Over the past one hundred years we have seen a heated debate about whether the radio spectrum is a scarce resource. To the public at large, the issue became particularly apparent during the heated demonstrations in the 1970s to provide legal AM broadcast frequencies for radio pirates transmitting from the international waters, or more recently during the Multi-Billion Euro auctions for UMTS mobile phone licenses, which causes an almost bankruptcy of some operators. Meanwhile technological progress has allowed us to transmit a number of bits per second in the total available spectrum that increased more rapidly that famous Moore's law for computer chip performance. Anyhow, it appears fair to conclude that efficient use of the radio spectrum is of critical importance to the deployment of new services and that it has drawn the attention of many researchers. The aim of this document is to review of common practice in the design of new radio networks. We will conclude that these practices are by and large inspired by circuit-switched networks, but do not necessarily support packet-switched traffic very well.

The cellular solution

The key design aspect behind cellular telephone networks is the regular reuse of radio frequencies in a certain geographic area. A crucial aspect in the evaluation and planning of such networks is the estimation of the effect of co-channel interference. The amount of interference that can be tolerated determines the required separation distance between transmit-receiver pairs that use the same radio resources. The performance of cellular telephone links was first studied around 1980. Initial analyses were limited to outage probabilities in continuous wave voice communication, i.e., the probability that the carrier-to-interference ratio drops below a minimum required value necessary to provide a reliable link. In the 1980's the technique for computing outage probabilities was refined step by step, considering among other things multipath fading, shadowing, multiple interfering signals, the modulation technique and error correction method.

The performance metric

To future data or multi-media services, queuing and retransmission delays are far more important performance measures than outage probability. Advanced methods to find link performance, as developed for cellular telephony, can still be used, but need to be extended to address the specific Quality of Service requirements for data or multi-media traffic. Messages lost due to interference from other cells can simply be retransmitted, so outage probabilities per sé may not be an appropriate criterion. The packet delay performance in the downlink, i.e., from base station to terminals, appears more appropriate.

The document is not intended as a report on, or justification of current approaches in the standardization of 3G or 4G wireless systems. In stead it shows that 'extrapolation' of the concepts used for GSM or cellular CDMA do not lead to efficient network for the traffic patterns and QoS expected in future.

For packet switched communication, it often is optimum to use the entire bandwidth in each cell. That is, it is suboptimum to split the spectrum in portions, and to avoid nearby reuse of the same portion of the spectrum. The corresponding high interference power levels from nearby transmitters in adjacent cells require a joint optimization and dynamic management of the spatial frequency reuse and the occupation of spectrum within cells.

Another aspect is the use of direct-sequence spread-spectrum modulation, which is generally accepted as an interesting option for telephone traffic. We will see that the advantages of CDMA for packet traffic are less evident.

Efficient mobile packet data transmission, as needed in networks for multimedia traffic or in an ambient intelligence setting, requires entirely different spectrum reuse than telephone nets. Several examples have been covered to support this insight. To optimize spectrum efficiency, user capacity and network performance, presumably a new class of access schemes is needed that dynamically combine random access within one cell with protection against interfering signals from other cells.

JPL Home