Mobile Networks and Applications 5 (2000) 5–16 5 Optimal channel assignment strategies for forced channel hopping in CDPD systems Chris Jedrzycki and Victor C.M. Leung Department of Electrical and Computer Engineering, The University of British Columbia, Vancouver, B.C., Canada V6T 1Z4 To provide an acceptable call blocking probability in circuit-switched cellular networks, such as the Advanced Mobile Phone System (AMPS) networks, a significant fraction of the channel capacity in each cell is normally unused. This “free” capacity can be effectively used for packet data transmissions that yield to voice traffic when necessary. Cellular Digital Packet Data (CDPD) is a packet-switched data service which may share radio channels with the AMPS service on a secondary basis to tap this “free” capacity. The length of time that a CDPD stream can occupy a channel is greatly influenced by the channel assignment strategies of both the AMPS and the CDPD systems. This paper investigates these channel assignment strategies and their effects on the CDPD channel holding times, in comparison with the optimal channel assignment strategy that interrupts the CDPD service only when a new AMPS call finds no other idle channels in the cell site. One such optimal strategy is the cooperative strategy in which the CDPD and AMPS networks actively communicate with each other. It is shown that other optimal strategies exist without the need of communications between the two systems. The effects of AMPS traffic levels, number of channels, and number of CDPD streams at the cell site on the CDPD channel holding time and channel utilization are also considered. 1. Introduction Cellular mobile communications have proven to be in- valuable for people on the move. The first generation ana- log cellular mobile radio networks use the Advanced Mo- bile Phone System (AMPS) [1] to provide circuit-switched voice service over 30 kHz FM/FDMA channels. In North America, the total bandwidth available to each of two ser- vice providers authorized to operate over specific geograph- ical areas is 25 MHz. In order to control co-channel in- terference, an N = 7 frequency reuse pattern is commonly used, such that each of 3 sectors in sectorized cell sites must use no more than 19 radio channels for user traffic. E.g., the network of B.C. Tel. Mobility, one of the cellular service providers in the Province of British Columbia in Canada, typically deploys 15 traffic channels per cell sector 1 . To provide a voice call blocking rate of between 1% and 3% during the peak traffic hour (a common grade of service ob- jective for most cellular telephony service providers) over these small channel sets, channel utilization of the order of 50%–60% is required. This leaves on average 40%–50% of the channel capacity idle or “free” during the busiest hour of the day. At other times of the day even more channels remain idle. Since analog cellular networks will remain in use for many years to come, there is continuous interest in new services which add values to these networks. Cellu- lar Digital Packet Data (CDPD) [2] is a recently developed packet-switched cellular mobile data service for the AMPS frequency band. It is intended as a secondary service over- laid on the primary circuit-switched voice service to take advantage of this free capacity to transmit data packets in idle cellular channels, which is the situation considered in this paper. Alternately, a CDPD network could also em- 1 Data on B.C. Tel. Mobility AMPS network were obtained in 1996. ploy dedicated cellular channels not shared with the voice service. Detailed descriptions of CDPD systems may be found in [2,3]. The system architecture consists of mobile data terminals, called Mobile End Systems (M-ESs), communi- cating over a CDPD network or with other M-ESs, and with Fixed End Systems (F-ESs) possibly via external data net- works interconnected with the CDPD network. The M-ESs in each cell exchange data packets with the Mobile Data Base Station (MDBS) at the cell site using 19.2 Kbps wire- less data channels, called CDPD streams. Several MDBS are interconnected in the CDPD network via a Mobile Data Intermediate System (MD-IS). Multiple CDPD streams may be available in each cell. Whereas the MDBS broadcasts data packets over the forward channel to all M-ESs in the cell sharing the same CDPD stream, the M-ESs contend for the reverse channel using the Digital Sense Multiple Access with Collision Detection (DSMA/CD) protocol. A descrip- tion of DSMA/CD and an analysis of its throughput-delay performance are given in [4]. When a CDPD stream is using an idle AMPS channel on a shared basis, it is required to give up the channel when the channel is assigned to a voice call, as determined by inter-system communications, or autonomously by a RF channel sniffer which is the preferred method. In this case transmission of the current data packet is aborted and a “forced-hop” of the CDPD stream to a different idle chan- nel, if one exists, is executed by the MDBS. If an idle channel is not available, the CDPD stream is suspended over a “blackout period” until a channel becomes avail- able. It has been shown that even at peak traffic load, a typical AMPS cell has plenty of free channel capacity with sufficiently long idle duration to be effectively utilized by CDPD streams for packet data transmissions [5].  Baltzer Science Publishers BV