ECE 4901 Fall 2015 Implementation of a MIMO Transceiver Using GNU Radio Ethan Aebli (EE) Michael Williams (EE) Erica Wisniewski (CMPE/EE) The MITRE Corporation 202 Burlington Rd Bedford, MA 01730 Department of Electrical and Computer Engineering University of Connecticut, Storrs 06269 United States September 28, 2015 This research was supported by the MITRE Corporation as an independent research project. The views, opinions, and findings are those of the authors and are not intended to convey or imply an official position of The MITRE Corporation.
1 Team 1604 - Proposal I. SUMMARY The MITRE Corporation is a not-for-profit organization which operates federally funded research and development centers known as FFRDCs. Their main offices are located in Bedford, Massachusetts and McLean, Virginia. MITRE provides valuable assistance to the United States government in many important areas of scientific research and engineering. In addition to fundamental research, MITRE also provides technical analysis of products and services the government may be considering purchasing from third parties. In so doing, the government is able to leverage MITRE s unbiased technical expertise into its own decision making process. An important area of research and development for MITRE is in wireless communications. Much of the work done in Bedford aims to improve existing technologies and communication systems. Software defined radio (SDR) is a relatively new development that allows engineers to design and test systems much more quickly that could have been done in the past, when much of the signal processing was done on dedicated hardware. GNU Radio is an open-source simulation and modeling platform that interfaces well with the USRPs and is used extensively at MITRE. Currently the SDRs have MIMO capability but GNU Radio does not, as yet, ship with that functionality. MITRE has many military communication systems modeled in GNU Radio and would like to be able to quickly determine if applying MIMO to a given system would enhance the current performance or reliability. We at UCONN have been tasked with writing the software to meet that need. A. Wireless Channel II. BACKGROUND To understand why MIMO attracts such interest, first we must understand the environment in which we are operating, the challenges such an environment presents and what may be done to overcome them. To this end we start with a brief description of the wireless channel and a few of its most important properties. The primary difference between wired and wireless communication channels is the number of paths a signal can take from the transmitter to the receiver. Transmitting over the wire leaves little choice, however, for a wireless signal the possibilities are endless. For a wireless signal, each path travelled from transmitter to receiver results in a uniquely modified version of the original signal, where each version is subjected to various levels of attenuation and phase changes. The type of path travelled can be characterized in one of two ways. The first is a line-of-sight (LOS) path, where there exists a direct unobstructed path from transmitter to receiver. Under these conditions received signal power follows the familiar inverse square law P r = βd v P t, (1) where the path loss variable v equals 2, the parameter β is dependent on frequency and other environmental factors. The second type is no-line-of-sight (NLOS) where no direct path exists and the signal must follow a round-a-bout route to the receiver. Conditions commonly experienced by electromagnetic waves include reflection, diffraction and scattering. Scattering occurs when a signal encounters objects in its path whose size is much smaller than the signals wavelength. Reflection takes place when the path is obstructed by objects much larger than the wavelength of the signal. Diffraction typically occurs due to sharp edges like those exhibited by an office building or doorway. Additionally, electromagnetic waves may experience refraction or absorption. All of these factors conspire to reduce signal power and whose effects may be experienced by the receiver as either large or small-scale fading. B. Large-scale Fading When signal power attenuates due to distance, antenna losses or filtering losses, this is referred to as large-scale fading. Interestingly, the power loss
2 experienced by the receiver due to these large-scale factors is often made irrelevant by the much more serious fading condition discussed below. C. Small-scale Fading At the receiver multiple attenuated phase-shifted copies of the original signal interfere with one another in a rapid and random fashion. These multipath waves combine to form a time varying signal. The effects of multipath interference are so great that they completely overwhelm the largescale effects. Depending on the relationship between the channel and the signal, small-scale fading can be subdivided under the following headings: fast fading, slow fading, flat fading or frequency selective fading. As is often the case in wireless communications, the receiver is a mobile device, therefore the distance between the transmitter and receiver is changing. This makes the channel not only a function of distance, but also a function of time. The receiver, having no concept of distance experiences a time varying channel. The velocity of the mobile device coupled with the movement of objects in the environment enable us to examine the relationship between the channel and signal in both time and frequency in a way that can be used to define the four types of small-scale fading. To begin, we may classify the channel as narrowband or wideband depending on the spectral and/or temporal relationship between the signal and the channel. It is important to note that narrowband or wideband channels do not exist in isolation but are only defined in relation to time and frequency parameter of the signal. A channel may be defined as narrow or wideband depending on the symbol period or transmission rate of the signal. Narrowband channels have a bandwidth larger than that of the signal. From a time domain perspective, the symbol period is larger than the impulse response of the channel which means that the channel is not changing for the duration of the symbol. Narrowband channels give rise to flat fading, because all frequency components of the signal are effected equally by the channel and there is no intersymbol interference. The channel is termed wideband if the bandwidth of the signal is greater than the bandwidth of the channel. In this case different spectral content effectively experiences a different channel. In the time domain the symbol period is less than the impulse response of the channel leading to intersymbol interference. We term this type of fading as frequency-selective. Doppler shift also plays an important role in determining the type of fading experienced at the receiver. Due to multipath propagation effects, signals are received at many different angles. Since Doppler shift is defined by f d = v cos θ (2) λ we see that the shift in frequency is directly proportional to the observed angle of the incoming wave. This leads to random time varying frequency modulation at the receiver. Say we transmit a signal at a given frequency, f c then due to the Doppler shift, the range of frequencies experienced by the receiver range from f c f s to f c + f s. This phenomena is known as spectral broadening or Doppler spread and the relationship between this range of frequencies compared to the baseband bandwidth of the signal leads to our final classification as either fast or slow fading. Slow fading occurs when the amount of Doppler spread is small compared to the signal bandwidth. Here, changes in the channel are slow in relation to the symbol duration. Fast fading occurs when the channel changes rapidly within a single symbol because the Doppler spread is significant in relation to the baseband bandwidth of the signal. The fast or slow fading nature of the channel can also be described in terms of the coherence time and coherence bandwidth of the channel. These parameters describe over what frequencies and for what time duration we can consider the channel to effectively remain static. D. Channel Models There are several models we might employ to describe the wireless channel. We have chosen to focus our attention on the Rayleigh model because it best describes a typical urban environment in which we are typically trying to communicate. The Rayleigh model claims no LOS between transmitter and receiver. If we have I versions of the transmitted signal each taking a separate path, the receiver then experiences a summing of these multipath components. A signal subjected to AWGN, transmitted
3 with carrier frequency, f c and having amplitude and phase, a and φ respectively, can be described as r(t) = a i cos(2πf c t + φ i ) + η(t). (3) If we expand this equation to reveal the in-phase component, A and quadrature component, B r(t) = cos(2πf c t) where and a i cos(φ i ) i=i sin(2πf c t) A = B = a i sin(φ i ) + η(t), (4) a i cos(φ i ), (5) i=i a i sin(φ i ), (6) then we see that the summation terms are actually random variables determined by random environmental factors. From the central limit theorem we know that no matter the underlying probability distribution, if enough samples are taken then the distribution of that curve tends to be Gaussian. This means that each summation term is an independent identically distributed random variable, whose envelope R = A 2 + B 2 follows a Rayleigh distribution, thereby making our received signal a Rayleigh random variable with probability distribution f R (r) = r r2 exp, r 0, (7) σ2 2σ2 which when applied to the received power takes the following form f(x) = 1 x exp, x 0. (8) 2σ2 2σ2 The above formula reveals the problem we are attempting to overcome through the use of our proposed MIMO configuration. Because of the random nature of the channel the receiver may encounter rapid changes in signal power to the point where the signal may not be recovered. Receivers are typically rated to reliably recover the transmitted signal at a given signal-to-noise-ratio (SNR) or higher. When the SNR drops below this threshold then we experience what is known as an outage. One possible solution to this problem is to increase diversity thereby decreasing the probability of outage. III. SOLUTION (TECHNICAL PROPOSAL) Diversity is a form of repetition where we transmit carefully crafted versions of our original signal. Because the different versions fade independently we decrease the probability that all versions of the signal will be below the SNR threshold for the receiver at a given time. We now have the opportunity to choose the signal with the highest SNR or to combine all versions into one strong signal. Similar to definitions describing our fading model, diversity can be defined in terms of a relationship. In this case the relationship exists between the received SNR, γ and the probability of error, P e and given by the following expression representing diversity gain, G d = lim γ log P e log γ. (9) Several methods exist to obtain a gain in diversity. We can separate our signal copies in terms of space, time, frequency or polarization. These methods may be combined to further improve the diversity gain, thereby reducing the outage probability. Also, it can be shown that as the number of transmit and receive antennas increase, so does the capacity of the channel. When the number of transmitters, N is greater than the number of receivers, M, the lower bound on the capacity is given by N C > log 2 (1 + ( γ /N) χ k ). (10) k=n M+1 Therefore, by implementing a MIMO scheme we have increased the channel capacity which enables us to increase our data rate while still being able to reliably decode our transmitted signal. We make certain assumptions and acknowledge certain limitations in choosing a MIMO scheme. First we assume no channel state information (CSI). This means that the receiver has no knowledge of the channel to help in the decoding process. A method of differential encoding can be carried out prior to transmission to overcome this obstacle. Secondly we assume a flat fading channel where the channel will remain static over the symbol duration. Given these design parameters we have chosen to implement a popular space-time block code, referred to as the differential Alamouti scheme. Due to the following recursion, [ ] x1 (n) = 1 [ ] [ ] x1 (n 1) x 2(n 1) s1 (n) x 2 (n) 2 x 2 (n 1) x 1(n 1) s 2 (n)
4 we are able to decode our message without prior knowledge of the channel. The differential method encodes information in the relationships between successive symbols, greatly reducing the effects of phase and frequency offsets. If the channel causes our transmitted signal to become inverted, the information it contains is still preserved. IV. PROJECT PHASES, TIMING, MILESTONES The plan for our project involves four major milestones to accomplish our goals. We have each milestone broken down into pieces and our timing for the each milestone vary based on the goals that need to be met. Our first major milestone is to be able to have a simulation for both our MIMO and SISO designs. We want to have each individual part working before we put together our simulation which is what we have been working on for the past few months leading up to our project proposal. From here we want our complete simulation finished by Thanksgiving break and we want to be able to start working on the rest of our project at that point in time. The second milestone that we want to meet is to be hardware ready and have a working model in GNU Radio. We plan to start this at the end of fall semester and work on it throughout the Winter Break. We intend to have the basic software for GNU Radio complete by the start of Spring Semester so that we can be hardware ready at the beginning of the semester and allow ourselves some time to get to know the USRP hardware better and make sure our software will work. This leads into our third milestone where we want to get the GNU Radio software working on the USRP. We will start out with the transmitter and receiver wired together to make sure we have a working design. We then will add the antennas and get our design working wirelessly. We want this milestone to be reached by Spring Break so we have time after to continue testing. The final milestone brings us to the end of our project. We need to test our design to see how it works and make changes and small edits to allow our design to include other designs and ideas and maybe allow us to accomplish more through testing. We will start our testing indoors, using the building as an obstacle to overcome. We are hoping to get our testing outdoors and incorporate a little more distance than what could be regularly found when testing indoors alone. This brings us to our final goal and where we want to be at the end of the year with our design project. V. BUDGET Thanks to MITRE and UCONN we have been supplied with all we need to complete our project. MITRE has supplied the USRPs and associated hardware, while UCONN has supplied the computers and lab space we need to carry out our work. VI. INFORMATION ABOUT PERSONNEL AND COLLABORATORS We would especially like to thank Michael Wentz, Lead Communications Engineer at MITRE for all his help and guidance during the course of this project. Our faculty advisors Professor John Chandy and Professor Peter Willett also deserve our respect and gratitude. VII. REFERENCES Hamid Jafarkhani, Space-Time Coding, Theory and Practice. Cambridge University Press, 2005. Georgios Giannakis, Space-Time Coding For Broadband Wireless Communications. Wiley, 2007.