Re: GPRS vs. W-CDMA
The following article provides an excellent overview on how GPRS, not to mention HSCSD and EDGE, works:
expatriot.itgo.com
Those seeking a more detailed run-down on GPRS in particular might want to go through this one:
comsoc.org
As stated in these articles, one of the great things about GPRS (at least in the eyes of carriers) is that, while it requires some hardware-related upgrades to take place within a GSM operator's core network, no such upgrades are necessary for its base stations. This, in turn, is why GPRS is so easy and cheap to implement, which, in turn, is why nearly every major GSM carrier under the sun has been so quick to embrace it.
However, these benefits do come with a major drawback: in order for GPRS to be overlayed in such a seamless manner over existing GSM base stations, it has to conform strictly to the current technical specifications of the GSM air interface, an air interface that wasn't created with high-speed wireless data support in mind.
To see why this is such a major weakness, all that's needed is to go over a few basic technological details (I'm not a techie, so feel free to correct any mistakes I make here).
GSM networks operate via the use of 200 khz. carriers. These carriers are each broken up into eight different time slots. When a GSM subscriber makes a phone call, one of these time slots is taken up for the duration of the call. The world's GSM operators have built out their networks with this requirement kept in mind. When used for data, these time slots traditionally serve up traffic at a rate of 9.6 kbps. However, in the case of GPRS, compression technologies allow this number to be increased by 50% to 14.4 kbps. Any further increase will require the implementation of new hardware, as is required for EDGE.
So how is GPRS able to attain data rates far higher than 14.4 kbps? Simply by using up more time slots. If all eight time slots within a 200 khz. GSM carrier are simultaneously allocated to an individual user, GPRS is able to operate at a speed of 115.2 kbps. The first GPRS handsets that'll be released won't be capable of handling this many time slots at once, and this is why GPRS services probably won't be able to offer such speeds until 2002 or later. In other words, the higher data rates that GPRS will eventually be able to scale to have nothing to do with technological improvements in terms spectral efficiency, and everything to do with the ability of handsets to take up greater amounts of capacity than they can right now; and this is where it gets interesting...
Let's assume that the combination of things such as always-on connections, position location technologies, large color screens, pen interfaces, and voice recognition software is enough to drive a DoCoMo-like spike in demand for a number of carriers. If such a scenario manifested itself, how would it be possible for a GSM operator to grant its GPRS subscribers anything close to 100 kbps speeds? Just think it over: a 100 kbps connection would take up 7x as much capacity as a GSM voice connection. Here, it has to be remembered that GPRS is a data-only upgrade, and doesn't do a damn thing for a carrier's voice capacity. Once again, today's GSM networks have been set up with single-time-slot voice traffic support in mind. In order for significant portion of, say, Vodaphone's subscriber base to be able to actively use GPRS at 100 kbps, it would require the company to significantly increase the number of base stations it operates, and spend billions in the process. It doesn't come across as a sound economic proposition to me. Even if the number of base stations needed is increased by "only" 25% or so, the deployment of W-CDMA, which will most likely offer 8x or greater data and voice capacity within a given cell area, and thus easily allow for at least 70-80 kbps average data rates, would make far more financial sense.
Also, power consumption issues can't be ignored. You don't need an engineering degree to figure out what the use of 7x as time slots within a connection will do to a phone's talk and standby times.
Of course, it could be argued that comparing GSM phone calls to GPRS data connections isn't an apples-to-apples comparision, due to the fact that the latter makes use of packet switching. I agree that, for standard data services (i.e. WAP, e-mail), in which, for every minute that a person happens to "use" the service, only a fraction of the time spent will involve the transmission or reception of data packets, the capacity benefits of packet switching help out greatly in boosting the data rates and limiting the power-related problems; but what about the types of applications which could actually make use of a full 100 kbps, or 70 kbps, for that matter? Things like online game playing, MP3 downloads, and streaming video will require packets to be sent to an individual user without interruption for extended periods of time, and are thus much like circuit-switched voice connections in nature. As a result, in such circumstances, my prior arguments regarding capacity and power-related problems still hold up.
In short, while GPRS should prove more than adequate for most general-purpose wireless data needs, it'd be a mistake to look at its potential peak data rates, and assume that, in the future, it'll also allow a large percentage of a GSM carrier's subscriber base to make use of bandwidth-intensive multimedia applications. Should demand for such applications take off in any substantial manner, as a result of both capacity and power consumption issues, they'll have no choice but to aggressively deploy W-CDMA.
Eric |
