Frequency Hopping on GSM Networks

Last Updated: 03-Mar-2003

Before I can adequately describe Frequency Hopping and its advantages, I need to explain how GSM divides up the available resources for each caller. Back in the days of analog that task was a fairly simple: one caller got one channel. When he handed off, he moved to another channel on another site. With digital technology it's nowhere near that simple, but it isn’t too difficult to understand either.

GSM uses a form of air interface called TDMA, which stands for Time Division Multiple Access. Don’t assume that it’s roughly the same as iDEN or IS-136 just because they too are TDMA-based technologies. They are no more alike than a Yugo is to a Ferrari, even though they are both examples automotive technology. So we'll forget about those other TDMA technologies, and we'll concentrate solely upon the GSM implementation of the concept.

GSM still uses physical channels, but each of those channels is divided into 8 time slots. One user consumes one slot, thus allowing 8 users to be on a GSM channel simultaneously. Each GSM channel is 200 kHz wide, thus giving a 30 MHz license-holder (such as Microcell Connexions) a grand total of 75 physicals channels within their spectrum allotment.

Obviously 75 channels isn’t enough to spread evenly among the 200 some odd cell sites around the GTA, each of which has 3 independent sectors. A sector is an area covering 120 degrees around the site. That’s a grand total of 600 sectors and only 75 channels. Obviously the idea is to reuse channels in multiple sites, and to keep those co-channels far enough apart that they don’t interfere with one another.

The most common type of interference suffered by a dense GSM network is therefore co-channel inference. This means that your phone call is interfered with by another site operating on the same physical channel and time slot. Unlike analog, where co-channel interference would often result in you actually hearing the other conversation, that never happens in GSM.

This is because each call is encrypted, and any attempt by your phone to decrypt another call would result a stream of bits that simply cannot be turned into audio. However, the signal from that other conversation can clobber the signal that you actually want to receive. This will result in audio dropouts, and generally poor audio quality overall.

Another problem facing narrowband radio systems is multipath. This happens when large objects such as buildings reflect your desired signal. The reflection can sometimes be just as strong as the direct signal, and the two can interfere with one another. Although you might think that the reflected signal and the direct signal are actually the same, the fact is they aren’t. The reflected signal had further to travel, and it is therefore out of phase with the direct signal.

It doesn’t sound like there is any viable way to avoid these problems, but that’s where the magic of Frequency Hopping comes in. Who says that our conversation must remain on the same physical channel and time slot for the entire time we are on a particular site? If the network were able to move us from slot to slot, and from frequency to frequency, then we could randomize the effects of interference.

Consider co-channel interference. Not all of the slots are in use on all of the physical channels on each site where they are reused, so although slot 4 on channel 522 might be clobbered by another conversation, slot 7 on channel 530 probably isn’t. So, if we can take each caller on a particular sector and jump them from slot to slot, and from frequency to frequency, then each user runs a far lower risk of suffering from co-channel interference. And when such interference does occur, chances are good that the error correction algorithms can take care of it.

Now consider multipath. Due to the very high frequency of PCS service the wavelength of the signals is extremely short (only a few inches in fact). That means the phase difference on one channel will be quite different than on another. By jumping from frequency to frequency we may only experience problematic multipath for very short periods of time, once again giving the error correction algorithms the chance to clean it up.

But does it work? From what I’ve been able to gather from testing Microcell Connexions in the areas where Frequency Hopping is implemented, and by comparing that to areas where it is not, I would have to say that it works exceedingly well. It obviously does best when site density is high, which explains why it isn't implemented everywhere. Rogers GSM also implements Frequency Hopping across their entire network.

Home