Madhusudhan Gurumurthy is a Senior Applications Specialist at Spirent Communications. He joined Spirent in 2006 and has contributed in roles in Product Development, Product Marketing and Standards Strategy. He also represents Spirent at 3GPP RAN meetings. Madhu earned a BTech-EE from Pondicherry Engineering College and an MSEE from the University of Cincinnati, where he developed a MIMO channel-estimation algorithm as part of his thesis
MWJ: Today, it often seems that beamforming is mentioned in the context of TD-LTE systems. Is there a technical reason for that? If not, why does that seem to be the case?
MG: MIMO and beamforming techniques are meant to improve system efficiency. TD-LTE, in common with other time-domain related wireless technologies like TD-SCDMA and WiMAX, is geared towards maximizing the utility of a single fixed frequency band. Because a TD-LTE provider is often starting with limited spectral resources, they’re charged with maximizing efficiency from the get-go. Urban areas in China are very densely populated, so the capacity gains offered by beamforming are especially attractive, making these areas a ripe market for early large-scale TD-LTE deployments.
There’s also an in-depth technical reason. Any beamforming technology requires accurate per-user “channel estimation” in order to properly form the beam and direct energy to a specific user. In time-multiplexed systems there is an inherent ability to estimate the downlink channel characteristics by measuring what is received from the mobile in the uplink. So for TD-LTE, since the duplex timing between downlink and uplink is relatively short, and since the downlink and uplink share the same frequency, the uplink signal is an excellent channel-sounding signal which can be used to estimate the downlink channel.
MWJ: Are 8-antenna systems simply an expansion of 4-antenna systems or is there more to it than that?
MG: Well yes, they could be, but in practice they are not. While we tend to speak of “MIMO beamforming” or “dual-layer beamforming” as one technique, it’s actually a combination of two related but different techniques. In practice, these techniques are first being combined in eight-antenna systems because that’s where the combination offers the most benefit.
Any multi-antenna array can be used to create beamforming. While implementation is complex, it’s a fairly simple concept: we control complex signal parameters (phases and gain) from a set of transmit antenna elements so that they add constructively at a desired physical location.
We could also use those same antenna elements to create MIMO transmission instead of beamforming. In this case, the physical separation between TX elements creates some independence between signal paths, allowing for MIMO gains to be achieved when using the right signaling format.
In practice, these two multi-antenna mechanisms are combined to create MIMO beamforming. If we start with eight antenna elements, four of them can be used to create one beamforming antenna while the others create a second beamforming antenna. This is essentially 2x2 MIMO except that the two transmissions come from beamforming antenna systems rather than single radiating elements. These eight antennas could also be used to do pure MIMO, or pure eight-antenna beamforming. The choice of transmission scheme is flexible, dynamic, and in many ways specific to the manufacturer implementing the eNodeB.
MWJ: 8-antenna LTE is said to be one of the more difficult things to test –why is that?
MG: One reason is that, due to the complexity of how the over-the-air channel is used, more realistic channel models are being used than in earlier well-known standards. Specifically, the use of geometric channel models, rather than the correlation-based models we’re used to, becomes a necessity. Further, the effects of antenna patterns, varying antenna spacing, polarization properties, and antenna configurations need to be modeled. A geometric channel approach models the radio channel as a set of rays, each of which has not only specific amplitude and phase properties, but correct spatial and temporal properties as well. The result is a much more accurate model for cases like MIMO beamforming, where simulating these aspects of the channel are critical to accurate testing.
For the testing we’re talking about, the industry has settled on the ITU’s geometric model, as documented in Report ITU-R M.2135. By the way, a number of leading researchers in this area have told us that Spirent’s implementation of this model is by far the most accurate they’ve seen.
Another challenge in testing these systems is that the phase characteristics of the test system need to be carefully controlled in order to properly test the beamforming function of the base station. Therefore, we as a test equipment maker, have to ensure optimal phase accuracy in both our model and our test system. This is no easy task when you consider that every element of the system, meaning every piece of equipment, every cable and every variation in temperature can all affect signal phase. We’ve had to do some pretty serious hardware development that specifically optimizes system phase accuracy while also allowing for automated phase calibration on a regular basis.
Finally, the scale of the system is itself a challenge. Modeling a SISO (Single Input, Single Output) channel is nothing more than modeling a signal link from a transmit antenna element to a receive antenna element. Modeling an 8 2 system means creating sixteen of those links in each direction and properly managing the relationships between them. The old test equipment paradigm didn’t work… although it could in principle be done, it took a considerable amount of time and effort and it increased the opportunity for setup errors.
For our new VR5 HD Spatial Channel Emulator, we’ve completely re-designed the way people interface with channel emulators. The software presents a lot of control and feedback graphically wherever possible. It guides the user through scenario configuration and simplifies the interaction required during setup. The engineers who’ve been using it are very relieved to not have to worry about a lot of arcane channel modeling details whenever they want to run testing—we do that job, allowing them to concentrate on executing their test plan.
MWJ: What is different about channel models for 8-antenna LTE?
MG: First of all, there are the geometric channel models we just mentioned. While we’ve been working with these models for years, TD-LTE deployments are the first occasions where the industry insists on their use. One reason is that they can correctly model the actual antenna patterns, which was a less critical issue in legacy technologies.
You also need to be able to model a wide variety of antenna configurations. Earlier I described a MIMO beamforming base station antenna made up of four pairs of orthogonally oriented antenna elements. These pairs might be equally spaced, but that’s not always the case. You may need to model eight vertically mounted elements, or pairs of elements that are not equally spaced, or any one of a variety of other configurations.
Finally, because phase relationships are so critical, you need to be able to properly model specific polarizations within the channel.
MWJ: Have you seen anything to indicate that these technologies will be able to deliver what they promised?
MG: Yes, actually. Originally there were some software simulations that seemed to bear results very close to what theory suggests. More recently, tests using our lab-based equipment, followed up by actual field-testing, bore similar results. Now, of course, there’s a very large pilot deployment in Shanghai, Xiamen, Nanjing, Hangzhou, Shenzhen and Guangzhou. Everyone involved seems to be confident that all of this work will pay off with the desired capacity and user gains.
MWJ: Most of the work in 8-antenna systems and beamforming seems to be going on in China. Why is this and how does this affect other parts of the world?
MG: On the technical front, pioneering work in 8n systems is being done where operators have single-band spectrum and where population density is high. Both of these factors bolster the business case for early investments in eight-antenna technologies. It’s also no secret that China Mobile and China Telecom, who serve about 650 million and 125 million subscribers respectively, are pursuing this technology. The connection to these operators alone makes the case for significant investment.
This work will eventually benefit industry players and subscribers all over the world. Even though the first target market is in China, network equipment for this market is being developed all over the world. Well-known infrastructure manufacturers in Sweden, Finland, Germany, Korea, and the US are all competing for business in this new space. There are other markets for the technologies as well. In India, all the broadband wireless access (BWA) spectrum winners, including Reliance, Airtel and Aircel are conducting TD-LTE trials. Here in the US, Clear (Clearwire) has announced its research partnership with China Mobile. In summary, it is clearly evident that single band spectrum currently used for other technologies such as WiMAX will ultimately be migrated to the TD-LTE technologies advancing today in China.