Gunes Karabulut Kurt - Istanbul Technical University, Wireless Communication Research Laboratory
Selahattin Gokceli - Istanbul Technical University, Wireless Communication Research Laboratory
The Istanbul Technical University (ITU) wireless communications research laboratory (WCRL) researches wireless communications topics including fading channels; multiple input, multiple output (MIMO) systems; multicarrier communication systems; orthogonal frequency division multiplexing (OFDM); signal processing for MIMO systems; coding; cooperative communications; network coding; and advanced security issues in communication systems. Ongoing projects involve theoretical/experimental works, as well tests and taking practical measurements.
Our group works on various aspects of the wireless communication; hence, we needed a common platform to easily share our expertise and get fast results out of algorithm prototypes. Due to the types of projects and deadlines associated, we needed a flexible yet reliable platform that could serve our needs.
Before using the NI platform, our group used open source tools and mathematical modelling languages to simulate and prototype on SDRs such as Ettus Research USRPs (Universal Software Radio Peripherals). With NI’s platform, we saw the opportunity to achieve faster results. We had dense trainings and workshops in the early stages that helped us achieve competency with certifications, which resulted in a high level of productivity in a short time frame. We also noticed the NI ecosystem, which included developers and users from academia and industry, a supporting layer of partners, and NI’s local branch offices that provide superior technical support.
We have presented our work using NI’s SDR platform in various international conferences with subjects such as “CP and Pilot Jamming Attacks on SC-FDMA,” in International Conferences on Signal Processing and Communication Systems/Australia; “Universal Filtered Multicarrier Systems: Testbed Deployment of a 5G Waveform Candidate,” in IEEE Sarnoff Symposium/USA; “Network Coded Cooperation in Delay Tolerant Networks,” in IEEE International Conference on Wireless for Space and Extreme Environments (WiSEE); and more. These conferences awarded us for our presentations and papers and included us in user groups and discussions in which we have shared our research and our ability and level of confidence in using NI’s graphical design platform.
We also coauthored a book chapter using our lab setup, “Cognitive Radio Testbeds: State of the Art and an Implementation, Spectrum Access, and Management for Cognitive Radio Networks,” (Springer).
About the Application Challenge
In one of our recent projects, we addressed a communication waveform technique challenge in 5G.
We expect the enormous increase in the usage of mobile devices to continue in the future, and we should adopt novel communication techniques. We tackled the encountered design problem within the scope of 5G research activities. Here, the communication waveform technique used is a crucial component due to its vital role in transmission. Specifically, the selected waveform technique should offer sufficient quality at less strict synchronization requirements and should also address dense usage cases, where a high number of users need to effectively share limited resources.
OFDM is an example of a waveform technique, and it is frequently used in recent communication standards. Until recently, OFDM met quality requirements. However, it is not suitable for 5G systems due to its drawbacks such as tight synchronization requirements, inefficient spectral properties like high spectral side lobe levels and cyclic prefix (CP) overhead. New waveform techniques have been proposed to overcome these drawbacks. Among these, universal filtered multicarrier (UFMC) is a leading candidate. Less strict synchronization requirements increase UFMC’s importance as a 5G candidate in which we can predict dense network scenarios with simple synchronization architecture.
In this project, we implemented the UFMC technique with SDRs in real time for the first time in the literature. Using efficient synchronization components and proper parameter selection (which is determined throughout experiments), we handled real-time design issues of the UFMC and achieved better spectral results over the OFDM usage. Furthermore, by exploiting the sub-band filtering approach of UFMC, we proposed enhanced versions of UFMC. These techniques mainly focus on peak-to-average power ratio (PAPR), I/Q imbalance, and synchronization issues, which we can also experiment and verify with our testbed.
We implemented our technique using two USRP-2921 SDR devices, a PXI-6683H timing module, and a PXIe-5644R transceiver module for hardware. We used LabVIEW, the Modulation Toolkit, and the Spectral Measurements Toolkit as software.
We conducted real-time tests over the USRP measurements with bit error rate (BER) and error vector magnitude (EVM) metrics to observe performance. As demonstrated with results, a proper implementation solution for UFMC makes it superior to OFDM.
USRP-2921 devices are SDR nodes, which function in 2.4–2.5 GHz and 4.9–5.9 GHz bands and can offer 20 MHz bandwidth as well as 100 MS/s. We configured these devices as a transmitter and a receiver. For synchronization of the system, we used a PXI-6683H timing and synchronization module, which is crucial for proper operation of multicarrier waveforms. This module delivers a 10 MHz clock signal, which is transmitted to nodes with RF cables and is capable of 1 ppm synchronization through a TCXO oscillator. We used the receiver channel of the PXIe-5644R to measure the spectrum of the modulated transmitter signals. With corresponding software configuration, we prepared the synchronization solution with these components.
We developed two different transceiver applications based on LabVIEW to implement the OFDM and UFMC techniques on our testbed equipment. Each application consists of a source VI and a destination VI, which realize transmitter and receiver functions, respectively.
NI SDR products helped us achieve our project goals faster and with fewer complexities due to reusability, existing examples, and the mature community. We had access to documentation around the examples, ready-to-run conceptual examples, and courseware and lab materials around the grounding wireless communication topics through the NI ecosystem. We took advantage of the graphical nature of LabVIEW to combine existing blocks of algorithms more easily compared to text-based options.
In OFDM’s source VI, we reserved the USRP configuration except the MIMO setting that we replaced with a single user setting. However, we implemented Data SubVI with quite different parameter pairs. According to determined frame structure, a symbol that has length of 600 samples is transmitted per transmission interval. We chose 240, 512, and 64 as the number of data subcarriers, FFT, and CP lengths, respectively. Due to insufficient synchronization sensitivity, we inserted a short packet that consists of only 24 samples at the beginning of the transmitted symbol for estimation of the TO and CFO. It is important to note that we used the same synchronization packet with the same setting in the UFMC for a fair comparison. Beyond these, we prepared other components and implemented the VI.
Similar to source VI, we implemented OFDM’s destination VI as in the previous study except OFDMA-related components. We did not changed the USRP configuration. However, we modified CFO estimation SubVI and used the synchronization packet for estimation rather than the CP portion.
In UFMC’s source VI, we used the same USRP and parameter configurations in the OFDM’s VI. We constituted the UFMC symbol in the Data SubVI. In this VI, we implemented the transmitter model with LabVIEW functions like array functions and Modulation Toolkit VI. We managed the Data SubVI on the basis of sub-bands. Specifically, we managed the signal after the sub-band generation process in Sub-Band SubVI. Figure 4 shows an exemplary view of the management loop of this SubVI in Data SubVI.
We used a short packet for timing and frequency synchronization. It is noteworthy that in both implementations, samples included in this packet are inserted from the last 24 samples of OFDM and UFMC symbols, which increases the estimation performance. With these additions, we implemented
UFMC source VI.
Comparing the measurements for OFDM and UFMC testbeds, the performance of UFMC is clearly superior. For instance, we obtained EVM as 18.50 percent at 1 dB transmitter power with UFMC. This is much less than OFDM, which we obtained as 34.96 percent.
Moreover, we conducted experiments to measure the performance of the proposed PAPR reduction method for UFMC waveform. Figure 6 shows that we can obtain a PAPR reduction of more than 1 dB, which demonstrates the effectivity of the considered method.
In addition to these facts, sidelobe suppression performance of a 5G waveform candidate is also important and requires proper analysis. In our waveform studies, we first create a simulation environment by using LabVIEW as simulation tool and compute numerical performance results. For OFDM and UFMC, we followed the same methodology and observed power spectral density graphs to analyze sidelobe suppression performance. Based on our observations, we proposed a technique based on digital correlative precoding for UFMC and we entitled it as Precoded-UFMC. We also measured its performance. Figure 7 shows a comparison of sidelobe suppression performance of OFDM, UFMC, and Precoded-UFMC. As observed, Precoded-UFMC has a superior suppression performance and improves UFMC’s spectral containment properties. Furthermore, interference levels between sub-bands decrease significantly with precoding; therefore, Precoded-UFMC may be more suitable for various 5G scenarios than UFMC.
In accordance with previous results, we validated the sidelobe suppression performances of corresponding waveforms in real time by using SDR nodes. Robustness of uplink communication was critical for mobile networks and SC-FDMA was the most used waveform in recent technologies. Because of DFT precoding, SC-FDMA based waveforms were also entitled as DFT-s-OFDM and some other types had similar names. To render UFMC suitable to uplink, we implemented DFT-s-UFMC—in other words, DFT precoded UFMC. Moreover, as a special type of DFT-s-OFDM, we adopted zero-padded (ZP) DFT-s-OFDM to UFMC and proposed ZP-DFT-s-UFMC. We implemented all these waveforms by using LabVIEW and PXIe-5644R module and measured their real-time spectral containment performances. This approach was important to understanding whether theoretical advantages were valid or not in real time. Without this observation, we could not evaluate these waveforms within 5G candidates. Figure 8 shows related results. It is clear that DFT-s-UFMC has superior suppression performance compared to DFT-s-OFDM and these results match simulation results. The ZP version of DFT-s-UFMC performs even better. Therefore, we can state that proposed UFMC methods are suitable to real time.
As the last component, we also studied multiple access techniques and analyzed them in real time. Among these techniques, non-orthogonal multiple access (NOMA) attracts attention because of its significant spectral efficiency benefits. Accordingly, symbols of different users were overlapped in the same resources, that is, frequency, time, and/or power. Symbols of users who have better channels were transmitted with less power. Users who do not have proper channels receive symbols having higher power. However, non-orthogonality renders NOMA’s applicability limited because of real-time issues and the need for proper self-interference cancellation (SIC). To analyze NOMA’s applicability, we implemented NOMA-OFDM in real time for the first time in the literature. We used SDR nodes to analyze channel conditions and estimation errors, and observe their destructive effects on NOMA. With a proper design methodology, we implemented NOMA properly in downlink with one transmitter and two receiver nodes. Because of usage of two receiver nodes, spectral efficiency is doubled. Different subcarriers overlap at the same frequency bins with different powers, and we used one frequency block rather than two blocks as used in an orthogonal fashion. Figure 9 and Figure 10 show the corresponding results. Figure 9 shows transmitted constellation diagrams. Figure 10 shows exemplary receptions. The first user receives degraded diagrams. The second user has a better reception performance and can decode its symbols properly after SIC. With proper gain adjustment, we can achieve a huge spectral efficiency improvement with a low error rate. For future work, we plan to implement NOMA in cooperative communications and propose new techniques by exploiting NOMA principals.
Thanks to our testbed, we presented SDR implementation of UFMC waveform in real time for the first time in the literature. With the proposed real-time implementation structure and the detailed synchronization methods, we showed UFMC’s suitability for 5G scenarios with tests based on EVM and BER metrics.
We expect that this study can aid 5G research activities with insight for the real-time deployment stages. In the near future, we plan to address the above-mentioned implementation issues and improve error performance of the UFMC waveform in real-time setup by proposing more robust synchronization methods. Furthermore, by adapting our experience on Gauss-Hermite Filtering, adaptive modulation, and synchronization methods, we will propose very robust UFMC enhancements and new waveform solutions that have been validated with our preliminary results. The development processes of these techniques are in progress with our testbed, and we have targeted suitability to the real-time scenarios.
We are also in the planning stage of deploying our testbed to multiple demonstration sites and partnering with an NI Alliance Partner. The NI Alliance Partner Network gives a clear picture of the capabilities and certification level of partner companies, which dramatically reduces the research and verification phase of the project partners. We are confident in outsourcing this part of our project and initial results are positive.
Gunes Karabulut Kurt
Istanbul Technical University, Wireless Communication Research Laboratory
Tel: +90 (212) 285 3509