Scholarly article on topic 'A flexible radio transceiver for TVWS based on FBMC'

A flexible radio transceiver for TVWS based on FBMC Academic research paper on "Computer and information sciences"

CC BY-NC-ND
0
0
Share paper
Academic journal
Microprocessors and Microsystems
OECD Field of science
Keywords
{"Cognitive radio" / "Flexible digital radio" / TVWS / ACLR / FBMC / "FPGA implementation"}

Abstract of research paper on Computer and information sciences, author of scientific article — Vincent Berg, Jean-Baptiste Doré, Dominique Noguet

Abstract In this paper a flexible radio approach for opportunistic access to the television white space (TVWS) is presented. Requirement stems from the coexistence scheme used in this band between opportunistic transmission and TV broadcast signals (or wireless microphones). To ensure nonharmful interference of the TVWS signal on the incumbent services, a high adjacent leakage power ratio (ACLR) is expected. Also, flexibility is required to address the vacant channels in the UHF spectrum. Flexibility and low ACLR specifications are difficult to obtain simultaneously. The approach proposed in this paper is based on filter bank multi-carrier modulation (FBMC) scheme and a flexible hardware platform to combine the digital filtering capability of FBMC with RF agility. A FBMC hardware architecture implementation is presented and its associated complexity is studied for this platform. Then the hardware implementation validates that both flexibility and ACLR performance of the system are preserved even when off-the-shelf component impairments are considered. An experimental setup validates the coexistence with a TV broadcast signal and a comparison with a classical approach shows the gain in performance.

Academic research paper on topic "A flexible radio transceiver for TVWS based on FBMC"

Microprocessors and Microsystems xxx (2014) xxx-xxx

Contents lists available at ScienceDirect

Microprocessors and Microsystems

journal homepage: www.elsevier.com/locate/micpro

A flexible radio transceiver for TVWS based on FBMC

Vincent Berg *, Jean-Baptiste Doré, Dominique Noguet

CEA-LETI, Minatec, Grenoble, France

ARTICLE INFO ABSTRACT

In this paper a flexible radio approach for opportunistic access to the television white space (TVWS) is presented. Requirement stems from the coexistence scheme used in this band between opportunistic transmission and TV broadcast signals (or wireless microphones). To ensure nonharmful interference of the TVWS signal on the incumbent services, a high adjacent leakage power ratio (ACLR) is expected. Also, flexibility is required to address the vacant channels in the UHF spectrum. Flexibility and low ACLR specifications are difficult to obtain simultaneously. The approach proposed in this paper is based on filter bank multi-carrier modulation (FBMC) scheme and a flexible hardware platform to combine the digital filtering capability of FBMC with RF agility. A FBMC hardware architecture implementation is presented and its associated complexity is studied for this platform. Then the hardware implementation validates that both flexibility and ACLR performance of the system are preserved even when off-the-shelf component impairments are considered. An experimental setup validates the coexistence with a TV broadcast signal and a comparison with a classical approach shows the gain in performance.

© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND

license (http://creativecommons.org/licenses/by-nc-nd/3XI/).

Article history: Received 6 December 2013 Revised 23 May 2014 Accepted 28 May 2014 Available online xxxx

Keywords:

Cognitive radio

Flexible digital radio

FPGA implementation

1. Introduction

In 2009, the United States radio regulator - the Federal Communication Commission (FCC) - authorized opportunistic unlicensed operation in the television (TV) bands [1,2]. Such opportunistic communication systems have to coexist with TV broadcast signals and wireless microphones (referred to as 'incumbent systems' hereafter). The coexistence scheme is enforced with a priority mechanism where opportunistic systems must guarantee that no 'harmful interference' will be incurred to the incumbents. Such rules are meant to allow the control of the deployment and use of the unlicensed service so as to avoid harmful interferences on incumbents, but not to restrict it [3].

With the FCC rules, harmful interference is defined in a twofold way. Firstly, co-channel communication between incumbent and opportunistic systems is prohibited. This means that opportunistic systems must be able to assess the presence of incumbent signals and only access channels vacant from any incumbent. Besides, opportunistic systems have a limited amount of time to evacuate the channel when an incumbent is switched on.

Secondly, adjacent channel leakage ratio (ACLR) is limited in order to prevent an opportunistic system from interfering with an incumbent operating in an adjacent channel. In [2], ACLR is

* Corresponding author. E-mail addresses: vincent.berg@cea.fr (V. Berg), jean-baptiste.dore@cea.fr (J.-B. Doré), dominique.noguet@cea.fr (D. Noguet).

restricted to be at least 55 dB. Similar requirements are about to be adopted in other countries (e.g. in the United Kingdom [4]). Such a high ACLR requirement is specific to this usage context also called TV White Space (TVWS). For instance, ACLR requirement is 10 dB stronger than the one set for LTE systems.

From these requirements, it can be concluded that an opportunistic system must be agile in frequency, as the operating channel may change according to the location and time of operation. Since the Ultra High Frequency (UHF) band spans across a large frequency range (typically from 450 MHz to 790 MHz), this agility constraint implies using a tunable mixer in the RF. Besides, ACLR must be large enough to meet the 55 dB requirement. Unfortunately when considering standard Orthogonal Multiplexing Frequency Division (OFDM) waveforms, high ACLR values are usually obtained with Surface Acoustic Wave (SAW) or bulk acoustic wave filters. Frequency agility is therefore very limited. Thus, another waveform architecture should be considered to address dynamic spectrum access in the TVWS band.

Filter bank multicarrier modulation (FBMC) has been considered as one of the well-suited waveforms for adoption to opportunistic transceiver usage [5]. However, the strict ACLR and flexible requirements of the TVWS bands require experimental implementation to further assess feasibility, implementation costs and practical performance. This paper presents the architecture of a flexible FBMC transceiver in this context. A hardware complexity based on a Field-Programmable Gate Array (FPGA) implementation

http://dx.doi.org/10.1016/j.micpro.2014.05.010 0141-9331/® 2014 The Authors. Published by Elsevier B.V.

This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

V. Berg et al. /Microprocessors and Microsystems xxx (2014) xxx-xxx

evaluation is provided and transceiver performance is measured for a flexible architecture.

The advantages of the FBMC air interface to tackle these challenges are addressed in Section 2. It is shown that the filtering scheme embedded in the modulation itself guarantees high ACLR and relaxes the constraints of the analog filtering. In order to meet the TVWS requirements, the hardware implementation is crucial. In this paper, the frequency agility is provided by the hardware platform and it is important to show that the hardware implementation does not degrade the ACLR performance too much, even when standard off-the-shelf components are selected. The architecture of the FBMC transceiver on the TVWS flexible platform T-FleX [6] is described in Section 3. The implementation complexity is then evaluated and compared to standard OFDM in Section 4. Finally, in order to prove the innocuousness of the opportunistic transmitter towards the incumbent systems, a hardware setup where adjacent coexistence with a TV broadcast service is demonstrated and quantified. This is presented in Section 5.

2. TVWS operation with FBMC

2.1. TVWS limitations with OFDM

Several scenarios have been investigated for TVWS operation. In [7,8], the need for broadband systems is emphasized, where the radio system shall guarantee few Mbps to few tens of Mbps. Usually, such scenarios are covered by multicarrier systems using frequency multiplexes transmitted simultaneously over several subcarriers. The most classical approach is to use orthogonal carriers. With such a technique, referred to as OFDM, the receiver can equalize each subcarrier independently. When the system is adequately specified, each subcarrier is processed as a narrow band signal under flat fading condition, despite the broadband nature of the overall multiplexed signal. This property leads to a simplified receiver architecture even when frequency selective channels are considered. As a consequence, OFDM has been the initial choice for almost all modern broadband wireless systems, such as WiFi, WiMAX, 3GPP LTE and DVB-T. The OFDM signal is represented in (1).

d[n, i] = -ffiY, s[n> k]e>2niiAfTs

where the subcarrier spacing Df = 1/Ts in order to preserve orthogonality between carriers. The major drawback of OFDM relies in its spectrum shape. Each rectangular shaped subcarrier in the time domain, results in an sinc(f) function in the frequency domain, which summation reveals high spectrum sidelobes (first lobe at -13 dB). Usually, this issue is solved by filtering sidelobes with analog SAW filters at the RF transmitter. Unfortunately, this results in non-agile radios as analog filters are not flexible enough. Other variants proposed to smooth the transitions between the OFDM symbols by using windowing techniques [9]. As the guard interval has to be increased by the duration of the window, this technique efficiently addressed the agility constraints but at the cost of an important spectral efficiency reduction. It can also be stressed that OFDM is applied to systems where the spectrum mask is smoother than the one of the TVWS.

In the case of the TVWS both frequency agility and sharp spectrum roll-off are expected. The agility requirement stems from the wide frequency span of opportunistic channels across the UHF band (this depends on the country but typically from 450 MHz to 790 MHz), and a sharp roll-off is requested to guarantee coexistence with adjacent incumbents, which translates into ACLR specifications. In order to avoid non agile RF filters, spectrum sharpness may be tackled by implementing the filter in the digital domain.

Then, the filter is implemented after the IFFT of the OFDM modulator. However, it was shown in [10] that the complexity of the filter increases dramatically as the guard band is reduced (i.e., the complexity increases with the filter frequency steepness). This is illustrated in Fig. 1, where the complexity of filtered OFDM under TVWS spectrum mask is compared to the one of a 1024-FFT alone as a function of the used portion of the channel, in terms of number of real multiplication. The calculation was achieved for an 8 MHz channel with 15 kHz subcarrier spacing and an equiripple filter (0.5 dB in band ripple) under 55 dB ACLR condition.

It can be observed that the complexity of the filter to meet the ACLR specification is very significant. The complexity of the filter is already similar to the complexity of the IFFT when only 40% of the channel is exploited. It is 10 times higher when the occupation reaches 85%.

2.2. TVWS fragmented spectrum access

Due to the TV broadcast frequency planning, the TVWS is heavily fragmented. This means that the available channel map cannot be predicted. In some places, contiguous available channels are virtually inexistent. The situation in London is highlighted in [11], but similar fragmented available spectrum can be observed in many locations, and even in rural areas. In order to enable broadband services under such conditions, the spectrum pooling approach is envisaged [12]. It consists of using the parallel nature of the multiplex to switch off the subcarriers to avoid interfering with an in-band incumbent. Ideally, this technique should enable the creation of deep notches and should not impact on the transmission of the active subcarriers as each subcarrier is processed individually. However, with the OFDM waveform, similar smooth roll-off is observed where notches are enforced. Unfortunately, the filter designed to steepen the edges of the spectrum (see above) does not have a positive impact on the notch [10]. Thus, on top of its large complexity the digitally filtered OFDM does not have the flexibility to address TVWS fragmented spectrum properly.

2.3. FBMC operation for TVWS

An alternative modulation issues by combining filtering niques while keeping spectral was introduced in the 60s by has been revisited recently in approach, also known as FBMC

scheme is proposed to solve these and multicarrier modulation tech-efficiency optimal. This technique Chang [13] and Saltzberg [14], but the context of TVWS in [10]. This allows the control of the frequency

Fig. 1. Complexity of filtered OFDM in terms of number of real multiplications to meet TVWS ACLR requirements.

V. Berg et al. /Microprocessors and Microsystems xxx (2014) xxx-xxx

response of each carrier by introducing a filter bank, centered on every active carrier and based on the same prototype response, which can be used to control adjacent leakage, and even to virtually null it. Because the prototype filter spans across several subcarriers, the neighboring subcarriers are no longer orthogonal but interference can be controlled by introducing an offset quadrature amplitude modulation (OQAM) [15]. It should be mentioned that FBMC and OFDM schemes exhibit similar Peak-to-Average Power Ratio (PAPR) performance [16]. Both multicarrier systems may be described by a trans-multiplexer structure, i.e., a synthesis-analysis filter bank. The synthesis filter bank is composed of a set of parallel transmit filters that shapes and modulates the signal on the different carriers. The analysis filter bank consists of all the matched filters of the receiver. OFDM can be seen as a particular case of FBMC where inverse and forward Discrete Fourier Transform are used for the analysis and the synthesis filter banks respectively. The prototype filter of OFDM is a rectangular window whose size is equal to the duration of the FFT. At the receiver, perfect signal recovery is possible under ideal channel conditions thanks to the orthogonality of the subchannel filters. Nevertheless, under more realistic multipath channels a data rate loss is induced by the mandatory introduction of a cyclic prefix (CP), longer than the impulse response of the channel.

FBMC waveforms utilize a more advanced prototype filter designed to better localize the subcarriers. The prototype filter used in this paper is based on the frequency sampling technique [17]. This technique provides the advantage of using a closed-form representation that includes only a few adjustable design parameters. The most significant parameter is the duration of the impulse response of the prototype filter also called overlapping factor. The orthogonality between sub-carriers is maintained by introducing half a symbol period delay between the in-phase and the quadrature components of every complex symbol.

Complexity comparison between FBMC and OFDM has already been studied in [18] and is summarized in [10]. Complexity comparison is done by estimating the amount of real multiply necessary to perform the transmultiplexer function both at the transmitter and at the receiver. Complexity for FBMC and OFDM is here recalled in (2).

Cfbmc = 4 .(N • 2 + (N(log2(N)- 3)+ 4)+ KN • 2)

Ccp-ofdm = 2 .(N(log2(N)-3)+4) ()

where N is the number of carriers considered in the multicarrier waveform and K the overlapping ratio (the duration of the prototype filter is then equal to KN). Assuming an overlapping factor of K =4 and N = 1024, the overhead of complexity between OFDMand FBMC using (2) is almost equal to a factor of 5 for FBMC implementation. This comparison is a first step to understand the overhead FBMC may introduce in terms of hardware cost but does not take into account some important parameters. ACLR spectral requirements are not considered. An important part of the complexity may be required for filtering as already mentioned in Section 1. Some important functions of a multicarrier transceiver are also omitted such as synchronization, equalization and forward error correction. The overhead of complexity using an implementation angle should therefore also be considered.

3. Flexible implementation of a TVWS transmitter

3.1. Hardware implementation platform

Although the results provided in [10] were very promising, it is important to assess the results with an actual implementation of the FBMC transmitter. Considering the very stringent requirements on ACLR considered, finite dynamic range computation in the

modulator must be accounted for in order to check its potential impact on ACLR degradation. Similarly, the noise introduced by analog to digital converters (ADC) must be considered. Finally, signal generation with an actual RF circuitry using affordable components is necessary to show that the approach has some realistic potential. A preliminary implementation of the FBMC transmitter on the TVWS Flexible radio system (T-FleX) was introduced in [6]. The T-FleX system is a TVWS flexible digital radio including digital processing capacity, ADC and DAC converters, and RF front end TRX stages. The block diagram of the T-FleX system is shown in Fig. 2.

A Xilinx Kintex 7 (XC7K325T) that is memory mapped to an ARM Cortex A8 core (Texas Instruments OMAP DM3730) forms the core of the design. Ethernet PHY/MAC and Wi-Fi 802.11b/g are provided for easy local area network (LAN) connectivity. Bluetooth is added to allow for instance connectivity with a Smart-phone. The on-board USB On-The-Go (OTG) gives another connectivity option particularly useful to interact with personal computers or other personal devices. Interface to the RF daughter board is done through 2-dual DAC and one Quad ADC.

The digital board is interfaced to the RF TX daughter board in order to generate UHF signal. The ARM processor controls the digital board and the RF Boards and interfaces to an external PC via either a USB, an Ethernet or a Wi-Fi connection. The transmitter covers a 40 MHz window (equivalent of 5 x 8 MHz TV channels) that can be tuned within the whole UHF band from 470 MHz up to 860 MHz in a flexible way.1 Inside the 40 MHz window, the signal channelization is performed directly with the FBMC modulation by nulling unwanted carriers, i.e. using the spectrum pooling method. There is no additional filtering on the channelization path, neither in the digital domain, nor on the RF board, leaving the entire subcarrier allocation flexibility to the digital FBMC modulator. The bottom view of the T-FleX system and the TX RF daughter module is shown in Fig. 3.

3.2. FBMC transmitter architecture

A flexible FBMC transmitter-receiver or transceiver has been implemented on the FPGA of the T-FLEX platform. The transmitter architecture is composed of three main elements: forward error correction, data mapping and modulation followed by a digital front-end (see Fig. 4).

Forward error correction (FEC) is implemented around a standard convolutional encoder of constraint length k = 7 and rate /. The code may be punctured to support variable encoding rates. In order to keep a modular approach and map the code to the mul-ticarrier modulation, the convolutional code is segmented by blocks of a few multicarrier symbols. The trellis is closed at the beginning and end of each FEC block.

The second module, Mapping and Modulation module, maps and modulates the encoded bits to the multicarrier modulation. The coded data are mapped to a QPSK, 16QAM or 64QAM. Symbols are then padded to complete the transmitted burst into an integer multiple of multicarrier symbols. A synchronization preamble is added to the burst structure. The preamble is used by the receiver to perform synchronization and channel estimation. The generated block of data and preamble symbols are mapped to the active carriers and modulated to an Offset-QAM before being inverse transformed into a time domain sequence using an N-point IFFT. A polyphase network (PPN) filter structure completes the data stream and shapes the output of the IFFT over the duration of the prototype filter. A digital front-end completes the transmitter

1 Note that TVWS band spans from 470 to 790 MHz, after the decision of the ITU to free the 790-860 MHz (the so-called 800 MHz band for LTE operation worldwide).

4 V. Berg et al. /Microprocessors and Microsystems xxx (2014) xxx-xxx

Fig. 2. Digital architecture of T-FleX.

Fig. 3. T-FleX hardware bottom view.

to allow data rate adaptation with the digital-to-analog (DAC) converter sampling rate. It consists of a set of filters and interpolators.

An OFDM transmitter would exhibit a very similar architecture except for the mapping and modulation module as in that case

neither the OQAM transform, nor the PPN structure is required. It should be noted that the OQAM modulation imposes that the IFFT is called twice as often as it would be for an OFDM transmitter. Indeed, the OQAM transformation effectively converts a block of complex data at the symbol rate into two blocks of real data at twice that rate (to keep the same overall bitrate). The carrier mapping module allows for a dynamic burst by burst selection of the active carriers. A data and control interface between the FPGA and the ARM processor allows for a dynamic software control of the main parameters of the transmitter.

3.3. FBMC receiver architecture

Fig. 5 gives the implemented architecture of the FBMC receiver. The burst synchronization algorithms have been realized in the frequency domain. The frequency domain processing of the receiver combined with the high stop-band attenuation of the FBMC prototype filter results in a receiver architecture that allows burst-by-burst and flexible configuration of active carriers. The overall architecture of the receiver follows the principles of the

Fig. 4. Architecture of implemented FBMC Transmitter.

V. Berg et al. /Microprocessors and Microsystems xxx (2014) xxx-xxx

Fig. 5. Architecture of implemented FBMC Receiver.

Table 1

Complexity of main transmitter modules - flexible FBMC transmitter.

Slice Regs LUTs DSP48E1 RAM BLKs

Forward error correction

Scrambler 29 22

FEC segmentation 95 110

CC and puncturing 61 26

Bit interleaver 134 167 1 2

319 325 1 2

Mapping and modulation

Mapper 70 62 2

Symbol padding 81 77

Preamble generation 3553 1880

Carriers mapping 1108 603

OQAM transform 461 561 1

IFFT 5131 3805 13 10

PPN 577 677 16 4

10,981 7665 29 17

frequency domain using information on the preamble. This process is used to generate the coefficients of a one-tap equalizer. Data is then equalized (the equalization process also corrects for time synchronization errors) and filtered by the prototype filter before demapping. Log-Likelihood Ratios (LLR) of the received bits are then estimated for soft Viterbi decoding of the FEC. Synchronization, Channel Estimation and Equalization algorithms implemented on this platform have been described in [20]. Performance of the algorithms is good even for channels exhibiting large delay spread. For instance, when using the same carrier spacing and multicarrier symbol duration as for the 10 MHz LTE mode, the receiver performs well for channels with delay spread of up to 8.3 is. This corresponds to 1/8th of the multicarrier symbol duration.

FS-FBMC approach documented in [19]. A digital front end adapts the sampling rate used by the ADC to the symbol rate at the input of the FFT. A KN-FFT is then performed at the receiver on the signal without any regard of frequency or time synchronization. A frequency domain synchronization module estimates the start of the transmission burst and the possible frequency error before correcting the signal. The channel response is then estimated in the

4. Complexity evaluation and comparison

The architecture described in section III has been implemented on the T-Flex platform and mapped to the Xilinx Kintex-7 FPGA available on this platform. Complexity results on the Xilinx give a refined measure of the implementation overhead required by FMBC transceivers in comparison to OFDM transceivers. The design is constrained by its clock frequency that is set to 130 MHz.

Slice Regs

DSP48E1

Fig. 6. FPGA hardware components usage for the transmitter.

V. Berg et al. /Microprocessors and Microsystems xxx (2014) xxx-xxx

Table 2

Complexity comparison: OFDM vs FBMC transmitter.

Slice Regs LUTs DSP48E1 RAM BLKs

Forward error correction 319 325 1 2

Mapping and modulation 9943 6427 13 12

10,262 6752 14 14

Forward error correction 319 325 1 2

Mapping and modulation 10,981 7665 29 17

11,300 7990 30 19

TX FBMC complexity overhead 10% 18% 114% 36%

Memory Usage (Block RAM)

Fig. 7. Memory Block usage for the transmitter.

Table 3

Complexity of main receiver modules - flexible FBMC receiver.

Slice Regs LUTs DSP48E1 RAM BLKs

FFT 6615 4394 19 35

Delay line 292 170 68

Synchronization 6968 7435 38 3

Channel estimation 13,915 9718 49 12

Equalization and demapping 11,535 9433 38 7

FEC decoder 2439 5493 1 8

Control 13,206 13,453 10 38

54,970 50,096 155 171

4.1. Transmitter complexity

processing (DSP) cells dedicated to multiplication and accumulation (MAC) operations. The amount of cells used by the transmitter is summarized in Table 1.

More than the absolute complexity values of each transmitter modules, a relative complexity comparison between these modules for the different hardware elements in the FPGA is interesting to analyze (see Fig. 6).

As expected, most of the complexity is to be found in the inverse FFT. OQAM and PPN account for between 10% and 15% of the overall complexity. Preamble insertion is not absolutely negligible, nor carrier mapping as both modules have been designed to be relatively flexible and programmable. Forward Error Correction encoding only amounts for around 5% of the overall complexity.

Although the amount of DSP48E1, Xilinx DSP modules, is relatively low, their usage is almost equally split between IFFT implementation and PPN filtering.

Overall memory usage is relatively more difficult to estimate from FPGA synthesis reports as memory usage is reported per block of 18kbits. Blocks of Memory have been used at the input of the forward error correction to allow for adaptation of data processing rate. Most of the memory usage of the transmitter is to be found in the inverse FFT.

The complexity may be compared to a similar flexible OFDM implementation. An OFDM transmitter based on the same architecture as described in Fig. 4 has been implemented and mapped using the same constraints (130 MHz clock) into the Kintex-7 FPGA. Forward error correction complexity is the same design, mapping and modulation is simplified as OQAM transform and PPN filters are removed. Memory block has been reduced accordingly. Using these hypotheses, the overhead of complexity between OFDM and FBMC at the transmitter is summarized in Table 2 (see Fig. 7).

Finally, the overall complexity overhead at the transmitter between OFDM and FBMC is very limited. Apart for DSP48E1 which have been doubled by the PPN, the complexity overhead is reasonably low. In comparison to closed form complexity calculation, this result comes from the fact that the inverse FFT is processing data at twice the throughput the OFDM inverse FFT would be processed. Also, it is worth mentioning that the OFDM complexity reported in Table 2 does not account for the additional filter described in [10] and which is compulsory to comply with the ACLR requirement of the TVWS bands. The implementation complexity of such a filter, highly depends on the filter architecture, which is the reason why it is not provided in this evaluation.

The complexity of the flexible FBMC transmitter is given by the amount of Slice Registers, LUT and DSP48E1 cells used by the different modules of the design. Slice Registers correspond to the amount of register cells used, while LUT to the amount of combinatorial logic in the design. DSP48E1 cells are digital signal

4.2. Receiver complexity

The implemented receiver complexity has also been studied and is summarized in Table 3. Without any particular effort of optimization less than 40% of the Kintex 7 is occupied by the receiver.

Slice Regs

DSP48E1

FEC Control

Fig. 8. FPGA hardware components usage for the receiver.

V. Berg et al. /Microprocessors and Microsystems xxx (2014) xxx-xxx

Memory Usage (Block RAM)

FEC Decoder 5%

Equalization 4%

Channel Estimation 7%

Fig. 9. Memory Blocks usage for the receiver.

Table 4

Complexity comparison: OFDM vs FBMC receiver.

Slice Regs LUTs DSP48E1 RAM BLKs

FFT 5131 3815 13 10

Inner receiver 24,440 19,540 75 22

FEC decoder 2439 5493 1 8

Control 10,564 10,762 8 9

42,574 39,600 97 49

FFT 6615 4394 19 35

Inner receiver 32,710 26,756 125 90

FEC decoder 2439 5493 1 8

Control 13,206 13,453 10 38

54,970 50,096 155 171

RX FBMC complexity overhead 29% 27% 60% 249%

even on channels exhibiting large delay spread but comes at the cost of a complexity overhead. However, in the receiver architecture this module takes only around 10% of the overall design cost. A significant amount of memory blocks have been used by the main delay line of the receiver (Memory module on Fig. 5). This memory stores the FFT output symbols for synchronization, channel Estimation and PPN filtering. Channel estimation, synchronization and demapping are using almost half of the receiver resources (see Fig. 8).

The most significant amount of memory usage introduced by FBMC comes from the memory of the delay line. This memory block, along with the FFT memory block, is a direct consequence of the architecture implementation choice. TVWS requires a large adjacent channel rejection and therefore a relatively large overlapping ratio. The amount of data necessary to temporarily store is directly proportional to the duration of the prototype filter impulse response (see Fig. 9).

As for the transmitter a flexible OFDM receiver has been implemented using the same parameters and constraints to further analyze resource usage. When OFDM and FBMC are compared, digital logic occupancy is similar while memory usage is significantly increased in the case of FBMC. In terms of digital logic, FBMC takes around 30% extra area in comparison to OFDM (as the overall amount of signal processing units such as DSP48E1 is relatively small in comparison to logical units such as Slice Registers and LUTs). However, an overhead of almost 300% of memory is necessary for the FBMC implementation. This corresponds to 4 times more memory as memory usage is directly proportional to the considered overlapping ratio (K = 4) for the FBMC waveform, due to the FS-FBMC architecture used at the receiver (see Table 4).

This transceiver easily maps to the Kintex 7 (XC7K325T) available on the T-Flex platform. For the receiver approximately 25% of the available LUTs are occupied, 13% of the slice registers, 18% of the DSP cells and 19% of the RAM blocks.

This includes the non-negligible overhead the flexible implementation has put on the design. Control is taking almost a quarter of the design area, in both OFDM and FBMC cases.

This is mainly because the choice of the active carrier, the choice of the preamble and the impulse response of the interpolation filters at the equalizer has been left fully programmable and allows for flexible usage of the TVWS.

The FFT is implemented as mentioned in section III, and considers a KN size implementation. This allows for good performance

5. FBMC validation testbed and results

5.1. Test setup

The validation of the flexible TVWS FBMC transmitter is carried out in a twofold way: through spectrum profile measurements and by observing the impact of adjacent opportunistic communication on a TV broadcast signal. Thus, the setup of the validation testbed involves the TVWS transmitter and a DVB-T modulator, which

DVB-T Broadcast signal generator Perceptual impact on TV broacast

ui-uivi Spectrum measurement

• FBMC

Fig. 10. Validation testbed setup.

V. Berg et al. /Microprocessors and Microsystems xxx (2014) xxx-xxx

signals are combined to share the same medium. Then, the compound signal is split to be visualized on a spectrum analyzer on one hand and to a TV demodulator on the other hand. This setup is illustrated in Fig. 10.

An attenuator is placed at the output of the RF transmitter in order to adjust the TVWS versus the DVB-T relative power. An interesting setup consists of adjusting the TVWS transmission power in order to be on the limit of the DVB-T receiver performance. This limit is sometimes called Quasi Error Free limit, when a DVB-T service is available in presence of adjacent signal. It also enables to compare the respective impact of OFDM and FBMC on the DVB-T television demodulator.

The TVWS opportunistic transmitter implemented on the T-FleX system can be configured either in CP-OFDM or FBMC mode.

With the digital flexible transmitter, subcarriers can be switched on or off by blocks of 2 MHz in a global window of 20 MHz. This enables the implementation of spectrum pooling features to avoid interfering with in-band incumbent transmission. In both the CP-OFDM and the FBMC case, the subcarrier spacing is set to 15 kHz (similar to the one of 3GPP LTE systems). The total number of active carriers depends on the configuration which can be set through a GUI displayed on a host device.

5.2. ACLR measurements

First, ACLR is measured without the DVB-T transmitter. This enables to avoid the saturation of the spectrum analyzer while maintaining its noise floor to a low level. This makes the spectrum

Teklionix RSA 3408B

12/7/2012 3:16:10 PM

Frequency: 703.75 MHz Span: 5 MHz Input Att: 20 dB

Trace 2:

10 kHz

(Freeze)

(Normal)

^Jfreq/chan Cancel - Back Center Freq (Hz)

Д1-2: 1.78046875 MHz

-55.601 dB (95.6 dBc/Hz) 1

Marker: 702.38515625 MHz

-28.84 dBm (-68.84 dBm/Hz)

M iWsH ¡НИМ

j w

Start Freq (Hz)

701.25M

Stop Freq (Hz)

706.25M

10 dB/

Channel Table..

Center Freq Step Same As C.F.

Center Freq Step Same As Span

-99 dBm

Center: 703.75 MHz

Step Size (Center Freq)

Spectrum Analyzer: Measurement Off f| Center Freq (MHz): 703.75

Fig. 11. ACLR measurement for the FBMC modulation.

Fig. 12. Spectrum pooling measurement with CP-OFDM and FBMC.

V. Berg et al. /Microprocessors and Microsystems xxx (2014) xxx-xxx

analyzer maximal dynamics available for the ACLR measurement. This is required for the high dynamic signal (>50 dB) that need to be measured. The result is shown in Fig. 11, where the upper curve corresponds to the transmitter switched in OFDM mode whereas the lower curve corresponds to the FBMC case. It can be observed that the roll-off of the FBMC modulation is far steeper than the one of CP-OFDM. This confirms the simulations presented in [10]. ACLR measured between the two markers that are visible on the plot reveals an ACLR value of 55.6 dB, which is in line with the FCC specification; although no RF filtering is added on top of the FBMC modulation.

The second test is to measure the ability of FBMC to carve deep notches in the spectrum profile. This is illustrated in Fig. 12, where OFDM and FBMC are compared again. FBMC exhibits very steep inband and out-of-band roll-offs. Moreover, the spectrum shape

shows a clean flat profile inside the notch. Therefore, potential interference in the notch is expected to be similar to the one in the adjacent channels. On the contrary, OFDM neither manage to reach the same flatness, nor a low interference level even at its lowest point, which is some 20 dB above the FBMC floor.

The measured ACLR in this case is not as good as in the previous case. This is due to the fact that twice the power is present at the input of the spectrum analyzer, which impacts the dynamic available for the measurement.

5.3. Impact on DVB-T incumbent signal

A comparison with OFDM is also performed to measure the impact on a DVB-T reception and the potential interference of the TVWS transceiver onto the incumbent signal. The

Fig. 13. CP-OFDM and FBMC in presence of a DVB-T signal.

Tektronix RSA 3408B 3/14/2013 6:00:16 PM FREE RUN

Frequency: 646 MHz Span: 4 MHz Input Att: 25 dB

À1-2: -542.96875 kHz

RBW: 10 kHz

Trace 1: (Average) 20/20

Trace 2: (Freeze)

Cancel - Back

Fig. 14. Zoom of the transition zone between the TVWS transmission and the DVB-T signal.

V. Berg et al. /Microprocessors and Microsystems xxx (2014) xxx-xxx

measurement has been performed with a low cost USB DVB-T demodulator centered at 642 MHz (Channel 42). This kind of demodulator is more sensitive to interference than better quality TV tuners. When using direct adjacent channels on both sides of the DVB-T signal (64QAM 7/6), the OFDM signal must be attenuated by 9 dB in comparison to the FBMC signal in order to get a quasi-error-free DVB-T service. Fig. 13 shows a measured plot where the DVB-T signal (at the center) is combined with either an OFDM signal (lower plot) or an FBMC signal (higher plot). In both cases, the power is adjusted to get the quasi-error-free situation.

The 9db margin measured on this setup can be used either to increase the transmitted power of the opportunistic TVWS system or as an extra protection margin for the DVB-T receiver, or as a combination of both.

A zoom of the transition zone between the TVWS signal and the DVB-T signal is shown in Fig. 14, for the CP-OFDM and the FBMC cases.

6. Conclusion

In this paper, the opportunistic use of TVWS in presence of incumbent systems was discussed. Two major requirements have been stressed. These are the stringent ACLR performance of the TVWS transmitter that must be guaranteed to avoid interference with adjacent incumbent services and the flexibility required to address available channel, which map varies across time and space. Both requirements are unusual in wireless systems, where operating frequencies are typically known in advance and where ACLR constraints are relaxed (e.g. WiFi, LTE).

Classical CP-OFDM modulation cannot meet the ACLR requirement, unless the transmitter flexibility is sacrificed or spectral efficiency is compromised. On the other hand, the FBMC modulation was recently revisited for TVWS application. The architecture for a FBMC transceiver has been presented and a complexity evaluation has been studied. The overhead of complexity introduced by the FBMC waveform is limited to around 30% in comparison to OFDM in terms of logic. A significant overhead in memory is however necessary but this overhead may not be as costly in silicon design as memory blocks are highly optimized for occupancy.

Through actual measurements using a hardware flexible TVWS transmitter, it was confirmed in this paper that the FBMC modulation can meet the ACLR and coexistence requirements. The gain in ACLR and coexistence was compared with CP-OFDM. First, the ACLR performance of FBMC is significantly better than the one of CP-OFDM. Then, the coexistence with incumbent TV signal is improved by an additional 9 dB with the FBMC modulator. This performance was measured with an actual RF implementation, therefore accounting for all the impairments of the transmission chain. Besides, as the modulated signal is shaped digitally at the baseband, no compromise on the transmitter frequency agility is required. Therefore, it can be concluded that ACLR requirement of TVWS bands can be met with FBMC at the cost of a reasonable complexity overhead.

Acknowledgment

The research leading to these results was derived from the European Community's Seventh Framework Programme (FP7) under Grant Agreement Number 248454 (QoSMOS).

References

[1] FCC proposed rule, Unlicensed Operation in the TV Broadcast Bands, US Federal Register, vol. 69, No. 117, pp. 34103-34111 (18.06.04).

[2] FCC final rule, Unlicensed Operation in the TV Broadcast Bands, US Federal Register, vol. 74, No. 30, pp. 7314-7332 (17.02.09).

[3] Electronic Communications Committee (ECC) within the European Conference of Postal and Telecommunications Administrations (CEPT) Report 132, Light Licensing, Licence-exempt and Commons, Moscow, June 2009.

[4] Digital Dividend: Cognitive Access, Statement on Licence Exempting Cognitive Devices using Interleaved Spectrum, OFCOM, July 2009.

[5] R Datta, M. Gautier, V. Berg, Y. Futatsugi, M. Ariyoshi, M. Schuhler, Z. Kollar, P. Horvath, D. Noguet, G. Fettweis, Flexible Multicarrier PHY Design for Cognitive Radio in White Space, Cognitive Radio Oriented Wireless Networks and Communications (CROWNCOM), 2011 Sixth International ICST Conference on, vol., no., pp. 141-145,1-3 June 2011.

[6] D. Noguet, V. Berg, X. Popon, M. Shuehler, M. Tessema, T-FleX: A Mobile SDR Platform for TVWS Flexible Operation, IEICE Technical Report - Software Radio, SR2012-51, Fukuoka, October 2012.

[7] M. Mueck, D. Noguet, TV White Space Standardization and Regulation in Europe, Wireless VITAE, Chennai, March 2011.

[8] P.H. Lehne, R. MacKenzie, D. Noguet, V. Berg, O. Grondalen, Mapping Cognitive Radio System Scenarios into the TVWS Context, WinnComm SDR Europe, Brussels, June 2012.

[9] B. Farhang-Boroujeny, OFDM Versus Filter Bank Multicarrier, Signal Processing Magazine, IEEE, vol. 28, no. 3, pp. 92-112, May 2011, doi: 10.1109/ MSP.2011.940267.

[10] D. Noguet, M. Gautier, V. Berg, Advances in opportunistic radio technologies for TVWS, EURASIP J. Wireless Commun. Network. 2011 (2011) 170.

[11] M. Nekovee, A Survey of Cognitive Radio Access to TVWhite Spaces, Int. J. Digital Multimedia Broadcast., volume, 2010, Hindawi.

[12] T.A. Weiss, F.K. Jondral, Spectrum pooling: an innovative strategy for the enhancement of spectrum efficiency, IEEE Commun. Mag. 42 (3) (2004) S8-S14.

[13] R.W. Chang, Synthesis of band-limited orthogonal signals for multichannel data transmission, Bell Syst. Tech. J. 45 (December) (1966) 1175-1796.

[14] B.R. Saltzberg, Performance of an efficient parallel data transmission system, IEEE Trans. Commun. Technol., vol. COM-15, December 1967, pp. 805-813.

[15] P. Siohan, C. Siclet, N. Lacaille, Analysis and design of OFDM/OQAM systems based on filterbank theory, IEEE Trans. Signal Process. 50 (5) (2002) 1170-1183.

[16] A. Skrzypczak, J. P. Javaudin, P. Siohan, Reduction of the Peak- to-Average Power Ratio for the OFDM/OQAM modulation, Vehicular Technology Conference, 2006. VTC 2006-Spring. IEEE 63rd, vol. 4, no., 7-10 May 2006, pp. 2018-2022.

[17] M.G. Bellanger, Specification and design of a prototype filter for filter bank based multicarrier transmission, in: Proc. of International Conference on Acoustics, Speech and Signal Processing (ICASSP '01), vol. 4, 2001, pp. 2417-2420.

[18] M. Bellanger, et al., FBMC Physical Layer: A Primer, 2010, <http://www.ict-phydyas.org>.

[19] M. Bellanger, Fs-fbmc: An alternative scheme for filter bank based multicarrier transmission, in: Communications Control and Signal Processing (ISCCSP), 2012 5th International Symposium on, 2012, pp. 1-4.

[20] J.B. Dore, V. Berg, N. Cassiau, D. Ktenas, FBMC Receiver for Multi-User Asynchronous Transmission on Fragmented Spectrum, EURASIP Journal on Advances in Signal Processing, 2014, Special Issue on Advances in Flexible Multicarrier Waveforms for Future Wireless Communications, 2014:41 (28.03.14).

Vincent Berg is a research engineer at CEA-LETI working on physical layer development and implementation for cognitive radio applications. He graduated from Supelec, a leading French ''Grande Ecole'' of engineering, and obtained a M.Sc. in Microwaves from University College London in 1995. After graduating, he joined the industry on the development of transceivers for wire and wireless applications. From 1998 to 2004, he has been involved in the development of HF, VHF and UHF radio modems. He then joined Massana Ltd, an Irish start-up company, in 1998 and participated at the development of a Gigabit Ethernet PHY (IEEE 802.3ab) integrated circuit. In 2004, he joined Philips Semiconductors as a physical layer architect and developed demodulators for terrestrial (DVB-T, ATSC), cable (DVB-C) and satellite (DVB-S, DVB-S2) digital TV receivers. Since 2010, he works for CEA-LETI at the Wireless and Security Communications Department. He is currently leading the Wireless Solutions and Digital Prototyping Laboratory at CEA-Leti.

V. Berg et al. /Microprocessors and Microsystems xxx (2014) xxx-xxx

Jean-Baptiste Doré obtained his Masters degree in 2004 from the Institut National des Sciences Appliquées (INSA) Rennes and his PhD in 2007 on ''Joint code design and decoding architecture for LDPC codes'' from France Telecom R&D (Orange Labs) and INSA Rennes. He started working at NXP semiconductors in Caen, home business unit, as a signal processing architect. He was in charge on the study and design of signal processing algorithms for DVB-T/DVB-T2 products (OFDM waveforms). In 2009 he joined the Centre for Atomic Energy (CEA-Minatec) in Grenoble as a research engineer. His main research topics are signal processing, hardware architecture optimizations, PHY and MAC layers for wireless network. Jean-Baptiste Doré has published around 20 papers in international conference proceedings and book chapters, and is the main inventor of more than 10 patents.

Dr. Dominique NOGUET graduated from the National Institute of Applied Sciences (INSA) in electrical engineering in 1992. He obtained an MSc in microelectronics of the University of Strasbourg in 1994, and a PhD from Polytechnic National Institute Grenoble (INPG) in microelectronics in 1998. Then, he joined LETI as digital communication ASIC designer. As a designer, design architect and project manager he has worked on DSSS, UMTS, OFDM, MIMO modems. He coordinated several national and international EU collaborative projects. For instance he was the coordinator of the FP6 ORACLE project, the first EU project on opportunistic radio. He also lead flexible radio activities (WPRC) within the Network of Excellence NEW-COM++. More recently he has been involved as Technical Manager in the European QoSMOS project where he contributed on FBMC PHY and platform design. He was awarded two best papers, and received the best PhD award from INPG. He authored or co-authored 50 + papers in peer reviewed journals and conferences and holds 12 patents. He has been in the TPC of some major wireless conferences including ICC, GLOBECOM, CRONWCOM, VTC... Since 2008, he has been the chair of the Euromicro DSD Special Session on Flexible Digital Radio. He has been involved in cognitive radio standardization activities with IEEE DYSPAN-SC, P1900.6, P1900.6a and P1900.7 since 2008. He is currently the head of the Communication and Security Dpt at CEA-LETI, where he also leads cognitive radio activities. His main fields of interest are reconfigurable radio, cognitive radio and white space communication.