CH5.

fm Page 47 Monday, November 22, 1999 10:42 AM

C H A P T E R

Mobile Station Access and Paging

DMA mobile stations transmit on the access channel according to a random access protocol. Detailed procedures of this random access protocol and ranges of various access parameters can be found in References [5.1] and [5.2]. This chapter, which consists of four sections, analyzes the protocol performance and discusses appropriate settings of mobile station access parameters. Section 5.1 describes the operation of the mobile station access protocol and its associated parameters. Section 5.2 presents persistence delays for access request attempt. Section 5.3 presents throughput and delay performance of the mobile station access protocol and discusses an approach to determine the access channel capacity. Section 5.4 evaluates paging channel capacity in terms of various paging messages related to call processing and special services such as short message services (SMSs) and voice mail services (VMSs).

5.1 Description of Mobile Station Access Protocol

As described in Reference [5.1], the mobile station transmits on the access channel using a random access procedure. Figure 51 shows a flowchart of the CDMA mobile station access protocol. The entire process of sending one message and receiving (or failing to receive) an acknowledgment for that message is called an access attempt. Each transmission in the access attempt is called an access probe (see Figures 52, 53, and 54). The mobile station transmits the same message in each access probe in an

47

CH5.fm Page 48 Monday, November 22, 1999 10:42 AM

48

Ch. 5 Mobile Station Access and Paging

Begin

PROBE=0 SEQ = 0

Hash using ESN and PROBE_PN_RAN to obtain RN

From B Access channel response

Yes

No If beginning of slot generate RP

RP < P

No

Yes Generate random number RA between 0 and ACC_CHAN

Initiate transmit power

To A

Figure 51 Access Procedure

CH5.fm Page 49 Monday, November 22, 1999 10:42 AM

Description of Mobile Station Access Protocol

49

From A

Transmit probe on access channel RA at system timing plus RN chaps. Set TA

Access successful

Yes

Acknowledgment received Increase transmit power by PWR_STEP dB No Wait RT slots

No

Timer TA expired

To B

Yes PROBE = PROBE + 1 Generate random number RT between 0 and PROBE_BKOFF + 1 Yes

Wait RS slots

PROBE < NUM_STEP

Generate random number RS between 0 and BKOFF + 1

No SEQ = SEQ + 1

Yes

SEQ < MAX_REQ_SEQ (MAX_RSP_SEQ

No

Access failure

Figure 51 Access Procedure (Continued)

Reproduced from [5.1] and [5.2] under written permission of the copyright holder (Telecommunications Industry Association)

CH5.fm Page 50 Monday, November 22, 1999 10:42 AM

50

Ch. 5 Mobile Station Access and Paging

Access Probe One Access Channel Slot

Access Channel Preamble 1 + PAM_SIZE (1 - 16 frames)

Access Channel Message Capsule 3 + MAX_CAP_SIZE (3 - 10 frames)

Access Channel Frame (20 ms)

4 + PAM_SIZE + MAX_CAP_SIZE (4 - 26 frames)

Actual Access Probe Transmission

PN Randomization Delay = RN chips

Figure 52 Access Channel Request and Response Attempts (Access Probe)

Reproduced from [5.1] and [5.2] under written permission of the copyright holder (Telecommunications Industry Association)

access attempt. Each access probe consists of an access channel preamble and an access channel message capsule. The number of twenty millisecond frames expresses the length of the preamble 1 + pam_sz as well as the length of message capsule 3 + max_cap_sz. Thus, the duration of an access probe (access channel slot) is 4 + pam_sz + max_cap_sz frames. Within an access attempt, access probes are grouped into an access probe sequence. Mobile stations send two types of messages on the access channel: a response message (one that is a response to a base station message, see Figure 53) or a request message (one that the mobile station sends autonomously, see Figure 54). Each access attempt consists of up to max_req_seq (for a request access) or max_rsp_seq (for a response access) access probe sequences.

CH5.fm Page 51 Monday, November 22, 1999 10:42 AM

Description of Mobile Station Access Protocol

51

Access Attempt Access Probe Sequence 1 Seq MAX_RSP_SEQ (15 max)

Seq 2

Seq 3

RS Message Ready

RS

Access Probe 1 + NUM_STEP (16 max) Access Probe 3 PI PI Access Probe 1 Access Probe 2

IP

TA

RT

TA

RT

TA

Select Access Channel (RA), initialize transmit power

Figure 53 Access Channel Response Attempts

Reproduced from [5.1] and [5.2] under written permission of the copyright holder (Telecommunications Industry Association)

CH5.fm Page 52 Monday, November 22, 1999 10:42 AM

52

Ch. 5 Mobile Station Access and Paging

Request Attempt Access Attempt Access Probe Sequence 1 Seq MAX_REQ_SEQ (15 max)

Seq 2

Seq 3

Message Ready

PD

RS

PD

RS

PD

Access Probe 1 + NUM_STEP (16 max) PI PI Access Probe 1 Access Probe 3 Access Probe 2

IP

TA

RT

TA

RT

TA

Select Access Channel (RA), initialize transmit power

Figure 54 Access Channel Request Attempts

Reproduced from [5.1] and [5.2] under written permission of the copyright holder (Telecommunications Industry Association)

Mobile stations pseudo-randomly determine the timing of the start of each access probe sequence. For every access probe sequence, a backoff delay, RS, from 1 to 1 + bkoff slots is generated pseudo-randomly. In the case for request access probe sequences, for each slot after the backoff delay RS, the mobile station performs a pseudo-random persistence test. If the persis-

CH5.fm Page 53 Monday, November 22, 1999 10:42 AM

Description of Mobile Station Access Protocol

53

tence test passes, the first probe of the sequence begins in that slot. Failure of the persistence test defers the access probe sequence until at least the next slot. Thus, the use of a persistence test imposes an additional delay, PD. For each access channel slot, the persistence test generates a random number and compares it with a predetermined threshold. The pre-computed threshold is different, depending on the nature of the request, the access overload class n, its persistence value psist(n), and its persistence modifier msg_psist, for message transmission or reg_psist for registrations. Each access probe sequence consists of up to 1 + num_step access probes, all transmitted on the same access channel. Mobile stations pseudo-randomly choose the access channel number RA (used for each access probe sequence) from 0 to acc_chan among all the access channels associated with the current paging channel. The mobile station shall use this access channel number for all access probes in that access probe sequence. The mobile station transmits the first probe in each access probe sequence at a mean output power level (referenced to the nominal CDMA channel bandwidth of 1.23 MHz) depending on the open-loop power estimate, the initial power offset for access init_pwr, and the nominal transmit power offset nom_pwr. Each subsequent access probe is transmitted at a power level of a specified amount PI (determined from pwr_step) higher than the previous access probe until it obtains an acknowledgment response or the sequence ends. Between access probes, the mobile station shall disable its transmitter. Expression of the timing of access probes and access probe sequences is in terms of access channel slots. The transmission of an access probe begins at the start of an access channel slot. A procedure called PN randomization determines the precise timing of the access channel transmissions in an access attempt. For the duration of each access attempt, the mobile station computes a delay, RN, from 0 to 2probe_pn_ran PN chips using a (non-random) hash function that depends on its electronic serial number ESN. The mobile station delays its transmit timing by RN PN chips. This transmit timing adjustment includes delay of the direct sequence spreading long code and of the quadrature spreading I and Q pilot PN sequence, so it effectively increases the apparent range from the mobile station to the base station. This increases the probability that the base station will be able to separately demodulate transmissions from multiple mobile stations in the same access channel slot, especially when many mobile stations are at a similar range from the base station. The mobile stations also pseudo-randomly generate timing between access probes of an access probe sequence. After transmitting each access probe, the mobile station waits a specified period, TA = 80 (2 + acc_tmo) milliseconds, from the end of the slot

CH5.fm Page 54 Monday, November 22, 1999 10:42 AM

54

Ch. 5 Mobile Station Access and Paging

to receive an acknowledgment from the base station. If it receives an acknowledgment, the access attempt ends. If not, the mobile station transmits the next access probe after an additional backoff delay RT from 1 to 1 + probe_bkoff slots.

5.2 Average Persistence Delay for Access Request Attempt

As discussed in the previous section, for access due to mobile station request, ANSI JSTD-008 [5.1] requires that a persistence test be performed prior to initiating the access probe sequence to control the rate at which the mobile station transmits requests. Assessing the appropriate range of persistence values to be assigned to the mobile stations requires information about the amount of delays due to persistence tests. This section includes calculations of average persistence delays as a function of persistence values for various types of request and access overload classes. For each access channel slot, the persistence test generates a random number RP (0 < RP < 1) and compares it with a predetermined threshold P. The access probe sequence initiates if the generated random number RP is smaller than the predetermined threshold P. Since the random number RP is generated from uniform distribution over the unity interval,
Pr {RP < P} = P .

(5.1)

In other words, a larger P implies a higher probability of initiating the access probe sequence. The pre-computed threshold P, in general, is different, depending on the nature of the request, the access overload class, and its persistence value psist(n) as well as its persistence modifier. As an example, for registration request of access overload classes 0 through 9, if psist(n) = 63, then P = 0. Thus the persistence test fails and therefore initiates no access probe sequence; if psist(n) is not equal to 63, for a given persistence modifier reg_psist, P is the monotonically decreasing function of psist(n); the larger the psist(n), the smaller the P, thus the smaller the probability of initiating the access probe sequence. Table 51 summarizes the persistence test thresholds for various types of requests and access overload classes. From Table 51, note that the maximum persistence value is 63 for access overload classes 0 through 9, and is 7 for access overload classes 10 through 15. If the base station assigns the maximum persistence value to the mobile station, the access attempts fail and the mobile station enters the system determination substate of the mobile station initialization state (see References [5.1] and [5.2]). For the persistence value not equal to the maximum, random persistence delays can be incurred to control the transmissions of mobile station requests. The persistence

CH5.fm Page 55 Monday, November 22, 1999 10:42 AM

Average Persistence Delay for Access Request Attempt

55

Table 51 Requests

Persistence Test Thresholds (P ) for Registration, Message, and Other

access overload classes n = 0, 1, , 9 psist (n) 63 psist (n) = 63 0 access overload classes n = 10, 11, , 15 psist (n) 7 2psist(n) reg_psist 2psist(n) msg_psist 2psist(n) psist (n) = 7 0

Registration Request Message Request Other Request

psist ( n ) reg_psist -----------------4

psist ( n ) msg_psist -----------------4 psist ( n ) -----------------4

delay PD is the number of times the tests have to be performed before the condition RP < P is satisfied. Thus, persistence delay PD is a random variable, and its discrete probability density is geometric with parameter P, that is,
Pr{PD = k slots} = (1 P )k P for k = 0, 1, 2, L .

(5.2)

One performance measure that can characterize the persistence test is the average persistence delay. Let E{} be the expectation operator. The average persistence delay E{PD}is
E{PD} = k Pr{PD = k slots} =
k =0

1 P . P

(5.3)

Using the values of persistence test thresholds P in Table 51 for various types of request and access overload class, Table 52 summarizes their average persistence delay E{PD}. We numerically evaluate the derived expressions of the average persistence delay in Table 52. For registration requests with persistence modifiers (reg_psist) 0 through 7, the average persistence delay as a function of persistence value is plotted for access overload classes 0 through 9 in Figure 55, and for access overload classes 10 through 15 in Figure 56, respectively. Figures 55 and 56 can also be used for message requests when persistence modifiers reg_psist are replaced by msg_psist. For other requests, Figure 57 plots the average persistence delay as a function of persistence value for access overload classes 0 through 9. Figure 58 plots the average persistence delay as a function of persistence value for access overload classes 10 through 15.

CH5.fm Page 56 Monday, November 22, 1999 10:42 AM

56

Ch. 5 Mobile Station Access and Paging

Table 52 Requests

Average Persistence Delay (E{PD}) for Registration, Message, and Other

access overload classes n = 0, 1, , 9 psist (n) 63 psist (n) = 63 access overload classes n = 10, 11, , 15 psist (n) 7 2psist(n) + reg_psist 1 2psist(n) + msg_psist 1 2psist(n) 1 psist (n) = 7

Registration Request Message Request Other Request

) psist ( n - + reg_psist -----------------4 psist ( n ) ------------------ + msg_psist 4 psist ( n ) -----------------4

re g _p sist = 0 re g _p sist = 4
10000000

re g _p sist = 1 re g _p sist = 5

re g _p sist = 2 re g _p sist = 6

re g _p sist = 3 re g _p sist = 7

Average Persistence Delay (slot)

1000000

100000

10000

1000

100

10

Figure 55 Average Persistence Delay for Registration Request of Access Overload Classes 0 through 9
Note: If msg_psist is substituted for reg_psist, the plot is for the Message Request.

In general, for each type of request and access overload class, maximum allowable persistence delay is expected to be unsurpassed. Thus, given the maximum allowable

0.1 0 5 10 15 20 25 30 35 40 45 50 55 60 65

Persistence e V a lu e P e rsist e n c Value

CH5.fm Page 57 Monday, November 22, 1999 10:42 AM

Average Persistence Delay for Access Request Attempt

57

re g _p sist = 0 re g _p sist = 4
10000

re g _p sist = 1 re g _p sist = 5

re g _p sist = 2 re g _p sist = 6

re g _p sist = 3 re g _p sist = 7

Average Persistence Delay (slot)

1000

100

10

Average Persistence Delay (slot)

1 0 1 2 3 4 5 6

Persistence a lu e P e rsist e n c e VValue

Figure 56 Average Persistence Delay for Registration Request of Access Overload Classes 10 through 15
Note: If msg_psist is substituted for reg_psist, the plot is for the Message Request.

100000

10000

1000

100

10

Figure 57 Average Persistence Delay for Other Requests of Access Overload Classes 0 through 9

0.1 0 5 10 15 20 25 30 35 40 45 50 55 60 65

Persistence Value P e rsist e n c e V a lu e

CH5.fm Page 58 Monday, November 22, 1999 10:42 AM

58

Ch. 5 Mobile Station Access and Paging

100

Average Persistence Delay (slot)

10

1 0 1 2 3 4 5 6

P e rsist e n c e Value Persistence V a lu e

Figure 58 Average Persistence Delay for Other Requests of Access Overload Classes 10 through 15

persistence delay for each type of request and access overload class, the appropriate range of persistence values can be assessed.

5.3 Access Channel Capacity

The access channel is used in the reverse link and supports origination calls, page responses, order messages, registrations, and short message services (SMSs). Even though the access channel uses a slotted random access protocol, a high throughput is possible in the access channel. Each access slot is of a fixed duration and consists of multiple frames, each of which is 20 msec in duration; its long code identifies it uniquely. An access probe is successful if the Eb /Io at the base station exceeds a threshold1 and there are no other probes received within an interval of two PN chips. In other words, if two access attempts are separated by an interval of two chips, in most cases, one of the access attempts will get through successfully. Therefore, even when the second access attempt comes before the first access attempt finishes its setup process, the first access attempt can still be successfully decoded without being impacted by the second access attempt.

1.

The threshold can be dynamic; a fixed signal-to-interference ratio (SIR) is assumed here for simplicity.

CH5.fm Page 59 Monday, November 22, 1999 10:42 AM

Access Channel Capacity

59

A large amount of access traffic may generate an unacceptable interference in the reverse link. As described in Section 5.1, after the first access probe fails to gain access to the system, the next access probe will increase its probing power by an amount specified by the access channel parameter PI. The aggregate effect will result in the increase of the total power received at the cell site and the reduction of the reverse-link capacity. To guarantee the capacity for the voice traffic channels, any excessive use of the capacity by the access channel should be limited. In this section, we develop an approach to evaluate the IS-95 access channel capacity. After investigating, in Sections 5.3.1 and 5.3.2, the impact of access traffic on the reverse-link interference through a simulation, we estimate the capacity of a single access channel in terms of the number of CDMA carriers that can be supported, which is the subject of Section 5.3.3. The results obtained in this section can be used as a guideline to estimate the capacity limitation and the impacts of other services, such as SMS, on the access channel. 5.3.1 Access Channel Simulation Model

To predict the capacity limit on the access channel, we use a simulation model of the access channel and a criterion based on the interference impact of the access channel on reverse link. This section describes the simulation model used to evaluate access channel performance. Within each access slot (seven frames for this case; see Table 53), we assume the arrival rate of the access attempt to have a Poisson distribution with a given mean. We also assume that the voice activity is 0.4 and the out-of-sector reverse-link interference is seventy-two percent of the in-sector interference. Since we are interested in the capacity of the access channel, we assume that voice traffic channel power is constant and the system is operating at fifty percent loading.2 The Eb /Io required for the access probes to be successful is typically higher than that for the voice traffic channel (see, for example, Reference [5.3]).3 We assume that the required Eb /Io for an access success is fifty percent higher (or 1.76 dB above) that of the voice traffic channel. The open-loop power control algorithm determines the initial mobile transmit power used for an access attempt. Because of the imbalance in the RF environment

2.

3.

Loading is in reference to the pole capacity (or pole point). The pole capacity, as discussed in Section 2.2.1.1, is a theoretical CDMA capacity by assuming infinite mobile transmission power. For example, the pole capacity of a CDMA system using an 8 Kbps vocoder is about forty radio channels. In this case, a fifty percent loading means that the system is operating at a capacity of twenty radio channels. For more discussion of the loading and pole capacity, see Section 8.1. We will define traffic channel Eb /Io in Chapter 7 (see, for example, (7.5)).

CH5.fm Page 60 Monday, November 22, 1999 10:42 AM

60

Ch. 5 Mobile Station Access and Paging

Table 53

Access Channel Parameters

Value 1 frame 2 frames 7 frames 3 2 Note preamble information 4+ pam_sz + max_cap_sz number of probes

Parameter pam_sz max_cap_sz message(slot) length num_step max_req_seq and max_rsp_seq

0.8 0.7 Throughput (accesses/slot) 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 PI = 3 dB PI = 4 dB PI = 5 dB PI = 7 dB

Arrival Rate (accesses/slot) Arrival Rate (accesses/slot) Figure 59 The Access Channel Throughput Performance versus Arrival Rate

between the forward- and reverse-links, the imperfections in the open-loop power control algorithm, and the loading in the system, the received signal-to-interference ratio resulting from the first access probe could be different from what is desired. We model the initial access channel power as a Gaussian random variable with mean that is 1.76 dB above the mobile transmit power needed to meet the traffic channel Eb /Io require-

CH5.fm Page 61 Monday, November 22, 1999 10:42 AM

Access Channel Capacity

61

16 14 12 Delay (slots) 10 8 6 4 2 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 PI = 3dB PI = 4dB PI = 5dB PI = 7dB

Arrival Rate (accesses/slot) Arrival Rate (accesses/slot) Figure 510 The Access Channel Delay Performance versus Arrival Rate

ment and variance of 8 dB. In this simulation, we do not explicitly consider the effect of channel variation (for example, fading) on the access channel. During the simulation, for a given number of access attempts per slot (or arrival rate), we count the number of successes per slot (or throughput) and the delays experienced by the successful attempts. The time from the completion of the transmission of the first access probe until the time the mobile station receives an acknowledgment defines the delay. We also measure the total received power on the access channel in terms of a fraction of the assumed sector (or cell) loading of fifty percent. For each value of the arrival rate and PI, we ran ten thousand trials to get average values of the throughput, delay, and power level. Table 53 lists all access channel parameters used in the simulation. 5.3.2 Access Channel Simulation Results

Figures 59 to 511 summarize some of the simulation results, showing the access channel performance in terms of throughput, delay, and interference impact on reverse link. Figure 59 plots the access channel throughput as a function of the access traffic arrival rate for a number of different values of PI. The plot shows that there is only very

CH5.fm Page 62 Monday, November 22, 1999 10:42 AM

62

Ch. 5 Mobile Station Access and Paging

120%

Access Channel Interference

100% 80% 60% 40% 20% 0% 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 PI = 3 dB PI = 4 dB PI = 5 dB PI = 7 dB

Arrival Rate (accesses/slot) Arrival Rate (accesses/slot) Figure 511 The Access Channel Interference Impact versus Arrival Rate

minor improvement by having a large value of PI (for example, 7 dB). Basically, the throughput is nearly insensitive to the choice of PI. In Figure 510, we plot the delay performance versus the amount of the access channel traffic. The improvement is less than two slots when increasing PI from 3 dB to 7 dB. The improvement decreases as the access traffic gets higher. Next, Figure 511 illustrates the amount of reverse-link interference generated by the access attempts. The access channel interference percentage shown in Figure 511 is with respect to the fifty percent traffic channel loading assumed in Section 5.3.1. Based on the results from Figure 59 and Figure 510, the PI of 7 dB may be chosen for the best performance of the throughput and delay. However, in Figure 511, we find that an excessive interference on the reverse link is possible for a big PI when the access traffic is very high. This implies that a smaller PI should be considered from the interference viewpoint. In addition, the throughput and delay performance of the access channel is not quite sensitive to the different PI. Therefore, a small PI should be chosen to minimize the interference impacts. This being the case, we choose PI of 3 dB when investigating the access channel capacity in Section 5.3.3. 5.3.3 Access Channel Capacity Analysis

As discussed before, it is important to consider the impact of interference from the access channel in investigating access channel capacity. With a high-access demand, the

CH5.fm Page 63 Monday, November 22, 1999 10:42 AM

Access Channel Capacity

63

Table 54

The Total Number of Subscribers Supported by a Sector*

Traffic (Erlangs) 13.2 31 49.6 68.7 88 107 Number of Subscribers (S) 660 (= 13.2/0.02) 1,550 (= 31/0.02) 2,480 (= 49.6/0.02) 3,435 (= 68.7/0.02) 4,400 (= 88/0.02) 5,370 (= 107.4/0.02)

Number of CDMA RF Carriers 1 (20 radio channels) 2 (40 radio channels) 3 (60 radio channels) 4 (80 radio channels) 5 (100 radio channels) 6 (120 radio channels)
*

Trunking efficiency has been considered for Erlang traffic. In a CDMA system, however, the air interface capacity does not necessarily inherit the trunking efficiency by having multiple carriers.

impact on reverse link might not be acceptable because, in this case, the reverse-link capacity decreases too much. Therefore we need to ensure a proper engineering of the access channel to limit the interference impact on reverse-link capacity. In general, we want to keep the access channel interference margin below ten percent of the traffic channel loading.4 For example, from Figure 511, in the case of PI of 3 dB, the interference margin at ten percent of the traffic channel loading corresponds to sixty percent access channel utilization.5 5.3.3.1 Erlang Traffic and Access Channel Traffic Assumptions

When a single frequency carrier of a CDMA system using an 8 Kbps vocoder can support twenty radio channels, it supports a capacity of 13.2 Erlangs with a two percent blocking probability.6 Assuming that the user traffic is 0.02 Erlang/subscriber (which is equivalent to each subscriber contributing one seventy-two-second call per hour), we are able to calculate the number of subscribers that a sector can support. By extending this concept to multiple-carrier cases, Table 54 shows the number of subscribers that a sector with multiple CDMA RF carriers can support. As mentioned before, the access channel supports call origination/termination, paging responses, orders, registrations, and SMSs. Statistically, each subscriber will

4. 5. 6.

With the assumed traffic channel loading of fifty percent, this corresponds to five percent of the pole capacity. In this case, the total sector loading on the reverse link will be fifty-five (= 50 + 5) percent. Note that the arrival rate in accesses/slot is translated into the access channel utilization here. This is from the Erlang B table; see, for example, [5.4].

CH5.fm Page 64 Monday, November 22, 1999 10:42 AM

64

Ch. 5 Mobile Station Access and Paging

Table 55

Access Channel Traffic Assumptions

Value 1/hr/subscriber* m/hr/subscriber 1 to 2/hr/subscriber 0.1 to 0.2/hr/subscriber 0 Note average number variable default value assume 10% of time-based registration cancel out with time-based registration

Access Channel Traffic Type call origination & termination (OT) short message service (SMS) time-based registration (Tr) zone-base registration (Zr) power on/off registration (Pr)
*

This is an average number. Some subscribers might make more than one call but others might make no calls during the time of interest.

contribute to the access traffic on different access traffic types. In Table 55, we list the values assumed for different access traffic types, which we will need in the next section. 5.3.3.2 Access Channel Capacity Calculation

The number of access messages per hour, A, required to support the access attempts of the types in Table 55 is given by
A = (OT + SMS + Tr + Zr + Pr ) S ,

(5.4)

where S is the number of subscribers supported by a sector shown in Table 54. Let M be the access channel capacity. Then, from Table 53, we find that the access channel capacity in terms of the number of messages that can be processed by an access channel is given by
3600 sec/hour M = ------------------------------------------------------------------------------------------------------------ = 25,714 messages/hour . 3 ( 20 10 sec/frame ) ( 7 frames/message )

(5.5)

We define the utilization of the access channel as

= A / M.
(5.6)

As discussed in Section 5.3.3, with a limit of a ten percent access channel interference margin, the value of value of should be less than sixty percent. In other words, the access traffic cannot exceed sixty percent of the access channel capacity. In Figures 512 and 513, using the numerical values in Table 54 and Table 55, we plot the access utilization as a function of m, the number of SMS messages. From Figure 512, when AR = 30 minutes (one autonomous registration per thirty minutes)

CH5.fm Page 65 Monday, November 22, 1999 10:42 AM

Access Channel Capacity

65

180% 160% 140% Utilization (%) 120% 100% 80% 60% 40% 20% 0% 0 1 2 3 4 5 6 Num ber of SMS/hr/subscriber 1 carrier 2 carriers 3 carriers 4 carriers 5 carriers 6 carriers

Figure 512

Access Channel Utilization with AR = 30 Minutes

160% 140% 120% Utilization (%) 100% 80% 60% 40% 20% 0% 0 1 2 3 4 5 6 Num ber of SMS/hr/subscriber 1 carrier 2 carriers 3 carriers 4 carriers 5 carriers 6 carriers

Figure 513

Access Channel Utilization with AR = 1 Hour

and SMS = 1/subscriber/hour, one access channel can support four CDMA RF carriers. In Figure 513, if we increase the AR time to one hour, then one access channel can support five CDMA RF carriers with SMS = 1/subscriber/hour. Clearly, higher demand on the SMS and the increased busy hour call attempt (BHCA) will reduce the number of CDMA RF carriers that each access channel can support. It can be shown, however, that even in the case of two SMSs and two call attempts per subscriber during the busy hour, one access channel can still support three CDMA RF carriers when AR = 30 minutes.

CH5.fm Page 66 Monday, November 22, 1999 10:42 AM

66

Ch. 5 Mobile Station Access and Paging

We can extend the above results for the 8 Kbps vocoder case to the 13 Kbps vocoder case. In systems using 13 Kbps vocoders, the number of subscribers supported by a single CDMA RF carrier is smaller, thus, users on a single RF carrier will generate less access traffic. Therefore, approximately one access channel of a system using 13 Kbps vocoders can support about one and a half times as many CDMA RF carriers as compared to the systems using 8 Kbps vocoders. Of course, this capacity could vary with different assumptions on the access traffic.

5.4 Paging Channel Capacity

In this section, we discuss an approach to evaluate the capacity of an IS-95A CDMA paging channel. The capacity model developed in this section incorporates most of the salient features of this channel and can be used to quickly answer a variety of questions related to paging channel sizing. We can, for example, calculate the residual capacity available for SMS and voice mail services (VMSs) after accounting for the call-processing load associated with a single paging channel. In a CDMA system, a paging channel conveys information from base stations to mobile stations. There are three major types of call-processing-related messages. The first is an overhead message. It contains information required for call setup (for example, system parameter messages, access parameter messages, neighbor list messages, channel list messages, and extended system parameter messages) and is updated periodically to ensure a successful call setup. The second is a page message (or general page message). The page message is used to page the mobile. The page message is sent when a mobile switching center (MSC) receives a call/service request for a mobile. Depending on the paging strategy, the page messages may be sent to a large area through the paging channel on all sectors. The third type is a channel assignment message and order message. These messages are important for interacting with a mobile to complete a call/service setup. The base station usually sends these messages only to a small area (a few sectors) during the call/service setup. In addition to paging messages related to call processing, there are messages associated with supplemental services such as SMS or VMS, which can be sent through the paging channel as well. In Section 5.4.1, we describe the characteristics of the paging channel that must be considered to formulate a relevant capacity model. In Section 5.4.2, we list all page

CH5.fm Page 67 Monday, November 22, 1999 10:42 AM

Paging Channel Capacity

67

messages and assumptions used for calculating the paging channel occupancy for each page message type. Section 5.4.3 provides the results for the paging channel capacity and the residual capacity for SMS and VMS after accounting for the call-processing load associated with a single paging channel. In Section 5.4.4, we briefly discuss the results and make some recommendations. 5.4.1 5.4.1.1 Paging Channel Characteristics Paging Channel Structure

As specified in IS-95A [5.2], the data rate for the paging channel can be either 4800 bps or 9600 bps. Unless otherwise specified, we will assume that the paging channel data rate is 9600 bps, but the capacity model to be developed will be applicable to both transmission rates. Figure 514 illustrates the paging channel structure. The paging channel is partitioned into 80 msec paging slots, and each slot consists of eight half frames, each of 10 msec duration. Each half frame begins with a synchronized capsule indicator (SCI) bit, and the first new message in a paging slot7 must begin immediately following an SCI bit that is set equal to 1. Paging channel messages are carried in paging channel capsules that consist of the message body, an eight-bit length field that indicates the length in bits of the entire capsule, and a CRC code of thirty bits. As we will describe later, paging messages, channel assignments, and orders have lengths in the 100 to 150 bits range, and since a paging slot consists of 760 bits (eight half frames, each having ninety-five potential payload bits), it can potentially carry multiple pages, channel assignments, and orders. A synchronized paging channel message capsule begins immediately following an SCI bit (in which case the SCI bit is set equal to 1). Since most page, channel assignment, and order message capsules occupy roughly one and a (typically small) fraction of a second half frame, insisting that all message capsules be synchronized would waste a good part of the second half frame. To avoid this inefficient use of a paging channel, the standard permits only the first new message capsules in a slot to be synchronized, and subsequent message capsules in the slot can be appended to the end of the preceding capsule. The message length field in a capsule indicates where the next message capsule in a slot begins. If a message capsule ends less than eight bits from an SCI bit, the standard dictates that the next message in the slot must be synchronized. The length of page and channel assignment capsules must be an integer number of bytes; both of

7.

Paging channel messages may spill over from a previous slot.

CH5.fm Page 68 Monday, November 22, 1999 10:42 AM

163.84 s 163.84 x R bits Maximum Paging Channel Slot Cycle 2048 Slots 80 ms 0.08 x R bits Paging Channel Slot n 8 Paging Channel Half Frames 10 ms 0.01 x R bits Paging Channel Half Frame

Paging Channel Slot 0

...

...

Paging Channel Slot 2047

... ...
SCI

Paging Channel Half Frame Paging Channel Half Frame Body =0

... ... ...

Paging Channel Message Capsule SCI =1

Paging Channel Half Frame

...
SCI =1 Paging Channel Half Frame Body
68

Paging Channel Half Frame Body

... ...
Abutted Messages Unsynchronized Capsules

...
First New Capsule in Slot Synchronized Capsule Paging Channel Message 8 x MSG_LENGTH

Paging Channel Message Capsule

Paging Channel Message Capsule

First New Capsule in Slot Synchronized Capsule Paging Channel Message Padding As required Paging Channel Message 8 x MSG_LENGTH
MSG_LENGTH 8 bits Message Body CRC 30 bits

Padding

As required

Figure 514 Paging Channel Structure Example

Reproduced from [5.1] and [5.2] under written permission of the copyright holder (Telecommunications Industry Association)

CH5.fm Page 69 Monday, November 22, 1999 10:42 AM

Paging Channel Capacity

69

these message types have zero to seven bits reserved for padding to satisfy this requirement. 5.4.1.2 Slotted and Non-Slotted Mode

A mobile may operate in non-slotted mode, in which case the mobile reads all page slots while in the mobile station idle state. Alternatively, a mobile may (choose to) operate in the slotted mode while in the mobile station idle state, in which case the mobile wakes up periodically in specified slots to check for paging messages directed to it. Slotted mode permits the mobile to power down until its pre-specified slot comes along. The mobile wakes up for one or two slots in a slot cycle8 the length of which can be negotiated between the mobile and the system. The minimum length of a slot cycle is sixteen slots (1.28 sec); for normal cellular/PCS service, this minimum length is likely to be widely used. Mobile stations may use longer slot cycles (maximum length permitted is 2,048 slots), but this would lead to significant delays in terminating a call to mobiles operating in the slotted mode. Both the mobile and the MSC use the same hash function (see References [5.1] and [5.2]) with the mobiles MIN (or TMSI) as an argument to determine the slot (in the slot cycle) in which the mobile will next awaken. This permits the cellular system to determine the correct slot in which to page mobiles in slotted mode. Mobiles in slotted mode do not generally read the entire paging channel slot in which they awaken because the system sends a _DONE message after all messages scheduled for the slot have been sent (this is an empty general page message with the appropriate field set to indicate that all messages in the slot have already been sent). This permits mobiles in slotted mode to conserve even more battery power. The requirement that a paging channel message must be contained in two successive slots (the message capsule cannot exceed 1,520 bits) constrains the size of a paging channel message. As indicated earlier, a mobile in slotted mode may read a message that continues onto the next slot; in this instance, the mobile may read the paging channel for as long as two slots. When IS-95A was approved, it defined three distinct page messages: a slotted page message, a page message, and a general page message. From the viewpoint of paging channel usage, there is little to distinguish among these three distinct page message types because they all have approximately the same length.

8.

In cases where a page message carries over to a second slot, a mobile would stay awake for the second slot to read the complete message.

CH5.fm Page 70 Monday, November 22, 1999 10:42 AM

70

Ch. 5 Mobile Station Access and Paging

Table 56

Paging Channel Messages

Message Name System Parameter Message Access Parameter Message Neighbor List Message CDMA Channel List Message Slotted Page Message Page Message Order Message Channel Assignment Message SSD Update Message Data Burst Message Authentication Challenge Message Feature Notification Message Extended System Parameters Message General Neighbor List Message Status Request Message Service Redirection Message General Page Message Global Service Redirection Message TMSI Assignment Message Null Message

5.4.2

Assumptions

Table 56 lists the twenty paging channel messages that the IS-95A (Reference [5.2]) defines. Of the twenty messages, the slotted page message and the page message are no longer in IS-95B (Reference [5.5]). Of the remaining eighteen messages, the following paging channel messages account for most paging channel usage:

CH5.fm Page 71 Monday, November 22, 1999 10:42 AM

Paging Channel Capacity

71

Table 57

Assumptions and Message Lengths

Numerical Values

General Assumptions a. Paging Channel Capacity b. Maximum Allowable Utilization c. Paging Strategy (number of pages per users) d. Termination Rate e. Busy Rate f. BHCA per subscriber g. Number of Sectors per MSC h. General Page Message i. Overhead Message j. System Parameter Message k. Access Parameter Message l. Neighbor List Message m. CDMA Channel List Message n. Extended System Parameter Message o. Channel Assignment Message p. Order Message Voice Mail Service q. Voice Mail Notification Short Message Services r. Data Burst Message (x = number of characters) s. _DONE Message 9600 bits/sec 0.9 1.5 0.35 0.03 2 200 136 bits =j+k+l+m+n 264 bits 184 bits 216 bits 88 bits 112 bits 144 bits 102 bits 720 bits (7x + 380) bits 72 bits

1. 2. 3. 4. 5.

General Page Message, Overhead Message, Channel Assignment Message, Order Message, and Data Burst Message for SMS.

The remaining messages are not constantly used. Table 57 lists all assumptions and message lengths used for the later calculation.

CH5.fm Page 72 Monday, November 22, 1999 10:42 AM

72

Ch. 5 Mobile Station Access and Paging

5.4.3

Paging Channel Capacity

In this section, we use ninety percent paging channel capacity to be the maximum allowable paging channel utilization, which is indicated in Table 57. We reserve the remaining ten percent to accommodate potential burst paging traffic. The paging channel occupancy for each message is then calculated as a fraction of the maximum allowable paging channel capacity. In other words, if the calculated occupancy is twenty percent, it means that the twenty percent of the ninety percent physical paging channel capacity will be needed to transport the message. In the following sections, we will evaluate the paging channel occupancy for each message. This will give us a quick answer to a variety of questions related to paging channel sizing. We will use the BHCA as a traffic reference for calculating paging channel capacity. 5.4.3.1 General Page Message (Call Termination)

Base stations use the general page message to page (find) a mobile when there is a terminating call or to notify the arrival of a special message service (for example, SMS or VMS). In general, the base station will send the message system-wide to locate the mobile. The paging channel occupancy (with respect to ninety percent paging channel utilization) for the general page message (Og) is calculated as9
Og = BHCA (d e ) c h . 3600 a b

(5.7)

From (5.7), we see that the paging channel occupancy for the general page message is linearly proportional to the growth of traffic demand (that is, BHCA). Figure 515 plots the paging channel occupancy for the general page message at different traffic demand. This shows that there is about thirty percent to forty percent occupancy at 150 K to 200 K BHCA. Note that if we can page a small area only (knowing where the mobile is), the reduction of paging channel occupancy can be substantial. For example, if the system pages only one-fourth of the areas, then there is about twenty percent to thirty percent paging channel usage reduction at the range of 150 K to 200 K BHCA traffic. 5.4.3.2 Overhead Message

As described in References [5.1] and [5.2], mobile stations need to update all overhead messages to have a successful call setup. Reducing call setup duration can improve successful call setup rate because the probability of losing a strong pilot is pro9. In equations (5.7) to (5.14), we use item indices in Table 57.

CH5.fm Page 73 Monday, November 22, 1999 10:42 AM

Paging Channel Capacity

73

45% 42.0% 40% 35% 30% Occupancy 25% 20% 15% 10% 5% 0% 50,000 100,000 BHCA 150,000 200,000 10.5% 21.0% 31.5%

Figure 515

Paging Channel OccupancyGeneral Page Message (Call Termination)

portional to the setup duration. In this section, we will evaluate the impacts of different overhead cycles on the paging channel capacity. The overhead message (which includes system parameter messages, access parameter messages, neighbor list messages, channel list messages, and extended system parameter messages) will be sent within every n slots10 to a mobile. The occupancy is independent of the traffic demand. The system will continuously send overhead messages. The paging channel occupancy for the overhead message (Oo) is calculated as
Oo = i [1 (0.08 Slot Cycle)] . ab

(5.8)

Equation (5.8) indicates that the paging channel occupancy for an overhead message is inversely proportional to the slot cycle duration. Figure 516 plots the paging channel occupancy versus varied overhead cycle slots. It shows that the paging channel occupancy grows exponentially with the reduction of the overhead cycle, which suggests that the overhead cycle slots should be greater than seven to avoid using too much

10.

Each slot is 80 msec.

CH5.fm Page 74 Monday, November 22, 1999 10:42 AM

74

Ch. 5 Mobile Station Access and Paging

45% 40% 35% 30% Occupancy 25% 20% 17.86% 15% 10% 5% 0% 3 5 7 9 11 13 15 Number of Slots per Cycle 13.89% 11.36% 9.62% 8.33% 25.00% 41.67%

Figure 516

Paging Channel Occupancy Overhead Message

paging channel capacity. Note that reducing the overhead message cycle from fifteen to seven results in approximately a nine and a half percent increase in paging channel usage. 5.4.3.3 Channel Assignment Message and Order Message

The base station sends channel assignment messages (CAMs) to the mobile during a call setup (either an originating or a terminating call). Order messages are used for registration reject and base station acknowledgments. The base station transmits CAMs and order messages, interacting during the call/service setup, to the mobile through current communicating sectors paging channels. The channel assignment message and order message almost come as a pair for each call setup. Since order messages also include registration reject, it has slightly higher rates. For simplicity, we assume that the rates for a channel assignment message and order message are the same. We also assume that the system (one MSC) has two hundred sectors. The paging channel occupancy for channel assignment and order messages, Oco, is calculated as

CH5.fm Page 75 Monday, November 22, 1999 10:42 AM

Paging Channel Capacity

75

2.5%

2.31% BHCA = 50k

2.0%

BHCA = 100k BHCA = 150k BHCA = 200k 1.85% 1.74%

1.5%

Occupancy

1.39%

1.39% 1.16%

1.0%

1.04% 0.93% 0.69% 0.69% 0.58% 0.46% 0.35% 0.23% 0.46% 0.93%

0.5%

0.0% 1 2 3 4

Number of CAM Repeat

Figure 517 es

Paging Channel OccupancyChannel Assignment and Order Messag-

Oco =

(BHCA g ) (N o + p ) ,
3600 a b

(5.9)

where N is the number of repeats for channel assignment messages. The first item on the right-hand side of (5.9) is a simple calculation for the rate of channel assignment and order messages. Clearly, it will tend to underestimate the areas with higher traffic. Figure 517 shows the paging channel occupancy for channel assignment messages and order messages with varied CAM repeats and different traffic demands (BHCA). The results show that the paging channel occupancy for channel assignment messages and order messages is low. Even with four repeats and 200 K BHCA, the occupancy is only 2.31 percent. 5.4.3.4 _DONE Message and SCI bit

As mentioned in Section 5.4.1.2, the base station sends the _DONE message after all messages scheduled for the slot have been sent. Here, we assume that there is a _DONE message in each slot. The SCI bit is inserted in each half frame. Therefore, the paging channel occupancy for _DONE message and SCI bits (Ods) can be calculated as

CH5.fm Page 76 Monday, November 22, 1999 10:42 AM

76

Ch. 5 Mobile Station Access and Paging

Table 58

Paging Channel Occupancy for _DONE Message and SCI bits

Message _DONE Message (s) SCI Ods Paging Channel Occupancy 10.42% 1.16% 11.58%

Ods =

(1 0.08) s + 1 (1 0.01)
ab

(5.10)

Table 58 lists the results for the paging channel occupancy due to _DONE messages and SCI bits. 5.4.3.5 Paging Channel Capacity (No SMS and No VMS)

Figure 518 plots the total paging channel occupancy (sum of those given by (5.7)-(5.10)) with different BHCA and overhead message cycles of seven and fifteen. Considering 150 K BHCA, the paging channel occupancy is about fifty-three percent for the overhead message cycle of seven and sixty-three percent for the overhead cycle of fifteen. Therefore, there is about thirty-seven percent to forty-seven percent residual capacity available for the services such as VMS and SMS. 5.4.3.6 Voice Mail Services (VMSs)

Voice mail service is a feature that allows callers to leave a voice mail message on a voice mail center. When a caller leaves a voice mail message, the voice mail center will then trigger a request to send a notification to the mobile. There are several ways to send voice mail notification messages. Here, we will introduce two methods of VMS. The first one is simply using a call setup procedure. The system will send a general page message periodically until the mobile responds. After following the regular call setup procedure, the base station will then send the voice mail through a traffic channel. The other method uses the SMS type transmission through the paging channel. In this method, no voice mail indication (VMI) message will be sent until the mobile powers on, registers, or sets up a call. The system will first send a general page message, which is followed by a VMI message (through the currently communicating sectors paging channel) upon receiving the response from the mobile. Let Ov1 denote the paging channel occupancy for the first method and Ov2 denote the paging channel occupancy for the second method. Then,

CH5.fm Page 77 Monday, November 22, 1999 10:42 AM

Paging Channel Capacity

77

80% OH Cycle = 7 slots 70% OH Cycle = 15 slots 62.65% 53.13% 73.72% 64.20%

60% 51.58% 42.05%

50% Occupancy 40.50% 30.98%

40%

30%

20%

10%

0% 50,000 100,000 BHCA 150,000 200,000

Figure 518

Paging Channel Occupancy (No VMS and No SMS)

Ov1 =

[(BHCA f ) NOrate 2 (d e ) (60 v ) + BHCA e] h

3600 a b
Ov2 = BHCA e (h + q/g ) , 3600 a b

(5.11) (5.12)

where v is the VMS cycle duration and NOrate is the NO PAGE RESPONSE rate. Figure 519 shows that, using method 1, the paging channel occupancy will substantially increase when the interval between two general page messages, v, is short. With the assumption of twenty percent no page response rate, there is a potential problem when the traffic demand is greater than 150 K BHCA. In this case, for a fifteenminute page cycle, the paging channel occupancy is close to twenty percent. Most of them are used to notify (page) those mobiles that will never respond. In method 2, with varied number of sectors in each MSC (that is, g), the SMS transmission can substantially reduce the paging channel occupancy from twenty percent to less than three percent as plotted in Figure 520. Note that it is almost insensitive to the number of sectors in one MSC.

CH5.fm Page 78 Monday, November 22, 1999 10:42 AM

78

Ch. 5 Mobile Station Access and Paging

40%

35%

30%

BHCA = 50k BHCA = 100k BHCA = 150k BHCA = 200k

25% Occupancy

20%

15%

10%

5%

0% 10 15 20 25 30 40 45 50 55 60 VMS Cycle Duration (minutes)

Figure 519

Paging Channel Occupancy for VMS (Call Setup Method)

5.0% 4.5% 4.0% 3.5%

BHCA = 50k BHCA = 100k BHCA = 150k BHCA = 200k

3.0% Occupancy 2.5% 2.0%

1.5% 1.0% 0.5%

0.0% 100 200 300 400 500 600 Number of Sectors per MSC

Figure 520

Paging Channel Occupancy for VMS (SMS Method)

CH5.fm Page 79 Monday, November 22, 1999 10:42 AM

Paging Channel Capacity

79

900 800 700

SMS Messages per Sector

OH Cycle = 7 OH Cycle = 15

600 500 400 300 200 100 0 50,000 100,000

BHCA

150,000

200,000

Figure 521

SMS Capacity (No VMS)

5.4.3.7

Data Burst Messages (Short Message Services)

The base station uses data burst messages to send SMS messages to a mobile. In addition to the seven bits sent for each character in the SMS message (the size of an SMS message in bits is denoted by r; each SMS message contains eighty characters), the data burst message capsule carries approximately 380 bits of overhead. (Of these bits, 232 are due to IS-637 [5.6] mandated overhead, 38 bits are due to the length field and the CRC, and the remaining 110 bits are due to fields in the data burst message). If we assume a flood-paging scheme, every CDMA paging channel in the MSC carries an SMS page for each delivered SMS message. In this case, every paging channel carries g M SMS pages/sec. The paging channel occupancy for SMS, Os, can then be calculated as
Os = M (r + h g ) . 3600 a b

(5.13)

Given a residual capacity, the number of SMS messages for each sector can be obtained by

CH5.fm Page 80 Monday, November 22, 1999 10:42 AM

80

Ch. 5 Mobile Station Access and Paging

M =

Os (3600 a b ) . r +hg

(5.14)

Figure 521 shows the residual capacity for SMS messages (without including VMS). Given that each sector supports about 400 subscribers11 for a system using 13 Kbps vocoders, the residual paging channel capacity can support about one SMS message per subscriber (at 150 K BHCA) during the busy hour. 5.4.4 Summary

At 150 K BHCA traffic, there is about a fifty-three to sixty-three percent paging channel occupancy used by messages related to call processing. The residual capacity for SMS is about one SMS message per subscriber during the busy hour. The SMS type transmission should be chosen for VMS to avoid unnecessary waste of paging channel capacity. Addition of new services such as SMS and VMS may have caused paging channel usage to reach its maximum allowable capacity. This situation will be even worse in IS-95B in which the overhead message size is larger and the paging channel loading is higher due to the new access handoff and the soft CAM features. To ensure the appropriate paging channel performance, we need to carefully choose a transmission algorithm for each paging message.

5.5 References
[5.1] [5.2] [5.3] [5.4] ANSI J-STD-008. Mobile Station - Base Station Compatibility Requirements for 1.8 and 2.0 GHz CDMA PCS, March 1995. EIA/TIA IS-95A. Mobile Station Base Station Compatibility Standard for Dual-Mode Wideband Spread Spectrum Cellular System, March 1995. A. J. Viterbi. CDMA: Principles of Spread Spectrum Communication, AddisonWesley Publishing Company, 1995. W.C.Y. Lee. Mobile Cellular Telecommunications Systems, McGraw-Hill, 1989.

11.

For a system using 13 Kbps vocoders, the Erlang traffic is about 7.4. If we assume that each call holding time is seventy-two seconds, the number of subscribers for each sector is 7.4/0.02 = 370.

CH5.fm Page 81 Monday, November 22, 1999 10:42 AM

References

81

[5.5]

TIA/EIA/SP-3693-1 (to be published as TIA/EIA-95-B). Mobile Station Base Station Compatibility Standard for Dual-Mode Wideband Spread Spectrum Cellular System, July 17,1998. TIA/EIA/IS-637. Short Message Services for Wideband Spread Spectrum Cellular Systems.

[5.6]

CH5.fm Page 82 Monday, November 22, 1999 10:42 AM