Unless someone can miraculously change Shannon’s Law,
wireless data will need increasing attention from engineering departments.

When the concept of using wireless networks for provision of commercial packet data services was first addressed 15 years ago, the generally held belief was that wireless data networks would be designed and engineered in much the same way as voice networks.

To some extent that process depends upon the air interface technology being used, but in general the rules for engineering of voice networks are pretty constant. First, users are where they are, not necessarily where it is most convenient to serve them. Second, in order to provide acceptable completeness of coverage, the RF path loss between base station and where users are must be held below a certain maximum. Third, the overlap in coverage from neighboring base stations must be minimized in order to limit mutual interference and thus maximize capacity. Finally, in order to manage costs, all of the above must be done with the fewest possible number of base stations.

The art of wireless network engineering isn’t nearly as well appreciated by the public as the latest cool handsets or price plans, but the overworked and underpaid folks who labor in the trenches of engineering departments have largely been responsible for the remarkable success of the industry.

They have delivered levels of coverage ubiquity and spectrum efficiency for voice service that were unimaginable when the first cellular networks were deployed. But in the past few years, most of the efforts of engineering staffs have been refocused on deployment and optimization of 3G data networks, and what they and the industry have learned is that there are in fact substantial differences between voice and data networks, and that the design practices that were perfected for voice may not suffice for optimization of data.

CHANNEL QUALITY
The critical difference between voice and data that so affects network design is in required channel quality. In voice service, particularly using digital technologies, there is a specific carrier-to-noise (C/N) or carrier-to-interference (C/I) ratio required to produce the best possible audio quality as perceived by the listener.

Any improvement in C/I or C/N above that level will yield no benefit in terms of service quality. This characteristic is a main reason why CDMA is such an ideal technology for voice service. In CDMA, transmit power levels are constantly adjusted so that the received C/N or C/I is just good enough to deliver the best possible audio, thus maximizing channel capacity.

But even with other air interface technologies like GSM, voice service capacity can be substantially increased by providing only the channel quality required for “perfect” audio performance.

In packet data networks, there is no similar “required” channel quality. Or, more accurately, as defined by Shannon’s Law, error-free data throughput (for a given channel bandwidth) is highly dependent on channel quality. Some arbitrary “acceptable” throughput rate for a given network will require a specific minimum C/I or C/N, but that number for 3G networks is almost certainly going to be much, much higher than what is required for “perfect” audio in voice service.

This differing relationship between channel quality and perceived performance has two huge implications for engineering of wireless data networks. First, higher path loss values that are nonetheless perfectly acceptable for voice service may yield agonizingly slow data speeds. In urban areas, this means that “acceptable” in-building data service will be much more difficult to achieve. Second, even moderate RF interference will depress data rates, putting even greater pressure on engineers to reduce overlap in propagation from neighboring base stations.

Unfortunately, as every engineer knows, increasing building signal penetration and reducing propagation overlap are pretty much conflicting goals in wireless network design.

This dilemma facing wireless data network engineers is only going to get worse with the eventual migration to 4G technologies. While some of the higher data rates they will nominally offer will come from greater channel bandwidths, much will depend upon use of higher maximum modulation levels (i.e. more bits per second per Hertz of channel bandwidth) than are used in 3G technologies. But there is no magic in 4G technologies that allows them to violate Shannon’s Law; the higher modulation levels will require even higher C/I or C/N levels for error-free reception.

Fortunately, developers of 4G technologies are well-aware of the need to provide very high channel quality in order to deliver promised data throughput rates. There is considerable hope that physical layer handling technologies will make the network engineer’s life a little easier.

Two technologies that are drawing particular attention are multiple input-multiple output (MIMO) antenna systems and “smart” base station antennas. The role of these technologies is to effectively increase receiver sensitivity (for improved C/N) and reduce interference (for improved C/I). In computer simulations and controlled field trials, these technologies appear to be quite promising, but nobody really knows how effective they will be in mature networks serving hundreds of millions of users.

At best, these and other physical layer technologies will probably only provide limited improvement in effective channel quality. Absent a miracle breakthrough that somehow repeals Shannon’s Law, wireless data network engineers will continue to face increasing challenges as throughput expectations grow.

If they eventually manage to achieve the levels of user satisfaction that we have today for voice networks, it will certainly be time to give them plenty of credit. Maybe even a raise.

CORRECTION: Last month, in discussing competitive 4G technologies LTE and UMB, I erroneously wrote that LTE retains some CDMA channel structure and that its future is uncertain. Actually, it’s UMB that uses CDMA in the uplink channel and that, because of the Verizon decision to go with LTE, has a cloudy future.

Drucker is president of Drucker Associates.
He may be contacted at edrucker@drucker-associates.com.