The 17th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’06) SOFT HANDOVER IN ADAPTIVE MIMO-OFDM CELLULAR SYSTEM WITH COOPERATIVE PROCESSING Antti T¨ olli, Marian Codreanu and Markku Juntti Centre for Wireless Communications, University of Oulu P.O. Box 4500, 90014 University of Oulu, Finland ∗ ABSTRACT The joint cooperative processing of transmitted signal from several multiple-input multiple-output (MIMO) base station (BS) antenna heads is considered for users located within a soft handover (SHO) region. Downlink space-frequency bit and power allocation problem with different BS power con- straints is studied for the considered adaptive MIMO-OFDM system. The performance of the proposed heuristic loading method is shown to be close to the optimal convex optimiza- tion method with per BS power constraints. It is shown that the highest SHO gains are achieved with a small power imbalance between the received BS powers at low signal-to-noise ratio (SNR), where the achievable rates can be even doubled. On the other hand, the gain from joint processing in SHO quickly diminishes as the imbalance increases, especially at low SNR. Moreover, the results indicate that the joint processing can be even detrimental for the system performance if a coarse phase synchronization between BS antenna head is not guaranteed, as some additional fading on the target SNR values is introduced. I. I NTRODUCTION Recently, there has been some research devoted to optimiz- ing the mutual information of a multiple-input multiple-output (MIMO) system with cooperative processing between base sta- tions (BS). For example, [1–4] studied the downlink sum rate and spectral efficiency optimization for MIMO cellular systems with perfect data cooperation between base stations. All the- oretical studies above assumed ideally that the inter-cell inter- ference seen at the receiver, the channel and transmitted covari- ance matrices from all the base stations were perfectly known at and shared between the transmitter(s). However, the required signalling feedback and/or practical channel estimation limita- tions would make the ideal approach unpractical in most appli- cations. In a time division duplex (TDD) orthogonal frequency divi- sion multiplexing (OFDM) system with adaptive MIMO trans- mission, the modulation parameters in uplink (UL) and down- link (DL) can be adapted according to the channel conditions. The reciprocal DL channel can be estimated accurately during the previous UL frame assuming that the frame length is shorter than the channel coherence time. In order to to attain the chan- nel information between all users and BS antennas in the cel- lular network the channels should be jointly estimated at each BS antenna head. In practical adaptive MIMO–OFDM cellular * This research was supported by Finnish Funding Agency for Technol- ogy and Innovation (TEKES), Nokia, the Finnish Defence Forces, Elektrobit, Tauno T¨ onning Foundation, Nokia Foundation and Infotech Oulu Graduate School. systems, however, UL transmissions from adjacent cells can be much more attenuated compared to the own cell users, thereby, the joint channel estimation may be difficult if not impossible to implement in practice. We consider somewhat more practical case where the joint cooperative processing of the transmitted signal from several MIMO BS antenna heads is restricted to an area where the users have comparable signal strengths from adjacent BS an- tenna heads. Moreover, we assume that the cooperative signal processing can be performed in a centralized manner so that the MIMO antenna heads are distributed over a larger geographi- cal area (e.g. hundreds of meters). The distributed antenna heads are then connected to the central processing unit (con- troller) via radio over fibre technology or wireless microwave links, for example. Similarly to the soft handover (SHO) fea- ture in (W)CDMA systems [5], SHO region is defined for users with similar received power levels from adjacent distributed BS antenna heads (±3 dB SHO window, for example). Since the signal processing of the BS antenna heads is concentrated at one central controller, joint beamforming from all the antennas belonging to the ”active set” can be performed to the user(s) within the SHO region. The transmissions for other users out- side the SHO region are seen as interference. In the following, the distributed BS antenna heads are denoted as base stations, for simplicity. In this paper, we concentrate on DL space-frequency bit and power allocation problem with different BS power constraints for the considered adaptive MIMO-OFDM system. Optimal power allocation with per BS power constraints using convex optimization tools is derived and compared to a more practical but sub-optimal heuristic power allocation. The performance of cooperative processing is studied both in terms of the mutual information and more practical achievable spectral efficiency per user. In order to perform joint transmission from all the BSs belonging to the active set, the baseband signals need to have a common carrier phase reference and the impact of the propagation delay from each of the transmitters to the intended user must be fully compensated for at the transmitter by using some feedback from terminals, for example. The impact of im- perfect synchronization between the BS antenna heads on the achievable gains is also addressed with some practical exam- ples. The emphasis is put on providing methods for throughput maximization of the users in a single SHO active set while the system level gains and trade-offs from cooperative SHO pro- cessing will be investigated further in [6]. The impact of the size of the SHO region, overhead from increased hardware and physical (time, frequency) resource utilization, different non- reciprocal inter-cell interference distributions due to SHO must be evaluated by system level simulations. 1-4244-0330-8/06/$20.00 c 2006 IEEE