This paper proposes the design of enabling technologies for practical wireless communication systems operating in the TV white space (TVWS). The main objective of this paper is to cover a macro perspective on the system design blocks including: (a) targeted use case applications and governing regulations, (b) channelization, physical (PHY) layer and medium access control (MAC) layer designs, and (c) achievable throughput and range. It is the intention of this paper to serve as a general guideline for designing wireless communication systems operating in TVWS. The core system design addresses both PHY and MAC layer issues with realistic system considerations. In the PHY layer, a channelization design that fits into the area-specific TV channels and a transceiver that enables data exchange in the TV bands are designed. In the MAC layer, a cognitive engine that manages access to vacant TV channels and MAC functionalities that facilitate effective medium access are also proposed. As a result, the system is capable of supporting up to a typical throughput of 80 Mbps, and a maximum number of 40 users, assuming all users performing the most bandwidth-hungry application in the use case scenario. The corresponding operating range is found reach up to 400 m.

1. Introduction

The recent regulatory development [13] has opened up new opportunities for unlicensed wireless communication systems to utilize unoccupied TV channels (a.k.a TV White Space or TVWS) in the very high frequency (VHF) and ultra-high-frequency (UHF) bands. TVWS is defined as the TV bands that can be occupied by secondary (often unlicensed) wireless systems with the condition that the related regulatory and technical requirements are met. The opening of TVWS has catalyzed more innovative and diversified wireless applications, and has invited overwhelming responses from the industrial community. Among others is the mobilization in international standardization initiatives [47].

It is, therefore, a timely window to provide to the technical community the enabling wireless technology for operation in the TVWS. The two core factors that drive the direction of the technology are: (a) potential application use cases that define and verify the necessity, and (b) governing regulations that dictate the “do's and do not's.” The enabling technology shall consist of a system design capable of delivering the characteristics and performance as required by the targeted application use cases, and at the same time, capable of complying to the rules set by the regulations. For the design of TVWS enabling technology, it is of no exception. In order to validate its existence, the system has to possess a “killer-application”. And in order to harvest and utilize the new spectrum resources, the system has to possess the appropriate enabling functionalities to operate in the vacant TV channels within the boundaries set by local primary-user protection regulations.

For this purpose, in this paper, we have proposed a full system design taking into consideration all three elements: the potential application use cases, the governing regulations, and the corresponding cross-layer system design. Firstly, the use cases highlight the prospective applications that outline several requirements in communication speed and range. Secondly, the governing regulations set conditions and boundaries that fractionally limits the liberty of the system to access the TVWS. Thirdly, our system design is specified to meet the use case requirements and to be able to comply to the regulations. The proposed cross-layer design is further divided into channelization design, physical (PHY) layer design, and cognitive medium access control (MAC) layer design. It should be emphasized that before detailed investigations and optimizations are conducted in each of these elements, it is essentially important to have the overall assessment that proves the feasibility of the technology in delivering the mandated task and to provide a high-level guidance in respective lower-level system design. To the best knowledge of the authors, such works are nonexistent in the current literature.

The contribution of this paper is threefold: (a) it verified that wireless technology operating in the TVWS does offer encouraging opportunities, (b) providing an overall cross-layer MAC and PHY cognitive system design that considers practical applications and regulations, and (c) providing realistic performance evaluation (e.g., throughput, coverage, etc.) in response to related requirements. It is our strong belief that this paper is of interest to both the academic and industrial community. The detailed design philosophy and step-by-step design procedure are given in Section 2. The combined discussion on how the requirements, design considerations, and system performance are related to each other is given in Section 6.

2. General Design Philosophy

This section gives an overview on the design philosophy and design flow of the TVWS enabling technology, as illustrated by Figure 1. The starting point of the TVWS often lies in the regulations and the use cases. In the case of TVWS, the development of regulations creates opportunities and requirements for the corresponding wireless technology. The opportunities have motivated the extended use cases of the technology, while the requirements affect the design of the system. Together, the regulations and the use case become the input factors that influence the system design. Sections 3 and 4 address these arenas.

In this paper, the system design consists of the channelization, the PHY layer, and the MAC layer design. The channelization design takes into consideration the specific TV channels allowed by different regulatory domains. In other words, the channelization engine has to be capable of supporting all the operable TV channels. In this work, a dual phase-locked loop (PLL) design is employed to produce both center frequencies and channel spacing. Section 5.1 provides a generic design with specific numerical examples.

The PHY layer intakes the output of the channelization design as input design parameters. The PHY design considers the transceiver design, the TV channel model, and the operating range. Among others, the operating range has to support the practical scenarios in the use cases. The output parameters of the PHY design such as error performance and PHY layer data rate will be sent to the MAC layer. Section 5.2 presents the practical PHY layer design for the communication system operating in the TVWS.

The MAC layer consists of the cognitive management engine and other MAC enabling protocols. The cognitive management engine deals with the occupancy of TVWS taking into consideration the required regulations. Section 5.3 describes the cognitive engine design. Besides the TVWS occupancy management engine, other MAC protocols are also essential to enable communication control and data communication. The output parameters of the MAC design such as the system throughput are crucial and will be the core reference to support the targeted use cases. Section 5.4 provides the MAC layer design with practical numerical values.

In a nutshell, this paper provides a practical cross-layer design for a practical communication system operating in the TVWS based on realistic numerical values in the real-world industrial design.

3. Use Cases

The typical use case of communication in TVWS is often associated with wireless coverage of wide areas (e.g., suburban and rural areas) where installation of facilities can be costly. In this perspective, the TVWS communication is especially a suitable candidate given its long reaching and high penetrating VHF/UHF signal characteristics. Figure 2 illustrates the typical scenario of a wide area high-speed backbone connectivity fed by respective local subnetworks. Sites A, B, and C are far-apart locations each with high-power devices (HPD) antenna towers linked via TVWS connection, while the low-power devices (LPD) may be portable/mobile devices linked to HPDs via TVWS connection. The clear definitions of HPD and LPD are outlined in accordance to regulations. This general illustration can be applied to, but not limited to the following. (i)Campus. HPD—rooftop antenna of buildings. LPD—computers and mobile devices. (ii)Industrial Site. HPD—antenna in control center. LPD—machineries and automation. (iii)Municipality. HPD—municipal building antenna. LPD—employee workstations. (iv)Enterprise. HPD—enterprise building antenna. LPD—employee workstations. (v)Utility Network. HPD—antenna of utility provider. LPD—data collectors and wireless meters.

The above use case scenarios have been constantly addressed in various initiatives [811] calling for enablement of communication in the TVWS. It is, therefore, a practical “killer-application” based on popular demand.

In these typical scenarios, the parameter requirements are given in Table 1. For backbone connectivity between HPDs in different locations (e.g., building rooftops) several hundreds of meters apart, the required throughput is typically 50 to 80 Mbps. These HPD will then be connected to LPDs (e.g., access points) that distributes the bandwidth to respective end users. Depending on the applications, each end user may require throughput of 100 kbps to 2 Mbps. For the 100 kbps requirement, hundreds of end users can be supported. As for the 2 Mbps requirement, tens of end users can be supported.

From the perspective of applications, the 100 kbps requirement mainly targets on sensor networks such as the smart utility networks [12], where low-power wireless meters are installed connected to wireless data collector. On the other hand, the 2 Mbps requirement mainly targets on broadband access including online video streaming, web browsing, and local file transferring. Among these applications, the online video streaming (e.g., MPEG4 format [13]) requires a typical throughput of 2 Mbps. The end users are connected to the backbone that supports a typical speed of several tens (e.g., 60 Mbps as given in Table 1) of Mbps.

4. TVWS Regulations

The occupancy of TVWS is closely tied to the policy and rule-making of the regulatory institutions. Up-to-date, the several regulators actively participating in TVWS regulations are the United States (US) Federal Communications Commission (FCC), the United Kingdom (UK) Office of Communications (OFCOM), the European Conference of Postal and Telecommunications Administrations (CEPT), and the Japanese Ministry of Internal Affairs and Communications (MIC). Some of these regulations have already reached a fairly matured specification sufficient for actual system deployment, while others are still in the stage of planning and brain-storming. This paper places weighted emphasis on the more complete and matured FCC regulations.

FCC has issued a report and order (R&O) [1] in Nov. 2008, and then later in Sept. 2010 [2], outlining the governing regulations for unlicensed usage in the TVWS. The TV band devices (TVBD) are categorized into two classes: fixed and personal/portable devices (hereon referred to as portable devices). The fixed device is a HPD operating at a fixed location with high-power outdoor antenna. The portable device is an LPD, and is further divided into Mode I and Mode II. A Mode I device is a client portable device controlled by a fixed device or a Mode II portable device. A Mode II device is an independent portable device with capability of accessing available TV channels.

Two cognitive radio capabilities are specified by the FCC for the TVBDs: the mandatory channel availability/geolocation awareness and the optional spectrum sensing. All devices, except Mode I portable devices, must include a location awareness capability by means of accessing the internet to obtain from a Geolocation Database (DB) information consisting of colocating incumbent TV broadcasting. Prior to operating, the unlicensed devices must access the geolocation DB to get the list of permitted channels currently not occupied by the incumbent services. Additionally, all devices may sense the spectrum for the presence of TV signals and wireless microphone signals. The sensitivity of the sensing is all the way down to a received power level of −114 dBm.

In the US, the TV channels usable by unlicensed users range from the VHF channels 2–13 to the UHF channels 14–51, while several limitations and exceptions apply [1]. Each TV channel is 6 MHz wide. Table 2 shows a fraction of the UHF TV channels in the US.

It should be noted that FCC is only one among the many regulatory organizations that outlined governing rules for operation in the TVWS. There are many other regulators actively working on similar rules and regulations. In this paper, although uses the examples and specific parameters of FCC, the same analysis can be repeated for other regulations.

5. System Design for Enabling TVWS Operation

In this section, system requirements inherent to the use cases as described in Section 3, and regulatory demands as described in Section 4, are taken into consideration for the overall system design. The cross-layer system design is divided into channelization, PHY layer design, cognitive management, and MAC layer design.

5.1. Channelization

The first essential parameter in enabling communication in TVWS is the channelization strategy. The communication system has to be equipped with the capability to operate at center frequencies and channel spacing of unused TV channels. For this purpose, a clock generator consisting of crystal oscillator and phase-locked loop (PLL) is designed. The clock generator employs a dual-PLL structure to produce both the center frequencies and symbol rate (i.e., Nyquist frequency) 𝑓nr.

Figure 3 shows the architecture of the clock generator. In Figure 3—center frequency generation, the phase detector compares the two input signals 𝑓ic and 𝑓bc to obtain the error signal that is proportional to their phase difference. The error signal is then low-pass filtered (LPF) and used to drive a voltage-controlled oscillator (VCO) which creates an output frequency, 𝑓cen. The output frequency is fed through a frequency divider back to the input system as 𝑓bc, thus producing a feedback loop. If 𝑓cen drifts, the error signal will increase, driving the VCO frequency in the opposite direction so as to reduce the error. Therefore, the output is locked to the input frequency 𝑓ic, which is the product of a stable frequency 𝑓XTAL derived from a crystal oscillator. This is the basic design to produce the center frequency of the TV channels. In Figure 3—symbol rate generation, the symbol rate 𝑓nr can also be obtained by using the same procedure, with a different set of parameters.

By employing the designed dual-PLL clock, the specific center frequencies 𝑓cen of the US TV channels given in Table 2 can be generated. In this numerical example, the 40 MHz crystal oscillator (i.e.,𝑓XTAL=40MHz), a popular choice for most 802.11 wireless local area network (WLAN) systems [14] is assumed. In Figure 3, by substituting 𝑃=𝑓XTAL, 𝑓ic of 1 MHz can be obtained. Then, by employing 𝑄=𝑓𝐿+(𝑓𝐻𝑓𝐿)/2 where 𝑓𝐻 and 𝑓𝐿 are upper and lower frequencies of the TV channels as shown in Table 2, 𝑓cen for each TV channel can be generated. Similarly, by substituting 𝐿=1, 𝑓is=𝑓XTAL can be produced. Next, by using 𝑀=8, 4, and 2, symbol rate 𝑓nr of 5, 10, and 20 MHz can be generated. Note that the reason to select symbol rates of 5, 10 and 20 MHz is also motivated by the attempt to align the TVWS communication system to the popular WLAN.

While the specific numerical results are derived assuming particular regulations and existing wireless systems, the generic clock architecture is capable of supporting frequency generation in any band and channel spacing.

5.2. PHY Layer Design

The channel plan provided by the generator in Section 5.1 gives the starting point of designing the PHY layer. In the following sections, the PHY layer design will be described, with the achievable PHY data rate and operating range as the output parameter.

5.2.1. Signal Model

Figure 4 shows an illustration of the orthogonal frequency division multiplexing (OFDM) transmission system for the TVWS operation. In the transmitter, the high-speed data is first encoded with a forward error correction (FEC) scheme with coding rate 𝑟𝑐, and then converted into parallel data of subchannels. The transmitted data can be expressed as [15] 𝑠(𝑡)=𝐿𝑘=FFT1𝑖=0𝑑𝑖(𝑘)exp𝑗2𝜋𝑓𝑖𝑡𝑘𝑇sym×𝑓cen𝑡𝑘𝑇sym,(1) where 𝐿FFT is the total number of OFDM subcarrier, 𝑇sym is the OFDM symbol duration, 𝑓𝑖 is the frequency of the 𝑖th subcarrier, and 𝑓𝑖=𝑓cen+𝑖/𝑇sym, 𝑓cen is the operating frequency. Note that each data 𝑑𝑖 is a complex number, 𝑑𝑖=𝑑𝐼𝑖+𝑗𝑑𝑄𝑖, where 𝑑𝐼𝑖 and 𝑑𝑄𝑖 are in the range of {±1,0},{±1,±1},{±3,±3} and {±7,±7} for binary phase shift keying (BPSK), quadrature phase shift keying (QPSK), 16 quadrature amplitude modulation (16QAM), and 64 quadrature amplitude modulation (64QAM), respectively. The modulation level 𝑚𝑙 for these schemes are 1, 2, 4, and 6, respectively. The chosen FEC scheme is the Reed-Solomon block codes.

Here, the subcarrier spacing 𝜁FFT of the OFDM signal and its equivalent time domain representation 𝑇FFT are given as 𝜁FFT=𝑓nr𝐿FFT,𝑇FFT=1𝜁FFT,(2) where 𝑓nr is the Nyquist frequency produced by the clock generator.

After the insertion of guard interval (GI), the OFDM signal is described as 𝑠(𝑡)=𝐿𝑘=FFT1𝑖=0𝑑𝑖(𝑘)exp𝑗2𝜋𝑓𝑖𝑡𝑘𝑇total𝑓𝑡𝑘𝑇total,(3) where 𝑇total=𝑇sym+𝑇GI, 𝑇GI is the GI interval, 𝑓(𝑡) is the modified pulse waveform of each symbol defined as 𝑓(𝑡)=1 if 𝑇GI𝑡𝑇sym, and 𝑓(𝑡)=0 if 𝑡<𝑇GI and 𝑡>𝑇sym.

5.2.2. Receiver Model

At the receiver, the received signal is given by 𝑟(𝑡)=0(𝜏,𝑡)𝑠(𝑡𝜏)𝑑𝜏+𝜂(𝑡),(4) where (𝜏,𝑡) is the impulse response of the radio channel at time 𝑡, and 𝜂(𝑡) is the complex additive white Gaussian noise (AWGN). The receiver configuration is described in Figure 4. The output of the receiver is expressed as the 𝑖th OFDM subcarrier as 𝑑𝑖=1𝑇sym𝑇sym+𝑘𝑇total𝑘𝑇total𝑟(𝑡)exp𝑗2𝜋𝑓𝑖𝑡𝑘𝑇totald𝑡.(5) Note that 𝑑𝑖 will be used to compare with the input data to obtain the error performance.

5.2.3. Design Parameter Considerations

This subsection presents the essential PHY parameters in designing the OFDM signaling. The output of the PHY layer design are the PHY data rate ΩPHY and the operating range Υ, that shall be sufficient to support the requirements in the targeted use cases, as given in Table 1. In order to achieve the sufficient ΩPHY and Υ, the critical parameters to be considered are modulation level 𝑚𝑙, FEC coding rate 𝑟𝑐, FFT size 𝐿FFT, and GI size 𝐿GI.

In the VHF/UHF band, existing channel propagation studies that are widely used are reported in [16, 17]. The important parameters extracted from these studies are the maximum delay spread 𝜖max that dictates the design of the OFDM signaling. In [16], 𝜖max is given as 10–20 μs for the range of 10–20 km. In [17], 𝜖max is 11–25 μs for 17–33 km and 25–60 μs for up to 100 km. As a result, it is concluded that within 10 km, 𝜖max can be set to 12 μs. Next, in order to effectively mitigate frequency selective fading in the OFDM signal, 𝑇GI should be set larger than 𝜖max.

For backbone connectivity, 𝐿FFT is chosen to be 1024 for the HPD, while for end-user connectivity, 𝐿FFT is chosen to be 256 for the LPD for lower complexity. For both scenarios, 𝑇GI is set to 𝑇FFT/4. The Nyquist frequency 𝑓nr for backbone connectivity and end-user connectivity are set to 5 and 20 MHz, respectively. These settings provide to both the backbone connectivity and end-user connectivity, 𝑇GI>𝜖max=12.8μs. In other words, the frequency selective fading can be mitigated with these settings.

5.2.4. Error Performance

The bit error rate (BER) can be quantified by comparing the transmitted bits 𝑑𝑖 and the bits produced by the receiver 𝑑𝑖 in (5), as 𝑑BER=𝑃𝑖𝑃𝑑=+1𝑖=1𝑑𝑖𝑑=+1+𝑃𝑖𝑃𝑑=1𝑖=+1𝑑𝑖,=1(6) where 𝑃() denotes the probability and 𝑑𝑖 and 𝑑𝑖 represent the 𝑖th transmitted and received data bit.

In similar way, the frame error rate (FER) can be described as 𝑑FER=𝑃𝑗=𝜚𝑗𝑃𝑑𝑗𝜚𝑗𝑑𝑗=𝜚𝑗,𝑗=1,2,,𝐿fr,(7) where 𝑑𝑗 and 𝑑𝑗 represent the 𝑗th transmitted and received data bit in a data frame, 𝜚𝑗 is the 𝑗th bit from the data bit sequence of length 𝐿fr, sent as a transmitted frame. The relationship between FER and BER can be expressed by (FER=11BER)𝐿fr.(8)

Specific numerical values are substituted into the PHY layer design to determine the error performance of the system. For this purpose, the FEC Reed-Solomon coding is designed to encode 239 to 255 block symbols. The chosen modulation schemes are QPSK, 16QAM, and 64QAM. The propagation channel is assumed to be a fading channel with Rayleigh distribution to consider the bottleneck of the channel condition. The signal to noise ratio (SNR) in the receiver is quantified as the ratio between energy per bit (𝐸𝑏) to noise power spectral density (𝑁0). In this paper, we neglect the intercarrier-interference between OFDM subcarriers for simplicity.

Figure 5 shows the error performance of the OFDM system in a fading channel with Rayleigh distribution. The minimum performance requirement for the system is application-dependent, as given in Table 1. By using (8), the required BER can be approximated as 105 with typical 𝐿fr=4096 [14]. Figure 5 gives the equivalent required 𝐸𝑏/𝑁0 for QPSK, 16QAM, and 64QAM to be 23, 25.5, and 29 dB, respectively.

5.2.5. PHY Data Rates

This section calculates the achievable PHY data rate ΩPHY based on the design parameters discussed in Section 5.2.3. The relationship between ΩPHY and relevant design parameters can be expressed as ΩPHY=𝑓nr𝑚𝑙𝑟𝑐𝐿FFT𝐿GI𝐿FFT.(9)

Employing (9), the list of supported ΩPHY are given in Table 3. In Table 3, Index 1 to 3 are designed for end-user connectivity which requires up to a total of 2 Mbps per user according as specified in Table 1. Index 4 to 6 are designed to fulfill the requirements of the backbone connectivity, suiting to various operating ranges Υ. The analysis on Υ will be conducted in Section 5.2.6. The combined analysis incorporating ΩPHY, Υ and the use case requirements will be conducted in Section 6.

5.2.6. Link Budget

Employing the PHY parameter settings in Section 5.2.5, the operating range Υ of data communication can be determined as log10𝑃(Υ)=TX+𝐺TX+𝐺RX𝑆𝑚𝐿PLn10𝜙,(10) where 𝑃TX and 𝐺TX are the transmitter power and antenna gain, 𝐺RX is the receiver antenna gain, and 𝜙 is the propagation loss index. Here, 𝐿PLn is the well-known urban path loss model [18] and is not repeated in this paper. Note that 𝐿PLn is a normalized parameter with center frequency 𝑓cen = 600 MHz (a typical value assumed in this example), HPD and LPD antenna height of 30 m and 1 m, respectively. All parameters are in dB scale. The minimum receiver sensitivity 𝑆𝑚 is calculated as 𝑆𝑚(dB)=174+10log10ΩPHY𝐸+NF+𝑏𝑁0,(11) where NF is the receiver noise figure, and 𝐸𝑏/𝑁0 is the required SNR to reach the targeted BER.

By substituting the relevant PHY parameters into (10), the effective operating range Υ can be determined. In this analysis, 𝑃TX is set to 100 mW for LPD and 4 W for HPD as specified by [1], 𝐺TX and 𝐺RX are set to 0 dBi to indicate omnidirectional transmission, NF is set to 8 dB, and 𝜙 is set to 3. Other relevant parameters are set similar to the previous sections.

The achievable Υ is given in Table 3. In Table 3, Index 1 to 3 are targeted for end-user connectivity and is thus confined within a shorter range between end users to the access point. On the other hand, Index 4 to 6 are intended for backbone connectivity which requires higher effective operating range to bridge across fixed stations. The combined analysis incorporating ΩPHY, Υ and the use case requirements will be conducted in Section 6.

5.3. Cognitive Management Plane

The cognitive management plane of the TVWS communication system consists of a master protocol, namely, the Dynamic Enabling Protocol (DEP) that controls the access to the targeted TV channel. The cognitive management engine includes into the DEP mechanism the features of the channel availability/geolocation awareness and spectrum sensing. Figure 6 shows the typical state machine implementation of the DEP.

For fixed and portable Mode II devices, the channel availability/geolocation awareness through accessing the DB is a mandatory requirement while the spectrum sensing is an optional feature. (i)Upon power on, a device is in the “unenabled” state. In the “unenabled” state, the device is not allowed to transmit any energy into the air. Next, the “unenabled” device accesses the DB containing information of incumbent users through the internet to inquire the vacancy of the targeted channel to be used. If the DB indicates that the targeted channel is not occupied by an incumbent user, the state transits to the “becoming enabled” state. (ii)In the “becoming enabled” state, if spectrum sensing is activated, the device may still not transmit energy into the air. The device shall perform spectrum sensing to check the availability of the targeted channel. Upon determining the targeted channel to be vacant through spectrum sensing, the state of the device can then transit to the “enabled” state. This step is not necessary if spectrum sensing is activated. Additionally, at any time in the “becoming enabled” state, if a de-enablement signal is received from the DB, indicating that an incumbent user has occupied the same channel, the device shall transit back to the “unenabled” state. (iii)In the “enabled” state, the device may start to set up networks and perform data communication by transmitting signals into the air. At any time in the “enabled” state, the device transits to the “becoming enabled” state if a de-enablement signal is received from any other “enabled” device within the same channel OR if the spectrum sensing reports new activities of incumbent systems in the channel.

For portable Mode I devices, the state machine starts from the “becoming enabled” state. The rest of the state transition processes are the same.

5.4. MAC Layer Design

Following the DEP described in Section 5.3, a device in the “enabled” state is allowed to transmit signal into the air. This section further specifies the basic MAC layer design necessary for enabling communication in the TVWS.

5.4.1. Network Topology and Resource

A typical network consists of 𝑁 devices, one of which is selected as the network coordinator (NC). The remaining 𝑁1 will be dependent devices associated with the NC. The NC provides the network reference timing and manages the resource sharing among the devices. A basic network unit is illustrated in the dotted box in Figure 2.

Figure 7 shows the reference timing for the network controlled by the NC. A full cycle of the network timing is known as the beacon interval (BI), which is the duration between two consecutive beacons. Following the beacon is a duration known as the contention access period (CAP). CAP is a duration that allows devices to access the wireless channel employing the carrier sense multiple access/collision avoidance (CSMA/CA) scheme. Following CAP is the contention free period (CFP). CFP allows the devices to access the wireless channel by using the time division multiple access (TDMA) scheme. In CFP, guaranteed time slots (GTSs) are allocated for each device to perform respective noncontentious communication.

The major reason of having both CAP and CFP for the MAC layer design is to support transmissions with and without stringent latency demands. As described in Section 3, the targeted use cases include local file transfer, web browsing, and video streaming. Particularly in video streaming, latency can be a delicate issue. Therefore, it is essential to have the hybrid CAP/CFP design to fulfill all needs.

5.4.2. MAC Functionalities

In order to enable communication in the TVWS, the MAC functionalities have to be specified. The basic set of functionalities to facilitate operation in the TVWS are starting a network, association/disassociation, GTS allocation, and data streaming, as shown in the message sequence charts (MSC) in Figure 8.

Starting a network involves the scanning process followed by the transmission of a beacon signal. In Figure 8(a), the MAC of a device receives request from the MAC Layer Management Entity (MLME) to scan all channels 1 to 𝑀. After the scan is completed, the MAC will report the results to the MLME. Next, a beacon frame will be sent by the device (now known as the NC) to seek other devices in the same location to form a network.

As shown in Figure 8(b), if a device receives a beacon frame and intends to join the network, the MLME will send an association request to the MAC. The MAC will then send an association request frame over the air to the NC. Upon receiving the request frame, the NC acknowledges with an acknowledgment (ACK) frame. The NC will then process the request and respond accordingly by sending the associate response frame. Similarly, the device acknowledges after receiving the associate response frame and reports to the MLME.

Figure 8(c) shows the GTS allocation for occupying the CFP. The device intending to occupy a GTS sends a GTS request frame to the NC. The NC acknowledges with an ACK frame, then sends to the MLME the request. In the next beacon transmission, the NC will insert into the beacon frame, the timing information of the GTS allocated for the device. The device upon receiving the beacon will wait until the CFP period and its turn to perform data streaming.

In Figure 8(d), data streaming is described. Assuming that the GTS notification is received by the preceding beacon, the device will send data frame to the destination (e.g., the NC) in the allocated GTS. The destination device will reply with an ACK frame if the data frame arrives.

5.4.3. Channel Access

As mentioned in Section 5.4.1, the two channel access methods to support data transmission in the TVWS are CSMA/CA in CAP and TDMA in CFP. This hybrid access method is a popular choice of major industrial standards specifications [14, 19, 20]. The following sections evaluates the achievable throughput for both schemes based on the data rate and other parameters provided by the PHY layer.

In the CSMA/CA scheme, a device attempting to transmit a frame first scan the TV channel. If the channel is idle for a period larger than 𝛼, the device will transmit. If the device detects that the channel is busy, it will persist to monitor the channel until it is idle. At this point, the device generates a random back-off interval to minimize collisions with frames from other devices. The back-off scale is discrete in time. The time following 𝛼 is slotted, and a device is allowed to transmit only at the beginning of each slot time 𝜎. Slot time 𝜎 is a value dependent on the PHY layer design, taking into consideration the radio design such as propagation delay, transmitter-receiver-turn-around time, and so on.

The CSMA/CA uses an exponential back-off scheme. At each frame transmission, the back-off time is uniformly chosen in the range (0,𝑤1), where 𝑤 is the contention window dependent of number of failed transmissions. At the first transmission attempt, 𝑤 is set to the minimum value CWmin. Each unsuccessful transmission increases 𝑤 until the maximum value CWmax.

The transmitting device has no means of detecting a potential collision of the frames sent out, it is essential to also specify a feedback loop for the receiving device to acknowledge the successful receipt of frames. The ACK frame is transmitted by the destination device to the source device after receiving the whole frame. The interval between the end of the frame and the beginning of the ACK frame is noted as 𝛽. The system designed should be tailored to have 𝛽<𝛼, so that no other devices is able to detect the channel as idle during the ACK frame feedback.

In order to calculate the effective system throughput, we firstly express Λ as the probability of a device transmitting in a randomly chosen slot time. Probability Λ can be expressed as [21, 22] Λ=2(12𝑝)(12𝑝)CWmin+1+𝑝CWmin(1(2𝑝)𝑚),(12) where 𝑝 is the probability that a transmitted frame encounters a collision and 𝑚 is the maximum back-off stage such that CWmax=2𝑚CWmin. In the simplified case of 𝑚=0 (i.e., no exponential backoff is considered), (12) can be rewritten as Λ(𝑚=0)=2CWmin+1.(13)

Let 𝑇𝐼 be the idle duration. When all devices are counting down, no frame transmissions take place; thus, 𝑇𝐼=𝜎.(14)

Let 𝑇SE be the success/error duration. When a frame is successfully transmitted or it is corrupted due to channel noise, the slot duration is given by 𝑇SE=𝑇pre+𝑇𝐻+𝑇fr+𝛽+𝛿+ACK+𝛼+𝛿,(15) where 𝑇𝐻 is the total header duration including the PHY and MAC headers, 𝑇pre is the duration of the preamble, 𝑇fr is the payload duration, and 𝛿 is the propagation delay.

Let 𝑇𝐶 be the collision duration. When two or more devices transmit at the same time and collision occurs, the sender will wait for the next transmission. So, 𝑇𝐶=𝑇pre+𝑇𝐻+𝑇fr+𝛼+𝛿.(16)

Let 𝑃𝐼, 𝑃SE, 𝑃𝐶 represents the probabilities of idle, success/error, and collision events. In a channel with a total of 𝑁 devices, 𝑃𝐼=(1Λ)𝑁,𝑃SE=𝑁1Λ(1Λ)(𝑁1),𝑃𝐶=1𝑃𝐼𝑃SE.(17)

The normalized system throughput Γ(nor)CAP (bps/Hz) can be given as Γ(nor)CAP=𝑃SE𝐿fr(1FER)𝑃𝐼𝑇𝐼+𝑃SE𝑇SE+𝑃𝐶𝑇𝐶,(18) where 𝐿fr=ΩPHY𝑇fr denotes the frame payload size in octets.

Finally, the effective system throughput (bps) can be expressed by ΓCAP=ΩPHYΓ(nor)CAP,(19) where ΩPHY is the data rate provided by the PHY layer design.

The specific MAC parameter settings for this analysis is given as follows [14]: for Index 1 to 3 where 𝑓nr=5MHz, 𝜎 is set as 21 μs, 𝛽 is set as 64 μs, and 𝑇pre and 𝑇𝐻 are 16 and 64 μs, respectively. As for Index 4 to 6 where 𝑓nr=20MHz, 𝜎 is set as 9 μs, 𝛽 is set as 16 μs, and 𝑇pre and 𝑇𝐻 are 4 and 16 μs, respectively. Other parameters such as 𝛼 is equivalent to 𝛽+2𝜎, 𝛿 is set as 1 μs, and CWmin and CWmax are set as 16 and 1023. Also note that the ACK frame is 14 octets while the average data frame 𝐿fr is 4096 octets.

Figure 9 presents the relationship between number of devices and achievable system throughput ΓCAP. In Figure 9, Index 1 to 3 shows ΓCAP for end-user connectivity, while Index 4 to 6 are particularly designed for backbone connectivity. With low number of devices sharing the channel (e.g., 𝑁=2), the efficiency of CAP is at least 75% for all cases. As 𝑁 increases, more collision among the users take place and ΓCAP decreases.

In the TDMA scheme, a device conducts communication only in the dedicated time slot allocated by the NC. The TDMA is employed to perform time-sensitive data streaming for better quality of service. Applications that may benefit from this scheme are the online video streaming and peer-to-peer video transmission.

Referring to Figure 7, the time slots are scheduled in the CFP, which is the period following the CAP. Devices intending to reserve the GTS in the CFP shall first request for time slot allocation from the NC in the CAP. Figure 8(c) shows the handshakes between the device and the NC for GTS allocation. The device first sends a GTS request frame to the NC, then awaits the ACK frame. If it receives the ACK frame, it will expect the next incoming beacon frame to carry the timing descriptor of the allocated GTS. The device will, therefore, conduct its collision-free streaming in the allocated GTS. The data streaming in GTS takes place as the time of the allocated GTS arrives. As shown in Figure 8(d), the source device will send data to the destination device and receives an ACK frame if successful.

The normalized system throughput Γ(nor)CFP (bps/Hz) can be given as the ratio between average payload information and the total frame length plus the frame-to-frame interval Γ(nor)CFP=𝐸𝐿fr𝐸𝐿fr+𝐿𝐻+𝐿pre+𝛽.(20)

Finally, the effective system throughput (bps) can be expressed by: ΓCFP=ΩPHYΓ(nor)CFP.(21)

The calculation of ΓCFP is more straightforward due to the collision-free transmission. The specific MAC parameter settings for this analysis is given as follows [14]: for Index 1 to 3 where 𝑓nr=5MHz, 𝛽 is 64 μs, and 𝑇pre and 𝑇𝐻 are 16 and 64 μs, respectively. As for Index 4 to 6 where 𝑓nr=20 MHz, 𝛽 is 16 μs, 𝑇pre and 𝑇𝐻 are 4 and 16 μs, respectively. The average data frame 𝐿fr is 4096 octets.

Employing (21) and the relevant parameters, ΓCFP calculated for Index 1 through 6 are given as 6.5, 13, 19.6, 27.5, 55, and 82.5 Mbps, respectively. Note that ΓCFP is the average system throughput for the system, where each user may occupy only a certain time slot for respective transmission.

6. A Combined Analysis

This section combines all the analysis from the previous sections to outline the entire design flow of the enabling technology for TVWS operation, as described in Section 2 and Figure 1. In Section 3, the use case scenarios are described highlighting both TVWS backbone connectivity and end-user connectivity with applications requiring throughput from 100 kbps up to 2 Mbps. The use case scenarios also specify other required parameters such as FER and supported number of users to be 0.08 and 30. In Section 4, the requirements in the regulation such as cognitive radio capability and TV channel plan are given. Together, the use case and the regulations define the “perimeter” of requirements for the system design.

The system design consists of the channelization design, the PHY and MAC layer designs. In Section 5.2, the PHY layer design employing OFDM signal is proposed with the output data rates of at least 7 Mbps. The minimum achievable operating range is 300 m. A total of six data rate modes (Index 1 to 6) are proposed to support different application demands, as illustrated in Figure 10.

Next, in Section 5.4, the cognitive management engine and other enabling MAC protocols are specified. The cognitive management engine specifies a state transition diagram based on the regulation requirements (i.e., geolocation awareness and spectrum sensing) to control the occupancy of TV channels. Other enabling MAC protocols such as channel access procedures specify the channel access mechanism that facilitates the sharing of the same channel by multiple users. A hybrid CSMA/CA and TDMA channel access method is proposed to support both effective multiple access and quality of service transmissions. In the CSMA/CA method, the minimum and maximum system throughput are 5.6 and 57.2 Mbps, respectively, whereas in the TDMA method, 6.5 and 82.5 Mbps, respectively.

Finally, according to Figure 1, the system throughput and operating range are fed back to the use case requirements. With the achievable system throughput, the communication system in TVWS is capable of supporting up to a typical number of 40 users (e.g., assuming simultaneous video streaming) in some 400 m radius. The supportable number of users that conduct web browsing or other less bandwidth hungry applications may even be higher. Additionally, for applications in need of higher power efficiency, devices with lower complexity and lower power may be employed at the expense of lower number of supported users and shorter range.

7. Conclusion

This paper proposes a cross-layer MAC/PHY layer design for wireless communication system operating in the TVWS, taking into consideration practical issues such as application use cases and governing regulations. The advantages offered by the proposed communication system in supporting the targeted use cases and complying to the regulations are verified. Future works include the specific optimization in each of the design block shown in this paper.