcarrier is On the other hand, S INR(k) is S INR(k) = S Re f (k) 2. () e(k) S INR(k) = Ŝ (k) P I+N (k) () where Ŝ (k) is the signal power estimated by the equation () and P I+N (k) is interference plus noise power level. The equation () is equal to the equation (), so P I+N (k) is represented in this equation (2). P I+N (k) = e(k) Ŝ (k) S Re f (k) 2 (2) Moreover, because notifying all the interference plus noise power levels can cause not only low frame efficiency but also low accuracy of notified information, we block all the subcarriers every M (hereinafter called block size) adjacent subcarriers and obtain the average value for the interference plus noise power level as its representative value for the block to reduce amount of information to be notified. The interference plus noise power level of m th block is P I+N (m) = M mm n=(m )M+ P I+N (k). () where P I+N (m) is the average value within m th block.

3 Fig.. Power S Ref S LLR #22 # Reference Interference Level R dbm Lower Limit Reference T dbm Spreading Code Guard # Power PN Code ( + ) Walsh Code () µsec µsec ( chip) µsec ( chip) ( chip) µsec # # t CDM Guard Interference Power # Interference Power # Interference Power # Walsh Code Assignment Reference LLR Interference Power # Interference Power #2 Walsh Code # Walsh Code #2 Walsh Code # Walsh Code # Interference Power #NB Walsh Code #(N B+2) A method of notifying interference plus noise power level QAM Power SRef SLLR Fig.. T R PI+N(m) (dbm) A setting of an analog QAM power C. notification of interference plus noise power level Notifying the interference plus noise power level with the digital modulation can cause the huge amount of information to be notified. Therefore, we employ an analog QAM and CDM sheme in notifying interference plus noise power level to achieve high frame efficiency and high accuracy of notified information. Fig. shows a method of notifying interference plus noise power level. In Fig., we employ (N B + 2) informations, which includes N B (= N S /M) interference plus noise power levels, a reference signal to know absolute values of interference level, and a lower limit for reference (LLR) signal to determine if the interference level information is notified, to the CDM scheme. After that, we employ a direct sequence/spread spectrum (DS/SS) scheme to grow resistans to delayed waves or noise. Fig. shows a setting of an analog QAM power, where the abscissa axis represents the actual interference plus noise level, and the ordinate axis represents the allocated power. R dbm is a reference to know the absolute value. Moreover, we employ a lower limit for reference (LLR) represented T dbm and the power is not allocated to the level information less than T dbm and that much power is allocated to other information to be notified because low level interference level has low affect to the performances. Consequently, the accuracy of whole interference level information improves. In Fig., the analog QAM power of m th block is S Re f S LLR (P I+N (m) T ) + S LLR (P I+N (m) > T ) S I+N (m) = R T () (P I+N (m) T ) where S I+N (m), S Re f and S LLR are the power allocated the interference information signal, the reference signal and the LLR signal. Each signal generated like this is applied CDM with the orthogonal Walsh sequence of chips. The lower right in Fig. shows code t Tab.. Simulation parameters Symbol Rate ksymbol/sec Modulation QPSK, QAM, QAM FEC Convolutional Coding (r = /, K = ) Num. of Subcarriers Target C/N 2 db Hz f D Hz Path Model Exponential Decaying Spike Rayleigh Shadowing Log normal distribution Standard Deviation σ = db Path Loss Model ITU-R M.22 outdoor to indoor and pedestrian test environment[] Cell Radius m Max. Tx Power dbm Num. of ICI station allocation to each information. In order to grow resistance to the delayed waves or noise, we apply a DS/SS to this signal with the spreading code added one chip to the pseudo noise (PN) sequence of chips period. These spreading codes employ guard chips backward and forward, which are selected to be cyclic shift []. This signal generated like this is transmitted to the base station (BS) side after a transmit power control to operate the same received power as data subframe. In the BS, after the delayed signals are separeted and combined by a rake receiver, and the interference level information of each block is retrieved by multiplying the Walsh code and correlating. The retrieved interference plus noise power level of m th block is R T (Ŝ I+N(m) Ŝ LLR) + T Ŝ ˆP I+N (m) = Re f Ŝ LLR T (Ŝ I+N(m) > Ŝ LLR) (Ŝ I+N(m) Ŝ LLR) () where Ŝ I+N (m), Ŝ Re f and Ŝ LLR is the received QAM power of m th block, the reference signal and the LLR signal. However, there are some parameters which change the power allocation for each information when the values of these parameters change in Fig.. These parameters are discribed as follows :. Reference : R dbm This parameter is the value to determine the absolute value of notified interference plus noise power level. When this becomes larger, the detection sensitivity of interference level detioriorates because the slope of line made by T and R is getting smaller. On the other hand, when this becomes smaller, the detection sensitivity of the reference detioriorates because the power allocated the reference signal is getting smaller. 2. LLR : T dbm This parameter is a threshold to determine if the interference plus noise power level is notified. When this becomes larger, the accuracy of notified information detioriorates because many interference plus noise power levels are less than T dbm, that is, amount of information is so little. On the other hand, when this becomes smaller, the power allocated to each information is getting low because there is the larger amount of informations to be notified,. Reference to LLR power ratio : S Re f /S LLR db When this value becomes larger, the accuracy of notified information deteoriorates because the slope of the line made T and R is small as the large reference. On the other hand, when this value becomes smaller, the power allocated the LLR signal is getting smaller, so the detection secsitivity of reference detioriorates. In this paper, we evaluate the performance of the proposed scheme after optimizing these parameters.

4 CE DPI ILI MLI µsec Normalized Interference Power [db] µsec µsec µsec.µsec Data Subframe (OFDM) µsec (2 OFDM Symbol) OFDM Symbol.µsec CE : common pilot for Channel Estimation DPI : Delay Profile Information ILI : Interference Level Information MLI : Modulation Level Information Fig.. GI TDMA slot format OFDM Symbol = µsec r.m.s. Delay Spread = 2ns Carrier Hole Subcarrier Fig.. Variance of interference level in frequency domain M= M=2 M= M= M= M=2 M= Perfect. r.m.s. Delay Spread = 2ns. 2 Fig.. A. Simulation Model Cumulative distribution of data throughput III. COMPUTER SIMULATION Tab. shows simulation parameters and Fig. shows TDMA slot format for the proposed scheme. Each slot includes common pilot for channel estimetion (CE) subframe to measure delay profile of the downlink channel, DPI subframe to notify the delay profile information for the uplink, interference level information (ILI) subframe to notify the interference plus noise level for the uplink, MLI subframe to notify the information of modulation parameter to the receiver, and data subframe. The porpose of this simulation is high accuracy of ILI subframe, channel state feedback and MLI notification are operated perfectly. B. Simulation Result ) estimation of interference plus noise power level: Fig. shows variance of interference level in frequency domain. As we can see from this figure, interference level varies dynamically in frequency domain, that is, when M (block size) becomes larger, it... M= M= M=. M=2 M= 2 Fig.. r.m.s Delay Spread [ns] M= M=2 delay spread v.s. % value of throughput is getting more difficult to block the interference plus noise power level due to lower channel correlation between subcarriers located at both end of block. Fig. shows cumulative distribution () of throughput when interference plus noise power level is notified perfectly, where Perfect is the case of known interference plus noise power level, and the abscissa axis represents throughput in data subframe and the ordinate axis represents the probability to be less than the value of throughput on the abscissa axis. Firstly, we can confirm that interference plus noise power level of each subcarrier is notified with high accuracy because M = is approximately equal to Perfect. Next, when M becomes larger, we can confirm the performance deteriorates. This is because the channel correlation between subcarriers within a block is getting low as M increases. Fig. shows the performance delay spread versus % of the throughput, where delay spread is changed by varying decaying level between adjacent arrival paths and maximal delay spread is the case of the equivalent power spike Rayleigh model. When delay spread becomes larger, the degradation is getting more serious. This is because channel correlation between adjacent subcarriers are getting low as delay spread becomes larger. When we design a one-cell reuse TDMA system using subcarrier adaptive modulation based OFDM scheme, we have to set the M considering trade-off between such throughput performance degradation and increase of frame efficiency due to less number of the amount of ILI to be fed back. In this paper, we will employ M = because it can guarantee at least Mbit/s of the throughput when delay spread is nearly ns. 2) notification of interference plus noise power level: Firstly, we optimize the parameters T, R, S Re f /S LLR. Fig. shows of the throughput and Fig. shows the throughput of and % when T changes. When T is set to dbm, % is good performance but % detioriorates so much in the comparison with T = dbm. On the other hand, when T is set to dbm % detioriorates somewhat in the comparison with T = dbm but % acheives the best performance. Therefore, we employ T opt = dbm as the optimum value of T. Similarly, Fig. 2 shows the throughput on and % when R changes and Fig. shows the throughput for and % when S Re f /S LLR changes. In these figure, we employ R opt = dbm and (S Re f /S LLR ) opt = 2 db as the optimum values of R and S Re f /S LLR. ) notification of interference plus noise power level: Fig. shows cumulative distribution of the throughput for the proposed scheme when we employ these optimum parameters. As shown in this figure, the performances in the high throughput region, where noise is diminant, detiroriorates due to differential between the value of T and the actual noise level, but we can suppress

5 Fig T=- dbm T=- dbm T=- dbm T=- dbm T=-2 dbm T=-2 dbm Fig.. of the throughput when T changes % value 2 % value Lower Limit for Reference T [dbm] The throughput of and % values when T changes the degradation quantity within %. Concequently, though the performance in low throughput region detioriorates interference plus noise level in high throughput region is notified with high accuracy. With these results, we can confirm that the proposed ILI notification scheme works well in the OFDM AMS / MTPC based TDMA systems. IV. CONCLUSIONS This paper proposed an ICI power level feedback technique for one-cell reuse TDMA systems using OFDM AMS/MTPC. As the result of the estimation, we confirmed that proposed scheme can estimate interference plus noise power level of each subcarrier with high accuracy by measuring vector error of the received symbol, and only the average value of it in adjacent subcarriers is available in transmission control. As the result of the notification, we confirmed that proposed scheme can notify interference plus noise power level with high frame efficency and high accuracy by applying an analog QAM and CDM scheme to it and optimizing the power allocation in ILI subframe. REFERENCES [] B.G.Evans and K.Baugham, Visions of G, IEE Electronics & Communication Engineering Journal, vol. 2, no., pp. 2-, Dec. 2. [2] H.Atarashi, S.Abeta, M.Sawahashi, Variable Spreading Facter - Orthogonal Frequency and Code Division Multiplexing, IEICE Trans. Commun., vol. EB-, no., Jan. 2. [] T.Keller and L.Hanzo, AdaptiveMulticarrier Modulation: A ConvenientFramework for Time-FrequencyProcessing in Wireless Communications, Proc. IEEE, vol., pp. -, May 2. [] M.R.Souryal and R.L.Pickholtz, Adaptive Modulation with Imperfect Channel Information in OFDM, IEEE ICC, vol., pp. -, June 2. [] T.Yoshiki, S.Sampei and M.Morinaga, High bit rate transmission scheme with multilevel transmit power control for OFDM based adaptive modulation scheme, IEEE VTC, vol., pp. 2-, May 2. Fig.. levels Fig.. % value 2 % value Reference R [dbm] Fig. 2. of throughput when T changes % value 2 % value 2 2. S Ref /S LLR [db] on % and % when T changes notification perfect of the throughput in notifying interference plus noise [] N.Nakanishi, S.Sampei and M.Morinaga, Performances of One Cell Reuse TDMA Systems Employing OFDM Based Adaptive Modulation Scheme and Multilevel Transmit Power Control, WPMC, vol., pp. -2, Oct. 2. [] N.Maeda, S.Sampei, N.Morinaga, Performance of the Delay Profile Information Channel based Subcarrier Transmit Power Control Technique for OFDM/FDD Systems, IEICE Trans., Commun., vol. J-B, no. 2, Feb. 2. [] H.Harada, Future Prospect and Reserch Theme for Next Generation Multimedia CDM Wireless Communications Based on R&D Results in Communications Reserch Laboratory, Technical Report of IEICE, SST-, Dec.. [] Guidance for Evaluation of Radio Transmission Technologies for IMT-2, Rec.ITU-R M.22,.