Tài liệu Adaptive WCDMA (P11) docx

11 CDMA packet radio networks 11.1 DUAL-CLASS CDMA SYSTEM In this chapter we consider some additional details of packet transmission in a Code Division Multiple Access (CDMA) radio network. The approach is very much based on Reference [1]. We start with dual-class traffic and then extend the analysis to multimedia systems. These two classes a re characterized by the following set of parameters: Class 1: The users in this class are delay intolerant. When transmitting information, they require support for a constant bit rate of R 1 bit s −1 ; they can tolerate a bit error rate (BER) of at most P b1 . Class 2: The users in this class are delay tolerant. When transmitting information, they require support for a bit rate of at least R min bit s −1 ; they can tolerate a BER of at most P b2 . When not transmitting information, it is assumed that the users still communicate with the base for synchronization purposes. The bit rate used in this synchronization mode is denoted as R 0 bit s −1 , and is referred to as the ‘idle rate’. One would e xpect that R 0 <R 1 ,R 0 <R min ; more detailed constraints on R 0 are given later. In an actual system, Classes 1 and 2 could represent voice and data users, respectively. Minimum rate R 0 would be used in a control channel. According to equation (10.10) a unique solution to a minimum total transmit power problem exists if and only if N 1  i=1 1  W R (1) i · 1 SIR (1) i + 1  + N 2  j=1 1  W R (2) i · 1 SIR (2) i + 1  < 1 − IW min i  P peak i h i  W R i · SIR i + 1  N 1 +N 2 i=1 (11.1) Adaptive WCDMA: Theory And Practice. Savo G. Glisic Copyright ¶ 2003 John Wiley & Sons, Ltd. ISBN: 0-470-84825-1 370 CDMA PACKET RADIO NETWORKS By using the following notation h 1 : (mobile to base) gain of the ith generic user • SIR (1) i = γ 1 = constant ∀ Class 1 users i; SIR γ 1 ⇔ BER P b1 • SIR (2) i = γ 2 = constant ∀ Class 2 users i; SIR γ 2 ⇔ BER P b2 minimum total transmit power solution is obtained from equation (11.1) for minimum required rate R min .P peak i →∞∀i, constraint (11.1) now becomes N 1  W R 1 · γ 1 + 1  + N max 2  W R min · γ 2 + 1  < 1 (11.2) Given that N 1 Class 1 users are present, one may support at most N max 2 =       1 + W R min · γ 2 − N 1 ·  W R min · γ 2 + 1   W R 1 · γ 1 + 1        Class 2 users (11.3) Given that N 1 Class 1 and up to N max 2 Class 2 users are present, the transmit powers assigned to the users will then be such that the following hold, (see also Chapter 10, Section 10.1). Each Class 1 user and Class 2 user will achieve an signal-to-interference ratio (SIR) of exactly γ 1 and γ 2 , respectively; this follows, as noted before, from the assumption of perfect power control. The BER requirements of users of both classes will therefore also be met with equality. The transmitted powers will be such that their sum is as small as possible; hence, the interference to other cells is minimized. It is assumed that admission control will handle the task of ensuring that the number of users of each class satisfies the constraints in equations (11.2 and 11.3). 11.1.1 Maximization of Class 2 throughput 1. Mode 1-Unscheduled Class 2 Transmissions: All N 2 users are allowed to transmit information, each at rate R 2 .TherateR 2 is chosen to be the largest possible so as to satisfy constraint (11.1). Since N 2 ≤ N max 2 , one has R 2 ≥ R min . This transmission mode is very similar to that f ollowed in the present systems. A more efficient (from the point of view of throughput) version of this scheme would allow each Class 2 user to transmit at an a ppropriate different rate. 2. Mode 2-Scheduled Class 2 Transmissions: The Class 2 transmissions are scheduled in such a way that at any given instant, only k 2 (<N 2 ) of users are transmitting information. The remaining (N 2 − k 2 ) are in contact DUAL-CLASS CDMA SYSTEM 371 with the base at the idle/synchronization rate R 0 bit s −1 . When transmitting information, a Class 2 user is allowed to transmit at a rate R ∗ 2 , which, again, is chosen so as to be the maximum value satisfying constraint (11.1). Thus, assuming a fair division of time, each Class 2 user has a ‘duty cycle’ of a fraction of time when it is transmitting information, given by  N 2 −1 k 2 −1    N 2 k 2  = (k 2 /N 2 ). The remaining fraction of time is spent in maintaining synchronization with the base at a rate R 0 . If the transmission rate T of Class 2 for k 2 = 1isT 2 , then the throughput gain G measured by the ratio of Mode 2 to Mode 1 throughputs is [1] G = T 2 R 2 = R 0 R 2  W R 2 γ 2 + 1  + N 2 W R 2 γ 2  1 − R 0 R 2  N 2  W R 2 γ 2 + 1  − N 2 2  1 − R 0 R 2  (11.4) Given R 0 , one has the following possibilities: Case 1 – R 0 ≤ R 0,upper,1 : In this case, one has T ≥ R 2 for any admissible value of k 2 , that is, ∀k 2 ∈ [1, ,N 2 − 1]. Case 2 – R 0,upper,1 <R 0 ≤ R 0,upper,2 : In this case, one has T ≥ R 2 for the set of k 2 values k 2 ∈{1, ,k max 2 }. Here, k max 2 ≤ (N 2 − 1);alsoN 2 ≥ 1 ⇒ k max 2 ≥ 1. Case 3 – R 0 >R 0,upper,2 : In this case, one has T<R 2 for any admissible k 2 . Parameters R 0,upper,1 ,R 0,upper,2 and k max 2 are given as R 0,upper,1 = R 2 2  W γ 2 + 2R 2  R 0,upper,2 = N 2 R 2 2  W γ 2 + [N 2 + 1]R 2  k max 2 =       N 2 R 2 (R 2 − R 0 ) R 0  W γ 2 + R 2        (11.5) For illustration purposes we use a system [1] with spreading bandwidth W = 1.23 MHz; Class 1 bit rate R 1 = 9.6kbs −1 , with a minimum SIR of γ 1 = 7 dB, minimum Class 2 bit rate R min = 14.4kbs −1 , with a minimum SIR of γ 2 = 8.5 dB (7.0795) and idle bit rate R 0 = 1.2kbs −1 . The Class 2 SIR requirement γ 2 is chosen under the conservative assumption that power control at higher rates might involve higher overheads. The number of Class 1 users N 1 was taken to be the primary variable; on the basis of this, the maximum number N max 2 of Class 2 users permitted was computed according to equation (11.3). 372 CDMA PACKET RADIO NETWORKS 1.5 2 2.5 Gain G 123456789 Number of Class 2 users, N 2 Figure 11.1 Gain G = T 2 /R 2 versus Class 2 population N 2 with N 1 = 8,N max 2 = 9. The number of Class 2 users in the system was then varied from 1 to N max 2 ,andthe corresponding gain G was computed from equation (11.4). The results are plotted in Figure 11.1. One should be aware that for N 1 smaller and smaller, G would be larger and larger. Variation of the synchronization rate limits R 0,upper,1 and R 0,upper,2 is shown in Figure 11.2. In the case of imperfect power control, the initial condition (11.1) should be replaced accordingly (see Section 10.1 of Chapter 10). We use again the spreading bandwidth W = 1.23 MHz, Class 1 bit rate R 1 = 9.6 kbit s −1 , with the power-controlled target SIR parameter γ 1 = 7 dB, minimum Class 2 bit rate R min = 14.4 kbit s −1 , with the power- controlled target SIR γ 2 = 8.5 dB (7.0975) and idle bit rate R 0 = 1.2 kbit s −1 . The results are shown in Figure 11.3. In the sequel, we consider a situation where an upper limit is placed on the peak interference that a particular cell can create in another. Clearly, such a constraint would translate to peak transmit-power limits on the mobiles in that cell. Also, mobiles located close to the boundary between the cells would have more stringent peak transmit-power limits than those in the interior. Considering the application of the scheduled transmission mode described earlier in such a situation, we note that the presence of constraints on the peak transmit powers translate to constraints on the peak transmission rate, which limits the throughput gains due to scheduling. Thus, in order to better exploit the looser constraints on the Class 2 users in the cell interior, it might be advantageous in such situations to schedule the transmissions of only a certain subset of the Class 2 users in the cell. We use notation (IW /h (2 ) j P peak j ) for the ratio of the power received from all DUAL-CLASS CDMA SYSTEM 373 40 10 20 30 0 123456789 R 0, upper, 2 R 0, upper, 1 R 0,upper,1, R 0,upper,2 in kb s −1 Number of Class 2 users, N 2 Figure 11.2 The variation of the synchronization rate limits. R 0,upper,1 and R 0,upper,2 versus N 2 for the N 1 = 8,N max 2 = 9. 0.8 1 2 2.6 s = 1.5 dB s = 2.5 dB s = 0.5 dB s = 0.0 dB 123456789 Gain G Number of Class users N 2 Figure 11.3 Gain T = T 2 /R 2 versus Class 2 population N2 for different values of power control error (τ )andN 1 = 8. 374 CDMA PACKET RADIO NETWORKS 1 2 1.2 1.4 1.6 1.8 0.8 Gain G 123456789 Number of Class 2 users, N 2 RPR = 0.1 RPR = 0.2 RPR = 0.3 RPR = 0.5 RPR = 0.7 Figure 11.4 The variation of the throughput gain G versus RPR and N 2 for N 1 = 8. sources outside the cell to that of the weakest link user at the base station (BS), assuming that the user is transmitting at its maximum power. It will be referred to as the received power ratio (RPR). Figure 11.4 shows gain G = T 2 /R 2 versus Class 2 population N 2 with N 1 = 8 and RPR being a parameter. 11.1.2 Adaptive and reconfigurable transmission The Class 2 mobiles must be capable of variable rate transmission, dependent on avail- able residual capacity and the base station capable of the corresponding reception. The information transmission rate to be used by the Class 2 users (which was denoted as R ∗ 2 ) will be decided by the base station, using information about the population distribution in the system, and in the constrained transmit power case, knowledge of the RPR parame- ter. In practice, the user-population distribution may be a slowly changing variable. This would imply that Class 2 information rate changes do not have to be effected too often. The Class 2 mobiles transmission rate alternates between the information rate R ∗ 2 and the synchronization rate R 0 ; hence, there must also be a mechanism to coordinate these rate changes with the base station. Mechanism to schedule the Class 2 transmission must be available. One option is for Class 2 mobiles to transmit in a ‘round-robin’ fashion. Another option is that each mobile has a fixed information transmission time of τ units within each cycle time of C units. The actual value of C depends on the amount of buffer space to be provided at ACCESS CONTROL FOR WIRELESS MULTICODE CDMA SYSTEMS 375 the mobile, the average number of Class 2 users expected and so on. Once C is known, an adaptive τ is simply given by τ = (C/N 2 ) where, as before N 2 is the number of Class 2 users. This form of scheduling will also have to be centrally controlled by the base station using information about the user-population distribution. There has to be a mechanism by which the base station informs a particular Class 2 mobile about the cycle time C, its information transmission time τ, as well as its ‘place’ within the cycle (this is called the slotting mechanism). Although it is desirable to have a fine slotting of the Class 2 users, that is, an arrangement of the slots such that their transmissions do not overlap, it may be difficult to achieve in practice without a significant increase in complexity. In that case, one could resort to coarse slotting in which some (as small as possible) part of the slot assigned to a user overlaps with that assigned to another user. The overlapping portions would then correspond to the case k 2 = 2 rather than the desired best case k 2 = 1. This would lead to a certain reduction in the throughput gains, but would simplify the implementation of the slotting mechanism. More sophisticated scheduling schemes, which exploit the traffic characteristics of the Class 2 mobiles, can be designed above the basic scheme. For example, a Class 2 mobile mayhavenoinformationtotransmitinitsassigned slot in which case that particular slot could be reassigned to some other user. This would require an adaptive reconfiguration of the upper layers in the network. Such schemes would lead to additional throughput gains. However, the base station would need additional knowledge about the state of the data in the mobiles, and would add some additional scheduling load to the system. Delay constraints need to be incorporated into the scheduling. A Class 2 mobile would need to use significantly higher transmit power in its information transmission slot time τ as compared to the rest of the time in a cycle, in order to maintain the same SIR. During its information transmission slot of time τ , a Class 2 mobile must increase its transmit power to correspond to the rate R 2 and lower it for the rest of the cycle to correspond to therateR 02 . 11.2 ACCESS CONTROL FOR WIRELESS MULTICODE CDMA SYSTEMS We assume a short-term traffic control in the uplink, that is, burst level traffic control. A multicode CDMA system (MC-CDMA) for the integration of multirate, multimedia services is considered [2,3]. In MC-CDMA, a code can be used to transmit information at a basic bit rate. Users ( video or data) who need higher transmission rates can use multiple codes in parallel. A two-phase congestion control is used for managing the data traffic and the Packet Error R ate (PER) of Real-Time (RT) traffic. The first congestion control phase makes sure that there is at most a prespecified number of data users who can transmit at a time. For those data users who have been granted the right to transmit (i.e. those who have been assigned CDMA codes), they further follow the second con- gestion control phase imposed by the BS so as to minimize the impact on the PER of RT traffic. 376 CDMA PACKET RADIO NETWORKS 11.2.1 Call level model The CDMA system consists of a large number of users who can generate voice calls or data calls. Users with voice calls or data messages will contact the central station (called the base station or BS in the context of c ellular networks) by sending reservation requests through the signaling channels. For details on the Universal Mobile Telecommunica- tion System (UMTS) standard, see Chapter 17. On successful reception of a reservation request, the central station will make a decision as to whether the request can be granted the admission control. Depending on whether the system has enough resources in the traffic channel to handle the incoming call, the reservation request will be accepted or rejected, and the user is notified via the downlink (DL) (the forward link). An accepted voice call is assigned a CDMA code immediately, but an accepted data call is first queued at the central station, and will be assigned CDMA codes subject to the congestion control to be elaborated. The time axis of the CDMA system is fully slotted, with a slot duration equal to the transmission time of a packet that is of the same size for both types of traffic. All users are synchronized at the packet level. The voice generation processes and data calls are Poisson distributed with the average rates of λ v and λ d calls per slot, respectively. It is assumed that the duration of a voice call and the data message length are geometrically distributed with an average of 1/µ and L slots, respectively. 11.2.2 Data congestion control scheme As it was already discussed in Chapter 10, owing to the on–off nature of accepted voice calls, the CDMA channel is not always fully utilized. Therefore, the central station can allow data traffic to dynamically ‘steal’ the unused channel capacity left by the voice users. It is the congestion control scheme that makes sure that the quality of service (QoS) of the established voice calls is not compromised, that is, the packet error probability remains less than a prespecified value. The congestion control scheme will make sure that among the accepted data message requests queued at the central station, only the first M d requests are assigned C d CDMA codes. The data users with assigned CDMA codes start to transmit their outstanding messages with probability p start , which can be determined by the central station according to the state of the system. Upon starting the transmission of its data message, the data user will not stop until the end of the message. Since the data message length is geometrically distributed with an average of L packets, the transmission time of the data message approximately follows a geometric distribution with an average of L/C d slots. When a data user advances to the first M d positions, it is in the standby state, where the user will enter the active state with probability p start . When in the active state, the user will transmit C d packets in parallel using the C d spreading codes assigned by the central station. Since the transmission time of a data message is geometrically distributed with mean L/ C d , the transition probability from the active state to the done state is C d /L.The state diagram of an admitted data user is shown in Figure 11.5. How p start is determined on the basis of some side information will be discussed next. ACCESS CONTROL FOR WIRELESS MULTICODE CDMA SYSTEMS 377 1 1− p start Done Active Standby p start 1− C d / L C d / L Figure 11.5 State transition diagram of an admitted data user. 11.2.3 Feedback-driven congestion control In order to fully utilize the unused channel capacity while maintaining the voice PER, p start needs to be dynamically updated (according to the extra information about the system state available to the central station) and broadcast to the data users in the standby state. To set p start , for time-slot t + 1, we assume that the central station can obtain some side information about the number of users transmitting in time-slot t. Although there exist many ways to update p start based on the side information I(N v t ,N d t ),whereN v t and N d t denote the number of voice and data users that transmits in time-slot t, respectively, we consider a threshold scheme with I(N v t ,N d t ) = (N v t ,N d t ).Define f(N v t ,N d t ) = N v t + N d t × C d (11.6) f(·) gives the number of active codes in the CDMA channel in time slot t, which can be used as a congestion index. Given the side information (N v t ,N d t ), the probability p start for time-slot t + 1is p start =            min  1, T − f(N v t ,N d t ) C d (min(M d ,N q t ) − N d t )  , if f(N v t ,N d t ) ≤ T 0, otherwise            (11.7) T is the threshold for congestion control N q t is the number of data requests accepted by the system, and min(x,y) is the minimum of the two arguments. In other words, if the system congestion index does not exceed the threshold T , the data users who are in the standby state should transmit with a probability selected so as to fill available ‘slots’ on an average. On the other hand, if the system congestion index exceeds T ,activedatausers are allowed to continue their transmission, but other data users standing by should refrain from transmitting. The rationale behind the selection of p start is that [T − f(N v t ,N d t )]/C d can be thought of as the residual channel capacity, and [min(M d ,N q t ) − N d t ] is the number of data users standing by for transmission. By choosing p start equal to the ratio of the above two factors, the expected number of data users who start to transmit will be equal 378 CDMA PACKET RADIO NETWORKS 1 2 3 M d One code per voice user C d Codes per data user Blocking f ( N t v , N t d ) ⇒ 1 2 M v max … … M d max − M d l v l d Figure 11.6 System level view of the traffic control mechanism. Table 11.1 System parameters Symbol Value Number of voice users admitted to the system M v 1–6 Transition probability of a voice source from On-to-Off state p on − off 1/17 Transition probability of a voice source from On-to-Off state p on − off 1/22 Slot duration (msec) 20 Data message arrival rate (message/slot) λ d Var i a b le Average length of data message (packets) L 10 Capacity of the data message request queue at BS M max d 10 or 15 Maximum number of data users in the standby state M d 2–8 Number of CDMA codes per data user C d 1, 2 Threshold for congestion control T 4–10 Packet length (bits) 255 Number of correctable bit errors per packet 4 Bit energy to (one-sided) noise spectral density (dB) E b /N 0 11 Processing gain G 64 Rician factor K R 5 −∞ to the left over (residual) capacity. The system level view of the traffic control mechanism is shown in Figure 11.6. For illustration purposes a set of the system parameters from Table 11.1 are used [3]. Data throughput versus data arrival rate is shown in Figure 11.7. Optimal threshold T as a function of M v is shown in Table 11.2 [ 3]. [...]... and expected delay for the specified cell population, the default cell population, are shown in Table 11.4 More details are given in Figures 11.17 to 11.19 The issue of multimedia packet transmission in WCDMA will be addressed in much more detail in Section 11.7 11.5 CDMA ALOHA NETWORK USING p-PERSISTENT CSMA/CD PROTOCOL 11.5.1 Network description A network consists of a number of terminals that are equally... information (FCSI) This section presents an analysis of implementation losses in such systems Such protocols can be used in different segments (or domains) of the future wireless IP networks including WCDMA cellular networks, wireless LANs and Ad Hoc networks The motivation behind these protocols is to control the channel access by statically or dynamically changing packet transmission permission probability... assumption is valid For instance, if TP = 20 ms and r = 30, then TP /r = 2/3 ms, that is identical to the random-access channel time-slot offset (contention slots) defined in the 3GPP standards for the 3G WCDMA cellular PCS (see Chapter 17) In Reference [10] for a similar feedback control protocol, the assumption that at most one event can occur in a time-slot with zero feedback and access delays was used, . − IW min i  P peak i h i  W R i · SIR i + 1  N 1 +N 2 i=1 (11.1) Adaptive WCDMA: Theory And Practice. Savo G. Glisic Copyright ¶ 2003 John Wiley. T 2 /R 2 versus Class 2 population N 2 with N 1 = 8 and RPR being a parameter. 11.1.2 Adaptive and reconfigurable transmission The Class 2 mobiles must be capable