I’ve found that often the smaller, focused conferences can be very enlightening. You are more likely to meet those working “at the coal face” than from the marketing department or in senior management. Conversations tend to me more open and relaxed, meaning you get a better understanding of what’s really going on. This event brought together radio planners from around the world, who shared their current best practices and concerns about how mobile networks are engineered to cope with growing voice and data traffic.
This mobile network optimisation conference attracted some 55 delegates, the majority being radio planners from mobile network operators - mostly from Europe with a few from Korea and the Middle East. Held in Vienna, it was conveniently located and well served for travel from Europe. The hotel, in the exhibition area of the city, was modern with large windows in the conference room – making a change from some of the basement dungeons we sometimes get. There weren’t any exhibition stands – with this size of event it was easier to mingle and network with others.
Rather than give a blow-by-blow account of each presentation, I thought I’d take a few days to reflect on the common themes, both said and unsaid, that were brought up.
Best practice radio planning and optimization
Several radio planners explained how they use monitoring tools and various criteria to determine when and where capacity is running out. They then use their experience and planning guidelines to determine where and when to add capacity to existing sites or commission new ones. Most 3G operators still seem to be at the stage of adding extra carriers (ie using up their full frequency allocation) at their existing cellsites, and starting to build out a few more when they are full. Telenor said they specifically now order a new cellsite at the same time as they install the 3rd carrier for high traffic sites.
It seemed to me that this might continue to work for another year or so, but then either LTE will need to be fully onstream or large numbers of small cells would be required (or both).
Kyivstar of Ukraine, who have clearly made the most from their 2G GSM network, explained that their key metric is the percentage of voice traffic sent using half-rate voice codec. When a cellsite gets busy, the bit rate used for the some voice calls is automatically halved. They have found that running their sites at around 45-55% half-rate is about right – more than that means customers become dissatisfied with the quality, less means that investment might be wasted. This triggers them to add capacity.
Speakers from the planning departments of Telenor Norway, Vodafone Spain, Mobitel Solvenia all provided deep insights into exactly how they do their job. In some cases, this seemed to me to be quite “siloed” with the radio planning teams focusing exclusively on the radio link performance. They used a lot of experience to determine how best to deal with traffic growth, mostly today by continuing to expand current cellsites with extra radio capacity and slowly increasing the number of cellsites.
A growing range of options and choices
The range of technical options facing radio planners seemed to me to be growing even further. Fast dormancy, MIMO, Multi-Carrier, spectrum refarming, LTE are but only a few of the options. Physical constraints such as the number and format of antenna also have to be borne in mind. Femtocells aren’t the only technology competing for the attention of radio planners, but frankly I was surprised not to see them being considered more prominently. It seemed to come as a shock to many delegates when one speaker told them how many femtocells were deployed in the US alone – outnumbering outdoor cellsites by some 2:1.
The first commercial voice and data LTE network
Matthias Sauder, head of Vodafone Germany’s Radio Access department, gave an excellent introduction of how they design and optimize their network. With their LTE network now in full commercial service, they are (to my knowledge) the first to offer commercial voice over LTE. This is because they have initially targeted areas without any wireline broadband (e.g. DSL service), and provided a mains-powered box to deliver internet broadband to those areas. Their existing DSL broadband services, which include a Voice over IP, are also included.
He listed a number of metrics by which their voice and broadband mobile data service is constantly measured and calibrated. As data traffic grows and smartphone takeup increases, the emphasis has moved from downlink speeds to the uplink (including signaling congestion). Vodafone Germany have been very pleased with the performance of LTE in the field, especially in the uplink, where 25Mbit/s is commonly achieved. He commented that the radio side of LTE had gone more smoothly than expected, whereas providing the Ethernet backhaul to these high capacity cellsites, and ensuring the end-to-end high speed transmission was properly configured, was perhaps more difficult than anticipated.
I asked him whether Vodafone Germany would soon be joining the other Vodafone group companies in offering femtocells to their customers or for public access. I understood his concern not to be whether femtocells are technically suitable, but that they may give a confused marketing message from a company that prides itself on outstanding coverage and capacity. It suggests that if a different marketing angle could be found, then there is no technical barrier to their deployment.
The demand for more than 1000MHz of new spectrum
One solution to providing much greater capacity for mobile networks is to allocate a huge amount of new spectrum. This would allow existing cellsites to expand their capacity simply by installing more transmission equipment (and possibly a few more antennae).
Uwe Löwenstein, Spectrum Manager from O2 Germany, gave a detailed report of how the regulators and industry are dealing with the issue.
However I can see at least two major problems with simply increasing spectrum:
- There isn’t a common, worldwide set of frequencies available. Unlike the GSM 900, 1800 and UMTS 2100 bands, there isn’t a common frequency band allocated for LTE. Already more than 30 different bands have been specified for LTE, with probably more to come. This means device vendors will either have to build different versions for these different frequencies, or they will become extremely complex, large and battery draining. Either way, this will prevent the low costs achieved through volume production.
- The combined radio transmission power of these cellsites would increase significantly. Site sharing/co-location with multiple operators would further increase the RF output, potentially exceeding regulatory limits.
WRC (World Radio Conference), due to meet early in 2012, is being asked to find and allocate these new spectrum bands. It’s a very complex problem, but I do hope that some further thought can be given to allocating low power spectrum for indoor use which could then be shared between operators in some way (ie allowing business owners to install just one box that provides great coverage and capacity).
The growing availability of live network measurements and reports
Several vendors spoke of the latest tools and techniques now being used to make better use of optimization. A major shift has occurred fairly recently because the basestation vendors have enabled large data feeds to be exported from their systems. These include measurements from mobile phones making calls together with fault reports of calls dropped and other important events. Several specialist vendors are able to take this data, process it and produce extremely useful reports and analysis.
In the past, radio optimizers would rely heavily on “drive tests” which literally drive a small test measurement lab around in a vehicle constantly recording measurements from nearby cellsites. This is still done today when new cellsites are commissioned. Telenor had a clever solution by installing equipment in the back of two taxis. These make calls every 2 minutes and cover most parts of the city in a weekly basis as part of their normal activity.
However, the trend seems to be moving away from expensive drive testing onto the analysis of live measurements from ordinary customers. With some 80% of mobile traffic originating indoors, this means drive testing outside is less relevant.
Accurate identification of hotspots and poor coverage is now more important
Arieso presented their solution which estimates very accurately where each call or data session is being made from – down to some 50 metres. They showed colour coded maps which highlighted the different geographic use between iPhones (used heavily in the suburbs) and Blackberrys (tightly concentrated in the business district). With the size of cellsites likely to get much smaller, this high accuracy of where the traffic and deadspots are located makes it much easier for radio planners to specify where new cellsites should go.
My own view is that with metro-femto, accurate placement is essential and that tools like this will be needed to achieve that.
Alternatively, why not optimize the traffic content at source?
Two vendors presented another technique to reduce traffic load dramatically – optimizing the video traffic which consumes over 50% of total traffic at times. Mobixell explained how they transcode all video sessions, adapting it to suit the type of device used. Comverse take this a step further by selectively identifying which users are suffering congestion (by calculating whether the video stream is keeping up or not) and video transcoding only those sessions. They argue this is a more cost effective mechanism, because its really only needed for some 10-20% of the time.
What was perhaps surprising, and counter intuitive, was that where these systems were installed, traffic levels actually increased! What seems to happen is that because the video streaming works – and customers find they can watch them properly – then they use it more. It depends what you are trying to achieve. By making the service work properly, you end up with more happy customers. It was said that more people churn to other providers if they have poor performance/service than due to price.
The event gave a good snapshot of where radio planners are today. There is no single solution worldwide, with operators choosing different technologies to meet their own requirements. Geographic topology and spectrum allocation strongly affect planning strategy. For example, the frequent use of RF repeaters in Japan would be much less common in Norway.
Today’s radio planning has to deal with providing the best coverage and capacity for both voice and data. With voice remaining the higher revenue stream, I sensed that it still gets a higher priority when optimizing networks.
New optimization tools are emerging, enabled by the streaming data feeds now available from the radio network equipment itself. The analysis these provide can pinpoint high traffic and poor coverage areas which will become ever more important as cell sizes shrink and smaller femtocells used.
Other techniques, such as video optimization, also have their place. I don’t believe that the radio planners are that interested in these external options though - these need to be sold and promoted elsewhere.
Despite the pre-event workshop given by the Femto Forum, it was disappointing to see the lack of knowledge of what the femtocell industry has achieved so far. Most people I spoke to thought they were residential only, hadn’t realized how many are already in use and didn’t factor them into their own public access network planning solutions.
Perhaps next year, there will be more than one speaker mentioning femtocells and demonstrating how smaller cells are an essential part of the longer term solution to provide both the capacity and coverage demanded by customers.
Sorry for the lengthy report. I hope this is of interest to readers and do please comment below to make any corrections or omissions.