Data vs. Voice

Wireless data always has been next year's big success. Although it has achieved some market penetration, data is only a fraction of the dominant voice market. A few years ago, CDPD was all the rage. Now talk is centering on wireless Internet access. This is one of the major aims of the ITU's IMT-2000 third-generation wireless initiative, which requires 144kb/s for mobile users (and 2,000kb/s for stationary users).

The Internet is closely associated with steadily increasing access speeds. Today, individual users generally are connecting with 28.8kb/s modems from homes and offices and migrating to 56kb/s. Although ISDN has not been a roaring success, one can expect the next big leap to be to xDSL services at multimegabit speeds over existing phone lines, which will become widely installed as availability increases and prices decline. This technology already is available in some areas, offering high-speed access to the Internet at a price that most small information-economy businesses can afford, although it still is expensive for the average consumer. The speeds offered by xDSL likely will eliminate the access technology as a constraining factor, leaving Internet routing and server capacity as limiting factors for some time.

Wireless, by contrast, can offer data speeds only in the 10kb/s to 20kb/s range. However, future developments in both TDMA and CDMA will make speeds of 30kb/s (TDMA) and 64kb/s (CDMA) a reality. The problem is that increasing wireless speeds comes at a direct cost of available capacity for voice users. Digital cellular and PCS carriers are in the business of selling bits, whether they like it or not. If data users require more bits than voice users, they will have to pay more.

Currently, voice users require raw bit rates of 9.6kb/s to 14.4kb/s for TDMA and CDMA vocoders, much less than normally used for Internet access, even today. Not only that, but the bit rates required for vocoders will decline over time, not increase. At the same time, the bit rates required to avoid the World Wide Wait on the Internet will be increasing, probably rapidly.

When digital cellular was born, data rates were of the same magnitude as the bit rate required for vocoders. The TDMA IS-54 vocoder (VSELP) ran at about a 9.6kb/s rate, while data rates on high-speed modems of the late '80s were 9.6kb/s, rising to 14.4kb/s.

Since then, common modem data rates for landline Internet access have risen to 28.8kb/s, 33kb/s and now 56kb/s, while vocoder bit rates have grudgingly been allowed to rise to 14.4kb/s. I say grudgingly because although the 14.4kb/s CDMA vocoder is generally agreed to have better voice quality than the 9.6kb/s alternatives, it also reduces the total capacity of a CDMA system by about one-third. There is little question that if a 9.6kb/s vocoder could be designed that would provide the same voice quality as the current 14.4kb/s vocoder, it would be adopted immediately. Similarly, if the original TDMA plans for a 4.8kb/s vocoder could provide the same voice quality as the current 9.6kb/s ACELP coder, it also would be adopted.

Data-to-Voice Ratio A useful guide to the cost of wireless data services is to look at the ratio of bit rates between data and voice users. Currently, desirable Internet bit rates are about 2 to 6 times the current vocoder requirements. This ratio will rise dramatically in the future. Assuming that xDSL will become widely available at speeds of 1,000kb/s, the ratio will become 100:1. If a practical 4kb/s vocoder is ever invented and xDSL speeds rise to 10,000kb/s, the ratio will be a spectacular 2,500:1. The most optimistic scenario is to assume, as IMT-2000 does, that a 144kb/s data rate would be a satisfactory Internet access speed and that more efficient vocoders will not be invented. Even this case provides a 10:1 data-to-voice ratio for CDMA; 15:1 for TDMA.

At an airtime rate of 10 cents per minute and a data-to-voice bit-rate ratio of 10:1, the data user will have to be charged $1 per minute. Worse yet, the cost ratio will continue to climb over time. It is difficult to believe that users will line up for wireless Internet service at those prices when they can get unlimited access from a wireline modem for around $10 per month.

Capacity Internet capacity requirements theoretically can be reduced by taking advantage of the generally unidirectional nature of sessions and of periods when the user is looking at the retrieved data and not transmitting data. If radio capacity is allocated only when needed, Internet requirements could be halved with unidirectional transmissions and halved again if the user spends equal amounts of time transmitting as looking at his screen.

However, the unidirectional nature of Internet access also is asymmetrical, and capacity could be exhausted in one direction before the other. Voice users need capacity in both directions to make calls, and voice users are more likely to be able to benefit from this approach because conversations generally are symmetrical. Even if a 4-times capacity reduction is achieved in this way, the advantage soon will be lost as Internet users continue to need increasing bit rates to access services that incorporate more graphics, sound and video.

Wireless data usage will continue to expand slowly, but data applications most likely will be those that run at speeds not much faster than a vocoder. This includes the CDPD IS-732 standard (19.2kb/s shared among multiple users) as well as the competing RAM and ARDIS data systems, which run at similar speeds and also are shared among multiple users. Furthermore, data uses that can take advantage of lower priced off-peak hours will be more desirable to carriers. These most likely will be providing computer-to-computer communications rather than human-to-computer.

Although wireline Internet access does have different peaks from voice usage, there is no guarantee that wireless Internet access will follow the same trend. If people were freed from accessing the Internet from the comfort of their own homes, they may well start accessing it more during the day and less in the evenings.

The fundamental problem of consumer-oriented wireless data is that the airwaves are a shared resource, and a phone line is not. If clever engineers can squeeze more capacity out of an existing phone line, the only user who can take advantage of the extra capacity is the person with equipment connected to that line.

The problem with cellular and PCS wireless systems is that even when the capacity of a cell site is increased through more advanced technology, carriers can sell the additional capacity to new customers. There is no advantage in providing the extra bits to their current customers. If wireless carriers were left with significant extra capacity, there would be no motivation to sell access to data users at a lower cost because that simply would drive users toward Internet telephony and erode their current income bases.