Archive for the ‘Industry Insights’ Category

Those of us initiated into the design and workings of cellular systems will know the truth of the matter. Those of us who have tried browsing on a GPRS mobile will know that we rarely get the promised maximum data rate of 171 kbps. Engineers are not to be blamed for this. They know for sure what’s possible and under what conditions. Misrepresentation, as we may call it, comes from the marketing guys.

These days, with the growing affluence of the Indian middle class and their spending power, I find more and more TV adverts for cars – sleek, stylish and spacious. These cars are fitted with ultra-modern features and latest gadgets. In these adverts, the cars zip across a lush countryside. The ride is smooth and noiseless. The roads are wide, free of traffic and spotlessly clean. Overall, the cars promise a great driving experience. Exactly the same cars on Bangalore’s congested, pot-holed, water-logged and narrow roads struggle to live up to expectations.

Just as these adverts have failed to point out that enjoyment of cars is only as good as the roads, phones are only as good as the network and its capacity. Operators have failed to inform subscribers that there is a great deal of difference between average data rates and peak data rates. Most of the time users will not get the data rates that they have come to expect, be it for GPRS or for HSDPA. Let’s take the example of GPRS.

Cellular systems were first designed for GSM. GSM networks were deployed such that there is enough C/N (Channel to Noise ratio) for the required BER at the cell edge. This meant that users closer to the BTS often had better quality. However, GSM did not utilize this to provide higher capacity, higher data rates and better QoS for subscribers. This didn’t matter much because GSM offered CS voice calls and CS data at only 9.6 kbps. Then GSM 14.4 kbps data service was introduced. The difference with the 9.6 kbps was that this had less error protection by puncturing more bits [TS 45.003]. The end result was that GSM 14.4 kbps did not offer the same cell coverage as 9.6 kbps although it did deliver the promised bit rate because of the circuit switched nature of the connection.

Enter GPRS. Four coding schemes were introduced, the difference being the level of error protection: CS-1, CS-2, CS-3 and CS-4. Cell coverage decreases as we move from CS-1 to CS-4 and the coverage is worse without frequency hopping. So only users close to the base station will be able to use CS-4 at bit rates as high as 171 kbps. Even this is an ideal situation which can happen only if neighbouring cells are lightly loaded (lower interference), current cell has sufficient spare capacity so that all the slots in a frame can be allocated to the user requesting the high data rate.

The situation is similar with HSDPA. When there are multiple users in the same cell, it is unlikely that a single user will be allocated all the 15 codes. Even when there is only one user, he has to be close to the Node-B for ideal channel conditions. At the cell edge, it is has been shown that HSDPA can offer only 250 kbps with 15 codes and HSDPA using 80% power [1, Chapter 11]. In fact, at cell edge DCH/DSCH can achieve 384 kbps mainly because of soft handover gain. Soft handover is not possible with HSDPA.

It is certain that GPRS, E-GPRS, HSDPA and HSUPA all increase network capacity. This improvement is advertised to the user as higher data rates. This is true in terms of average bit rates which are significantly lower than peak rates. Peak data rates are possible only under good channel conditions that enable higher modulation and lower FEC coding. So although spectral efficiency has improved with newer access technologies, the difference between average and peak efficiency has also increased. The exceptions are AMPS and GSM for which cells were planned for what they were meant to deliver. Figure 1 summarizes the efficiency gap [2].

Figure 1: Average vs Peak Spectral Efficiency over Time

Average vs Peak Spectral Efficiency

This may be one way seeing why the cost of mobile/wireless data has always been not as competitive as fixed data. If it were cheap, lots of subscribers would try it for a start. This would increase the load, which would reduce the effective data rate per user. Reduction in data rate per user occurs because available capacity has to be shared and more users means more interference. The user would be disillusioned with the service when he had been promised a much higher rate. Thus, it is only natural that data is for premium users and not for the mass market. If all mobile users were to start using data at the moment and expect speeds possible with their ADSL modems, today’s cellular networks are not ready to deliver. We have the technology but not the delivery mechanism. Operators have to perform cell splitting and provide HSDPA at the level of microcells rather than at the macrocell level. This would result in more handovers. They would have to upgrade their backhauls (Iub, Iur and Iu). Naturally, link budgeting has to be revised along with other factors associated with cell planning.

Will it make business sense to make these upgrades? To answer this we need to look at the predicted growth of data when compared to voice so that a proper cost-benefit analysis can be made. Hopefully, this will be another post. But then, there are also competing technologies such as WiMAX/WiFi (microcell) and femtocells (picocell).


  1. Harri Holma and Antti Toskala, WCDMA for UMTS, Second Edition, John Wiley & Sons, 2002.
  2. “What Next for Mobile Telephony?”, Agilent Measurement Journal, Issue 3, 2007.

Read Full Post »

« Newer Posts