Researchers at CATT and NYU WIRELESS have built the world’s first wireless emulator suitable for 5G systems that feature massive bandwidths and hundreds of antenna elements. In this unique patented design, the Emulator emulates not only the wireless channel, but also the beamformers (or phased-arrays) on both the transmitter and receiver devices under test. This joint emulation of the beamformer and the wireless channel is the key enabling technology that allows for faster development cycles, while significantly lowering the hardware cost and complexity of the emulator. The project was led by post-doctoral fellow Dr. Aditya Dhananjay, and was supervised by faculty members Dr. Sundeep Rangan and Dr. Dennis Shasha. The project was enabled by a generous hardware donation of commercial off-the-shelf (COTS) components from National Instruments. The NYU researchers are also making the Emulator software available to academic researchers for free.
As the 5G standardization process gathers steam, one of the critical challenges lies with the design and testing of the new generation of wireless systems. As Dr. Dhananjay explained, “a key component of this testing process is channel emulation, where-in a channel emulator is used to simulate a configurable wireless channel between the transmitter and receiver devices under test”. An emulator can be configured to reproduce a variety of wireless scenarios such as urban micro-cell, rural macro-cell, mobility patterns, different weather conditions, and so on. Channel emulation offers reproducibility and enables the validation and testing of designs under worst-case scenarios, and is therefore an essential step before the time-consuming and expensive over-the-air (OTA) and field testing.
In the existing emulation paradigm, the transmitter and receiver devices under test (TX and RX DUTs) are connected using cables to a channel emulator as shown in Fig. 1. The RF signal vector x from the TX DUT (one signal from each antenna element) is cabled through to the emulator. The wireless channel to be emulated is generally described via multipath fading profile which can be configured to reproduce measured traces or standard profiles such as in the 3GPP models. The output from the emulator is the signal vector y, which is then cabled through to the RX DUT using one cable for each antenna element. Dr. Dhananjay explained, “basically, the job of a channel emulator is to transform the input signal vector x into the output signal vector y”. Due to the benefits offered, channel emulators have been widely used for WiFi, 3G, and 4G LTE development, and are indeed a staple on any wireless research lab bench.
However, there are no commercial emulators available that are suitable for upcoming 5G millimeter wave (mmWave) systems. This is because the upcoming 5G mmWave systems will differ from existing 4G systems in two important ways: a) the number of antenna elements are increased by an order of magnitude due to the use of mmWave phased arrays; and b) the bandwidths that these systems operate over is also increased by at least an order of magnitude. These differences make the existing emulation paradigm unsuitable for 5G mmWave systems for a variety of reasons. First, phased-array antenna elements cannot be connected to cables. Second, the large number of antenna elements makes the hardware cost of building the emulator prohibitively expensive. Third, the computational complexity is increased by at multiple orders of magnitude due to the large number of antenna elements and large bandwidth. As Dr. Dhananjay explained, “it is precisely for these reasons that no commercial 5G mmWave emulators exist today, which can support hundreds of antenna elements and several gigahertz of real-time bandwidth without resorting to prohibitively expensive frequency stitching”.
The NYU Emulator solves these challenges by defining a fundamentally new Emulation paradigm for 5G mmWave systems. In this paradigm, the Emulator not only emulates the wireless channel, but also the multi-antenna beamformers on both the TX and RX DUTs as contrasted in Fig. 2(a) and 2(b). The TX and RX DUTs share their instantaneous beamforming vectors wtx and wrx with the Emulator, so that the beamforming operations can be emulated. The TX DUT sends the Emulator the pre-beamformed signal u as opposed to the post-beamformed signal vector x. Symmetrically, the Emulator sends the RX DUT the post-beamformed signal z as opposed to the output pre-beamformed signal vector y. The Emulator in this new paradigm transforms u to z, as opposed to x to y. The Emulator is flexible, and can support the signals u and z being in baseband, IF, or in RF, depending on the DUT configuration. By combining the emulation of the multi-antenna beamformers on the DUTs with the emulation of the wireless channel (the key patented technology), the computational complexity and hardware cost of the Emulator are in fact greatly reduced. The intuition behind the reduction in complexity can be explained by the fact that the dimensions of u and z are significantly lower than that of x and y, respectively. Dr. Dhananjay further explained, “another key benefit in this new emulation paradigm is that researchers can experiment over different theoretical phased-array designs, further accelerating the R&D in protocols at all layers of the protocol stack”. The NYU team has already demonstrated this Emulator with 64-element DUTs and more than 2GHz bandwidth at two major trade-shows: the Brooklyn 5G summit, and at a workshop at NI Week in Austin.
The implementation of this Emulator was supported by a very generous hardware donation from NI to NYU. The software that runs the Emulator, along with reference TX and RX DUT software designs will be made available to academic researchers for free.
Fig. 1. The existing emulation paradigm where the TX and RX DUTs interface with the emulator over RF with one cable per antenna element. This method of emulation is unsuitable for mmWave systems due to the prohibitive hardware cost, high computational complexity, and the inability to connect phased-array antennas to cables.
Fig. 2(a). The existing emulation paradigm, with the internal structure of the TX and RX DUTs illustrated. Note that the beamforming operations are performed by the DUTs themselves.
Fig. 2(b). The proposed emulation paradigm, where the Emulator performs not only the emulation of the wireless channel, but also of the beamforming antenna arrays on both the TX and RX. This design enables the hardware cost and computational complexity to be manageable, even when the number of antenna elements and bandwidth are both increased by an order of magnitude.