ABSTRACT DOWNLINK TRANSMIT POWER ISSUES IN A WCDMA CELLULAR SYSTEM CHARLES NOBLET, MANOS FANDRIDIS, RAY OWEN MOTOROLA UK

Transcription

1 244 DOWNLINK TRANSMIT POWER ISSUES IN A WCDMA CELLULAR SYSTEM CHARLES NOBLET, MANOS FANDRIDIS, RAY OWEN MOTOROLA UK ABSTRACT After many years of standardisation Third generation cellular systems have now reached the deployment phase. Prior to this deployment, a dimensioning and planning stage is necessary to assess the radio access network coverage and capacity. A critical downlink issue of radio capacity and coverage planning is the power allocation required per user to fulfil network GoS requirements. Indeed, each Node B within a WCDMA system has a finite power resource that has to be shared amongst a number of users, or more accurately links elements (including users in soft handover). The consequent capacity of a 3G cellular system is dependant on the adequate resource sharing of the Node B transmit power. Three fundamental parameters describe this power: 1. The total transmit power 2. The pilot power 3. The allocated transmit power per dedicated traffic channel (DCH). This paper discusses the impact of these tuneable parameters on capacity, coverage, QoS and consequently GoS in a WCDMA network. INTRODUCTION For WCDMA cellular networks the system capacity is fundamentally limited by multiple access interference. This is created by the variable BS and MS transmit powers and non-orthogonal channel characteristics. The required peak and average MS transmit powers and BS transmit powers are directly linked to coverage and capacity of a cell and consequently the network. This is shown through the derivation of the uplink and downlink load factor as demonstrated in Burley (1,2). Given that the total base station power can limit the capacity of any cell and consequently network the approximated total base station power can be expressed as a function of the average system load. It is commonly found that CDMA networks tend to be coverage limited on the upldc, capacity limitation occurs on the downlink as explained in Holma and Toskala (3). In any given cell, the downlink available power can be shared between dedicated traffic channel (DCH), downlink shared channel (DSCH) and common channel, such as the pilot channel (CPICH). It is common to allocate approximately 10%-15% of the total available PA power to the common pilot channels as explained in Love et a1 (4). An example is shown in Figure 1. Out of a total available average power of 20W per sector, 2W (33dBm) is assigned to the pilot. After establishmg the fixed pilot power each established link is allocated a maximum power allocation for the remaining data channel. When all pre-allocated DCH power resource have been utilised, different strategies can be employed to free extra power, enabling additional radio link connections. As an example one such strategy is to reduce the maximum allocated power assigned to the traffic channels. Altematively, both DCH and pilot powers can be reduced equally. The primary objective of this paper is to concentrate on optimising downlink performance using the following parameters. On the uplink the network reliability can be based on criteria such as noise rise level above thermal background and connection outages. In the downlink GoS can be based on the power allocated per cell and cell-to-mobile frame erasure rate. UMTS poses some fundamental differences between uplink and downlink operation. From a downlink perspective, the transmission is point-to multi-point (cell-to-manymobiles). Consequently each mobile observes a different downlink interference level over and above thermal noise. In addition to this, the level of out-of-cell interference differs from each mobile unit due to position. This is commonly expressed as an F-factor, which provides a measure of the out-of-cell to in-cell interference level. This paper addresses the average downlink PA power split required to operate a network at the most affective GoS. The paper is organised as follows: the first section presents a theoretical approach to the allocated Node B transmit power requirements. The second section analyses the impact of the pilot power on capacity and coverage, showing cell-size reduction effects under heavy load. Finally the last section address the issues related to the maximum allowable power per DCH lmk to give the optimum network GoS, maximum range and outage are analysed on a per service basis. 3G Mobile Communication Technologies, 8-10 May 2002, Conference Publication No IEE 2002

2 245., Pilot ' 2w Traffic Power taken by users w '?,\ \ Dedicated Traffic Chznnel law. AvailablePower Figure 1 - Node B Power Distribution per sector NODE B TRANSMIT POWER ANALYSIS A major hardware constraint in a wireless network is the amount of air-interface resource available. For WCDMA systems this 'resource' can be measured in the uplink noise rise and downlmk total transmit power. Whilst a Node B has a finite power to be shared amongst users, optimum operation occurs when each Node B allocates only sufficient power (DCH link) to maintain the QoS requirements of each mobile. In order to formulate the total Node B transmit power, we first consider the received energy per bit to total interference density at the ith mobile owing the link from the kth Node B; that is E,, I Noi = pgi ' PDCK i "i K (l-a).plli +cpil,! +Nth k=l (1) Note that Eq.(2) is a simplified form, based on the following assumption: (a) Fixed pilot power (b) Uniform user and traffic distribution (c) Hexagonal cell deployment of identical cells. (d) Hata pathloss model Eq. (2) is the mean received E&, balance equation, where PGi is the link processing gain, PDCHi is the power allocated to the link, li is the link pathloss, a is the orthogonality factor, K is the number of interfering Node Bs, lik is the pathloss, and Nth is the thermal noise. Eq. (2) establishes the average Node B power as the addition of the pilot power plus the sum of powers allocated to each link multiplied by their activity factor Vi. By using (1) and (2), the power per DCH link PDCHi can be derived. Eq. (3) is comprehensively derived in Burley (1,2). Assuming that all users have the same service (i.e. bit rate and QoS requirement), and that the average Node B transmit power at any instant is known as P, Eq. (3) can be simplified. f represents the average f-factor for all mobiles. Although a very simplistic assumption thls value can be derived using numerical methods. It is mainly dependent on the Node B antenna type, spatial traffic distribution and the propagation model considered. pdch = --(P.(l- E, 1 No, a + fi) + N,.li)(4) PG, PILOT POWER In a WCDMA system the pilot channel (CPICH) signal strength can be directly related to system performance. Its role is three-fold: it defines the cell boundaries and provides each mobile with a 'lock' signal to determine the ownership cell. In addition the signal strength comparisons between base stations can be used to determine when to go into soft handover between two cells. Finally it may also be used for initial system acquisition and to aid the channel estimation for the dedicated channels. To keep the initial cell boundaries fvred the Node B transmits the pilot at a fixed level. From a mobile perspective, the pilot is perceived as the ratio between the received energy per chip to total interference or EJI,. In order to keep the mobile referenced to a cell, this received ratio must exceed a minimum threshold at the mobile at all times. This threshold is often referred to as fmger locking, for which a typical value is between -16 and -2OdB for UMTS. The strongest received signals above the threshold are the best served cells and are placed in the Active Set. Any other pilots received above the threshold are placed in the candidate set. If the current QoS target is not met satisfactorily, upon certain conditions, the next strongest candidate can be admitted in the active set to initiate a 2-way soft handover, or multi-way handover as described in 3GPP (5). The pilot power, the cell coverage and capacity share an intimate relation, which is discussed in this section. The allocated Node B transmit power is shared between pilot and traffic channel, as shown on Figure 1. In WCDMA, the downlink traffic power transmitted to other users is viewed as a partial noise, or interference moderated by the downlink channel orthogonality described by a. Consequently, a significant increase of the composite and individual link downlink traffic powers will start to reduce the pilot to traffic power ratio. Figure 3 depicts the ratio between pilot and traffic channel power in a 37-cells configuration. The simulations, run for a mix of services (60% 12.2 Kbps voice, 25% LCD 64Kbps, 10% LCD 144Kbps and 5%

3 246 LCD 384Kbps}, have been performed on a static simulation. All services have a QoS target of FER 1%. The simulation environment was set to represent dense urban, assuming a distance of 600 meters site-to-site, and an average 20dB in-building penetration loss. The pilot power was set to 2W, while the maximum allowed power on the downlink dedicated traffic channel was set to 2W. As discussed in section IV, the setting of the maximum DCH power value can have a great impact on coverage and QoS through either pilot detection or traffic channel. All the statistics collected from the simulations are based on the seven inner sites. A description of the simulation modelling approach can be found in Owen (6). As shown in Figure 2, (at low load), the pilot power is predominant compared to the traffic channel power, which could lead to pilot pollution as demonstrated in Love et a1 (4). As depicted on the graph, there are some benefits using the downlink transmit diversity in this case. It provides an average transmit power reduction from 6 to 3 Watts at around 50% cell load (7-10 users) or alternatively an increase in the capacity for a fixed average cell power allocation. Indeed, the downlink transmit diversity decreases the EO, requirement at the mobile receiver; hence decreasing the required transmit power. Nlocstlon d1ot.l PAPW~ I sector demonstrates the close relation between coverage and through the pilot signal strength in a WCDMA system. However a weak pilot could yield very high rate users to surpass their maximum llnk power limit just to compensate for the increased pilot interference. The next section will discuss in detail the setting of the maximum limit of the downlink DCH power. Increasing the cell load has the effect of reducing the area over which the pilot is strong enough to demodulate. Figure 4 provides another illustration of the cell breathing effect. The left picture shows a system with a 1dB noise rise load, whereas the right picture is a 5dB noise rise system. At the fixed 2Watts pilot power, the cell reduction is drastic, reflecting the statistics presented on Figure 3. In some cases the pilot power should be reduced. For instance if the level of the CPICH is too high, the coverage is extended and more mobile users are invited, fust to include the cell in their candidate set, and to establish a link if necessary. However, this will increase the number of users in Soft Handover (SHO) and will lead to a load increase in the Node B. Hence, the number of links issued from the same cell will increase whilst keeping the same number of physical mobile users. Indirectly, this might decrease the system, capacity. However, in downlink soft handover the E@l, ratio can be significantly reduced, countering the effect of increasing the number of connections. 15,N' 50% cell load zone P"' Ec lo M. Distance Node B - UE Ustance (m) 50 1W 150 ax, 250 3W Nuder of Usen per cell Figure 2 - Pilot/DCH Resource Sharing in a cell As the load increases, the pilot signal strength at the mobile receiver reduces (EJI,) this will in turn decrease the system downlink capacity and range. The range decreases with the load using up the available power at the BTS and the exponential increase in the power required maintaining each links QoS. In order to illustrate this effect, Figure 3 shows the EJI, as a function of the distance between the MS and BS for different uplink loads assuming symmetrical services, as described above. It can be observed that increasing the cell capacity has an effect on the pilot coverage especially at large cell distances. As the cell range increases the level of adjacent cell interference also increases (as all cells are operating at the same frequency). For instance, 13 users load and a distance of 250 meters, the received EJI, has a value of dB against -6.5 db for a 5 user load This effect I t 60% volnl LCD 64KI lo)( LCD 1UKI S% LCD 3&(K Figure 3 - CapacityICoverage trade-off Figure 4 - Coverage - Pilot E&, (1 de3 / 5dB) load

4 247 '. DOWNLINK DCH POWER In WCDMA, each application has a specific quality of service (QoS) to fulfil. At the physical layer, the QoS target is set by the outer loop power control at the Radio Network Controller (RNC). The QoS target is typically represented by the energy-to-noise ratio (ED,) at both ends, i.e. receiver and transmitter. After demodulation, the received signal should satisfy this target in order to maintain a desired BER or FER. The En, target used to achieve the required QoS is function of several parameters, such as bit rate, orthogonality/diversity, interference level, FER target and mobile speed. WCDMA fast power control ensures that the transmitted signal is at a level that achieves the QoS target. Although this has the advantage of ensuring QoS is met it can also mean that a user in a very deep fade will introduce a disproportionate amount of interference to all other users. In effect it would be worth (in cell capacities terms) to sacrifice the QoS of one user to improve the QoS of many users. As such a maximum threshold (MAX-DCH) to the downlink DCH power is set to avoid such situations. The choice of this value has an impact on the performance of different services and service mixes in any one cell. The MAX-DCH transmit power parameter must be judiciously set at the base station. Indeed, If the maximum DCH power is set too high the interference to other users in the cell will be high, resulting in the required Ea, target not being obtained or pilot coverage failure. If the maximum DCH power is set too low the mobiles (especially the higher rate-sets mobiles that require larger amounts of transmit power to achieve QoS targets) will not be able to achieve the required EJN, and consequently QoS targets. The next section will investigate the impact of the DCH max limit power setting from a coverage and outage perspective, considering different bearer rates. Range-Limited System The QoS of a particular service depends on whether or not the post rake receiver received is equal or greater then the ED, target. In the fist section, the power balance equation Eq.( 1,2) shows that this ratio is function of the transmit power on the DCH, as well as the path loss. Since the mean path loss is directly related to the distance between the base station and the mobile, at a fixed EJN, target, the required DCH Power can be derived as a function of the mobile-to-bts distance. Results presented in Figure 5 have been generated using a combined upllnwdownlink llnk budget and analytxal model for load estimation as derived in Burley (1,2). Figure 5 depicts the required max power on a DCH for each type of services, as a function of the distance between mobile and base station, for two different environments: dense urban and sub-urban. The large difference between dense urban and sub urban environment translate the significance of the DCH maximum limit relative to the range. In Dense Urban conditions (20 db penetration loss, 10 db for suburban), for a radius of 400 meters, the minimum max setting of the DCH power requirement for voice is 0.16W to maintain 95% connection success rate. Equivalently the 384 kbps service requires 1.6W or ten times the average power required for voice. Figure 5 illustrate the high requirements on downlink transmit power in dense urban environment E 5 z5 e 2 s Requirsl DCHPomnrs.NBxRange DmreUlDa lsld Urban 0 om om rm *ELI ) 3m 3.61 R.ng. Irml Figure 5 - DCH Power Limit vs. Range Outage Limitation Based on the simulation configuration described in the pilot power study section, a similar approach has been adopted to present the performance of a cell in terms of bad connections, or so-called outage. The network performances have been evaluated considering single service as shown in Figure 6, and a mix of services (Figure 7). The logic behind the simulation exercise is to reach the maximum system load for 100% good connection. Then for the same load, the MAX-DCH power limit is changed and outage re-assessed. It can be anticipated that a degradation of the QoS on the downlink will be noticeable when decreasing the DCH power. Both Figure 6 and 7 show the outage versus the DCH max limit allocated on a per link basis, not the average transmit power. As it can be seen in Figure 6, the higher bit rate services and consequently low processing gain llnks require proportionally more DCH power than the lower rate counterparts. Proportionally high rate connections actually require a lower EbNo ratio than lower rate connections due to the coding type used and the interleaving depth - this offsets this phenomenon to a small extent but they still require more power since the difference in rate is a more significant factor). It is interesting to observe that the system performance degrades not only when the limit is decreased but also increasing. Indeed, allowing higher

5 248 power on per link basis increases the total transmit power in the cell, which in tums increases, the downlmk interference. Consequently the E&, (Pilot) decreases, increasing the number of connections. Based on Figure 6, at 5% outage point, the minimum requirement for the DCH max power limit is given in Table 1. However, these values are valid for single service operating in a network. Table 1 - Requirements for the DCH max limit. Figure 7 illustrates the performance for a network using the same assumptions as in pilot power section, i.e. a mix of services. As it was shown on figure 6, each rate has a corresponding MAX-DCH limit optimum. Firstly, it is interesting to note that the outage has a minimum value corresponding to an optimum MAX-DCH limit value, 0.9 Watts. It should be noted that at 0.9 watts, most affected mobiles corresponds to 384K mobiles. To obtain a near 0%-outage, it is recommended to use 2.6 watts as depicted in Table 1. This optimum value will change depending on the mix of services, e.g. a higher proportion of high bit rates would shift this requirement towards 2 watts in order to maintain the same Grade of Service (GoS), characterised by the 5% outage. Following the same trend as in Figure 6, the outage starts increasing again when the MAX-DCH limit is set too high. To overcome this phenomenon, the possibility to set a MAX-DCH limit per rate would optimise the outage. A common threshold has been used for all rates, which implies that a trade-off between QoS of the services should be found. A strategy based on individual threshold would allow the operator to benefit from more flexibility. interference in a cell. Therefore the nature of the services in a cell will heavily influence the choice of the pilot. REFERENCES (1) S. Burley, Capacity of an integrated LCD/UDD WCDMA cellular network, IEE 3G 2001, London. (2) S. Burley, Downlink Capacity Estimation of an Integrated LCDAJDD WCDMA Cellular Network, PIMRC 2001, Atlantic City. (3) Holma, Toskala, WCDMA for UMTS, Wiley & Sons, (4) Love, R.T.; Beshir, K.A.; Schaeffer, D.; Nikides, R.S., A pilot optimization technique for CDMA cellular systems, VTC Fall. IEEE VTS 50*, Volume: 4, 1999 Page(s): vo1.4 (5) 3GPP, TS , Radio Resource Management Strategies, (6) R.Owen, P.Jones, S.Dehgan, D.Lister, WCDMA capacity and planning issues, IEE Electronics & Communication Engineering Journal, Volume 12 Number 3, pp , June (7) C. Noblet, R. Owen, N. Wahid, Assessing the GSM Cell-Site re-use for 3G Networks, IEE 3G 2001, London. It\ t 144 * 384 CONCLUSION This paper has reviewed the impact of the pilot power and allocated power per dedicated downlink channel in a WCDMA system. The Echo measurements collected at the mobile, as depicted in Figure 4 illustrates the effect of cell breathing. Also linked to the coverage, from a downlmk perspective is the amount of DCH power available for each service, function of the environment. It will be the choice of the operator to anticipate which services will be deployed in the network to adjust accordingly the Node B power parameters. The pilot power value has a critical impact on the coverage, and certainly varies with the cell load. This is also known as the cell breathing effect. Since each individual service requires different downlink transmission power, it directly relates to the amount of Mm-DCH Power Umlt 1 Figure 6 - Max DCH Power Limit vs. Outage

6 f 20 i 15. g 0 9 Was for 95% GoS 1 8 io MU OW P0w.r [wraa] Figure 7 - Max DCH Power Limit vs. Outage