Resumen de Superimposed training for channel estimation in next-generation wireless multicarrier techniques
The wireless communications have changed in recent decades. The data requirements of the users have changed exponentially with the improvements of the mobile terminal technologies. The user data rate necessity is related with the new applications appearance. The required bandwidth has increased in function of these trends. Following this tendency, the academic community has developed a huge interest in response to the needs, then, several wireless technologies generations were developed proposing solutions to the requirements of the users. The communications generation that first was developed is 1G, which bases its functionality in offering the voice transmission. This standard (considered as analog) was first proposed in the 80s and 90s. The most relevant technologies that were commercially presented were as part of the 1G were total access communication systems (TACS) and advanced mobile phone system (AMPS).
The next evolution of the mobile communication was presented at the beginning of the 90s and were called the second generation (2G). Several services as picture and text messaging at low data rates and digital voice were offered in this technology generation. Another interesting evolution was the 2.5G which provides, using general packet data rate (GPRS) technology, an improved throughput. The technology can offer data rates of up to 144 kbps. The services applications known as web browsing, instant messaging and email transmission can be provided by 2.5G technology.
The following generation was the third generation (3G) which begins to offer services in 2000's as web, roaming, and TV, services with data rates up to 2Mbps. Several new user equipment’s appear here known as smart phones allowing services as real time video, web services, reaching bandwidths up to 20 MHz.
In the case of the 4G generations new data rates up to 100 Mbps are offered which includes the possibility of supporting new services as broadband multimedia messaging and broadcasting services. In this case the access technique that allows provide this throughput is orthogonal frequency division multiple access (OFDMA) or code division multiple access (CDMA). The offered bandwidth can vary between 5 to 20 MHz.
For the next generation known as 5G, it is expected that a latency of 1 mS can be guaranteed and that 500 Mbps can be provided by this mobile technology. In this generation several techniques as small cells, non-orthogonal multiple access (NOMA), beamforming, massive multiple input multiple output (MIMO), and edge computing, have been researched to offer high data rates that can face the new users demands. Some new techniques as millimeter waves (mmWave) has been proposed where ultra-high frequency (UHF) bands are implemented to transmit information at very high frequencies.
A common technology that is implemented are the multicarrier techniques which can increase the bandwidth channels with narrow bands. In this case, avoiding the interference the signals can be transmitted ideally overlapping nearby signals thanks to the orthogonality that is provided by the transmission system.
OFDM is the most widely employed multicarrier technique and it is going to be considered as part of 5G mainly because it offers compatibility with previous OFDM-like standards. It uses a cyclic prefix (CP) to eliminate the interference provoked by the nearby OFDM symbols in time-domain. Other multicarrier technique that has been studied into the literature is filterbank multicarrier offset quadrature amplitude modulation (FBMC-OQAM) which can eliminate the use of CP increasing the spectral efficiency compared to OFDM. This technique reduces the inter-symbol interference (ISI) utilizing greater symbol lengths and certain level of overlapping. One of its drawbacks is the intrinsic interference caused by the surrounding symbols due to the lack of subcarrier orthogonality.
Another popular way to increase the bandwidth to meet the users demand is using a promising technique known as visible light communication (VLC) which conveys information at the visible light spectrum. This technique assures greater data rates and bandwidth thanks to the lack of interference and the frequency band where it works. VLC uses DCO-OFDM (DC offset OFDM) as a multicarrier technique to provide to the light emitting diode (LED) with a real-valued signal.
In coherent communications, channel estimation is a necessary step to recover the transmitted information. Pilot symbol amplitude modulation (PSAM) is a very popular technique which uses dedicated resources to transmit pilot signals to estimate the channel. This technique obtains good channel estimation error at the cost of spending dedicated resources in the transmission of pilots. Another attractive and promising technique is superimposed training (ST), which can be used for 5G and beyond, because it increases the spectral efficiency which is highly desirable in current and future standards. This technique improves the spectral efficiency due to the fact that no dedicated pilot subcarriers are used at the cost of a reduced quality of channel estimation error. The literature of ST has been extended to relay networks and using MIMO systems. Also, some investigations for the analysis of peak-to-average power ratio have been presented for ST.
ST methods have attracted much attention because of the advantages obtained in spectral efficiency compared with PSAM techniques. This one arithmetically sums the information signal with the pilot signal that later is used for estimating the channel. ST do not waste bandwidth, unlike PSAM, due to no subcarrier is used exclusively for pilot transmission. ST schemes are promising technologies especially for next-generation 5G wireless networks where higher demands of data rates are expected. These techniques have been proposed in 5G to exceed the existing limits like pilot contamination interference in massive Multiple Input Multiple Output (MIMO) or combined with modern proposals such as Non-Orthogonal Multiple Access (NOMA) but using only simple methods of superimposition.
Several ST schemes are proposed to be used in orthogonal frequency division multiplexing (OFDM). Here, a technique called symbol-blanking superimposed training (SBST) retires data symbols where pilot symbols will be transmitted. Then, the retired data symbols are spread over a group of data symbols next to them, inducing interference and degradation in the system performance.
In ST over OFDM transmission, some authors proposed to precode the information and then to null data in the subcarrier positions where the pilots are going to be transmitted. Then, a channel estimation error like in PSAM is reached, but with a reduction of the symbol error rate (SER) performance. This effect can be seen specially in high order constellations. Some iterative mechanisms have been proposed to recover the system from the nulling process.
In practice, some mechanism tries to emulate the performance of a PSAM mechanism reducing the interferences and, in some cases, redistributing it until the point there is no superimposition.
Some hybrid mechanisms have been proposed in the literature for OFDM, where a data detection and channel estimation mechanism are combined in an iterative loop. Here, a bit interleaved coded modulation (BICM) coder accompanied with an interference cancellation technique are used to improve the data detection performance but with a prohibitive complexity.
In this thesis, we analyze novel channel estimation techniques called superimposed training (ST) for different multicarrier waveforms for future wireless systems, 5G and beyond. The channel estimation consists in statistically obtaining the channel state information to recover the transmitted signal at the receiver. This process can be done applying mainly pilot-based techniques called pilot symbol assisted modulation (PSAM) and ST. PSAM techniques consists in the transmission of pilot training symbols as reference signals using dedicated resources for obtaining the CSI. The channel estimation error in this case is dependent on the noise at the cost of reduced average channel capacity due to the fact that dedicated pilot is transmitted instead of data. This one is the most used scheme in current standards. ST is a promising technique that can provide higher spectral efficiency compared with dedicated pilot proposals because it does not use dedicated resources for channel estimation at the cost of increasing the channel estimation error. A new data interference signal is added to the noise increasing the degradation effect in the transmitted signal.
New ST proposals have been presented in literature for single carrier and multicarrier techniques. These ones take advantage of the averaging and the size of the multicarrier symbol blocks that are going to be transmitted to try to reduce the channel estimation error.
In this context, a new proposal called partial-data superimposed training (PDST), proposed by the authors, is addressed for orthogonal frequency division multiplexing (OFDM) systems. Some subcarriers are used to transmit data and others to transmit a combination of data and pilot signals. The transmission power is divided accordingly to the power allocation factor that is required. The novelty of PDST is that incorporates an additional power control factor that, unlike previous proposals, allows to improve performance with an accurate control of data and pilot interference. Based on this proposal the channel estimation error is derived and compared with the simulation results. The analytical fits the simulations result which validates the analysis. The signal-to-interference and noise ratio (SINR) and average channel capacity are later introduced based on the channel estimation error that was deduced. The partial superimposition can overcome classical overlay schemes reducing the number of resources affected by the superimposition. This scheme is tested using QPSK and 16QAM modulation techniques where it is possible to see that denser constellations cannot support a high level of superimposition. Here, it is possible to see that PDST can get a better performance under the same conditions in comparison with other ST schemes. Then, the average channel capacity is compared and it can be seen that PDST overcome the others using the superimposed level obtained in SER simulations.
Secondly, channel estimation proposals and their performance using a promising multicarrier technique known as filterbank multicarrier offset quadrature amplitude modulation (FBMC-OQAM) are evaluated using channel estimation error and average channel capacity. In this contribution, we model the introduction of a ST sequence for channel estimation in an FBMC-OQAM system. Other PSAM-like channel estimation techniques are evaluated for comparison purposes. Due to the superimposition, new parameters appear in the formulation known as intrinsic interference and data interference that combined degrade the transmitted signal, the channel estimation error, and the average channel capacity. An analytical expression for the channel estimation error, which fits with the simulation results, is depicted. Then, the SINR and its average channel capacity expressions are presented using the previously deduced expression of channel estimation error. Here, the distribution of the pilot and data power is controlled by the power allocation factor. Furthermore, the factor that maximizes the average channel capacity is found. It is shown that this proposal overcomes the performance of FBMC-OQAM with PSAM and other multicarrier techniques as OFDM with PSAM and ST. Simulations results are presented to validate the analysis.
Finally, the channel estimation in a promising technology known as visible light communication (VLC) is analyzed. In this proposal, we model an VLC scenario where ST is used for channel estimation. The multicarrier technique DC-offset OFDM is considered to adapt the electrical signal to the optical one. Then, multiple input single output (MISO) is used as the most adequate spatial multiplexing technique to work with VLC. In this novel technique an extensive analysis is done for deduction of the channel estimation error, SINR and spectral efficiency. Simulations results generated by ray-tracing techniques validate the performance of each proposal in a quasi-stationary environment compared with PSAM-based channel estimation techniques. Simulations results validates the analysis that is presented.
This channel estimation techniques are evaluated in different scenarios where new analytical expressions are found and used to deduce new expressions of average channel capacity. These expressions are compared with other multicarrier techniques and channel estimation mechanism to evaluate their effectiveness. The superimposed training modifies the analytical expressions in a different way depending on the multiplexing technique that was used.
High spectral efficiency and data rate can be expected of implementing these proposals which successfully could satisfy the new data-rate requirements. All of these proposals consider multicarrier techniques and novel technologies to be used in 5G and beyond systems thus offering new lines of investigation.
This thesis is organized as follows:
In chapter I, we present a general overview of the research topic to be analyzed. Then, some literature of the state of the art aiming the research topic is reviewed. Thus, the scientific contributions of the author are denoted. Finally, the thesis chapters are briefly described.
In chapter II, some concepts of the multicarrier techniques are explained introducing the state of the art that in the following chapters will be used for analysis. Some new techniques as OFDM, FBMC, universal filtered multicarrier (UFMC), and generalized frequency division multiplexing (GFDM) are introduced. Some variations of OFDM for VLC as DC-biased optical OFDM (DCO-OFDM) and asymmetrically clipped optical OFDM (ACO-OFDM) are presented. The block diagrams at the transmitter and at the receiver is analyzed.
In chapter III, a new superimposed training technique called PDST is described. Here, the generated interference signal is analyzed, and a power control factor is introduced for the adequate pilot and data power assignment. Some expressions are deduced to describe the data interference signal, the channel estimation error and the average channel capacity. The SER of the ST channel estimation techniques is also figured out, it is shown that PDST is better than the other techniques. Taking into account the power allocation information the average channel capacity expression is tested, and it is shown the superiority of the proposed technique. PDST overcomes the rest of the proposals presented in literature with an specific power allocations factor. Several simulation scenarios are tested to show the performance of the proposal in comparison with PSAM-like and ST.
In chapter IV, the application of ST for channel estimation over a new waveform known as FBMC-OQAM is analyzed. FBMC-OQAM presents better average channel capacity than orthogonal frequency division multiplexing (OFDM) due to the fact that it avoids the use of CP using well localized filters which reduce the spectral leakage between subcarriers. The main drawback of this technique consists in the generation of an intrinsic self-interference provoked by the neighboring OQAM symbols and an additional signal known as data interference that is the product of the data superimposition. In this case, to avoid the effects of inter-symbol interference this technique transmits longer multicarrier symbols and it takes advantage of the averaging and the stationarity of the system. The analytical expression of the channel estimation error is deduced. Later, this expression is used to denote the SINR and average channel capacity. Simulation results validate the analysis and show the performance of the system under different stationarity parameters.
In chapter V, the combination of VLC applying ST technique for channel estimation is investigated, considering a MISO system configuration which is very common in VLC. In this case, ST cannot be easily applied to VLC, then, an analysis of this combination is carried out. Here, an analytical expression of the mean squared error (MSE) using a least squares (LS) estimator which matches with the MSE obtained with simulations is obtained. Additionally, spectral efficiency expressions are obtained to compare the performance of PSAM and ST techniques in this scenario. Furthermore, the optimal values for the maximum spectral efficiency are provided. Finally, simulation results are presented at the end of the chapter.
In chapter VI, the conclusions of the analysis of the proposal is presented highlighting the main ideas of the investigations and future lines of research that can be derived from the previous proposals are provided.
Finally, the references that were taken into account in this thesis and in the papers of our proposals are presented.
