SideBar

Introduction

Wireless systems are communication networks which use the electromagnetic spectrum propagating over free space as a communication medium. Discovered and popularized by Marconi in the early 20th century, it has revolutionized the world over the last 100 years. Stunning advances in communication (cellular telephony, walkie talkies, international calling via satellite, wireless data using WiFi), navigation (GPS, radar), control (wireless sensors) and personal usage (radio based remote controls, RFID) have been made possible today.

In this article we shall discuss the most interesting technical areas and problems in the wireless domain. For a general discussion of wireless networks or introductory information on the same, a good starting point is

Wireless receiver design

A very large part of wireless network engineering is the optimal design of wireless receivers. Indeed, the principal challenge in designing a wireless air interface is to ensure that the complexity imposed upon the receiver (in terms of power, processing requirement, testing complexity, etc.) is manageable. This has become especially important in the era of personal communication systems, where the receiver is embedded inside a terminal whose physical dimensions, power consumption and cost are constrained.

A wireless receiver typically involves multiple components. These include a synchronization block for timing recovery, a demodulator for recovering the transmitted bytes and an FEC decoder for decoding the transmitted bytes and extracting the user payload. In the subsequent sections, we shall review some of the new research areas in receiver design.

DCR

Direct Conversion Radio is a new technique which gets rid of the intermediate frequency stage of a receiver. The received RF signal goes through a high-speed sampler and is directly delivered to a high speed digital receiver, which executes the entire receiving process at that sample rate, without downconversion of any sort. The idea is to get rid of the expensive IF components, which are typically custom hardware and thus, expensive. The challenge is in sampling an RF signal at a high enough rate, that the Nyquist criterion is honoured and then being able to handle this in the digital receiver. For example, for a GSM receiver operating at the 900Mhz range, the receiver is required to handle at least 2 billion samples per second, i.e. one sample per nanosecond. For interesting articles in DCR, please see DCR at UCLA

SDR

Software Defined Radio is a similar process where the entire baseband receiving logic is implemented in general purpose, reprogrammable CPUs/DSPs and other components. The point here is not to eliminate IF, but to use general purpose boards and COTS products and to have a standard computing platforms for all wireless networks and waveforms.

The principal challenge in SDR is to achieve flexible partitioning. Consider an SDR board with two CPUs and a single DSP, which is being designed for a certain network. The receiver logic consists of three processes, a PLL for carrier recovery, a demod function and a decoder. Due to the CPU requirements, the functionality is distributed as follows:
CPU1 : PLL, CPU2: demod DSP:decoding. The corresponding data path was CPU1->CPU2->DSP

All of a sudden, the radio environment deteriorates, or the network changes to a new coding technique i.e from Viterbi to Turbo. Now, the load on the DSP is too high and some functionality must be offloaded to one of the CPUs. This will require a reconfiguration of the functional breakup, as well as the data path. The new data path looks like CPU1->CPU2->DSP->CPU1->DSP. A true SDR system would be able to instantaneously achieve the functional and data path reconfiguration. For more details, see SDR Forum for more details.

Capacity of interference constrained networks

A second large issue in wireless system design is the capacity management of wireless networks. Since wireless networks have significant opex cost, it is imperative to manage the network so as to maximize the capacity.

Why is this specifically a problem in wireless networks? It is because wireless cellular networks are the quintessential shared networks. Effects of a single transmitter cannot be isolated to that user or even that cell, due to the phenomenon of interference. Optimization of capacity of a network thus becomes a global optimization problem. Add to the complexity the fact that different users have different QoS requirements and see different channel conditions and the complexity of the problem rises manifold.

Two interesting works in this area are by [Gupta00], [Gupta01]. Here, they show that the capacity per user in any interference constrained wireless network diminishes. The exact relations are:
where

This result is something akin to Shannon's Limit for wireless systems. It is interesting to see how close modern networks get to this limit. Capacity models for 2nd generation networks [Wigard96] are fairly well understood. Capacity models for 3rd generations are much more difficult, see [Karmani01, Karlsson00], but give interesting insight into these technologies.

Power Control

The close companion of capacity estimation is power control. Power is the one most important resource in modern cellular networks, which are all interference constrained. Power control refers to the optimal allocation of transmit power to all entities in the network such that the aggregate network capacity is maximized. The constraints are that the SIR for all entities must be such that they meet their QoS requirements.

Traditional power control methods involved a network controller receiving feedback from all remote entities and creating an optimal allocation vector for all entities. This system was easier, but with significant drawbacks in terms of signaling requirements and lack of adaptibility to changing conditions. Modern systems use distributed power control. Distributed power control is one of the principal applications of distributed control theory to telecommunications. In DPC, each entity does its own power setting using its own local measurements - the nature of algorithm is such that it satisfies key properties of convergence, adaptibility and fairness.

For more details, check the entries on Distributed Power Control and Distributed Control.

Scheduling in wireless networks

Scheduling problems in wireless networks are tied to issues of wireless QoS. Scheduling for wireless systems is fundamentally different from scheduling in standard terrestrial systems due to the dual requirements of meeting QoS requirements and maximizing capacity utilization. Many modifications of existing scheduling algorithms include, such as Channel State Dependent Scheduling [Bhagwat96]. None of them work very well in wireless networks. This is an area of active research, especially in the emerging WLAN and Wimax networks.

Main article: qos for broadband wireless access networks

Optimal link control

Optimal link control refers to the optimal management of the radio channel in a changing environment. Modern wireless networks use adaptive coding and modulation schemes, where the physical layer can be dynamically adapted to meet changing system requirements. A related subject is hybrid ARQ, which uses physical layer retransmissions and retained state of packets to effectively get double the FEC gain for specific packets which are errored. Hybrid ARQ is one technique in the field of cross-layer optimization techniques.

New Wireless Topologies

Traditional wireless networks were built around a hub and spoke concept. A single network entity acted as the central controller as well as the single point of attachment to the terrestrial network. Modern wireless networks are doing away with this and bringing new topologies to the fore-front.

Adhoc networks involve a flat network of entities which dynamically discover each other's presence and use collaboration to maintain the network.

A more interesting topology is wireless mesh networking. Here a group of entities is configured into a single network with multiple points of attachment to the terrestrial network. The key feature is that all entities can act as relays for each other and relay data streams in the most optimal fashion to the correct attachment point. Also, network control is separate from network routing. Thus the entity which is managing use of wireless resources may be completely different from the one providing ingress/egress services or the entity providing authentication services. In extreme cases, most of the network control functions are distributed and not in any single point of control.

References

[Gupta00] Gupta, Kumar, "The capacity of Wireless Networks", IEEE Transactions on Information Theory, 2000

[Gupta01] Gupta, Kumar, "Internets in the sky: the capacity of 3 dimensional cellular networks", Communications in Information Systems, 2001

[Karmani01] Karmani, Sivarajan, "Capacity evaluation for cdma cellular systems", Proceedings of the IEEE Infocom, 2001

[Karlsson00] Karlsson, Everitt, "Teletraffic capacity of hierarchical cellular cdma networks"

[Bhagwat96] Bhagwat, Bhattacharya, Krishna, Tripathi, "Enhancing Throughput over wireless LANs using channel state dependent packet scheduling", 15th Annual Joint Conference of IEEE, 1996

[Wigard96] Wigard, J.   Mogensen, P.   Johansen, J.   Vejlgaard, B. "Capacity of a GSM network with fractional loading and randomfrequency hopping", 7th IEEE International Symposium on Personal, Indoor and Mobile Radio Communication, 1996

readers have visited this page
Maintainer: abheek.saha@hsc.com