NI Wireless Testbed Solution for Novel Future Generation Communication Systems

  Read in   |   Print Print

"A key advantage of the NI testbed is that it can function in either real-time or offline mode, whereas many other testbed solutions offer only one of these functions."

- Pat Chambers, Heriot-Watt University

The Challenge:
Testing a wide range of fourth generation (4G) and beyond fourth generation (B4G) communications systems for multiple-input, multiple-output (MIMO), long-term evolution (LTE), spatial modulation (SM) with supporting channel measurements, orthogonal frequency division multiplexing (OFDM) and channel emulation.

The Solution:
Using NI PXI hardware and NI LabVIEW software to develop a suite of test programs which emphasis the versatility of the NI testbed solution.

Pat Chambers - Heriot-Watt University
Xuemin Hong - Heriot-Watt University
Nikola Serafimovski - University of Edinburgh
Abdelhamid Younis - University of Edinburgh
Raed Mesleh - University of Tabuk
Zengmao Chen - Heriot-Watt University
Yu Fu - Heriot-Watt University
William Thompson - University of Bristol
Cheng-Xiang Wang - Heriot-Watt University
Harald Haas - University of Edinburgh
Mark Beach - University of Bristol


Channel models and measurements used in conjunction with system simulations and system metric calculations can assess the viability of novel communications systems. However, they do not provide a complete picture of system performance. These approaches do not generally address issues such as hardware complexity, transmit and receive synchronization and implementation impairments. For example, many simulations only consider the baseband channel, which in turn could ignore the effect of frequency offset at radio frequency (RF) upconversion and downconversion stages. This type of performance assessment also often neglects other aspects of the communication link, such as the quality of the antenna at either end.

A wireless testbed solves these analytical shortcomings. The acquisition of this testbed forms part of the UK-China Joint R&D Centre for Future Wireless Networks and the UK-China Science Bridges: R&D on B4G Wireless Mobile Communications (UC4G), which both aims to encourage academic and industrial collaboration between the UK and China. The project partners chose the NI solution because it offered the possibility of both real-time and offline function, modular hardware design and versatility.

Testbed Design:

MIMO techniques exploit multiple antennas at both transmit and receive ends of a communications link. The testbed consists of a 4-channel transmitter chassis and a 2-channel receiver chassis to test these systems.

PXI System for RF Signal Transmission

  • We used a modular PXI System with 18 instrument slots for the hardware architecture for RF signal transmission (Tx). The chassis comprises three major components: an embedded PC controller, an NI FlexRIO FPGA module and a 4-channel radio frequency signal generator (RFSG). Each of these components is interconnected via the NI PXI backplane bus architecture.
    An NI PXI embedded PC controller controls the Tx Chassis and provides networking interfaces. It includes an Intel quad-core i7 1.73 GHz processor running Windows 7 with National Instruments LabVIEW™ software and The MathWorks, Inc. MATLAB® software, to control the operation of the testbed. The NI FlexRIO FPGA module includes a Xilinx FPGA chip and provides real-time signal processing programmed using LabVIEW. The 4-channel RFSG is broken down further into several NI devices: a single RF local oscillator (LO), four arbitrary waveform generators and four 6.6 GHz RF signal up-converters. The LO generates an RF reference signal and a 10 MHz reference clock. The four RF signal upconverters share the RF reference signal and the 10 MHz clock for synchronised transmission. The RFSG offers an operational frequency range of 85 MHz to 6.6 GHz and can facilitate a maximum bandwidth of 100 MHz and Tx power of 10 dBm. Each RFSG can  connect to an antenna as the final transmission stage.

PXI System for RF Signal Reception

  • The PXI system we use for RF signal reception (Rx) has the same 18-slot NI PXI chassis comprising an embedded PC controller and NI FlexRIO FPGA as the transmitter. However, we incorporate a 2 channel RF signal analyser (RFSA) instead of the 4-channel RFSG at the receiver. In addition, we can connect a 6 Terabyte hard drive array, known as a redundant array of independent disks (RAID), to the receiver.
    Similar to the RFSG, the RFSA includes several NI modules: one RF local oscillator, two digitisers and two 6.6 GHz RF signal down converters. Again, the LO generates an RF reference signal and a 10 MHz reference clock. The two RF signal downconverters share the RF reference signal and the 10 MHz clock for synchronised reception. The RFSA can operate in a frequency range of 10 MHz to 6.6 GHz with an operational bandwidth of 50 MHz.

Real-Time or Offline Operation

A key advantage of the NI testbed is that it can function in either, real-time, or, offline mode, whereas many other testbed solutions offer only one of these functions.

Using LabVIEW to write baseband signal processing steps to NI FlexRIO FPGA boards (at Tx and Rx) primarily facilitates real-time operation. After this, the RFSGs (at Tx) and RFSAs (at Rx) can interact with the FPGA boards to upconvert and transmit the signal before receiving and downconverting it. The LabVIEW FPGA module encourages user-friendly FPGA programming is based on the dataflow programming paradigm of LabVIEW. With LabVIEW we easily incorporate code we had previously written in VHDL or with MATLAB. Researchers on this project were able to develop a form of continuous data transmission (at Tx) as well as continuous data acquisition and processing (at Rx) using LabVIEW. Overall, the real-time operation is highly appropriate for live demonstration purposes and highly effective for system testing.

For offline operation, the testbed host PCs use LabVIEW to interact with their respective RFSGs and RFSAs. Each baseband signal processing stage is first written in code like C++or by using third-party software such as MATLAB®. In the case of transmission, this code results in a waveform written to a binary file. LabVIEW then interprets this file to upconvert the transmit waveform by the RFSGs. While at the Rx, LabVIEW controls the RFSAs so the resulting receive signal is downconverted and recorded before it is written to another binary file. After this, again, software such as MATLAB® opens the resulting binary file and appropriate baseband signal processing steps are performed on the baseband signal stored in the file. Offline mode has the advantage of conveniently facilitating minor system reconfigurations when testing.


The following systems have been tested: MIMO long term evolution (LTE), spatial modulation with supporting channel measurements, orthogonal frequency division multiplexing (OFDM) and channel emulation.

  • MIMO LTE: For the MIMO LTE standard, the Vienna University of Technology have recently provided an entire suite of The MathWorks, Inc. MATLAB® software functions that incorporates all of the baseband transmit and receive signal processing as well as a selection of channel models. Frame recovery, synchronization and frequency offset compensation algorithms were developed into the simulator’s baseband signal processing stages at the Rx in order to successfully facilitate the simulator’s operation with the testbed. Two MIMO LTE signal processing variants were used namely ‘Multiplexing’, which is a technique that emphasizes higher data transmission rates, and, ‘Diversity’, which is a technique that lowers data transmission rate in a tradeoff for greater reliability. The testbed was set up in a Line-of-sight (LOS) channel. Measurement data was used to establish the fact that the channel was Rician distributed. The same data was used in conjunction with SM and will be described later. Bit error rate (BER) curves for the two MIMO schemes were recorded using the testbed and, for comparative purposes, using software generated Rayleigh channels in Fig. 1. The Testbed and Rayleigh curves match reasonably well for diversity but are offset in the case of multiplexing. This however is entirely consistent with MIMO theory as Rayleigh channels statistically incorporate more signal scattering than LOS channels and since multiplexing is designed to work better in rich scattering (Rayleigh) channels, it is reasonable that a LOS channel would provide slightly poorer results than a Rayleigh one, while the performance of diversity would be less dependent on channel type.
  • Spatial modulation (SM): This is a novel MIMO communications technique where two mechanisms for conveying information across a wireless link are employed together. The first mechanism is the conventional one where data is modulated by means of mapping segments of the transmit information stream to a series of transmit symbols. The second more novel mechanism is to take other segments of the transmit information stream and map them onto specific choices of transmit antenna from an array of antennas. Then, by activating a certain choice of transmit antenna at any given moment in time, this will lead to the conveying of information if the receiver can determine which transmit antenna was active, i.e. determine the choice of transmit antenna that was made. In relation to the testbed, a challenge here is for the receiver to determine which RFSG out of the set is active at various points of time for successful demodulation at the receiver.

    Although researchers here have shown that the system overall exhibits reasonably low complexity hardware requirements, all theoretical analyses of the performance characteristics of SM have revealed it to exhibit highly complex behaviour in this regard. As a result, an experimental validation of SM is a highly valuable exercise and the NI testbed has been the first apparatus in the world to do this.

    Furthermore, since an important part of analysing the behaviour of this system is to understand the wireless channel behaviour, i.e. statistics, a system for ascertaining this was complementarily developed in tandem here. Fig. 2 shows the BER performance behaviour of SM, which was acquired in offline mode. The experimental environmental is similar to the one reported for MIMO LTE and Fig. 2 indicates that the experimental performance of SM matches the theoretical performance when two important properties are incorporated. These properties are that (i) the channel exhibit Rician statistics with high K-factor and (ii) all RF chains, i.e. RFSGs and the RFSAs, exhibit power imbalances (PIs). PIs are effectively differences in gain that occur in RF chains to due to component design tolerances. For SM, a set of 2 (Tx) x 2 (Rx) MIMO RF chains were used and in order to determine properties (i) and (ii), algorithms were developed to measure the wireless channel and the results in Fig.3 confirm that the resultant 4 channels, each denoted, ‘h’, with appropriate sub-script, was indeed Rician with similarly high K-factors. Not shown due to space constraints, are the results showing the PIs. These were obtained using the same measurement algorithms but where the wireless channels were removed from the measurement and substituted for transmission lines to perform an effective system calibration.   



Figure 1. Comparison of BER Curves for MIMO Diversity and Multiplexing Schemes Using The testbed and Software Generated Rayleigh i.i.d Channels.


Figure 2. BER Performance Characteristics of Spatial Modulation. (Experimental, refers to testbed performance; while, Sim, and Ana, refer to software simulation and mathematical analyses respectively; PI refers to power imbalance).



Figure 3. Channel Statistics of Testbed Channels in Testing Spatial Modulation and MIMO LTE (Rician distributed with similarly high K-factors but different means; each channel is assigned an appropriately subscripted, h).


  • OFDM: Here the Tx and Rx signal processing architectures have been developed in LabVIEW FPGA for a single RFSG/RFSA link. Segments of a transmit information stream are mapped onto symbols and are then assigned subcarriers. Pilot sequences are then generated to estimate the channel and a guard bit ensures robust OFDM symbol transmission. An Inverse Fast Fourier Transform (IFFT) is then applied to the sequence where typically a cyclical prefix is also used to ensure circular convolution with the channel and hence a reduction in inter-carrier interference (ICI). These processes are then undone at the Rx end to recover the original bit sequence. The system runs in real-time and a constellation diagram from this system is depicted in Fig.4 showing the demodulated 16 QAM symbols as well as two pilot tones. For demostration purposes, the user can, for example, block or move the antenna into a different polarisation state and observe in real-time the effect on the constellation diagram and hence on system performance.


Figure 4. LabVIEW FPGA OFDM System Showing Received 16 QAM Symbols With Two Pilot Tones.


  • Channel emulation: The concept of channel emulation is that of being able to take an arbitrary channel model and convolve it with transmit data thus emulating the effect of any desired channel model on a given set of transmit data. The main advantage over software approaches is the higher (real-time) speed of such a device. The FPGA capability of the testbed was used to develop this system. In the testbed, the PC reads the data from a local hard drive, in which the in-phase and quadrature (IQ) components of baseband modulated signals and the channel impulse response (CIR) coefficients of the channel model, are stored. Then, these data are passed through dynamic memory access channels (DMA) channels to an FPGA module. Once the data has been combined correctly, convolution between the CIR and the baseband IQ data is performed mathematically in FPGA. The following FPGA functionality was developed in order to do this: (i) Shift registers: These are used to store and shift the IQ baseband samples (acting as tap delays) and channel coefficients, (ii) Control logic to load channel coefficients: These are to enable serial to parallel conversion and load channel coefficients into multipliers regularly, (iii) Multipliers: These are implemented to complete the complex multiplication of baseband signals and channel coefficients, and (iv) Adders: The pipelined adders are to improve the processing clock rate. In order to verify the correct functioning of the channel emulator, its channel frequency response was evaluated, i.e., dividing the output data of the channel emulator by the corresponding input data in frequency domain.. In Fig. 5, the result was compared with the channel model that was desired to be implemented and an exact match can be observed. A C4 WINNER II channel model was used.


Figure 5. Comparison of Sample-by-Sample Division of Channel Emulator Output and Input Data Streams (Channel Emulator) with Desired Implemented Channel (Reference Channel).


Versatile, Flexible Testbed Solution

We created a suite of programs to test 4G and B4G technologies including: MIMO LTE, SM, OFDM and a channel emulator. SM was the most novel system we tested. We also developed channel measurement algorithms to verify SM performance characteristics and justify MIMO LTE system results. We used both the real-time and offline mode of operation of the NI testbed, which highlighted the versatility of the NI testbed solution.

MATLAB® is a registered trademark of The MathWorks, Inc.

Author Information:
Pat Chambers
Heriot-Watt University
Heriot-Watt University, Riccarton
Edinburgh EH14 4AS
Tel: 07557130286

Bookmark and Share

Explore the NI Developer Community

Discover and collaborate on the latest example code and tutorials with a worldwide community of engineers and scientists.

‌Check‌ out‌ the‌ NI‌ Community

Who is National Instruments?

National Instruments provides a graphical system design platform for test, control, and embedded design applications that is transforming the way engineers and scientists design, prototype, and deploy systems.

‌Learn‌ more‌ about‌ NI