The user clustering problem in an uplink MIMO Non-Orthogonal Multiple Access (NOMA) scheme is considered here. The receiver is assumed to operate in two sequential stages that employ Linear Minimum Mean Squared Error (LMMSE) receivers. At the first stage, the receiver is designed to recover the transmission from a cluster of selected users/nodes. The contribution of these users is then subtracted from the received signal and the remaining user transmissions are then linearly recovered. The determination of which users should be detected during the first stage is formulated as a deep learning based multiple classification problem. In order to guarantee that the selection is robust to fast fading, the input to the neural network is based on second order channel statistics. Furthermore, the training process is simplified by using a large system approximation of the resulting sum-rates. Simulation results indicate that the proposed deep learning-based solution is able to achieve a significant rate advantage with respect to other lazy approaches, such as fixed or random cluster assignments.

In this paper we present results on the time dynamics of the most used cellular bands, from the measurement campaign conducted in seven European cities. Special attention was given on acquiring comparable datasets between the different measurement locations. We provide detailed examples of how much variability there is in the measured activity and we also test the validity of common assumptions usually found in the literature. In particular, we show that the spectrum occupancy metrics exhibit significant degree of stationarity in all the measurement locations and in all of the frequency bands studied. We also characterize in detail the influence of the diurnal cycle, i.e., the differences in spectrum use during different times of day. In addition to raw spectrum occupancy metrics, we study the structure of the ON/OFF activity patterns for cellular bands. Our results show that the ON and OFF period durations in our data sets have significantly heavier tails than predicted by the commonly used Markov models.

Cognitive Radios provide communication devices with the flexibility to adjust to varying network and channel conditions. For this to be fully realizable spectrum sensing and signal reception have to happen simultaneously and have to require as little power as necessary to function in handheld devices. This work argues for the need of flexible digital-front ends as indispensable building block, able to perform control operations over the analog front-end and to perform sensing and synchronization procedures without the need of power consuming baseband processors. A low power, reconfigurable digital front-end that supports concurrent synchronization and sensing of high-throughput wireless standards is presented. Multiple operating modes, useful for various communication standards, such as LTE, WLAN and DVB-T are introduced and analyzed. The digital front-end has been implemented in 65 nm CMOS technology resulting in a chip area of 6.4 mm2. Fine grain clock gating allows synchronization at 4 mW and sensing at 7 mW power consumption. Experiments in combination with a reconfigurable analog front-end show that a 1.7 GHz wide frequency band can be scanned based on energy detection in an exceptionally low time window of 10 ms while consuming 13 mW power and that coarse energy detection can speed-up the sensing process. Furthermore, advanced feature detection for DVB-T and LTE signals is implemented and measured. Low power sensing of DVB-T signals shows that a target false alarm rate of 10 % and a detection probability of 90 % at an input power level of 106 dBm while consuming 7 mW power are possible. Synchronization-aided FFT-based LTE sensing with leakage cancellation was experimentally validated for various bandwidths showing a power consumption of maximum 20 mW.

In recent years, many SDR base band processors have been proposed to meet the high performance and programmability requirement for emerging wireless communications. To be able to support hundreds of Mbps or even Gbps wireless communications, such SDR base band processors often have massive parallel computation capability. This promising processing capability may also be exploited for other types of signal processing tasks. Our work explores the feasibility of performing challenging media processing on SDR base band processors. In this paper, we will show exploratory experiments for supporting full HD H.264/AVC media decoding on a recent version of ADRES based SDR base band processor. Two computational dominant tasks, motion compensation and deblocking filter, have been selected to experiment on the processor. These two blocks account about 80% of the total execution time. Since the processor is designed to be wireless domain specific, algorithm and architecture co-optimizations are crucial to make the goal feasible. Results show that, with limited architecture extension, the ADRES based base band processor achieves very competitive performance and efficiency even when compared with several architectures that are specifically optimized for the media decoding.

We report initial results from 48 hour spectrum occupancy measurement campaign that was done in time-correlated fashion in seven different European locations. We give a description of the measurement campaign and provide results from our preliminary analysis of overall duty cycle in the frequency range 110-3000 MHz. The paper particularly focuses on traffic and duty cycle patterns in GSM 900 and GSM 1800 bands. For the sake of completeness, we also discuss ISMband utilization in two of our measurement sites. We show that spectrum utilization is generally low, but that variance between frequency bands and locations is significant, which means that any statistical claims on occupancy statistics need to be done carefully before regulatory claims are made.

The design of multi-Gbps LDPC decoder has become a hot topic in recent years as the demand of the transformation towards 4G. In this paper, we describe an energy efficient multi-Gbps LDPC decoder engine based on ASIP using Target tool suite. The ASIP core can be configured as half-layer paralleled or quarter-layer paralleled decoding, which offers a good trade-off between the throughput and power/area efficiency when compared to the state-of-art fully paralleled ASIC based multi-Gbps LDPC decoder. When the ASIP core is instantiated for 802.11ad, it achieved a throughput up to 5.3 Gbps at 5 iterations with a latency of less than 150 ns and a record energy efficiency of 4.3 pJ/bit/iteration in 40G TSMC technology for the coding rate 13/16, showing to be competitive versus published ASIC solutions.

The Category-5 (Cat-5) UE defined by LTE, as the most demanding category, requires processing 20Mhz bandwidth and 4×4 MIMO transmissions. Very little progress has been reported for its feasibility on programmable processors. In fact, most related work focus on lower categories with much less throughput. Since MIMO signal processing complexity increases non-linearly even with the simplest linear MIMO detectors, 4×4 MIMO transmissions combined with 20Mhz bandwidth is much more challenging when compared to lower UE categories. Our work explores the feasibility of software defined baseband for the most demanding UE category. On a customized SDR baseband processor, we have recently accomplished a software defined downlink inner receiver for Cat-5 LTE UE. The implemented inner receiver includes fully fledged synchronization and data detection functionalities, including coarse CFO estimation/compensation, I/Q imbalance estimation/compensation, OFDMA demodulation, channel estimation, fine SCO/CFO estimation/compensation, MIMO channel processing, MIMO data detection and LLR generation. Both linear MIMO detectors and more advanced MIMO detectors have been studied. To the best of our knowledge, this is the first work experimenting practical Cat-5 LTE receivers on baseband processors.

We present results from the comparison of measurements between dedicated embedded spectrum sensing chip targeting low-cost and low-power applications and a high-end spectrum analyzer. We use different signal types, including actual spectrum usage measured simultaneously with the two devices. We analyze the typical problems such devices suffer from. Finally we also study how to estimate the spectrum sensing results once the device type is taken into account.

Lattice Reduction aided soft output MIMO detectors (LR-SOMD) have been demonstrated to offer a promising gain. This work explores the potential of implementing a LR-SOMD on a parallel programmable baseband processor. In this paper, first a LR algorithm called the Data Regularized Parallel Lattice Reduction algorithm (DRP-LR) is proposed. Afterwards, a low-complexity LR-SOMD, Radius Constrained Multi-Tree Selective Spanning (RC-MTSS) is presented. RC-MTSS uses a novel multiple-tree search approach for LR-SOMD, while combining the benefits of Sphere Detection (SD) and Selective Spanning with Fast Enumeration (SSFE). A fixed complexity LR-SOMD, Multi-Tree Selective Spanning (MTSS) is also proposed for implementation. Both the algorithms, DRP-LR and MTSS, are enabled to exploit data level parallelism (DLP) and instruction level parallelism (ILP). In order to evaluate performance, the proposed DRP-LR and MTSS are implemented on the ADRES baseband processor for a 4 × 4 LTE system using QAM-64. DRP-LR achieves an average throughput of 33.33 M LR per second, which is comparable to recently reported ASIC implementations, while MTSS shows an average throughput of 730 Mbps on the same processor. To the best of authors' knowledge, this is the first reported implementation of a LR-SOMD algorithm on a parallel programmable baseband processor.

Multi-gigabit LDPC decoders are demanded by standards such as IEEE 802.11ad and IEEE 802.15.3c. In order to achieve high throughput, most published multi-gigabit designs use row-paralleled architecture. In this paper, we proposed a half-row paralleled LDPC decoder with half layer level pipeline and single permutation network for the 802.11ad standard, which reduces the hardware resources almost by half compared to the state-of-the-art row-paralleled LDPC decoder, achieving a good trade-off between energy efficiency and area efficiency. The decoder achieves a throughput of 5.6 Gbps and consumes only 99 mW for the highest coding rate 13/16 at 5 iterations, working at 500 MHz by using 40nm G technology, yielding an energy efficiency of 3.53 pJ/bit/iteration and area efficiency of 35 Gbps/sqmm.

Emerging high throughput wireless communication standards, such as LTE/LTE-A and IEEE 802.11ac, impose exciting challenges on SDR baseband implementations. Our work explores the feasibility of SDR baseband for the most demanding-modes in those emerging high throughput standards. On a customized C programmable SDR baseband processor (with compiler support), we have accomplished realtime inner receiver implementations for Cat-4/5/7 LTE/LTE-A UE and up to the 80MHz 4 × 4 mode of IEEE 802.11ac. The implemented inner receiver includes all essential synchronization and data detection functionalities, including coarse CFO estimation/compensation, I/Q imbalance estimation/compensation, OFDM(A) demodulation, channel estimation, fine SCO/CFO estimation/compensation, channel tracking, MIMO channel processing, MIMO data detection, LLR generation, etc..

The Chinese Digital Television Terrestrial Broadcasting System has a complex PHY layer definition with many different modes including two different block transmission schemes (OFDM and SC) and three different known symbol padding cyclic extensions, some of which with phase rotation between blocks that break the cyclicity. The block sizes with or without cyclic extension are “non power of two” numbers. This plurality of modes and the unusual block sizes make the design of a signal processing architecture very difficult. In addition, the known symbol padding extensions are intended for channel estimation but have poor auto-correlation properties; hence the channel estimation in long multipath channels is degraded and not suitable for high order constellations. We have designed a novel unified receiver architecture supporting all modes of this broadcasting system, capable to start from a poor initial channel estimation. We describe in detail this architecture and provide simulation results supporting our system choices.

Rotated constellations have shown promising potentials for harsh channel conditions and it becomes an crucial feature of DVB-T2. Several other standards, such as new generations of ATSC and DTMB, are also considering it. How-ever, the superiority of performance comes at the cost of orders of magnitudes higher complexity for demodulation, which is a crucial part of the receiver. Constellations of conventional QAM modulations stay on completely regular integer points, so that the demodulation can be as simple as quantization and table lookup. However, the rotation and independent transmissions of I/Q branches break the above properties, so that soft demodulation becomes much more complex. Previous efficient solutions are mostly based on reduced search instead of performing a full search. We take a totally different approach that derives closest constellation points with geometrical transformations. Search is completely avoided. This results in orders of magnitudes of complexity reduction.

Cognitive Radio requires the architecture of radio systems to combine reception and spectrum monitoring functionality efficiently. We propose a flexible digital front end that supports concurrent synchronization and sensing of high-throughput wireless standards. The chip is implemented in 65 nm CMOS technology resulting in a chip area of 6.4 mm2. Fine grain clock gating allows synchronization at 4 mW and sensing at 7 mW power consumption. Experiments with the chip in combination with a reconfigurable analog front end show that a 1.7 GHz wide frequency band can be scanned based on energy detection in an exceptionally low time window of 10 ms while consuming 13 mW power. Feature detection of DVB-T signals is implemented and measured as well and achieves for a single autocorrelation step a performance target false alarm rate of 10% and detection probability of 90% at an input power level of-106 dBm while consuming 7 mW power.

We address the problem of precoding in downlink multi-user MIMO communications using SDMA. In this setup, multiple antennas are used at the base station and at the terminals. A double level of spatial multiplexing is present: the users are spatially multiplexed (SDMA) and each user receives spatially multiplexed symbol streams (SDM). The cumulative multi-user and multi-stream interference is a potential performance limitation in this scheme. To mitigate this, we propose a linear precoder that has the following properties: it can be computed analytically; it can be computed without iterations; it improves on the state-of-the-art solution based on the generalized eigenvectors; it is computed separately for each user. Although the formulation of the multi-user precoding can be separated from the multi-stream precoding, our solution optimizes the multi-user separation taking the multi-stream processing into account. The performance of our precoding scheme is assessed by simulations.

In this demonstration paper we describe a prototype of an LTE system deployment that opportunistically exploits the spectral white spaces in the upper UHF TV bands, intelligently guided in its spectum access by a radio environment map (REM). The architecture is modular in the sense that interfaces are generic and minimal. In the proposed demo we will illustrate how information of primary transmitters and other secondary transmitters as well as estimates of the radio field strength over frequency, time and space can be made available and exploited by a secondary TDD-LTE base station to make judicious decisions on its spectral occupation.

In this paper we report on spectrum use measurements carried out synchronously over the period of 48+ hours in seven European cities. Special care has been placed on harmonizing the measurement settings and equipment so as to obtain as comparable data as possible.We describe in detail the measurement setup, including the coordinated preparation activities carried out across the different measurement sites.We present preliminary analysis of the obtained data set, and particularly highlight similarities and differences in spectrum use between selected measurement locations.We plan to release later the full data set for research community for the further research.

Mobile data traffic is increasing dramatically, essentially due to the exploding use of wireless applications. To cope with this, one solution is the deployment of many small base station sites to cover the hot spots. With the deployment of more heterogeneous networks, the total energy expenditure will increase. Traditionally, the techniques to improve the energy efficiency of small base stations exploit the network load variation and the idle time. However, they do not consider that the resource scheduling can have an important impact on energy consumption when the network is not fully loaded. In fact, forcing the system to transmit a packet when the channel conditions are poor, results in low energy efficiency. In this paper we propose an adaptive scheduling algorithm for pico base stations that applies the optimal scheduling algorithm of “water-pouring” in time in a full LTE framework using consolidated pico base station power models from EARTH project. Our approach reduces the energy per information bit by transmitting at full load during good channel conditions and going into sleep mode the rest of the time. We quantify the achievable gains with realistic channel and network conditions and we show energy savings of up to 61% in the energy per information bit with a marginal impact on packet delay.