Database-Assisted Spectrum Sharing — Way to go or No?
Overview
Recently, I came across this link which spells out the challenges the Sigfox company has faced (I am not sure if it is still facing) to stay afloat. To the level of the use of words “…gone rapidly south”, that shows a serious battle a company must be facing. For those unfamiliar with Sigfox, it is a French company (also the name of the technology) founded by two individuals — Ludovic Le Moan and Christophe Fourtet to expand Internet of Things (IoT) connectivity across the globe based on Low Power Wide Area Network (LPWAN) technology. Notably, Sigfox and LoRaWAN, while they are both unlicensed LPWANs (operating within the range of 862 to 928 MHz globally), they have a few distinct variations including their business models. For instance, in a great summary done by Link Labs, looking at both technology and business models, the Sigfox technology comes off as a sophisticated LPWAN approach to conserve battery power and limit the effect of noise. This means it has more constraints on the network performance (link budget limitations). On the other hand, its business model is an “all-in-one box” where Sigfox owns all of its technology, from the backend data and cloud server as well as the end point softwares. To be honest, as a developer/maker, I struggled understanding the necessity of this model if I am to run any pilot. It meant, I have to contact a Sigfox country partner to get started. Ofcourse, there is a segment left to the chip manufacturers and vendors based on agreed business terms to make Sigfox radios, but still, not as flexible. LoRaWAN, on the other hand (through the LoRa Alliance) seems to have adopted relatively open specifications making it more flexible and inviting masses in form of makers, hardware and gateway manufacturers to build a lot on top of LoRa/LoRaWAN. This has allowed the network to scale and more IoT deployments tested, although it is also worth noting that there are also closed elements to the technology. Story of another day!
This Sigfox/LoRaWAN models got me thinking on the opportunity that Spectrum Sharing presents and the model of the Database Assisted Spectrum Sharing. In the past decade, a lot of research has gone into DSA leading to very great publications and unlocking the avenue for National Regulatory Authorities (NRAs) to gradually embrace spectrum sharing and enable connectivity for both the unserved and underserved. For starters, Spectrum Sharing (SS) comes in as a promising approach to solve the problem of spectrum congestion by allowing a flexible access to more spectrum in order to meet the rapidly growing bandwidth demand in this era of ubiquitous use of smartphones and the new innovations of IoT. SS takes away the concern that we are running out of usable radio frequencies (RF). Spectrum measurements conducted across different countries tell a story that — at any given location and time, much of the prized spectrum lies idle. This calls for an evaluation of the existing spectrum management policies to allow various wireless service providers to encroach spectrum not primarily allocated to them. This was the foundation of the buzz word “TV White Spaces” when the US became the first country to authorise “unlicensed” access into the TV broadcasting band for provision of wireless Internet access back in 2010. Many countries followed suit with Singapore publishing their regulations in 2014, Canada and UK in 2015 and South Africa (the first country on the African continent) in 2018. Kenya, officially published its regulations in 2021.
Although TV White Spaces (TVWS) created an impression of exclusivity to the name “White Spaces”, White Spaces actually refer to those spectrums that are unused in a particular location at a particular time — which can be found in any allocated RF band. Therefore TV White Spaces (TVWS) refer to the idle frequencies in the VHF (e.g. 54–216 MHz for US) and UHF (e.g. 470–694 MHz for Kenya) TV broadcast bands that are either unassigned or unused by existing TV broadcast licensees — in short TVWS are the unused TV channels assigned to TV broadcasting. Spectrum sharing through TVWS became an important topic as it was the first step towards efficient use of spectrum in an opportunistic and dynamic manner. The propagation characteristics (longer communication distance and better penetration through obstacles) of the TV band presented an opportunity through TVWS that was (is) potentially seen to contribute to the efforts for Internet access to the last mile and enable more rural areas to be connected. In fact, the regulatory framework for TVWS in Kenya heavily points out on the usage of TVWS to connect the rural areas compared to a city country like Singapore. With such an opportunity that can contribute to bridging the digital divide, it is paramount that the governments and all the stakeholders ought to pay attention on how it can really work for the economic progress of the areas not adequately served. Unfortunately, this has not been the case. Which raises the question of Why?
To unpack that, it is important that we assess the architecture of deploying White Spaces. Since the incumbents do not need to be relocated, rules have to be developed such that they are protected i.e. the incumbents are not harmfully interfered with. Three principal models have been considered thus far as the basis of the rules (with the consideration of TVWS as the founding opportunistic spectrum usage band) — 1. The use of Spectrum Sensing, 2. Use of Beacons and 3. The use of Geolocation Databases. The overall idea of these three models is to ensure that the white space devices (call them the Secondary devices or radios) dynamically identify and use different frequencies within a defined band (such as the TV broadcasting band for TVWS) based on what frequency is available and in what geographic location for interference-free operation. The secondary radios need to have flexibility and agility to locate and operate on the unused channels, no matter where they are located in the country.
Spectrum Sensing
The use of Spectrum Sensing model is based on the concept of Cognitive Radios (CRs). In their book — Cognitive Radio Engineering, Charles W. Bostian, Nicholas J. Kaminski and Almohanad S. Fayez define a CR as a transceiver which is (a) aware of its environment, its own technical capabilities and limitations and those of the radios with which it may communicate, the rules governing its operation, and its user’s needs, priorities, and limitations; (b) capable of acting on that awareness and past experience to configure itself in a way that optimises its performance subject to some set of constraints; and © capable of learning from experience. In a real sense, a cognitive radio is an intelligent communication system that designs and redesigns itself in real time. While most research work for DSA has been described to focus on Spectrum Sensing, the reliability of it in multipath fading, blocking and weak signal conditions have been said to be most challenging. Energy detection (as well as cyclostationary detection), key to detecting signal presence has been said to lack the appropriate level of accuracy. In UK, for example, the broadcast industry was concerned that under certain conditions even the sensitivity (of a capability to detect a TV signal at a thousandth power level) was insufficient to allow opportunistic access to TVWS by secondary radios. The need to enhance the desired sensitivity meant that white space devices would be very difficult or expensive to implement which would translate to the cost of the secondary radios being high to inspire adoption of DSA. As such, even in the new bands such as the 3.5 GHz in the United States for Citizens Broadband Radio Service (CBRS) as well as the latest consideration of the 6 GHz band for RLAN, there has been very little consideration of Spectrum Sensing or at least not spoken actively about. Although, regulatory authorities such as the Federal Communications Commission of the US, the National Communications Authority (NCA) of Ghana and the Malawi Communications Regulatory Authority (MACRA) propose usage of spectrum sensing for TVWS, albeit with more constraints.
The Beacon Model
Another way for the secondary radios to determine which frequencies are available for opportunistic use is for a pilot or marker signal to be provided. These transmissions can be provided in a given area to enable the secondary radios to know which frequencies are safe. Robert Stewart, David Crawford and Andrew Sterling in their collection of articles under a book called “TV White Space Communications and Networks” note that ensuring that there is a beacon coverage over the United Kingdom would be an expensive undertaking. OPEX costs would also be an issue adding to the spectrum inefficiency that the beacons would bring up again as they would to occupy spectrum without direct value to the users. Back in 2010, Alexander Wyglinski, Maziar Nekovee and Thomas Hou in their book “Cognitive Radio Communications and Networks“ had also pointed out challenges that would emerge from the use of beacons such as:
- Who provides the beacon signal and what would be the commercial arrangements?
2. How is the information the beacons transmits updated in a heavily dynamic incumbent environment?
3. What spectrum would the beacons use?
4. How can the bacon signal be prevented from being received in an unintended area of coverage? Among other reasons.
The Geolocation Database Model (The Database Assisted Spectrum Sharing Model)
With the foregoing challenges of spectrum sensing and beacons model, spectrum regulators, together with some standards bodies and industrial organisations such as Google, Microsoft etc. advocated a database-assisted dynamic spectrum sharing architecture. In such an architecture as explained by Yuan Luo, Lin Gao and Jianwei Huang in their book “Economics of Database-Assisted Spectrum Sharing” a database will assist unlicensed devices to opportunistically exploit the white space spectrum. The database takes responsibility for intensive data processing (e.g., identifying the white space spectrum and computing the allowable transmission parameters) that requires significant energy consumption. Secondary radios only need to perform the necessary local computations (e.g., identifying their current locations). Hence, such a network architecture can effectively reduce the overall energy consumption of mobile devices. I share an image of the Ofcom’s model of this operation in Figure 1 here.
The Woes of the Database-Assisted Spectrum Sharing
While the viability of the database-assisted spectrum sharing has been proven by many trials and several commercial deployments to the extent that all the regulatory frameworks adopted it for utilisation of TVWS, things haven’t just been seen to take off to practically deliver implementation of DSA — beginning with the TVWS itself. For instance, all the pilots done in Kenya on Nominet and Fairspectrum databases are yet to present commercial developments — in short nothing is running yet. In fact, on a global scale, Nominet abandoned the business and sold its WaveDB (the TVWS spectrum management database) to RED technologies (I need to update myself on the status of WaveDB at the moment though).
Notably, the challenges affecting this model seem to stem from the regulatory and the economic models to sustain it. Regulatory in the sense that regulators (from a TVWS perspective) have been uncertain on how to provide an effectively competitive environment based on who should operate the database and how can the rules be set on a level field such that both in-country and multinational companies participate in the business. Further, the need of the regulator to manage various databases from a single point also calls for new unplanned software implementations from their side. While Ofcom (UK) and the CA (Kenya) suggested the use of a listing server which can list all the operating databases — this is still blurry as the proposition of the process of database discovery (initial base station transmission for TVWS) needs to contact the listing server first before getting to use a geolocation database — a bit too much to get the network running. The economic challenge stems from the fact that the service providers (i.e. Google, Microsoft, Fairspectrum etc) are uncertain of how the users of their databases need to actually pay them.
A recent study on “Gap Analysis in Spectrum Sharing” in Kenya has also brought another challenge — the stakeholders (the Wireless Internet Service Providers) seem unfamiliar with how the geolocation databases approach will operate in practice and allow them to extend their service and generate profits at the same time. Therefore even a collaborative model of the radio hardware manufacturers and the geolocation databases for TVWS might not suffice to make everyone happy in terms of profit/return on investment in the whole chain — the WISPs, the radio hardware manufacturers and the geolocation database service providers.
A bigger concern, however, is the future of DSA through the database-assisted spectrum sharing model. The Wireless Innovation Forum in their document WINNF-TR-1002 “Propagation Models and Interference Protection Criteria for Sharing between Fixed Services and Unlicensed Devices in the 6 GHz Band” note that the use of Spectrum Access System (similar to Geolocation databases in TVWS) in the CBRS has had some members pointing out that it is complex and prone to unnecessarily prolonged standardisation and testing processes. While it is already on the table to adopt Automated Frequency Coordination (again similar to geolocation databases) in the 6 GHz Wi-Fi (the latest Wi-Fi technology), the same document also points concerns on AFC viability in some scenarios.
Conclusion
To solve the spectrum scarcity problem, many researchers have proposed the use of dynamic spectrum sharing (or Dynamic Spectrum Access), which enables unlicensed devices to opportunistically share the spectrum with licensed companies. Specifically, unlicensed devices would access the white space spectrum, which refer to those spectrums that are unused in a particular location at a particular time, to avoid the harmful interference to licensed devices. Spectrum Sharing, therefore seems promising to contribute to the efforts of connecting the unconnected. The approach of using geolocation databases to effectively manage opportunistic spectrum access, perhaps is one that needs to be examined in detail. Is it really sustainable? Should we continue pursuing it or how can we effectively pursue it? Should the regulations be adjusted or then just allow all Spectrum Shared Networks to be classed under Community Networks? Should geographical locations be demarcated for Spectrum Sharing? A key technical question can also extend to study the approach of spectrum sensing for sustainable spectrum sharing…. In his blog post on Saving TVWS, Steve Song begins by posing a question of “How do technologies die?” — to which I think it is worth thinking about as DSA expands across many RF bands under the aegis of geolocation databases. While his conclusions on the same blog post highlights the approach of “Regulation Now, Database Later” for TVWS, I tend to think that needs to be examined across all the developments of DSA including the potential bands being considered for spectrum sharing. A more unique path of consideration though — whether now or in the future, is the Database-Assisted Spectrum Sharing Way to Go or No? Should we consider the path of LoRaWAN or “go south rapidly” like Sigfox?
