Why Are African Operators Still Rolling Out 3G Networks in 2017?

I attended the Telecom Infra Project Summit last week in Santa Clara where the OpenCellular Project (yes, a project within a project) announced a new grants program to make their OpenCellular base stations available for testing and piloting to organisations researching and/or implementing affordable access solutions.  This is a great opportunity for groups who may already be doing work with WiFi or dynamic spectrum but see the need for affordable GSM services in the areas in which they work, also for universities in the Global South to research cutting-edge low-cost access technologies.  For the time being, this will only apply to the 2G OpenCellular base stations that are currently in production but will soon be extended to LTE.  The prototype LTE OpenCellular (OC-LTE) base station was on display at the event.  The LTE platform is currently undergoing testing in preparation for production within a few months.

The fact that the OpenCellular project has chosen to develop 2G and LTE platforms but not 3G is what I want to dwell on in this article.  2G networks are mostly oriented around voice services although they are capable of data services at basic speeds and capacity.  There are still millions and millions of people without access to simple voice services in the world and the OpenCellular 2G base station directly addresses the economic limitations of extending access into sparsely populated areas where traditional mobile network operator models are unsustainable.  The decision to develop an LTE version of the OpenCellular base station makes perfect sense as LTE is the current standard for mobile broadband and offers better digital performance than 3G networks.

In the course of the event, I watched Navindran Naidoo present MTN’s plans for OpenCellular in which he emphasised the importance of legacy mobile technologies in Africa and the continuing need for 2G networks.  He also urged the OpenCellular project to develop a 3G version of their technology. Interesting!  And it is not just MTN that continue to see 3G as a relevant technology.  Only last week, Airtel in Tanzania announced a significant investment in expanding their 3G network coverage in the 900MHz band into rural areas.  In September, the Ghanaian regulator actively encouraged licensed operators to offer 3G services in the 900MHz band.  Airtel in Uganda completed a massive upgrade to 3G on their network in 2016, again in the 900MHz band.  Perhaps you’re like me and wondering, why would you choose 3G when you could have LTE? The answer is not as simple as it appears.

Slide from MTN presentation at TIP Summit 2017. Red ellipse emphasis is mine.

LTE Is A Leap

LTE stands for Long Term Evolution and in evolutionary terms it is a significant change.  The voice components of 2G and 3G networks are based on circuit-switched networks which allocate a dedicated ‘circuit’ for every call on the network.  The part of 2G/3G networks that serves data is separate from the voice and SMS side of the network.  LTE on the other hand is a completely packet-switched (IP-based) network where all services voice, SMS, data are delivered via IP packets. This allows for much more efficient use of the wireless network but means that a new protocol, Voice over LTE (VoLTE) is needed to deliver voice services over the network.  This can present a problem for network operators who still want to deliver basic voice and SMS services to the millions of users in African countries who still have feature phones that are only designed to operate on 2G or 3G networks. In many cases, operators deploying LTE networks are obliged to maintain 2G networks in parallel in the same areas in order to ensure full service delivery.

Deploying a 3G network might seem like a better option to operators in that 3G technology is usable by feature phones while still providing data services that can be considered broadband; a cost-effective compromise.  They downside is that it may end up costing the operators more in the future when they inevitably have to upgrade to LTE as smartphones proliferate and demand for broadband increases.

Voice call quality may also be a factor in decision-making on LTE vs 3G in that VoLTE is not regarded by some as delivering the same level of quality and reliability of voice services as 2G/3G circuit-switched networks.  Naturally opinions vary on this and VoLTE is only likely to improve over time but for operators that are already under pressure from their regulators on Quality of Service, this may be a near-term factor.

Spectrum Re-Use

Another motivation for operators to deploy 3G networks, especially in the 900MHz band, is that it is typically spectrum they already have a license for.  Regulators have been slow to make new spectrum frequencies available to operators to enable more broadband services.  Re-using existing GSM spectrum allows the operator to move forward in upgrading their broadband services without having to wait for regulatory approval of new frequencies.  Given the millions of dollars in fees that are being levied for new spectrum bands, re-using existing frequencies also represents a significant cost savings for operators.  The rub is that using the 900MHz band may mean taking down some of their existing 2G network.  Some operators will need to implement a technology that offers backward compatibility for feature phones and 3G fits that bill.  I also have the impression that new 3G technologies can be upgraded to LTE at a lower cost than older 3G networks.

It Depends

Thus there is no right or wrong as to whether low-cost 3G or LTE is the right technology to develop and deploy.  It depends on the operator’s existing networks, how old they are, what technologies dominate in terms of handsets, and what radio frequencies are available to the operator.  It also depends to some degree on how bold and disruptive the view of the operator is.  Consider Reliance Jio in India who have bet everything on LTE and an all IP-based platform. This forward-looking approach has allowed them huge efficiencies in moving billing and other services into the cloud.  It also positions them for future network growth and evolution.  This is an easier decision for a newer operator entering into a massive build-out than it is for an older operator with a vast legacy network.  Now, a dual-radio OpenCellular base station that can serve 2G and LTE, that would be an interesting idea.

The Failure of Spectrum Auctions in Africa

Spectrum auctions are widely regarded as “best practice” in the assignment of wireless frequencies where demand exceed availability yet, as of 2013  among African countries, only Nigeria had successfully conducted a spectrum auction. This is perhaps not surprising as spectrum auctions are notoriously difficult to run well from the point of view of ensuring fair play and even more so from the point of view of ensuring the growth of competition. The last four years has seen a number of African countries embracing spectrum auctions while others have engaged in processes that appear more ad hoc and negotiated.



In late 2013 the Nigerian communication regulator (NCC) announced a spectrum auction for 30MHz of 2.3GHz spectrum. The auction attracted only two bidders and was won by a new entrant consortium called Bitflux. Bitflux paid just over the reserve price of 23 million USD for the spectrum license. At the time this was lauded as a success in bringing a new market entrant into the field of LTE services in Nigeria. By mid-2015 however pundits began to wonder why Bitflux had yet to offer services. In late 2016, the commencement of commercial rollout was announced but, as of early 2017, little evidence of widespread roll-out appears to exist.  There is speculation that they have struggled to find investment to underwrite the roll-out of their network.


Later in 2014, the regulator attempted to launch an auction in 2.6GHz. This was withdrawn and re-attempted in 2015 but again withdrawn. Finally in March of 2016, the regulator announced an auction of spectrum in the 2.6GHz band. In total 14 lots of 2x5MHz spectrum (140MHz of spectrum in total) were put up for bid. By the end of the bidding process only one operator, MTN Nigeria, was willing to meet the reserve prices of 16 million USD per lot. MTN bid for 6 lots, paying a total of 96 million USD for 60MHz of spectrum. The remaining spectrum remains unsold.



In April of 2013, the Mozambiquan regulator announced the auction of five lots of 2x5MHz (a total of 50MHz) of 800MHz spectrum with a reserve price of 30 million USD per lot. The auction did not attract any bids and was widely perceived to have an excessively high reserve price. The auction was quietly withdrawn and no subsequent auction has been attempted by the regulator.

South Africa

2.6GHz, 3.5GHz, 800MHz

This is a saga that is almost too painful to tell.  Since 2010, the South African regulator has attempted to convene a spectrum auction three times. The first two attempts were in 2.6GHz and 3.5GHz and the most recent in 800MHz, 2.6.GHz, and 3.5GHz. Each time the auction has been withdrawn with the most recent one being cancelled in February of 2017. There are multiple causes of these serial auction failures. In part, it can be attributed to industry push-back sparked by the regulator’s embedding of Black Economic Empowerment objectives into the spectrum auction design. It can also be attributed to a lack of coherent vision from the Ministry of Communications which has seen seven different ministers since 2009 that has, in turn, led to disputes between the regulator and the ministry. The Ministry’s current vision of removing all exclusive-use spectrum in favour of a national wholesale network has attracted widespread criticism leaving the current policy and regulatory environment in turmoil. As of early 2017, no resolution is in sight.



In 2015, the Ghanaian regulator announced an auction of 800MHz spectrum, offering 2 lots of 2x10MHz spectrum ( a total of 40MHz) with a reserve price of 67.5 million USD per lot. While local companies were encouraged to participate, none of the three Ghanaian companies that registered for the auction were able to meet the reserve price. The only company to meet the reserve price for a single lot was Scancom (MTN) resulting in an effective monopoly for MTN in the 800MHz band. The regulator has announced plans to attempt to auction the remaining spectrum with intention of using auction proceeds to fund the roll-out of digital terrestrial broadcasting infrastructure.



Kenya is a very interesting story which did not involve a spectrum auction. The assignment of spectrum in the 800MHz band in Kenya began in 2014 with a request from the Kenyan government to the largest operator, Safaricom, to build a national police communications network. Safaricom initially agreed to pay 56.2 million USD and build the requested network in exchange for access to 2x15MHz of 800MHz spectrum. After complaints from Airtel and Telkom Kenya, the regulator compelled Safaricom to relinquish 2x5Mhz of spectrum so that each of the three incumbent mobile network operators would be assigned 2x10MHz of 800MHz spectrum each for a total of 60Mhz of spectrum. The three operators have agreed to each pay 25 million USD for the spectrum licenses. The cost of the national police network has now been disaggregated from the spectrum sale.


800MHz, 1800MHz

In late 2015, the Senegalese regulator announced an invitation to apply for LTE spectrum in the 700MHz, 800Mhz, and 1800MHz bands: specifically, 3 blocks of 2x30MHz in 800MHz, 4 blocks of 2x20MHz in 700MHz and 3 blocks of 2x30MHz spectrum in 1800MHz. The reserve price for a 20 year license was set at XOF 30 billion or approximately 50 million USD. The reaction from operators in December 2015 was to draft a collective letter to the regulator to express their concern over the high reserve price for the spectrum. Their effective boycott of the auction resulted in a stand-off between the regulator and operators. This was resolved in June of 2015 when the regulator restarted the licensing process having negotiated a deal with the former fixed-line incumbent operator, Sonatel, to pay XOF 32 billion or 53 million USD for 2x10MHz of spectrum in the 800Mhz band and 2x10Mhz in the 1800Mhz band. The 20 year license commits Sonatel to the provision 70% population coverage within five years and 90% coverage within ten years.


900MHz, 1800MHz

In mid-2016, the Egyptian regulator announced the availability of 40MHz of spectrum to existing operators at price of approximately 50 million USD per MHz. Operators protested the high price as well as the relatively small allocation of spectrum as well as the requirement that 50% of the license fee be paid in USD. Only Telecom Egypt accepted the terms set by the regulator agreeing to pay 7.08 billion Egyptian pounds (797 million USD) for 5MHz in 900MHz and 2x5MHz in 1800MHz. The regulator held firm with the other operators and ultimately came to agreement with all four operators, with Orange and Etisalat each receiving 10MHz of spectrum and Vodafone 5MHz. Total revenue from the spectrum sale exceeds 1.9 billion USD.


If the purpose of spectrum auctions is to make wireless spectrum available in an equitable manner that promotes competition, I think it is fair to say that they have not worked terribly well in Africa in the last five years.  The 2.6GHz auction in Nigeria and the 800MHz spectrum auction in Ghana resulted in the sole successful bidder in both cases being network giant, MTN.  This was clearly not the optimal outcome in terms of increasing competition. Even a seemingly positive case such as the 2.3GHz auction in Nigeria requires closer analysis.  The auction was won by Bitflux, a new market entrant, ostensibly at a relatively low cost per MHz for the spectrum acquired, compared to other auctions in the region.  Three years on, Bitflux does not appear to have rolled out substantial infrastructure. It may be that the financial burden of spectrum auctions is only bearable by incumbent operators with greater resources. The GSMA have urged caution in the use of set-asides and other mechanisms to encourage participation of new market entrants in spectrum auctions because the high rate of failure of such strategies. Kenya’s somewhat chaotic path to the assignment of 800MHz spectrum seems to have resulted in an outcome that has worked out well for the incumbent operators. Other countries seem to have had less optimal outcomes. South Africa’s ongoing attempts to integrate its transformation agenda with spectrum has resulted in a roadblock which does not have any obvious resolution at this point. The auction failure in Mozambique in 2013 due to the high reserve price for a small amount of spectrum has ensured that 800MHz spectrum will like fallow for at least five years. The economic cost of that failure is arguably higher than the original expected revenue from the auction. In Egypt, the hard line taken by the Egyptian regulator may be an economic windfall for the Egyptian treasury but time will tell whether it results in more affordable access in the country. The lack of willing participation of operators in spectrum auctions suggests that spectrum auction reserve prices need more careful evaluation.

The focus on the assignment of long term, exclusive-use licensed spectrum also ignores other critical metrics of spectrum efficiency such as cell size. Reduction in cell size can result in significantly greater increases in spectrum efficiency than simple assignment of new spectrum frequencies. Once seen as a limitation of WiFi, its small cell size of approx. 100m makes for extremely efficient use of a very limited frequency, allowing for its re-use over multiple cells in a given area. Similarly, new generation, low-cost GSM technologies typically operate at much lower power output and cover smaller cell areas than traditional mobile networks. Yet the efficiencies represented by these more granular approaches to access typically do not receive the strategic attention that monolithic licensing approaches do. The large amounts of money associated with spectrum assignments may be part of the problem, attracting the attention of governments and network operators alike. Civil society groups are often challenged both financially and technically to engage in national strategic debates on spectrum although there are signs that this is beginning to change as spectrum is increased recognised as a critical roadblock to affordable access.

From a technological infrastructure perspective we are seeing a shift to more pervasive availability of affordable backhaul infrastructure both through the spread of fibre infrastructure as well as new generation satellite technology. Combined with dramatically less expensive wireless access technology, there is an opportunity for a more granular and dynamic approach to spectrum management as a complement to traditional long term license strategies. It is possible to envision a set of regulations that enable local access providers through the use of a combination of unlicensed, dynamic and traditional spectrum licensing aimed at increasing access in unserved regions.

Long term, exclusive spectrum licenses continue to play a critical role in increasing affordable access to communication. The large amounts of money associated with spectrum auctions and licensing has focused a great deal of public attention on them. Observation of recent spectrum assignments in Sub-Saharan Africa reveals that spectrum auctions are often not successful in their aims of increasing competition. However, the flaw may not be with the auctions themselves but with the reserve prices set by regulators. Other economic priorities may tempt decision-makers to focus on an immediate and lucrative financial windfall from a spectrum instead of the longer term (and more diffuse) economic benefits that increased affordable access can bring.

How can this problem be addressed?   It is evident that there is no one-size-fits-all answer but here are some general principles that might be embraced based on the above:

  • Convince policy-makers and regulators that the long term economic benefits of making spectrum available affordably and immediately are dramatically greater than any short-term windfalls from spectrum auctions.  Keep spectrum fees modest.
  • Skip the auction and assign roughly 3/4 of popular spectrum bands to existing operators.  This obviously needs to be adjusted based on the amount of spectrum available in a given band and the number of operators.  Each operator needs to get enough spectrum to deliver effective broadband services.
  • Enforce a use-it-or-share-it policy, whereby licensed spectrum that is not in use by the primary spectrum holder is made available.  This could be through spectrum leasing and/or secondary use policies.
  • Keep a quarter of the band for innovation.  This could be a national wholesale network, dynamic spectrum assignment, industry self-regulation within a band or something else.  The current spectrum management paradigm is too static to keep pace with technological change.  Regulators need to make space in order to innovate what comes next in spectrum management practice.

Dynamic Spectrum in 2015 Wrap-up of the Dynamic Spectrum Alliance Global Summit

nsrc-logoThe Calm Before The Storm” was the subtitle I used to describe my experience of the Dynamic Spectrum Alliance’s Global Summit in 2014. Fast forward to 2015 where I have just returned from this year’s Global Summit in Manila, which I attended on behalf of the Network Startup Resource Center (NSRC). Is there a storm of manufacturing and deployments, one year on? No, not yet but interesting things are happening. Here’s what I learned.


The success of dynamic spectrum is an interlocking puzzle with each piece from manufacturing to regulation to deployment dependent on each other. Dynamic spectrum hardware is a keystone puzzle piece though without which the rest of the picture doesn’t make sense. This year was notable in that virtually every manufacturer in the dynamic spectrum market was represented at the Summit. This along with the dramatic increase in participant turnout at the event was a very positive indicator.

Last year we heard about a new tri-band chipset that was being developed through a partnership between Mediatek and Aviacomm. While that is still on the cards, it has proved to be a bit of a disappointment as Mediatek don’t sound as committed to it as they did last year, in spite of it still being scheduled for a Q4 release this year.

In contrast, while continuing to partner with Mediatek, Aviacomm have also announced their own product based on their AF3010 chipset. The AF3010 was in trials in 2014 but now in full production. Aviacomm equipment based on the AF3010 is being used as part of a national connectivity project by the government of the Philippines. It is interesting that while this equipment is compliant with FCC standards for TVWS operation, it is based on the more common 802.11an WiFi standard rather than the emerging 802.11af or 802.22 Television White Spaces (TVWS) standards. This simplifies manufacturing for Aviacomm and lowers costs. The trade-off is that it is not interoperable with other TVWS equipment. As the market evolves, standards are highly desirable for consumers as it prevents vendor lock-in and encourages good behaviour among manufacturers. However, in the early stages of market development, it may be that the perfect is the enemy of the good and having affordable equipment ready to deploy may trump the need for standards which can evolve in tandem with the market.

Other vendors at the summit, including Carlson, Adaptrum, 6Harmonics, and Japan’s National Institute of Information & Communication Technology (NICT), all announced that they would have new products shipping in 2015. Significant here is that many of these 3rd generation products are moving from FPGA to ASIC  manufacturing. Without getting too technical, this is an indication of the maturation of the technology from advanced prototyping for small markets to mass market. ASIC manufacturing represents a significant drop in production costs.

Carlson appear to have partnered with chip manufacturer Saankhya Labs for the next generation of their equipment. Even more interesting is their announcement of a Crowd Investment campaign seeking a broad range of investors to support their new platform.

I came away from the manufacturer’s session at the Summit somewhat humbled by their vision and willingness to stake their reputation and finances on a dynamic spectrum enabled future. I too believe this is inevitable.

Geo-location Databases

As with the manufacturer’s session, I was impressed with the turnout of companies including ATDI from France, FairSpectrum from Finland, SpectrumBridge from the US and the Council for Scientific and Industrial Research (CSIR) from South Africa.
The big question I came into the session on geo-location databases with was whether they were necessary for countries with loads of empty UHF spectrum and comparatively modestly-resourced regulators. Many countries in sub-Saharan Africa fall into this category. Do they need the all-singing-and-dancing dynamic spectrum management solution before deploying dynamic spectrum at scale? The consensus seemed to be no, that regulators could proceed without having to have a complete geo-location database solution in place. A sensible suggestion was made that regulators that choose this route should still implement a type-approval regime that insists that all dynamic spectrum equipment should have support for geo-location so that, as the regulator adopts geo-location database technology, existing deployments can easily be adapted to a database approach.

Peter Stanforth of SpectrumBridge emphasised the extent to which emerging markets could leverage existing investments in geo-location database technologies to be at the forefront of spectrum management as a means of future-proofing spectrum regulation. He spoke of the work that had been done by the DSA in development “model rules”  for regulators to build on combined with the learning that has been gained from places like the US, UK, and Singapore in the implementation of geo-location databases.

Ntsibane Ntlapa gave an African perspective on the problem recognising the capacity issue for regulators that may not have the initial spectrum occupancy data needed to populate a geo-location database. He also spoke about their bilateral partnership with the government of Ghana in supporting a TVWS pilot there.

Most interesting for me, was Heikki Kokkinen of FairSpectrum who gave me an insight into the future hybridised world of spectrum management where licensed and unlicensed spectrum are more of a continuum than a binary option for regulators. To illustrate this he asked us to imagine how a geo-location database might enhance a licensed spectrum regime. If those with spectrum licenses were obliged to report to a geo-location database, one could imagine the dynamic assignment of spectrum within a licensed band such that exclusivity is maintained but specific location within a spectrum band could be managed by the regulator to make most efficient use of the band. This hints at the way that geo-location databases could solve many spectrum management challenges.

Silicon valley investor Marc Andreesen famously said “software is eating the world”. It seems inevitable to me that software will eat the problem of spectrum management and those countries that are in the forefront of this are going to have a huge leg-up.

The Regulators

As in other areas, the turnout from regulators at the Summit was impressive with representatives from Singapore, Indonesia, the Philippines, Benin, Bhutan, Bolivia, Ghana, Nigeria, Botswana, Nigeria, the United Kingdom and the United States. It was fascinating to see how leadership is shifting in this space. Ira Keltz, Deputy Chief, Office of Engineering and Technology for the FCC gave a talk in which he clearly framed dynamic spectrum as the future of spectrum management. He made the point the scope for traditional spectrum licensing in the US is almost exhausted. By contrast, Rachel Clark, Director of Spectrum Policy for OFCOM in the UK gave a talk that was timid by comparison. Previously the global leader in dynamic spectrum regulation, OFCOM seems to be moving ahead with dynamic spectrum regulation but in an extremely cautious manner. One could almost imagine that the broadcast lobby, so active in the US, has caught up with the UK.

Regulators from South Africa and Malawi, both of whom who were present last year, were conspicuous by their absence this year. For South Africa, this can be at least partially attributed to the government’s failure to provide competent leadership to the sector. Dynamic spectrum is still part of the regulator’s plans but is competing with various other urgent priorities. Malawi is still on track for implementing dynamic spectrum regulation this year.  Microsoft has provided equipment for new pilots in Malawi and has offered support to the regulator in the development of regulation.

The most impressive regulator from Africa this year was Henry Kanor, Director of Engineering for the National Communication Authority (NCA) in Ghana. Ghana has enthusiastically embraced dynamic spectrum with no less than four pilots currently underway, one of them already operating commercially. Kanor emphasised that the market was simply not providing services to certain key areas and that dynamic spectrum offered the potential to address this gap.

Botswana was not far behind with Tebogo Ketshabile of the Botswana Communications Regulatory Authority (BOCRA) making the key point that current spectrum regulation is not keeping up with the pace of technological change. With support from Microsoft, they have authorised a 3-year pilot.


There were two things that really stood out from me coming away from the event.  One was the announcement of the Philippines government of a national broadband roll-out that would embrace WiFi, dynamic spectrum technologies, and fibre. A 30 million dollar project designed to provide public access to Internet. Both Aviacomm and NICT are involved in this project. Its vision lies not just in the goal of affordable access for all but also in the recognition that there is no one-size-fits-all technology. Affordable access will be achieved by technological diversity which allows the right technology to fit the purpose.
The second big takeaway for me was just how important a role Microsoft is playing in this landscape. Their support for pilots across the world and their support for regulators is playing an absolutely pivotal role in unlocking the potential of dynamic spectrum as an affordable access technology. This is really making a difference. It is on a par with Google’s vision for metro fibre in Africa.

Thanks to NSRC who made my participation in the Dynamic Spectrum Global Summit possible.
Photo copyight Dynamic Spectrum Alliance, reproduced with permission.

How To Let GSM Serve The People That Other Networks Can’t Reach

Mobile telephony has transformed access to communication in emerging markets.  In 1994, there were more telephone lines in New York City than in the whole of Africa.  These days, about two thirds of the population in sub-Saharan Africa is covered by a mobile phone signal.  That’s an amazing and profound change yet there are still millions of people who don’t have access to mobile networks.  Why is that?  Typically, the reason boils down to the fact that a business case doesn’t exist for mobile network operators to extend coverage to that area.  That might be because it is sparsely populated,  or in a hilly region that is expensive to cover, or because ability to pay is very low or some combination of the above.  Here are some low-risk things that regulators can do to help GSM services reach the unserved.

Social-Purpose Licenses For Rural Access

Low-cost alternative GSM technologies have existed for some time and there are a variety of startups in this space including Range Networks, Vanu, Endaga, Fairwaves, and others.   Some are based on popular Open Source software projects like OpenBTS and OpenBSC.  Manufacturers like NuRan Wireless are producing low-cost radio platforms that can serve as robust platforms for these Open Source initiatives.  The result is that it is possible to put up a DIY GSM base station for a few thousand dollars.  What holds these startups back though is the fact that the popular GSM spectrum bands have largely been assigned to existing Mobile Network Operators (MNOs).   Low-cost GSM startups like those mentioned above are left with the option of trying to sell their technology to incumbents, whose supply chains are pretty closely tied to big equipment suppliers like Huawei et al.  Or they can try to go it alone and find a sympathetic regulator but, to date, sympathetic regulators have been a bit thin on the ground.

That all changed last week, when the Mexican communications regulator, IFETEL, published their frequency plan for 2015.  IFETEL have set aside 2 x 5MHz of spectrum in the 800MHz band (in the range of 824-849 and 869-894 MHz) for “social” use.  The criteria are that the communities being served must be less than 2500 people or be designated as an indigenous region or designated as a priority zone.  Small cell GSM to rural areas has already been going on for some time.  Rhizomatica, a non-profit organisation that has been providing GSM services to indigenous communities around Oaxaca since 2012.  Until last week they were doing it under a special dispensation from IFETEL but now the allocation of spectrum to this purpose is official and any organisation may apply for access to this spectrum under the conditions specified.  2 x 5MHz spectrum is not a huge amount of spectrum compared to the big operators where 2 x 10MHz or more is common but it is plenty for smaller communities.

Rhizomatica have played a critical role both in engaging with the regulatory change process and in providing the regulator with a successful model to base their regulatory reform on.  What IFETEL have done is groundbreaking in terms of regulation. This opens the doors to community entrepreneurial efforts to solve their own connectivity problems.

Rhizomatica are spreading rapidly in the Oaxaca region serving over 15 communities now. Peter Bloom, the founder of Rhizomatica, had this to say about IFETEL’s announcement:

 Rural areas have traditionally been no-go areas for incumbent telcos, and this forward-looking approach by the IFETEL allows other actors, including the communities themselves, to provide affordable access to communication services by having direct access to spectrum. This will lead to more people being connected which will bring both social and economic benefits to underserved areas.  If the role of regulators is to maximize the benefit that society obtains from the use of radio spectrum, then this is a step in the right direction.

When I look at the spectrum allocations for countries in sub-Saharan Africa, most of them have sufficient wiggle room in their GSM allocations that they could do this.   The allocations of spectrum are small enough that this is unlikely to infringe on the requirements of any existing operators.  What if remote communities all over Africa began to manage their own communication infrastructure?

Unlicensed GSM

Licensed spectrum is not the only way to take better advantage of GSM technology.  Communication regulators in the Netherlands, Sweden, and the UK have created unlicensed spectrum bands for GSM use in the 1800MHz range. According to this OECD report, more than 3,000 organisations (hospitals, stadiums, etc) are using this spectrum to build their own private networks. That’s pretty remarkable. There is a trade-off with  unlicensed spectrum in that the power output limits have to be much lower on the GSM technology so that the signal does not propagate very far.   This makes it an ideal technology to deploy at a single building or complex, e.g. hospital, office building, stadium, etc. By keeping the power limits low, they don’t have to worry about interference.  1800MHz is also high enough up the spectrum band to limit propagation as well.  It would be worth pursuing unlicensed GSM regulation in emerging markets as a key to stimulating innovation and filling in gaps in coverage.

Dynamic GSM

Television White Spaces Spectrum (TVWS) or what is now known as Dynamic Spectrum is an approach to spectrum management in which frequencies are dynamically assigned via a geo-location database.  While geo-location databases were designed with television spectrum in mind, there is nothing to stop them being used for any frequency range.  Google are pursuing just this sort of approach in the 3.5GHz band.   It is plausible to suggest that a geo-location database approach might be used to make GSM spectrum available on a secondary basis in underserved communities.  This would really open the flood-gates to rural GSM solutions.  I suggested as much last year.  There are some potential obstacles to this happening quickly though. The first is the possibility of existing spectrum holders fighting the idea tooth and nail because the end-game implications of this is conceivably the end of traditional spectrum licenses.  Another possible problem is a more technical one. With television spectrum it is fairly easy to build a spectrum database. The location of television towers and transmitters as well as their power outputs are usually well-known by the regulator and can be used to build a good picture of spectrum occupancy.  For GSM spectrum, it would mean getting all that information from MNOs. This entails many more towers, technologies, etc. In theory this shouldn’t be that hard. In Canada, for instance, you can get that information from the regulator and build your own map of mobile towers but in most countries in sub-Saharan Africa, I suspect that information may be harder to extract from operators.  I have never seen a map of mobile towers but I sure would love to.

Finally I would be remiss if I failed to mention the brilliant innovation that researchers from UC Berkeley came up in using mobile handsets themselves to do dynamic spectrum sensing and automatically choose unused GSM spectrum. In this scenario, a geo-location database is not required as each basestation would self-regulate based on data reported from phones on the network. An idea so clever in its simplicity, it took my breath away the first time I read about it. Getting regulator buy-in though is likely to be tougher than the geo-location database approach even if this approach may be cheaper and more effective in the short-term.

Image courtesy Rhizomatica.org.  Title courtesy of Heineken

Spectrum Safari!

Previously I’ve written about why experts find it so hard to agree on spectrum regulation.  However there is one aspect of spectrum that I haven’t touched on and that is the fact that radio waves are invisible, which turns out to be a basic practical challenge.  Radio waves are all around us yet they are imperceptible.  We intuit their existence through technology but we have no means of seeing what frequencies are in use or how busy it is.  There has been at least one interesting art project aimed at making WiFi visible but in general we rely on experts to tell us what is in use and what isn’t.

And indeed if we look at a chart of spectrum allocation from the FCC, the United States communication regulator, things do look fairly full up.  One of the first insights I had into the potential of dynamic spectrum allocation was work carried out by the New America Foundation and the Shared Spectrum Company back in 2003 that simply set out to measure spectrum occupancy from New America’s offices in Washington DC.  This early study revealed a lot of unused spectrum.  Shared Spectrum has gone on to publish measurements from several spectrum occupancy studies since then.  This information is incredibly useful as it revealed that only a small fraction of the spectrum that has been allocated by regulators is actually in use.

Back in 2009 when I was advocating with others for TVWS regulation, we looked into getting a spectrum occupancy scan done in South Africa.  Simply finding a company with the expertise and equipment to do it was difficult enough but even more challenging was the price tag: many thousands of dollars.  A lot has changed since then and the last few years have seen the arrival of low-cost DIY spectrum scanners from companies like Nuts About Nets who make the RF Explorer series of spectrum scanners which range in price from USD119 dollars upwards.  This is a real breakthrough for both researchers and policy advocates.

In 2012, when Marco Zennaro and Ermanno Pietrosemoli from the Abdus Salam International Centre for Theoretical Physics (ICTP) worked with the Prof Chomora Mikeka of the University of Malawi to explore the potential of a TVWS pilot there, they were able to take advantage of the RF Explorer to carry out affordable spectrum occupancy scans that provided critical evidence in order to allow a TVWS pilot to proceed.

But things haven’t stopped there.  Around the same time that Marco, Ermanno, and Chomora were make plans to scan spectrum in Malawi, wireless hackers elsewhere were discovering that a inexpensive usb dongle designed for digital television broadcast reception, which sold for less than USD20, contained a Software Defined Radio chip that could operate on any frequency under 1GHz.  This magic combination of affordability and openness spawned a new community of wireless hackers developing applications that take advantage of this dongle, known as the RTL-SDR.  The development of Open Source software by Osmocom that could drive the RTL-SDR chips provided a platform for a variety of applications to be developed on this platform.  There is now a vast variety of software applications for the RTL-SDR that can do things ranging from television or radio broadcast reception to tracking ships at sea to monitoring satellites.

However, one of the challenges with spectrum monitoring is that it is highly dependent on location and on the sensitivity of the both the monitoring device and its antenna.  Being inside or outside or on top of a mountain or down in a valley will have a significant effect on how spectrum is measured.  Inexpensive equipment like the RTL-SDR dongle may not register spectrum that a more expensive, more sensitive device might pick up.  Thus spectrum scanning devices need to be calibrated in order to provide credible evidence.  Or do they?

In recent years, a host of new participatory research initiatives that allow ordinary citizens to participate in serious science projects have begun to change the way we think about research.  Known as citizen science or crowd science, these projects harness the interest of amateur or non-professional scientists often leveraging the power of the Internet to gather input from a huge range of participants.  Citizen science has spawned non-profit organisations like the Public Lab in the United States that were founded specifically to develop and apply open-source tools to enable citizen involvement in environmental science.

A great example of crowd science is the Quake-Catcher Network.  This is a research network that takes advantage of the fact that most modern laptops have hard-drives with built-in accelerometers that detect movement.  Designed to stabilise the hard drive in the event that the laptop is shaken or dropped, the accelerometers can also detect motion as a result of an earthquake.  However, a hard drive accelerometer has never been calibrated for use in monitoring earthquakes and a single hard drive accelerometer on its own would not provide meaningful information about an earthquake but hundreds or thousands of these devices networked together can provide more comprehensive information than the most sensitive seismograph.

You see where I’m going here?  Low-cost devices like the RTL-SDR dongle could be used for a citizen science project that would not only provide a wealth of critical information on spectrum occupancy in African countries (well anywhere actually) but would also be a practical introduction to radio spectrum for electrical engineering students across the continent.  Education and research directly connected to proactive regulatory reform: who could ask for more?  The more institutions or individuals that participate, the more accurate the research becomes and the more potential there is to compare regions and countries.  Something like this could also be used to validate and inform the spectrum propagation models used by geo-location databases.

Emmanuel Togo and Ato Yawson compare readings.

Emmanuel Togo and Ato Yawson compare readings.

To test this idea, Emmanuel Togo of the University of Ghana Legon and I arranged to do some test spectrum scanning in Accra last month.  We set out for the top of the hill that dominates the UG Legon campus armed with both RF Explorers and RTL-SDR dongles.  It took a little while to get everything working but below you can see some screenshots from the RF Explorer using their Touchstone spectrum scanning software.


This slideshow requires JavaScript.

And by comparison, here are a similar set of screenshots taken from the RTL-SDR dongle using the Open Source RTLSDR Scanner software.

This slideshow requires JavaScript.

You can see that the the readings are similar across the two devices.  It isn’t possible to make an absolute comparison between them the Touchstone software is displaying signal strength in dBm and the RTLSDR Scanner software is displaying in dB/√Hz.  To be honest I am still learning about the different ways in which the signal strength is being expressed but it is clear that they are both registering spectrum occupancy on the same frequencies.

You can see that the UHF band is moderately occupied with plenty of unused bands.  This is in the highest point in Accra, arguably a point of highest spectrum occupancy in the country.  With inexpensive spectrum scanning technologies, we have captured information that can directly inform a debate on how UHF spectrum ought to be utilized.  I don’t particularly believe in the old adage that “you can’t manage what you can’t measure” but in this case, having direct evidence of spectrum occupancy puts any conversation with regulator or ministry of communication on a firmer footing.

ofcom_spectrumWith some work towards standardizing how spectrum scans are done, I can imagine these inexpensive devices producing a rich picture of spectrum occupancy over an entire country.  The UK regulator, OFCOM, has done a great job of visualising spectrum allocation in the UK.  Most interesting for me is a little link at the bottom where you can download the data source for the map in JSON format.  Imagine being able to take a data file with the spectrum allocations and overlay spectrum occupancy data on top of it.  That would be pretty useful in terms of identifying spectrum bands lying fallow.

So imagine an army of students, researchers, concerned citizens all scanning spectrum with an inexpensive device (I paid $17 each for the ones I bought) and uploading the data where it could be used to paint a comprehensive picture of spectrum occupancy across a country or even across the continent.   Time for everyone to go on a spectrum safari!