The COMONSENS Multiuser MIMO Testbed

The COMONSENS Multiuser MIMO Testbed is a wireless network hardware demonstrator for the practical evaluation of multiuser multiantenna transmission techniques developed in the realm of the COMONSENS project. The COMONSENS Multiuser MIMO Testbed was jointly designed and implemented by the GTAS research group at the University of Cantabria and by the GTEC research group at the University of A Coruña.

The testbed consists of three transmit and three transmit/receive nodes each equipped with Multiple Input Multiple Output (MIMO) capabilities. The testbed is suitable for assessing the performance of different transmission methods over wireless network channels, such as multiple access channels, broadcast channels, relay channels or interference channels. Obviously, it is also useful to evaluate transmission methods for point-to-point channels.

Both transmit and receive testbed nodes are equipped with a Quad Dual-Band front-end from Lyrtech, Inc

  • Up to eight antennas connected to four direct conversion transceivers
  • Based on Maxim MAX2829 chip (also found in frontends like Ettus XCVR2450 or Sundance SMT911)
  • Dual-band operation: 2.4 to 2.5 GHz and 4.9 to 5.875 GHz band

  • Programmable variable attenuator to control the transmit power value up to 25 dBm per transceiver
  • We use dual-band antennas Mobile Mark PSKN3-2455S

Baseband hardware of the three receivers from Lyrtech

  • VHS-DAC and VHS-ADC modules with eight DACs and eight ADCs
  • 8 x DACs: 14 bit @ 480 Msample/s
  • 8 x ADCs: 14 bit @ 105 Msample/s
  • Receiver nodes can also be used as transmitters
  • Each pair of DAC/ADC is connected to a single transceiver of the RF front-end
  • The signals are passed in I/Q format

Baseband hardware of the three transmitters from Sundance

  • Custom design based on the SMT8036E (SMT310Q + SMT365E + 2 x SMT370-AC)
  • 4 x DACs: 16 bit @ 400 Msample/s
  • A single DAC is connected to each transceiver of the RF front-end
  • The signals are passed in IF, employing only the I or the Q branch of the RF front-end
  • RF carrier is shifted at the receiver and appropriate filtering is also applied to remove undesired signal replica

All six testbed nodes are synchronized in frequency from the same 40 MHz oscillator. We employ an accurate oscillator Miteq XTO-05-040-G-15P-OP15 with a frequency stability of 0.1 ppm. Additionally, all transmit nodes are synchronized in time, thus allowing us to precisely control transmission instants.

Experimental Evaluation of Interference Alignment based on IEEE 802.11a

In wireless communications, an interference channel is obtained when various transmitter-receiver pairs operate simultaneously, sharing the same transmission medium.

Employing the COMONSENS Multiuser MIMO Testbed we assess the feasibility of Interference Alignment ([Maddah] [Cadambe]) for the case of three users equipped with two antennas per user at each side of the link and sending a single data stream per user, thus resulting in nine different 2x2 MIMO channels that are recreated in a measurement scenario located in the GTAS laboratory as shown below.

With the aim of facilitating the experimental evaluation of interference-channel problems under real-world channels, we make 5844 channel realizations freely available to the research community. Such channel realizations were collected during a measurement campaign carried out in Santander, July 2011, by means of transmissions based on the IEEE 802.11a standard, but extended to support 2x2 MIMO. Except for the organization of the pilots, all parameters corresponding to the signals transmitted by each antenna are standard-compliant, including the number of subcarriers, subcarrier spacing, guard subcarriers, guard interval (800 ns), channel bandwidth, etc. The carrier frequency was fixed at 5.610 GHz, employing different antenna combinations at the transmitter as well as at the receiver nodes to obtain all channel realizations.

Each of the 5844 channel realizations consists of 52 channel matrices (one per data/pilot subcarrier) with 2 rows and 2 columns (2x2 MIMO). Those 5844 channel realizations are divided in four different groups consisting of 1461 channel realizations each. Such groups correspond to four different Modulation and Coding Scheme (MCS) values for the transmitted signals. Although we follow exactly the same automated measurement procedure for each group, such measurements were carried out at different time intervals, thus not guaranteeing that they correspond to exactly the same realization. Additionally, we consider a channel realization consisting of nine 2x2 MIMO matrices. Therefore, if any of the 2*2*6 = 24 antennas is changed, then a different channel realization is obtained. Consequently, changing a single antenna in the measurement setup does not guarantee that all resulting 2x2 MIMO links will change in at least one channel coefficient. For this reason, if we only look at the 1461 channel realizations corresponding to a specific MIMO link, we will find out that several of them are repeated. In order to facilitate the access of all different 2x2 MIMO channel realizations, thus not restricting the measurements to the 3-user 2x2 MIMO interference network, we have incorporated three different methods to extract all different 2x2 MIMO channel realizations from the whole measurement data.

Download channel measurements

The file contains the channel measurements ready to use in Matlab as well as detailed instructions of how to access them.

MIMO multiuser testbed

Have a look to the demonstration of the MIMO multiuser testbed build and integrated in the COMMONSENS project.

COMONSENS UPM Wireless-Sensor-Networks (WSN) Testbed

The smart grid, vehicular networks, ambient intelligence or the Internet-of-Things are emergent domains of applications in which myriads of interconnected devices sense and control the environment. In many of these applications, energy-efficient devices and robust networks are mandatory. Therefore, transmitting packets to central-stations is discouraged and the devices have to rely on learning from small sets of local data. Using distributed algorithms the devices communicate only within their neighborhoods, but they process the data in a manner such that the information is diffused and they can benefit from experience gathered the whole network (even when the nodes are not directly connected). This kind of in-network processing has received much attention in the last years and several theoretical contributions have appeared. One popular approach is based on the 'consensus' algorithm. Nevertheless, when we consider real-world deployments of WSN, many issues could arise. In particular, packet loss, time delays, clock drifts, and changes in the topology make the diffusion of the information a difficult task. Therefore, it is necessary to validate theoretical contributions in more realistic scenarios. The COMONSENS UPM WSN Testbed is a wireless sensor network for rapid prototyping and validation of distributed signal processing algorithms in realistic midsize networks. It has been designed and implemented by the GAPS research group at the Technical University of Madrid (UPM). Our main goal is to provide a flexible and modular tool so the users can port their theoretical contributions to a real deployment with minimum effort. The COMONSENS UPM WSN Testbed has already been successfully used to validate distributed robust estimation algorithms developed by Barcelona Tech (UPC) and University of Vigo, as well as for a distributed voice detection algorithm developed by the Technical University of Madrid. WSN devices have typically low-memory, with low-computing resources and with constrained power-supply. This is an extra challenge when deploying real-world distributed signal-processing algorithms. The COMONSENS UPM WSN testbed is composed of commercially available econotag motes that include a Freescale ARM7 microcontroller with 802.15.4 radio. The motes run Contiki operating system that provides multithreading and low-power Internet communication even for memory constrained devices. It also includes the Cooja simulator that emulates the whole functionality of a network before burning the program into the hardware. It is important to remark that many distributed algorithms require strict synchronization and perfect data transmissions to guarantee correct results. We have developed a set of functions that abstract the signal processing algorithm from the actual within-neighborhood-communication, so synchronization and data coherence is achieved transparently, which speeds up the development cycle.

Related Publications

Publications related to the COMONSENS Multiuser MIMO Testbed and UPM Wireless-Sensor-Networks Testbed containing the details of the measurement campaigns carried out for evaluating the feasibility of Interference Alignment in indoor environments employing single-carrier transmissions:

  • O.González, D. Ramírez, I. Santamaría, J. A. García-Naya, and L. Castedo,"Experimental validation of interference alignment techniques using a multiuser MIMO testbed," in Proc. International ITG Workshop on Smart Antennas (WSA 2011), Aachen, Germany, Feb. 2011, DOI: 10.1109/WSA.2011.5741921

  • J. A. García-Naya, L. Castedo, O. González, D. Ramírez, and I. Santamaría, "Experimental evaluation of interference alignment under imperfect channel state information," in Proc. 19th European Signal Processing Conference (EUSIPCO). Special session on Interference Alignment, pp. 1085-1089, Barcelona, Spain, Aug. 2011. Available online.

  • S. Valcárcel Macua, C. Moreno León, J. S. Romero, S. Silva Pereira J. Zazo, A. Pagès-Zamora, R. López-Valcarce, and S. Zazo, "How to Implement Doubly-Stochastic Matrices for Consensus-Based Distributed Algorithms," in Proc. IEEE Sensor Array and Multichannel Signal Processing Workshop (SAM2014), A Coruna, Spain, June 2014.