US 6,278,701 B1Grant

Capacity enhancement for multi-code CDMA with integrated services through quality of services and admission control

Issue Date:2001-08-21

•10 Claims

•10 Drawing Sheets

Abstract

With the addition of high-speed data traffic to traditional CDMA cellular networks, there is a need to efficiently utilize system capacity so that the quality of service of existing voice and low-speed data users is maintained while new high-speed data users are added to the network. Methods and systems are presented that control allocation of power to users, quality of service requirements, and/or user activity levels to enhance capacity utilization. These methods and systems are based on a method for estimating the capacity of a CDMA carrier with both voice and data users using an interference-based analysis of the reverse link. In particular, the methods enhance capacity utilization in a multi-code CDMA network architecture, in which several codes are allocated to a single high-speed data user for parallel transmission.

Metadata

Assignees

Verizon Laboratories Inc.
University of Maryland

Inventors

Deepak Ayyagari
Samuel Resheff
Anthony Ephremides

Application Information

Application Number:US 09/113,551

Filing Date:1998-07-10

Priority Date:1998-07-10

Art Unit:7

Classifications

IPC:

H04B 7216

Field of Search:

370335370336370341370342370431370441370468370320370318370310455522455446455403455422455517455 69455 88

Patent Drawings (10 sheets)

Description

BACKGROUND OF THE INVENTION

[0002] The present invention relates to cellular networks and, more particularly, to methods of improving capacity utilization of a cellular carrier with integrated voice and data traffic through methods involving cellular network admission control, power control, and quality of service control.

[0003] Until recently, cellular telephone networks provided service for voice users only and did not address high-speed data applications such as electronic mail, facsimile, image transmission, and FTP transfer. Currently, network engineers are designing cellular networks, in particular Code Division Multiple Access (CDMA) networks, to carry integrated voice and data traffic. For example, the Telecommunications Industry Association (TIA) TR-45.5 cellular CDMA standard committee is developing CDMA networks that can support high-speed data (HSD) along with conventional voice telephony, described in “TR-45.5 Working Committee for CDMA Phase 1 High Speed Data (HSD) Stage 1 Services Description for HSD Services,” October, 1996, which is hereby incorporated by reference. The systems proposed by TR-45.5, for both the 1.25 MHz CDMA channel and the broadband 5 MHz channel, achieve high data rates by implementing a multi-code architecture, which allocates several codes to a single HSD user for parallel transmission, in contrast with a variable gain CDMA system where HSD users use a single code to transmit at different data rates. This allows the user using m codes to transmit at a multiple rate of m times the nominal rate of a single code, e.g., 9.6 kbps. Such a multi-code architecture was proposed in I. Chih-Lin and R. Gitlin, “Multi-Code CDMA Wireless Personal Communications Networks,” ICC '95 Conference Record, pp. 1060-1064, June, 1995, and further described in I. Chih-Lin, G. Pollini, L. Ozarow, and R. Gitlin, “Performance of multi-code CDMA wireless personal communications network, ” Proceedings of IEEE Vehicular Technology Conference, pp. 907-911, 1995, which are hereby incorporated by reference.

[0004] A major issue in the design of a CDMA system with both voice and data users is the efficient use of system capacity. Capacity can generally be defined as the number of voice users and data users at different transmission rates (also referred to herein as different data classes) that can be supported by a CDMA channel with specified Quality of Service (QoS) attributes, specifically frame error rate (FER) and service availability probability (SAP), where the SAP is the probability that the FER does not exceed a predefined threshold. The QoS for a CDMA user is directly related to the ratio of the received bit energy to interference power. In conventional CDMA systems that carry only voice traffic, the FER and SAP targets are identical for all users, and therefore system capacity is maximized when all voice frames are received at the same power level. In a multi-class system, however, different users have different FER and SAP targets, and the relationship between capacity and QoS is more difficult to analyze for two reasons. First, the actual QoS experienced by a voice or HSD user depends on the activity of interfering users and the tightness of the algorithm controlling received power. Second, QoS targets are often “soft” for certain classes of data users (e.g., Internet access), and these users can accept a lower QoS in exchange for enhanced capacity.

[0005] Once the intricate relationship between capacity and parameters, such as QoS of users, the mix of voice and data users, their activity level, and their power levels, is established, control methods can be developed that control power allocation, QoS levels, and/or the activity levels of existing users to give additional users access to the system, thereby enhancing capacity utilization.

[0006] One previous effort to model CDMA network capacity is K. S. Guihousen, I. M. Jacobs, R. Padovani, L. A. Weaver and C. A. Wheatley, “On the capacity of a cellular CDMA system,” IEEE Transactions on Vehicular Technology, vol. 40, pp. 301-312, May, 1991. This interference-based model addresses voice-only systems and assumes perfect power control and a uniform distribution of users in the cell. It has been shown that the interference resulting from HSD mobile users located close to cell boundaries and transmitting at a multiple of the basic rate is much higher than the interference generated by a uniform distribution of an equivalent number of mobiles transmitting at the basic rate (Z. Liu, M. J. Karol, M. E. Zarki, and K. Eng, “Interference issues in multi-code CDMA networks,” PIMRC1996, pp. 98-102, October, 1996).

[0007] It has also been shown that capacity limitations result from imperfect power control (e.g., A. J. Viterbi and A. M. Viterbi, “Erlang capacity of a power controlled CDMA system,” IEEE Journal on Selected Areas in Communications, vol. 11, pp. 892-899, August, 1993; R. Cameron and B. Woemer, “Performance analysis of CDMA with imperfect power control,” IEEE Transactions on Communication Theory, vol. 44, pp. 777-781, July, 1996; F. D. Priscoli and F. Sestini, “Effects of imperfect power control and user mobility on a CDMA cellular network,” IEEE Journal of Selected Areas in Communication, vol. 14, pp. 1809-1817, December, 1996). One model with imperfect power control was studied in N. B. Mandayam, J. Holtzman, and S. Barberis, “Erlang capacity for an integrated voice/data DS-CDMA wireless system with variable bit rate source,” Proceedings of PIMRC, vol. 3, pp. 1078-1082, 1995. This model, however, addresses only single code data users, and does not analyze a multi-code CDMA architecture.

[0008] The capacity of a CDMA system is also limited by the coverage area of handsets, which depends on the limitations on handset power. In a CDMA system based on the IS-95A standard, voice handsets have a peak output power of 200 mW. An HSD handset with identical power output transmitting at four times the basic rate outputs a maximum of 50 mW per code/channel. A proper analysis of CDMA system capacity should also consider the possible received power limitations for voice and data users.

[0009] It is desirable, therefore, to provide a method for detennining the capacity of a CDMA system with a mix of voice and multi-code high-speed data users with varying QoS requirements. It is more desirable to provide such a method that considers capacity limitations resulting from imperfect power control. It is also desirable to provide such a method that also considers handset power limitations. Having established such capacity estimation methods, it is ultimately desirable to provide methods for admission control to a CDMA network based on the determined relationship between system capacity and QoS, power allocation, and user activity level, thereby enhancing capacity utilization.

SUMMARY OF THE INVENTION

[0010] Methods and systems consistent with the present invention address these needs by controlling admission of new users to the CDMA system through control of power allocation, QoS requirements, and/or the activity level of existing users. Such methods and systems are predicated on methods and systems consistent with the present invention that estimate the capacity of a single CDMA carrier with a mixture of voice and HSD users through an interference-based analysis of the reverse link (mobile to base station).

[0011] A method consistent with the present invention for determining the capacity of a CDMA cellular carrier supporting voice and multi-code data users, wherein each user belongs to a user class, and there may be multiple data user classes, comprises the steps of determining a received power level for each user class, and determining a number of users in each user class that satisfies a quality of service requirement based on the received power levels and the activity variables for all user classes.

[0012] A method consistent with the present invention for enhancing the capacity of a CDMA cellular carrier supporting voice and multi-code data users comprises the steps of setting a quality of service requirement for the data users based on the traffic load and the quality of service requirement for the voice users, adjusting the received power level of the voice users to be equal, adjusting the received power level of the data users to be equal to that of the voice users, decreasing the received power level of the data users until their quality of service requirement is satisfied, and adjusting the quality of service requirement for the data users in response to a request from a new data user to access the cellular carrier.

[0013] Another method consistent with the present invention for enhancing the capacity of a CDMA cellular carrier supporting voice and multi-code data users comprises the steps of determining a measure of traffic from voice users, determining a measure of traffic from data users, assigning each data user the same number of codes, adjusting the activity factor for each data user based on the traffic measures and the quality of service requirement, and controlling the activity level of the data users using the adjusted activity factor.

[0014] Yet another method consistent with the present invention for enhancing the capacity of a CDMA cellular carrier supporting voice and multi-code data users comprises the steps of assigning a feasible power level to the users, determining and assigning a minimum power level to the users, determining the maximum received power level, determining the ratio of the maximum received power level to the minimum power level of each user, selecting the smallest ratio, scaling the minimum power level of each user by the smallest ratio, and assigning the scaled power levels to the users.

[0015] Systems are also provided for carrying out methodologies consistent with the present invention.

[0016] The above desires, and other desires, features, and advantages of the present invention will be readily appreciated by one of ordinary skill in the art from the following detailed description of the preferred implementations when taken in connection with the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

[0017] FIG. 1 is a high-level block diagram of a CDMA cellular network consistent with the present invention;

[0018] FIG. 2 is a block diagram of the Mobile Switching Center and base station in FIG. 1;

[0019] FIG. 3 illustrates cell sector geometry consistent with the present invention;

[0020] FIG. 4 illustrates the results of a method for determining cellular network capacity consistent with the present invention;

[0021] FIG. 5 illustrates the results of a second method for determining cellular network capacity consistent with the present invention;

[0022] FIG. 6 illustrates another set of results of the second method for determining cellular network capacity consistent with the present invention;

[0023] FIG. 7 is a flowchart illustrating the steps of a method for enhancing cellular network capacity consistent with the present invention;

[0024] FIGS. 8A-8B are flowcharts illustrating the steps of a second method for enhancing cellular network capacity consistent with the present invention; and

[0025] FIGS. 9A-9B are flowcharts illustrating the steps of a third method for enhancing cellular network capacity consistent with the present invention.

DETAILED DESCRIPTION

[0026] The following detailed description of the invention refers to the accompanying drawings. The same reference numbers in different drawings identify the same or similar elements. Also, the following detailed description does not limit the invention. Instead, the scope of the invention is defined by the appended claims and their equivalents.

[0027] Methods and systems consistent with the present invention provide a methodology for analyzing system capacity for CDMA networks with both voice users and multi-code data users. The methodology uses an interference-based model for the reverse link (mobile to base station) that determines the number of voice users N_vand the number of data users N_d,ibelonging to each of C data classes i that a CDMA carrier can support. Methods and systems consistent with the present invention also provide admission control schemes that maximize system capacity by controlling power allocation, QoS levels, and/or user activity levels.

[0028] FIG. 1 is a system level block diagram illustrating portions of a CDMA cellular network consistent with the present invention. Mobile Switching Center (MSC) 10 is a cellular telephone network switch that provides processing and switching functions to allow mobile units to communicate with other telephones or devices (cellular and wireline). MSC 10 is connected to one or more base stations (also referred to as cell sites), one of which, base station 20, is shown by way of example. Base station 20 provides radio communication with mobile units. Consistent with the present invention, a CDMA cellular network includes both voice users and multi-code data users, which transmit high-speed data using two or more codes in parallel. By way of example, FIG. 1 shows voice mobile units 30 and data mobile units, each of which belongs to one of C classes of data users. For example, data mobile units 40 belong to class 1, and data mobile units 50 belong to class C. As will be described below, each data user class is defined by a transmission rate and may have QoS requirements independent of those for the voice users or other data user classes.

[0029] As shown in FIG. 2, MSC 10 includes memory 12 and processor 14, and base station 20 includes memory 22 and processor 24. Methods for admission control and capacity analysis consistent with the present invention are stored as software in memory 12, memory 22, or both and can be executed by processor 14 and/or processor 24. Memory 12 and memory 22 may be implemented with any type of computer-readable medium, such as any electronic, magnetic, or optical read/write storage device. If necessary, signaling for implementation of admission control methods consistent with the present invention may occur via the communication link between MSC 10 and base station 20 or via a signaling link (not shown).

CDMA Carrier Capacity

[0030] Methods and systems consistent with the present invention analyze the capacity of a single CDMA carrier with voice users and C classes of data users, where C is an arbitrary number, as shown in FIG. 1. The C classes are defined by the transmission rate requirement. If B is the basic transmission rate of a single code (e.g., 9.6 kbps for IS-95A), data users of class i have a transmission rate of r_d,i*B. Thus, r_d,irepresents the number of codes a data user in class i uses for parallel multi-code transmission. The voice users use a single code, i.e., r_v=1. The suffix d herein denotes all variables related to data users, and the suffix v denotes the variables corresponding to voice users. Each code used by a data user of class i experiences interference from the voice users, other data users and the user's own (r_i−1) codes (self-interference). Chih-Lin et al. describe a coding method referred to as “sub-code concatenation” whereby all the codes of a single user can be made orthogonal to each other, thus eliminating self-interference. However, in some systems, including IS-95 systems, the codes arc non-orthogonal and necessarily contribute to self-interference.

[0031] The capacity of the reverse link (i.e., transmission from the mobile unit to the base station) is defined as the vector {N_v, N_d}, N_d=[N_d,1, N_d,2, . . . , N_d,C], the number of voice users N_v, and the number of data users of class i, N_d,i, that can be supported on a single CDMA carrier so that their FER and SAP requirements are met. The Service Availability Probability is defined as SAP≡Prob(FER≦Threshold). The SAP for voice and the C classes of data users is determined as a function of the capacity (N_v,N_d).

[0032] A model for analyzing carrier capacity consistent with the present invention expands on the voice-only model in Gilhousen et al., which is hereby incorporated by reference, to account for integration of voice and multi-code data users. The model makes the following assumptions:

[0033] (1) The model assumes a simple correlation receiver. The performance is determined in terms of

\frac{E_{b}}{N_{0}},

[0034] the ratio of bit energy to the noise power (N₀represents thermal noise plus interference). The interference experienced by a CDMA user depends on the correlation properties of the spreading codes. However, the total interference is modeled as additive noise without considering the special correlation properties. This is an accepted approximation since the spreading codes on the reverse link form a non-orthogonal set. Mobility further destroys possible synchronization among the various coding chips.

[0035] (2) A log-normal attenuation model is used to define shadowing or slow fading. The path loss between the subscriber and the cell site is proportional to 10^ξ/10d⁻⁴, where d is the distance from the subscriber to the cell site and ξ is a Gaussian random variable with standard deviation σξ=8 db and zero mean. The model assumes that Rayleigh fading errors (due to multipath propagation) are accounted for by bit-interleaving, error correcting codes, and multi-antenna diversity. The effects of Rayleigh fading are subsumed in the E_b/N₀requirement.

[0036] (3) The received power levels of the voice users and data users (S_vand S_d,i, ∀i=1, . . . C, respectively) in a fully loaded system with maximally shrunk cells (no cell-breathing) are allowed to differ. As is known in the art, different received power levels cause the problem of capture due to the near-far effect in a packet radio scenario, in which the receiver despreads the strongest frame received. However, in a system such as IS-95, once a code has been assigned to a user, the receiver despreads the user signal with the designated spreading code alone.

[0037] (4) The activity variables, i.e., the proportion of time that a user is active, are referred to herein as ψ_vfor voice users and ψ_d,ifor class i data users and are modeled as binomially distributed random variables with P(ψ_v=1)=α_vand P(ψ_d,i=1)=α_d,i, ∀i=1 . . . C.

[0038] (5) Processing gains for voice and data users is given by

\frac{W}{B},

[0039] where W is the spread spectrum bandwidth and B is the nominal transmission rate for a single code, e.g., 9.6 kbps. The processing gain for all codes is assumed to be equal.

[0040] (6) The sectorization in all cells is perfect.

Perfect Power Control and Uniform User Distribution

[0041] In a method for analyzing CDMA capacity consistent with the present invention, the users are assumed to be uniformly distributed over the area shown in FIG. 3. The method also assumes perfect power control. Because the FER is directly related to the received power ratio Eb/N₀, the SAP can be computed by quantifying the statistical variations in Eb/N₀. The interference power is comprised of thennal noise power η, total interference due to voice users in other cells I_v, total interference due to data users in other cells I_d,i, i=1 . . . C, and interference from voice and data users within the reference cell. The Eb/N₀ratio for class i data users is:

\begin{matrix} \begin{matrix} {(\frac{E_{b}}{N_{0}})}_{d, i} = (\frac{W}{B}) / (\sum_{j = 1}^{N_{v}} ψ_{v, j} \frac{S_{v}}{S_{d, i}} + \sum_{k = 1, k \neq i}^{C} \sum_{j = 1}^{N_{d, k}} ψ_{d, k, j} r_{d, k} \frac{S_{d, k}}{S_{d, i}} + \\ \sum_{j = 1}^{N_{d, i}} ψ_{d, i, j} r_{d, i} + (r_{d, i} - 1) + \sum_{k = 1}^{C} \frac{I_{d, k}}{S_{d, i}} + \frac{I_{v}}{S_{d, i}} + \frac{η}{S_{d, i}}) \end{matrix} & (1) \end{matrix}

[0042] The first term in the denominator represents interference from voice users in the reference cell. The second term represents interference due to data users within the reference cell. The third term represents self-interference. The last term includes data and voice interference from other cells and the thermal noise power. Similarly, the Eb/N₀ratio for voice users is:

\begin{matrix} \begin{matrix} {(\frac{E_{b}}{N_{0}})}_{v} = (\frac{W}{B}) 1 / (\sum_{j = 1}^{N_{v - 1}} ψ_{v, j} + \sum_{k = 1}^{C} \sum_{j = 1}^{N_{d, k}} ψ_{d, k, j} r_{d, k} \frac{S_{d, k}}{S_{v}} + \\ \sum_{k = 1}^{C} \frac{I_{d, k}}{S_{d, i}} + \frac{I_{v}}{S_{v}} + \frac{η}{S_{v}}), \end{matrix} & (2) \end{matrix}

[0043] which is similar to Eb/N₀for data users, but without self-interference.

[0044] All quantities in equations (1) and (2) are deterministic, except the interference from other cells I_d/S and I_v/S. From the log-normal assumption and assuming that path loss varies as the inverse fourth power of distance, d, the interference from other cells can be expressed as

\begin{matrix} \frac{I_{d, i}}{S_{d, i}} (d_{0}, d_{m}) = {r_{d, i} (\frac{d_{m}}{d_{0}})}^{4} \times 10 \frac{ξ_{d, 0} - ξ_{d, m}}{10} & (3) \\ \frac{I_{v}}{S_{v}} (d_{0}, d_{m}) = {(\frac{d_{m}}{d_{0}})}^{4} \times 10 \frac{ξ_{d, 0} - ξ_{d, m}}{10} & (4) \end{matrix}

[0045] where d₀is the distance from a mobile unit in the interfering cell to the base station in the reference cell, and d_mis the distance from a mobile unit in the interfering cell to the base station in the interfering cell. ξ_d,0, ξ_v,0and ξ_d,m, ξ_v,mare independent so that the differences have zero mean and variance 2σ_ξ². The densities of the interfering codes are given by: ρ_v=2N_v/{square root over (3)} voice codes per unit area and ρ_d,i=2N_d,ir_d,i/{square root over (3)} data codes of class i per unit area. The total interference ratio due to other cells can now be expressed as:

\begin{matrix} \begin{matrix} \frac{I_{d, i}}{S_{d, i}} = r_{d, i} \int \int_{Area} ψ_{d, i} ({(\frac{d_{m}}{d_{0}})}^{4} (10 \frac{ξ_{d, 0} - ξ_{d, m}}{10} φ) ξ_{d, 0} - \\ ξ_{d, m,} \frac{d_{0}}{r_{m}})) ρ_{d, i} ⅆ A \end{matrix} & (5) \\ \begin{matrix} \frac{I_{v}}{S_{v}} = \int \int_{Area} ψ_{v} ({(\frac{d_{m}}{d_{0}})}^{4} (10 \frac{ξ_{v, 0} - ξ_{v, m}}{10} φ) ξ_{v, 0} - \\ ξ_{d, m,} \frac{d_{0}}{r_{m}})) ρ_{v} ⅆ A \end{matrix} & (6) \end{matrix}

\begin{matrix} \begin{matrix} where φ (ξ_{0} - ξ_{m,} \frac{d_{0}}{d_{m}}) = 1 if {(\frac{d_{m}}{d_{0}})}^{4} \times (10 ξ_{0} - \frac{ξ_{m}}{10}) \leq 1 \\ = 0 otherwise \end{matrix} & (7) \end{matrix}

[0046] where, d_m=min_k≠0d_k, is the shortest distance between base station k and a mobile unit. The area of integration is the sector shown in FIG. 3.

[0047] As described in Gilhousen, the moments of the other cell interference random variables defined in equations (4) and (5) are:

\begin{matrix} \begin{matrix} E (\frac{I_{d, i}}{S_{d, i}}) = \int \int_{area} {(\frac{d_{m}}{d_{0}})}^{4} E [ψ_{d, i} 10^{χ_{d} / 10} φ (χ_{d}, \frac{d_{0}}{d_{m}})] ρ_{d, i} ⅆ A \\ = \int \int_{area} {α_{d, i} (\frac{d_{m}}{d_{0}})}^{4} f (\frac{d_{m}}{d_{0}}) ρ_{d, i} ⅆ A \end{matrix} & (8) \\ \begin{matrix} E (\frac{I_{v}}{S_{v}}) = \int \int_{area} {(\frac{d_{m}}{d_{0}})}^{4} E [ψ_{v} 10^{χ_{v} / 10} φ (χ_{v}, \frac{d_{0}}{d_{m}})] ρ_{v} ⅆ A \\ = \int \int_{area} {α_{v} (\frac{d_{m}}{d_{0}})}^{4} f (\frac{d_{m}}{d_{0}}) ρ_{v} ⅆ A \end{matrix} & (9) \end{matrix}

\begin{matrix} \begin{matrix} where f (\frac{d_{m}}{d_{0}}) = \exp [{(σξ \ln \frac{10}{10})}^{2}] [1 - Q [40 \log_{10 γ0} / \\ {\frac{r_{m}}{\sqrt{2 σ_{ξ}^{2}}} - \sqrt{2 σ_{ξ}^{2}} \ln \frac{10}{10}]}^{-} . \end{matrix} & (10) \end{matrix}

[0048] where χ_d,i=ξ_d,0−ξ_d,mand χ_v=ξ_v,0−ξ_v,mare Gaussian random variables with zero mean and 2σ_ξ²variance, with φ(χ,d₀/d_m) defined by equation (7). Q[.] is the complementary error function.

[0049] The integrals in equations (8) and (9) are over the 2-D area shown in FIG. 3. As described in Gilhousen et al., the integrals may be numerically evaluated for σ_ξ=8 db. The resulting bounds for α_v=⅜, which has been shown by experiments and field trials to be a suitable value are:

\begin{matrix} E [\frac{I_{v}}{S_{v}}] \leq 0.247 N_{v} ≅ m_{v}; E [\frac{I_{v}}{S_{d, i}}] \leq \frac{S_{v}}{S_{d, i}} \times m_{v} E [\frac{I_{d, i}}{S_{d}}] \leq \frac{0.247}{α_{v} / α_{d, i}} N_{d, i} r_{d, i} ≅ m_{d, i}; E [\frac{I_{d, i}}{S_{v}}] \leq \frac{S_{d, i}}{S_{v}} \times m_{d, i} & (11) \end{matrix}

[0050] The calculations of second moments of I_d,i/S_d,iand I_v/S_vare similar to the calculation described in Gilhousen:

\begin{matrix} var (\frac{I_{v}}{S_{v}}) \leq \int \int_{Area} {(\frac{d_{m}}{d_{0}})}^{8} [α_{v} \times g (\frac{d_{m}}{d_{0}}) - α_{v}^{2} \times f^{2} (\frac{d_{m}}{d_{0}})] ρ_{v} ⅆ A & (12) \\ var (\frac{I_{d, i}}{S_{d, i}}) \leq \int \int_{Area} {(\frac{d_{m}}{d_{0}})}^{8} [α_{d, i} \times g (\frac{d_{m}}{d_{0}}) - α_{d, i}^{2} \times f^{2} (\frac{d_{m}}{d_{0}})] ρ_{v} ⅆ A & (13) \end{matrix}

\begin{matrix} where, g (\frac{r_{m}}{r_{0}}) = \exp [{(σ_{ξ} \ln \frac{10}{5})}^{2}] [1 - Q [\frac{40 \log_{10} r_{0} / r_{m}}{\sqrt{2 σ_{ξ}^{2}}} - \sqrt{2 σ_{ξ}^{2}} \ln \frac{10}{5}]] & (14) \end{matrix}

[0051] Evaluating numerically over the area for σ=8 db, bounds may be computed for the variances:

\begin{matrix} var (\frac{I_{v}}{S_{v}}) \leq 0.078 N_{v} ≅ σ_{v}^{2}; var [\frac{I_{v}}{S_{d, i}}] \leq \frac{S_{v}^{2}}{S_{d, i}} \times σ_{v}^{2} var [\frac{I_{d, i}}{S_{d, i}}] \leq \frac{0.078}{a_{v} / a_{d, i}} N_{d, i} r_{d, i} ≅ σ 2_{d, i}; var [\frac{I_{d, i}}{S_{v}}] \leq \frac{S_{d, i}^{2}}{S_{v}^{2}} \times σ_{d, i}^{2} & (15) \end{matrix}

[0052] Note that the bound on variance for interference due to data variables is valid for α(d, i)≧α_v. Since I_d,i/S_d,iand I_v/S_vare linear functionals of a 2-D white random process, they are well modeled by a Gaussian random variable with means and variances bounded by equations (11) and (15).

[0053] The SAP for voice and data is defined as follows:

\begin{matrix} \begin{matrix} {SAP}_{v} [γ_{v}] ≅ P_{v} = Prob [{FER}_{v} \leq Threshold] \\ = Prob [{(\frac{E_{b}}{N_{0}})}_{v} \geq γ_{v}] \end{matrix} & (16) \end{matrix}

\begin{matrix} \begin{matrix} {SAP}_{d, i} [γ_{d, i}] ≅ P_{d, i} = Prob [{FER}_{d, i} \leq Threshold] \\ = Prob [{(\frac{E_{b}}{N_{0}})}_{d, i} \geq γ_{d, i}] \end{matrix} & (17) \end{matrix}

[0054] or,

\begin{matrix} 1 - P_{v} = Prob [\sum_{j = 1}^{N_{v} - 1} ψ_{v, j} + \sum_{k = 1}^{C} \sum_{j = 1}^{N_{d, k}} ψ_{d, k, j} r_{d, k} \frac{S_{d, k}}{S_{v}} + \sum_{k = 1}^{C} \frac{I_{d, k}}{S_{d, i}} > δ_{v}] & (18) \end{matrix}

\begin{matrix} where δ_{v} = \frac{W / B}{γ_{v}} - \frac{η}{S_{v}} & (19) \end{matrix}

\begin{matrix} \begin{matrix} 1 - P_{d} = Prob [\sum_{j = 1}^{N_{v}} ψ_{v, j} \frac{S_{v}}{S_{d, i}} + \sum_{k = 1, k \neq i}^{C} \sum_{j = 1}^{N_{d, k}} ψ_{d, k, j} r_{d, k} \frac{S_{d, k}}{S_{d, i}} + \\ \sum_{j = 1}^{N_{d, i}} ψ_{d, i, j} r_{d, i} + \sum_{k = 1}^{C} \frac{I_{d, k}}{S_{d, i}} + \frac{I_{v}}{S_{d, i}} + \frac{η}{S_{d, i}} > δ_{d}] \end{matrix} & (20) \end{matrix}

\begin{matrix} where δ_{d, i} = \frac{W / R_{d}}{Υ_{d, i}} - \frac{η}{S_{d, i}} - (r_{d, i} - 1) & (21) \end{matrix}

[0055] Since ψ_d,iand ψ_vare binomially distributed, and

\sum_{i - 1}^{C} \frac{I_{d, i} + I_{v}}{S}

[0056] is the sum of independent Gaussian random variables, equations (18) and (20) can be written as:

\begin{matrix} \begin{matrix} 1 - P_{v} = \sum_{k_{d, C} = 0}^{N_{d, C}} \dots \sum_{k_{d, 1} = 0}^{N_{d, 1}} \sum_{k_{v} = 0}^{N_{v} - 1} \prod_{j = 1}^{C} F_{d, j} (k_{d, j}, N_{d, j}) F_{v} (k_{v}, N_{v} - 1) \times \\ Q [\frac{δ_{v} - k_{v} - \sum_{j = 1}^{C} k_{d, j} \frac{S_{d, j}}{S_{v}} - (m_{v} + \sum_{j = 1}^{C} m_{d, j} \times \frac{S_{d, j}}{S_{v}})}{\sqrt{σ_{v}^{2} + \sum_{f = 1}^{C} \frac{S_{d, j}^{2}}{S_{v}^{2}} \times σ_{d, j}^{2}}}] \end{matrix} & (22) \\ \begin{matrix} 1 - P_{d, i} = \sum_{k_{d, C} = 0}^{N_{d, C}} \dots \sum_{k_{d, 1} = 0}^{N_{d, i} - 1} \sum_{k_{v} = 0}^{N_{v}} \prod_{j = 1, j \neq i}^{C} F_{d, j} (k_{d, j}, N_{d, j}) F_{v} (k_{v}, N_{v}) \times \\ F_{d, i} (k_{d, i} N_{d, i} - 1) \times \\ Q [\frac{δ_{d, i} - k_{v} \frac{S_{v}}{S_{d, i}} - \sum_{j = 1}^{C} k_{d, j} \frac{S_{d, j}}{S_{d, i}} - (m_{v} \frac{S_{v}}{S_{d, i}} + \sum_{j = 1}^{C} m_{d, j} \times \frac{S_{d, j}}{S_{d, i}})}{\sqrt{σ_{v}^{2} + \sum_{f = 1}^{C} \frac{S_{d, j}^{2}}{S_{v}^{2}} \times σ_{d, j}^{2}}}] \end{matrix} & (23) \end{matrix}

where F_{d, i} (a, b) = \frac{b!}{(a!) (b!)} {α_{d, i}^{a} (1 - α_{d, i})}^{b - a} .

[0057] In methods consistent with the present invention, equations (22) and (23) can be used in detennining the capacity (N_v,N_d). As will be described below in the Coverage and Received Power Levels section, the received power level settings S_v, and S_d,ican be determined based on a simple link budget analysis that accounts for coverage limitations resulting from limited handset transmission powers. After determining the received power level settings, the SAP measures in equations (22) and (23) depend only on the number of voice and data users in the system.

Imperfect Power Control and Distribution of Users at Sector Edges

[0058] In another method for analyzing CDMA capacity consistent with the present invention, the effects of imperfect power control and a non-uniform user distribution are considered. As shown in Viterbi et al., Cameron et al., and Priscoli et al., variations in the received power levels in multi-code CDMA systems, due to imperfections in the control loop, are well modeled as log-normal random variables. Therefore, if S is the required received power level for a voice user or a HSD user (from any of C classes of HSD users), the actual received power is given by S×e^βX, where β=ln(10)/10 and X is a Gaussian random variable with zero mean and variance σ²_X. The mean of the log-normal random variable e^βXis defined to be m₁and the variance σ₁².

[0059] The method assumes a user distribution with all voice and data users located along the edges of the sector shown in FIG. 3. The number of users of each class is assumed to be the same in every sector. The method also assumes a fixed base station assignment for each user. Since the distance of the mobiles located at the boundaries from the interfering base station and the reference base station are the same, the received power from the interfering mobiles at the reference base station is assumed to be distributed identically to the power received at the interfering base.

[0060] The SAP equation for the class of voice users or any of the C classes of data users is:

\begin{matrix} 1 - P = Prob [\frac{\frac{W}{B} ⅇ^{(β X)} S}{I_{in} + I_{out} + η} < γ] & (24) \end{matrix}

[0061] where, I_in, is the in-cell interference, I_outis the out-of-cell interference, and η is the thermal noise power. Note that I_in, depends on the class of the user, but I_outis the same for all classes of users. I_inand I_outare random variables. Defining

g = \frac{γ}{W / B}

[0062] and A=I_in+I_out+η, equation (24) can be written as:

\begin{matrix} \begin{matrix} 1 - P = Prob [X < \frac{1}{β} \ln (\frac{g}{S} A)] \\ = 1 - \int_{0}^{\infty} Q [\frac{\frac{1}{β} \ln (\frac{g}{s} a)}{\sqrt{σ_{X}^{2}}}] f_{A} (a) ⅆ A \end{matrix} & (25) \end{matrix}

[0063] It has been shown that if θ is a random variable and P(θ) is a real function of θ, then the mean E[P(θ)] is well approximated by E(P(θ)]≈⅔P(μ)+⅙P(μ+{square root over (3+L )}σ)+⅙P(μ+{square root over (3+L )}σ) (J. M. Holtzman, “A simple, accurate method to calculate spread spectrum error probabilities,” IEEE Transactions on Communications, vol. 40, pp. 461-464, March, 1992, which is hereby incorporated by reference). This approximation is fairly robust and holds under very general conditions. The approximation is exact when the function is a fifth degree polynomial and the random variable A is Gaussian. Using this approximation on the Q function in equation (26):

\begin{matrix} \begin{matrix} 1 - P \approx 1 - \frac{2}{3} Q [\frac{\frac{1}{β} \ln (\frac{g}{S} μ_{A})}{\sqrt{σ_{X}^{2}}} - \frac{1}{6} Q \frac{1}{β} \ln \frac{(\frac{g}{S} (μ_{A} + \sqrt{3} σ_{A}))}{\sqrt{σ_{X}^{2}}}] - \\ \frac{1}{6} Q [\frac{\frac{1}{β} \ln (\frac{g}{S} μ_{A} - \sqrt{3} σ_{A}))}{\sqrt{σ_{X}^{2}}}] \end{matrix} & (27) \end{matrix}

where, μ_A=E[I_in]+E[I_out]+η (28)

[0064] σ_A²=Var[I_in]+Var[I_out] (29)

[0065] The in-cell interference for voice users I_v,inand for class i data users I_d,inis given by:

\begin{matrix} I_{v, in} = \sum_{j = 1}^{N_{v} - 1} ψ_{v, j} S_{v} ⅇ^{β X_{v, j}} + \sum_{k = 1}^{C} \sum_{j = 1}^{N_{d, k}} ψ_{d, k, j} r_{d, k} S_{d, k} ⅇ^{β X_{d, k, j}} & (30) \\ \begin{matrix} I_{d, in} = \sum_{j = 1}^{N_{v}} ψ_{v, j} S_{v} ⅇ^{β X_{v, j}} + \sum_{k = 1, k \neq i}^{C} \sum_{j = 1}^{N_{d, k}} ψ_{d, k, j} r_{d, k} S_{d, k} ⅇ^{β X_{d, k, j}} + \\ \sum_{j = 1}^{N_{d, i} - 1} ψ_{d, i, j} r_{d, i} S_{d, i} ⅇ^{β X_{d, i, j}} + (r_{d, i} - 1) S_{d, i} ⅇ^{β X_{d}} \end{matrix} & (31) \end{matrix}

[0066] The out-of-cell interference is:

\begin{matrix} I_{out} = \sum_{j = 1}^{N_{v}} ψ_{v, j} S_{v} ⅇ^{β X_{v, j}} + \sum_{k = 1}^{C} \sum_{j = 1}^{N_{d, k}} ψ_{d, k, j} r_{d, k} S_{d, k} ⅇ^{β X_{d, k, j}} & (32) \end{matrix}

[0067] The SAP in equation (26) depends on the means and variances of the random variable A. The SAP determination for the multi-class case is simple, as the random variables constituting I_inand I_outare all independent. The X variables modeling the power control imperfections are independent and identically distributed.

Coverage and Received Power Levels

[0068] In a fully loaded system with maximally shrunk cells (i.e., no cell breathing), the received power levels needed to determine system capacity for a given SAP are constrained by two factors: the peak transmit power (per code) for each user and the maximum loss due to fading that can occur within the cell. The power limitations are usually determined by equipment specifications, while the maximum loss is usually recorded through propagation measurements.

[0069] A method consistent with the present invention determines the received power levels through a simple link analysis under maximum loading and uniform coverage conditions for all classes of users. Let PL denote the maximum loss that can occur in a cell. This includes loss due to fading, as well as antenna and transmission losses. The peak transmit power of all voice handsets S_v(T_x.) is assumed to be fixed. The received voice frame power S_v(R_x.) is derived from PL and the maximum transmit power S_v(T_x.): S_v(R_x.)(dbm)=S_v(T_x.)(dbm)−PL(dbm). The class i data user received power levels S_d,i(R_x.) are preferably set at some multiple of this received voice frame power.

CAPACITY EXAMPLE

[0070] Examples of methods for capacity analysis of a system with voice users and a single class of data users consistent with the present invention will now be presented. If perfect power control and a uniform user distribution is assumed, then SAP equations (22) and (23) become:

\begin{matrix} \begin{matrix} 1 - P_{v} = \sum_{k_{d} = 0}^{N_{d}} \sum_{k_{v} = 0}^{N_{v} - 1} F_{d} (k_{d}, N_{d}) F_{v} (k_{v}, N_{v} - 1) \times \\ Q [\frac{δ_{v} - k_{v} - k_{d} r_{d} \frac{S_{d}}{S_{v}} - (m_{v} + m_{d} \times \frac{S_{d, j}}{S_{v}})}{\sqrt{σ_{v}^{2} + \frac{S_{d}^{2}}{S_{v}^{2}} \times σ_{d}^{2}}}] \end{matrix} & (33) \\ \begin{matrix} 1 - P_{d} = \sum_{k_{d} = 0}^{N_{d}} \sum_{k_{v} = 0}^{N_{v} - 1} F_{v} (k_{v}, N_{v}) F_{d} (k_{d}, N_{d} - 1) \times \\ Q [\frac{δ_{d} - k_{v} (\frac{S_{v}}{S_{d}}) - k_{d} r_{d} - (m_{v} (\frac{S_{v}}{S_{d}}) + m_{d})}{\sqrt{σ_{d}^{2} + \frac{S_{v}^{2}}{S_{d}^{2}} \times σ_{v}^{2}}}] \end{matrix} & (34) \end{matrix}

[0071] For an imperfectly power controlled system and with all the users located at the edge of the cells, the SAP for the voice and data users can be computed using equations (27)-(29). The means and variances are:

[0072] E[I_v,in]=(N_v−1)α_vS_vm₁+N_dα_dS_dr_dm₁ (35)

[0073] E[I_d,in]=N_vα_vS_vm₁+(N_d−1)α_dS_dr_dm₁+(r_d−1S_dm₁ (36)

[0074] Var[I_v,in]=(N_v−1)(α_v(1−α_v)+α_v²)(σ₁²+m₁²)S_v²+(N_v−1)σ_v²S_v²m₁²+

[0075] N_d(α_d(1−α_d)+α_d²)(σ₁²+m₁²)S_d²r_d²+N_dα_d²m₁²S_d²r_d² (37)

[0076] Var[I_d,in]=N_v(α_v(1−α_v)+α_v²)(σ₁²+m₁²)S_v²−N_vα_v²S_v²m₁²+

[0077] (N_d−1)(α_d(1−α_d)+α_d²)(σ₁²+m₁²)S_d²r_d²−(N_d−1)α_d²m₁²S_d²r_d²+ (38)

[0078] The mean and variance of the out-of-cell-interference is computed as:

[0079] E[I_out]=N_vα_vS_vm₁+N_dα_dS_dr_dm₁ (39)

[0080] Var[I_out]=N_v(α_v(1−α_v)+α_v²)(σ₁²+m₁³)S_v²+N_vα_v²m₁²S_v²+

[0081] N_d(α_d(1−α_d)+α_d²)(σ₁²+m₁²)S_d²r_d²+N_dα_d²m₁²S_d²r_d² (40)

[0082] In a method consistent with the present invention, the capacity of a CDMA system, i.e., the number of voice users N_vand the number of data users N_d,ifor each data class i, can be determined by equations (22) and (23) (uniform distribution, perfect power control) or by equations (27)-(29) (users at cell edges, imperfect power control). For the perfect power control method, N_vand N_d,ican be determined when the activity variables and QoS requirements are known for each user class and the maximum received power levels are determined. For the imperfect power control method, the variance of the random variable X used to estimate the actual received power must also be known.

[0083] FIGS. 4-6 illustrate the results of methods consistent with the present invention for determining capacity of a system with the following parameters and settings. Conventional cellular systems such as IS-95A use a 1.228 MHz CDMA channel with a basic transmission rate for a single code of 9.6 kbps or 14.4 kbps. For exemplary purposes, the basic rate will be assumed to be 9.6 kbps, which gives the system a processing gain of 128. The maximum loss (due to fading and transmission) is set at −130 db. For a path loss of −130 db, the empirical estimate of cell size obtained from a well-known model (the Hata model for urban environments with base antenna height=30 m and mobile antenna height=3 m) would be about 1.5 km. Conventional CDMA voice handsets have a peak transmit power capability of 200 mW. For a cell with maximum losses of −130 db, the maximum received power from a voice user located at the edge of the cell would be −106.98 dbm. The required received power of all voice users is set at −106.98 dbm for exemplary purposes to ensure uniform coverage for all users in a maximally loaded cell. The required received powers for data transmissions are assumed to be equal to voice power (assuming the same SAP and FER requirements for voice and data transmissions). The FER and SAP for both voice and data users are set at 1% and 90%, respectively. Experimental results have shown that for this QoS, an E_b/N₀of 7 db is required (R. Padovani, “Reverse link performance of IS-95 based cellular systems,” IEEE Personal Communications, no. 3, pp. 28-34, 1994, which is hereby incorporated by reference). Finally, the voice activity factor is assumed to be ⅜ while the data activity factor is assumed to be 1 (dedicated bandwidth). The noise power at 800 MHz is approximately −108 dbm.

[0084] FIG. 4 shows the results of the method with perfect power control and uniform user distribution. Equations (33) and (34) are used to determine the number of voice users versus the number of data users of a single class. Results are shown when the number of codes per data user is set at 1,2,4, and 8.

[0085] FIG. 5 and 6 provide the capacity estimate derived using a method consistent with the present invention with imperfect power control and user location along cell boundaries, for standard deviations of 0.5 db and 1.5 db in the power control, respectively. Equations (35)-(40) are used to determine the number of voice users versus the number of data users of a single class. Results are shown when the number of codes per data user is set at 1, 2, and 4.

[0086] The effect of introducing a high speed data service on the voice capacity is of primary concern to most cellular service providers. The loss in voice capacity due to imperfections in power control and user distribution can be high as seen in FIGS. 5 and 6. The slope of the decline in available voice capacity increases as the rate of the data users in the system increases, implying that HSD users using m codes affect the voice capacity more adversely than m data users using a single code.

Admission Control

[0087] Consistent with the present invention, an admission control methodology utilizes the relationship between capacity and QoS requirements described above. Admission control is the process by which the cellular network determines whether a new user can be admitted to the network and controls certain aspects of the network in order to allow admittance of new users. As shown in equations (22) and (23) and (27)-(29), the maximum number of voice and data users that the network can support is determined by the users' QoS requirements, their received power levels, and their activity levels. As will now be explained, the cellular network can control each of these to enhance the capacity of the network.

QoS Control

[0088] FIG. 7 illustrates a method for enhancing CDMA cellular network capacity consistent with the present invention by controlling the QoS of data users. First, the QoS (FER and SAP) of voice users is fixed (step 100). Usually step 100 will be performed at MSC 10 (executed by processor 14) (FIG. 2), where software for changing the FER and SAP will usually reside. Alternatively, step 100 may be performed at base station 20 (executed by processor 24) (FIG. 2). Next, the method chooses acceptable FER and SAP values for data users (step 102) using the equations provided by the capacity model. The current voice load and data traffic level are inputs to the capacity model. Next, the method controls the received power of voice users so that all voice frames are received at equal power (step 104). This is accomplished by executing standard up/down power control commands from the base station. Then the received power of data users is controlled so that all data frames are received at the same power as the voice frames (step 106), again accomplished by executing standard up/down power control commands from the base station. Next, the received power for data users is decreased until the FER target from step 102 is reached (step 108). The capacity model equations ((22)-(23) or (27)-(29)) are used to estimate the decrease in power necessary to satisfy the FER target. Whenever a new data user requests access to the cellular network (step 110), the FER and SAP requirements for the data users can be loosened (step 112). The received power for data users is decreased until the new FER target is reached (step 108), thereby providing spare capacity in the network and allowing more data traffic to be carried. An additional benefit of reducing the received power for data users in response to loosened QoS requirements is handset power savings, and therefore an enhanced handset battery life.

User Activity Control

[0089] In wireline networks, the allocation of bandwidth on a burst basis, through medium access and resource allocation mechanisms that control user activity, has resulted in significant improvements in capacity utilization. A method consistent with the present invention for enhancing capacity in a CDMA cellular network similarly controls user activity, specifically data user activity, in order to free up capacity for additional users desiring access to the network.

[0090] FIG. 8A illustrates a method for enhancing CDMA cellular network capacity consistent with the present invention by controlling the activity level of data users. First, MSC 10 collects requests for network access from data users (step 120). Next, MSC 10 estimates the current voice load based on measured voice traffic and expected traffic patterns for the time of day (step 122). Next, all data users are assigned the same number of codes (step 124). The number of codes is determined by the rate offering, e.g., 9.6 kbps, 38.6 kbps, 64 kbps. The rate offering in turn depends on the voice load and the number of data requests. For example, if the voice load is low and there are few data user requests, data users can be assigned a higher rate offering and therefore a larger number of codes. Thus the data rate offering (the number of codes assigned to each data users) is discretionary at the behest of the network controller. Next, an appropriate data activity factor is determined from the capacity model (equations (22)-(23) or (27)-(29)), given the estimated voice load and data load and the QoS targets for voice and data (step 126). Once a data activity factor has been determined, MSC 10 or base station 20 controls data user activity (step 128). This can be accomplished, for example, by transmitting a message to all data mobile units to use their codes only a percentage of time, determined by the data activity factor. Alternatively, data user activity may be controlled by randomly sending permission bits (1 or 0) from base station 20 to the data users. 1 indicates a data user can transmit using all codes. 0 indicates a data user does not have permission to transmit. The base station sends a 1 a percentage of the time, where the percentage is based on the data activity factor determined in step 126.

[0091] When the system receives a request for access from a new voice user (step 130), processing returns to step 122, where the new voice load is estimated. Data code assignment and/or data activity is then adjusted to accommodate the new voice user (steps 124-128).

[0092] When the system receives a request for access from a new data user (step 132), the system determines whether to admit the new data user based on a capacity estimate determined by the capacity model (step 134). If a limitation, such as a minimum activity factor, has been reached, the system will no longer admit new users. If the system decides to admit a new user, processing returns to step 126 to set a new data activity factor that allows new users to enter the network.

[0093] FIG. 8B illustrates a related method for enhancing CDMA cellular network capacity consistent with the present invention by controlling the activity level of data users. Steps 140-154 are identical to steps 120-134 in FIG. 8A. If the system decides to admit a new data user (step 154), the system may reduce the number of codes per data user (step 156), and processing returns to step 146. This provides an additional way to enhance free capacity for new users.

Power Control

[0094] In another method consistent with the present invention, power management is used to maintain QoS levels of existing users and maximize the free capacity available for new users. This results in admission control process that includes two main phases:

[0095] (1) The power of the active users is scaled appropriately to maximize the tolerance to additional interference. The tolerance margin for interference provides an estimate of the maximum data rate the network can support, given the current loading of the network. This data rate is also dependent on the QoS requirements of the links (frame error rate and outage probability). For example, users with lower FER and outage probability have a lower signal to interference ratio (SIR) requirement and contribute less interference to other active users. Therefore, the overall data rate that can be supported by the system at the lower QoS is higher, allowing more low rate users or increased data rate of a single user.

[0096] (2) The interference margin available from the scaling process is used to introduce new users into the system. The new users are selected from the pool of data users requesting service and allocated transmit power levels in a manner that optimizes some network wide measure, such as the overall throughput, the number of active users, the energy cost per bit, or an arbitrary revenue function.

[0097] A power allocation method consistent with the present invention allocates power to existing users in an optimal manner to maximize free capacity for additional users. In order to determine the optimal power allocation, a system with M active codes/channels is considered. A code/channel is active if it is currently being used by a user for transmission. Each active user may be holding one or more active codes. This represents a multi-code CDMA system where voice users and low rate data users use a single code transmitting a fixed fundamental speed, while high-speed data (HSD) users use a number of basic rate codes in parallel. Let s=[s₁s₂. . . s_M] be the vector of received powers from the M active codes. Let s^max=[s₁^maxs₂^max. . . s_M^max] be the peak received power vector possible. s_i^maxis a function of the handset transmitter peak output power and the propagation gain from the user's current location. Note that s_i^maxcan change with the location and velocity of the user, as the channel conditions between the user and the base station vary. All M codes are used to communicate with the same base station, i.e., there is a fixed base station assignment.

[0098] On the uplink of a CDMA system, the signals of interfering transmissions can be modeled as interfering noise signals. The SIR requirements of the code are denoted by T_i. For each of the M active codes, the SIR must remain above the T_ilevel in order to ensure a certain frame error rate (FER). This requirement can be written as:

\begin{matrix} \frac{S_{i}}{\sum_{j \neq i} S_{j} + η + B} \geq T_{i} \forall i = 1, 2, \dots, M & (41) \end{matrix}

[0099] where η is the other cell interference plus thermal noise. Note that there are M codes that are currently active. B is the interference that will be injected by the new codes to be activated. B is also referred to herein as the interference “slack” that the current loading of the system (M active codes) permits. Let s^min=[s₁^mins₂^min. . . s_M^min] denote the vector that satisfies the inequalities in equation (41) in equality with B=0, i.e.,

\begin{matrix} \frac{S_{i}^{\min}}{\sum_{j \neq i} S_{j}^{\min} + η} = T_{i} \forall i = 1, 2, \dots, M & (42) \end{matrix}

[0100] For a non-trivial solution s>0 of the inequalities in equation (42), it is straightforward to show that 0<T_i<1. In a CDMA system, T_iis the ratio of the required E_b/N₀and the spreading gain.

[0101] The conditions for the existence of such a vector s^minand its uniqueness have been described in several references, including G. J. Foschini and Z. Milanjic, “A simple distributed autonomous power control algorithm and its convergence,” IEEE Transactions on Vehicular Technology, vol. 42, pp. 641-646, November, 1993, and R. D. Yates, “A framework for uplink power control in cellular radio systems,” IEEE Journal on Selected Areas in Communications, vol. 13, pp. 1341-1346, September, 1995. The following analysis assumes that such a vector exists and is unique.

[0102] The relationship between any feasible vector s′ and s^minis given by:

[0103] s′>s^min(elementwise) Lemma (1)

[0104] By way of proof, Yates has demonstrated that starting with any feasible power vector s′, a monotonically decreasing sequence of feasible power vectors can be constructed that converges to a unique fixed point as follows: Let s(0)=s′. Define the vector I_i(s)=T_i(Σ_j≠is_j+η). Let s(n+1)=I(s(n)). Since the sequence is monotonically decreasing and the fixed point, s^minis unique, s(0)≧s^min→s′≧s^min.

[0105] Therefore, s′ can be represented as s′=[α₁s₁^minα₂s₂^min. . . α_Ms_M^min], where α_iis referred to as the “scaling” factor, and α_i≧1 ∀i=1,2, . . . M.

[0106] To ensure that the quality of active links is maintained while maximizing the interference “slack” B that can be used to admit new data users, a new power vector s* for the active users is determined that satisfies the inequalities in equation (41) for the largest possible value of B.

[0107] In order to formulate the problem, the following functions are defined:

[0108] g_i(s):=s_i−s_i^max

[0109]

h_{i} (s) : = - (\frac{s_{i}}{T_{i}} - \sum_{j \neq i} s_{j} - η)

f_{i} (s) : = \frac{s_{i}}{T_{i}} - \sum_{j \neq i} s_{j} - η

The set Ω:={s: s≧0, g_i(s)≦0 and h_i(s)≧0 ∀i=1,2, . . . ,M} (43)

[0110] The set of feasible power vectors is determined by the maximum power limitations (g_i(s)≦0 ∀i=1,2 . . . M) and the active link quality constraints (h_i(s)≦0 ∀i=1,2 . . . M). Since the constraints are all linear functions of the power vector s, the feasible set is a bounded polyhedral cone. The optimal power vector s* that lies within the feasible set and results in the maximum value of B is also the power vector at which the function min_i𝑓_i(s) has the maximum value, because any value of B>min_i𝑓_i(s) will violate at least one of the M constraints in equation (41). Therefore, the problem may be stated as follows:

[0111] P₀: max_smin_i𝑓_i(s) subject to g_i(s)≦0 and h_i(s)≦0 ∀i=1,2 . . . M (44)

[0112] This problem is a non-smooth optimization problem as the objective function min_i𝑓_i(s) is not smooth. It is shown in R. Fletcher, Practical Methods of Optimization, p. 361, John Wiley & Sons, 1987, that problem P₀has the following equivalent formulation:

[0113] P₁: max_s,vv subject to v−𝑓_i(s)≦0 ∀i=1, . . . , M (45)

[0114] This leads to the following theorem:

[0115] The two problems P₀and P₁are equivalent in the sense that if s* is a local solution of the problem P₁, s* also maximizes the objective function in problem P₀, min_i𝑓_i(s). Theorem (1)

[0116] By way of proof by contradiction, let sl be the optimal solution for Problem P₀and let 𝑓1 be the corresponding value of min_i𝑓_i(s1). Suppose that (s1,𝑓1) is not optimal for the equivalent formulation, problem P₁. Thus, there exists a vector s2 not equal to s1 and a corresponding optimal value v2, such that v2 is the maximum value satisfying v2<𝑓_i(s2)∀i. By the definition of optimality for problem P₁, v2>v1 where v1 is the value of the objective function max_vv such that v1<𝑓_i(s1)∀i. However, this contradicts the fact that s1 is an optimal value for the problem P₀because the objective function in problem P0, min_i𝑓_i(s), has a higher value at s2 than at s1. A similar argument can be made in the reverse direction. Since a local optimum in problem P1 is also a global optimum (convex objective function defined over a convex set), such a point is also the global optimal solution for problem P₀.

[0117] In order to determine the optimal power vector s*, consider how the objective function min_i𝑓_i(s) varies over the set Ω:

[0118] Given any point s1 in the strict interior of the bounded polyhedral cone Ω, there exists at least one point s2 on the boundary of this set, where the objective function of problem P₀, min_i𝑓_i(s), has a higher value than at s1, i.e., min_i𝑓_i(s1)<min_i𝑓_i(s2). Lemma (2)

[0119] By way of proof, since s1 is a feasible power vector, it may be represented as s1=[α_1,1s₁^minα_1,2s₂^min. . . α_1,Ms_M^min].

\begin{matrix} \begin{matrix} \min_{i} f_{i} (s1) = \min_{i} (\frac{α_{1, i} s_{i}^{\min}}{T_{i}} - \sum_{j \neq i} α_{1, i} s_{j}^{\min} - η) \\ = \min_{i} (\sum_{j \neq i} (α_{1, i} - α_{1, j}) s_{j}^{\min} + (α_{1, i} - 1) η) \end{matrix} & (46) \end{matrix}

[0120] Let s2 be a new point constructed as follows: s2=[α_2,1s₁^minα_2,2s₂^min. . . α_2,Ms_M^min]. Let k denote the mobile which has the minimum value of the ratio of peak power to minimum required power i.e.,

\frac{s_{k}^{\max}}{s_{k}^{\min}} = \min_{i} \frac{s_{i}^{\max}}{s_{i}^{\min}} .

[0121] Define

Δ = \frac{s_{k}^{\max}}{s_{i}^{\min}} - α_{1, k}

[0122] and let α_2,i=α_1,i+Δ. Then s2 is a point on the boundary of the set Ω. This is true because at s2, mobile k is transmitting at its peak power capability s_k^maxwhile all other mobiles are at or below their peak transmit power thresholds, and the constraints g(s)<0 and h(s)<0 are satisfied.

\begin{matrix} \min_{i} f_{i} (s2) = \min_{i} (\sum_{j \neq i} (α_{2, i} - α_{2, j}) s_{j}^{\min} + (α_{2, i} - 1) η) & (47) \end{matrix}

[0123] It is known that α_2,i−α_2,j=(α1,i−a1,j) since the scaling factors α₁are adjusted by a constant factor Δ>0 for all mobiles. Since δ>0, α_2,i>α_1,iand therefore min_i𝑓_i(s2)>min_i𝑓_i(s1). This implies that the optimal power vector s* lies on the boundary of the set Ω, where one or more of the constraints h_i(s)<0 and g_i(s)<0 are binding. The search for s* can be further focused by the following results.

[0124] If g_i(s^min)<0 ∀i=1, . . . M, then at the optimal power vector s*, the constraints h_i(s*) are nonbinding, i.e., h_i(s*)<0. Lemma (3)

[0125] By way of proof, let s* be the power vector that is optimum for problem P₀, such that at least one of the constraints h_i(s*), say h_k(s*), is binding, i.e., h_k(s*)=0. Then the optimal value of problem P₀is 0, i.e., B=0, because min_i𝑓_i(s*)=𝑓_k(s*)=0. If g_i(s^min)<0 ∀i=1, . . . ,M, then there exists a feasible point s1 which is obtained by simply scaling the power vector s^minby a suitable constant α>1 i.e., s1=αs^min. Now, 𝑓_i(s^min)=0 ∀i . Also

f_{i} (s^{\min} < \frac{1}{α} f_{i} (α s^{\min})

[0126] <𝑓_i(s1) since α>1. Therefore, min_i𝑓_i(s1)>0 and hence, min_i𝑓_i(s1)>min_i𝑓_i(s*). This implies that s′ cannot be the optimal power vector that solves problem P₀, thereby resulting in a contradiction.

[0127] This result implies that as long as none of the M codes are required to transmit at the peak power capability of the corresponding mobile in order to be active, i.e., s_i^min≠s_i^max, the “slack” B that exists for admission of new users is non-zero.

[0128] At the optimal power vector s*, 𝑓_i(s*) is given by:

[0129] 𝑓_i(s*)=𝑓_i(s*), i,jε{1,2, . . . , M}. Lemma (4)

[0130] From Theorem (1), it is known that the power vector s that solves problem P₁also solves problem P₀. The problem P₁is a smooth optimization problem, and since the feasible set Ω is bounded and convex, and the objective function is concave, the Kuhn-Tucker conditions for the global optimal solution s* can be written. The K-T conditions are both necessary and sufficient conditions for an optimal solution s* to problems P₀and P₁.

[0131] The appropriate Lagrangian function for the problem P₁would be:

\begin{matrix} L (s, v, a, b, c) = v - \sum_{i}^{M} a_{i} g_{i} (s) - \sum_{i}^{M} b_{i} h_{i} (s) - \sum_{i}^{M} c_{i} (v - f_{i} (s)) & (48) \end{matrix}

[0132] where the vectors a, b, and c are the Lagrange multipliers. The K-T conditions yield:

\begin{matrix} \frac{δ}{δ v} L (s^{*}, v^{*}, a, b, c) = 0 \Rightarrow \sum_{i}^{M} c_{i} = 1 & (49) \end{matrix}

\begin{matrix} \nabla_{s} L (s^{*}, v^{*}, a, b, c) = 0 \Rightarrow a_{i} + \sum_{j \neq i} b_{j} - \frac{b_{i}}{T_{i}} + \sum_{j \neq i} c_{j} - \frac{c_{i}}{T_{i}} = 0 & (50) \end{matrix}

a_i, b_i, c_i≧0; ∀i=1, . . . , M (51)

[0133] a_ig_i(s*)=0; ∀i=1, . . . , M (52)

[0134] b_ih_i(s*)=0; ∀i=1, . . . , M (53)

[0135] c_i(v*−𝑓_i(s*))=0; ∀i=1, . . . , M (54)

[0136] Equations (52-54) are the complementary slackness conditions. From Lemma (3), it is known that h_i(s*)≠0, and therefore, from equation (53), b_i=0. Substituting for b_iin equation (50),

\begin{matrix} a_{i} + \sum_{j \neq i} c_{j} = \frac{c_{i}}{T_{i}} & (55) \end{matrix}

[0137] If for any k, c_kis 0, then from equations (51) and (55), it is seen that for all i, c_i=0, and hence Σ_i^Mc_i=0. This is a contradiction of the condition in equation (49). Hence, c_i>0 for all i. Then, equation (54) implies that 𝑓_i(s*)=v for all i or 𝑓_i(s*)=fj(s*) for any i and j.

[0138] This result implies that at the optimal received power vector s*, all mobiles have the same “slack” B or tolerance margin for interference to be introduced by new users. In order to determine s*, the appropriate scaling vector α* must be determined. The following result translates the equal slack requirement in Lemma (4) to the corresponding condition on the optimal scaling vector α.

[0139] Let s be any feasible vector, represented by s=[α₁s₁^min. . . α_M^min]. If 𝑓_k(s)=𝑓_i(s), then α_k=α_i. Lemma (5)

[0140] By way of proof, equations (42) and (43) yield:

\begin{matrix} f_{i} (s) = \sum_{j \neq i} (α_{i} - α_{j}) s_{j} + (α_{i} - 1) η f_{k} (s) = \sum_{j \neq j} (α_{k} - α_{j}) s_{j} + (α_{k} - 1) η & (56) \end{matrix}

[0141] Since 𝑓_i(s)=𝑓_k(s), the following is true:

\begin{matrix} α_{i} [\sum_{j \neq i, k} s_{j} + s_{k}] - [\sum_{j \neq k, i} α_{j} s_{j} + α_{k} s_{k}] + (α_{i} - 1) η = α_{k} [\sum_{j \neq i, k} s_{j} + s_{i}] - [\sum_{j \neq k, i} α_{j} s_{j} + α_{i} s_{i}] + (α_{k} - 1) η & (57) \end{matrix}

[0142] From equation (57), it follows that α_i=α_k.

[0143] This result allows us to determine the optimal solution s* of the original problem P₀. The result in Lemma (5) indicates that “uniform” scaling (same scaling factor, α_i=α_j∀i,j) of the minimum power vector s^minyields s*. The constant scaling factor α* is limited by the peak transmit powers of the mobiles at the current instant.

[0144] The optimal power vector is given by the following theorem:

[0145] The optimal solution of problem P₀is the power vector s* given by s*=α*s^min, where

α *= \min_{i} (\frac{s_{i}^{\max}}{s_{i}^{\min}}) .

[0146] Further, the value of the optimal solution (the maximum “slack”B) is (α*−1)η. Theorem (2)

[0147] Since the M codes are active, there exists at least one feasible power vector and therefore there exists a unique minimum power vector s^min. The maximum received power vector s^maxis assumed to be known and fixed at the admission epoch. From Lemma (5), a necessary and sufficient condition for the optimal solution to problem P₀is that s*=αs^min, where α is a constant. As this constant scaling factor increases, the value of the slack B also increases, i.e., if α1>α2 then min_i𝑓_i(α1s^min)>min_i𝑓_i(α2s^min). Therefore, the optimal vector s* is obtained by scaling s^minby the largest constant. Since s* is also a feasible power vector, the largest such value is

α *= \min_{i} (\frac{s_{i}^{\max}}{s_{i}^{\min}}) .

[0148] If the scaling factor is greater than α*, at least one of the M codes will be peak power limited and unable to attain its QoS.

[0149] The solution of problem P₀provides the maximum slack B that the system can tolerate with M active codes. Based on the above solution s* and using equation (42) and Lemma (4),

\begin{matrix} \begin{matrix} B = \frac{s_{i}^{*}}{T_{i}} - \sum_{j \neq i} s_{j}^{*} - η \\ = α \cdot [\sum_{j \neq i} s_{j}^{\min} + η] -- \sum_{j \neq i} s_{j}^{*} - η \\ = (α^{*} - 1) η . \end{matrix} & (58) \end{matrix}

[0150] The optimal power allocation given by Theorem (2) is used in a method consistent with the present invention to enhance CDMA cellular network capacity by controlling power levels. FIG. 9A illustrates a method consistent with the present invention for power control. The method implements the optimal solution given by Theorem (2). Initially, a feasible power vector is assigned to the active users (step 160). The initial power vector may be determined analytically or through the standard process through which users ramp-up their power. Next, the minimum power vector is computed using the active links (step 162). The power levels of this minimum power vector are then assigned to the active users using standard up/down power control commands from the base station (step 164). Next, the maximum received handset power is determined as the product of the peak handset transmission power, which is known, and the path gains, which is measured using standard power control procedures (stop 166). Then the ratio of peak received power to the minimum received power computed in step 162 is determined (step 168). The link with the smallest such ratio is determined (step 170), and all active link powers (i.e., the minimum powers previously assigned) are scaled by the ratio to determine new received power levels, as shown in Theorem (2) (step 172). Finally, the new received power levels are set on the active links using standard up/down control commands (step 174).

[0151] FIG. 9B illustrates another power control method consistent with the present invention. The method is similar to that in FIG. 9A, except that the minimum power vector is obtained in a different manner (steps 182-184). In step 182, the received power of each user is set to the interference power. If this is repeated an infinite number of times, it has been shown that the achieved power vector is the same as the minimum power vector determnined in step 162 in FIG. 9A. Since it is physically impossible to repeat step 182 an infinite number of times, a vector close to the minimum power vector may be achieved by iterating a large enough number of times (step 184). Once an approximation of the minimum power vector is achieved, the remaining steps (steps 186-194) are identical to steps 166-174 in FIG. 9A.

[0152] It will be apparent to those skilled in this art that various modifications and variations can be made to the capacity enhancement schemes of the present invention without departing from the spirit and scope of the invention. Other embodiments of the invention will be apparent to those skilled in this art from consideration of the specification and practice of the invention disclosed herein. For example, a power allocation method consistent with the present invention may be implemented anywhere in the cellular network where power control processing occurs. Similarly, a QoS control method consistent with the present invention may be implemented anywhere in the network where QoS parameters are controlled. It is intended that the specification and examnples be considered exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

Claims

We claim:

1. A method for enhancing the capacity of a CDMA cellular carrier supporting traffic from a plurality of users, including a plurality of voice users and a plurality of multi-code data users, each user having an adjustable received power level, and the plurality of voice users having a fixed quality of service requirement, the method comprising the steps of:

setting a quality of service requirement for the plurality of data users based on the traffic from the plurality of users and the voice user quality of service requirement;

adjusting the received power level of each of the plurality of voice users so that their received power levels are identical;

adjusting the received power level of each of the plurality of data users so that their received power levels are identical to the received power levels of the plurality of voice users;

decreasing the received power level of each of the plurality of data users until their quality of service requirement is satisfied; and

adjusting the quality of service requirement for the plurality of data users in response to a request from a new data user to access the cellular carrier.

2. A method for enhancing the capacity of a CDMA cellular carrier supporting traffic from a plurality of users, including a plurality of voice users and a plurality of multi-code data users, each data user having an adjustable activity factor and capable of transmitting at an adjustable activity lcvel, and each user having a fixed quality of service requirement, the method comprising the steps of:

determining a measure of the traffic from the plurality of voice users;

determining a measure of the traffic from the plurality of data users;

assigning to each of the data users an identical number of codes;

adjusting the activity factor for each data user based on the measures of traffic from the plurality of voice users and the plurality of data users and the quality of service requirement of each of the plurality of users; and

controlling the activity level of each of the plurality of data users in response to the adjusted activity factor.

3. The method of claim 2 wherein the controlling step includes the substep of transmitting a message to each of the plurality of data users indicating a percentage of time during which each data user may transmit.

4. The method of claim 2 wherein the controlling step includes the substep of continuously transmitting an indicator to each of the plurality of data users, a value of the indicator indicating that the data user has permission to transmit, wherein the indicator is set to the value randomly based on the adjusted activity factor.

5. The method of claim 2 further comprising the step of reducing the number of codes assigned to each of the plurality of data users in response to a request from a new data user to access the cellular carrier.

6. A method for enhancing the capacity of a CDMA cellular carrier supporting traffic from a plurality of users, including a plurality of voice users and a plurality of multi-code data users, each data user having an adjustable received power level, and each user having a quality of service requirement, the method complising the steps of:

assigning a feasible power level to each of the plurality of users;

determining a minimum power level to each of the plurality of users;

determining a maximum received power level from among the received power levels of the plurality of users;

determining a ratio of the maximum received power level to the minimum power level determined for each of the plurality of users;

selecting smallest ratio from the determined ratios;

scaling the minimum power level determined for each of the plurality of users by the smallest ratio; and

assigning the scaled power level to each of the plurality of users.

7. A computer-readable medium containing instructions for enhancing the capacity of a CDMA cellular carrier supporting traffic from a plurality of users, including a plurality of voice users and a plurality of multi-code data users, each user having an adjustable received power level, and the plurality of voice users having a fixed quality of service requirement, by:

setting a quality of service requirement for the plurality of data users based on the traffic from the plurality of users and the voice user quality of service requirement;

adjusting the received power level of each of the plurality of voice users so that their received power levels are identical;

adjusting the received power level of each of the plurality of data users so that their received power levels are identical to the received power levels of the plurality of voice users;

decreasing the received power level of each of the plurality of data users until their quality of service requirement is satisfied; and

adjusting the quality of service requirement for the plurality of data users in response to a request from a new data user to access the cellular carrier.

8. A computer-readable medium containing instructions for enhancing the capacity of a CDMA cellular carrier supporting traffic from a plurality of users, including a plurality of voice users and a plurality of multi-code data users, each data user having an adjustable activity factor and capable of transmitting at an adjustable activity level, and each user having a fixed quality of service requirement, by:

determining a measure of the traffic from the plurality of voice users;

determining a measure of the traffic from the plurality of data users;

assigning to each of the data users an identical number of codes;

controlling the activity level of each of the plurality of data users in response to the adjusted activity factor.

9. A computer-readable medium containing instructions for enhancing the capacity of a CDMA cellular carrier supporiing traffic from a plurality of users, including a plurality of voice users and a plurality of multi-code data users, each data user having an adjustable received power level, and each user having a quality of service requirement, by:

assigning a feasible power level to each of the plurality of users;

determining a minimum power level for each of the plurality of users;

determining a maximum received power level from among the received power levels of the plurality of users;

determining a ratio of the maximum received power level to the minimum power level determined for each of the plurality of users;

selecting a smallest ratio from the determined ratios;

scaling the minimum power level determined for each of the plurality of users by the smallest ratio; and

assigning the scaled power level to each of the plurality of users.

10. The method of claim 2, wherein the step of adjusting the activity factor is performed by a base station; and wherein the step of controlling the activity level is performed by each of the plurality of data users.

Patent Citations (20)

Patent	Date	Inventor	Cited By
US5107487	1992-04-01	Vilmur et al.	Applicant
US5257283	1993-10-01	Gilhousen et al.	Applicant
US5299226	1994-03-01	Schilling	Applicant
US5341397	1994-08-01	Gudmunson	Examiner
US5457813	1995-10-01	Poutanen	Applicant
US5481561	1996-01-01	Fang	Applicant
US5485486	1996-01-01	Gilhousen et al.	Applicant
US5548616	1996-08-01	Mucke et al.	Applicant
US5566165	1996-10-01	Sawahashi et al.	Applicant
US5570353	1996-10-01	Keskitalo et al.	Applicant
US5590409	1996-12-01	Sawahashi et al.	Applicant
US5621723	1997-04-01	Walton	Examiner
US5623484	1997-04-01	Muszynski	Applicant
US5623486	1997-04-01	Dohi et al.	Applicant
US5722051	1998-02-01	Agrawal	Examiner
US5734646	1998-03-01	I	Examiner
US6038452	2000-03-01	Strawczynski	Examiner
US6044072	2000-03-01	Ueda	Examiner
US6069883	2000-05-01	Ejzak	Examiner
US6070085	2000-05-01	Bender	Examiner

Non-Patent Literature (27)

Bambos, N. et al., Radio Link Admission Control Algorithms for Wireless Networks with Power Control and Active Link Quality Protection, Tech. Report UCLA-ENG-94-25, ULCA School of Engg., p. 1-22, 1994.Applicant
N. Bambos et al., Power Control Based Admission Policies in Cellular Radio Networks, Proc. of IEEE Globecom, pp. 863-867, 1992.Applicant
Evans, J. et al., Effective Interference: a Novel approach for Interference Modelling and Traffic Analysis in CDMA Cellular Networks, Proc. of IEEE Globecom, vol. 3, pp. 433-442, 1995.Applicant
Evans, J. et al., Call Admission Control in Multiple Service DS-CDMA Cellular Networks, Proc. Of IEEE Vehicular Tech. Conf., vol. 1, pp. 227-231, 1996.Applicant
Zander, J., Distributed Cochannel Interference Control in Cellular Radio Systems, IEEE Transactions on Vehicular Technology, vol. 41, pp. 305-311, Aug. 1992.Applicant
Grandhi, S.A. et al., Distributed Power Control in Cellular Radio Systems, IEEE Transactions on Communications, vol. 42, pp. 226-228, Feb./Mar./Apr. 1994.Applicant
Grandhi, S.A. et al., Centralized Power Control in Cellular Radio systems, IEEE Transactions on Vehicular Technology, vol. 42, pp. 466-468, Nov. 1993.Applicant
Grandhi, S.A. et al., Constrained Power Control in Cellular Radio Systems, Proc. of IEEE Vehicular Tech. Conference, 1994.Applicant
Foschini, G.J. et al., A Simple Distributed Autonomous Power Control Algorithm and its Convergence, IEEE Transactions on Vehicular Technology, vol. 42, pp. 641-646, Nov. 1993.Applicant
Yates, R.D., A Framework for Uplink Power Control in Cellular Radio systems, IEEE Journal on Selected Areas in Communication, vol. 13, pp. 1341-1346, Sep. 1995.Applicant
Huang, C.Y. et al., Call Admission in Power Controlled CDMA Systems, Proc. of the IEEE Vehicular Technology Conference, vol. 3, pp. 1665-1669, 1996.Applicant
Yates, R.D. et al., Integrated Power Control and Base Station Assignment, IEEE Trans. on Vehicular Technology, vol. 44, pp. 638-644, Aug. 1995.Applicant
Hanly, S.V., An Algorithm for Combined Cell-site Selection and Power Control to Maximize Cellular Spread Spectrum Capacity, IEEE Journal on Selected Areas in Communication, vol. 13, pp. 1332-1340, Sep. 1995.Applicant
Mitra, D., An Asynchronous Distributed Algorithm for Power control in Cellular Radio systems, 4th WINLAB workshop in 3rd Generation Wireless Info. Networks, 1993.Applicant
Fletcher, R. Practical Methods of Optimization, John Wiley and Sons, 1987.Applicant
TR 45.5 Working Committee for CDMA, Service Description for Third Generation CDMA Systems applicable to IMT-2000 (Version 0.07) Aug. 5, 1997.Applicant
Chin-Lin, I. et al., Multi-code CDMA Wireless Personal Communications Networks, in ICC '95 Conference Record, pp. 1060-1064, Jun. 1995.Applicant
Chih-Lin, I. et al., Performance of Multi-Code CDMA Wireless Personal Communications Network, Proc. of IEEE Vehicular Technology Conference, pp. 907-911, 1995.Applicant
Chih-Lin, I. et al., Variable Spreading Gain CDMA with Adaptive Power Control for Integrated Traffic in Wireless Networks, Proc. of IEEE Vehicular Technology Conference, pp. 794-798, 1995.Applicant
Gilhousen et al., On the Capacity of a Cellular CDMA System, IEEE Transactions on Vehicular Technology, vol. 40, pp. 301-312, May 1991.Applicant
Liu, Z. et al., Interference Issues in Multi-Code CDMA Networks, PIMRC 1996, pp. 98-102, Oct. 1996.Applicant
Viterbi, A.J. et al., Erlang Capacity of a Power controlled CDMA System, IEEE Journal on Selected Areas in Communications, vol. 11, pp. 892-899, Aug. 1993.Applicant
Cameron, R. et al., Performance Analysis of CDMA with Imperfect Power Control, IEEE Transactions on Communication Theory, vol. 44, pp. 777-781, Jul. 1996.Applicant
Priscoli, F.D. et al., Effects of Imperfect Power Control and User Mobility on a CDMA Cellular Network, IEEE Journal of Selected Areas in Communication, vol. 14, pp. 1809-1817, Dec. 1996.Applicant
Mandayam, N.B. et al. Erlang Capacity for an Integrated Voice/Data DS-CDMA Wireless System with Variable Bit Rate Sources, Proc. of PIMRC, vol. 3, pp. 1078-1082, 1995.Applicant
Holtzman, J.M., A Simple, Accurate Method to Calculate Spread Spectrum Error Probabilities, IEEE Transactions on Communications, vol. 40, pp. 461-464, Mar. 1992.Applicant
Padovani, R., Reverse Link Performance of IS-95 Based Cellular Systems, IEEE Personal communications, No. 3, pp. 28-34, 1994.Applicant

Abstract